Nri V04 N07 2004

FOREWORD
TOLL-LIKE RECEPTORS AND THEIR PLACE IN IMMUNOLOGY

Where does the immune response to infection begin?
And for that matter, where does all immunity begin, including autoimmunity? Is there a first cause of such complex phenomena? A molecular spark that lights the fire? The answer is much simpler than many might have thought even a short time ago. For more than a century, we have known of the existence of two basic types of immunity in mammals. There is adaptive immunity (known as specific immunity), so named because it reacts specifically to the situation, gathering strength once a particular micro-organism attacks the host. And there is innate immunity (known as natural immunity or innate resistance), so named because we are born with it, and it defends us against microbes we have never seen before. Both of these immune systems (in fact, any immune system imaginable) must have a means of sensing infection, discriminating self from non-self, and controlling or eradicating infection. Although effector mechanisms were defined for both systems decades ago, the key molecular sensors used by the mammalian innate immune system to detect microorganisms were identified only in the late 1990s. Their discovery was presaged by work in Drosophila (discussed in this special issue by Bruno Lemaitre), which led to the realization that a cell-surface protein known as Toll previously better known for its role in development is a key component of a biochemical pathway that alerts flies to infection, prompting the production of antimicrobial peptides. This discovery coincided with steady advances in genetic technology that allowed others to show that mammals use Toll-like receptors (TLRs) as direct sensors of microbial infection. The discovery of TLR function was a triumph of reductionism and filled a great void in immunology. Immune responses are tremendously complex when fully developed, yet it now appears they are ignited by only a few molecules. Mammalian responses to virtually all microorganisms depend upon TLRs ab initio. The TLRs are the key site of direct and consequential contact between the cellular immune system and its microbial quarry. If the TLRs, their adaptors, or the kinases to which these adapters are linked are lacking, the host is severely immunocompromised. Macrophages do not make the requisite cytokines to initiate the full repertoire of innate immune responses. They do not summon neutrophils to aid in the destruction of microorganisms. Neither do neutrophils respond directly to microorganisms in the way that they should. And without the innate immune response that precedes it, the adaptive immune response is feeble. The TLRs are clearly at the top of the immune-system pyramid. This level of understanding is new, and the field has developed quickly. Both random and targeted germline mutagenesis have helped to resolve the complexity of the TLRs and their signalling pathways, and we are beginning to elucidate a conceptual framework or roadmap that will be essential to understanding just how the TLRs work and what they do. Here, Shizuo Akira reviews recent advances in signalling, taken from strong genetic data. Luke ONeill has organized what is known about TLR signalling into poster format, so that we might immediately understand how mammalian innate immune sensing and signalling operate. These presentations provide a foundation on which anyone might build. Although any scientific advance is a fine thing in its own right, most of us would like to know how we might best put it to use. Here, Richard Ulevitch considers how the TLRs might be targeted therapeutically. We have imagined (and hoped) that sterile inflammation, as it occurs in autoimmune diseases, for example, could depend on activation of the same biochemical pathways as those triggered by microorganisms, by yet-to-beidentified endogenous stimuli. With this in mind, quasiinfectious stimuli, such as lipopolysaccharide, have long been used to induce, model and study inflammation. We have also known that, where infection is concerned, terrible consequences sometimes arise from the hosts earnest attempt to defend itself. The TLRs offer a clear entry point into these problems. Although we cannot yet know where our new knowledge will lead, we know where the trail begins.
Bruce Beutler The Scripps Research Institute, Department of Immunology, IMM-3110550, N. Torrey Pines Road, La Jolla, California 92037, USA. e-mail: bruce@scripps.edu
498
| JULY 2004 | VOLUME 4
www.nature.com/reviews/immunol
FOCUS ON TLR SIGNALLING
REVIEWS
TOLL-LIKE RECEPTOR SIGNALLING
Shizuo Akira* and Kiyoshi Takeda
One of the mechanisms by which the innate immune system senses the invasion of pathogenic microorganisms is through the Toll-like receptors (TLRs), which recognize specific molecular patterns that are present in microbial components. Stimulation of different TLRs induces distinct patterns of gene expression, which not only leads to the activation of innate immunity but also instructs the development of antigen-specific acquired immunity. Here, we review the rapid progress that has recently improved our understanding of the molecular mechanisms that mediate TLR signalling.
All living organisms are exposed constantly to microorganisms that are present in the environment and need to cope with invasion of these organisms into the body. The vertebrate immune response can be divided into innate and acquired immunity, with innate immunity being the first line of defence against pathogens. By contrast, acquired immune responses are slower processes, which are mediated by T and B cells, both of which express highly diverse antigen receptors that are generated through DNA rearrangement and are thereby able to respond to a wide range of potential antigens. This highly sophisticated system of antigen detection is found only in vertebrates and has been the subject of considerable research. Far less attention has been directed towards innate immunity, as it has been regarded as a relatively nonspecific system, with its main roles being to destroy pathogens and to present antigen to the cells involved in acquired immunity. However, recent studies have shown that the innate immune system has a greater degree of specificity than was previously thought and that it is highly developed in its ability to discriminate between self and foreign pathogens1. This discrimination relies, to a great extent, on a family of evolutionarily conserved receptors, known as the Toll-like receptors (TLRs), which have a crucial role in early host defence against invading pathogens1,2. Furthermore, accumulating evidence indicates that activation of the innate immune system is a prerequisite for the induction of acquired immunity, particularly for the induction of a T helper 1 (TH1)-cell response3,4. This marked shift in our thinking has changed our ideas about the pathogenesis and treatment of cancers, and infectious, immune and allergic diseases. In the past few years, our knowledge of TLR signalling and the responses these receptors control has greatly increased. In this review, we discuss the TLRs, focusing on their signalling pathways.
TLR/IL-1R superfamily: structure and function
*Department of Host Defense, Research Institute for Microbial Diseases, Osaka University, and ERATO, Japan Science and Technology Agency, 3-1 Yamada-oka, Suita, Osaka 565-0871, Japan. Department of Molecular Genetics, Medical Institute of Bioregulation, Kyushu University, 3-1-1 Maidashi, Higashi-ku, Fukuoka 812-8582, Japan. Correspondence to S.A. e-mail: sakira@biken.osaka-u.ac.jp
doi:10.1038/nri1391
The discovery of the TLR family began with the identification of Toll, a receptor that is expressed by insects and was found to be essential for establishing dorsoventral polarity during embryogenesis5. Subsequent studies revealed that Toll also has an essential role in the insect innate immune response against fungal infection6. Homologues of Toll were identified through database searches, and so far, 11 members of the TLR family have been identified in mammals. The TLRs are type I integral membrane glycoproteins, and on the basis of considerable homology in the cytoplasmic region, they are members of a larger superfamily that includes the interleukin-1 receptors (IL-1Rs). By contrast, the extracellular region of the TLRs and IL-1Rs differs markedly: the extracellular region of TLRs contains leucine-rich repeat (LRR) motifs, whereas the extracellular region of IL-1Rs contains three immunoglobulin-like domains (FIG. 1a). Toll/IL-1R domain. TLRs and IL-1Rs have a conserved region of ~200 amino acids in their cytoplasmic tails, which is known as the Toll/IL-1R (TIR) domain7. Within the TIR domain, the regions of homology comprise three conserved boxes, which are crucial for
NATURE REVIEWS | IMMUNOLOGY
VOLUME 4 | JULY 2004 | 4 9 9
REVIEWS
signalling (FIG. 1a). Amino-acid sequence conservation among the TIR domains is generally 2030%, and these domains vary in size. The crystal structures of the TIR domains of human TLR1 and TLR2 have been obtained and analysed; they contain a central fivestranded parallel -sheet, which is surrounded by five -helices on each side8. These two secondary structural elements are connected by loops: for example, the BB loop connects the strand -B and the helix -B. The conserved boxes 1 and 2 and the BB loop are adjacent and display most of their side chains for interaction with adaptor molecules. C3H/HeJ mice have a defect in their ability to respond to lipopolysaccharide (LPS) because of a missense mutation in the Tlr4 gene9, which alters the sequence located at the tip of the BB loop, farthest from the rest of the TIR domain. This indicates that the mutation abrogates LPS signalling not because it disrupts the TIR domain structure itself, but rather because it disrupts a direct point of contact with another molecule or molecules, specifically with other TIR-domain-containing molecules. Leucine-rich repeats. The extracellular domain of TLRs contains 1925 tandem copies of the LRR motif. Each repeat consists of 2429 amino acids and contains the leucine-rich sequence XLXXLXLXX, and another conserved sequence XXXX4FXXLX (REF.10), where X denotes any amino acid and a
Figure 1 | TLR structure and signalling. a | Toll-like receptors (TLRs) and interleukin-1 receptors (IL-1Rs) have a conserved cytoplasmic domain, that is known as the Toll/IL-1R (TIR) domain. The TIR domain is characterized by the presence of three highly homologous regions (known as boxes 1, 2 and 3). Despite the similarity of the cytoplasmic domains of these molecules, their extracellular regions differ markedly: TLRs have tandem repeats of leucine-rich regions (known as leucine rich repeats, LRR), whereas IL-1Rs have three immunoglobulin (Ig)like domains. b | Stimulation of TLRs triggers the association of MyD88 (myeloid differentiation primary-response protein 88), which in turn recruits IRAK4 (IL-1R-associated kinase 4), thereby allowing the association of IRAK1. IRAK4 then induces the phosphorylation of IRAK1. TRAF6 (tumour-necrosis-factorreceptor-associated factor 6) is also recruited to the receptor complex, by associating with phosphorylated IRAK1. Phosphorylated IRAK1 and TRAF6 then dissociate from the receptor and form a complex with TAK1 (transforming-growthfactor--activated kinase), TAB1 (TAK1-binding protein 1) and TAB2 at the plasma membrane (not shown), which induces the phosphorylation of TAB2 and TAK1. IRAK1 is degraded at the plasma membrane, and the remaining complex (consisting of TRAF6, TAK1, TAB1 and TAB2) translocates to the cytosol, where it associates with the ubiquitin ligases UBC13 (ubiquitinconjugating enzyme 13) and UEV1A (ubiquitin-conjugating enzyme E2 variant 1). This leads to the ubiquitylation of TRAF6, which induces the activation of TAK1. TAK1, in turn, phosphorylates both mitogen-activated protein (MAP) kinases and the IKK complex (inhibitor of nuclear factor-B (IB)-kinase complex), which consists of IKK-, IKK- and IKK- (also known as IKK1, IKK2 and nuclear factor-B (NF-B) essential modulator, NEMO, respectively). The IKK complex then phosphorylates IB, which leads to its ubiquitylation and subsequent degradation. This allows NF-B to translocate to the nucleus and induce the expression of its target genes.
a
IL-1R TLR
Ig-like domain
LRRs
Box 1 Box 2 TIR domain
Box 3
b
TLR
MyD88
IRAK1
IRAK4
UEV1A
UBC13
TRAF6
TAB2 TAK1 TAB1
IKK- IKK- IKK- MAP kinases
IB p50 p65
NF-B
NF-B-binding motif
Nuclear membrane
500
Table 1 | Toll-like receptors and their ligands

Receptor TLR1 TLR2 Ligand Triacyl lipopeptides Soluble factors Lipoprotein/lipopeptides Peptidoglycan Lipoteichoic acid Lipoarabinomannan Phenol-soluble modulin Glycoinositolphospholipids Glycolipids Porins Atypical lipopolysaccharide Atypical lipopolysaccharide Zymosan Heat-shock protein 70* Double-stranded RNA Lipopolysaccharide Taxol Fusion protein Envelope protein Heat-shock protein 60* Heat-shock protein 70* Type III repeat extra domain A of fibronectin* Oligosaccharides of hyaluronic acid* Polysaccharide fragments of heparan sulphate* Fibrinogen* Flagellin Diacyl lipopeptides Lipoteichoic acid Zymosan Imidazoquinoline Loxoribine Bropirimine Single-stranded RNA Imidazoquinoline Single-stranded RNA CpG-containing DNA N.D. N.D. Origin of ligand Bacteria and mycobacteria Neisseria meningitidis Various pathogens Gram-positive bacteria Gram-positive bacteria Mycobacteria Staphylococcus epidermidis Trypanosoma cruzi Treponema maltophilum Neisseria Leptospira interrogans Porphyromonas gingivalis Fungi Host Viruses Gram-negative bacteria Plants Respiratory syncytial virus Mouse mammary-tumour virus Chlamydia pneumoniae Host Host Host Host Host Bacteria Mycoplasma Gram-positive bacteria Fungi Synthetic compounds Synthetic compounds Synthetic compounds Viruses Synthetic compounds Viruses Bacteria and viruses N.D. Uropathogenic bacteria References 112 113 114 115,116 116 117 118 119 120 121 122 123 124 125 52 9 126 127 128 129,130 131 132 133 134 135 136 137 116 138 139 12 12 140,141 142 140 143 144
TLR3 TLR4
TLR5 TLR6
TLR7
TLR8 TLR9 TLR10 TLR11
*It is possible that these ligand preparations, particularly those of endogenous origin, were contaminated with lipopolysaccharide and/or other potent microbial components, so more-precise analysis is required to conclude that TLRs recognize these endogenous ligands. N.D., not determined; TLR, Toll-like receptor.
hydrophobic amino acid. The repeats comprise a -strand and an -helix connected by loops. The LRR domains of TLRs form a horseshoe structure, and it is thought that the concave surface of the LRR domains is involved directly in the recognition of various pathogens. The main ligands recognized by different TLRs are summarized in TABLE 1. Remarkably, despite the conservation among LRR domains, different TLRs can recognize several structurally unrelated ligands1,3,4. The subcellular localization of different TLRs correlates to some extent with the molecular patterns of their ligands. TLR1, TLR2 and TLR4 are located on the cell surface and are recruited to phagosomes after activation by their respective ligands. By contrast, TLR3, TLR7 and TLR9, all of which are involved in the recognition of nucleic-acid-like structures, are not expressed on the cell surface1113. For example, TLR9 has recently been shown to be expressed in the endoplasmic reticulum, and it is recruited to endosomal/lysosomal compartments after stimulation with CpG-containing DNA14.
TLR/IL-1R-superfamily signalling cascade
After ligand binding, TLRs/IL-1Rs dimerize and undergo the conformational change required for the recruitment of downstream signalling molecules. These include the adaptor molecule myeloid differentiation primary-response protein 88 (MyD88), IL-1R-associated kinases (IRAKs), transforming growth factor- (TGF-)-activated kinase (TAK1), TAK1-binding protein 1 (TAB1), TAB2 and tumour-necrosis factor (TNF)-receptor-associated factor 6 (TRAF6)15,16. MyD88. MyD88 was isolated originally as a gene that is induced rapidly during the IL-6-stimulated differentiation of M1 myeloleukaemic cells into macrophages17. The encoded protein has an amino (N)-terminal death domain (DD), which is separated from its carboxy (C)-terminal TIR domain by a short linker sequence. MyD88 was subsequently cloned as an adaptor molecule that functions to recruit IRAK to the IL-1R complex following stimulation with IL-1 (REFS 1820). The association between MyD88 and IRAK is mediated
VOLUME 4 | JULY 2004 | 5 0 1
REVIEWS
through a DDDD interaction. MyD88 forms homodimers through DDDD and TIR-domainTIR-domain interactions and exists as a dimer when recruited to the receptor complex21. Therefore, MyD88 functions as an adaptor linking TLRs/IL-1Rs with downstream signalling molecules that have DDs. IRAK family. Four IRAKs IRAK1, IRAK2, IRAK4 and IRAK-M showing distinct gene-expression patterns have been identified in mammals22. IRAKs contain an N-terminal DD and a central serine/threonine-kinase domain. IRAK1 and IRAK4 have intrinsic kinase activity, whereas IRAK2 and IRAK-M have no detectable kinase activity. The kinase activity of IRAK1 increases strongly following TLR/IL-1R stimulation, and its kinase domain is essential for signalling through nuclear factor-B (NF-B). However, kinase activity itself is not essential for signalling, because in IRAK1-deficient cells, the overexpression of a kinase-defective mutant of IRAK1 can strongly induce NF-B activation23. By contrast, the overexpression of IRAK4 does not result in robust NF-B activation, yet the expression of a kinase-inactive mutant of IRAK4 inhibits IL-1-mediated NF-B activation. It has also been shown that IRAK1 is a direct substrate of IRAK4 but not vice versa24. In IRAK1deficient mice, cytokine production in response to IL-1 and LPS was diminished but not abolished2527, whereas IRAK4-deficient mice showed virtually no response to IL-1, LPS or other bacterial components, demonstrating that IRAK4 has an important role in IL-1R/TLR signalling28. Recently, patients with an inherited IRAK4 deficiency have been identified29. These patients failed to respond to IL-1, to IL-18 or to stimulation of each of five TLRs (TLR2, TLR3, TLR4, TLR5 and TLR9). Together, these results show that IRAK4 and its kinase activity are required for TLR signalling and that IRAK4 functions upstream of IRAK1. TRAF6. TRAFs constitute a family of evolutionarily conserved adaptor proteins30. So far, six members of the TRAF family have been identified in mammals, and they are characterized by the presence of an N-terminal coiled-coil domain (known as TRAF-N) and a conserved C-terminal domain (known as TRAF-C). The N-terminal portion of most TRAF proteins contains a RING (really interesting new gene)-finger/zinc-finger region, which is essential for downstream signalling events, whereas the TRAF-C domain mediates selfassociation and interactions with upstream receptors and signalling proteins. TRAF6 functions as a signalling mediator for both the TNF-receptor superfamily and the TLR/IL-1R superfamily, interacting directly with members of the TNF-receptor superfamily (CD40 and TNF-related activation-induced cytokine receptor, TRANCER) or being coupled indirectly to TLR/IL-1R superfamily members through its association with IRAKs. The consensus sequence for the TRAF6-binding domain has been identified as P-X-E-X-X-(D/E/F/W/Y)31. This motif is found in CD40, TRANCER, and the IRAKs; three of these TRAF6-binding motifs are found in IRAK1, two in IRAK2 and one in IRAK-M. TAK1 and TABs. The activation by TRAF6 of the transcription factors NF-B and activator protein 1 (AP1) involves TAK1 and two adaptor proteins TAB1 and TAB2. TAK1 is a member of the mitogen-activated protein kinase kinase kinase (MAPKKK) family32, which has been shown to be essential for both IL-1/LPS- and TNF-induced NF-B activation33. Two TAK1-binding proteins, TAB1 and TAB2, have been identified34,35. When co-expressed ectopically, TAB1 enhances the kinase activity of TAK1, indicating that TAB1 functions as an activator of TAK134. By contrast, TAB2 functions as an adaptor, linking TAK1 to TRAF6 and thereby facilitating TAK1 activation35. However, embryonic fibroblasts obtained from TAB2-deficient mice show no impairment in either IL-1/LPS- or TNFinduced activation of NF-B36. Furthermore, a new TAB2-like molecule, TAB3, has been identified recently, and similar to TAB2, it has been shown to associate with TAK1 and activate NF-B37. Co-transfection of SMALL INTERFERING RNAS (siRNAs) directed against both TAB2 and TAB3 inhibited both IL-1- and TNFinduced activation of TAK1 and NF-B, indicating that TAB2 and TAB3 function redundantly as mediators of TAK1 activation. It has been shown that UBIQUITYLATION has an important role in TAK1 activation and that TRAF6 functions as an E3 ubiquitin ligase38. TRAF6 can interact through its RING-finger domain with ubiquitin-conjugating enzyme 13 (UBC13), and this UBC13TRAF6 complex triggers TAK1 activation through the assembly of a lysine63-linked polyubiquitin chain39. NF-B. The NF-B family of transcription factors is composed of five members p65 (REL-A), REL-B, cytoplasmic (c) REL, p50 and p52 which function as homo- and heterodimers. NF-B dimers are usually sequestered in the cytoplasm in an inactive form by molecules of the inhibitor of NF-B (IB) family. Activation of NF-B involves the phosphorylation and proteolysis of the IB proteins and the concomitant release and nuclear translocation of the NF-B factors. This acute activation process is mediated by the IB kinase (IKK) complex, which comprises two catalytic subunits IKK- and IKK- (also known as IKK1 and IKK2) and a regulatory subunit, IKK- (also known as NF-B essential modulator, NEMO) 40. After activation by upstream signals, IKK phosphorylates the IBs, leading to their polyubiquitylation and proteasome-mediated degradation. Recently, an alternative pathway of NF-B activation has been suggested, in which NF-B-inducing kinase (NIK) activates IKK-, which then phosphorylates the NF-B2 precursor protein p100 (REFS 41,42). After phosphorylation, p100 is recognized by an SCF (S-phase kinase-associated protein 1Cullin1F box)-family E3 ubiquitin ligase, -transducin repeat-containing protein (-TRCP), which catalyses the polyubiquitylation of p100 and thereby triggers its processing by the proteasome to the transcriptionally active p52 form. This proteolytic event is tightly regulated by sequences located in the C-terminal region of p100 (REF. 43).
SMALL INTERFERING RNAS
(siRNAs). Synthetic doublestranded RNA molecules of 1923 nucleotides, which are used to knockdown (silence the expression of) a specific gene. This is known as RNA interference (RNAi) and is mediated by the sequencespecific degradation of mRNA.
UBIQUITYLATION
The attachment of the small protein ubiquitin to lysine residues present in other proteins. This tags these proteins for rapid cellular degradation.
502

This phosphorylation leads to the degradation of IB and consequently the release of NF-B. Activation of TAK1 also results in the activation of MAPKs, including JUN N-terminal kinase (JNK) (FIG. 1b).
MyD88-dependent and -independent pathways
MyD88
TLR4
p50
p65 Late-phase NF-B
IRF3
Early-phase NF-B
Inflammatory cytokines MyD88-dependent response
IFN-, IFN-inducible gene products MyD88-independent response
Figure 2 | TLR4 signalling: MyD88-dependent and independent pathways. Stimulation of Toll-like receptor 4 (TLR4) facilitates the activation of two pathways: the MyD88 (myeloid differentiation primary-response protein 88)-dependent and MyD88-independent pathways. The MyD88-dependent pathway involves the early phase of nuclear factor-B (NF-B) activation, which leads to the production of inflammatory cytokines. The MyD88-independent pathway activates interferon (IFN)-regulatory factor (IRF3) and involves the late phase of NF-B activation, both of which lead to the production of IFN- and the expression of IFN-inducible genes.
Despite these findings, the nuclear translocation of NF-B alone might not be sufficient for activation of NF-B-dependent transcription. Some NF-B proteins, particularly p65, are post-translationally regulated, and various kinases including cyclic AMP (cAMP)dependent protein kinase (PKA), casein kinase II, protein kinase C- (PKC-) and IKK itself have been implicated in this process40. Fitting the TLR/IL-1R signalling pathway together. After TLR/IL-1R stimulation, MyD88 is recruited to the cytoplasmic TIR domain, where it facilitates the association of IRAK4 with the receptor complex through a homophilic DD interaction. The binding of MyD88 to IRAK4 facilitates IRAK4-mediated phosphorylation of a crucial residue or residues in the kinase-activation loop of IRAK1, which induces the kinase activity of IRAK1. Activated IRAK1 then autophosphorylates residues in its N-terminus, and this hyperphosphorylation of IRAK1 enables TRAF6 to bind to this complex. The IRAK1TRAF6 complex then disengages from the receptor and interacts at the plasma membrane with another preformed complex consisting of TAK1, TAB1, and TAB2 or TAB3. This interaction induces phosphorylation of TAB2/TAB3 and TAK1, which then translocate together with TRAF6 and TAB1 to the cytoplasm. TAK1 is subsequently activated in the cytoplasm, leading to the activation of IKKs, which then phosphorylate the IBs.
MyD88-deficient mice do not produce TNF or IL-6 when exposed to IL-1 or microbial components that are recognized by TLR2, TLR4, TLR5, TLR7 or TLR9 (REFS 15,44). So, MyD88 is essential for responses against a broad range of microbial components. However, closer study of MyD88-deficient cells has revealed the existence of MyD88-dependent and -independent pathways, both of which mediate signalling in response to LPS45 (FIG. 2). For example, the activation of NF-B in response to mycoplasmal lipopeptide, a TLR2 ligand, is completely abolished in MyD88-deficient macrophages, whereas NF-B activation still occurs in response to LPS, a TLR4 ligand, although with delayed kinetics. MAPK activation is also delayed in LPS-stimulated MyD88-deficient macrophages. The MyD88-independent pathway was further characterized by determining the genes expressed in MyD88-deficient macrophages following exposure to LPS46. A number of genes known to be interferon (IFN)-inducible genes were identified, such as glucocorticoid-attentuated response gene 16 (GARG16), immunoresponsive gene 1 (IRG1) and the gene encoding CXC-chemokine ligand 10 (CXCL10; the product of which is also known as IFN--inducible 10 kDa protein, IP10). As expected, genes encoding inflammatory cytokines, such as TNF, IL-6 and IL-1, were not expressed. Induction of the IFN-inducible genes was completely abolished in TLR4-deficient macrophages, demonstrating that CXCL10, GARG-16 and IRG1 are produced in a TLR4-dependent but MyD88-independent manner. By contrast, stimulation with TLR2 ligands was found not to upregulate the expression of IFN-inducible genes, which is consistent with the idea that TLR2 does not use this MyD88-independent pathway. Further studies that used mice deficient in the IFN-/ receptor subsequently showed that the production of CXCL10 in response to LPS is mainly a secondary consequence of IFN- production47,48. In addition to inducing the expression of IFNinducible genes, the MyD88-independent pathway leads to the LPS-mediated maturation of dendritic cells (DCs)49. When cultured with LPS, MyD88-deficient bone-marrow-derived DCs upregulate the cell-surface expression of co-stimulatory molecules, such as CD40, CD80 and CD86, and induce the proliferation of T cells. By contrast, TLR4-deficient DCs fail to mature in response to LPS, indicating that DC maturation proceeds in a MyD88-independent manner49. However, because stimulation of wild-type DCs, with either TLR2 ligands (through the MyD88-dependent pathway) or TLR4 ligands (through the MyD88-dependent and -independent pathways), was observed to increase the cell-surface expression of co-stimulatory molecules, then either the MyD88-dependent or -independent
VOLUME 4 | JULY 2004 | 5 0 3
REVIEWS
pathway is sufficient for the induction of co-stimulatory molecules. For TLR4, the MyD88-independent production of co-stimulatory molecules is mainly secondary to the production of IFN-47. By contrast, the TLR-mediated expression of genes that encode inflammatory cytokines is controlled by the MyD88-dependent pathway, although both the MyD88-dependent and -independent pathways are involved in TLR4-mediated production of inflammatory cytokines50,51. Therefore, the expression of the genes that encode inflammatory cytokines and co-stimulatory molecules is differentially regulated during TLR signalling. Although MyD88 has been reported to be involved in TLR3 signalling52, TLR3 seems to transduce its signals mainly through the MyD88-independent pathway, because stimulation with the TLR3 ligand polyinosinicpolycytidylic acid (poly I:C) does not result in impaired production of inflammatory cytokines and co-stimulatory molecules in MyD88-deficient mice (S.A and K.T., unpublished observations).
Adaptor family
MyD88 TIRAP TRAM TRIF
TLR5/7/9
TLR1/6
TLR2
CD14 MD2
TLR4
TLR3
IKK- IRAK1 TBK1
TRAF6
IKK- IKK- IKK-
IRF3
The discovery of the MyD88-independent pathway led researchers to characterize the signalling pathways of the various TLRs, the activation of which leads to different patterns of gene expression. As a result, the molecular mechanisms underlying such differences can now be explained, at least in part, by the existence of several adaptors, which are used by different TLRs. These adaptors, which all have TIR domains, include the following (in order of identification): MyD88; TIRAP (TIR-domain-containing adaptor protein; also known as MyD88-adaptor-like protein, MAL); TRIF (TIR-domain-containing adaptor protein inducing IFN-; also known as TIR-domain-containing molecule 1, TICAM1); and TRAM (TRIF-related adaptor molecule; also known as TIR-domain-containing molecule 2, TICAM2) (FIG. 3). TIRAP. Identification of the MyD88-independent pathway of TLR4 signalling led to the discovery of the second TIR-domain-containing adaptor, which is known as TIRAP53,54. Unlike MyD88, TIRAP does not have a DD and was initially thought to mediate the MyD88-independent pathway of TLR4 signalling. However, the physiological role of TIRAP was revealed by generating knockout mice: the production of inflammatory cytokines in response to LPS was found to be defective in TIRAP-deficient mice, but the expression of IFN-inducible genes and the delayed activation of NF-B was still observed54,55. This phenotype was similar to that of MyD88-deficient mice, and it indicated that TIRAP is essential for the TLR4mediated, MyD88-dependent signalling pathway but not the MyD88-independent pathway. The possibility that MyD88 and TIRAP might function redundantly in the MyD88-independent pathway was excluded by generating mice deficient in both MyD88 and TIRAP. Enforced overexpression of MyD88 in TIRAP-deficient embryonic fibroblasts, but not vice versa, activates the NF-B-dependent promoter, indicating that TIRAP probably acts upstream of MyD88 (S.A. and K.T., unpublished observations). Interestingly, TIRAP-deficient mice also show impaired cytokine production in response to TLR2 ligands, despite having normal responses to TLR3, TLR7 and TLR9 ligands55,56. Therefore, TIRAP is essential for MyD88-dependent signalling through TLR2 and TLR4.
IB p50 p65 Nuclear membrane
NF-B
NF-B-binding motif
Figure 3 | Involvement of TIR-domain-containing adaptors in TLR-signalling pathways. The Toll/interleukin-1 (IL-1)-receptor (TIR)-domain-containing adaptor molecule MyD88 (myeloid differentiation primary-response protein 88) mediates the Toll-like receptor (TLR)-signalling pathway that activates IRAKs (IL-1-receptor-associated kinases) and TRAF6 (tumour-necrosis-factorreceptor-associated factor 6), and leads to the activation of the IKK complex (inhibitor of nuclear factor-B (IB)- kinase complex), which consists of IKK-, IKK- and IKK- (also known as IKK1, IKK2 and nuclear factor-B (NF-B) essential modulator, NEMO, respectively). This pathway is used by TLR1, TLR2, TLR4, TLR5, TLR6, TLR7 and TLR9 and releases NF-B from its inhibitor so that it translocates to the nucleus and induces expression of inflammatory cytokines. TIRAP (TIRdomain-containing adaptor protein), a second TIR-domain-containing adaptor protein, is involved in the MyD88-dependent signalling pathway through TLR2 and TLR4. By contrast, TLR3- and TLR4-mediated activation of interferon (IFN)-regulatory factor 3 (IRF3) and the induction of IFN- are observed in a MyD88-independent manner. A third TIR-domain-containing adaptor, TRIF (TIRdomain-containing adaptor protein inducing IFN-), is essential for the MyD88-independent pathway. The non-typical IKKs IKK- and TBK1 (TRAF-family-member-associated NF-B activator (TANK)-binding kinase 1) mediate activation of IRF3 downstream of TRIF. A fourth TIR-domaincontaining adaptor, TRAM (TRIF-related adaptor molecule), is specific to the TLR4-mediated, MyD88-independent/TRIF-dependent pathway.
504

TRIF. Because analyses of TIRAP-deficient mice indicated that TIR-domain-containing molecules might mediate the specificity of different TLR signalling pathways, further database searches for such proteins were conducted, leading to the identification of a third TIR-domain-containing adaptor, TRIF57. The same molecule was also identified as a TLR3-binding molecule in a yeast two-hybrid screen, but in this report, it was called TICAM158. The enforced expression of TRIF, but not of MyD88 or TIRAP, led to activation of the IFN- promoter in HEK293 (human embryonic kidney 293) cells, whereas a dominant-negative form of TRIF inhibited TLR3-dependent activation of the IFN- promoter. These in vitro studies indicate that TRIF functions in the MyD88-independent pathway to induce IFN-. The physiological role of TRIF was subsequently revealed through the targeted deletion of Trif in mice. In response to TLR3 and TLR4 ligands, these TRIF-deficient mice showed both impaired activation of IFN-regulatory factor 3 (IRF3) and decreased expression of IFN-inducible genes50. Consistent with this, analysis of LPS-hyporesponsive mice, which were generated by random germline mutagenesis, also led to the identification of Trif as a gene responsible for TLR3- and TLR4-mediated responses51. Therefore, studies that used two independently generated strains of Trif-mutant mice demonstrated that TRIF is essential for the TLR3- and TLR4-mediated activation of the MyD88-independent pathway, which subsequently leads to the production of IFN-. In addition, TRIF-deficient mice showed defective production of inflammatory cytokines in response to TLR4 ligands but not to other TLR ligands. However, TLR4-mediated activation of the MyD88-dependent pathway was not impaired, as determined by phosphorylation of IRAK1 and early-phase activation of NF-B50,51. These findings indicate that TLR4 requires both MyD88dependent and MyD88-independent/TRIF-dependent signals to induce the expression of inflammatory cytokines. By contrast, activation of the MyD88dependent pathway alone is sufficient to induce inflammatory-cytokine production in response to the ligation of TLR2, TLR5, TLR7 or TLR9 none of which activate the MyD88-independent/TRIF-dependent pathway. It remains unclear why TLR4-mediated signalling pathways use both MyD88-dependent and MyD88-independent pathways to induce the expression of inflammatory cytokines. However, these findings might indicate that, as well as NF-B, an unidentified molecule or molecules activated by the MyD88independent/TRIF-dependent pathway is required for inflammatory-cytokine induction. TRAM. A fourth TIR-domain-containing adaptor, TRAM, was recently identified through sequence homology in database searches5962. In vitro studies indicated that TRAM associates with TRIF and TLR4 but not with TLR3 (REFS 61,62), and the inhibition of TRAM expression by siRNA demonstrated its important role in the TLR4- but not TLR3-mediated induction of IFN- and IFN-inducible genes61,62. Analysis of TRAM-deficient mice further established that TRAM has an essential role in the MyD88-independent cascade of TLR4-induced signals. In response to TLR4 ligands, TRAM-deficient mice, similar to TRIF-deficient mice, showed impaired activation of IRF3 and reduced expression of IFN-inducible genes. However, unlike TRIF-deficient mice, TRAM-deficient mice showed a normal response to TLR3 stimulation60. So, TRAM is involved specifically in the activation of the MyD88independent/TRIF-dependent signalling pathway through TLR4. In addition, similar to TRIF-deficient mice, TRAM-deficient mice are defective in their production of inflammatory cytokines in response to LPS, despite the fact that the activation of IRAK1 and earlyphase NF-B is normal60. This indicates that TRAM and TRIF are involved in the TLR4-mediated induction of inflammatory cytokines, although the precise mechanisms remain unknown. Differential use of adaptors in TLR signalling. The characterization of TIR-domain-containing adaptors has established the essential roles of these adaptors in TLR signalling (FIG. 3). MyD88 is essential for all TLR-mediated production of inflammatory cytokines. However, stimulation of TLR3 or TLR4 results in induction of type I IFNs (IFN-/) in a MyD88-independent manner. This MyD88-independent response is entirely dependent on TRIF. In addition, TIRAP is involved specifically in TLR2and TLR4-mediated activation of the MyD88-dependent pathway, and TRAM is a specific adaptor in the TLR4mediated, MyD88-independent/TRIF-dependent pathway. Therefore, TIRAP and TRAM provide the specificity for different TLR-signalling pathways. Because the cytoplasmic portion of TLR4 binds directly to TRAM but not TRIF, as shown by in vitro experiments, TLR4 might acquire the ability to induce type I IFNs by associating with TRAM, which bridges TLR4 and TRIF62. Interestingly, all of these adaptors are involved in the TLR4-signalling pathway; however, it remains unknown why only TLR4 requires all of the TIRdomain-containing adaptors to induce gene expression. But this use of various adaptors, and the synergistic activation of both the MyD88-dependent and MyD88independent/TRIF-dependent pathways, might explain why the TLR4 ligand LPS is such a strong immunostimulator, sufficient to induce ENDOTOXIC SHOCK. In contrast to the induction of type I IFNs through TLR3 and TLR4, TLR7 and TLR9 mediate the production of type I IFNs through a MyD88-dependent signalling cascade. However, it remains unclear which molecule or molecules provides specificity in these signalling pathways. There is one further TIR-domain-containing molecule, which is known as SARM (sterile - and armadillo-motif-containing protein)63. An orthologue of mammalian SARM, the Caenorhabditis elegans TIRdomain-containing protein (TIR1), has recently been shown to mediate the expression of genes that encode antimicrobial peptides. However, this response is independent of the C. elegans TLR64. Nonetheless, elucidation of the role of mammalian SARM might improve our understanding of TLR signalling.
ENDOTOXIC SHOCK
A serious systemic disorder that leads to multiple organ failure and death. It is caused by an excessive release of lipopolysaccharide (also known as endotoxin) during Gramnegative bacterial infection.
VOLUME 4 | JULY 2004 | 5 0 5
REVIEWS
IRF3 in response to viral infection and stimulation of TLR3 (REFS 71,72). Overexpression of IKK- or TBK1 activates the promoter of the IFN- gene and of IFNinducible genes and induces the phosphorylation and nuclear localization of IRF3. In addition, siRNA targeting of IKK- and TBK1, but not of IKK-, was observed to considerably reduce the TRIF-dependent activation of a reporter gene containing an IRF DNA-binding motif and to decrease the viral induction of IFN-4 and IFN- reporter genes71,72. Furthermore, analysis of TBK1deficient embryonic fibroblasts confirmed the essential role of TBK1 in IRF3-dependent gene expression mediated by TLR3 and TLR4 signalling73. NF-B activation. MyD88-deficient mice still activate NF-B in response to stimulation with LPS, although the kinetics are delayed compared with wild-type mice. However, the activation of NF-B in response to LPS is completely abrogated in mice lacking both MyD88 and TRIF, demonstrating that TRIF is essential for NF-B activation through the MyD88-independent pathway50. Transcriptional activation of the IFN- gene requires activation of both NF-B and IRF3, and in contrast to inflammatory cytokine production which requires both early- and late-phase NF-B activation, mediated by MyD88 and TRIF, respectively production of IFN- can be induced by TRIF-mediated late-phase NF-B activation alone (FIG. 4). In vitro analyses showed that both the N-terminal and C-terminal regions of TRIF can independently activate an NF-B-responsive promoter, whereas only the N-terminal region is involved in the activation of the IFN- promoter57. The mechanism of NF-B activation through the N-terminal region of TRIF was further studied in a yeast two-hybrid screen, which led to the finding that TRAF6 physically interacts with TRIF74. The TRAF-C domain of TRAF6 is reported to bind to a consensus motif P-X-E-X-X-(D/E/F/W/Y) as discussed earlier31. Interestingly, both human and mouse TRIF contain three TRAF6-binding motifs in the N-terminal region. Although mutating each of the three TRAF6-binding motifs (TRIF3A mutant) abolished the association of TRIF with TRAF6, activation of NF-B by this TRIF3A mutant was only partially reduced. The activation that still occurs is probably a result of the C-terminal region of TRIF, which activates NF-B independently of the N-terminal region: indeed, a version of the TRIF3A mutant that lacks the C-terminal region was found to lose its ability to activate NF-B. In addition, it was shown that TBK1, which activates IRF3 and thereby induces IFN-, associates with the N-terminal region of TRIF72,74. Therefore, the N-terminal region of TRIF directly associates with TRAF6 and TBK1, leading to activation of NF-B and the IFN- gene, respectively (FIG. 4). Furthermore, TRIF uses at least two pathways for NF-B activation. The first involves its N-terminal region and is mediated by TRAF6, and the second involves its C-terminal region. A recent study indicates that TRIF-dependent NF-B activation is dependent on receptor-interacting protein 1 (RIP1), which associates with the C-terminus of TRIF75.
MyD88-independent/TRIF-dependent pathway
IRF3 activation. Previous studies have shown that activation of the gene encoding IFN- and the IFNinducible genes requires IRF3 (REF. 65). The IRFs are a family of transcription factors that are involved both in the induction of type I IFNs and in the response to IFNs66. So far, of the nine known members of the IRF family, IRF3, IRF5 and IRF7 have been shown to function as direct transducers of virus-mediated signalling and have a crucial role in the expression of type I IFNs67. IRF3 is expressed constitutively by various cells, and in response to viral infection, its C-terminal regulatory domain is activated by phosphorylation, which allows the formation of IRF3 dimers. After dimerization, IRF3 translocates rapidly to the nucleus, where despite lacking intrinsic transcriptional activity, it activates transcription of the type I IFN genes by recruiting the co-activators p300 and CBP (cAMPresponsive-element-binding protein (CREB)-binding protein). IRF3 mediates the initial induction of type I IFNs during viral infection, and these secreted type I IFNs activate the expression of IFN-inducible genes, such as CXCL10 and IRG1, through the JAK (Janus activated kinase)STAT (signal transducer and activator of transcription) signalling pathway68. In contrast to IRF3, mRNA that encodes IRF7 is produced in most cell types in response to IFN and viral infection. During viral infection, the activation of the constitutively expressed IRF3 molecules and the consequent production of type I IFNs leads to the induction and activation of IRF7 through the JAKSTAT signalling pathway. Subsequently, both IRF3 and IRF7 are involved in the production of delayed-type IFNs (IFN-/), thereby amplifying the expression of IFNs69. Stimulation with LPS also activates IRF3, and because LPS can induce the expression of both IFN- and IFN-inducible genes in a MyD88-independent manner, this indicates that IRF3 activation does not require MyD88. In fact, IRF3 activation, as shown by dimer formation and nuclear translocation, can be observed in MyD88-deficient cells47. The activation of IRF3 by signalling through TLR3 is more rapid and potent than that triggered by TLR4 signalling, and this correlates with increased production of IFN- (S.A. and K.T., unpublished observations). Whereas activating TLR3 with poly I:C results in the C-terminal phosphorylation of IRF3, as detected using a phosphospecific antibody, stimulation with LPS does not induce detectable C-terminal phosphorylation 70. However, we think that TLR3and TLR4-mediated IRF3 activation probably differ quantitatively, rather than qualitatively, because it is more plausible that the phosphospecific antibody is not sensitive enough to detect IRF3 phosphorylation following TLR4 activation. Two IKK-related proteins IKK- (also known as inducible IKK, IKKi) and TBK1 (TRAF-family-memberassociated NF-B activator (TANK)-binding kinase 1; also known as NF-B-activating kinase, NAK) have recently been identified as the kinases that phosphorylate
506

(REF. 77). In resting cells, TOLLIP forms a complex with members of the IRAK family, thereby preventing NF-B activation by blocking the phosphorylation of IRAK1. After receptor activation, TOLLIPIRAK1 complexes are recruited to the receptor, which results in the rapid autophosphorylation of IRAK1 and its dissociation from the receptor. At the same time, IRAK1 phosphorylates TOLLIP, which might then lead to the dissociation of TOLLIP from IRAK1 and its rapid ubiquitylation and degradation. TOLLIP is therefore thought to function mainly to maintain immune cells in a quiescent state and to facilitate the termination of TLR/IL-1R-induced cell signalling during inflammation and infection.
TLR3
RIP1 TRIF TBK1 TRAF6
IB p50 p65 IRF3
NF-B
Nuclear membrane
NF-B-binding motif
Figure 4 | TRIF-dependent induction of IFN-. The aminoterminal region of TRIF (Toll/interleukin-1-receptor (TIR)domain-containing adaptor protein inducing interferon (IFN)-) interacts with both TRAF6 (tumour-necrosis-factor-receptorassociated factor 6) and TBK1 (TRAF-family-memberassociated nuclear factor-B (NF-B) activator (TANK)-binding kinase 1). TRIF-dependent activation of TBK1 leads to the phosphorylation of IRF3 (IFN-regulatory factor 3), and TRAF6 mediates NF-B activation. RIP1 (receptor-interacting protein 1) mediates the NF-B activation that is induced through the carboxy-terminal region of TRIF. Activation of both NF-B and IRF3 contributes to the activation of the IFN- gene. IB, inhibitor of NF-B; TLR, Toll-like receptor.
Pellino. Pellino was first identified in Drosophila as a protein that binds to Pelle, the Drosophila homologue of the IRAKs78. Three mammalian homologues of Pellino have since been identified pellino-1, pellino-2 and pellino-3 and these show a high degree of evolutionary conservation7981: human pellino-2 is 60% identical to Drosophila Pellino. Mammalian pellino-1 and -2 interact with IRAK1 and have been shown to be required for NF-B activation in the TLR/IL-1R-signalling pathways79,80. Under steady-state conditions, IRAK1 and pellino-2 remain separate, but following TLR/IL-1R stimulation, they form a complex79. Because of their ability to interact with IRAKs and their lack of any domain capable of enzymatic activity, it is likely that the pellinos function as scaffolding proteins that facilitate the release of phosphorylated IRAK from the receptor. PI3K. PI3Ks are activated during TLR/IL-1R signalling, as a result of the direct interaction of the PI3K p85 regulatory subunit with the receptor82. This interaction involves the SRC homology 2 (SH2) domain of the p85 subunit and a domain in the receptor containing the motif Tyr-Xaa-Xaa-Met. The subsequent association of the p110 catalytic subunit of PI3Ks results in complete activation, leading to the phosphorylation and activation of its downstream target, AKT. Interestingly, the PI3K-binding motif Tyr-XaaXaa-Met, where Xaa denotes any amino acid, is present only in a subset of TLRs: TLR1, TLR2 and TLR6, but not TLR3, TLR4 or TLR5 (REF. 83). However, a putative PI3K-binding site (Tyr257-Lys258-Ala259-Met260) is present in the C-terminus of MyD88, and LPS stimulation has been shown to result in the tyrosine phosphorylation of MyD88 and the formation of a PI3KMyD88 complex84. MyD88 also interacts directly with AKT, and a dominant-negative mutant of AKT causes a defect in MyD88-dependent NF-B transcriptional activity. However, the binding of NF-B to DNA is not affected by inhibiting AKT, indicating that AKT might be involved in the phosphorylation of the p65 transactivation domain. A dominant-negative mutant of MyD88 was shown to block the kinase activity of AKT generated in response to LPS and IL-1, and a dominant-negative mutant of p85 inhibited the NF-B activity elicited by LPS and IL-1 but not that elicited by TNF85. These findings indicate that PI3K is a positive mediator of the signalling induced by LPS and IL-1 that
This region of TRIF contains a RIP homotypicinteraction motif, which is required for association with RIP1. A dominant-negative form of RIP1 inhibits TRIF-mediated NF-B activation, and embryonic fibroblasts from RIP1-deficient mice showed impaired TLR3-mediated NF-B activation. So, RIP1 probably mediates NF-B activation through the C-terminal region of TRIF.
Other molecules involved in TLR signalling
After ligand binding, TLRs activate various intracellular signalling molecules in addition to those discussed earlier. These include Toll-interacting protein (TOLLIP), the pellinos, phosphatidylinositol 3-kinase (PI3K), AKT (also known as protein kinase B, PKB), evolutionarily conserved signalling intermediate in Toll pathways (ECSIT), the SRC-family tyrosine kinases and MAPKKKs. These molecules are potentially involved in TLR-signalling pathways and are discussed briefly here. TOLLIP. Originally, TOLLIP was cloned as a protein that interacts with the IL-1R accessory protein76. Subsequently, it has been shown to associate directly with the cytoplasmic TIR domain of IL-1Rs, TLR2 and TLR4, following the stimulation of these receptors, and to inhibit TLR-mediated cellular responses by suppressing the phosphorylation and kinase activity of IRAK1
VOLUME 4 | JULY 2004 | 5 0 7
REVIEWS
leads to NF-B activation. However, recent studies using mice that lack the p85 regulatory subunit showed an increased production of IL-12 by DCs, possibly because of enhanced activation of the p38 MAPK, indicating that PI3K might have a negative role in TLR signalling in DCs86. ECSIT. ECSIT has no homology with any known protein and was cloned as a TRAF6-interacting protein by yeast two-hybrid screening87. ECSIT interacts with the conserved TRAF domain of TRAF6. A Drosophila homologue of ECSIT has been identified, and the interaction between TRAF6 and ECSIT is also conserved in Drosophila. ECSIT also interacts with MEKK1 (MAPK/ERK (extracellular signal-regulated kinase) kinase kinase 1), which can phosphorylate and activate the IKK complex. Expression of a dominantnegative mutant of ECSIT blocks signalling through TLR4, indicating that ECSIT might transduce TLR signals by bridging TRAF6 and the IKK complex. Furthermore, the inhibition of ECSIT expression, using siRNA in a macrophage cell line, resulted in impaired LPS-induced, but not TNF-induced, NF-B activation88. The physiological function of ECSIT was studied by generating ECSIT-deficient mice, which were found to die on about embryonic day 7.5 (REF. 88). Further characterization showed that ECSIT is an obligatory intermediate in bone morphogenetic protein (BMP) signalling, and therefore ECSIT is an essential component in both the TLR- and BMP-signalling pathways. SRC family of tyrosine kinases. Brutons tyrosine kinase (BTK) is a member of the SRC-related TEC-family of protein tyrosine kinases and has an essential role in B-cell development and B-cell receptor (BCR)-mediated signalling. Macrophages from X-linked-immunodeficient mice that lack BTK show reduced responses to LPS, and BTK has also been found to associate with the TIR domain of TLR4,TLR6, TLR8 and TLR9 (REF. 89). It has also been shown to associate with MyD88, TIRAP and IRAK1, and to be tyrosine phosphorylated in response to LPS, whereas a dominant-negative form of BTK inhibits LPS-induced activation of NF-B, indicating that BTK is involved in the TLR-mediated signalling pathway. During BCR-mediated signalling, BTK interacts with and is activated by the SRC family of tyrosine kinases, such as FYN, LYN and haematopoietic-cell kinase, HCK; however, SRC-family kinases have only a minor role in LPS signalling90. Therefore, the involvement of BTK in TLR signalling needs to be further investigated. MAPKKK. Members of the MAPKKK family such as TAK1, MEKK1, MEKK2, MEKK3, TPL2 (tumourprogression locus 2; also known as cancer Osaka thyroid, COT) and NIK are implicated in IKKNF-B and MAPK activation. Among these members, MEKK3 has been shown to be involved in signalling through TLR4 but not through TLR9 (REF. 91): in response to stimulation with a TLR4 ligand but not a TLR9 ligand, embryonic fibroblasts from MEKK3-deficient mice were shown to have impaired IL-6 production and defective activation of NF-B, JNK and the p38 MAPK. Stimulation of TLR4 also induced association of MEKK3 with TRAF6. So, MEKK3 is involved in the TLR4-mediated signalling pathway. Another member of the MKKK family, TPL2 has been shown to be involved in the TLR4-mediated activation of ERK92. In response to TLR4 ligand, TPL2-deficient mice showed impaired TNF production and defective activation of ERK. Taken together, it is clear that several MAPKKKs mediate TLR-signalling pathways.
SIGIRR
ST2 CD14 TLR4 MD2 TIRAP MyD88s MyD88
IRAK1
TRAF6
IKK- IKK- IKK-
IB p50 p65
NF-B
Nuclear membrane
NF-B-binding motif
Figure 5 | Negative regulation of TLR signalling. Toll-like receptor (TLR)-signalling pathways are negatively regulated by several molecules that are induced by the stimulation of TLRs. IRAK-M (interleukin-1-receptor (IL-1R)-associated kinase M) inhibits the dissociation of the IRAK1IRAK4 complex from the receptor. SOCS1 (suppressor of cytokine signalling 1) probably associates with IRAK1 and inhibits its activity. MyD88s (myeloid differentiation primary-response protein 88 short) blocks the association of IRAK4 with MyD88. The TIR (Toll/IL-1R)-domaincontaining receptors SIGIRR (single immunoglobulin IL-1Rrelated molecule) and ST2 have also been shown to negatively modulate TLR signalling. IB, inhibitor of NF-B; IKK, IB kinase; NF-B, nuclear factor-B; TIRAP, TIR-domaincontaining adaptor protein; TRAF6, tumour-necrosis-factorreceptor-associated factor 6.
508
IRAK-M
SOCS1
IRAK4

IRAK1 (REF. 98), the precise mechanism by which SOCS1 inhibits TLR signalling remains unclear. MyD88s, an alternatively spliced variant of MyD88 that lacks the intermediary domain, is induced in monocytes following stimulation with LPS. Unlike MyD88, MyD88s does not bind IRAK4, and overexpression of MyD88s does not induce IRAK1 phosphorylation. Therefore, MyD88s inhibits LPS-induced NF-B activation because of its inability to bind to IRAK4 and promote IRAK1 phosphorylation99. In addition to these cytoplasmic molecules, the negative effects of which are induced by TLR signalling, membrane-bound molecules that contain a TIR domain such as SIGIRR and ST2 have recently been shown to be involved in the negative regulation of TLR signalling. SIGIRR-deficient mice were found to be highly sensitive to LPS-induced endotoxic shock100. Following TLR stimulation, SIGIRR has also been shown to interact transiently with TLR4, IRAK1 and TRAF6, thereby negatively regulating TLR-signalling pathways. Similarly, ST2-deficient mice showed increased production of inflammatory cytokines in response to LPS; moreover, they also showed defective induction of LPS tolerance101. Overexpression of ST2 was found to inhibit NF-B activation, because ST2 associated with, and probably sequestered, MyD88 and TIRAP. Therefore, TIR-domain-containing orphan receptors, such as SIGIRR and ST2, are implicated in the negative regulation of TLR signalling.
Conclusions and future prospects
LPS TOLERANCE
Negative regulation of TLR signalling
A transient state of hyporesponsiveness to subsequent stimulation with lipopolysaccharide (LPS), which is induced by the administration of Toll-like receptor ligands in vivo and in vitro.
The inflammatory cytokines produced as a result of TLR signalling, when released in excess, induce serious systemic disorders that are associated with a high mortality rate such as endotoxic shock, which can be induced by the TLR4 ligand LPS. It is therefore not surprising that organisms have evolved mechanisms for modulating their TLR-mediated responses (FIG. 5). The molecules thought to negatively regulate TLR signalling are discussed briefly here; these include IRAK-M, SOCS1 (suppressor of cytokine signalling 1), MyD88 short (MyD88s), SIGIRR (single immunoglobulin IL-1R-related molecule) and ST2. Unlike the other IRAKS, which are ubiquitously expressed, the expression of IRAK-M is restricted to monocytes and macrophages and increases following stimulation with TLR ligands. IRAK-M also lacks kinase activity22. In response to TLR ligands, IRAK-M-deficient mice show increased production of inflammatory cytokines and defective induction of LPS TOLERANCE93. Biochemical analysis has revealed that IRAK-M prevents the dissociation of the IRAK1 IRAK4 complex from MyD88, thereby preventing the formation of the IRAK1TRAF6 complex. These findings indicate that IRAK-M negatively regulates TLR-signalling pathways. SOCS1 is a member of the SOCS family of proteins, which are induced by cytokines and negatively regulate cytokine-signalling pathways 94. LPS and CpG-containing DNA have been shown to induce the expression of SOCS1 in macrophages95,96, and SOCS1deficient mice have been shown to be hypersensitive to LPS-induced endotoxic shock (that is, to show increased production of inflammatory cytokines)97,98. Furthermore, LPS tolerance was not induced in SOCS1-deficient mice and the ectopic introduction of SOCS1 into macrophages inhibited LPS-induced NF-B activation. These findings indicate that SOCS1 directly downmodulates TLR-signalling pathways. Although SOCS1 has been shown to associate with
Box 1 | TLR-independent recognition of microorganisms The intracellular recognition of certain pathogens seems to involve a Toll-like receptor (TLR)-independent system. The most well-characterized is the nucleotidebinding oligomerization domain (NOD) family of proteins, which includes NOD1 and NOD2 proteins that recognize the core structures of bacterial peptidoglycan in the cytoplasm107. In addition, TLR-independent mechanisms have been shown to be involved in the recognition of viral products. For example, TLR3-deficient mice retain responsiveness to double-stranded (ds) RNA, indicating that dsRNA is recognized by both TLR3-dependent and -independent mechanisms50,52,108. Furthermore, the introduction of dsRNA into the cytoplasm of dendritic cells leads to the induction of type I interferons (IFNs) (IFN-/) through a mechanism that is partly dependent on PKR (IFN-inducible dsRNA-dependent protein kinase) but independent of TLR3 (REF. 109). PKR was originally proposed to mediate the cellular recognition of and response to dsRNA; however, PKR-deficient mice do not show considerable impairment in their response to dsRNA or viral infection110111. It is still unclear whether PKR is involved in the TLR3-independent response to dsRNA, and it is possible that another molecule or molecules mediates this recognition. The generation of mice lacking both TLR3 and PKR will further elucidate the role of PKR in the recognition of dsRNA.
The molecular mechanisms by which the TLRs activate innate immunity are being elucidated by analysing mice that lack either individual TLRs or other molecules involved in TLR signalling. TIRdomain-containing adaptors, such as MyD88, TIRAP, TRIF and TRAM, have been found to have crucial roles in TLR-signalling pathways, because they provide specificity to the response generated by signalling through each TLR. However, several questions remain to be answered. For example, TLR7 and TLR9, but not TLR2, induce type I IFNs in a MyD88-dependent manner, indicating that TLR7 and TLR9 have a unique signalling pathway. It is possible that an additional TIR-domain-containing adaptor, SARM63, is involved in this TLR7/TLR9-mediated pathway. Alternatively, some TLR signalling might be regulated by molecules that do not contain a TIR domain, and elucidation of this unique TLR7/TLR9 pathway should improve our understanding of the mechanisms of the TLR-mediated activation of innate immunity. In addition, we are now able to study mice that are deficient in both MyD88 and TRIF, which therefore lack all of the TLR-signalling pathways that have been characterized so far. We now need to intensively analyse the role of TLR signalling in host defence against various infectious microorganisms. For example, MyD88-deficient mice have been shown to be sensitive to Gram-negative bacterial and Grampositive bacterial infections102, yet they still generate immune responses against intracellular bacteria
VOLUME 4 | JULY 2004 | 5 0 9
REVIEWS
(Listeria monocytogenes and mycobacteria)103106 and viruses. Consistent with this, TLR-independent mechanisms for the recognition of intracellular bacteria and viruses have been proposed (BOX 1). Analysis of mice that lack both MyD88 and TRIF should reveal the extent to which MyD88-dependent and TRIF-dependent/MyD88-independent pathways of TLR signalling contribute to host defence. Mutant mice that lack components of the TLR-signalling pathways should provide powerful models for the in vivo analysis of immune responses, host defence against infectious diseases and anticancer responses.
1. 2.
3. 4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15. 16.
17.
18.
19.
20.
21.
22.
23.
24.
Janeway, C. A. Jr & Medzhitov, R. Innate immune recognition. Annu. Rev. Immunol. 20, 197216 (2002). Beutler, B. & Rietschel, E. T. Innate immune sensing and its roots: the story of endotoxin. Nature Rev. Immunol. 3, 169176 (2003). Medzhitov, R. Toll-like receptors and innate immunity. Nature Rev. Immunol. 1, 135145 (2001). Akira, S., Takeda, K. & Kaisho, T. Toll-like receptors: critical proteins linking innate and acquired immunity. Nature Immunol. 2, 675680 (2001). Hashimoto, C., Hudson, K. L. & Anderson, K. V. The Toll gene of Drosophila, required for dorsalventral embryonic polarity, appears to encode a transmembrane protein. Cell 52, 269279 (1988). Lemaitre, B., Nicolas, E., Michaut, L., Reichhart, J.-M. & Hoffmann, J. A. The dorsoventral regulatory gene cassette spatzle/Toll/cactus controls the potent antifungal response in Drosophila adults. Cell 86, 973983 (1996). Slack, J. L. et al. Identification of two major sites in the type I interleukin-1 receptor cytoplasmic region responsible for coupling to pro-inflammatory signaling pathways. J. Biol. Chem. 275, 46704678 (2000). Xu, Y. et al. Structural basis for signal transduction by the Toll/interleukin-1 receptor domains. Nature 408, 111115 (2000). Poltorak, A. et al. Defective LPS signaling in C3H/HeJ and C57BL/10ScCr mice: mutations in Tlr4 gene. Science 282, 20852088 (1998). The first report that TLR4 is involved in the recognition of bacterial components. Bell, J. K. et al. Leucine-rich repeats and pathogen recognition in Toll-like receptors. Trends Immunol. 24, 528533 (2003). Ahmad-Nejad, P. et al. Bacterial CpG-DNA and lipopolysaccharides activate Toll-like receptors at distinct cellular compartments. Eur. J. Immunol. 32, 19581968 (2002). Heil, F. et al. The Toll-like receptor 7 (TLR7)-specific stimulus loxoribine uncovers a strong relationship within the TLR7, 8 and 9 subfamily. Eur. J. Immunol. 33, 29872997 (2003). Matsumoto, M. et al. Subcellular localization of Toll-like receptor 3 in human dendritic cells. J. Immunol. 171, 31543162 (2003). Latz, E. et al. TLR9 signals after translocating from the ER to CpG DNA in the lysosome. Nature Immunol. 5, 190198 (2004). Takeda, K., Kaisho, T. & Akira, S. Toll-like receptors. Annu. Rev. Immunol. 21, 335376 (2003). Dunne, A. & ONeill, L. A. The interleukin-1 receptor/Tolllike receptor superfamily: signal transduction during inflammation and host defense. Sci. STKE [online], 2003, re3 (2003). Lord, K. A., Hoffman-Liebermann, B. & Liebermann, D. A. Nucleotide sequence and expression of a cDNA encoding MyD88, a novel myeloid differentiation primary response gene induced by IL6. Oncogene 5, 10951097 (1990). Muzio, M., Ni, J., Feng, P. & Dixit, V. M. IRAK (Pelle) family member IRAK-2 and MyD88 as proximal mediators of IL-1 signaling. Science 278, 16121615 (1997). Wesche, H., Henzel, W. J., Shillinglaw, W., Li, S. & Cao, Z. MyD88: an adapter that recruits IRAK to the IL-1 receptor complex. Immunity 7, 837847 (1997). Burns, K. et al. MyD88, an adapter protein involved in interleukin-1 signaling. J. Biol. Chem. 273, 1220312209 (1998). Dunne, A., Ejdeback, M., Ludidi, P. L., ONeill, L. A. & Gay, N. J. Structural complementarity of Toll/interleukin-1 receptor domains in Toll-like receptors and the adaptors Mal and MyD88. J. Biol. Chem. 278, 4144341451 (2003). Janssens, S. & Beyaert, R. Functional diversity and regulation of different interleukin-1 receptor-associated kinase (IRAK) family members. Mol. Cell 11, 293302 (2003). A review of the functional roles of the IRAK family. Li, X. et al. Mutant cells that do not respond to interleukin-1 (IL-1) reveal a novel role for IL-1 receptor associated kinase. Mol. Cell. Biol. 19, 46434652 (1999). Li, S., Strelow, A., Fontana, E. J. & Wesche, H. IRAK-4: a novel member of the IRAK family with the properties of an IRAK-kinase. Proc. Natl Acad. Sci. USA 99, 55675572 (2002).
25. Kanakaraj, P. et al. Defective interleukin (IL)-18-mediated natural killer and T helper cell type 1 responses in IL-1 receptor-associated kinase (IRAK)-deficient mice. J. Exp. Med. 187, 20732079 (1998). 26. Thomas, J. A. et al. Impaired cytokine signaling in mice lacking the IL-1 receptor-associated kinase. J. Immunol. 163, 978984 (1999). 27. Swantek, J. L., Tsen, M. F., Cobb, M. H. & Thomas, J. A. IL-1 receptor-associated kinase modulates host responsiveness to endotoxin. J. Immunol. 164, 43014306 (2000). 28. Suzuki, N. et al. Severe impairment of interleukin-1 and Tolllike receptor signalling in mice lacking IRAK-4. Nature 416, 750756 (2002). 29. Picard, C. et al. Pyogenic bacterial infections in humans with IRAK-4 deficiency. Science 299, 20762079 (2003). 30. Bradley, J. R. & Pober, J. S. Tumor necrosis factor receptorassociated factors (TRAFs). Oncogene 20, 64826491 (2001). 31. Ye, H. et al. Distinct molecular mechanism for initiating TRAF6 signalling. Nature 418, 443447 (2002). 32. Yamaguchi, K. et al. Identification of a member of the MAPKKK family as a potential mediator of TGF- signal transduction. Science 270, 20082011 (1995). 33. Takaesu, G. et al. TAK1 is critical for IB kinase-mediated activation of the NF-B pathway. J. Mol. Biol. 326, 105115 (2003). 34. Shibuya, H. et al. TAB1: an activator of the TAK1 MAPKKK in TGF- signal transduction. Science 272, 11791182 (1996). 35. Takaesu, G. et al. TAB2, a novel adaptor protein, mediates activation of TAK1 MAPKKK by linking TAK1 to TRAF6 in the IL-1 signal transduction pathway. Mol. Cell 5, 649658 (2000). 36. Sanjo, H. et al. TAB2 is essential for prevention of apoptosis in fetal liver but not for interleukin-1 signaling. Mol. Cell. Biol. 23, 12311238 (2003). 37. Ishitani, T. et al. Role of the TAB2-related protein TAB3 in IL-1 and TNF signaling. EMBO J. 22, 62776288 (2003). 38. Deng, L. et al. Activation of the IB kinase complex by TRAF6 requires a dimeric ubiquitin-conjugating enzyme complex and a unique polyubiquitin chain. Cell 103, 351361 (2000). 39. Wang, C. et al. TAK1 is a ubiquitin-dependent kinase of MKK and IKK. Nature 412, 346351 (2001). 40. Karin, M. & Ben-Neriah, Y. Phosphorylation meets ubiquitination: the control of NF-B activity. Annu. Rev. Immunol. 18, 621663 (2000). 41. Senftleben, U. et al. Activation by IKK of a second, evolutionary conserved, NF-B signaling pathway. Science 293, 14951499 (2001). 42. Dejardin, E. et al. The lymphotoxin- receptor induces different patterns of gene expression via two NF-B pathways. Immunity 17, 525535 (2002). 43. Amir, R. E., Haecker, H., Karin, M. & Ciechanover, A. Mechanism of processing of the NF-B2 p100 precursor: identification of the specific polyubiquitin chain-anchoring lysine residue and analysis of the role of NEDD8-modification on the SCF (TrCP) ubiquitin ligase. Oncogene 23, 25402547 (2004). 44. Adachi, O. et al. Targeted disruption of the MyD88 gene resulted in impaired IL-1- and IL-18-mediated function. Immunity 9, 143150 (1998). 45. Kawai, T., Adachi, O., Ogawa, T., Takeda, K. & Akira, S. Unresponsiveness of MyD88-deficient mice to endotoxin. Immunity 11, 115122 (1999). Provides the first evidence for the existence of the MyD88-independent pathway. 46. Kawai, T. et al. Lipopolysaccharide stimulates the MyD88-independent pathway and results in activation of IFN-regulatory factor 3 and the expression of a subset of lipopolysaccharide-inducible genes. J. Immunol. 167, 58875894 (2001). 47. Hoshino, K., Kaisho, T., Iwabe, T., Takeuchi, O. & Akira, S. Differential involvement of IFN- in Toll-like receptorstimulated dendritic cell activation. Int. Immunol. 14, 12251231 (2002). 48. Toshchakov, V. et al. TLR4, but not TLR2, mediates IFN-induced STAT1/-dependent gene expression in macrophages. Nature Immunol. 3, 392398 (2002). 49. Kaisho, T., Takeuchi, O., Kawai, T., Hoshino, K. & Akira, S. Endotoxin-induced maturation of MyD88-deficient dendritic cells. J. Immunol. 166, 56885694 (2001).
50. Yamamoto, M. et al. Role of adaptor TRIF in the MyD88independent Toll-like receptor signaling pathway. Science 301, 640643 (2003). 51. Hoebe, K. et al. Identification of Lps2 as a key transducer of MyD88-independent TIR signalling. Nature 424, 743748 (2003). References 50 and 51 demonstrate that TRIF is an essential adaptor in the TLR3- and TLR4-mediated MyD88-independent pathway. 52. Alexopoulou, L., Holt, A. C., Medzhitov, R. & Flavell, R. A. Recognition of double-stranded RNA and activation of NF-B by Toll-like receptor 3. Nature 413, 732738 (2001). 53. Horng, T., Barton, G. M. & Medzhitov, R. TIRAP: an adapter molecule in the Toll signaling pathway. Nature Immunol. 2, 835841 (2001). 54. Fitzgerald, K. A. et al. Mal (MyD88-adaptor-like) is required for Toll-like receptor-4 signal transduction. Nature 413, 7883 (2001). 55. Yamamoto, M. et al. Essential role of TIRAP/Mal for activation of the signaling cascade shared by TLR2 and TLR4. Nature 420, 324329 (2002). 56. Horng, T., Barton, G. M., Flavell, R. A. & Medzhitov, R. The adaptor molecule TIRAP provides signaling specificity for Toll-like receptors. Nature 420, 329333 (2002). References 55 and 56 show that TIRAP is essential for the TLR2- and TLR4-mediated MyD88-dependent pathway. 57. Yamamoto, M. et al. A novel TIR domain-containing adaptor that preferentially activates the interferon- promoter. J. Immunol. 169, 66686672 (2002). 58. Oshiumi, H., Matsumoto, M., Funami, K., Akazawa, T. & Seya, T. TICAM-1, an adaptor molecule that participates in Toll-like receptor 3-mediated interferon- induction. Nature Immunol. 4, 161167 (2003). 59. Bin, L. H., Xu, L. G. & Shu, H. B. TIRP, a novel Toll/interleukin-1 receptor (TIR) domain-containing adapter protein involved in TIR signaling. J. Biol. Chem. 278, 2452624532 (2003). 60. Yamamoto, M. et al. TRAM is specifically involved in the Toll-like receptor 4-mediated MyD88-independent signaling pathway. Nature Immunol. 4, 11441150 (2003). 61. Fitzgerald, K. A. et al. LPSTLR4 signaling to IRF-3/7 and NF-B involves the Toll adapters TRAM and TRIF. J. Exp. Med. 198, 10431055 (2003). 62. Oshiumi, H. et al. TIR-containing adapter molecule (TICAM)-2: a bridging adapter recruiting to Toll-like receptor 4 TICAM-1 that induces interferon-. J. Biol. Chem. 278, 4975149762 (2003). References 60 to 62 show the involvement of the TIRdomain-containing adaptor TRAM in the TLR4mediated, MyD88-independent/TRIF-dependent pathway. 63. ONeill, L. A., Fitzgerald, K. A. & Bowie, A. G. The TollIL-1 receptor adaptor family grows to five members. Trends Immunol. 24, 286290 (2003). 64. Couillault, C. et al. TLR-independent control of innate immunity in Caenorhabditis elegans by the TIR domain adaptor protein TIR-1, an ortholog of human SARM. Nature Immunol. 5, 488494 (2004). 65. Yoneyama, M. et al. Direct triggering of the type I interferon system by virus infection: activation of a transcription factor complex containing IRF-3 and CBP/p300. EMBO J. 17, 10871095 (1998). 66. Taniguchi, T., Ogasawara, K., Takaoka, A. & Tanaka, N. IRF family of transcription factors as regulators of host defense. Annu. Rev. Immunol. 19, 623655 (2001). 67. Barnes, B., Lubyova, B. & Pitha, P. M. On the role of IRF in host defense. J. Interferon Cytokine Res. 22, 5971 (2002). 68. Taniguchi, T. & Takaoka, A. The interferon-/ system in antiviral responses: a multimodal machinery of gene regulation by the IRF family of transcription factors. Curr. Opin. Immunol. 14, 111116 (2002). 69. Sato, M. et al. Distinct and essential roles of transcription factors IRF-3 and IRF-7 in response to viruses for IFN-/ gene induction. Immunity 13, 539548 (2000). 70. Servant, M. J. et al. Identification of the minimal phosphoacceptor site required for in vivo activation of interferon regulatory factor 3 in response to virus and double-stranded RNA. J. Biol. Chem. 278, 94419447 (2003).
510

71. Sharma, S. et al. Triggering the interferon antiviral response through an IKK-related pathway. Science 300, 11481151 (2003). 72. Fitzgerald, K. A. et al. IKK and TBK1 are essential components of the IRF3 signaling pathway. Nature Immunol. 4, 491496 (2003). References 71 and 72 describe the identification of non-canonical IKKs as kinases that activate IRF3. 73. McWhirter, S. M. et al. IFN-regulatory factor 3-dependent gene expression is defective in Tbk1-deficient mouse embryonic fibroblasts. Proc. Natl Acad. Sci. USA 101, 233238 (2004). 74. Sato, S. et al. Toll/IL-1 receptor domain-containing adaptor inducing IFN- (TRIF) associates with TNF receptorassociated factor 6 and TANK-binding kinase 1, and activates two distinct transcription factors, NF-B and IFN-regulatory factor-3, in the Toll-like receptor signaling. J. Immunol. 171, 43044310 (2003). 75. Meylan, E. et al. RIP1 is an essential mediator of Toll-like receptor 3-induced NF-B activation. Nature Immunol. 5, 503507 (2004). 76. Burns, K. et al. Tollip, a new component of the IL-1RI pathway, links IRAK to the IL-1 receptor. Nature Cell Biol. 2, 346351 (2000). 77. Zhang, G. & Ghosh, S. Negative regulation of Toll-like receptor-mediated signaling by Tollip. J. Biol. Chem. 277, 70597065 (2002). 78. Grosshans, J., Schnorrer, F. & Nusslein-Volhard, C. Oligomerisation of Tube and Pelle leads to nuclear localisation of Dorsal. Mech. Dev. 81, 127138 (1999). 79. Yu, K. et al. Cutting edge: Mouse pellino-2 modulates IL-1 and lipopolysaccharide signaling. J. Immunol. 169, 40754078 (2002). 80. Jiang, Z. et al. Pellino 1 is required for interleukin-1 (IL-1)mediated signaling through its interaction with the IL-1 receptor-associated kinase 4 (IRAK4)IRAKtumor necrosis factor receptor-associated factor 6 (TRAF6) complex. J. Biol. Chem. 278, 1095210956 (2003). 81. Jensen, L. E. & Whitehead, A. S. Pellino3, a novel member of the pellino protein family, promotes activation of c-Jun and Elk-1 and may act as a scaffolding protein. J. Immunol. 171, 15001506 (2003). 82. Koyasu, S. The role of PI3K in immune cells. Nature Immunol. 4, 313319 (2003). 83. Arbibe, L. et al. Toll-like receptor 2-mediated NF-B activation requires a Rac1-dependent pathway. Nature Immunol. 1, 533540 (2000). 84. Ojaniemi, M. et al. Phosphatidylinositol 3-kinase is involved in Toll-like receptor 4-mediated cytokine expression in mouse macrophages. Eur. J. Immunol. 33, 597605 (2003). 85. Li, X. et al. Phosphoinositide 3 kinase mediates Toll-like receptor 4-induced activation of NF-B in endothelial cells. Infect. Immun. 71, 44144420 (2003). 86. Fukao, T. et al. PI3K-mediated negative feedback regulation of IL-12 production in DCs. Nature Immunol. 3, 875881 (2002). 87. Kopp, E. et al. ECSIT is an evolutionarily conserved intermediate in the Toll/IL-1 signal transduction pathway. Genes Dev. 13, 20592071 (1999). 88. Xiao, C. et al. Ecsit is required for BMP signaling and mesoderm formatio n during mouse embryogenesis. Genes Dev. 17, 29332949 (2003). 89. Jefferies, C. A. et al. Brutons tyrosine kinase is a Toll/interleukin-1 receptor domain-binding protein that participates in nuclear factor B activation by Toll-like receptor 4. J. Biol. Chem. 278, 2625826264 (2003). 90. Meng, F. & Lowell, C. A. Lipopolysaccharide (LPS)-induced macrophage activation and signal transduction in the absence of Src-family kinases Hck, Fgr, and Lyn. J. Exp. Med. 185, 16611670 (1997). 91. Huang, Q. et al. Differential regulation of interleukin 1 receptor and Toll-like receptor signaling by MEKK3. Nature Immunol. 5, 98103 (2003). 92. Dumitru, C. D. et al. TNF- induction by LPS is regulated posttranscriptionally via a Tpl2/ERK-dependent pathway. Cell 103, 10711083 (2000). 93. Kobayashi, K. et al. IRAK-M is a negative regulator of Toll-like receptor signaling. Cell 110, 191202 (2002). 94. Yasukawa, H., Sasaki, A. & Yoshimura, A. Negative regulation of cytokine signaling pathways. Annu. Rev. Immunol. 18,143164 (2000). 95. Stoiber, D. et al. Lipopolysaccharide induces in macrophages the synthesis of the suppressor of cytokine signaling 3 and suppresses signal transduction in response to the activating factor IFN-. J. Immunol. 163, 26402647 (1999). 96. Dalpke, A. H., Opper, S., Zimmermann, S. & Heeg, K. Suppressors of cytokine signaling (SOCS)-1 and SOCS-3 are induced by CpG-DNA and modulate cytokine responses in APCs. J. Immunol. 166, 70827089 (2001). 97. Kinjyo, I. et al. SOCS1/JAB is a negative regulator of LPS-induced macrophage activation. Immunity 17, 583591 (2002). 98. Nakagawa, R. et al. SOCS-1 participates in negative regulation of LPS responses. Immunity 17, 677687 (2002). 99. Burns, K. et al. Inhibition of interleukin 1 receptor/Toll-like receptor signaling through the alternatively spliced, short form of MyD88 is due to its failure to recruit IRAK-4. J. Exp. Med. 197, 263268 (2003). 100. Wald, D. et al. SIGIRR, a negative regulator of Toll-like receptorinterleukin 1 receptor signaling. Nature Immunol. 4, 920927 (2003). 101. Brint, E. K. et al. ST2 is an inhibitor of interleukin 1 receptor and Toll-like receptor 4 signaling and maintains endotoxin tolerance. Nature Immunol. 5, 373379 (2004). 102. Takeuchi, O., Hoshino, K. & Akira, S. Cutting edge: TLR2-deficient and MyD88-deficient mice are highly susceptible to Staphylococcus aureus infection. J. Immunol. 165, 53925396 (2000). 103. Sugawara, I., Yamada, H., Mizuno, S., Takeda, K. & Akira, S. Mycobacterial infection in MyD88-deficient mice. Microbiol. Immunol. 47, 841847 (2003). 104. Seki, E. et al. Critical roles of myeloid differentiation factor 88dependent proinflammatory cytokine release in early phase clearance of Listeria monocytogenes in mice. J. Immunol. 169, 38633868 (2002). 105. Way, S. S., Kollmann, T. R., Hajjar, A. M. & Wilson, C. B. Cutting edge: Protective cell-mediated immunity to Listeria monocytogenes in the absence of myeloid differentiation factor 88. J. Immunol. 171, 533537 (2003). 106. Serbina, N. V. et al. Sequential MyD88-independent and dependent activation of innate immune responses to intracellular bacterial infection. Immunity 19, 891901 (2003). 107. Inohara, N. & Nunez, G. NODs: intracellular proteins involved in inflammation and apoptosis. Nature Rev. Immunol. 3, 371382 (2003). 108. Hoebe, K. et al. Upregulation of costimulatory molecules induced by lipopolysaccharide and double-stranded RNA occurs by Trif-dependent and Trif-independent pathways. Nature Immunol. 4, 12231229 (2003). 109. Diebold, S. S. et al. Viral infection switches non-plasmacytoid dendritic cells into high interferon producers. Nature 424, 324328 (2003). 110. Abraham, N. et al. Characterization of transgenic mice with targeted disruption of the catalytic domain of the doublestranded RNA-dependent protein kinase, PKR. J. Biol. Chem. 274, 59535962 (1999). 111. Honda, K. et al. Selective contribution of IFN-/ signaling to the maturation of dendritic cells induced by doublestranded RNA or viral infection. Proc. Natl Acad. Sci. USA 100, 1087210877 (2003). 112. Takeuchi, O. et al. Cutting edge: Role of Toll-like receptor 1 in mediating immune response to microbial lipoproteins. J. Immunol. 169, 1014 (2002). 113. Wyllie, D. H. et al. Evidence for an accessory protein function for Toll-like receptor 1 in anti-bacterial responses. J. Immunol. 165, 71257132 (2002). 114. Aliprantis, A. O. et al. Cell activation and apoptosis by bacterial lipoproteins through Toll-like receptor 2. Science 285, 736739 (1999). 115. Takeuchi, O. et al. Differential roles of TLR2 and TLR4 in recognition of Gram-negative and Gram-positive cell wall components. Immunity 11, 443451 (1999). 116. Schwadner, R. et al. Peptidoglycan- and lipoteichoic acid-induced cell activation is mediated by Toll-like receptor 2. J. Biol. Chem. 274, 1740617409 (1999). 117. Means, T. K. et al. Human Toll-like receptors mediate cellular activation by Mycobacterium tuberculosis. J. Immunol. 163, 39203977 (1999). 118. Hajjar, A. M. et al. Cutting Edge: Functional interactions between Toll-like receptor (TLR) 2 and TLR1 or TLR6 in response to phenol-soluble modulin. J. Immunol. 166, 1519 (2001). 119. Coelho, P. S. et al. Glycosylphosphatidylinositol-anchored mucin-like glycoproteins isolated from Trypanosoma cruzi trypomastigotes induce in vivo leukocyte recruitment dependent on MCP-1 production by IFN--primedmacrophages. J. Leukoc. Biol. 71, 837844 (2002). 120. Opitz, B. et al. Toll-like receptor-2 mediates Treponema glycolipid and lipoteichoic acid-induced NF-B translocation. J. Biol. Chem. 276, 2204122047 (2001). 121. Massari, P. et al. Cutting edge: Immune stimulation by Neisserial porins is Toll-like receptor 2 and MyD88 dependent. J. Immunol. 168, 15331537 (2002). 122. Werts, C. et al. Leptospiral liopolysaccharide activates cells through a TLR2-dependent mechanism. Nature Immunol. 2, 346352 (2001). 123. Hirschfeld, M. et al. Signaling by Toll-like receptor 2 and 4 agonists results in differential gene expression in murine macrophages. Infect. Immun. 69, 14771482 (2001). 124. Underhill D. M. et al. The Toll-like receptor 2 is recruited to macrophage phagosomes and discriminates between pathogens. Nature 401, 811815 (1999). 125. Asea, A. et al. Novel signal transduction pathway utilized by extracellular HSP70: role of Toll-like receptor (TLR) 2 and TLR4. J. Biol. Chem. 277, 1502815034 (2002). 126. Kawasaki, K. et al. Mouse Toll-like receptor 4MD-2 complex mediates lipopolysaccharide-mimetic signal transduction by taxol. J. Biol. Chem. 275, 22512254 (2000). 127. Kurt-Jones, E. A. et al. Pattern recognition receptors TLR4 and CD14 mediate response to respiratory syncytial virus. Nature Immunol. 1, 398401 (2000). 128. Rassa, J. C. et al. Murine retroviruses activate B cells via interaction with Toll-like receptor 4. Proc. Natl Acad. Sci. USA 99, 22812286 (2002). 129. Bulut, Y. et al. Chlamydial heat shock protein 60 activates macrophages and endothelial cells through Toll-like receptor 4 and MD2 in a MyD88-dependent pathway. J. Immunol. 168, 14351440 (2002). 130. Ohashi, K., Burkart, V., Flohe, S. & Kolb, H. Cutting edge: Heat shock protein 60 is a putative endogenous ligand of the Toll-like receptor-4 complex. J. Immunol. 164, 558561 (2000). 131. Vabulas, R. M. et al. HSP70 as endogenous stimulus of the Toll/interleukin-1 receptor signal pathway. J. Biol. Chem. 277, 1510715112 (2002). 132. Okamura, Y. et al. The extra domain A of fibronectin activates Toll-like receptor 4. J. Biol. Chem. 276, 1022910233 (2001). 133. Termeer, C. et al. Oligosaccharides of hyaluronan activate dendritic cells via Toll-like receptor 4. J. Exp. Med. 195, 99111 (2002). 134. Johnson, G. B., Brunn, G. J., Kodaira, Y. & Platt, J. L. Receptor-mediated monitoring of tissue well-being via detection of soluble heparan sulfate by Toll-like receptor 4. J. Immunol. 168, 52335239 (2002). 135. Smiley, S. T., King, J. A. & Hancock, W. W. Fibrinogen stimulates macrophage chemokine secretion through Toll-like receptor 4. J. Immunol. 167, 28872894 (2001). 136. Hayashi, F. et al. The innate immune response to bacterial flagellin is mediated by Toll-like receptor-5. Nature 410, 10991103 (2001). 137. Takeuchi, O. et al. Discrimination of bacterial lipopeptides by Toll-like receptor 6. Int. Immunol. 13, 933940 (2001). 138. Ozinsky, A. et al. The repertoire for pattern recognition of pathogens by the innate immune system is defined by cooperation between Toll-like receptors. Proc. Natl Acad. Sci. USA 97, 1376613771 (2000). 139. Hemmi, H. et al. Small anti-viral compounds activate immune cells via the TLR7 MyD88-dependent signaling pathway. Nature Immunol. 3, 196200 (2002). 140. Heil, F. et al. Species-specific recognition of singlestranded RNA via Toll-like receptor 7 and 8. Science 303, 15261529 (2004). 141. Diebold, S. S., Kaisho, T., Hemmi, H., Akira, S. & Reis e Sousa, C. Innate antiviral responses by means of TLR7mediated recognition of single-stranded RNA. Science 303, 15291531 (2004). 142. Jurk, M. et al. Human TLR or TLR8 independently confer responsiveness to the antiviral compound R-848. Nature Immunol. 3, 499 (2002). 143. Hemmi, H. et al. A Toll-like receptor recognizes bacterial DNA. Nature 408, 740745 (2000). 144. Zhang, D. et al. A Toll-like receptor that prevents infection by uropathogenic bacteria. Science 303, 15221526 (2004).
Acknowledgements
We thank M. Lamphier for careful appraisal of the manuscript and M. Hashimoto for excellent secretarial assistance. This work was supported by grants from the Special Coordination Funds of the Japanese Ministry of Education, Culture, Sports, Science and Technology, the Uehara Memorial Foundation, Japan, and the Naito Foundation, Japan.
Competing interests statement

The authors declare that they have no competing financial interests.
Online links
DATABASES The following terms in this article are linked online to: Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene AKT | BTK | ECSIT | IFN- | IB | IKK- | IKK- | IKK- | IL-1R | IRAK1 | IRAK2 | IRAK4 | IRAK-M | IRF3 | MEKK3 | MyD88 | pellino-1 | pellino-2 | pellino-3 | PI3K | RIP1 | SARM | SIGIRR | SOCS1 | ST2 | TAB1 | TAB2 | TAK1 | TBK1 | TIRAP | TLR1 | TLR2 | TLR3 | Tlr4 | TLR5 | TLR7 | TLR9 | TOLLIP | TPL2 | TRAF6 | TRAM | TRIF | UBC13 Access to this interactive links box is free online.
VOLUME 4 | JULY 2004 | 5 1 1
REVIEWS
THERAPEUTICS TARGETING THE INNATE IMMUNE SYSTEM

Richard J. Ulevitch
Proteins that recognize the components and products of microorganisms have an important role in innate immunity. Here, I focus on recent advances in our understanding of the function of several such protein families. In particular, I consider how members of the TLR (Toll-like receptor), NOD (nucleotide-binding oligomerization domain)-protein and MyD88 (myeloid differentiation primaryresponse protein 88) families are providing emerging opportunities for the development of new therapeutics that modify innate immune responses in ways which benefit the host.
Department of Immunology, The Scripps Research Institute, La Jolla, California 92037, USA. e-mail: ulevitch@scripps.edu
doi:10.1038/nri1396
The modulation of immune responses achieved by targeting cell-surface receptors or intracellular pathways is one of the main goals in the development of new therapeutics for human immune or inflammatory diseases. Traditionally, the focus has been on targets that are considered to be part of the adaptive immune system. In this way, major efforts have been made to develop therapeutics for transplantation, autoimmunity and cancer. Now, the ever-expanding wealth of information about the innate immune system, including the identification of cognate ligands of innate immune receptors and the elucidation of their downstream signalling pathways, provides a new set of targets for drug development, which might impact on various human diseases19. The innate immune system sits at the intersection of the pathways of microbial recognition, inflammation, microbial clearance and cell death, thereby offering diverse targets for therapeutics (FIG. 1). This review limits its scope to gene families that encode the key receptors of the innate immune system the Toll-like receptor (TLR)- and nucleotide-binding oligomerization domain (NOD)-gene families their ligands and their downstream signalling pathways. Here, the term ligand is used to indicate a molecule that binds to its cognate receptor, although for TLRs and NOD proteins proof of the formation of a ligandreceptor complex is often lacking. TLRs comprise a family of cell-surface and endosomally expressed receptors that recognize conserved products unique to microorganisms, such as lipopolysaccharide (LPS),
leading to the activation of the innate and adaptive immune systems. The proteins that are encoded by the NOD genes (also known as members of the caterpillar family) are cytosolic proteins that have a role in various innate and adaptive immune responses. Of particular interest here are NOD1 and NOD2, which recognize distinct structures derived from peptidoglycan that are not ligands for TLRs1012. In mammals, the main function of the innate immune system is to detect the presence of invading microorganisms. Importantly, another concept of the function of innate immunity is emerging that it also has a role in sterile inflammation. Sterile inflammation is not driven by the response to infection with microorganisms but by ligands derived from damaged cells, which are not usually present in the extracellular environment: for example, heat-shock proteins, -defensins and oxidized lipids. It can also be speculated that proteins modified by oxidation or nitration might be TLR ligands and thereby might promote sterile inflammation. So, in both infection and sterile inflammation, the innate immune system uses receptors that include members of the TLR and NOD-protein families9. Ligands for these receptor families can be divided into several groups: naturally occurring molecules that are released from microbial sources, synthetic structures based on those of microbial products, small molecules with no obvious structural relationships to naturally occurring ligands and endogenous ligands of host origin. The downstream signalling pathways for the TLRs
512

are involved; however, here, only the steps that include the MyD88-adaptor family are considered. Ultimately, we need to define the signalling events and gene-activation patterns that occur downstream of each of the TLRs so that we can identify appropriate targets for the selective activation or inhibition of these pathways. In contrast to the TLRs, NOD1 and NOD2 seem to signal through pathways that are independent of members of the MyD88 family. Instead, after engagement of their cognate ligands, NOD proteins activate NF-B through a pathway that requires the kinase RICK (receptor-interacting serine/threonine kinase; also known as receptor-interacting protein 2, RIP2) 16,17. RICK and NOD1 or NOD2 interact through the caspase-recruitment domains (CARDs) present in each protein. Further details about other proteins involved in NOD-protein signalling are unknown at present. The central question then is how to use the current information about the signalling pathways of the innate immune system to design new therapeutics. The desired feature of any therapeutic that modifies the innate immune response either as an agonist or antagonist is, of course, specificity. In this review, examples of agonists and antagonists are provided to illustrate the potential utility of drugs targeting innate immunity. Complications that could arise from either blocking or activating innate immunity are also discussed, as well as issues that result from targeting innate immune pathways in acute and chronic settings.
TLR family
Microbial-clearance and killing pathways
Apoptotic and necrotic cell-death pathways
Innate immune system
Microbial-recognition pathways TLR Exogenous ligands NOD protein
Sterile-inflammation pathways TLR Endogenous ligands
Figure 1 | Innate immunity at the crossroads. The innate immune system is at the intersection of several pathways that influence the balance between health and disease. These pathways, which are responsible for the recognition of microorganisms and endogenous host-derived ligands, trigger the clearance and/or killing of microorganisms, as well as apoptotic and necrotic cell-death pathways that depend on pro-inflammatory mediators. NOD, nucleotidebinding oligomerization domain; TLR, Toll-like receptor.
and NOD proteins show one main difference. TLRs couple ligand binding to cell activation through members of the myeloid differentiation primary-response protein 88 (MyD88)-adaptor family (BOX 1), whereas NOD proteins do not seem to use MyD88-dependent signal-transduction pathways. In the signalling that occurs in an innate immune response, two central events are the activation of nuclear factor-B (NF-B) and interferon regulatory factor 3 (IRF3) . The NF-B pathway controls the production of pro-inflammatory molecules, such as interleukin-1 (IL-1) and tumournecrosis factor (TNF), whereas the IRF3 pathway leads to the production of type I interferons (IFN- and IFN-). In this review, I discuss interesting drug targets that have emerged from recent studies of TLRs and/or NOD proteins, without considering as targets the numerous downstream cytokines and other mediators that are produced following the activation of the innate immune system. Signalling from TLRs involves the Toll/IL-1 receptor (TIR) domain of the receptor interacting with one or more members of the MyD88-adaptor family, which also each contain a TIR domain. These interactions are known as TIRTIR domain interactions4,1315. The engagement of a TLR by its cognate ligand most probably facilitates the formation of a hetero- or homodimeric receptor, which then interacts with cytosolic TIR-domain-containing adaptor proteins. Numerous studies have investigated the signalling pathways immediately downstream of the TIRdomain interactions, showing how the molecules IRAK1 (IL-1 receptor (IL-1R)-associated kinase 1), IRAK4 and TRAF6 (TNF-receptor-associated factor 6)
Over the past five years, there have been remarkable advances in the identification and characterization of the TLRs. This information provides the basis for considering new ways of modulating the innate immune response by designing new therapeutic approaches (FIG. 2). The Drosophila studies of Hoffmann and colleagues 1821 provided the basis for the discovery of the TLRs, the mammalian homologues of the Drosophila protein Toll. The TLR family now consists of 13 members: although only TLR1 to TLR10 have been identified in humans. TLR11 does not seem to be present in humans22, and TLR10 is not present in mice. TLRs are members of the Toll/IL-1R (TIR)domain-containing superfamily, and there is considerable homology between the cytoplasmic domains of all family members. In particular, a high degree of similarity exists within the 200 amino-acid region that comprises the TIR domain; within this domain are up to three conserved regions that are crucial for the assembly of downstream signalling complexes containing one or more members of the MyD88adaptor family. Other proteins that transduce signals or otherwise regulate receptor function probably also form part of these complexes. Despite the similarity of TIR-domain-containing superfamily members, there are substantial differences in the extracellular regions of these proteins. The extracellular regions of the IL-1Rs consist of immunoglobulin-like domains. By contrast, the TLRs have two highly conserved
VOLUME 4 | JULY 2004 | 5 1 3
REVIEWS
structural features, the most important of which is the leucine-rich repeat (LRR)-containing ectodomain. This domain is thought to be involved in ligand recognition and usually comprises 2429 amino-acid repeats that contain the canonical repeat XLXXLXLXX, where X denotes any amino acid. Both the LRRs and the TIR domains could be exploited as drug targets. In addition to the conserved structural features among the TLRs, there seems to be another level of organization. The TLRs that are involved in the recognition of microbial products TLR1, TLR2, TLR4, TLR5, TLR6 and TLR11 are displayed on the cell surface. By contrast, TLR7, TLR8 and TLR9 are localized intracellularly, and their natural ligands might only be found within acidic compartments, such as phagolysosomes. The natural ligand for TLR9 is bacterial DNA, and recent studies using TLR7-deficient mice, as well as investigations of human TLR8, have shown that these TLRs recognize guanosine and uridine (G+U)-rich single-stranded RNA. Whether TLR3 functions at the cell surface or intracellularly is still unresolved, although some reports indicate that signalling from TLR3 does require endosomal maturation23. These differences between TLRs indicate that the design of synthetic agonists will need to take cellular localization into consideration. For example, if high-throughput binding assays are to be used to identify novel TLR ligands, then the pH of the binding solutions will need to reflect the potential differences between the plasma membrane and intracellular compartments. The precise mechanism by which synthetic ligands enter cells and engage
Box 1 | MyD88-adaptor family The figure shows the domain structure of the five known MyD88 DD TIR members of the MyD88 (myeloid differentiation TIRAP TIR primary-response protein 88) -adaptor family: MyD88, TRIF TIR TIRAP (Toll/interleukin-1 receptor (TIR)-domainTRAM TIR containing adaptor protein; also known as MyD88SARM SAM SAM TIR adaptor-like protein, MAL), TRIF (TIR-domaincontaining adaptor protein inducing interferon-), TRAM (TRIF-related adaptor molecule) and SARM (sterile - and armadillo-motif-containing protein)15. Evidence from genetic and biochemical studies has linked signalling from Toll-like receptor 7 (TLR7), TLR8 and TLR9 to a pathway that requires only MyD88. By contrast, signalling from TLR2, after the formation of a heterodimer with either TLR1 or TLR6, is linked to a pathway that requires a complex of MyD88 and a second adaptor, TIRAP. The activation of interferon-regulatory factor 3 (IRF3) and the subsequent induction of interferon-, which are elicited by signalling from TLR3 or TLR4, are MyD88-independent and involve a third adaptor, TRIF. A fourth adaptor, TRAM, is also involved in MyD88-independent signalling through TLR4 (REF. 7477), and a fifth adaptor, SARM, has been described, but its role in TLR signalling is unclear at present. At the most proximal step after ligand binding, these proteins of the MyD88-adaptor family provide specificity for the outcome of TLR signalling that is initiated following the engagement of different types of ligands by different TLRs.
DD, death domain; SAM, sterile -motif.
their cognate intracellular TLRs also remains to be determined. This is of particular interest for the synthetic TLR9 ligand, which is based on non-methylated CpG sequences that are found in bacterial DNA. The report by Medzhitov, Janeway and PrestonHurlburt98 that described a human homologue of Drosophila Toll showed that the expression of human Toll (TLR) in human cell lines leads to the activation of NF-B and the expression of cytokines. This data led to the first model for TLR-dependent cell activation, in which the expression and multimerization of TLRs is sufficient to support cell activation. Subsequently, when the cognate ligands for the TLRs were defined, it became clear that TLRs associate with other proteins to form a heteromeric receptor complex. These additional proteins might be TLRs that form either homo- or heterodimeric complexes or proteins unrelated to the TLR family. There are several examples of each type of receptor complex, including the CD14MD2TLR4 complex24,25 and the TLR2TLR1 (REFS 2629), TLR2TLR6 (REFS 2629) and TLR4TLR5 (REF. 30) complexes. Although heteromeric complexes have only been identified for TLRs that are present at the plasma membrane, it seems probable that further components of these complexes will be identified for both the cell-surface expressed and intracellular TLRs. There is little doubt that the heteromeric nature of the TLR complexes reflects the downstream physiological pathways. So, can this knowledge be exploited to identify new drug targets that do not require the development of novel agonists or antagonists that directly target the TLR-ligand-binding sites? To achieve this, agents that disrupt proteinprotein interactions will be required.
NOD-protein family
Recently, a family of mammalian proteins that contain a nucleotide-binding oligomerization domain (NOD) and LRRs has been shown to have an important role in innate immunity, and substantial progress has been made in identifying the ligands for the two family members that have known innate immune functions: NOD1 and NOD2 (REFS 1012) (FIG. 3). Members of this family also have structural features that are found in the large family of R proteins (also known as plant disease-resistance proteins)3133. There are more than 20 members of the NOD-protein family, including proteins of known and unknown functions. One common feature of these proteins is the presence of specific structural domains at the amino (N)-terminus, including the CARD and pyrin sequences. The latter allow for unique proteinprotein interactions and most probably, the propagation of some downstream signals. The tissue-distribution patterns of NOD proteins range from ubiquitous to highly restricted. Mutations in several proteins including NOD2, MHC class II transactivator (CIITA), neuronal apoptosis inhibitory protein 5 (NAIP5) and cryopyrin are associated with increased disease susceptibilities, chronic inflammatory disease and hypersensitivity to some bacterial infections. The NOD-protein family initiates cellular activation through a multi-protein
514

complex known as an inflammasome. The composition and properties of these multi-protein complexes are discussed in detail by Martinon and Tschopp34 and are not reviewed here. NOD-protein-family members seem to be involved in a number of autoinflammatory human diseases, such as Blau syndrome, Muckle-Wells syndrome and familial cold urticaria, thereby indicating their importance in inflammation. This probably also depends, in part, on members of the inflammatorycaspase family. The blockade of caspases by selective inhibitors is a well-developed research area and is not discussed further here.
Block TIR domain TIR domain interactions CD14 MD2
TLR4 agonists: MPL, RC-529 Block receptor ligand or protein protein interactions TLR4
Two of the NOD-protein-family members, NOD1 and NOD2, are intracellular sensors for bacterial products that include subunit structures found in bacterial peptidoglycans1012. Similar to the TLRs, the LRR domains of NOD1 and NOD2 are thought to contain the ligand-binding sites. The N-terminal domains might also provide additional sites for drug interactions. A recent study by Nunez and colleagues35 highlights the complexity of the domains of NOD1 and NOD2 that are required for ligand binding. The fact that NOD1 and NOD2 recognize ligands derived from bacterial peptidoglycan, are linked to NF-B and contain CARDs indicates that these proteins are involved in pathways that regulate inflammation and cell death. The cell-death pathways are almost certainly mediated by members of the caspase family, including those known as inflammatory caspases34.
TLR and NOD-protein ligands
Endosome
TIRAP
TRAM
Agonists: loxoribine (TLR7), resiquimod (TLR7), imiquimod (TLR7, TLR8) CpG ODNs (TLR9)
MyD88
IRAK1 RIP1 TRAF6 TRAF6
IKK- IKK- IKK-
IB p50 Nuclear membrane p65 NF-B
NF-Bbinding motif
Figure 2 | TLRs as targets for therapy. Toll-like receptor (TLR) signalling is initiated by plasma membrane and intracellular (endosomal) TLRs following ligand recognition. TLR4 is shown as a representative membrane receptor and the endosomal TLRs (TLR7, TLR8 and TLR9) are also shown. Stimulation of TLRs triggers interaction with the adaptor molecule MyD88 (myeloid differentiation primary-response protein 88). Other adaptor molecules, TIRAP (Toll/interleukin-1 receptor (TIR)-domain-containing adaptor protein), TRIF (TIR-domain-containing adaptor molecule inducing interferon-) and TRAM (TRIF-related adaptor molecule) are also involved in TLR4 signalling. The MyD88-dependent signalling pathway leads to the activation of nuclear factor-B (NF-B), which regulates the expression of target genes that encode pro-inflammatory mediators. Several laboratories have developed agonists of TLR4 and the endosomal TLRs that can stimulate TLR signalling and should be useful therapeutics for treating various human diseases. The key steps that could be targeted in the development of molecules to inhibit TLR signalling are also indicated. IB, inhibitor of NF-B; IKK, IB kinase; IRAK, interleukin-1receptor-associated kinase; MPL, monophosphoryl lipid A; ODN, oligodeoxynucleotide; RIP, receptor-interacting protein; TRAF, tumour-necrosis-factor-receptor-associated factor 6.
MyD88
TRIF
TLR7/8/9
There are at least four broad types of ligand for the TLRs, and for NOD1 and NOD2: naturally occurring molecules that are constituents of microorganisms, synthetic analogues of naturally occurring substances, small-molecule fully synthetic compounds, and endogenous components that are released from host cells during processes such as necrosis, apoptosis and inflammation. Ligands are generally thought of as agonists, but several classes of antagonist have been developed using the structure of agonists as a template for synthetic analogues. A survey of TLR ligands is shown in TABLE 1. Most of the ligands are derived from microbial sources, and they include polymeric molecules such as peptidoglycans (which bind TLR2), bacterial LPS (TLR4), bacterial and viral DNA (TLR9)36,37 and double-stranded RNA (TLR3). Nonetheless, there are many naturally occurring ligands, these are small molecules and include lipoteichoic acid (TLR2), lipoarabinomannan (TLR2) and taxol (TLR4). Large polymeric molecules are not ideal drug candidates because of many factors, including their pharmacodynamic and pharmacokinetic properties. By contrast, because small-molecule ligands have been identified for many of the TLRs, the design of new small molecules with drug-like properties seems to be within reach. There are already good examples of synthetic analogues of naturally occurring ligands, such as molecules based on the structure of lipid A, CpG oligodeoxynucleotides (ODNs) modelled on bacterial DNA sequences, peptidoglycan subunits containing muramic acid, and unique peptide sequences38. Compounds of the imidazole quinoline series, which function as immunostimulants, provide proof-of-concept that fully synthetic small molecules are recognized by TLRs. The ligands for NOD1 and NOD2 are distinct and the minimal structural requirements have been defined. Both are derived from bacterial peptidoglycan: NOD2 recognizes ligands with muramyl-dipeptidic structures, whereas NOD1 binds to tripeptide structures in which the terminal amino acid is diaminopimelic acid. Interestingly, the binding of tripeptides to NOD1
VOLUME 4 | JULY 2004 | 5 1 5
REVIEWS
alter the prevailing T helper 2 (TH2)-cell-type response in sensitized individuals. Therefore, longer-lasting treatments afforded by manipulating the innate immune system would provide a great advance. Immunotherapy can result in decreased symptoms, reduced medication use and enhanced quality of life. It can also prevent the transition from allergic disease to asthma. So, immunotherapy might have therapeutic as well as preventive effects. However, the present design of treatments is limited by our lack of understanding of the essential mechanisms that are involved in triggering chronic inflammatory disease. For example, in asthma, we do not have a mechanistic explanation for the selective activation of TH2 cells3941. Nonetheless, it is clear that one aim of treating asthma with immunomodulators would be to try to change the TH-cell balance in asthmatic individuals: that is, from the dominant TH2-type inflammation to a TH1-type response. Central to this approach is the use of synthetic TLR ligands. In particular, there has been a sustained effort to use oligonucleotides that contain non-methylated CpG motifs to shift the balance to a TH1-type response40. This strategy stems from the hygiene hypothesis, which states that the marked reduction in infection in children in modern society produces decreased TH1-type responses and concomitant TH2-type inflammation. Approaches to treat patients using TH1-type cytokines have failed and, in some cases, have caused serious side effects. There are numerous preclinical studies showing that administration of the TLR9 ligand CpG ODN induces the production of TH1-type cytokines, blocks the production of TH2 cytokines and prevents the symptoms of asthma. Treatment with CpG ODNs seems to reduce disease, without the serious side effects produced following the administration of TH1-type cytokines. However, at present, there is too little attention paid to the potential side effects of chronic administration of immunostimulatory molecules; potential drawbacks of this type of therapy are discussed later. Adjuvants/immunostimulation. Another area that could benefit from targeting the innate immune response is the development of new adjuvants for vaccines. Vaccines are crucial for the management and prevention of infectious disease. Although our endogenous defences mount effective responses to microorganisms, it seems that enhanced innate immunity could be beneficial in specific settings. Previously, such enhancement of immune responses has been achieved using complex mixtures of microbial products, which most probably function through the innate immune system to enhance adaptive immune responses42,43. In addition, we now know that several types of TLR ligand are efficacious as vaccine adjuvants44,45. This has been demonstrated for synthetic TLR7 agonists (imiquimod and resiquimod), TLR9 agonists (CpG ODNs) and TLR4 agonists (lipid A analogues), thereby indicating the merits of identifying synthetic TLR agonists. Here, I discuss a few important examples that involve TLR4, TLR7 and TLR9 (TABLE 2). There are now considerable data indicating that the use of synthetic TLR4 agonists can be beneficial. Some of
Agonists/ antagonists Ligandrecognition domain NOD protein NOD Potential inhibitory site Inhibit CARDCARD interactions
CARD
RICK Kinase domain NF-B activation and caspase activation
Secretion of pro-inflammatory cytokines (TNF and IL-1)
Immune responses
Figure 3 | NOD1 and NOD2 as targets for therapy. Signalling pathways mediated by the nucleotide-binding oligomerization domain (NOD)-protein-family members NOD1 and NOD2 are dependent on the receptor-interacting serine/threonine kinase (RICK) to initiate downstream signalling. RICK interacts with NOD1 and NOD2 through homophilic caspase-recruitment domain (CARD)CARD interactions. Key steps that could be targeted for the development of therapeutics are shown. IL, interleukin; NF-B, nuclear factor-B; TNF, tumour-necrosis factor.
does not require the presence of a muramic acid backbone. Neither NOD1 nor NOD2 seem to recognize polymeric peptidoglycan structures, regardless of the bacterial source.
Targeting innate immunity in human disease
Targeting the innate immune system could provide therapeutic benefit in both acute and chronic human diseases. Efforts are also being directed towards using TLR ligands in vaccines. Examples of each of these applications are discussed here. Asthma. Considerable effort is going into the design of new approaches to treat asthma. The prevalence of allergic diseases, including asthma, has increased markedly in the past three decades. Unfortunately, there are few new drugs that offer cures for these problems, or even provide control or treatment. We still rely on glucocorticoids and bronchodilators, used in combination and in various forms of administration, as the standard practice. Although they are effective, these drugs need to be administered at least twice daily and are not without long-term effects, especially in the increasingly young patient population. Furthermore, corticosteroids that block inflammatory cells do not
516

the most compelling data have been obtained using monophosphoryl lipid A (MPL)4648. This compound is a chemically modified derivative of the lipid A moiety of LPS. MPL is considerably less toxic than LPS but has a similar immunostimulatory activity. MPL has been used extensively in clinical trials, in formulations of both prophylactic and therapeutic vaccines that have targeted various diseases, including cancer, infectious disease and allergies. It seems to have a strong safety profile and good efficacy. A set of related compounds that are also lipid A mimetics, aminoalkylglucosaminide 4-phosphates (AGPs), have been found to be TLR4 ligands and some molecules from this new group of immunostimulants will undoubtedly move into clinical trials. In particular, one derivative, RC-529, has been used in a hepatitis B vaccine trial, replacing the standard alum-based adjuvant. In nearly all of those vaccinated with an MPL adjuvant, protective antibody levels were reached after only two doses of the vaccine, in contrast to the current protocol in which three doses are required to obtain similar levels of protection. Recently, single-stranded RNA has been shown to be a ligand for mouse TLR7 and human TLR8 (REFS 49,50). However, it is well known that several small-molecule immunostimulants are also ligands for TLR7 and TLR8. The prototypical compound, resiquimod, induces IFN- and other cytokines. Resiquimod is one of a class of imidazole quinolines and has been the subject of several clinical investigations. It has been shown to be an immune-response modifier that enhances cutaneous immune responses: it is effective against external genital warts and other diseases
Table 1 | Toll-like receptor ligands
Receptor TLR1 TLR2 Naturally occurring N.D. Lipoproteins/lipopeptides Peptidoglycan Lipoteichoic acid Lipoarabinomannan Atypical lipopolysaccharide Double-stranded RNA Lipopolysaccharide Taxol* HSP60 (Chlamydia pneumoniae) Bacterial flagellin N.D. (G+U)-rich single-stranded RNA* (G+U)-rich single-stranded RNA Bacterial DNA Viral DNA Other DNA with low content of non-methylated CpG sequences HSP70 HSP60 Oligosaccharides of hyaluronic acid Synthetic analogues Triacyl lipopeptides Di- and triacyl lipopeptides Fully synthetic small molecules Exogenous ligands
caused by human papillomavirus51,53. The imidazole quinolines are the first class of TLR agonists to be used for several clinical applications in humans. Another compound, loxoribine (7-allyl-7,8-dihydro8-oxo-guanosine), has antitumour activity and functions through a TLR7 MyD88-dependent pathway, which was demonstrated using TLR7- and MyD88deficient mice as a source of cells5456. The activity of loxoribine and resiquimod depend on endosomal maturation and presumably acidification. In this regard, TLR7 and TLR8 are similar to TLR9 and seem to form a subgroup within the TLR family, which functions in endosomal/lysosomal compartments. TLR7 also recognizes bropirimine, an immunostimulant known to have antitumour effects in superficial transitional-cell carcinoma of the bladder. Bropirimine might have direct effects on tumour cells rather than functioning through inducible mediators; however, more study is required to understand its mechanism of action. Infection. Last, one might consider targeting innate immune pathways in the acute setting, where there is local or systemic infection. Here, blunting the proinflammatory cascade might be achieved using receptor antagonists based on inhibitory antibodies or antagonistic ligand mimetics. Another approach could be to induce protection by activating TLRs and/or NOD1/2 to elicit nonspecific resistance effects and enhance host resistance to bacterial or viral infection. Workers at Corixa Corporation have extended the previous studies of AGPs (lipid A mimetics) and showed
TLR3 TLR4
Poly I:C LPS/lipid A mimetics, such as MPL
Synthetic lipid A, E5564
TLR5 TLR6 TLR7 TLR8 TLR9
Discontinuous 13-amino-acid peptide Diacyl lipopeptides Oligonucleotides CpG oligodeoxynucleotides
Imidazole quinolines (imiquimod, resiquimod) Guanosine nucleotides (loxoribine) Imidazole quinolines (imiquimod)
Endogenous ligands TLR2 TLR4
*Ligand for mouse TLR only. Ligand for human TLR only. HSP, heat-shock protein; LPS, lipopolysaccharide; MPL, monophosphoryl lipid A; N.D., not determined; Poly I:C, polyinosinicpolycytidylic acid; TLR, Toll-like receptor.
VOLUME 4 | JULY 2004 | 5 1 7
REVIEWS
non-selectively block IKK activity using non-steroidal anti-inflammatory drugs. At present, the most progress towards developing selective inhibitors is in the development of selective small-molecule inhibitors of IKK-. These might have applications in controlling events downstream of TLR activation in chronic disease. In particular, it is possible to envisage the use of such inhibitors to treat chronic inflammatory diseases, in which sterile inflammation that results from TLR activation is important. The use of other selective kinase inhibitors could also potentially control TLRdriven inflammation. The kinase targets might include IRAK4, RICK and the mitogen-activated protein kinases; this topic is not considered further here. The current progress in defining the specific roles of members of the MyD88-adaptor family in TLR signalling offers new possibilities for the selective blockade of pathways downstream of the TLRs. We now require the development of new therapeutics to target specific proteinprotein interactions. This has proven to be a formidable barrier, with only limited progress being made. Screening of conventional chemical libraries comprising small-molecule drug-like structures has not provided compounds able to block proteinprotein interactions. In part, this results from the typical flatness of the interface that characterizes proteinprotein interactions. Large surface areas are involved, usually 7501500 2, and there are no deep cavities present that could function as small-moleculebinding sites. Many researchers, including Wells and his co-workers70 have defined hot spots that represent the most important regions for high-affinity binding. As we accumulate knowledge about each type of molecule, the TLRs, NOD-protein-family members and adaptor proteins should become more tractable targets. There are several strong examples that support the contention that progress is being made. One is in the unrelated area of p53 interactions but the other provides data about the targeting of TIR domains. In the former, Vassilev et al.71 have discovered a group of synthetic compounds, known as nutlins, that specifically block p53MDM2 interactions. This finding followed a series of elegant studies by others, which were based on designing peptidic inhibitors and then identifying the specific binding site of p53 that mediates MDM2 binding. This work could well provide a model for future efforts in the design of drugs to target innate immune pathways. In fact, Rebek and colleagues72 have provided the much needed proof-of-concept that TIR domains could be targets for selective inhibitors. Using the TIR-domain interactions between MyD88 and the type 1 IL-1R (IL-1R1) as a model, they synthesized and characterized a low-molecular-weight MyD88 mimic, hydrocinnamoyl-l-valyl-pyrrolidine. This compound was developed by modelling a tripeptide sequence of the BB-loop of the TIR domain (F/Y)-(V/L/I)-(P/G). The compound seems to be selective in both cell-based and in vivo systems. Development of this low-molecular-weight inhibitor of the TIR-domain interaction between MyD88 and
Table 2 | Toll-like receptors as drug targets in clinical trials

Receptor TLR4 Ligand MPL adjuvant MPL adjuvant Ribi 529 E5564 TLR7 Loxoribine Resiquimod Imiquimod TLR9 CpG ODNs CpG ODNs CpG ODNs Application Adjuvant allergy treatment Vaccines Vaccines Endotoxaemia, liver disease Cancer Increased cutaneous immune responses Basal-cell carcinoma Vaccine adjuvant Melanoma Non-Hodgkin lymphoma 8689 9092 9396 97 97 References 78,79 46,80,81 81 8284 85
MPL, monophosphoryl lipid A; ODN, oligodeoxynucleotide; TLR, Toll-like receptor.
that they provide remarkable protection in Listeria or influenza-virus challenge models, in which there is no administration of the microorganism itself or additional antigen57. Considerable data indicate a role for the innate immune system in antiviral responses. The work of Finberg and colleagues58 supports the contention that viral proteins might directly activate TLRs, such as TLR4, thereby resulting in an antiviral response. Subsequently, it was shown that stimulation of TLR3 and TLR4 results in the induction of a subset of genes, including those that encode type I interferons, which inhibit viral replication59. Signalling from TLR3 and TLR4 has been shown to involve both the NF-B and IRF3 pathways. More recently, a detailed understanding of these pathways has emerged from the efforts of Beutler and others6065. In particular, it has been shown that pathways linked to TLR3 and TLR9 contribute to innate responses to systemic viral infection. Although neither pathway alone has a dominant role, mutations affecting both result in markedly decreased antiviral responses. Therefore, targeting the TLR3, TLR4 and TLR9 pathways allows for an enhanced response to certain viral infections. There are numerous small-molecule agonists that have the potential to enhance resistance to viral infection. A new strategy involving covalent linkage of antigen to CpG ODNs promises to provide an effective way of improving the immunomodulatory activities of CpG, by enhancing antigen uptake and stimulation of antigen-presenting cells66,67. This concept has been extended to clinical trials with the use of allergenCpG ODN conjugates68.
Targeting intracellular pathways
The important role of NF-B in TLR signalling is now well established. The NF-B-signalling pathways are one of the main targets for new drug discovery and development69. Numerous studies have shown that involvement of the IKK complex (inhibitor of NF-B (IB)-kinase complex), a downstream element in TLR-signalling pathways, is essential. The two kinases present in the IKK complex, IKK- and IKK-, are potential targets, and recent studies have been able to
518

IL-1R1 indicates that the design of further selective inhibitors of TIR-domain-containing proteins is within reach. This promises to yield an entirely new generation of potential anti-inflammatory compounds to modulate the innate immune system.
Complications with therapeutics
functions and the potentially harmful effects that result from TLR activation remains a serious challenge for those designing new therapeutics.
Conclusions
As therapeutics, CpG ODNs might require administration over long time periods. The effect of chronic administration has not yet been addressed in detail. However, a recent publication reported the consequences of chronic CpG ODN administration73: mice given daily injections of CpG ODNs developed considerable pathology in their lymphoid organs, with changes in both structure and function. Specifically, the report showed destruction of lymphoid-follicle architecture and suppression of follicular dendritic cells and germinal-centre B lymphocytes. Furthermore, after three weeks, multi-focal liver necrosis and haemorrhagic ascites developed. These effects were dependent on both CpG ODNs and TLR9. Additional studies will be required to determine whether the administration other TLR ligands induces similar effects and how the CpG-ODN-induced injury correlates with dosage and its more-beneficial effects. The use of TLR antagonists might have applications in acute settings, in which dampening the proinflammatory response of the innate immune system is beneficial. In particular, TLR antagonists could be used to treat septic shock. However, the important role of the innate immune system in host defence against microorganisms indicates that only limited uses of antagonists will be found. Certainly, their use in any chronic settings will require a high degree of selectivity among the various TLRs. Determining how to maintain the balance between host-defence
We now have detailed knowledge of the ligands, receptors and intracellular-signalling pathways that control innate immune responses during infection and inflammation, and this provides new approaches for highly selective therapeutics. New information about the role of NOD2 in human disease exemplifies how genetic approaches will increase our understanding of disease pathogenesis and how best to target individual proteins or pathways. The potential to modulate the innate immune system is supported by the first generation small-molecule ligands of TLR7 and TLR8, the imidazole quinolines. This class of drug was developed for topical use, targeting papillomavirus-induced warts, without understanding its exact mechanism of action. Furthermore, the ongoing trials based on targeting TLR9 with CpG ODNs will enhance our knowledge of the effects of blocking or enhancing selective pathways of innate immunity. It should now be possible to use our combined knowledge of the TLR- and NOD-protein-family members, together with the recently developed tools for drug discovery, to identify new therapeutic molecules with selective actions. Our progress will only be limited by our ability to express the key proteins of the innate immune system for use in high-throughput screening assays, and our ability to develop small-molecule libraries containing chemical scaffolds that are directed towards the hot spots of proteinprotein interactions. Hopefully, the fruits of ongoing basic research are within reach of the drugdevelopment community, and we will soon see how our newly found knowledge can be used to better treat human disease.
Yamamoto, M., Takeda, K. & Akira, S. TIR domaincontaining adaptors define the specificity of TLR signaling. Mol. Immunol. 40, 861868 (2004). 2. Beutler, B., Hoebe, K., Du, X. & Ulevitch, R. J. How we detect microbes and respond to them: the Toll-like receptors and their transducers. J. Leukoc. Biol. 74, 479485 (2003). 3. Underhill, D. M. Toll-like receptors: networking for success. Eur. J. Immunol. 33, 17671775 (2003). 4. ONeill, L. A. The role of MyD88-like adapters in Toll-like receptor signal transduction. Biochem. Soc. Trans. 31, 643647 (2003). 5. Dunne, A. & ONeill, L. A. The interleukin-1 receptor/Tolllike receptor superfamily: signal transduction during inflammation and host defense. Sci. STKE [online] 2003, re3 (2003). 6. Takeda, K., Kaisho, T. & Akira, S. Toll-like receptors. Annu. Rev. Immunol. 21, 335376 (2003). 7. Chamaillard, M.,Girardin, S. E., Viala, J. & Philpott, D. J. Nods, Nalps and Naip: intracellular regulators of bacterialinduced inflammation. Cell. Microbiol. 5, 581592 (2003). 8. Inohara, N. & Nunez, G. NODs: intracellular proteins involved in inflammation and apoptosis. Nature Rev. Immunol. 5, 371382 (2003). 9. Akira, S. & Takeda, K. Toll-like receptor signalling. Nature Rev. Immunol. 4, 499511 (2004). 10. Girardin, S. E. et al. Peptidoglycan molecular requirements allowing detection by Nod1 and Nod2. J. Biol. Chem. 278, 4170241708 (2003). 11. Chamaillard, M. et al. An essential role for NOD1 in host recognition of bacterial peptidoglycan containing diaminopimelic acid. Nature Immunol. 4, 702707 (2003). 12. Girardin, S. E. et al. Nod1 detects a unique muropeptide from Gram-negative bacterial peptidoglycan. Science 300, 15841587 (2003).
1.
13.
14.
15.
16.
17.
18. 19.
20. 21.
22.
Provides the first definition of the structural requirements of the NOD1 ligand. ONeill, L. A. et al. Mal and MyD88: adapter proteins involved in signal transduction by Toll-like receptors. J. Endotoxin Res. 9, 5559 (2003). Martin, M. U. & Wesche, H. Summary and comparison of the signaling mechanisms of the Toll/interleukin-1 receptor family. Biochim. Biophys. Acta 1592, 265268 (2002). ONeill, L. A., Fitzgerald, K. A. & Bowie, A. G. The TollIL-1 receptor adaptor family grows to five members. Trends Immunol. 24, 286290 (2003). Ogura, Y. et al. Nod2, a Nod1/Apaf-1 family member that is restricted to monocytes and activates NF-B. J. Biol. Chem. 276, 48124818 (2001). Kobayashi, K. et al. RICK/Rip2/CARDIAK mediates signalling for receptors of the innate and adaptive immune systems. Nature 416, 194199 (2002). Provides important evidence implicating RICK in both TLR- and NOD-protein signalling pathways. Hoffmann, J. A. The immune response of Drosophila. Nature 426, 3338 (2003). Gobert, V. et al. Dual activation of the Drosophila Toll pathway by two pattern recognition receptors. Science 302, 21262130 (2003). Imler, J. L. & Hoffmann, J. A. Toll signaling: the TIReless quest for specificity. Nature Immunol. 4, 105106 (2003). Gottar, M. et al. The Drosophila immune response against Gram-negative bacteria is mediated by a peptidoglycan recognition protein. Nature 416, 640644 (2002). Provides a new paradigm that describes different arms of the innate immune response in insects. Zhang, D. et al. A Toll-like receptor that prevents infection by uropathogenic bacteria. Science 303, 15221526 (2004).
23. Nishiya, T. & DeFranco, A. L. Ligand-regulated chimeric receptor approach reveals distinctive subcellular localization and signaling properties of the Toll-like receptors. J. Biol. Chem. 279, 1900819017 (2004). 24. Miyake, K. Endotoxin recognition molecules, Toll-like receptor 4MD-2. Semin. Immunol. 16, 1116 (2004). 25. Akashi, S. et al. Lipopolysaccharide interaction with cell surface Toll-like receptor 4MD-2: higher affinity than that with MD-2 or CD14. J. Exp. Med. 198, 10351042 (2003). 26. Wetzler, L. M. The role of Toll-like receptor 2 in microbial disease and immunity. Vaccine 21, S55S60 (2003). 27. Kirschning, C. J. & Schumann, R. R. TLR2: cellular sensor for microbial and endogenous molecular patterns. Curr. Top. Microbiol. Immunol. 270, 121144 (2002). 28. Hajjar, A. M. et al. Cutting edge: Functional interactions between Toll-like receptor (TLR) 2 and TLR1 or TLR6 in response to phenol-soluble modulin. J. Immunol. 166, 1519 (2001). 29. Ozinsky, A. et al. The repertoire for pattern recognition of pathogens by the innate immune system is defined by cooperation between Toll-like receptors. Proc. Natl Acad. Sci. USA 97, 1376613771 (2000). Provides key data showing how TLRs comprise heteromeric subunits that determine ligand specificity. 30. Mizel, S. B., Honko, A. N., Moors, M. A., Smith, P. S. & West, A. P. Induction of macrophage nitric oxide production by Gram-negative flagellin involves signaling via heteromeric Toll-like receptor 5/Toll-like receptor 4 complexes. J. Immunol. 170, 62176223 (2003). 31. Schulze-Lefert, P. Plant immunity: the origami of receptor activation. Curr. Biol. 14, R22R24 (2004).
VOLUME 4 | JULY 2004 | 5 1 9
REVIEWS
32. Nimchuk, Z., Eulgem, T., Holt, B. F. & Dangl, J. L. Recognition and response in the plant immune system. Annu. Rev. Genet. 37, 579609 (2003). 33. Martin, G. B., Bogdanove, A. J. & Sessa, G. Understanding the functions of plant disease resistance proteins. Annu. Rev. Plant Biol. 54, 2361 (2003). 34. Martinon, F. & Tschopp, J. Inflammatory caspases; linking an intracellular innate immune system to autoinflammatory diseases. Cell 117, 561574 (2004). 35. Tanabe, T. et al. Regulatory regions and critical residues of NOD2 involved in muramyl dipeptide recognition. EMBO J. 23, 15871597 (2004). 36. Lund, J., Sato, A., Akira, S., Medzhitov, R. & Iwasaki, A. Toll-like receptor 9-mediated recognition of Herpes simplex virus-2 by plasmacytoid dendritic cells. J. Exp. Med. 198, 513520 (2003). 37. Krug, A. et al. Herpes simplex virus type 1 activates murine natural interferon producing cells through Toll-like receptor 9. Blood 103, 14331437 (2004). 38. Kaisho, T. & Akira, S. Regulation of dendritic cell function through Toll-like receptors. Curr. Mol. Med. 3, 373385 (2003). 39. El Biaze, M. et al. T cell activation, from atopy to asthma: more a paradox than a paradigm. Allergy 58, 844853 (2003). 40. Hussain, I. & Kline, J. N. CpG oligodeoxynucleotides: a novel therapeutic approach for atopic disorders. Curr. Drug Targets Inflamm. Allergy 2, 199205 (2003). 41. Kuchroo, V. K., Umetsu, D. T., DeKruyff, R. H. & Freeman, G. J. The TIM gene family: emerging roles in immunity and disease. Nature Rev. Immunol. 3, 454462 (2003). 42. Melief, C. J. Strategies for immunotherapy of cancer. Adv. Immunol. 75, 235282 (2000). 43. Hess, J., Schaible, U., Raupach, B. & Kaufmann, S. H. Exploiting the immune system: toward new vaccines against intracellular bacteria. Adv. Immunol. 75, 188 (2000). 44. Kaisho, T. & Akira, S. Toll-like receptors as adjuvant receptors. Biochim. Biophys. Acta 1589, 113 (2002). 45. Jiang, Z. H. & Koganty, R. R. Synthetic vaccines: the role of adjuvants in immune targeting. Curr. Med. Chem. 10, 14231439 (2003). 46. Evans, J. T. et al. Enhancement of antigen-specific immunity via the TLR4 ligands MPL adjuvant and Ribi. 529. Expert Rev. Vaccines 2, 219229 (2003). 47. Persing, D. H. Taking Toll: lipid A mimetics as adjuvants and immunomodulators. Trends Microbiol. 10, S32S37 (2002). 48. Stover, A. G. et al. Structure-activity relationship of synthetic Toll-like receptor 4 agonists. J. Biol. Chem. 279, 44404449 (2004). 49. Heil, F. et al. Species-specific recognition of single-stranded RNA via Toll-like receptor 7 and 8. Science 303, 15261529 (2004). Identifies a naturally occurring ligand for TLR 7 and TLR 8. 50. Diebold, S. S., Kaisho, T., Hemmi, H., Akira, S. & Reis e Sousa, C. Innate antiviral responses by means of TLR7-mediated recognition of single-stranded RNA. Science 303, 15291531 (2004). Identifies a naturally occurring ligand for TLR7. 51. Skinner R. B. Jr. Imiquimod. Dermatol. Clin. 21, 291300 (2003). 52. Hengge, U. R. & Cusini, M. Topical immunomodulators for the treatment of external genital warts, cutaneous warts and molluscum contagiosum. Br. J. Dermatol. 149 (Suppl. 66), 1519 (2003). 53. Bernard, H. U. Established and potential strategies against papillomavirus infections. J. Antimicrob. Chemother. 53, 137139 (2004). 54. Akira, S. & Hemmi, H. Recognition of pathogen-associated molecular patterns by TLR family. Immunol. Lett. 85, 8595 (2003). 55. Heil, F. et al. The Toll-like receptor 7 (TLR7)-specific stimulus loxoribine uncovers a strong relationship within the TLR7, 8 and 9 subfamily. Eur. J. Immunol. 33, 29872997 (2003). 56. Hemmi, H. et al. Small anti-viral compounds activate immune cells via the TLR7 MyD88-dependent signaling pathway. Nature Immunol. 3, 196200 (2002). 57. Baldridge, J. R. et al. Immunostimulatory activity of aminoalkyl glucosaminide 4-phosphates (AGPs): induction of protective innate immune responses by RC-524 and RC-529. J. Endotoxin Res. 8, 453458 (2002). 58. Kurt-Jones, E. A. et al. Pattern recognition receptors TLR4 and CD14 mediate response to respiratory syncytial virus. Nature Immunol. 1, 398401 (2000). Among the first reports to indicate a role for the innate immune system in antiviral responses. Doyle, S. et al. IRF3 mediates a TLR3/TLR4-specific antiviral gene program. Immunity 17, 251263 (2002). Hoebe, K. & Beutler, B. LPS, dsRNA and the interferon bridge to adaptive immune responses: Trif, Tram, and other TIR adaptor proteins. J. Endotoxin Res. 10, 130136 (2004). Tabeta, K. et al. Toll-like receptors 9 and 3 as essential components of innate immune defense against mouse cytomegalovirus infection. Proc. Natl Acad. Sci. USA 101, 35163521 (2004). Hoebe, K. et al. Upregulation of costimulatory molecules induced by lipopolysaccharide and double-stranded RNA occurs by Trif-dependent and Trif-independent pathways. Nature Immunol. 4, 12231229 (2003). Beutler, B. et al. Lps2 and signal transduction in sepsis: at the intersection of host responses to bacteria and viruses. Scand. J. Infect. Dis. 35, 563567 (2003). Honda, K. et al. Selective contribution of IFN-/ signaling to the maturation of dendritic cells induced by doublestranded RNA or viral infection. Proc. Natl Acad. Sci. USA 100, 1087210877 (2003). Matsumoto, M., Funami, K., Oshiumi, H. & Seya, T. Toll-like receptor 3: a link between Toll-like receptor, interferon and viruses. Microbiol. Immunol. 48, 147154 (2004). Heit, A. et al. Cutting edge: Toll-like receptor 9 expression is not required for CpG DNA-aided cross-presentation of DNA-conjugated antigens but essential for cross-priming of CD8 T cells. J. Immunol. 170, 28022805 (2003). Heit, A. et al. CpG-DNA aided cross-priming by crosspresenting B cells. J. Immunol. 172, 15011507 (2004). Speigelberg, H. L., Horner, A. A., Takabayashi, K. & Raz, E. Allergen-immunostimulatory oligodeoxynucleotide conjugate: a novel allergoid for immunotherapy. Curr. Opin. Allergy Clin. Immunol. 2, 547551 (2002). Karin, M., Yamamoto, Y. & Wang, Q. M. The IKK NF-B system: a treasure trove for drug development. Nature Rev. Drug Discov. 3, 1726 (2004). Arkin, M. R. & Wells, J. A. Small-molecule inhibitors of proteinprotein interactions: progressing towards the dream. Nature Rev. Drug Discov. 3, 301317 (2004). Vassilev, L. T. et al. In vivo activation of the p53 pathway by small-molecule antagonists of MDM2. Science 303, 844848 (2004). Bartfai, T. et al. A low molecular weight mimic of the Toll/IL-1 receptor/resistance domain inhibits IL-1 receptor-mediated responses. Proc. Natl Acad. Sci. USA 100, 7971-7976 (2003). Provides the first evidence indicating that it is possible to target TIRTIR domain interactions with a small molecule. Heikenwalder, M. et al. Lymphoid follicle destruction and immunosuppression after repeated CpG oligodeoxynucleotide administration. Nature Med. 10, 187192 (2004). Hoebe, K. et al. Identification of Lps2 as a key transducer of MyD88-independent TIR signalling. Nature 424, 743748 (2003). This seminal paper describes the use of forward genetics to decipher TLR-signalling pathways and identifies Lps2 as a gene encoding TRIF. Kawai, T., Adachi, O., Ogawa, T., Takeda, K. & Akira, S. Unresponsiveness of MyD88-deficient mice to endotoxin. Immunity 11, 115122 (1999). Yamamoto, M. et al. TRAM is specifically involved in the Toll-like receptor 4-mediated MyD88-independent signaling pathway. Nature Immunol. 4, 11441150 (2003). Fitzgerald, K. A. et al. LPSTLR4 signaling to IRF-3/7 and NF-B involves the Toll adapters TRAM and TRIF. J. Exp. Med. 198, 10431055 (2003). Drachenberg, K. J., Heinzkill, M., Urban, E. & Woroniecki, S. R. Efficacy and tolerability of short-term specific immunotherapy with pollen allergoids adjuvanted by monophosphoryl lipid A (MPL) for children and adolescents. Allergol. Immunopathol. (Madr.) 31, 270277 (2003). Mothes, N. et al. Allergen-specific immunotherapy with a monophosphoryl lipid A-adjuvanted vaccine: reduced seasonally boosted immunoglobulin E production and inhibition of basophil histamine release by therapy-induced blocking antibodies. Clin. Exp. Allergy 33, 11981208 (2003). Bienzle, U. et al. Immunization with an adjuvant hepatitis B vaccine after liver transplantation for hepatitis B-related disease. Hepatology 38, 811819 (2003). 81. Vernacchio, L. et al. Effect of monophosphoryl lipid A (MPL) on T-helper cells when administered as an adjuvant with pneumocococcal-CRM197 conjugate vaccine in healthy toddlers. Vaccine 20, 36583667 (2002). 82. Lynn, M. et al. Extended in vivo pharmacodynamic activity of E5564 in normal volunteers with experimental endotoxemia. J. Pharmacol. Exp. Ther. 308, 175181 (2004). 83. Wong, Y. N. et al. Safety, pharmacokinetics, and pharmacodynamics of E5564, a lipid A antagonist, during an ascending single-dose clinical study. J. Clin. Pharmacol. 43, 735742 (2003). 84. Liang, E. et al. Pharmacokinetics of E5564, a lipopolysaccharide antagonist, in patients with impaired hepatic function. J. Clin. Pharmacol. 43, 13611369 (2003). 85. Agarwala, S. S., Kirkwood, J. M. & Bryant, J. Phase 1, randomized, double-blind trial of 7-allyl-8-oxoguanosine (loxoribine) in advanced cancer. Cytokines Cell. Mol. Ther. 6, 171176 (2000). 86. Sauder, D. N., Smith, M. H., Senta-McMillian, T., Soria, I. & Meng T. C. Randomized, single-blind, placebo-controlled study of topical application of the immune response modulator resiquimod in healthy adults. Antimicrob. Agents Chemother. 47, 38463852 (2003). 87. Jones, T. Resiquimod 3M. Curr. Opin. Investig. Drugs 4, 214218 (2003). 88. Imiquimod (Aldara) for actinic keratoses. Med. Lett. Drugs Ther. 46, 4244 (2004). 89. Prinz, B. M. et al. Treatment of Bowens disease with imiquimod 5% cream in transplant recipients. Transplantation 77, 790791 (2004). 90. Kreuter, A. et al. Treatment of anal intraepithelial neoplasia in patients with acquired HIV with imiquimod 5% cream. J. Am. Acad. Dermatol. 50, 980981 (2004). 91. Huber, A. et al. Topical imiquimod treatment for nodular basal cell carcinomas: an open-label series. Dermatol. Surg. 30, 429430 (2004). 92. Herbert, W. C. Imiquimod and the treatment of cutaneous T-cell proliferative diseases: at the threshold. Skinmed. 2, 273274 (2003). 93. Leifer, C. A., Verthelyi, D. & Klinman, D. M. Heterogeneity in the human response to immunostimulatory CpG oligodeoxynucleotides. J. Immunother. 26, 313319 (2003). 94. Tulic, M. K. Amb a 1-immunostimulatory oligodeoxynucleotide conjugate immunotherapy decreases the nasal inflammatory response. J. Allergy Clin. Immunol. 113, 234241 (2004). 95. Bohle, B. AIC. Dynavax. Curr. Opin. Investig. Drugs 4, 603607 (2003). 96. Immunostimulatory DNADynavax. AIC, Amb a 1 immunostimulatory conjugate, HBV-ISS, ISS 1018, ISS DNA, ISS DNAdynavax, ISS1, ISS2. Drugs R D 3, 193196 (2002). 97. Paul, S. Technology evaluation: CpG-7909, Coley. Curr. Opin. Mol. Ther. 5, 553559 (2003). 98. Medzhitov, R., Preston-Hurlburt, P. & Janeway, C. A. Jr. A human homologue of the Drosophila Toll protein signals activation of adaptive immunity. Nature 388, 394397 (1997).
59. 60.
61.
62.
63.
64.
65.
66.
67. 68.
69.
70.
71.
72.
73.
74.
75.
Acknowledgements
I thank J. C. Mathison for comments and editorial support and P. Rutledge for administrative support. This work was supported by the National Institutes of Health, The Charles Dana Foundation and The Novartis Foundation.
76.
77.

The author declares competing financial interests: See Web version for details.
78.
Online links
DATABASES The following terms in this article are linked online to: Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene CD14 | IKK- | IKK- | IL-1R1 | IRAK1 | IRAK4 | IRF3 | MD2 | MyD88 | NOD1 | NOD2 | RICK | SARM | TIRAP | TLR1 | TLR2 | TLR3 | TLR4 | TLR5 | TLR6 | TLR7 | TLR8 | TLR9 | TLR11 | TRAF6 | TRAM | TRIF
79.
80.
520
PERSPECTIVES
adopted to identify the NF-B-like transcription factor that regulates the genes encoding antibacterial peptides. The first strategy was to use biochemical techniques to purify the B-motif binding factor(s) from extracts of Drosophila cell lines. For this approach, the objective was to identify the binding factor and then to work backwards, step by step, to identify the upstream elements of the signalling cascade that activate the factor. This tactic was motivated by the successful characterization of mammalian NF-B-signalling pathways that use similar strategies9. Considerable efforts were made to use biochemical techniques to isolate an NF-B-like molecule that functions in the Drosophila immune response; however, this approach was ultimately unsuccessful. The second strategy for identifying a Drosophila NF-B-like transcription factor that regulates immune responses was a genetic approach, which I undertook with colleagues in the Hoffmann laboratory. The power of using genetic techniques to dissect complex biological processes had previously been illustrated by the mutant screens that Eric Wieschaus and Christiane Nusslein-Volhard carried out (in Heidelberg, Germany) to identify Drosophila genes that regulate early embryogenesis. In the early 1990s, several research groups identified parallels between the establishment of the dorsoventral axis by the Toll pathway in Drosophila embryos (BOX 1) and the cytokineinduced expression of several immune genes by the interleukin-1 receptor (IL-1R)NF-Bsignalling cascade in mammals1012. These groups noted that in both pathways, a Toll/IL-1R (TIR)-domain-containing transmembrane receptor Drosophila Toll or mammalian IL-1R (FIG. 1) activates intracellular signalling, which culminates in the nuclear translocation of an NF-B/NF-Blike transcription factor. In Drosophila, the NF-B-like factor regulated by the Toll pathway during embryonic patterning is known as Dorsal13,14, and Dorsal regulates target genes through B-binding motifs1517.
VOLUME 4 | JULY 2004 | 5 2 1
LANDMARK
The road to Toll

Bruno Lemaitre
A few years ago, it would have been difficult to argue that elucidating the mechanisms of disease resistance in the fruit fly, Drosophila melanogaster, would provide new insights into mammalian immunity. Yet the finding that the Drosophila protein Toll mediates immune responses to fungal infection had a pioneering role in the identification of Toll-like receptors as essential regulators of mammalian host defence, and it fundamentally altered our understanding of innate immunity. In this Landmark article, I describe the thought processes and the experimental steps that defined Toll as a key regulator of Drosophila immune responses.
Given their relatively short lifespans, it is not obvious that insects have, or even require, a powerful immune system for fighting microbial pathogens. Nevertheless, insects are highly resistant to microbial infection. Until recently, however, the mechanisms behind this resistance were poorly understood, because insects do not have an equivalent of the vertebrate adaptive immune system. An important discovery regarding insect immunity was made in 1981, when Hans G. Boman and associates1,2 (in Stockholm, Sweden) characterized the inducible antibacterial peptides Cecropin and Attacin from the moth Hyalophora cecropia. Following a septic injury, these small peptides are produced rapidly in large amounts by the insect fat body (an analogue of the mammalian liver) and then secreted into the haemolymph (insect blood) where they kill invading bacteria. By the early 1990s, several genes encoding antibacterial peptides (such as Diptericin
and the Cecropins) were also identified in Drosophila35 after they were found to be strongly induced at the transcriptional level following the injection of bacteria into the body cavity of the fly. The next challenge in the field was then to determine the molecular mechanisms that regulate these genes in response to microbial infection. Because nothing was known of the steps that lead from the recognition of microorganisms to the expression of genes that encode antibacterial peptides, my colleagues and I (in Jules Hoffmanns laboratory in Strasbourg, France) called this process the black box. In this article, I describe the experiments that initiated the elucidation of the signalling pathways that control the expression of genes encoding antimicrobial peptides in Drosophila (TIMELINE).
Strategies for opening the black box
The first clue to what was inside the black box of antimicrobial-peptide expression was provided by the sequences of genes encoding several antibacterial peptides. Their upstream regulatory regions contain sequence motifs that are similar to the binding sites recognized by the mammalian nuclear factor-B (NF-B)/REL family of transcription factors6. Subsequently, in 1993, the use of fly lines carrying a reporter gene under the control of wild-type or mutated B-binding motifs demonstrated that these binding sites confer the immune inducibility of the Diptericin and Cecropin A1 genes in Drosophila7,8, indicating that Diptericin and Cecropin A1 are regulated by an NF-B-like transcription factor. On the basis of these observations, two distinct strategies were
PERSPECTIVES
Timeline | Discoveries relevant to the function of the Toll and Imd pathways in Drosophila immunity
Link between Toll and IL-1R pathways established Characterization of the Toll mutation and other dorsoventral mutations Cloning of dorsal, homology with the REL oncogene Cloning of Cecropin A and Diptericin in Drosophila Concept of patternrecognition receptors The Toll pathway regulates the antifungal response Identification of the Relish gene B motifs are required for Diptericin and Cecropin induction Dorsal and DIF are linked to the Drosophila immune response Genetic characterization of Dif, Relish and necrotic Drosophila antimicrobial response is adapted to the aggressors Demonstration that PGRPs regulate the Drosophila antimicrobial response Demonstration that GNBP1 regulates Drosophila immune response. Toll and IMD pathways are activated by specific PG recognition (2000 to 2004) Identification of new components of the Toll and IMD pathways by forward and reverse genetic screens and cell-culture studies (IRD5, Kenny, DREDD, TAK1, IMD, Myd88, Fadd)
Identification of the dorsal mutation
1979
1981
1984
1986
1987
1988
1989
1990
1991
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
Sequences of the first inducible antibacterial peptides in insects
NF-B identified in mammals Cloning of Toll
Cloning of NF-Bencoding gene
Identification of Drosomycin Identification of the immune deficiency (imd) mutation
Cloning of insect PGRP A human Toll linked to NF-B TLR4 is a receptor for LPS
Completion of the Drosophila-genome sequence. Molecular characterization of Drosophila PGRP-, GNBP- and Toll-gene families
B motifs are found in the promoters of insect genes that encode antimicrobial peptides
Purification of PGRP and molecular cloning of GNBP from silkworms
(1999 to 2004) Identification of microbial specificities of several TLRs, including TLR2 (PG, lipoprotein), TLR3 (dsRNA), TLR5 (flagellin) and TLR9 (CpG-containing DNA)
The discovery of the role of Toll in the Drosophila immune response (yellow) was influenced mainly by research in three fields: insect immunity (red), vertebrate immunology and signalling (blue) and developmental genetics (green). DIF, Dorsal-related immunity factor; DREDD, Death-related ced-3/Nedd2-like protein; ds, double-stranded; GNBP, Gram-negative-bacteriabinding protein; IB, inhibitor of NF-B; IKK, IB kinase; IL-1R, interleukin-1 receptor; IMD, Immune deficiency; IRD5, Immune-response deficient 5; LPS, lipopolysaccharide; NF-B, nuclear factor-B; PG, peptidoglycan; PGRP, PG-recognition protein; serpin, serine-protease inhibitor; TAK1, Transforming growth factor--activated kinase 1; TLR, Toll-like receptor.
The parallels between the Toll pathway and the IL-1R pathway raised the obvious question of whether the Toll pathway, in addition to its role in dorsoventral polarity, controls the expression of antibacterial peptides in differentiated tissues. Furthermore, although the genes encoding components of the Toll pathway were initially described as maternal-effect genes (which regulate early embryogenesis), it was soon apparent that these genes are also expressed in larvae and adults1820. With Jean Marc Reichhart and other colleagues21 in the Hoffmann laboratory, we determined that the expression of the dorsal gene is upregulated and that in fatbody cells, the Dorsal protein translocates rapidly to the nucleus in response to bacterial infection. These results suggested that Dorsal was the NF-B-like factor that mediated Drosophila immune responses, and they stimulated a wave of enthusiasm for further studies of Dorsal. The next goal was to determine whether Dorsal regulated the expression of the genes encoding antimicrobial peptides and whether this Dorsal activity was linked to the Toll pathway. This project was facilitated by the numerous dorsal and Toll mutants from the WiechausVolhard screens, which were available at the Tbingen stock centre (Germany). The results that I obtained, however, were frustrating: although, in fat-body cells, Dorsal
was activated by the Toll pathway in response to bacterial infection, none of the mutations affecting either Dorsal or the Toll pathway significantly altered the induction of Diptericin expression after infection22. During this period, Tony Ip and colleagues (in Michael Levines laboratory in San Diego, United States) identified a second Drosophila NF-B-like gene, which they called Dorsal-related immunity factor (Dif ). However, they stopped studying Dif when they realized that it was not involved in dorsoventral patterning of the embryo. Further studies of Dif started when Ylva Engstrm (in Stockholm) pointed out a potential link between DIF and the expression of the genes encoding antimicrobial peptides. They showed that DIF is expressed by the fat body, that it can bind to the B-like sequence motifs in the Cecropin A promoter and that its translocation into the nucleus is regulated by Toll23. However, because there were no fly lines carrying a mutation in the Dif gene, they were unable to test for DIF function in immune responses. A clue to how antimicrobial-peptide expression is regulated was provided eventually by Dan Hultmarks group (in Stockholm), when they observed that overexpression of an active form of Toll increased the expression of a Cecropin A transgene in a cell-culture assay24. However, although our studies of
Dorsal, the identification of DIF, and the Toll overexpression studies all indicated a link between the Toll pathway and the expression of antimicrobial peptides, they still did not identify the exact function of Toll in the immune response.
There are two pathways
The initial lack of success using genetic strategies refocused attention on the biochemical approach, and it left me at an impasse in my attempts to use genetics to decipher the signalling pathway that regulates the expression of antimicrobial-peptide genes. The way out of this quandary was revealed by two unexpected discoveries made in the Hoffmann laboratory. The first breakthrough happened in 1994, when Philippe Bulet (a talented biochemist) and his colleagues carried out a differential screen to identify Drosophila peptides that are induced specifically by bacterial infections. In their screen, one of the proteins that was highly induced by bacterial infection was a peptide, which they called DIF-30. DIF-30 did not, however, have any antibacterial activity, and interest in DIF-30 declined until Bulet and colleagues (with the help of the plant pathologist Willem F. Broekaert, from Leuven, Belgium) showed that DIF-30 has strong inhibitory effects against various filamentous fungi. DIF-30 was renamed Drosomycin, and
522

publication of these studies provided the first description of an inducible antifungal peptide identified in insects25. Drosomycin then became a key molecule in my attempts to identify immune-signalling pathways in fruit flies (discussed later). The second breakthrough was my serendipitous discovery of the Drosophila immune deficiency (imd ) mutation. After failing to link Diptericin expression with mutations in the Toll pathway, I decided to stop focusing solely on Toll and broaden my search. Consequently, I began to measure the level of Diptericin expression after infection of any fly line that carried a mutation in a gene that could be linked to immune responses. To my delight, I found that Diptericin expression was significantly reduced in stock 1046 from the Bloomington Stock Center (United States). This fly line originates from an EMS (ethyl methane sulphonate) mutagenesis carried out by Ellsworth Grell in 1969 and has a mutation, Black cells (Bc), that affects Drosophila blood cells known as a crystal cells26. Crystal cells are implicated in the prophenoloxidase cascade, an enzymatic reaction that leads to the deposition of melanin around invading pathogens and has an important role in arthropod immune defence27. Our first analysis indicated that the Bc mutation affected Diptericin expression because a chromosomal deficiency that spanned Bc also reduced Diptericin expression. A link between a melanization cascade and the expression of antibacterial peptides was an attractive concept, and we were preparing a publication on the role of Bc in Diptericin regulation when we realized that we were on the wrong track: first, I noticed that other fly lines carrying mutations that reduced the melanization reaction did not affect Diptericin expression; and second, Michael Levine informed Jules Hoffmann that in their studies, the Bc mutation did not affect Diptericin expression. So, faced with the absence of detectable defects in the anti microbial-peptide expression of Toll-deficient mutants, and after hearing about my results during a visit from Jules Hoffmann, Joe Corbo and Michael Levine had also started to study stock 1046 (the Bc mutant). Using a different set of deficiencies than I had used, they identified a deficiency spanning the mutation that reduced Diptericin expression but not the Bc mutation, demonstrating that these two mutations were at different loci. Using this new information, we then determined that the mutation that blocked Diptericin expression actually mapped to a locus 3.5 centimorgans from Bc, which we
Box 1 | Toll and establishment of polarity in the Drosophila embryo Semisaturating screens carried out by Follicle cell Christiane Nsslein-Volhards group (in Tbingen, Germany), as well as by Trudy ? Schpbach and Eric Wieschaus (in Princeton, SerpinUnited States), identified numerous Gastrulation 27A defective maternal-effect mutations that disrupt embryonic polarity. Some of these mutations affect 12 genes that are involved in Easter Snake establishing dorsoventral polarity67,68. Among these, numerous mutations in the Toll (which means cool in German) locus were analysed in further detail by Kathryn V. Anderson in Sptzle Nsslein-Volhards group69. In 1988 , the Toll Perivitelline Toll space gene was cloned in the Anderson laboratory (in Berkeley, United States) and shown to Cytoplasm encode a transmembrane receptor70. The Myd88 TIR TIR molecular characterization of the other DD dorsoventral patterning genes, carried out by several research groups, has defined the Tube Pelle components of a signalling pathway. Recently, two additional genes, Myd88 and Serpin-27A, which were identified initially for their ? immune phenotype, were shown to function in dorsoventral patterning7174. The Toll Cactus pathway is also used at later developmental Dorsal Dorsal stages, including morphogenetic movement, muscle attachment75 and haemocyte (blood cell) proliferation76. During oogenesis, a molecular cue localized Nucleus on the ventral side of follicle cells initiates a proteolytic cascade in the perivitelline space Twist, Snail outside the fertilized embryo, which is mediated by the proteases Gastrulation B-binding motif defective, Snake and Easter. The activity of Easter is inhibited by the serine-protease inhibitor, Serpin-27A. This cascade results in the ventral processing of Sptzle in a graded manner. The cleaved form of Sptzle then functions as a ligand for Toll. Localized activation of Toll leads to the activation of an intracellular pathway that involves the adaptors Tube and Myd88 and the kinase Pelle. The result of this activation is the phosphorylation and degradation of the IB orthologue Cactus. Cactus interacts with and inhibits the transcription factor Dorsal. Degradation of Cactus allows Dorsal to enter the nucleus, where it regulates the genes that organize dorsoventral patterning, such as twist and snail.
DD, death domain; Serpin, serine-protease inhibitor; TIR domain, Toll/interleukin-1 receptor domain.
called imd, and we used genetic recombination to generate an imd-mutant line that lacked the Bc mutation28. When published in 1995 (REF. 28), imd was the first reported mutation to affect the expression of genes encoding antibacterial peptides, and its identification demonstrated the potential of genetic approaches for analysing immune-signalling pathways. Interestingly, although imd-mutant flies are perfectly viable, they are highly susceptible to Gram-negative bacterial infection, and this phenotype provided the first functional evidence that antimicrobial peptides are important for fighting infections in vivo. I also made the crucial observation that the level of
Drosomycin expression is nearly normal in imd mutants, which indicated that more than one signalling pathway regulates the expression of antimicrobial peptides and that the expression of the genes encoding the antibacterial peptide Diptericin and the antifungal peptide Drosomycin are regulated by separate pathways. Corbo and Levine29 published their own work on the imd mutation one year later, but they did not analyse Drosomycin expression in the imd mutant. The ultimate identification of imd as a second mutation in stock 1046 was, for me, a good lesson in genetics, and it also reminded me that science often progresses when unexpected help and stimulation is provided by potential competitors.
VOLUME 4 | JULY 2004 | 5 2 3
PERSPECTIVES
transgenes encoding DIF or Dorsal into this deletion line, they demonstrated that DIF, but not Dorsal, was the main regulator of the expression of genes encoding antimicrobial peptides in adult fruit flies. Subsequently, the identification of mutations that only affected the Dif locus confirmed this result33. Another important finding that we made after our return to studying Toll, was that in contrast to imd mutants, which die after Gram-negative bacterial infection, fruit flies carrying mutations in the Toll pathway are highly susceptible to fungal infection (FIG. 2). In addition, flies that lack both IMD and Toll fail to express any antimicrobial peptides and are susceptible to both bacterial and fungal infections. The marked and complementary phenotypes of the Toll and imd mutants indicated that Toll and IMD were components of the two main signalling pathways that regulate both the expression of antimicrobial peptides and resistance to bacterial and fungal infection. Our demonstration of Toll function in the antifungal immune response was published in 1996 (REF. 30) and provided the first evidence that Toll has an important role in animal host defence. In this paper, we suggested a basic model in which Toll and IMD control the expression of genes encoding antimicrobial peptides, and we extended the parallels between the cytokine-induced activation of NF-B and the Toll pathway, thereby showing that the regulation of NF-B/NF-Blike molecules is an ancient mechanism for fighting infection.
An adapted innate immune response?
LRR Ig domain N-flank
C-flank
Extracellular Intracellular
TIR domain
Toll Drosophila
IL-1R Mammals
Figure 1 | Structure of the Toll and IL-1 receptors. The ectodomain of Toll comprises leucine-rich repeats (LRRs) that are flanked by cysteine-rich motifs (known as the N- and Cflanks). The ectodomain of the interleukin-1 receptor (IL-1R) comprises three immunoglobulin (Ig) domains. The intracellular Toll/IL-1R (TIR) domain of both Toll and the IL-1R interacts with TIR-domain-containing adaptor proteins (for example, Drosophila Myd88 or the mammalian MyD88) and signals through NF-B or NF-B-like molecules (FIG. 3).
Yet, little did I know how competitive and stimulating the genetic studies of Drosophila immunity would become.
Toll regulates the antifungal response
Freshly armed with the discovery that imd regulates Diptericin but not Drosomycin expression, I postulated that the Toll pathway could be a regulator of Drosomycin. So, in my previous experiments, I had selected the wrong target gene! This time, by checking the expression of a series of genes encoding antimicrobial peptides in Toll and Tollpathway mutants, my colleagues and I determined that this prediction was correct: after microbial infection, Drosomycin expression is regulated by the Toll pathway, whereas Diptericin expression is regulated by imd 30. It was also shown that not all of the components of the Toll pathway that regulate embryogenesis (BOX 1) have an immune function. For example, Easter, a serine protease that cleaves Sptzle and activates the Toll pathway during embryonic development31, does not regulate Toll during immunity. Similarly, several lines of evidence indicated that the NF-B-like factor Dorsal, which is activated by the Toll pathway during development, is not required for the expression of antimicrobial peptides. Initially, Tony Ip and colleagues32 (in Worcester, United States) generated a fly line carrying a small deletion that spans both Dif and Dorsal. By re-introducing
When it was first shown that Drosomycin and Diptericin are not regulated by the same signalling pathways, I wondered whether the expression of each gene is induced in response to different types of infection. To test this hypothesis, the levels of Drosomycin and Diptericin expression were compared after infecting fruit flies with different types of microorganism. The results were clear: the gene encoding the antibacterial peptide, Diptericin, was most highly induced by Gram-negative bacteria, whereas Grampositive bacteria and fungi were the strongest inducers of the gene encoding the antifungal peptide, Drosomycin. The simplest interpretation of these results is that Toll and imd are activated differentially by different types of microorganism. I consider that the best experiment supporting this interpretation was the demonstration that flies dusted with spores of Beauveria bassiana (a fungus that infects insects) specifically express the antifungal peptide Drosomycin but do not express antibacterial
peptides, indicating the selective activation of the Toll pathway34. Furthermore, Toll-deficient flies succumb rapidly to B. bassiana infections. This experiment demonstrates that flies mount immune responses that are adapted to the invading microorganism. It was also the first demonstration, using a natural route of infection, that showed that Toll signalling is required to combat a true insect pathogen. These observations that the Toll pathway is more responsive to Gram-positive bacteria and fungi, whereas IMD regulates responses to Gram-negative bacteria challenged the prevailing dogma that innate immune mechanisms provide an entirely nonspecific response to infection. On the contrary, the separation of Toll- and IMDmediated responses enables the fly to mount an immune response that is, to some extent, adapted to the species of aggressor. The existence of a degree of specificity in innate immune responses is not restricted to Drosophila, and determining how infections by distinct microorganisms shape the innate immune responses of vertebrates is currently the focus of intense study. When we published the data on selective Toll and imd activation in 1997 (REF. 34), we used the term adapted immune response, rather than specific immune response, to indicate how Drosophila uses several signalling pathways to discriminate between microorganisms and mount microorganism-specific immune responses. Our experiments on selective Toll or imd activation also taught us that the types of microorganism used, as well as the infection procedure (natural versus artificial infection), influence immune responses. Therefore, some inconsistencies in the reports on Drosophila immunity can probably be attributed to the way immune responses are triggered a common observation in immunology. For example, I now realize that our success in identifying the function of Toll
Figure 2 | Toll mutants are highly susceptible to fungal infection. Toll-deficient fruit flies (shown), but not wild-type fruit flies, succumb rapidly to infection with the fungus Aspergillus fumigatus. This image is reproduced with permission from REF. 30 (1996) Cell Press.
524

in the Drosophila immune response was partly because we routinely used a mixture of Gram-negative and Gram-positive bacteria to infect flies, whereas other groups only used Gram-negative bacteria. The Gram-positive bacteria strongly activated the Toll pathway and enabled us to discern the role of Toll in inducing Drosomycin expression.
The Toll and IMD paradigm
Toll pathway Fungi ? Sptzle Necrotic (serpin) PSH Proteases GNBP1 PGRP-SA Gram-positive bacteria (Lys-type peptidoglycan)
IMD pathway Gram-negative bacteria (DAP-type peptidoglycan)
All of these findings established a model of two potentially independent pathways that regulate the expression of the Drosophila genes encoding antimicrobial peptides. This model was tested rapidly when several groups began to use the power of Drosophila genetics to identify new factors that regulate Drosophila immune responses. After imd, the next gene identified to control antibacterial responses was characterized in Dan Hultmarks laboratory (in Umea, Sweden) in 1999 (REFS 35,36). His group demonstrated that a deletion of the Relish gene which encodes a third Drosophila NF-B-like protein produces phenotypes that are similar to those of fruit flies carrying the imd mutation35,36. Subsequently, several successful forward genetic screens identified other mutations that, similar to the imd and Relish mutations, render flies highly susceptible to Gram-negative bacterial infections3743. Surprisingly, none of these mutations affect any detectable functions of the Toll pathway. Genetic epistasis studies and molecular analysis of gene function show that imd, Relish and these other genes encode components of a signalling pathway, which is completely distinct from the Toll pathway and is essential for combating Gram-negative bacterial infection38,39,42,4448 (FIG. 3). Today, the Toll and IMD pathways have emerged as a simple paradigm of innate immune-response regulation in animals, showing how two distinct signalling cascades can modulate the expression of a complex transcriptional programme in response to different pathogens (FIG. 3). This model disputes prevailing views of innate immunity by indicating the existence of specificity and the absence of redundancy in Drosophila innate immune responses. There was initially some resistance to a simple genetic model of two separate NF-B-like signalling pathways in fruit flies, perhaps because studies of mammalian NF-B regulation (in cultured cells) indicate intricate and convergent networks of signalling cascades. Although it is possible that genetic analysis has simplified our vision of the Drosophila immunoresponsive-signalling pathways, I suspect that complexity exists in the capacity of these two pathways to integrate
Extracellular Intracellular Myd88 TIR (MyD88) DD
Toll (TLR)
PGRP-LC
? IMD (RIP)
Tube
Pelle (IRAK) D E D
Fadd (FADD)
TAK1 (TAK1) Cactus (IB) ANK Dorsal (NF-B) ? DIF (NF-B) IRD5 (IKK-) KEY (IKK- / NEMO)
DREDD (caspase-8)
Relish (NF-B) REL ANK
Nuclear membrane Antimicrobial peptides Bbinding motif
Figure 3 | The Toll and Imd pathways. The genes that encode antimicrobial peptides are regulated by a balance between two signalling pathways: the Toll pathway, which is activated mainly by fungi and Grampositive bacteria, and the Immune deficiency (IMD) pathway, which is activated mainly by Gram-negative bacteria. Depending on the variant of the B-binding motif present in the promoter, the genes encoding antimicrobial peptides are more sensitive to the TollDIF cascade (for example, Drosomycin), the IMDRelish cascade (for example, Diptericin) or are co-regulated. Toll is activated by binding to a cleaved form of Sptzle, which is processed by proteolytic cascades that are activated by secreted recognition molecules (such as the peptidoglycan-recognition protein PGRP-SA and Gram-negative-bacteria-binding protein 1, GNBP1). PGRP-SA might bind to a lysine-type peptidoglycan found in Gram-positive bacteria. Intracellular signal transduction (BOX 1) regulates the nuclear translocation of the nuclear factor-B (NF-B)like proteins DIF and Dorsal. The IMD pathway is probably triggered by an interaction between the transmembrane receptor PGRP-LC and peptidoglycan from Gram-negative bacteria (diaminopimelate (DAP)-type peptidoglycan). Following PGRP-LC activation, the death-domain (DD) adaptor protein, IMD, is recruited and binds to Fadd, which interacts with the caspase DREDD (Death-related ced-3/Nedd2-like protein). DREDD has been shown to associate with Relish (which it might cleave directly) after Relish has been phosphorylated by the Drosophila IKK (inhibitor of NF-B (IB)-kinase) complex, which comprises Immune response deficient 5 (IRD5) and Kenny (KEY). The IKK complex is itself activated by TAK1 (Transforming-growth-factor--activated kinase 1) (a mitogen-activated protein kinase kinase kinase) in an IMD-dependent manner. After cleavage, the REL domain of Relish moves to the nucleus, where it regulates the transcription of genes with immune function77,78. Vertebrate homologues are indicated in parentheses. ANK, ankyrin-repeat domain; DED; death-effector domain; DIF, Dorsal-related immunity factor; FADD, FASassociated death domain; IRAK, interleukin-1-receptor-associated kinase; MyD88, myeloid differentiation primary-response protein 88; NEMO, NF-B essential modulator; PSH, Persephone; RIP, receptorinteracting protein; serpin, serine-protease inhibitor; TIR domain, Toll/interleukin-1-receptor domain; TLR, Toll-like receptor.
VOLUME 4 | JULY 2004 | 5 2 5
PERSPECTIVES
many external factors (nature of infectious agents, mode of infection) and internal factors (tissue identity, physiological state) and to transduce these variables into a complex output (the sequential expression of many genes with immune function) that is far from understood.
Distinct functions for Toll and TLRs
TLRs, although they both have important roles in innate immune responses that are mediated by NF-B/NF-B-like molecules.
Continued value of Drosophila models
1.
2.
The studies of Toll stimulated research on Toll-like receptors (TLRs) in mammals. In 1997, the identification of a human Toll homologue in expressed sequence tag (EST) databases and the analysis of its function indicated that, similar to Toll, TLRs were linked to NF-B signalling and were probably important regulators of immunity49. Subsequently, a series of remarkable studies with mutant mice clearly demonstrated that TLRs function as recognition receptors for many microbial and viral ligands and control numerous aspects of both the innate and adaptive immune responses50,51. Interestingly, although TLRs function as direct recognition receptors for microbial components, the recognition of lipopolysaccharide by TLR4 is mediated by a complex that includes CD14 and MD2, in addition to TLR4. The identification of TLRs as receptors for microbial and viral ligands was not predicted by studies carried out in flies because Toll clearly does not interact directly with microbial products but is activated by an endogenous ligand, Sptzle, during the immune response30,52. Although the Drosophila genome encodes eight additional Toll molecules, surprisingly, none of these has yet been shown to participate directly in microbial sensing, and only one (Toll-9) has been implicated in the regulation of immune responses (using a cell-culture assay)5355. In contrast to the rapid identification of TLRs and their ligands, the Drosophila molecules involved in microbial recognition remained unknown until recently. At present, we know that some aspects of microbial recognition in flies are mediated by peptidoglycan-recognition proteins (PGRPs) and Gram-negative-bacteria-binding proteins (GNBPs)43,5660 two protein families identified, initially in other insects, by their capacity to bind microbial components6164 (FIG. 3). Furthermore, unlike mammals, the Drosophila immune system recognizes Gram-negative bacteria by detecting a specific form of peptidoglycan and not by sensing lipopolysaccharide65. Therefore, it seems that insects and mammals might use different strategies to detect microorganisms and that functional differences exist between the Drosophila Toll molecules and the human
The conservation of some immune responses in insects and mammals has produced an exchange of results and ideas that have invigorated the field of innate immunity. The discovery of TLRs was a turning point in the study of the mammalian immune system and opened numerous avenues of research. This discovery also validated the fruit fly as a model for analysing immune-response pathways. One advantage of the Drosophila model is that it offers a different perspective on the battle between pathogens and their hosts, and it allows fundamental questions in immunity to be addressed, without the added complexity of an adaptive immune system. A second advantage of the Drosophila model is illustrated by the studies of Toll and imd that have been described here, during which about 100,000 flies were infected. This seems reasonable when one understands that 300400 flies can be infected in one hour. Clearly, such exhaustive studies are not possible in vertebrate immunology. Furthermore, comparing the strategies that different species have developed to fight microbial infection is essential if we want to fully understand the immune system and not become confused by its intrinsic complexity. These comparisons, however, also need to consider the varied methods that are used to study immune responses in different species, because different approaches can influence the models that we build. As discussed in this Landmark, the immune-signalling pathways in Drosophila were elucidated as an extension of the discovery of antimicrobial peptides. However, it is not difficult to imagine that a simple screen for mutations that caused susceptibility to infection by pathogens would also have identified the Drosophila NF-B-like pathways. Such a project was technically possible several decades ago, but at that time, fly geneticists had mostly deserted the field of physiology and were focusing their attention on Drosophila development. Finally, although the black box of antimicrobial-peptide gene expression has lost some of its mystery, antimicrobial peptides are only part of the large arsenal of insect immune responses to pathogens. As a result, the future holds the promise of many exciting discoveries that will probably further impact on mammalian studies.
Bruno Lemaitre is at the Centre de Gntique Molculaire, Centre National de la Recherche Scientifique, F-91198 Gif-sur-Yvette, France. e-mail: lemaitre@cgm.cnrs-gif.fr
doi:10.1038/nri1390
3.
4.
5.
6.
7.
8.
9. 10. 11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
Steiner, H., Hultmark, D., Engstrom, A., Bennich, H. & Boman, H. G. Sequence and specificity of two antibacterial proteins involved in insect immunity. Nature 292, 246248 (1981). Hultmark, D. et al. Insect immunity. Attacins, a family of antibacterial proteins from Hyalophora cecropia. EMBO J. 2, 571576 (1983). Samakovlis, C., Kimbrell, D., Kylsten, P., Engstrom, A. & Hultmark, D. The immune response in Drosophila: pattern of Cecropin expression and biological activity. EMBO J. 9, 29692976 (1990). Kylsten, P., Samakovlis, C. & Hultmark, D. The cecropin locus in Drosophila; a compact gene cluster involved in the response to infection. EMBO J. 9, 217224 (1990). Wicker, C. et al. Insect immunity. Characterization of a Drosophila cDNA encoding a novel member of the Diptericin family of immune peptides. J. Biol. Chem. 265, 2249322498 (1990). Sun, S. C., Lindstrom, I., Lee, J. Y. & Faye, I. Structure and expression of the attacin genes in Hyalophora cecropia. Eur. J. Biochem. 196, 247254 (1991). Kappler, C. et al. Insect immunity. Two 17bp repeats nesting a B-related sequence confer inducibility to the Diptericin gene and bind a polypeptide in bacteriachallenged Drosophila. EMBO J. 12, 15611568 (1993). Engstrom, Y. et al. B-like motifs regulate the induction of immune genes in Drosophila. J. Mol. Biol. 232, 327333 (1993). Li, Q. & Verma, I. M. NF-B regulation in the immune system. Nature Rev. Immunol. 2, 725734 (2002). Gay, N. & Keith, F. Drosophila Toll and IL-1 receptor. Nature 351, 355356 (1991). Schneider, D. S., Hudson, K. L., Lin, T. Y. & Anderson, K. V. Dominant and recessive mutations define functional domains of Toll, a transmembrane protein required for dorsalventral polarity in the Drosophila embryo. Genes Dev. 5, 797807 (1991). Heguy, A., Baldari, C. T., Macchia, G., Telford, J. L. & Melli, M. Amino acids conserved in interleukin-1 receptors (IL-1Rs) and the Drosophila Toll protein are essential for IL-1R signal transduction. J. Biol. Chem. 267, 26052609 (1992). Nusslein-Volhard, C. in Determinants of Spatial Organization (eds Subtelny, S. & Konigsberg, I. R.) 185211 (Academic, New York, 1979). Steward, R. Dorsal, an embryonic polarity gene in Drosophila is homologous to the vertebrate protooncogene c-rel. Science 238, 692694 (1987). Ip, Y. T., Kraut, R., Levine, M. & Rushlow, C. A. The Dorsal morphogen is a sequence-specific DNA-binding protein that interacts with a long-range repression element in Drosophila. Cell 64, 439446 (1991). Thisse, C., Perrin-Schmitt, F., Stoetzel, C. & Thisse, B. Sequence-specific transactivation of the Drosophila twist gene by the dorsal gene product. Cell 65, 11911201 (1991). Pan, D. J., Huang, J. D. & Courey, A. J. Functional analysis of the Drosophila twist promoter reveals a dorsal-binding ventral activator region. Genes Dev. 5, 18921901 (1991). Gerttula, S., Jin, Y. S. & Anderson, K. V. Zygotic expression and activity of the Drosophila Toll gene, a gene required maternally for embryonic dorsalventral pattern formation. Genetics 119, 123133 (1988). Letsou, A., Alexander, S., Orth, K. & Wasserman, S. A. Genetic and molecular characterization of tube, a Drosophila gene maternally required for embryonic dorsoventral polarity. Proc. Natl Acad. Sci. USA 88, 810814 (1991). Hecht, P. M. & Anderson, K. V. Genetic characterization of tube and pelle, genes required for signaling between Toll and Dorsal in the specification of the dorsalventral pattern of the Drosophila embryo. Genetics 135, 405417 (1993). Reichhart, J. M. et al. Expression and nuclear translocation of the rel/NF-B-related morphogen Dorsal during the immune response of Drosophila. C. R. Acad. Sci. III, Sci. Vie 316, 12181224 (1993). Lemaitre, B. et al. Functional analysis and regulation of nuclear import of Dorsal during the immune response in Drosophila. EMBO J. 14, 536545 (1995). Ip, Y. et al. Dif, a dorsal-related gene that mediates an immune response in Drosophila. Cell 75, 753763 (1993). Rosetto, M., Engstrom, Y., Baldari, C. T., Telford, J. L. & Hultmark, D. Signals from the IL-1 receptor homolog, Toll, can activate an immune response in a Drosophila hemocyte cell line. Biochem. Biophys. Res. Commun. 209, 111116 (1995).
526

25. Fehlbaum, P. et al. Insect immunity: septic injury of Drosophila induces the synthesis of a potent antifungal peptide with sequence homology to plant antifungal peptides. J. Biol. Chem. 269, 3315933163 (1994). 26. Rizki, T., Rizki, R. & Grell, E. A mutant affecting the crystal cells in Drosophila melanogaster. Rouxs Arch. Dev. Biol. 188, 9199 (1980). 27. Sderhll, K. & Cerenius, L. Role of prophenoloxidaseactivating system in invertebrate immunity. Curr. Opin. Immunol. 10, 2328 (1998). 28. Lemaitre, B. et al. A recessive mutation, immune deficiency (imd), defines two distinct control pathways in the Drosophila host defense. Proc. Natl Acad. Sci. USA 92, 94659469 (1995). 29. Corbo, J. C. & Levine, M. Characterization of an immunodeficiency mutant in Drosophila. Mech. Dev. 55, 211220 (1996). 30. Lemaitre, B., Nicolas, E., Michaut, L., Reichhart, J. & Hoffmann, J. The dorsoventral regulatory gene cassette sptzle/Toll/cactus controls the potent antifungal response in Drosophila adults. Cell 86, 973983 (1996). 31. Chasan, R. & Anderson, K. V. The role of Easter, an apparent serine protease, in organizing the dorsalventral pattern of the Drosophila embryo. Cell 56, 391400 (1989). 32. Meng, X., Khanuja, B. S. & Ip, Y. T. Toll receptor-mediated Drosophila immune response requires Dif, an NF-B factor. Genes Dev. 13, 792797 (1999). 33. Rutschmann, S. et al. The Rel protein DIF mediates the antifungal but not the antibacterial host defense in Drosophila. Immunity 12, 569580 (2000). 34. Lemaitre, B., Reichhart, J. & Hoffmann, J. Drosophila host defense: differential induction of antimicrobial peptide genes after infection by various classes of microorganisms. Proc. Natl Acad. Sci. USA 94, 1461414619 (1997). 35. Dushay, M., Asling, B. & Hultmark, D. Origins of immunity: Relish, a compound Rel-like gene in the antibacterial defense of Drosophila. Proc. Natl Acad. Sci. USA 93, 1034310347 (1996). 36. Hedengren, M. et al. Relish, a central factor in the control of humoral but not cellular immunity in Drosophila. Mol. Cell 4, 827837 (1999). 37. Wu, L. P. & Anderson, K. V. Regulated nuclear import of Rel proteins in the Drosophila immune response. Nature 392, 9397 (1998). 38. Leulier, F., Rodriguez, A., Khush, R. S., Abrams, J. M. & Lemaitre, B. The Drosophila caspase Dredd is required to resist Gram-negative bacterial infection. EMBO Rep. 1, 353358 (2000). 39. Rutschmann, S. et al. Role of Drosophila IKK in a Tollindependent antibacterial immune response. Nature Immunol. 1, 342347 (2000). 40. Elrod-Erickson, M., Mishra, S. & Schneider, D. Interactions between the cellular and humoral immune responses in Drosophila. Curr. Biol. 10, 781784 (2000). 41. Lu, Y., Wu, L. P. & Anderson, K. V. The antibacterial arm of the Drosophila innate immune response requires an IB kinase. Genes Dev. 15, 104110 (2001). 42. Vidal, S. et al. Mutations in the Drosophila dTAK1 gene reveal a conserved function for MAPKKKs in the control of rel/NF-B-dependent innate immune responses. Genes Dev. 15, 19001912 (2001). 43. Choe, K. M., Werner, T., Stoven, S., Hultmark, D. & Anderson, K. V. Requirement for a peptidoglycan recognition protein (PGRP) in Relish activation and antibacterial immune responses in Drosophila. Science 296, 359362 (2002). 44. Silverman, N. et al. A Drosophila IB kinase complex required for Relish cleavage and antibacterial immunity. Genes Dev. 14, 24612471 (2000). 45. Stven, S., Ando, I., Kadalayil, L., Engstrm, Y. & Hultmark, D. Activation of the Drosophila NF-B factor Relish by rapid endoproteolytic cleavage. EMBO Rep. 1, 347352 (2000). 46. Georgel, P. et al. Drosophila immune deficiency (IMD) is a death domain protein that activates antibacterial defense and can promote apoptosis. Dev. Cell 1, 503514 (2001). 47. Leulier, F., Vidal, S., Saigo, K., Ueda, R. & Lemaitre, B. Inducible expression of double-stranded RNA reveals a role for dFADD in the regulation of the antibacterial response in Drosophila adults. Curr. Biol. 12, 9961000 (2002). 48. Naitza, S. et al. The Drosophila immune defense against Gram-negative infection requires the death protein dFADD. Immunity 17, 575581 (2002). 49. Medzhitov, R., Preston-Hurlburt, P. & Janeway, C. A human homologue of the Drosophila Toll protein signals activation of adaptive immunity. Nature 388, 394397 (1997). 50. Akira, S. Mammalian Toll-like receptors. Curr. Opin. Immunol. 15, 511 (2003). 51. Beutler, B., Hoebe, K., Du, X. & Ulevitch, R. J. How we detect microbes and respond to them: the Toll-like receptors and their transducers. J. Leukoc. Biol. 74, 479485 (2003). 52. Weber, A. N. et al. Binding of the Drosophila cytokine Spatzle to Toll is direct and establishes signaling. Nature Immunol. 4, 794800 (2003). 53. Tauszig, S., Jouanguy, E., Hoffmann, J. A. & Imler, J. L. Toll-related receptors and the control of antimicrobial peptide expression in Drosophila. Proc. Natl Acad. Sci. USA 97, 1052010525 (2000). 54. Imler, J. L. & Hoffmann, J. A. Toll receptors in Drosophila: a family of molecules regulating development and immunity. Curr. Top. Microbiol. Immunol. 270, 6379 (2002). 55. Ooi, J. Y., Yagi, Y., Hu, X. & Ip, Y. T. The Drosophila Toll-9 activates a constitutive antimicrobial defense. EMBO Rep. 3, 8287 (2002). 56. Michel, T., Reichhart, J. M., Hoffmann, J. A. & Royet, J. Drosophila Toll is activated by Gram-positive bacteria through a circulating peptidoglycan recognition protein. Nature 414, 756759 (2001). 57. Ramet, M., Manfruelli, P., Pearson, A., Mathey-Prevot, B. & Ezekowitz, R. A. Functional genomic analysis of phagocytosis and identification of a Drosophila receptor for E. coli. Nature 416, 644648 (2002). 58. Gottar, M. et al. The Drosophila immune response against Gram-negative bacteria is mediated by a peptidoglycan recognition protein. Nature 416, 640644 (2002). 59. Gobert, V. et al. Dual activation of the Drosophila Toll pathway by two pattern recognition receptors. Science 302, 21262130 (2003). 60. Pili-Floury, S. et al. In vivo RNA interference analysis reveals an unexpected role for GNBP1 in the defense against Gram-positive bacterial infection in Drosophila adults. J. Biol. Chem. 279, 1284812853 (2004). 61. Yoshida, H., Kinoshita, K. & Ashida, M. Purification of a peptidoglycan recognition protein from hemolymph of the silkworm, Bombyx mori. J. Biol. Chem. 271, 1385413860 (1996). 62. Lee, W., Lee, J., Kravchenko, V., Ulevitch, R. & Brey, P. Purification and molecular cloning of an inducible Gramnegative bacteria-binding protein from the silkworm, Bombyx mori. Proc. Natl Acad. Sci. USA 93, 78887893 (1996). 63. Ochiai, M. & Ashida, M. A pattern recognition protein for peptidoglycan. Cloning the cDNA and the gene of the silkworm, Bombyx mori. J. Biol. Chem. 274, 1185411858 (1999). 64. Kang, D., Liu, G., Lundstrom, A., Gelius, E. & Steiner, H. A peptidoglycan regonition protein in innate immunity conserved from insects to humans. Proc. Natl Acad. Sci. USA 95, 1007810082 (1998). 65. Leulier, F. et al. The Drosophila immune system detects bacteria through specific peptidoglycan recognition. Nature Immunol. 4, 478484 (2003). 66. Rubin, G. M. et al. Comparative genomics of the eukaryotes. Science 287, 22042215 (2000). 67. Anderson, K. V. & Nusslein-Volhard, C. in Pattern Formation: A Primer in Developmental Biology (eds Malacinski, G. M. & Bryant, S. V.) 269289 (Macmillan, New York, 1984). 68. Anderson, K. V., Bokla, L. & Nusslein-Volhard, C. Establishment of dorsalventral polarity in the Drosophila embryo: the induction of polarity by the Toll gene product. Cell 42, 791798 (1985). 69. Anderson, K. V., Jurgens, G. & Nusslein-Volhard, C. Establishment of dorsalventral polarity in the Drosophila embryo: genetic studies on the role of the Toll gene product. Cell 42, 779789 (1985). 70. Hashimoto, C., Hudson, K. & Anderson, K. The Toll gene of Drosophila, required for dorsalventral embryonic polarity, appears to encode a transmembrane protein. Cell 52, 269279 (1988). 71. Hashimoto, C., Kim, D. R., Weiss, L. A., Miller, J. W. & Morisato, D. Spatial regulation of developmental signaling by a serpin. Dev. Cell 5, 945950 (2003). 72. Ligoxygakis, P., Roth, S. & Reichhart, J. M. A serpin regulates dorsalventral axis formation in the Drosophila embryo. Curr. Biol. 13, 20972102 (2003). 73. Kambris, Z. et al. DmMyD88 controls dorsoventral patterning of the Drosophila embryo. EMBO Rep. 4, 6469 (2003). 74. Charatsi, I., Luschnig, S., Bartoszewski, S., NussleinVolhard, C. & Moussian, B. Krapfen/dMyd88 is required for the establishment of dorsoventral pattern in the Drosophila embryo. Mech. Dev. 120, 219226 (2003). 75. Belvin, M. P. & Anderson, K. V. A conserved signaling pathway: the Drosophila TollDorsal pathway. Annu. Rev. Cell Dev. Biol. 12, 393416 (1996). 76. Qiu, P., Pan, P. C. & Govind, S. A role for the Drosophila Toll/Cactus pathway in larval hematopoiesis. Development 125, 19091920 (1998). 77. Hultmark, D. Drosophila immunity: paths and patterns. Curr. Opin. Immunol. 15, 1219 (2003). 78. Ferrandon, D., Imler, J. L. & Hoffmann, J. A. Sensing infection in Drosophila: Toll and beyond. Semin. Immunol. 16, 4353 (2004).
Acknowledgements
I thank Jules Hoffmann and all of my former colleagues for the stimulating and encouraging environment in Strasbourg, France, and Nicolas Vodovar for providing the original illustration of figure 3. I am also indebted to the generous spirit of the Drosophila-research community, including Kathy Matthews at the Bloomington Stock Center, Bloomington, United States, and Iris Koch and Dirk Beuchle at the Tbingen Stock Centre, Tbingen, Germany, for providing the many fly stocks that made this work possible. My laboratory is supported by the Centre National de la Recherche Scientifique, France, and the Schlumberger and Bettencourt Foundations, France.

The author declares that he has no competing financial interests.
Online links
DATABASES The following terms in this article are linked online to: Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene CD14 | IB | IL-1R | MD2 | TLR4 FlyBase: http://flybase.bio.indiana.edu/ Bc | Cactus | Cecropin A1 | Dif | Diptericin | Dorsal | Drosomycin | Easter | Gastrulation defective | imd | Myd88 | Pelle | Relish | Serpin-27A | snail | Snake | Sptzle | Toll-9 | Tube | twist FURTHER INFORMATION Bruno Lemaitres lab: http://www.cnrs-gif.fr/cgm/immunity/enindex.html Access to this interactive links box is free online.
VOLUME 4 | JULY 2004 | 5 2 7
REVIEWS
CD4/CD8-LINEAGE DIFFERENTIATION IN THE THYMUS: FROM NUCLEAR EFFECTORS TO MEMBRANE SIGNALS
Rmy Bosselut
During thymocyte development, immature thymocytes that express both CD4 and CD8 genes must choose either a helper CD4+ or cytotoxic CD8+ T-cell fate. Over the past two years, there have been some important advances regarding T-cell lineage choice, including the identification of transcription factors required for CD4 gene silencing by CD8-lineage cells (RUNX3) or for CD4+ T-cell differentiation (GATA3), and a better understanding of how T-cell receptor (TCR) signalling correlates CD4/CD8-lineage differentiation to MHC specificity. This review summarizes these recent advances and highlights potential links between TCR signals and nuclear effectors of lineage differentiation.
POSITIVE SELECTION
A process in the thymus that selects thymocytes expressing T-cell receptors (TCRs) that are of intermediate avidity for self-peptideMHC complexes. TCR signals generated by this weak interaction cause thymocyte survival and differentiation into mature T cells, the TCRs of which can recognize foreign peptides bound to self-MHC molecules. Positive selection establishes the MHC-restricted T-cell repertoire.
Laboratory of Immune Cell Biology, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, Maryland 20892, USA. e-mail: remy@helix.nih.gov
doi:10.1038/nri1392
T cells meditate effective immune responses by recognizing antigen through cell-surface receptors (T-cell receptors, TCRs) composed of two variable (clonotypic) chains, which form the interface that is responsible for antigen binding, and four invariant chains1. These invariant chains (CD3, , and ) associate with the clonotypic components2,3 and contain specialized tyrosine-based motifs that are responsible for signal transduction. The most prevalent T-cell population expresses TCR - and -clonotypic chains, develops in the thymus and recognizes peptide antigens that are associated with classical MHC class I or class II molecules4. Most TCR-+ T cells also express one of two cell-surface glycoproteins that are important for TCR signalling, CD4 or CD8. These proteins, which are referred to as co-receptors, bind invariant regions of MHC class II (CD4) and class I (CD8) molecules, and contribute to TCR-signal transduction through their interactions with membrane-associated signalling molecules59. Expression of CD4 or CD8 defines two distinct T-cell lineages that differ both by their MHC specificity and by their function. Most CD4+ T cells are MHC class II restricted and function as T helper (TH) cells when activated, assisting effector components of the immune system (such as B cells) through cytokine secretion and
upregulation of expression of specific membrane ligands. By contrast, CD8+ T cells are MHC class I restricted and after activation, acquire cytotoxic properties that allow them to directly kill cells expressing their target antigen. As CD4+ and CD8+ T cells are derived from a common precursor pool of doublepositive (DP) thymocytes, so called because they express both CD4 and CD8 (FIG. 1a), the question arises as to how the concordance between MHC specificity and lineage is established. That is, which signals, if any, direct thymocytes to either lineage? What are the cytosolic intermediates and nuclear effectors that transform these signals into CD4- or CD8-specific gene-expression programmes? And how do these mechanisms match lineage choice to MHC specificity? Before addressing these issues, it is important to highlight that multiple developmental events accompany lineage choice during the differentiation of DP thymocytes into CD4+CD8 or CD4CD8+ single-positive (SP) thymocytes a process referred to as POSITIVE SELECTION. Of all of these events, the most crucial is the rescue of DP thymocytes from programmed cell death, and this is determined by the avidity of the expressed TCR for intrathymic self-peptideMHC complexes. Most DP thymocytes are short-lived, and die by neglect within a few days of their generation because their TCR
VOLUME 4 | JULY 2004 | 5 2 9
REVIEWS
fails to be engaged and to transduce signals10,11. By contrast, thymocytes that express a TCR of high avidity for self-peptideMHC complexes are actively eliminated by TCR-induced cell death negative selection. So, only a few DP thymocytes, the TCRs of which are of intermediate avidity for self-peptideMHC complexes, are rescued from cell death and differentiate into CD4+ or CD8+ SP T cells1216. The overlap of TCR-induced thymocyte survival and lineage differentiation has both set the stage for concepts and models of lineage choice and hindered their experimental assessment. Early investigations1719 of lineage choice focused on how the link between CD4/CD8 lineage and MHC specificity is established. It was initially proposed that lineage-differentiation signals are mechanistically coupled to positive-selection signals; that is, engagement of both the TCR and the co-receptor by MHC molecules determines (instructs) lineage choice20,21. In this case, there is no differentiation of thymocytes with mismatched lineage and MHC specificity: MHC class-II-induced signals promote positive selection and differentiation into CD4+ T cells, whereas MHC class-I-induced signals promote positive selection and differentiation into CD8+ T cells. An alternative model, referred to as the selective model, proposes that lineage choice is independent from MHC/TCR signals (that is, lineage choice is stochastic or induced by ligands other than MHC molecules). As a result, some thymocytes are mismatched; however, the selective model proposes that such cells are eliminated as they fail to carry out a subsequent intrathymic signalling step that requires MHC co-engagement of the TCR and the appropriate co-receptor2227.
Positive selection
CIS-REGULATORY ELEMENTS
DNA sequences located within or next to transcribed genes and that either increase (enhancers) or decrease (repressor or silencer, depending on their mechanism of action) gene transcription. Cis-regulatory elements act by recruiting trans-acting transcriptional activator or repressor proteins.
Distinguishing between these alternatives has been difficult for many reasons, of which two are directly linked to features of positive selection. First, mismatched thymocytes are difficult to detect because they die rapidly by neglect. Second, although cell-surface expression of CD4 and CD8 molecules defines the lineage of mature thymocytes and T cells, it cannot be used to assess lineage choice in immature thymocytes2830. For example, CD8 expression is downregulated by both CD4- and CD8-differentiating cells, causing thymocytes that are undergoing selection into either lineage to adopt a transitional CD4+CD8low cellsurface phenotype31,32 (FIG. 1b). Attempts to validate the instructive or selective models have generated a large amount of evidence that cannot be explained by either model as originally formulated33,34. Instead, as is discussed in the last part of this review, the concept that has emerged is that quantitative attributes of TCR signals determine lineage choice. One approach to by-pass the difficulties met in searching for lineage-choice signals stems from the fact that lineage differentiation eventually leads to the termination of CD4 or CD8 gene expression. Consequently, identifying the CIS-REGULATORY ELEMENTS and transcription factors that control CD4 and CD8 gene transcription is expected to lead to identification of the downstream end of signal-transduction cascades, from which pathways coming from the membrane can be elucidated. The goal of this review is to highlight recent findings of such an upward approach35, and to place them in the broader context of the cellular and molecular mechanisms that regulate CD4 and CD8 expression in the thymus.
b
SP CD4+ CD8 CD4+ CD8low DP CD4+ CD8+
a
TCR-
TCR-gene rearrangement TCR-
TCR- Pre-TCR CD4 CD8 DN CD4 CD8 Pre-TCR signal DP CD4+ CD8+ TCR signal
SP CD4+ CD8
MHC class II restricted CD4 CD4low CD8+
DN No TCR signal Neglect Short-lived Long-lived CD8 TCR signal SP CD4 CD8+ MHC class I restricted SP CD4 CD8+
Figure 1 | A simplified model of intrathymic differentiation. a | Thymocyte precursors enter the thymus and rearrange their T-cell receptor- (TCR-), TCR- and TCR- genes. Thymocytes that productively rearrange TCR- express the product of the rearranged allele on the cell surface as part of a pre-TCR complex, which also includes the pre-T- invariant chain and CD3 components, and differentiate into double-positive (DP) thymocytes. DP thymocytes are short-lived cells that rearrange their TCR- gene. The product of the rearranged TCR- allele associates with the TCR- chain and CD3 components, and a TCR- complex is expressed at the cell surface. The avidity of the TCR- for self-peptideMHC complexes determines the fate of the DP thymocyte. Thymocytes with high avidity for selfpeptideMHC complexes are eliminated by negative selection (not shown) whereas those with low avidity die by neglect in the thymic cortex. Cells with intermediate avidity for self-peptideMHC complexes survive and differentiate into mature CD8+ T cells (if MHC class I restricted) or CD4+ T cells (if MHC class II restricted). This process is known as positive selection and can be followed by characteristic phenotypic changes, such as the upregulation of CD5 or CD69 expression. b | Both MHC class-I- and class-II-induced TCR signals have the potential to downregulate CD8 gene expression before lineage choice31,32. As a result, the CD4+CD8low transitional population contains CD4- and CD8-lineage cells. By contrast, the CD4lowCD8+ transitional population is thought to contain only CD8-lineage cells28,29.
530
REVIEWS
being inactive on its own in transgenic reporter assays43. However, enhancer deletion studies also indicated marked redundancy among some elements. Single deletion of either E8I or E8II does not affect CD8 expression by DP thymocytes or thymic-derived CD8+ T cells40,41, whereas deletion of both E8I and E8II reduces CD8 expression levels in mature CD8+ T cells42 and causes VARIEGATED CD8 expression in DP thymocytes a phenotype similar to that observed in mice lacking E8V. The variegated pattern is characterized by a large reduction in the level of CD8 expression on a fraction of DP thymocytes, apparently stochastically chosen, contrasting with normal levels of CD8 expression on the remaining DP cells. In summary, the regulation of CD8 gene expression involves multiple stage- and lineagespecific enhancers; identifying the trans-acting factors they recruit will provide insight into the mechanisms of CD8-lineage differentiation37,44. CD4 gene cis-regulatory elements. Despite the symmetric patterns of CD4 and CD8 gene expression that is, simultaneous expression by immature thymocytes and mutually exclusive expression by mature T cells the mechanisms directing lineage-specific expression of CD4 and CD8 genes are distinct37,45. So, although several enhancers with T-cell-specific activity have been identified within or near the CD4 locus, of which one located ~10 kb upstream of the CD4 promoter seems to be crucial for CD4 expression in the T-cell lineage46 (FIG. 3), none of these elements has CD4-lineage-specific activity. Instead, a distinct DNA segment prevents CD4 expression by CD8-lineage T cells and by early CD4CD8 (double negative, DN) precursors of DP thymocytes47,48. Transgenic analyses and targeted deletion have identified the minimal element that has these activities as a 434 bp sequence located within the first intron of the CD4 gene, referred to as the CD4 silencer (FIG. 3a). Analyses of CD4 silencer function indicate that two distinct mechanisms, referred to as repression and silencing, operate to prevent inappropriate CD4 expression in the T-cell lineage49. Although these concepts are rapidly evolving and cannot be given a definitive definition, repression can be seen as an active inhibition of transcription, operating through the sequence-specific recruitment of repressor proteins that act on the transcription machinery. This definition is based on the fact that repression requires the integrity of both the cisregulatory elements and the repressor proteins, and that repression is not an all-or-nothing phenomenon, presumably because gene transcription at any given time depends on the balance of repressor and activator protein activities. By contrast, silencing is proposed to suppress gene expression by closing the gene to access by transcription machinery. Two corollaries of that definition are that, once established, silencing is expected to be an all-or-nothing phenomenon (the locus is either open or closed) and not to require sequence-specific repressor proteins. These two mechanisms share the task of controlling CD4 gene expression in a stage-specific manner: repression seems to work in DN thymocytes, whereas silencing operates in post-selection CD8-lineage cells.
ENHANCERS
Control of CD4 and CD8 gene expression
Control elements within DNA to which regulatory proteins bind, thereby influencing the rate of gene transcription; enhancers function in an orientation- and positionindependent manner (that is, they can act either upstream or downstream of or in an intron).
VARIEGATED
A phenomenon characterized by the expression of a gene by only a fraction of the cells, apparently randomly chosen, in a population of cells that are otherwise of the same developmental and functional status. Variegation is thought to reflect all-or-nothing changes in chromatin organization.
CD8 gene cis-regulatory elements. Mammals have two CD8 genes that are closely linked within a single genomic locus, referred to as CD8A and CD8B, encoding CD8 and CD8, respectively36. DP thymocytes and most peripheral CD8 + T cells transcribe both CD8A and CD8B and express cell-surface CD8 heterodimers. By contrast, some specialized T-cell populations, such as intraepithelial lymphocytes, transcribe only CD8A and express cell-surface CD8 homodimers. In CD8+ T cells, the lineage specificity of CD8 gene expression is thought to result from activity at a minimum of five ENHANCERS (referred to as E8I to E8V), located within the CD8 locus and colocalizing with DNase hypersensitivity sites3743 (FIG. 2). When evaluated individually, by their ability to direct expression of heterologous reporter genes in transgenic mice, all five CD8 enhancers, with the possible exception of E8IV, are CD8-lineage specific that is, they direct reporter-gene expression in DP thymocytes or CD8+ T cells, but not in CD4+ T cells. Furthermore, they are active during defined developmental stages: E8III is active in DP thymocytes41, E8I in mature CD8 SP thymocytes and CD8+ T cells38,39, and E8II in both DP thymocytes and mature CD8+ T cells41. Enhancer deletion studies have confirmed the importance of CD8 enhancers and extended the findings of transgenic analyses. For example, deletion of the enhancer associated with DNase hypersensitivity cluster II (referred to here as E8V43) affects CD8 expression by both DP thymocytes and CD8+ T cells, despite E8V
Mature Thymocytes T cells
Enhancer activity
CD4 CD8 CD4 SP CD8 SP DP
+ + + +
+ + +
+ +
+ + +
Enhancer location E8IV E8III DNase hypersensitivity sites CD8 locus 5 CD8B CD8A 3 E8II E8I Cluster III E8V Cluster II Cluster IV
Figure 2 | Cis-regulatory elements that control CD8 gene expression. CD8A and CD8B genes are depicted (vertical bars: exons). Genetic studies using segments (horizontal bars) of the CD8 locus inserted upstream of a reporter gene have identified five CD8 enhancers referred to as E8I to E8V (REFS 3841). These enhancers co-localize with three clusters of DNase hypersensitivity sites (which are used as the basis for enhancer designation in an alternative nomenclature37,39,43). An additional cluster (not shown) located 3 of the CD8A gene has not yet been associated with a cis-regulatory element. The developmental stage of activity of these enhancers, as determined by transgenic and deletion studies, is indicated above each bar. The activity of E8IV is deduced from analyses of transgenic constructs containing E8III or E8II and E8III in addition to E8IV41; it is unknown whether E8IV has autonomous enhancer activity and what its lineage specificity is, if any. E8V lacks enhancer activity on its own in transgenic reporter analyses but has been shown by homologous recombination to promote CD8 expression by both thymocytes and CD8+ T cells43. Enhancer-deletion studies40,41 have shown that E8I is dispensable for CD8 expression by thymocytes or thymus-derived CD8+ T cells, but is required for CD8 expression by CD8+ intraepithelial lymphocytes.
VOLUME 4 | JULY 2004 | 5 3 1
REVIEWS
level of CD4 expression. Second, conditional silencer deletion indicates that this element is not required for the maintenance of CD4 silencing in mature CD8lineage cells50. That is, although silencer deletion at or before the DP stage causes persistent CD4 expression by mature CD8+ T cells, silencer deletion in mature peripheral CD8+ T cells does not cause CD4 re-expression. So, the silencer is required to establish silencing during CD8+ T-cell development in the thymus, but not to maintain silencing in mature CD8+ T cells. Importantly however, the establishment of silencing in CD8-lineage thymocytes remains silencer dependent until the completion of positive selection50.
Silencer-binding proteins
CD4
E4Pro 5
Exon 1 Silencer
Exon 2
E4DP 3
434 Transgenic reporter analysis 194 172190 244300 Deletion analysis
51130 CD4 silencer binding factors RUNX Unknown Unknown MYB HES1 SAF c 2 + 2 1+2 1+3 2
167257 2 1 3
279429
Figure 3 | Cis-regulatory elements controlling CD4 gene expression. a | Several cisregulatory elements have been identified within and next to the CD4 locus by transgenic reporter studies. With the exception of the silencer, none of these elements has been shown by gene-deletion analyses to be required for the control of CD4 expression. CD4 expression by T cells requires its promoter as well as a 5 enhancer element referred to as the proximal enhancer (E4Pro). A 5 distal enhancer (not shown) has also been implicated in the control of CD4 expression145, and a double-positive (DP)-specific 3 enhancer, E4DP, cooperates with the E4Pro to promote CD4 expression in DP thymocytes51,146148, although it lacks enhancer activity on its own. The CD4 silencer, located in the first intron of the mouse and human CD4 genes46,47,149, is required to confine CD4 expression to the CD4 lineage, as CD4 enhancers are active in both CD4- and CD8-lineage cells. b | The 434 bp minimal silencer is shown. Regions of functional importance for CD4 repression (in double-negative thymocytes) and silencing (in CD8-lineage cells), as determined by transgenic reporter assays51 and by gene deletion49,52, are shown. Single deletion of the 5 (51130) or central (167257) regions completely abolished silencing, whereas deletion of the 3 regions caused variegated CD4 expression. Point mutations in four DNA sites that affect silencing are indicated: two of them, 2 and 2 (REF. 49), match the consensus sequence for the binding of Runt-related transcription factor (RUNX). Factors binding to the other two sites identified by knocked-in mutations are not yet known. Although site 1 overlaps with a MYB-binding site, mutagenic studies indicate that the binding site recognized by the site 1-specific factor is distinct from that of MYB. The HES1-binding site overlaps with the upstream RUNX site150. Two binding sites for the putative transcription factor silencer-associated factor (SAF) are indicated53. c | Combinations of point mutations in the CD4 silencer that completely disrupt silencer activity.
KNOCKED-IN MUTATIONS
Mutations introduced into a gene locus using homologous recombination techniques. Unlike knockout mutations, which use large deletions or insertions to eliminate the function of the target gene, knock-in mutations are intended to change the function of the target locus, generally through point mutation.
The distinction between repression and silencing relies on two pieces of evidence. First, mutations within the 434 bp minimal silencer segment have distinct consequences in CD8-lineage T cells and DN thymocytes49. Silencing CD4 by CD8-lineage cells is an all-or-nothing phenomenon: partial silencer deletions result in a variegated CD4-expression pattern, characterized by CD4 expression by an apparently random fraction of CD8+ T cells and by complete silencing in the remaining CD8+ T cells. By contrast, the same partial silencer deletions cause unimodal CD4 expression by DN thymocytes and, when combined, have additive effects on the
After characterization of the CD4 silencer, the next step in the upward strategy was to identify protein-binding sites within this element and to assess the function of the protein(s) they recruit a task independently carried out by two groups. One approach defined three protein-binding sites within the 434 bp core silencer sequence, the importance of which was assessed in transgenic reporter analyses51 (FIG. 3b). These studies indicated redundancy in silencer function, as mutation of at least two of these sites (including the central one) was required to abolish silencing. A second, more recent study49 has characterized three functionally important regions within the CD4 silencer that partly overlap with those identified in the earlier study51. Further experiments using KNOCKED-IN MUTATIONS defined four elementary active sites within the silencer49,52 (FIG. 3b,c). Point mutations at any of these sites result in inappropriate CD4 expression by CD8+ T cells. Interestingly, single mutations cause variegated CD4 expression, confirming the discrete nature of silencing, whereas combined mutations abolish silencer function and cause uniform CD4 expression by all CD8+ T cells. An important step towards the identification of silencing factors was the realization that two of the four sites bind RUNX (Runtrelated transcription factor) transcriptional regulators52. Besides RUNX proteins, several other molecules bind the silencer in vitro at sites that are defined as functionally important in vivo. A factor known as silencerassociated factor (SAF) binds to the central region of the silencer53 (FIG. 3b); however, the function of SAF in silencing remains unclear, as isolated mutations of the SAF-binding site do not impair silencing49,53. The role of two additional proteins in CD4 silencing, the transcriptional repressor HES1 (a target of Notch signals) and MYB (the MYB proto-oncogene product), will be discussed further later. RUNX proteins as CD4-silencing factors during CD8lineage differentiation. The RUNX family of proteins (BOX 1) is defined by a DNA-binding motif known as the Runt domain and includes three proteins in mammals (RUNX13), all of which are expressed in the thymus52. Whereas binding of the Runt domain to the silencer element can be shown in vitro, the role of each Runx gene in CD4 expression had to be investigated using RUNX1-, RUNX2- or RUNX3-deficient thymocytes. These studies
532
REVIEWS
According to this hypothesis, the functional outcome of RUNX recruitment to the silencer is determined by other factors being recruited to the CD4 locus. In support of this, one silencer-binding site (site 3; FIG. 3b) seems to be more important for CD4 silencing in CD8+ T cells than for CD4 repression in DN thymocytes49. Second, it is possible that the proteins themselves carry distinct functions: that is, in DN thymocytes, where all three RUNX proteins are expressed, RUNX1 but not RUNX2 or RUNX3 molecules could repress CD4 expression. The NotchHES1 connection. Although the discovery of RUNX proteins as CD4-silencing factors is an example of using the upward approach to understand CD4/CD8-lineage differentiation, the possibility that HES1 has a role in CD4 silencing came from research on CD4 expression and Notch signalling59. Notch proteins form a family of four related receptors that are crucial to many developmental processes, although their role in CD4/CD8-lineage differentiation remains controversial60,61. An initial gain-of-function study showed that constitutively active Notch-1 promotes CD8-lineage differentiation62. However, this conclusion was disputed by subsequent gain-of-function analyses6365, and more recently by gene-deletion studies showing that neither Notch-1 nor Notch-2 gene disruption detectably alters CD4/CD8-lineage differentiation66,67. As thymocytes express all four Notch genes68 it is possible that functional redundancy explains the latter findings. One way to by-pass this difficulty is to assess the role of Notch-target genes in CD4/CD8-lineage differentiation. One such Notch target is the transcription factor HES1. Hes1 gene transcription is stimulated by complexes associating a transactivating moiety that is, a Notch intracellular domain generated by proteolytic cleavage after Notch-receptor ligation to a DNAbinding moiety composed of a protein known as CSL69. It was an exciting finding that the CD4 silencer has a binding site for HES1, and that, in cooperation with another transcription factor, MYB, HES1 promotes the activity of the CD4 silencer in vitro59. Mutation of the HES1-binding site impairs silencing in transgenic reporter constructs, further suggesting a role for HES1 in CD4 silencing. However, this conclusion has been questioned by two subsequent studies. First, the putative HES1-binding site overlaps with one of the RUNXbinding sites, making it difficult to determine whether in the mutagenesis study, HES1 binding or RUNX3 binding was disrupted49. Second, analyses of T-cell differentiation in THYMIC ORGAN CULTURES has shown that Hes1 gene disruption does not affect CD4 silencing or CD8-lineage differentiation68, arguing against a role of HES1 in either process. Perhaps more important to the role of Notch signalling in lineage choice will be studies on the DNAbinding protein CSL itself, as it mediates Notch-induced transcriptional activation in most cell types analysed so far. Conditional disruption analyses have shown that CSL is required for the differentiation of bone-marrow
Box 1 | RUNX proteins The RUNX family of transcriptional regulators includes three proteins in mammals, designated RUNX1 (also known as AML1, CBF-2 or PEBP2B), RUNX2 (also known as AML3, CBF-1 or PEBP2A) and RUNX3 (also known as AML2, CBF-3 or PEBP2C)142. Previous designations for these proteins reflect the fact that they were initially characterized as the products of genes that are involved in chromosomal translocations common in acute myeloid leukaemias (AML designations) and by their ability to bind viral enhancers, and so the polyomavirus enhancer binding protein (PEBP) or core-binding factors (CBF) nomenclatures. The family is defined by a 128-amino-acid motif known as the Runt domain, which mediates sequence-specific DNA binding and heterodimerization with the CBF- protein, which is essential for RUNX protein function. RUNX protein target sites include the consensus sequence YGYGGT, where Y is a pyrimidine143. RUNX proteins function as transcriptional activators or transcriptional repressors, presumably depending on the DNA context surrounding their binding sites and on the simultaneous recruitment of co-repressor factors142144. For example, RUNX proteins mediate repression of some genes by binding members of the Groucho family of repressors, at least in part through a conserved Val-Trp-Arg-Pro-Tyr motif located in their carboxy-terminal domain143,144. Although the sequences of RUNX proteins diverge outside of the Runt domain, there is evidence that these molecules are functionally interchangeable57,58.
HEMIZYGOUS
A genotype characterized by the presence of a wild-type and a non-functional allele.

THYMIC ORGAN CULTURES
A technique allowing the culture of thymic lobes taken from fetal or neonatal mice. Thymic organ culture does not disrupt thymocytestromal-cell interactions and allows manipulation of the extracellular milieu in which thymocytes develop, thereby combining advantages of in vivo and in vitro approaches.
showed impaired silencing in RUNX3-deficient CD8lineage thymocytes and mature peripheral T cells, whereas silencing is normal in their RUNX1 or RUNX2 counterparts52,54. So, RUNX3 is the most important RUNX factor for CD4 silencing, a finding that was reproduced by knocking-down RUNX3 expression in developing thymocytes with Runx3 antisense oligonucleotides55. However, RUNX3-deficient peripheral CD8+ T cells do not uniformly express CD4, but rather display a variegated CD4 expression pattern similar to that caused by mutations at single RUNX-binding sites. As germline disruption of both the RUNX-binding sites in the silencer results in unimodal CD4 expression by all peripheral CD8+ T cells, the variegated phenotype caused by Runx3 disruption indicated that there might be compensation by other RUNX-family members. Confirmation of this was provided by Runx3/Runx1+/ mice, in which all CD8+ T cells expressed CD4, indicating partial compensation of RUNX3 deficiency by homozygous, but not HEMIZYGOUS, RUNX1 expression54. As germline Runx1 disruption prevents the establishment of definitive haematopoiesis56, the role of RUNX1 in CD4 expression was studied using conditional gene disruption. Although Runx1 disruption did not affect CD4 silencing in CD8+ T cells, it caused CD4 de-repression in DN thymocytes, in which Runx3 disruption had no effect52. As Runx2 disruption did not affect T-cell development, the tentative picture emerging from these analyses suggests that RUNX3 is mainly a CD4-silencing factor during CD8-lineage differentiation and RUNX1 is mainly a CD4-repressing factor in DN thymocytes. Despite their similarity, the functional redundancy between these proteins is only partial, which can be explained by two non-exclusive hypotheses. First, it is possible that RUNX proteins are functionally interchangeable57,58 and that their apparent specificity of function simply reflects that, in the thymus, RUNX1 is preferentially expressed by DN cells52.
VOLUME 4 | JULY 2004 | 5 3 3
REVIEWS
precursors into the T-cell lineage70, as is Notch-1 itself 71; consequently, targeting CSL deletion to DP thymocytes should shed light on the role of Notch in CD4/CD8-lineage differentiation.
Targeting RUNX3 action to CD8-lineage cells
Several factors cooperate with RUNX3 to promote CD4 silencing. As RUNX3 is at present the only sure lead to silencer function, it is important to understand how RUNX3 targets CD4 silencing to CD8-lineage T cells. As indicated by conditional silencer deletion analyses50, the mechanisms that establish silencing in CD8-lineage thymocytes are probably different from those that maintain silencing in mature CD8+ T cells. In fact, as the CD4 silencer element is dispensable for the maintenance of silencing50, it is probable that RUNX3 has no role in this effect (although conditional Runx3 gene disruption will be necessary to demonstrate this). So, a discussion of RUNX3-mediated CD4 silencing must focus on thymocytes. Although still incomplete, experimental evidence indicates that RUNX3-mediated CD4 silencing involves two mechanisms: cooperation with other DNA-binding factors, such as those recruited to the two non-RUNX functional sites of the silencer (sites 1 and 3; FIG. 3b,c), and recruitment of chromatin-modifying enzymes. Chromatin remodelling and CD4 expression. Besides sequence-specific DNA-binding proteins, there is evidence that the control of CD4 gene expression involves CHROMATIN-REMODELLING COMPLEXES. These complexes are multi-protein assemblies that use cellular ATP to promote nucleosome displacement and are thereby thought to control the binding of transcriptional regulators to promoter and enhancer elements37,72,73. One such complex in mammals, known as BAF (BRG1-associated factor) and related to the yeast SWISNF complex, is formed of multiple subunits, the number and nature of which depends on the cell type analysed74. Among these components are a catalytic subunit with ATPase activity, known as BRG, and a DNA-binding subunit, BAF57, that does not seem to recruit BAF complexes to DNA but instead bends DNA during chromatin remodelling75. The link between BAF complexes and CD4 expression was established by impairing BAF function in the T-cell lineage, through targeted disruption of the Brg gene or overexpression of a BAF57 mutant that lacks DNA-binding activity. Besides revealing the pleiotropic role of BAF complexes in many aspects of thymocyte development, including differentiation steps mediated by pre-TCR signalling76,77, these analyses showed that BAF complexes are required to repress CD4 expression by DN thymocytes78,79. The effects of BAF inactivation are similar to those caused by Runx1 gene disruption, pointing to a functional link between RUNX1 and BAF complexes52,78. Indeed, CHROMATIN IMMUNOPRECIPITATION ASSAYS showed that BAF components bind to CD4 silencer DNA in vivo, although it is not known whether such binding involves RUNX proteins78. Is there then a role for BAF complexes in the establishment of CD4 silencing, in addition to their role in repression? As Brg gene disruption in the T-cell lineage
arrests T-cell differentiation before the DP stage79, this question was addressed in mice with partial BAF inactivation, which impairs but does not abolish T-cell development78. Although most mature CD8 SP thymocytes and T cells in these mice properly silence CD4 transcription, a fraction fail to do so, indicating a role for BAF activity in CD4 silencing. In addition, it must be considered that the cells that overcome the block in positive selection imposed by BAF inactivation might have developed compensatory mechanisms that also promote CD4 silencing and thereby mask the function of BAF in silencing. So, BAF activity is likely to be important not only for CD4 repression in DN thymocytes but also for CD4 silencing in CD8-lineage cells. A model for the establishment of silencing during CD8+ T-cell differentiation. The recent findings described earlier suggest a multi-step scenario for CD4 silencing, which, albeit hypothetical, illustrates how lineage and stage specificity can be achieved by factors with much broader expression spectra (FIG. 4). In DP thymocytes undergoing selection, CD8-differentiation signals would transiently repress CD4 gene expression through the recruitment of RUNX3 and other as-yet-unknown factors to the silencer. Importantly, although RUNX3 is expressed at similar levels by both mature CD4+ and CD8+ T cells, Runx3 gene expression during positive selection is mainly CD8-lineage specific52,54,55,80 and precedes the cessation of CD4 expression80. So, at this developmental stage, cell specificity could result from the CD8-lineage-specific expression of RUNX3, combined with the thymocyte-specific expression of other silencer-binding factors. The assembly of this repressing complex on the silencer would antagonize enhancerdriven transcription, possibly through the recruitment of chromatin-modifying enzymes, and cause the termination of CD4 expression37,72. In addition to chromatin remodelling, such chromatin modifications could include histone deacetylation, which is thought to hinder transcription by strengthening histone association to DNA81. Indeed, histone deacetylases are involved in gene repression by Runt, the Drosophila homologue of RUNX proteins82, implicating another link between RUNX3-mediated silencing and chromatin modification. In addition to these two levels of regulation, a further level may include the control of BAF activity itself. Indeed, recent results indicate that the activity of BAF complexes can be controlled by cell-specific signalling8386, notably by phosphoinositides and inositol polyphosphates. As phosphoinositides give rise to inositol polyphosphates after activation of phospholipase C (PLC) a target of TCR signalling in thymocytes8789 these findings suggest a direct, albeit hypothetical, link between TCR signalling and chromatin organization at the CD4 locus. How such reversible changes in silencer activity result in stable silencing remains unclear. It is unlikely that the maintenance of silencing involves histone deacetylase activity, as CD4 silencing in mature CD8+ T cells is resistant to trichostatin A, an inhibitor of these enzymes50. However, there is evidence that another type
CHROMATIN-REMODELLING COMPLEXES
ATP-dependent multi-protein complexes that mediate the repositioning or reorganization of nucleosomes over a singleor multi-gene locus, resulting in either increased or reduced gene transcription.
CHROMATIN IMMUNOPRECIPITATION ASSAYS
A technique allowing the detection of in vivo interactions between a DNA-associated protein and candidate DNA target sequences. It involves the fragmentation of genomic DNA chromatin after chemical crosslinking of proteins to DNA, the immunoprecipitation of proteinDNA complexes using an antibody against the protein of interest, followed by PCR amplification of candidate DNA sequences after reversal of the crosslink.
534
REVIEWS
of histone modification, methylation, could be important for silencing. Results in heterologous systems indicate that variegated expression, as observed for CD4 expression by CD8+ T cells with silencer point mutations, is the phenotypic manifestation of locus relocation into inactive chromatin areas known as heterochromatin, in which gene silencing is maintained in an EPIGENETIC manner, inheritable through cell division90. Heterochromatin location seems to involve methylation of histone H3 and the subsequent recruitment of molecules such as heterochromatin protein 1 (HP1)91. That such a mechanism could operate for CD4 silencing is suggested by the partial rescue of silencer mutations by forced expression of HP1 molecules49.
RUNX proteins and CD8-lineage differentiation?
EPIGENETIC
Any heritable influence (in the progeny of cells or of individuals) on chromosome or gene function that is not accompanied by a change in DNA sequence. Examples of epigenetic events include mammalian X-chromosome inactivation, imprinting, centromere inactivation and position-effect variegation.
There is little evidence to support this possibility. Although Runx3 gene disruption decreases the size of mature CD8+ T-cell populations, this probably reflects a role in the terminal differentiation or survival of CD8+ T cells, as Runx3 disruption has either no52 or minimal54 consequences on the size of CD8-lineage thymocyte populations. Furthermore, Runx3/Runx1+/ mice generate mature CD8+ T cells that uniformly express CD4, showing distinct RUNX requirements for CD8-lineage differentiation and for CD4 silencing54. So, with the caveat of potential gene redundancy, loss-of-function analyses indicate that RUNX activity is not involved in CD8-lineage choice itself, although it might promote the survival or differentiation of CD8-lineage cells92.
A role for GATA3 in CD4-lineage differentiation?
Nuclear proteins that control CD4 or CD8 gene expression in thymocytes are not necessarily involved in lineage choice itself, as they might instead function as downstream effectors after lineage decisions have been made. So, the question arises as to whether RUNX activity, beyond its role in CD4 silencing, participates in CD8-lineage differentiation or in CD8-lineage choice.
CD4 silencer DP thymocytes Silencer activity
Because there is no known cis-regulatory DNA element that promotes CD4-lineage-specific expression, the strategy that revealed the role of RUNX proteins in CD8-lineage differentiation could not be applied to identify CD4-lineage differentiating factors. Instead, progress has come from converging approaches that address the role of the transcription factor GATA3 during positive selection. GATA3 is a member of a family of
CD4 promoter RNA polymerase Nucleosome CD4 Transcription
a
RUNX X RUNX Y
Acetyl
b
CD8-lineage thymocytes
Histone deacetylase H3 methylation RUNX X RUNX Y
+/
Chromatin-remodelling complex
HP1
c
CD8+ T cells RUNX X RUNX All sites inaccessible for binding Y
CD4+ T cells
d
RUNX X RUNX Y
RNA polymerase
CD4
Figure 4 | Model for the establishment of silencing in CD8-lineage thymocytes. A hypothetical model for CD4 silencer function during CD8-lineage differentiation. CD4 gene regulatory elements are schematically shown as horizontal bars (left: silencer, right: promoter). Known DNA motifs that are required for silencing are depicted as rectangles (red: RUNX-binding sites; blue or green: binding sites for non-RUNX factors). a | In double-positive (DP) thymocytes that have not upregulated RUNX3 expression, non-RUNX factors bind the silencer but are insufficient to promote silencing. As a result, the basal transcription machinery (including RNA polymerase and associated general transcription factors) is active on the CD4 promoter. Active transcription at the CD4 promoter is associated with appropriate positioning of nucleosomes and with the presence of acetylated histones. b | In CD8-lineage thymocytes, RUNX3 expression, induced by T-cell receptor (TCR) signals or as a result of CD8 differentiation80, causes the formation of a repressing complex that reversibly antagonizes CD4 gene expression, by directly interacting with the transcription machinery or by recruiting chromatin-modifying enzymes, such as histone deacetylases or chromatin-remodelling complexes. Both chromatin modifiers reduce transcription by deacetylating histones and repositioning nucleosomes. In addition, the accumulation of deacetylated histones and the absence of transcription promote histone H3 methylation at the CD4 locus and the subsequent recruitment of heterochromatin protein 1 (HP1), establishing silencing by transferring the CD4 locus to transcriptionally inactive heterochromatin. c | These modifications are complete and irreversible in mature CD8 single-positive thymocytes and CD8+ T cells, in which silencing is maintained and inherited in a silencer-independent manner50. d | In mature CD4+ T cells, CD4 transcription is active despite the presence of RUNX3, possibly because RUNX proteins fail to affect CD4 transcription in the absence of non-RUNX co-silencing factors. RUNX, Runt-related transcription factor.
VOLUME 4 | JULY 2004 | 5 3 5
REVIEWS
zinc-finger transcriptional regulators that are crucial for the differentiation of multiple tissues and organs93. Genedisruption studies have indicated that GATA3 is indispensable for early T-cell-lineage differentiation, and have confirmed that it is also required for the differentiation of mature CD4+ T cells into TH2-type effector cells94,95. A role for GATA3 during CD4+ T-cell selection was first implied from its expression pattern in the thymus, with high expression in CD4 SP thymocytes and in the transitional CD4+CD8low thymocyte population enriched in thymocytes undergoing selection, and low expression in CD8 SP and DP thymocytes96,97. Furthermore, upregulation of GATA3 expression by transitional cells is MHC class II dependent, indicating that it is specific for, and possibly involved in, CD4-lineage differentiation97,98. These suspicions have now been confirmed using two independent strategies. A first approach used reaggregate thymic organ cultures99 in which thymocytes were retrovirally transduced to express GATA3, or to impair GATA3 expression by RNA INTERFERENCE. In this system, increased GATA3 expression favoured CD4+ over CD8+ T-cell differentiation, whereas reduced GATA3 expression had the opposite effect, therefore indicating a role for GATA3 as a CD4-lineage differentiation factor97. However, there were a few significant concerns with this approach. GATA3 expression markedly impaired thymocyte survival. Furthermore, the retroviral expression system resulted in GATA3 expression levels that were considerably higher than those normally observed during positive selection. This latter aspect was of particular concern, as previous analyses had found that a Gata3 transgene, the expression of which was closer to physiological levels, had no effect on CD4/CD8-lineage differentiation100. It was therefore important that the role of GATA3 in CD4-lineage differentiation was also investigated in vivo. This was done recently using a conditional gene knockout strategy to target Gata3 disruption to DP thymocytes, thereby by-passing GATA3 requirements for earlier T-cell-differentiation steps101. In agreement with the retroviral expression study, Gata3 gene disruption essentially abolished CD4+ T-cell differentiation but did not affect CD8-lineage differentiation. So, both gain- and loss-of-function approaches confirm a crucial role for GATA3 during CD4+ T-cell development. Furthermore, the fact that GATA3 overexpression both increases CD4+ T-cell differentiation and reduces CD8+ T-cell differentiation97 indicates that GATA3 might promote CD4-lineage choice per se. That Gata3 gene disruption fails to promote the generation of MHC class-II-restricted CD8+ T cells101 does not exclude this possibility: it is conceivable that Gata3 disruption causes MHC class-II-restricted thymocytes to choose the CD8 lineage, but that such MHC classII-restricted CD8-lineage cells die as they downregulate CD4 and fail to undergo MHC class-II-induced TCR and CD4 co-engagement required for their survival. A similar argument applies to the observations that enforced GATA3 expression fails to redirect MHC class-I-restricted thymocytes into CD4+ T cells97.
Matching lineage choice to MHC specificity
As the upwards approach to lineage differentiation has resulted in the identification of RUNX3 as crucial to the proper development of CD8+ T cells and as recent analyses have revealed the role of GATA3 in CD4+ T-cell development, it is legitimate to examine if and how these findings affect our understanding of the lineage choice process. Duration of signalling and lineage choice. A large body of recently reviewed evidence33,34,102 argues against the original instructive or selective models and instead supports the concept that CD4/CD8-lineage choice is determined by quantitative attributes of MHC class-I-induced TCR signals. The most commonly held view is that the strength (or intensity) of TCR signals directs lineage choice, with strong signals promoting CD4-lineage choice and weak signals promoting CD8-lineage choice. This concept stems from experiments directly or indirectly manipulating the activity of LCK, a tyrosine kinase that associates with the cytosolic tails of the CD4 and CD8 co-receptors103,104, and that is crucial for TCR signalling105. Because LCK binds with greater affinity to CD4 than to CD8 (REFS 5,106), the strength-of-signal model proposes that in DP thymocytes, TCR and CD4 co-ligation generates signals of greater intensity than TCR and CD8 co-ligation, and that this difference in signalling intensity is translated into a lineage decision33,102. Supporting this view, transgenic expression of CD8 molecules with CD4 cytoplasmic tails allows MHC class-I-restricted thymocytes to differentiate into the CD4 lineage107,108, whereas CD4 disruption causes MHC class-II-restricted thymocytes to differentiate into CD8+ T cells109,110. Similar effects were achieved by using genetic approaches to augment or reduce intracellular LCK activity, respectively111, and additional, albeit indirect, support for the strength-of-signal hypothesis comes from other analyses112,113. However, several results conflict with this concept. Only in a few circumstances has it been observed that altered TCR signalling causes lineage redirection that is, CD8-lineage choice by MHC class-II-restricted thymocytes or CD4-lineage choice by MHC class-I-restricted thymocytes108,111,114. Instead, mutations that alter TCR signalling in DP thymocytes affect the number of thymocytes undergoing selection but generally not their lineage direction115122. Moreover, transgenic CD8 molecules with CD4 cytoplasmic tails promote the differentiation of MHC class-I-restricted T cells into both the CD4+ and CD8+ T cells122. This indicates that, rather than determining lineage choice, the co-receptor tails control the number of thymocytes being selected, thereby ensuring the numerical predominance of MHC class-II-restricted T cells (signalled through CD4) over MHC class-I-restricted T cells (signalled through CD8). An alternative view proposes that the duration of intrathymic TCR signals, rather than their intensity, determines CD4/CD8-lineage direction31,123. The validity of this hypothesis was recently tested directly in vivo, by restricting to DP thymocytes the expression of ZAP70
RNA INTERFERENCE
A technique to inhibit (or knock-down) the expression of a specific target gene using short oligonucleotides that are complementary to the sequence of the gene mRNA.
536
REVIEWS
Duration of signal: mechanistic aspects. If the duration of TCR signalling directs lineage differentiation, why would MHC class-II-induced signals be of longer duration than MHC class-I-induced signals? One answer to this question, proposed by the kinetic signalling model of lineage choice31,130, is that TCR signalling itself contributes to create this asymmetry by downregulating CD8 but not CD4 expression, thereby shortening CD8dependent (MHC class-I-induced) signals. The kinetic signalling model relies on three postulates. First, that TCR signalling in DP thymocytes, regardless of its MHC specificity, downregulates CD8 gene expression31,32. Second, that CD4-lineage commitment (understood as the loss of CD8-developmental potential) does not occur in DP thymocytes, but does occur after TCR signalling has transcriptionally downregulated CD8, as in CD4+CD8low thymocytes. Finally, that in CD4+CD8low cells the persistence of TCR signalling causes CD4-lineage choice, whereas the cessation or downregulation of TCR signalling promotes CD8-lineage choice (thereby causing the cessation of CD4 expression and the re-expression of CD8, a phenomenon known as co-receptor reversal31). The kinetic signalling model proposes that MHC classI-induced signals would be of shorter duration than MHC class-II-induced signals, because only MHC class I signals are sensitive to the downregulation of CD8 expression. Of note, the model does not exclude the possibility that factors other than CD8 downregulation contribute to make MHC class-II-induced signals of longer duration than MHC class-I-induced signals. The fact that such mechanisms exist is indicated by the observation that MHC class-I-restricted thymocytes lacking CD8 molecules, when driven into positive selection by high-avidity ligands, choose the CD8 lineage despite the lack of CD8 expression128,131; and also that CD8 transgenes expressed throughout positive selection, even in CD4-differentiating cells, are relatively inefficient at promoting the CD4-lineage differentiation of MHC class-I-restricted thymocytes, opposing the predictions of the kinetic signalling model26,108,122. So, the kinetic signalling model proposes a novel, non-instructive, non-selective answer to the question of how signal attributes match lineage choice to MHC specificity. Additional analyses will be required to discriminate between the strength-of-signal and kinetic signalling/duration-of-signal models. Of note, as both models propose more stringent TCR-signalling requirements for CD4- than for CD8-lineage choice, some experiments that are generally interpreted in favour of the strength-of-signal model are equally supportive of the kinetic signalling model132,133. Duration of signal and nuclear effectors of lineage differentiation. Although it would be premature to establish firm connections between the kinetic signalling/duration-of-signal concept and nuclear effectors of lineage differentiation, there are a few interesting correlates (FIG. 5). The first one relates to GATA3. Even though its role in lineage choice per se is still hypothetical, GATA3 is selectively required for CD4+ T-cell development and therefore can be considered as a
Protein GATA3 RAG1 and RAG2
Expression levels + + + +++ TCR signal TCR signal DP CD4+ CD8+ DP CD4+ CD8+ TCR signal CD4+ CD8low/ SP CD4+ CD8 GATA3 deletion or Hd mutation +++ CD4 lineage
Uncommitted thymocyte
No TCR signal results in death by neglect CD4CD8+ TCR signal
Attenuated TCR signals cause CD8-lineage choice CD8 lineage
SP CD4 CD8+
Figure 5 | Matching lineage differentiation to MHC specificity during positive selection. T-cell receptor (TCR)-signal-induced and TCR-independent (dotted arrows) developmental stages are shown, based on a hypothetical model of thymocyte development. In the absence of TCR signalling, double-positive (DP) thymocytes die by neglect prior to lineage differentiation. Initial TCR signalling in DP cells causes the termination of recombinase-activating gene 1 (RAG1) and RAG2 gene expression151 (thereby fixing TCR- specificity), the upregulation of CD5 and CD69 expression152, and the downregulation of CD8 gene expression31,32. The kinetic signalling hypothesis proposes that such initial TCR signalling makes thymocytes competent for lineage choice. Persistent TCR signalling, despite downregulation of CD8 expression, causes CD4-lineage differentiation, through the upregulation of GATA3 expression. Downregulation of TCR signalling before or during CD8 downregulation causes CD8-lineage differentiation31,80. A terminal TCR signalling step proofreads CD8-lineage differentiation80,127 and contributes to the terminal differentiation of CD8-lineage thymocytes, notably by allowing interleukin-7 signalling153. This model predicts that MHC class-II-restricted CD8 single-positive (SP) populations develop when TCR and CD4 signalling, and thereby thymocyte survival, is maintained by enforcing CD4 expression after CD8-lineage differentiation. Indeed, such populations are detected when CD4 expression is maintained in CD8-lineage cells using either CD4 transgenes24,27 or germline deletion of the CD4 silencer154. Hd, helper cell deficient.
(-chain-associated protein of 70 kDa), a tyrosine kinase required for TCR-signal transduction124126, thereby genetically limiting the duration of TCR signalling during positive selection80. Under such conditions, peptideMHC-induced TCR signals initiate positive selection, but the premature cessation of TCR signalling arrests thymocyte development before the emergence of CD4 and CD8 SP populations, at the transitional CD4+CD8low or CD4lowCD8+ stages127. CD4- and CD8-lineage differentiation was assessed in these arrested DP and transitional populations by measuring the expression of lineage-specific genes, namely those encoding GATA3 for the CD4 lineage97 and perforin or cathepsin W for the CD8 lineage128,129. Using these approaches, it was shown that TCR signals confined to DP thymocytes allow CD8-lineage differentiation (although not the development of CD8 SP populations), but prevent CD4-lineage differentiation and redirect MHC class-II-restricted thymocytes to the CD8 lineage, indicating that the duration of signal directly affects lineage choice.
VOLUME 4 | JULY 2004 | 5 3 7
REVIEWS
CD4-lineage marker. Consequently, the kinetic signalling model predicts that persistent TCR signalling should promote GATA3 expression. In vivo analyses fulfill this prediction by showing that upregulation of GATA3 expression occurs in transitional CD4+CD8low cells 97 and requires these cells to be competent for signalling80. The mechanistic link between TCR signalling and GATA3 expression remains to be clarified. Studies in mature T cells have shown that many factors regulate Gata3 gene expression, including signal transducer and activator of transcription 6 (STAT6)-mediated cytokine signalling, co-stimulatory molecules and positive autoregulation by GATA3 itself 94. However, none of these mechanisms seem to work in the thymus. Although it is conceivable that transient activation of STAT6 contributes to the upregulation of GATA3 expression in the thymus, STAT6-deficient mice have normal populations of naive CD4+ T cells, indicating that, unlike GATA3, STAT6 is not required for CD4+ T-cell differentiation134,135. Furthermore, there is evidence that the autoregulatory loop that enhances GATA3 expression during TH2-cell differentiation136 does not operate in the thymus97, indicating that GATA3 upregulation by CD4differentiating thymocytes is not self-sustained. Finally, TCR signalling in mature T cells per se is not a major stimulus of GATA3 expression. So, although some of the pathways that activate GATA3 expression by mature T cells are presumably targeted by intrathymic TCR signalling137, it remains unclear how TCR signals in thymocytes promote GATA3 expression. Unlike for GATA3, the kinetic signalling hypothesis predicts that the upregulation of RUNX3 expression characteristic of CD8-lineage thymocytes, should not require persistent TCR signals. In vivo analyses in mice with ZAP70 expression confined to DN thymocytes support this view: although TCR signalling is required for upregulated RUNX3 expression in thymocytes in vivo, RUNX3 expression in CD8-lineage cells persists even if TCR signalling is interrupted80. A key question is whether TCR-induced RUNX3 upregulation precedes or results from CD8-lineage choice. Preliminary experiments argue against the first possibility, as TCR stimulation per se does not seem to promote RUNX upregulation in thymocytes (X. Liu and R.B., unpublished observations).
Conclusion and perspectives
As highlighted throughout this review, there are still many uncertainties about lineage-choice mechanisms. Furthermore, questions continue to emerge, as it has been suggested that CD4- or CD8-lineage choice by individual thymocytes might be affected by the developmental fate of their thymocyte neighbours138. Nonetheless, the past two years have seen decisive progress in the identification of nuclear effectors of lineage differentiation and have clarified how the duration of TCR signals affects lineage choice. The goal for the years to come is to bridge these two fields of investigation, to understand the basis for the link between lineage and MHC specificity. Progress in elucidating the mechanisms of CD4 silencing, including the potential role of TCR signals in modulating the activity of chromatin-remodelling complexes and the search for additional silencerbinding factors, will hopefully improve our understanding of CD8 differentiation. Furthermore, recent findings indicate that overexpression of the highmobility-group DNA-binding protein TOX promotes CD8-lineage differentiation, and does so in the absence of MHC class-I-induced TCR ligation, indicate that TOX activity might be important for CD8-lineage choice139. Conversely, deciphering the mechanisms of TCR-induced upregulation of GATA3 expression, including analyses of Gata3 gene regulatory elements, is expected to shed light on the mechanisms of CD4lineage differentiation. In this regard, observations in mature T cells raise the intriguing possibility that RUNX proteins inhibit GATA3 expression 140, even though the opposite does not seem to be true 97. Finally, it is expected that the identification of the mouse gene known as Hd (helper-cell deficient), a mutation in which results in impaired CD4-lineage differentiation and redirects MHC class-II-restricted thymocytes into the CD8 lineage, will provide a new avenue of research in the field141.
Note added in proof
It was recently shown that disruption of the CSL gene in DP thymocytes does not affect the generation of mature CD4+ or CD8+ T cells, further supporting the notion that Notch signalling is not involved in CD4/CD8-lineage differentiation155.
1. 2.
3.
4. 5.
6.
Weiss, A. & Littman, D. R. Signal transduction by lymphocyte antigen receptors. Cell 76, 263274 (1994). Kearse, K. P., Roberts, J. L. & Singer, A. TCRCD3 association is the initial step in dimer formation in murine T cells and is limiting in immature CD4+CD8+ thymocytes. Immunity 2, 391399 (1995). Call, M. E., Pyrdol, J., Wiedmann, M. & Wucherpfennig, K. W. The organizing principle in the formation of the T cell receptorCD3 complex. Cell 111, 967979 (2002). Janeway, C. A. J. & Bottomly, K. Signals and signs for lymphocyte responses. Cell 76, 275285 (1994). Veillette, A., Ziga-Pflcker, J. C., Bolen, J. B. & Kruisbeek, A. M. Engagement of CD4 and CD8 expressed on immature thymocytes induces activation of intracellular tyrosine phosphorylation pathways. J. Exp. Med. 170, 16711680 (1989). Shaw, A. S. et al. The lck tyrosine protein kinase interacts with the cytoplasmic tail of the CD4 glycoprotein through
its unique amino-terminal domain. Cell 59, 627636 (1989). 7. Bosselut, R. et al. Association of the adaptor molecule LAT with CD4 and CD8 coreceptors identifies a new coreceptor function in T cell receptor signal transduction. J. Exp. Med. 190, 15171526 (1999). 8. Brdikov, N. et al. LIME: a new membrane raftassociated adaptor protein involved in CD4 and CD8 coreceptor signaling. J. Exp. Med. 198, 14531462 (2003). 9. Hur, E. M. et al. LIME, a novel transmembrane adaptor protein, associates with p56lck and mediates T cell activation. J. Exp. Med. 198, 14631473 (2003). 10. Egerton, M., Scollay, R. & Shortman, K. Kinetics of mature T-cell development in the thymus. Proc. Natl Acad. Sci. USA 87, 25792582 (1990). 11. Huesmann, M., Scott, B., Kisielow, P. & von Boehmer, H. Kinetics and efficacy of positive selection in the thymus of
12.
13.
14.
15.
16. 17.
normal and T cell receptor transgenic mice. Cell 66, 533540 (1991). Kisielow, P., Teh, H. S., Bluthmann, H. & von, B. H. Positive selection of antigen-specific T cells in thymus by restricting MHC molecules. Nature 335, 730733 (1988). Teh, H. S. et al. Thymic major histocompatibility complex antigens and the T-cell receptor determine the CD4/CD8 phenotype of T cells. Nature 335, 229233 (1988). Sha, W. C. et al. Positive and negative selection of an antigen receptor on T cells in transgenic mice. Nature 336, 7376 (1988). Kaye, J. et al. Selective development of CD4+ T cells in transgenic mice expressing a class II MHC-restricted antigen receptor. Nature 341, 746749 (1989). Alam, S. M. et al. T-cell-receptor affinity and thymocyte positive selection. Nature 381, 616620 (1996). Janeway, C. A. J. T-cell development. Accessories or coreceptors? Nature 335, 208210 (1988).
538
REVIEWS
18. von Boehmer, H. Positive selection of lymphocytes. Cell 76, 219228 (1994). 19. von Boehmer, H. CD4/CD8 lineage commitment: back to instruction? J. Exp. Med. 183, 713715 (1996). 20. Borgulya, P., Kishi, H., Muller, U., Kirberg, J. & von, B. H. Development of the CD4 and CD8 lineage of T cells: instruction versus selection. EMBO J. 10, 913918 (1991). 21. Robey, E. A. et al. Thymic selection in CD8 transgenic mice supports an instructive model for commitment to a CD4 or CD8 lineage. Cell 64, 99107 (1991). 22. Itano, A., Kioussis, D. & Robey, E. Stochastic component to development of class I major histocompatibility complexspecific T cells. Proc. Natl Acad. Sci. USA 91, 220224 (1994). 23. Chan, S. H., Cosgrove, D., Waltzinger, C., Benoist, C. & Mathis, D. Another view of the selective model of thymocyte selection. Cell 73, 225236 (1993). 24. Davis, C. B., Killeen, N., Crooks, M. E., Raulet, D. & Littman, D. R. Evidence for a stochastic mechanism in the differentiation of mature subsets of T lymphocytes. Cell 73, 237247 (1993). 25. van Meerwijk, J. P. & Germain, R. N. Development of mature CD8+ thymocytes: selection rather than instruction? Science 261, 911915 (1993). 26. Chan, S. H., Waltzinger, C., Baron, A., Benoist, C. & Mathis, D. Role of coreceptors in positive selection and lineage commitment. EMBO J. 13, 44824489 (1994). 27. Baron, A., Hafen, K. & von Boehmer, H. A human CD4 transgene rescues CD4CD8+ cells in 2-microglobulindeficient mice. Eur. J. Immunol. 24, 19331936 (1994). 28. Lundberg, K., Heath, W., Kontgen, F., Carbone, F. R. & Shortman, K. Intermediate steps in positive selection: differentiation of CD4+CD8intTCRint thymocytes into CD4CD8+TCRhi thymocytes. J. Exp. Med. 181, 16431651 (1995). 29. Suzuki, H., Punt, J. A., Granger, L. G. & Singer, A. Asymmetric signaling requirements for thymocyte commitment to the CD4+ versus CD8+ T cell lineages: a new perspective on thymic commitment and selection. Immunity 2, 413425 (1995). 30. Lucas, B. & Germain, R. N. Unexpectedly complex regulation of CD4/CD8 coreceptor expression supports a revised model for CD4+CD8+ thymocyte differentiation. Immunity 5, 461477 (1996). 31. Brugnera, E. et al. Coreceptor reversal in the thymus: signaled CD4+CD8+ thymocytes initially terminate CD8 transcription even when differentiating into CD8+ T cells. Immunity 13, 5971 (2000). An in vitro study exploring the role of TCR-signal duration in lineage differentiation. The authors propose the initial formulation of the kinetic signalling model. 32. Bosselut, R., Guinter, T. I., Sharrow, S. O. & Singer, A. Unraveling a revealing paradox: why major histocompatibility complex I-signaled thymocytes paradoxically appear as CD4+CD8lo transitional cells during positive selection of CD8+ T cells. J. Exp. Med. 197, 17091719 (2003). 33. Germain, R. N. T-cell development and the CD4CD8 lineage decision. Nature Rev. Immunol. 2, 309322 (2002). 34. Bosselut, R. & Singer, A. CD4/CD8 coreceptors in thymocyte development, selection, and lineage commitment: analysis of the CD4/CD8 lineage decision. Adv. Immunol. 83, 91131 (2004). 35. Hedrick, S. M. T cell development: bottoms-up. Immunity 16, 619622 (2002). 36. Parnes, J. R. Molecular biology and function of CD4 and CD8. Adv. Immunol. 44, 265311 (1989). 37. Kioussis, D. & Ellmeier, W. Chromatin and CD4, CD8A and CD8B gene expression during thymic differentiation. Nature Rev. Immunol. 2, 909919 (2002). 38. Hostert, A. et al. A region in the CD8 gene locus that directs expression to the mature CD8 T cell subset in transgenic mice. Immunity 7, 525536 (1997). 39. Ellmeier, W., Sunshine, M. J., Losos, K., Hatam, F. & Littman, D. R. An enhancer that directs lineage-specific expression of CD8 in positively selected thymocytes and mature T cells. Immunity 7, 537547 (1997). 40. Hostert, A. et al. Hierarchical interactions of control elements determine CD8A gene expression in subsets of thymocytes and peripheral T cells. Immunity 9, 497508 (1998). 41. Ellmeier, W., Sunshine, M. J., Losos, K. & Littman, D. R. Multiple developmental stage-specific enhancers regulate CD8 expression in developing thymocytes and in thymusindependent T cells. Immunity 9, 485496 (1998). References 40 and 41 present a detailed dissection of the CD8 locus cis-regulatory elements that are active in DP thymocytes and mature T cells. 42. Ellmeier, W., Sunshine, M. J., Maschek, R. & Littman, D. R. Combined deletion of CD8 locus cis-regulatory elements affects initiation but not maintenance of CD8 expression. Immunity 16, 623634 (2002). 43. Garefalaki, A. et al. Variegated expression of CD8 resulting from in situ deletion of regulatory sequences. Immunity 16, 635647 (2002). Using homologous recombination techniques, references 42 and 43 demonstrate the role of two CD8 enhancers for CD8 gene expression, and indicate that these enhancers act, at least in part, to prevent chromatin modifications that would silence the CD8 locus. 44. Harker, N. et al. The CD8 gene locus is regulated by the Ikaros family of proteins. Mol. Cell 10, 14031415 (2002). 45. Ellmeier, W., Sawada, S. & Littman, D. R. The regulation of CD4 and CD8 coreceptor gene expression during T cell development. Annu. Rev. Immunol. 17, 523554 (1999). 46. Sawada, S. & Littman, D. R. Identification and characterization of a T-cell-specific enhancer adjacent to the murine CD4 gene. Mol. Cell. Biol. 11, 55065515 (1991). 47. Siu, G., Wurster, A. L., Duncan, D. D., Soliman, T. M. & Hedrick, S. M. A transcriptional silencer controls the developmental expression of the CD4 gene. EMBO J. 13, 35703579 (1994). References 46 and 47 report the initial identification and characterization of the CD4 silencer. 48. Sawada, S., Scarborough, J. D., Killeen, N. & Littman, D. R. A lineage-specific transcriptional silencer regulates CD4 gene expression during T lymphocyte development. Cell 77, 917929 (1994). 49. Taniuchi, I., Sunshine, M. J., Festenstein, R. & Littman, D. R. Evidence for distinct CD4 silencer functions at different stages of thymocyte differentiation. Mol. Cell 10, 10831096 (2002). This study uses knocked-in point mutations to characterize functional regions within the CD4 silencer. It also dissects the repressor from the silencing function of the minimal silencer element. 50. Zou, Y. R. et al. Epigenetic silencing of CD4 in T cells committed to the cytotoxic lineage. Nature Genet. 29, 332336 (2001). This paper shows that the CD4 silencer is required to establish silencing in CD8-lineage thymocytes but not to maintain silencing in mature CD8+ T cells. 51. Duncan, D. D., Adlam, M. & Siu, G. Asymmetric redundancy in CD4 silencer function. Immunity 4, 301311 (1996). 52. Taniuchi, I. et al. Differential requirements for Runx proteins in CD4 repression and epigenetic silencing during T lymphocyte development. Cell 111, 621633 (2002). This study identifies the crucial role of RUNX1 and RUNX3 proteins in preventing inappropriate expression of CD4 by non-CD4-lineage T cells. 53. Kim, W. W. & Siu, G. Subclass-specific nuclear localization of a novel CD4 silencer binding factor. J. Exp. Med. 190, 281291 (1999). 54. Woolf, E. et al. Runx3 and Runx1 are required for CD8 T cell development during thymopoiesis. Proc. Natl Acad. Sci. USA 100, 77317736 (2003). 55. Ehlers, M. et al. Morpholino antisense oligonucleotidemediated gene knockdown during thymocyte development reveals role for Runx3 transcription factor in CD4 silencing during development of CD4CD8+ thymocytes. J. Immunol. 171, 35943604 (2003). 56. Okuda, T., van, D. J., Hiebert, S. W., Grosveld, G. & Downing, J. R. AML1, the target of multiple chromosomal translocations in human leukemia, is essential for normal fetal liver hematopoiesis. Cell 84, 321330 (1996). 57. Ducy, P., Zhang, R., Geoffroy, V., Ridall, A. L. & Karsenty, G. Osf2/Cbfa1: a transcriptional activator of osteoblast differentiation. Cell 89, 747754 (1997). 58. Geoffroy, V., Ducy, P. & Karsenty, G. A PEBP2 /AML-1related factor increases osteocalcin promoter activity through its binding to an osteoblast-specific cis-acting element. J. Biol. Chem. 270, 3097330979 (1995). 59. Kim, H. K. & Siu, G. The notch pathway intermediate HES-1 silences CD4 gene expression. Mol. Cell. Biol. 18, 71667175 (1998). 60. Radtke, F., Wilson, A., Mancini, S. J. & MacDonald, H. R. Notch regulation of lymphocyte development and function. Nature Immunol. 5, 247253 (2004). 61. Allman, D., Punt, J. A., Izon, D. J., Aster, J. C. & Pear, W. S. An invitation to T and more: notch signaling in lymphopoiesis. Cell 109, Suppl. S1S11 (2002). 62. Robey, E. et al. An activated form of Notch influences the choice between CD4 and CD8 T cell lineages. Cell 87, 483492 (1996). 63. Deftos, M. L., Huang, E., Ojala, E. W., Forbush, K. A. & Bevan, M. J. Notch1 signaling promotes the maturation of CD4 and CD8 SP thymocytes. Immunity 13, 7384 (2000). 64. Fowlkes, B. J. & Robey, E. A. A reassessment of the effect of activated notch1 on CD4 and CD8 T cell development. J. Immunol. 169, 18171821 (2002). 65. Izon, D. J. et al. Notch1 regulates maturation of CD4+ and CD8+ thymocytes by modulating TCR signal strength. Immunity 14, 253264 (2001). 66. Wolfer, A. et al. Inactivation of Notch 1 in immature thymocytes does not perturb CD4 or CD8 T cell development. Nature Immunol. 2, 235241 (2001). 67. Saito, T. et al. Notch2 is preferentially expressed in mature B cells and indispensable for marginal zone B lineage development. Immunity 18, 675685 (2003). 68. Kaneta, M. et al. A role for pref-1 and HES-1 in thymocyte development. J. Immunol. 164, 256264 (2000). 69. Jarriault, S. et al. Signalling downstream of activated mammalian Notch. Nature 377, 355358 (1995). 70. Han, H. et al. Inducible gene knockout of transcription factor recombination signal binding protein-J reveals its essential role in T versus B lineage decision. Int. Immunol. 14, 637645 (2002). 71. Radtke, F. et al. Deficient T cell fate specification in mice with an induced inactivation of Notch1. Immunity 10, 547558 (1999). 72. Smale, S. T. The establishment and maintenance of lymphocyte identity through gene silencing. Nature Immunol. 4, 607615 (2003). 73. Vaquero, A., Loyola, A. & Reinberg, D. The constantly changing face of chromatin. Sci. Aging Knowledge Environ. 2003, RE4 (2003). 74. Wang, W. et al. Diversity and specialization of mammalian SWI/SNF complexes. Genes Dev. 10, 21172130 (1996). 75. Wang, W. et al. Architectural DNA binding by a high-mobilitygroup/kinesin-like subunit in mammalian SWI/SNF-related complexes. Proc. Natl Acad. Sci. USA 95, 492498 (1998). 76. Mombaerts, P. et al. Mutations in T-cell antigen receptor genes and block thymocyte development at different stages. Nature 360, 225231 (1992). 77. Fehling, H. J., Krotkova, A., Saint-Ruf, C. & von Boehmer, H. Crucial role of the pre-T-cell receptor gene in development of but not T cells. Nature 375, 795798 (1995). 78. Chi, T. H. et al. Reciprocal regulation of CD4/CD8 expression by SWI/SNF-like BAF complexes. Nature 418, 195199 (2002). This paper identifies the role of BAF chromatinremodelling complexes in the control of CD4 and CD8 gene expression. 79. Chi, T. H. et al. Sequential roles of Brg, the ATPase subunit of BAF chromatin remodeling complexes, in thymocyte development. Immunity 19, 169182 (2003). 80. Liu, X. & Bosselut, R. Duration of TCR signaling controls CD4-CD8 lineage differentiation in vivo. Nature Immunol. 5, 280288 (2004). Using a transgenic approach to dissect positive selection, this paper shows that the duration of intrathymic TCR signalling controls CD4/CD8-lineage choice. 81. Eberharter, A. & Becker, P. B. Histone acetylation: a switch between repressive and permissive chromatin. Second in review series on chromatin dynamics. EMBO Rep. 3, 224229 (2002). 82. Wheeler, J. C. et al. Distinct in vivo requirements for establishment versus maintenance of transcriptional repression. Nature Genet. 32, 206210 (2002). 83. Zhao, K. et al. Rapid and phosphoinositol-dependent binding of the SWI/SNF-like BAF complex to chromatin after T lymphocyte receptor signaling. Cell 95, 625636 (1998). 84. Shen, X., Xiao, H., Ranallo, R., Wu, W. H. & Wu, C. Modulation of ATP-dependent chromatin-remodeling complexes by inositol polyphosphates. Science 299, 112114 (2003). 85. Steger, D. J., Haswell, E. S., Miller, A. L., Wente, S. R. & OShea, E. K. Regulation of chromatin remodeling by inositol polyphosphates. Science 299, 114116 (2003). 86. Rando, O. J., Chi, T. H. & Crabtree, G. R. Second messenger control of chromatin remodeling. Nature Struct. Biol. 10, 8183 (2003). 87. June, C. H. et al. Inhibition of tyrosine phosphorylation prevents T-cell receptor-mediated signal transduction. Proc. Natl Acad. Sci. USA 87, 77227726 (1990). 88. Secrist, J. P., Karnitz, L. & Abraham, R. T. T-cell antigen receptor ligation induces tyrosine phosphorylation of phospholipase C-1. J. Biol. Chem. 266, 1213512139 (1991). 89. Wiest, D. L. et al. A spontaneously arising mutation in the DLAARN motif of murine ZAP-70 abrogates kinase activity and arrests thymocyte development. Immunity 6, 663671 (1997). 90. Dillon, N. & Festenstein, R. Unravelling heterochromatin: competition between positive and negative factors regulates accessibility. Trends Genet. 18, 252258 (2002).
VOLUME 4 | JULY 2004 | 5 3 9
REVIEWS
91. Festenstein, R. et al. Heterochromatin protein 1 modifies mammalian PEV in a dose- and chromosomal-contextdependent manner. Nature Genet. 23, 457461 (1999). 92. Hayashi, K. et al. Overexpression of AML1 transcription factor drives thymocytes into the CD8 single-positive lineage. J. Immunol. 167, 49574965 (2001). 93. Weiss, M. J. & Orkin, S. H. GATA transcription factors: key regulators of hematopoiesis. Exp. Hematol. 23, 99107 (1995). 94. Murphy, K. M. & Reiner, S. L. The lineage decisions of helper T cells. Nature Rev. Immunol. 2, 933944 (2002). 95. Pai, S. Y., Truitt, M. L. & Ho, I. C. GATA-3 deficiency abrogates the development and maintenance of T helper type 2 cells. Proc. Natl Acad. Sci. USA 101, 19931998 (2004). 96. Hendriks, R. W. et al. Expression of the transcription factor GATA-3 is required for the development of the earliest T cell progenitors and correlates with stages of cellular proliferation in the thymus. Eur. J. Immunol. 29, 19121918 (1999). 97. Hernandez-Hoyos, G., Anderson, M. K., Wang, C., Rothenberg, E. V. & Alberola-Ila, J. GATA-3 expression is controlled by TCR signals and regulates CD4/CD8 differentiation. Immunity 19, 8394 (2003). This study identifies the role of GATA3 in CD4-lineage differentiation using thymic reaggregation cultures. 98. Nawijn, M. C. et al. Enforced expression of GATA-3 in transgenic mice inhibits TH1 differentiation and induces the formation of a T1/ST2-expressing TH2-committed T cell compartment in vivo. J. Immunol. 167, 724732 (2001). 99. Jenkinson, E. J., Anderson, G. & Owen, J. J. Studies on T cell maturation on defined thymic stromal cell populations in vitro. J. Exp. Med. 176, 845853 (1992). 100. Nawijn, M. C. et al. Enforced expression of GATA-3 during T cell development inhibits maturation of CD8 single-positive cells and induces thymic lymphoma in transgenic mice. J. Immunol. 167, 715723 (2001). 101. Pai, S. Y. et al. Critical roles for transcription factor GATA-3 in thymocyte development. Immunity 19, 863875 (2003). This study uses conditional gene disruption to target GATA3 deletion to DP thymocytes, thereby bypassing blocks resulting from the lack of GATA3 at earlier developmental stages. The result shows that GATA3 is required for CD4+ T-cell differentiation. 102. Hogquist, K. A. Signal strength in thymic selection and lineage commitment. Curr. Opin. Immunol. 13, 225231 (2001). 103. Barber, E. K., Dasgupta, J. D., Schlossman, S. F., Trevillyan, J. M. & Rudd, C. E. The CD4 and CD8 antigens are coupled to a protein-tyrosine kinase (p56lck) that phosphorylates the CD3 complex. Proc. Natl Acad. Sci. USA 86, 32773281 (1989). 104. Veillette, A., Bookman, M. A., Horak, E. M. & Bolen, J. B. The CD4 and CD8 T cell surface antigens are associated with the internal membrane tyrosine-protein kinase p56lck. Cell 55, 301308 (1988). 105. Anderson, S. J., Levin, S. D. & Perlmutter, R. M. Involvement of the protein tyrosine kinase p56lck in T cell signaling and thymocyte development. Adv. Immunol. 56, 151178 (1994). 106. Wiest, D. L. et al. Regulation of T cell receptor expression in immature CD4+CD8+ thymocytes by p56lck tyrosine kinase: basis for differential signaling by CD4 and CD8 in immature thymocytes expressing both coreceptor molecules. J. Exp. Med. 178, 17011712 (1993). 107. Seong, R. H., Chamberlain, J. W. & Parnes, J. R. Signal for T-cell differentiation to a CD4 cell lineage is delivered by CD4 transmembrane region and/or cytoplasmic tail. Nature 356, 718720 (1992). 108. Itano, A. et al. The cytoplasmic domain of CD4 promotes the development of CD4 lineage T cells. J. Exp. Med. 183, 731741 (1996). 109. Matechak, E. O., Killeen, N., Hedrick, S. M. & Fowlkes, B. J. MHC class II-specific T cells can develop in the CD8 lineage when CD4 is absent. Immunity 4, 337347 (1996). References 108 and 109 report the observations that gave rise to the strength-of-signal hypothesis. 110. Tyznik, A. J., Sun, J. C. & Bevan, M. J. The CD8 population in CD4-deficient mice is heavily contaminated with MHC class II-restricted T cells. J. Exp. Med. 199, 559565 (2004). 111. Hernandez-Hoyos, G., Sohn, S. J., Rothenberg, E. V. & Alberola-Ila, J. Lck activity controls CD4/CD8 T cell lineage commitment. Immunity 12, 313322 (2000). This paper documents that interfering with LCK activity redirects thymocytes into the wrong lineage: transgenic LCK molecules with constitutive or disrupted catalytic activity promote CD4- or CD8lineage differentiation, respectively. 112. Bommhardt, U., Cole, M. S., Tso, J. Y. & Zamoyska, R. Signals through CD8 or CD4 can induce commitment to the CD4 lineage in the thymus. Eur. J. Immunol. 27, 11521163 (1997). 113. Sharp, L. L., Schwarz, D. A., Bott, C. M., Marshall, C. J. & Hedrick, S. M. The influence of the MAPK pathway on T cell lineage commitment. Immunity 7, 609618 (1997). 114. Watanabe, N., Arase, H., Onodera, M., Ohashi, P. S. & Saito, T. The quantity of TCR signal determines positive selection and lineage commitment of T cells. J. Immunol. 165, 62526261 (2000). 115. Shores, E. W. et al. Role of TCR chain in T cell development and selection. Science 266, 10471050 (1994). 116. Fischer, K. D. et al. Defective T-cell receptor signalling and positive selection of Vav-deficient CD4+ CD8+ thymocytes. Nature 374, 474477 (1995). 117. Tarakhovsky, A. et al. Defective antigen receptor-mediated proliferation of B and T cells in the absence of Vav. Nature 374, 467470 (1995). 118. Zhang, R., Alt, F. W., Davidson, L., Orkin, S. H. & Swat, W. Defective signalling through the T- and B-cell antigen receptors in lymphoid cells lacking the vav proto-oncogene. Nature 374, 470473 (1995). 119. Hashimoto, K. et al. Requirement for p56lck tyrosine kinase activation in T cell receptor-mediated thymic selection. J. Exp. Med. 184, 931943 (1996). 120. Turner, M. et al. A requirement for the Rho-family GTP exchange factor Vav in positive and negative selection of thymocytes. Immunity 7, 451460 (1997). 121. Pages, G. et al. Defective thymocyte maturation in p44 MAP kinase (Erk1) knockout mice. Science 286, 13741377 (1999). 122. Bosselut, R., Feigenbaum, L., Sharrow, S. O. & Singer, A. Strength of signaling by CD4 and CD8 coreceptor tails determines the number but not the lineage direction of positively selected thymocytes. Immunity 14, 483494 (2001). This paper compared the efficiency of the CD4 and CD8 cytoplasmic tails in promoting CD4- and CD8lineage differentiation. The results indicate that, when fused to the CD8 molecule, the CD4 tail promotes equally the differentiation of MHC class-I-restricted thymocytes into CD4- and CD8-lineage T cells. 123. Yasutomo, K., Doyle, C., Miele, L., Fuchs, C. & Germain, R. N. The duration of antigen receptor signalling determines CD4+ versus CD8+ T-cell lineage fate. Nature 404, 506510 (2000). An in vitro study using an original two-step culture system to propose a role for TCR-signal duration in lineage differentiation. Unlike the kinetic signalling model, lineage choice in this system seems to occur before any change in co-receptor gene expression. 124. Chan, A. C. et al. ZAP-70 deficiency in an autosomal recessive form of severe combined immunodeficiency. Science 264, 15991601 (1994). 125. Elder, M. E. et al. Human severe combined immunodeficiency due to a defect in ZAP-70, a T cell tyrosine kinase. Science 264, 15961599 (1994). 126. Negishi, I. et al. Essential role for ZAP-70 in both positive and negative selection of thymocytes. Nature 376, 435438 (1995). 127. Liu, X. et al. Restricting Zap70 expression to CD4+CD8+ thymocytes reveals a T cell receptor-dependent proofreading mechanism controlling the completion of positive selection. J. Exp. Med. 197, 363373 (2003). 128. Sebzda, E., Choi, M., Fung-Leung, W. P., Mak, T. W. & Ohashi, P. S. Peptide-induced positive selection of TCR transgenic thymocytes in a coreceptor-independent manner. Immunity 6, 643653 (1997). 129. Bhandoola, A., Kithiganahalli, B., Granger, L. & Singer, A. Programming for cytotoxic effector function occurs concomitantly with CD4 extinction during CD8+ T cell differentiation in the thymus. Int. Immunol. 12, 10351040 (2000). 130. Singer, A. New perspectives on a developmental dilemma: the kinetic signaling model and the importance of signal duration for the CD4/CD8 lineage decision. Curr. Opin. Immunol. 14, 207215 (2002). 131. Goldrath, A. W., Hogquist, K. A. & Bevan, M. J. CD8 lineage commitment in the absence of CD8. Immunity 6, 633642 (1997). 132. Schmedt, C. et al. Csk controls antigen receptor-mediated development and selection of T-lineage cells. Nature 394, 901904 (1998). 133. Sohn, S. J., Forbush, K. A., Pan, X. C. & Perlmutter, R. M. Activated p56lck directs maturation of both CD4 and CD8 single-positive thymocytes. J. Immunol. 166, 22092217 (2001). 134. Shimoda, K. et al. Lack of IL-4-induced TH2 response and IgE class switching in mice with disrupted Stat6 gene. Nature 380, 630633 (1996). 135. Takeda, K. et al. Essential role of Stat6 in IL-4 signalling. Nature 380, 627630 (1996). 136. Ouyang, W. et al. Stat6-independent GATA-3 autoactivation directs IL-4-independent TH2 development and commitment. Immunity 12, 2737 (2000). 137. Das, J. et al. A critical role for NF-B in GATA3 expression and TH2 differentiation in allergic airway inflammation. Nature Immunol. 2, 4550 (2001). 138. Canelles, M., Park, M. L., Schwartz, O. M. & Fowlkes, B. J. The influence of the thymic environment on the CD4-versusCD8 T lineage decision. Nature Immunol. 4, 756764 (2003). 139. Aliahmad, P. et al. TOX provides a link between calcineurin activation and CD8 lineage commitment. J. Exp. Med. 199, 10891099 (2004). 140. Komine, O. et al. The Runx1 transcription factor inhibits the differentiation of naive CD4+ T cells into the TH2 lineage by repressing GATA3 expression. J. Exp. Med. 198, 5161 (2003). 141. Keefe, R., Dave, V., Allman, D., Wiest, D. & Kappes, D. J. Regulation of lineage commitment distinct from positive selection. Science 286, 11491153 (1999). 142. Speck, N. A. & Gilliland, D. G. Core-binding factors in haematopoiesis and leukaemia. Nature Rev. Cancer 2, 502513 (2002). 143. Wheeler, J. C., Shigesada, K., Gergen, J. P. & Ito, Y. Mechanisms of transcriptional regulation by Runt domain proteins. Semin. Cell. Dev. Biol. 11, 369375 (2000). 144. Javed, A. et al. Groucho/TLE/R-esp proteins associate with the nuclear matrix and repress RUNX (CBF/AML/PEBP2) dependent activation of tissue-specific gene transcription. J. Cell Sci. 113, 22212231 (2000). 145. Wurster, A. L., Siu, G., Leiden, J. M. & Hedrick, S. M. Elf-1 binds to a critical element in a second CD4 enhancer. Mol. Cell. Biol. 14, 64526463 (1994). 146. Salmon, P., Boyer, O., Lores, P., Jami, J. & Klatzmann, D. Characterization of an intronless CD4 minigene expressed in mature CD4 and CD8 T cells, but not expressed in immature thymocytes. J. Immunol. 156, 18731879 (1996). 147. Uematsu, Y., Donda, A. & De, L. G. Thymocytes control the CD4 gene differently from mature T lymphocytes. Int. Immunol. 9, 179187 (1997). 148. Adlam, M. & Siu, G. Hierarchical interactions control CD4 gene expression during thymocyte development. Immunity 18, 173184 (2003). 149. Donda, A., Schulz, M., Burki, K., De, L. G. & Uematsu, Y. Identification and characterization of a human CD4 silencer. Eur. J. Immunol. 26, 493500 (1996). 150. Allen, R. D., Kim, H. K., Sarafova, S. D. & Siu, G. Negative regulation of CD4 gene expression by a HES-1-c-Myb complex. Mol. Cell. Biol. 21, 30713082 (2001). 151. Bhandoola, A. et al. Positive selection as a developmental progression initiated by TCR signals that fix TCR specificity prior to lineage commitment. Immunity 10, 301311 (1999). 152. Swat, W., Dessing, M., von, B. H. & Kisielow, P. CD69 expression during selection and maturation of CD4+CD8+ thymocytes. Eur. J. Immunol. 23, 739746 (1993). 153. Yu, Q., Erman, B., Bhandoola, A., Sharrow, S. O. & Singer, A. In vitro evidence that cytokine receptor signals are required for differentiation of double positive thymocytes into functionally mature CD8+ T cells. J. Exp. Med. 197, 475487 (2003). 154. Leung, R. K. et al. Deletion of the CD4 silencer element supports a stochastic mechanism of thymocyte lineage commitment. Nature Immunol. 2, 11671173 (2001). 155. Tanigaki, K. et al. Regulation of / T cell lineage commitment and peripheral T cell responses by Notch/RBP-J signaling. Immunity 20, 611622 (2004).
Acknowledgements
I apologize to colleagues whose work could not be mentioned or cited because of space limitations. I am grateful to A. Singer and A. Bhandoola for continued and invaluable exchange of ideas about issues discussed in this review. I thank W. Ellmeier, X. Liu and S. Sarafova for helpful discussions, and J. Ashwell, A. Bhandoola, A. Ggonne and A. Singer for critical reading of the manuscript. Work in my laboratory is funded by the National Cancer Institute.
Online links
DATABASES The following terms in this article are linked online to: Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db =gene CD4 | CD8 | GATA3 | HES1 | HP1 | LCK | MYB | Notch-1 | Notch-2 | RUNX1 | RUNX2 | RUNX3 | STAT6 | TOX | ZAP70 Access to this interactive links box is free online.
540
REVIEWS
CLASS-SWITCH RECOMBINATION: INTERPLAY OF TRANSCRIPTION, DNA DEAMINATION AND DNA REPAIR

Jayanta Chaudhuri and Frederick W. Alt
Class-switch recombination (CSR) of immunoglobulin heavy chains is the genetic process by which a B cell switches from the production of IgM to the production of IgG, IgE or IgA. Although the general characteristics of CSR have been known for some time, the detailed molecular mechanism of this process is only now emerging. CSR is unique, in that it seems to involve transcription-generated, higher-order RNADNA structures, specific DNA deamination and several DNA-repair pathways. In this review, we discuss our current knowledge of the mechanism of CSR and highlight the important unanswered questions.
NON-HOMOLOGOUS END JOINING
(NHEJ). The process that joins broken DNA ends without depending on extended homology. Components of this pathway include the DNAend-binding proteins Ku70 and Ku80, the endonuclease ARTEMIS, X-ray repair crosscomplementing protein 4 (XRCC4), DNA ligase IV and the catalytic subunit of DNAdependent protein kinase (DNA-PKcs).
Howard Hughes Medical Institute, Center for Blood Research and Department of Genetics, Harvard University Medical School, Boston, Massachusetts 02115, USA. Correspondence to F.W.A. e-mail: alt@enders.tch.harvard.edu
doi:10.1038/nri1395
The great diversity of antigen receptors that is produced by the mammalian immune system depends on the unique ability of B and T cells to somatically alter their genomes. B cells are particularly remarkable in that they can undergo three distinct genetic alterations. Developing B-lineage cells in the bone marrow or foetal liver assemble the exons that encode immunoglobulin heavy-chain (IgH) and light-chain (IgL) variable regions from component variable (V), diversity (D) and joining (J) segments, through a process known as V(D)J recombination. This site-directed, antigen-independent recombination event is initiated by the sequence-specific recombination-activating gene 1 (RAG1)RAG2 endonuclease complex and is completed by components of the NON-HOMOLOGOUS END-JOINING (NHEJ) machinery1,2. Productive assembly of IgH and IgL V-region exons allows the expression of IgH and IgL chains as cellsurface IgM by newly generated B cells. IgM+ B cells migrate to secondary lymphoid organs (for example, the spleen and lymph nodes), where they can undergo further antigen-driven immunoglobulin-gene diversification through somatic hypermutation (SHM) and class-switch recombination (CSR) (FIG. 1). SHM introduces point mutations at high rates, into the V-region exons of both IgH and IgL genes, and this enables the selection of B cells that produce higheraffinity antibodies3,4. By contrast, CSR exchanges the
initially expressed IgH constant (C)-region (C) exons for an alternative set of downstream IgH C-region exons, such as C, C or C, known as CH genes. Therefore, CSR allows the expression of antibodies that have the same antigen specificity but are of a secondary IgH isotype (IgG, IgA or IgE) and thereby have a different effector function5,6. Despite being distinct processes targeted to distinct immunoglobulin regions (TABLE 1), CSR and SHM are similar in that they both occur in antigen-stimulated B cells, require transcription through the target region and require the activated B-cell-specific enzyme activation-induced cytidine deaminase (AID)37. The mouse immunoglobulin locus contains eight CH genes, which encode proteins that are capable of different effector functions. Each CH gene, except C, is preceded by specialized DNA sequences called switch (S) regions. CSR involves a recombination event between two S regions, with the intervening sequences, including C, being deleted. As a result, the VDJ exon is juxtaposed with a different downstream C H gene 5,6 (FIG. 1). Mechanistically, CSR is a deletional recombination reaction that, similar to V(D)J recombination, could best be explained by a cut and paste mechanism, in which breaks are introduced to the DNA of the two participating S regions followed by their fusion. However, CSR differs from
VOLUME 4 | JULY 2004 | 5 4 1
REVIEWS
(n = 2501,000) VH1 VHn
DH112
JH14 E
CH
DNA-PKcs Ku70, Ku80 ARTEMIS XRCC4 DNA Ligase IV V DJ E S C C C3
RAG1/2 NHEJ
V(D)J recombination
Excised intervening DNA
C1
C2b
C2a S
3 RR
IgM Recombined CH locus
D JC
At S regions: DNA breaks DNA repair Transcription
AID
Class-switch recombination
DJ E
3 RR
C2a
C C
C2b IgE V D J C C1 C3
Excised intervening DNA
Figure 1 | Rearrangement at the immunoglobulin heavy-chain locus. The variable region of the immunoglobulin heavy chain is assembled from component variable (VH), diversity (DH), and joining (JH) gene segments by V(D)J recombination. The process of rearrangement involves cleavage of the recombination signal sequences in the DNA, which flank the rearranging gene segments, which is carried out by the recombination-activating gene 1 (RAG1)RAG2 complex. Joining of the DNA ends requires nonhomologous end-joining (NHEJ) proteins, including Ku70, Ku80, ARTEMIS, X-ray repair cross-complementing protein 4 (XRCC4), DNA ligase IV and the catalytic subunit of DNA-dependent protein kinase (DNA-PKcs). Transcription across the locus is driven by a promoter upstream of the rearranged VDJ segment (blue arrow), which facilitates the synthesis of a heavy chain. This then associates with a light chain, thereby forming an IgM molecule, which is displayed on the cell-surface of a B cell. Subsequently, secondary isotypes are produced by class-switch recombination (CSR), a process that exchanges the constant region of the heavy chain (CH) with a set of downstream constant-region genes (CSR to IgE is shown). This deletional-recombination reaction, which requires the enzyme activation-induced cytidine deaminase (AID), involves the generation of DNA breaks at switch (S) regions, which precede the constant-region genes, followed by the repair of DNA. This leads to a rearranged CH locus and deletion of the intervening sequence as an episomal circle. Cytokines stimulate transcription (red arrows) through the CH gene and determine the immunoglobulin isotype that the B cell will switch to. The rearranged variable regions of both the heavy and light chains also undergo a high rate of point mutation through the process of somatic hypermutation (SHM) (not shown). The E and 3-regulatoryregion (3 RR) enhancers influence V(D)J recombination and CSR, respectively.
V(D)J recombination in several ways: first, CSR does not require RAG8, and second, CSR has no well-defined consensus target sequence9. Recent studies have indicated a working model for the initiation of CSR, in which transcription through the S regions renders these regions substrates for modification by AID10,11. Other work has implicated several DNA-repair pathways in the processing and joining of AID-initiated S-region breaks that complete CSR12. In this context, several lines of evidence indicate that CSR uses a DNA double-stranded-break (DSB) intermediate5,6. In this review, we discuss the current understanding of CSR and the role of S regions, transcription, AID, DSBs and DNA-repair proteins.
Transcription-generated S-region structures
CYTOKINES
Biologically active molecules that are released by cells and affect the function of other cells.
S regions consist of highly repetitive 112-kilobase (Kb) sequences with G-rich non-template strands13. The S-region repeats can be divided into two general categories: S, S and S are composed of variations of pentameric sequences, whereas S1, S2a, S2b and S3 mostly contain 4952-base-pair repeats (FIG. 2a). Deletion
of most of the S region has been shown to severely impair CSR to downstream CH genes14, whereas deletion of the entire S1 region essentially blocks switching to IgG1 (REF. 10), indicating that the S regions are specialized targets for the CSR process. However, there are certain caveats associated with this interpretation, as neither S-region-deletion study replaced the deleted sequences with an unrelated sequence of equal size. Transcription through a particular S region is required for CSR to the corresponding CH gene5. In response to specific activators and CYTOKINES (TABLE 2), transcription is initiated from an activation/cytokineresponsive promoter upstream of an exon (known as the I exon) that precedes all CH genes that undergo CSR (FIG. 2b). Transcription proceeds through the S region and terminates downstream of the corresponding CH gene. The primary transcript is spliced to remove the intronic S region and to join the I exon to the CH exons, producing a seemingly non-coding transcript5 (FIG. 2b). This requirement for transcription and transcription-control elements for CSR has been demonstrated directly in gene-targeting experiments, in which deletion of I-exon
542
REVIEWS
Table 1 | Main differences between CSR and SHM

Parameters Target Transcription-generated structures AID activity Completion residue can lead to SHM; SHM Variable-region coding DNA in both the heavy and light chains Transient transcription bubbles, other structures? Single deaminated residue can lead to SHM Replication across deaminated intermediates probably does not require DSB intermediates CSR Switch-region non-coding DNA in the constant region of heavy chains Stable RNADNA hybrids, other structures? High density of deamination might be required, at least in one strand Probably requires DSB
AID, activation-induced cytidine deaminase; CSR, class-switch recombination; DSB, double-stranded break; SHM, somatic hypermutation.
promoters or certain 3 IgH-enhancer elements abolished or greatly reduced germline transcription of the corresponding CH genes and CSR to these genes5,15. Furthermore, constitutively active heterologous promoters inserted in place of endogenous I-region promoters can also drive CSR5,16. Therefore, transcription directly influences CSR. The role of germline transcription in CSR has long been a subject of speculation. Transcription has been postulated to function at several non-mutually exclusive levels, including rendering S-region chromatin structures accessible to the enzymes involved in CSR17,18 and generating DNA structures able to be substrates of a CSR recombinase1922. Support for the DNA-structure model arises from the notable feature of all mammalian S regions that the non-template strand is G-rich. Moreover, when S regions are transcribed in vitro in the normal physiological orientation, the transcripts stably associate with the template strand of the DNA to form RNADNA hybrids1924 (FIG. 2c). More recently, these RNADNA hybrids were demonstrated to form R loops, in which the displaced non-template strand exists as single-stranded (ss) DNA, both in vitro 22,25 and in vivo 25. Theoretically, the non-template strand of the RNADNA hybrid could assume other structures in addition to being singlestranded in R loops: for example, stem loops, from the abundant palindromic motifs in S regions26, or parallel four-stranded guanosine quartets, stabilized by Hoogstein pairing between G residues27,28 (FIG. 2c). However, although these latter structures have been postulated to have a role in CSR26,28, there is no direct evidence to support this hypothesis, and the existence of stem loop and G-quartet structures during in vivo CSR has not been demonstrated. Thermodynamic properties predict that stable R loops should form only when mammalian S regions are transcribed in the physiological orientation29, and this prediction has been confirmed by in vitro-transcription studies1925. Therefore, if RNADNA hybrids have a mechanistic role in CSR, the transcriptional orientation of S regions should influence CSR considerably. Studies carried out using extrachromosomal substrates have provided the initial evidence that recombination at S regions is orientation dependent21; although this idea has been challenged by other studies that used chromosomally integrated, short S-region substrates30. Recent
studies directly examined orientation in vivo and found that inversion of the 12-Kb S1 significantly reduced CSR to IgG1, although it was not completely abolished10. In addition, a random synthetic 1-Kb sequence that was G-rich on the non-template strand and formed R loops in vitro, but lacked the characteristic S-region repeat elements, could replace native S1 in supporting CSR in vivo, although at a reduced level. Importantly, the activity of this synthetic sequence was again dependent on transcriptional orientation, because it targeted almost no CSR activity when in the inverted (C-rich), non-R-loop forming configuration10. These findings provided evidence for the orientation dependence of S-region transcription in vivo and led to the proposal that transcription-generated DNA structures, in particular ssDNA, might be substrates for the enzymes involved in CSR10. In addition, the finding that CSR can be supported by a synthetic G-rich sequence with no repeat motifs characteristic of endogenous S1 provides further evidence that not only primary S-region structures but also higher-order secondary structures have an important role in targeting CSR. The finding that inverted S1 supports some CSR (25% of the normal level), even though this orientation does not lead to readily detectable R-loop formation, indicates that other mechanisms might exist to reveal ssDNA structures during CSR10. Consistent with this, Xenopus S regions are (A+T)-rich31 so would not be predicted to show extensively displaced non-template ssDNA strands. However, similar to mammalian S regions, Xenopus S regions are rich in palindromic repeats, which if rendered single-stranded, could form stem loops. Indeed, recombination breakpoints in Xenopus S regions are observed frequently at positions where a computer program, which models ssDNA folding, predicts a transition from a stem to a loop structure31. Consistent with this, sequences with the potential to form palindromes supported low-level recombination within substrates introduced into cell lines26. Notably, other work has implicated transcriptionally generated ssDNA substrates as having a role in targeting SHM to sequences that do not form R loops3234, indicating additional, potentially factor-dependent mechanisms for generating substrates that do not readily reveal extensive ssDNA structures. Given that the initiating enzyme, AID, is the same for SHM and CSR, this finding might also have implications for this aspect of CSR. In support of
VOLUME 4 | JULY 2004 | 5 4 3
REVIEWS
a
Region S S3 S1 S2b S2a S S Non-template strand GAGACTGAGCTGGGGTAGCT GGGGACCAGGCTGGGGCAGCTCTNGGGGAGCTGGGGTAGGTTGGGAGTGT GGTGACCCAGCAGAGCAGCTCCAGGGGCGCCAGGACAGGTGGAAGTGT AGGGGACCAGWCCTAGCAGCTRTGGGGGAGCTGGGGAWGGTGGGAATGTG GGGACCAGGCAGTACAGCTCTGGGTRGGGRNCAGGCAGTACAGCTCTGNGTG GGGCTGGGCTGAGCTGRGCTGAGCTGRGCTGAGCTRARNT ATGAGCTGGGATGRRCTGAGCTAGGCTGGAATAGGCTGGGCTGGGCTG GTGTGAGCTGGGTTAGGCTGAGCTGAGCTGGA Repeat length (bp) 1040 49 49 49 52 4050 80 Approximate total length of core (bp) 3.2 2.5 12.0 5.0 2.5 1.0 4.2
b
Ix P
c
G-rich
Sx Cx exons Transcription unit Transcription Demonstrated in vivo Sx RNADNA hybrids Single-stranded DNA R loop RNA Stem loops Hypothetical Primary transcript
G-rich
Sx
G G
Hypothetical
G G
G quartet
AAAA
Processed germline transcript
Hypothetical
Factor-dependent AID transcriptioncoupled structures RPA
Figure 2 | Switch-region sequence, transcription and transcription-generated structures in CSR. a | The switch (S)-region repeat elements, which are similar but not identical to each other, are shown. All S regions are G-rich in the non-template (top) strand. N, any nucleotide; R, a purine; W, A or T; Y, a pyrimidine. The table has been compiled from previous studies15. b | The heavy-chain constant-region (CH) genes are transcription units, in which transcription initiates from cytokine-inducible I-promoters (P) upstream of the I-exon (I) and proceeds through S regions and the CH exons. The primary transcript is processed to generate mature germline transcripts that do not code for proteins. c | S-region transcripts can stably associate with the template strand to form RNADNA hybrids, in which the non-template strand can theoretically assume several structures (including G quartets or stem loops) or can remain single stranded (R loops). Although R loops have been detected in vivo and are strongly implicated to have a mechanistic role, at present, there is no direct experimental evidence for the formation of stem loops and G-quartet structures in vivo. AID, activationinduced cytokine deaminase; bp, base pairs; RPA, replication protein A; UNG, uracil-DNA glycosylase; x, any constant-region gene.
RGYW MOTIFS
The nucleotide sequence RGYW (where R denotes A or G, Y denotes C or T, and W denotes A or T) is considered a hot spot for somatic hypermutation, as it is considerably more mutable than other random sequences.
this, in Xenopus S regions, the areas where most CSR breakpoints occur are rich in RGYW MOTIFS, which are also hot spots for SHM31. Therefore, these sequences might function somehow to target AID activity in both processes. In this regard, recent studies have implicated replication protein A (RPA), a ssDNA-binding protein, as a factor that can target AID to transcribed sequences in the context of RGYW motifs (Chaudhuri, J., Khong, C. & Alt, F. W., unpublished observations). Overall, current findings indicate that transcription might function in the context of various sequence motifs, and several mechanisms might exist to provide ssDNA substrates that are probably the targets of AID activity.
The Function of AID in CSR
AID and DNA deamination. The AID cDNA was cloned from a subtractive screen for genes that were upregulated in a B-cell line induced to undergo CSR35.
AID is only expressed by activated B cells35 and is essential for CSR and SHM36,37. Ectopic AID expression by non-lymphoid cells enabled CSR and SHM on artificial substrates3840, implicating AID as the only lymphoidspecific factor required for these processes. However, the efficiency of CSR was low in these ectopic-expression studies, and whereas SHM is restricted primarily to V genes in B cells, several transcribed DNA sequences were shown to serve as AID substrates during SHM in nonlymphoid cells, implying that additional factors might be required for the efficiency and specificity of CSR and SHM. AID shares sequence homology with the RNAediting cytidine deaminase APOBEC1 (apolipoprotein B mRNA-editing enzyme, catalytic polypeptide 1)35, which led to initial speculation that AID edits an mRNA such that it encodes a novel protein required to activate CSR and SHM36,37 (FIG. 3). Although this is still considered a possibility by some41, no direct experimental
544
REVIEWS
that are activated for CSR50. A deficiency in H2AX also leads to S-region translocations51. Furthermore, a deficiency in p53-binding protein 1 (53BP1), an early component of the DSB response, severely impairs CSR52. Finally, some components of the NHEJ pathway, Ku and the catalytic subunit of DNA-dependent protein kinase, DNA-PKcs, are required for effective CSR in vivo, which also indicates a role for DSB intermediates5355. Although such evidence points strongly to the involvement of DSB intermediates in CSR, this interpretation requires some caution. LM-PCR studies have detected DSBs in V genes, both in the presence and absence of AID5658, and although -H2AX foci clearly form at the IgH locus in B cells undergoing CSR, it has not been formally proven that this occurs in response to DSBs in S regions, rather than other regions of the IgH locus59. However, despite these potential caveats, all available data are consistent with the idea that CSR proceeds through DSB intermediates. Given that DSBs are essential intermediates in CSR, how does AID-mediated DNA deamination lead to DSBs? It was proposed that during CSR, the AIDintroduced deoxyuridine in S-region DNA is removed by the base-excision-repair (BER) enzyme uracilDNA glycosylase (UNG) to generate an abasic site, the processing of which, by the apurinic/apyrimidinic endonuclease 1 (APE1), creates a nick43. A closely spaced, similarly created nick on the opposite strand could lead to a staggered DSB. In addition, the deoxyuridinedeoxyguanosine mismatch could also be processed by components of the mismatch repair (MMR) machinery to generate staggered DSBs43. Accordingly, both UNG deficiency60,61 and MMR deficiency6264 result in CSR defects. Therefore, deaminated DNA might be processed by several DNA-repair pathways, which ultimately lead to the generation of a DNA DSB (FIG. 4). The AIDUNGAPE1 pathway for the generation of DSBs predicts that AID needs to deaminate both the template and the non-template strands during CSR. In vitro studies using a G-rich synthetic substrate showed that although the non-template ssDNA is the preferred AID substrate, the template strand is also deaminated11. How the template strand is rendered a substrate is not known, but several mechanisms have been considered. One possibility is that AID acts at ssDNA exposed by the transition between duplex DNA and the R-loop structure22. In addition, it has been proposed that collapse of R loops, perhaps secondary to RNA removal by RNase H, could lead to misalignment of repeats on opposite strands, leading to single-stranded loops on both the template and the non-template strands of the DNA65. Furthermore, there must be mechanisms other than R loops that allow AID to have access to the DNA (discussed earlier); this includes the possibility that additional factors, such as RPA, target AID to non-R-loop-forming sequences in the context of transcription-coupled mechanisms. In theory, several mechanisms could work together to generate lesions on both S-region strands. For this to occur, a mechanism that generates a
Table 2 | In vitro activation of CSR

Cytokine LPS LPS and IL-4 CD40-specific antibody and IL-4 LPS and IFN- LPS, TGF- and IL-5 Ig isotype IgG2b, IgG3 IgG1, IgE IgG1, IgE IgG2a IgA
CSR, class-switch recombination; IFN, interferon; IL, interleukin; LPS, lipopolysaccharide; TGF, transforming growth factor
LIGATION-MEDIATED PCR (LM-PCR). This method, which
involves the ligation of unphosphorylated adaptors to broken DNA followed by PCR, can detect DNA doublestranded breaks in defined regions of the genome. It has been used extensively to analyse recombination-activating gene (RAG)-dependent breaks during V(D)J recombination, and more recently, in variable (V) gene segments and switch (S) regions during somatic hypermutation and class-switch recombination, respectively.
evidence has been obtained to support the RNA-editing activity of AID. Moreover, it now seems that APOBEC1 is an outlier in a family of DNA-cytidine deaminases that evolved to mediate innate immunity42. Therefore, for the remainder of the review, we focus on AID as a DNA-modifying enzyme, because this activity has been clearly documented. Several lines of evidence indicate that AID acts directly on DNA as a DNA-cytidine deaminase. First, enforced expression of AID in Escherichia coli, which are not expected to contain any putative mRNA substrates of AID, induced mutations in several bacterial genes43. Second, in vitro studies showed that purified AID is a ssDNA deaminase that converts deoxycytidine residues to deoxyuridine on target DNA11,33,44,45. Last, chromatinimmunoprecipitation experiments demonstrated that ectopically expressed AID is directly associated with the transcribed S-region DNA in cells that are undergoing CSR46. The finding that AID deaminates only ssDNA in vitro supports the argument that CSR might be enhanced by ssDNA targets, such as those provided by R-loop formation. Indeed, the G-rich DNA that supported CSR in vivo 10 was efficiently deaminated by AID in vitro, mainly on the non-template strand (that is the ssDNA of the R loop), when the reaction was coupled to transcription. Conversely, C-rich DNA that did not target CSR in vivo was a poor AID substrate when transcribed in vitro11. Overall, the current findings provide a plausible model by which transcription through mammalian S regions could generate ssDNA R-loop substrates for the cytidine deaminase activity of AID and thereby generate the initiating DNA lesion that is ultimately processed into a DNA break in the S region. In addition, it is probable that there are factors that function in the context of transcription of non-R-loop-forming sequences to generate ssDNA structures that are also AID targets, potentially focusing somehow on RGYW motifs for SHM. Moreover, such a mechanism might also function in CSR, in conjunction with R-loop formation in mammalian S regions (perhaps to target the template strand), or in place of R loops for Xenopus S regions. DNA DSBs in CSR. The generation of an episomal circle in CSR strongly suggests that CSR proceeds through DNA DSB intermediates. Indeed, in LIGATION-MEDIATED PCR (LM-PCR) assays, DSBs have been detected in S regions47,48, and these appear to be AID dependent48. In addition, the foci of the phosphorylated histone H2AX (-H2AX), which form around DNA DSBs49, occur at the IgH locus in an AID-dependent fashion in B cells
VOLUME 4 | JULY 2004 | 5 4 5
REVIEWS
ability to cleave ssDNA in transcribed S regions in vitro22. However, mutations in XPF or in XPG do not have a readily detectable effect on the rate of CSR71,72; although this finding does not rule out redundancy with other repair pathways.
Cell-cycle regulation of CSR
a
Cytidine-deaminase motif (5594)
9 20 110
APOBEC1-like C-terminal domain (119181) Helix

1
Pseudoactive site
148 183 198
N NLS NLS
Active site
Linker
NES NES
b
H NH2 O Zn AID OH N O N O Enz AID OH
H H H2N HN N O O O Zn AID AID OH HN N O H O Zn AID
Cytidine
[Transition state]
Uridine
Figure 3 | AID and DNA deamination. a | The primary structure of activation-induced cytidine deaminase (AID) is shown, depicting the nuclear-localization sequence (NLS), nuclear-export sequence (NES) and other predicted domains based on the structure of apolipoprotein B mRNA-editing enzyme, catalytic polypeptide 1 (APOBEC1)103. Some of the mutations that affect AID function are shown: mutations in the active site and linker region (blue arrows) impair both class-switch recombination (CSR) and somatic hypermutation (SHM), whereas those at the carboxyl (C)-terminus (orange arrows) and a ten amino-acid C-terminal deletion (not shown) impair CSR without affecting SHM. N, amino terminus. b | Mechanism of cytidine deamination, based on a bacterial cytidine deaminase that is homologous to APOBEC1 and AID. The reaction proceeds through a direct nucleophilic attack at position 4 of the pyrimidine ring by zinc ions (Zn2+) coordinated with AID12.
high density of nicks on one strand (for example, by R-loop formation) would only require a mechanism that generates a low density of nicks on the other strand, to allow formation of a set of closely opposing nicks that would create a DSB (FIG. 5). Processing of DSBs during CSR. Irrespective of the process that generates AID substrates, the deamination model predicts that most DSBs would result from staggered cleavages. Indeed, studies using LM-PCR assays have indicated that most AID-dependent S lesions are staggered single-stranded breaks66. Such lesions might need to be filled or resected to generate blunt DSBs or overhangs short enough for efficient ligation6. This could be achieved by either nuclease-mediated recession or fill-in through short-patch DNA synthesis, using error-prone polymerases, and could potentially account for the deletions and mutations observed at S-region junctions9 (FIG. 4). Consistent with this, the absence of the error-prone polymerase has been shown to affect the mutation profile of S-region junctions67. The MMR protein mutS homologue 2 (MSH2) has also been implicated in end processing during CSR63,64,68,69, and because yeast MSH2 homologues can recruit nucleases, such as the RAD1RAD10 complex, to cleave branched intermediates70, it is possible that MSH2 could similarly recruit the mammalian homologues of RAD1 and RAD10, the ERCC1XPF complex (excision repair cross-complementing protein 1xeroderma pigmentosum, complementation group F), in B cells to process DNA ends. Both ERCC1XPF and XPG (xeroderma pigmentosum, complementation group G) have been speculated to participate in CSR as a result of their
V(D)J recombination occurs in the G1 phase of the cell cycle where NHEJ is the main DSB-repair pathway1, and understanding the point at which CSR occurs in the cell cycle is crucial for elucidating its mechanism. CSR requires proliferation73, and inhibitors of DNA synthesis block CSR, indicating a role for progression into S phase and/or replication16. In this context, H2AX deficiency on a p53-mutant background leads to B-lineage tumours with S-region translocations that might have been initiated in S phase51. By contrast, on the basis of the point in the cell cycle at which AID-dependent, IgH-associated -H2AX foci appear in B cells activated for CSR, it has been argued that AID deamination and generation of CSR DSBs occurs in G1 (REF. 50). We note that none of the existing pieces of evidence relevant to the cell-cycle dependence of CSR are mutually exclusive, and it is not impossible that processing of lesions initiated in G1 might occur in S phase. In terms of the mechanism, replication, if involved, might function to convert single-stranded S-region nicks into DSBs74. Alternatively, CSR might use pathways or factors that are only active, or mainly active, in a particular phase of the cell cycle. Additional work on this important aspect of CSR is required.
Switch region synapsis
SYNAPSIS
Non-covalent juxtaposition of two non-adjacent stretches of DNA.
Internal deletions in S in the absence of CSR. CSR involves recombination between two S regions, and an efficient recombination reaction probably requires close juxtaposition, or SYNAPSIS, between the two recombining sequences. During V(D)J recombination, the RAG1RAG2 endonuclease complex cleaves its recognition sites only in the context of a preformed synaptic complex involving two complementary RAG-recognition signal sequences1. For CSR, coordinate interaction between two different donor and acceptor S regions does not seem to be required to target AID activity, as indicated by the observation that constitutively transcribed S or S2b regions integrated randomly into a pro-B cell line undergo a high rate of internal deletions following AID expression75. In splenic B cells and in B-cell lines activated to undergo CSR, large deletions occur frequently within the endogenous S region7577 (FIG. 5). These deletions are AID-dependent and therefore seem to represent the same general process as CSR75; however, there has been no direct demonstration that the S breaks that lead to internal deletions are the same as those channelled into CSR with a downstream S region. In theory, unique structural or transcriptional properties of S (for example, proximity to the E intronic enhancer) render it a preferred AID substrate, with AID-targeting to downstream S regions constituting the rate-limiting step in CSR78. Therefore, in the
546
REVIEWS
H2AX might work in a similar way to facilitate the joining of two S regions that contain DSBs80. Correspondingly, deficiency for 53BP1, an H2AXassociated DSB response factor nearly abrogates CSR, due to the impairment of CSR steps downstream of AID induction and germline transcription; this is consistent with a role for this factor in the joining phase of CSR52. Although the idea that -H2AX and its associated factors might function in the context of S-region synapsis during CSR is intriguing, other interpretations are possible. To fully evaluate the synapsis model, it will be important to demonstrate that S deletions and mutations represent a true CSR-initiation reaction, rather than an SHM-like process promoted by proximity to the elements that specify SHM. The ataxia telangiectasia mutated (ATM) protein, which is a master regulator of the DSB response, has been suggested to have a role in CSR, as ataxia telangiectasia patients, who have mutations in ATM, have modestly reduced levels of IgA, IgE and IgG subclasses81 and their SS junctions show significantly increased microhomology82,83. The precise role of ATM in CSR is unknown, and it is not clear if it has a direct role in CSR or whether the decreased CSR potential of ATM-deficient B cells is secondary to other defects in B-cell development. It is to be noted, however, that H2AX and 53BP1 are activated in the DSB response by ATM and other phosphatidylinositol 3-kinase (PI3K)related kinases, such as ATM-related protein (ATR) and DNA-PKcs8486. The CSR defect of 53BP1-deficient cells seems more severe than in ATM-deficient cells52,87, potentially indicating roles for other PI3K-related kinases in CSR. Finally, DNA-PKcs has the ability to synapse DNA ends in vitro 88 and, theoretically, could have such a function in CSR. The heterodimer PMS2MLH1 (post-mitotic segregation 2MutL homologue 1) participates in MMR downstream of MSH2-mediated recognition of DNA mismatches. Mice deficient in PMS2 and MLH1 show a 25-fold decrease in the efficiency of CSR, and their S junctions have increased microhomology relative to normal mice68,89. These findings, along with the observation that yeast homologues of PMS2 and MLH1 can bind directly to two different DNA molecules90, led to the proposal that these proteins might participate in S-region synapsis during CSR91. MLH1 interacts with ATM both in vitro and in vivo, and the interaction is stabilized by DSBs92, providing a further potential link. Finally, the DNA binding protein lipopolysaccharideresponsive factor 1 (LR1) has been proposed to participate in S-region synapsis28. LR1 is a heterodimer of nucleolin and heterogeneous nuclear ribonucleoprotein D (HNRD), and each subunit can individually bind to G-quartet DNA in vitro, leading to the proposal that a single heterodimer of LR1 can bind to two separate S regions to juxtapose donor and acceptor S regions during CSR28; however, this has not yet been formally tested. In summary, although several factors have been suggested to be involved in the synapsis of donor and acceptor S regions, none of the factors tested have
5 3
C G AID
G C
3 5
S-region DNA
5 3
U G
G U
3 5
Deaminated DNA
(Other minor pathways?)
(Main pathway)
BER UNG1
MMR
Abasic site
5 3
G G APE1 Nick
3 5
MSH2 MSH6
5 3
MLH1 PMS2 EXO1
3 5 5 3
G G
3 5
Staggered DSB
5 3 3 5
Staggered DSB
5 3 3
G
5
Nuclease
5 3
End processing Fill-in Or

5 3
Nuclease Blunt DSBs

5 3
End processing Fill-in Or

5 3
Figure 4 | Generation of DNA double strand breaks in S regions. Activation-induced cytidine deaminase (AID) deaminates cytidine residues in DNA converting them to uridine residues. The GU mismatch can then be processed by either the base-excision repair (BER) pathway or the mismatch-repair (MMR) machinery which includes the mutS homologue 1 (MSH1), MSH6, exonuclease 1 (EXO1), mutL homologue 1 (MLH1) and postmitotic segregation (PMS) proteins to introduce gaps or nicks on opposite strands of the switch (S)-region DNA. The nicks induced by the BER pathway are thought to be generated by the following process: uracil-DNA glycosylase (UNG) removes the AID-introduced deoxyuridine in S-region DNA, creating an abasic site (blue circles) that is processed by the apurinic/apyrimidinic endonuclease 1 (APE1), which creates the nick. Processing of the staggered ends by unknown exonucleases or by error-prone end-filling reactions (orange lines) can lead to blunt double-stranded breaks (DSBs) that can then be ligated to similarly created breaks on downstream S-region DNA to complete class-switch recombination (CSR).
absence of an acceptor S region, DSB S lesions would be resolved by intra-S-region recombination, leading to internal S deletions. Alternatively, when the acceptor S region DSBs are available, S breaks could be efficiently channelled into a genuine CSR reaction. This scenario is consistent with the observation that downstream S-region deletions, in the absence of productive CSR, have rarely been observed75, unless S was effectively inhibited from being a donor77. DSB response and other factors potentially involved in synapsis. Several proteins have been potentially implicated in S-region synapsis. Deficiency in H2AX reduces CSR by 5080%50, but mutations and deletions within S occur at normal levels, indicating that targeting of AID to S and assembly of DNA-repair components is not affected and that the CSR defect might be due to an inability to synapse S breaks with downstream S regions79. In this regard, -H2AX has been proposed to function as an anchor to recruit and assemble lattices of DNA-repair factors at sites of DSBs and thereby to hold broken chromosomal ends in close proximity to allow proper NHEJ80. In theory,
VOLUME 4 | JULY 2004 | 5 4 7
REVIEWS
been found to be absolutely required for CSR, and their precise roles remain to be determined. So, elucidation of the S-region-synapsis mechanism remains an important challenge.
a
V DJ S Transcription AID V DJ Stable R-loop formation C C C C AID C C C C C C C S C
DNA end joining
APE1 UNG V DJ U U U U
RNA polymerase U
Cytidine deamination
UNG APE1 U U C
U U
Base-excision repair or mismatch repair V DJ C C
Closely spaced nicks ( ) on opposite strands converted into double-stranded breaks
Synapsis (possibly inducing H2AX, 53BP1, ATM, LRI, MLH1, DNA-PKcs) and end joining (possibly by NHEJ) V DJ C Productive CSR
b
AID targeted to S only V DJ U U U U U C C C C C C
DJ
C S
DJ
C S
C S internal deletion
Figure 5 | Current model for CSR. a | Transcription through the two participating switch (S) regions generates R loops (or other secondary structures, not shown) that provide singlestranded DNA substrates for activation-induced cytidine deaminase (AID). Subsequently, the activities of uracil-DNA glycosylase (UNG) and apurinic/apyrimidinic endonuclease 1 (APE1) can introduce a high density of nicks on the non-template strand, and this, combined with a closely spaced nick on the opposite strand (*), can generate double-stranded breaks (DSBs) in the S regions. The breaks are synapsed in a process that requires the putative participation of several proteins, including histone 2A family-member X (H2AX), p53 binding protein 1 (53BP1), lipopolysaccharide-responsive protein1 (LR1), ataxia telangiectasia mutated (ATM), mutL homologue 1 (MLH1) and DNA-PKcs (catalytic subunit of DNA-dependent protein kinase). Class-switch recombination (CSR) is completed by fusion of two S regions DSBs, possibly by the non-homologous end-joining pathway (NHEJ) which has been strongly implicated to have a role in CSR, although this has not been unequivocally proven. The pink and orange oval at the recombination site represents deletions and mutations that are found at S junctions. b | In addition to productive CSR, the S region can undergo a high rate of internal deletions (S) when this region is targeted by AID but the downstream S region is not. It is not clear if such intraswitch recombination reactions have the same requirements as productive CSR. C, constant region of IgM heavy chain; D, diversity gene segment; J, joining gene segment; V, variable gene segment.
Completion of CSR seems to involve the joining of two broken S regions. The two main pathways that join broken DNA ends in mammalian cells are homologous recombination and NHEJ. The limited or complete lack of homology at S-region junctions rules out the possibility that the former pathway has a main role, leaving NHEJ as the most probable candidate. A role for NHEJ in CSR was indicated by the observation that Ku70- or Ku80-deficient B cells reconstituted with rearranged IgH and IgL chains had severely impaired CSR53,54,79. However, such an effect on CSR could be indirect as Ku-deficient B cells had proliferative defects and died when induced to undergo CSR54,79. The latter effect might result from an inability to repair DSBs generated during CSR, because cells activated to divide, but not undergo CSR, proliferated relatively normally54. It is also possible that Ku-deficient cells could die before undergoing productive CSR owing to abortive S breaks that are uncoupled from CSR. Although this argument has been countered by the finding that Ku80-deficient B cells that have undergone several rounds of cell division still show impaired CSR79, such cells could represent those that have escaped activation. Therefore, although Ku activity is required for CSR, it has not been formally proven that its role is through NHEJ. Unlike Ku-deficient B cells, B cells from mice with a targeted disruption of DNA-PKcs, which completely eliminates expression of any DNA-PKcs protein, showed no proliferation defects55. Yet DNA-PKcs deficiency severely impaired CSR to all IgH isotypes except IgG1 (REF. 55), strongly indicating a direct role for DNAPKcs in CSR. However, analysis of B cells derived from mice homozygous for the severe combined immunodeficiency (SCID) mutation (which inactivates the DNA-PKcs kinase domain but leaves most of the protein intact) has complicated this interpretation. In one study, B cells from SCID mice were found to undergo nearly wild-type levels of CSR93, whereas another study reported reduced, but significant, levels of CSR to most CH genes94. There could be several possible interpretations of these results: an intriguing one being that the kinase-deficient DNA-PKcs protein in B cells from SCID mice could provide a necessary function for CSR. In this scenario, the role of DNA-PKcs in CSR would be separate from its role in NHEJ, which requires a functional kinase domain. Accordingly, one of the main roles of DNA-PKcs, in the NHEJ component of V(D)J recombination, is end processing through phosphorylation of the ARTEMIS endonuclease95, and ARTEMIS is dispensable for CSR (Rooney, S., Manis, J. P. & F.W.A. unpublished observations). One potential kinase-independent role for DNAPKcs could be its previously mentioned function as a synaptic factor, because the DNA-bridging property of DNA-PKcs in vitro is independent of its kinase activity88. Other potential DNA-PKcs roles could be the activation of other factors involved with AID recruitment or in the response to DSBs downstream of AID. The observed DNA-PKcs-independent CSR activity of S1 might reflect the differential ability of this S region,
548
REVIEWS
which is considerably larger than the others, to be synapsed or targeted for AID activity. Finally, genetic experiments have provided clear evidence that Ku and DNA-PKcs have NHEJ-independent functions in DNA repair, which might overlap with ATM96. So, in this way, even if the CSR defects observed in Ku and DNA-PKcs reflect a true requirement of the proteins in CSR, this does not formally prove that their effects on CSR are mediated through the NHEJ pathway. So far, X-ray repair cross-complementing protein 4 (XRCC4) and DNA Ligase IV are the only proteins required for all types of NHEJ reactions that have no reported roles outside NHEJ. Therefore, although most available evidence points to a role for NHEJ factors in CSR, elucidation of the role of XRCC4 or DNA ligase IV would provide the most unequivocal proof. Potential Roles of AID in CSR downstream of deamination. Recently, it has been argued that AID itself might participate in the joining phase of CSR. Mutational analyses revealed the C-terminus of AID to be crucial for CSR but dispensable for SHM and gene conversion97,98 (FIG. 3). Mechanistically, CSR and SHM pathways diverge with respect to targeting V genes versus S regions and in the requirement of proteins downstream of deamination for example, H2AX, 53BP1, and DNAPKcs for CSR50,52,55 but not SHM52,79,99, probably because of the mechanistic requirement for DSBs in CSR
Table 3 | Role of proteins involved in CSR and SHM
Protein B-cell specific AID Base-excision repair UNG Reduced transversion mutations at GC base pairs Decrease in AT mutations Decrease in AT mutations Decrease in AT mutations None Severely impaired Removes deoxyuridine from S regions to create abasic sites 60,61 Blocked Blocked Single-stranded cytidine deaminase, role in recruiting DNA-repair protein? 11,3237, 44,45,97,98 SHM phenotype in absence of activity CSR phenotype in absence of activity Activity in CSR References
but not SHM (TABLE 3). C-terminal AID mutations do not impair S mutations, indicating that targeting of AID, at least to S, is not affected, leading to the suggestion that the C-terminus of AID is involved in the recruitment or assembly of CSR DNA-repair proteins or in the synapsis of the acceptor and donor S regions97,98. However, an alternative interpretation of these interesting findings could be that these AID mutants, owing to an altered conformation or an inability to interact with CSR-specific cofactors, cannot target R loops or other CSR-associated DNA structures with a frequency high enough to generate DSBs in S regions. In this scenario, S mutations might not represent a CSR-initiation reaction, as they might occur through an SHM-like mechanism (for example, because of their proximity to the IgH intronic enhancer). Unlike CSR, SHM probably does not proceed through DSB intermediates, and therefore might not require a high density of AID-mediated target-DNA deaminations. Concerning the hypothesis that CSR might require a higher incidence of deamination relative to that required for SHM, the observation that V genes have a much higher rate of SHM mutation than S regions would need to be accounted for. It is conceivable, given that the proteins and pathways required downstream of deamination are clearly different between CSR and SHM, that during CSR, but not SHM, most deaminated residues in S regions are converted into abasic sites and then into nicks.
Mismatch repair MSH2 PMS2, MLH1 EXO1 NHEJ DNA-PKcs/ DNA-PKcs inactivated (SCID) Ku70, Ku80 Other DNA-repair proteins H2AX 53BP1 ATM None None None 5080% reduced, no effect on S mutations and deletions Severely impaired S junctions have increased microhomology Synapsis? Repair, synapsis? End processing? synapsis? 50,79 52 82,83 None Severely reduced except to IgG1 Reduced CSR to all isotypes Severely impaired End joining? End joining? synapsis? 54,99 93,94 53,54,79 25-fold reduction 25-fold reduction Severely impaired Mismatch recognition, end processing Mismatch processing, end-region synapsis Mismatch processing 6264,68, 69,104 68,89,91 105
53BP1, p53 binding protein 1; AID, activation-induced cytidine deaminase; ATM, ataxia telangiectasia mutated; CSR, class-switch recombination; DNA-PKcs, catalytic subunit of DNA-dependent protein kinase; EXO1, exonuclease 1; H2AX, histone 2A family, member X; MLH1, mutL homologue 1; MSH2, mutS homologue 2; NHEJ, non-homologous end-joining; PMS2, post-mitotic segregation 2; S, switch; SCID, severe combined immunodeficiency; SHM, somatic hypermutation; UNG, uracil-DNA glycosylase.
VOLUME 4 | JULY 2004 | 5 4 9
REVIEWS
The CSR phenotype of C-terminal AID mutants closely resembles that found in a group of patients with hyper IgM type IV syndrome. B cells from these patients are considerably impaired in CSR, but SHM and the accumulation of mutations in S are not affected100. Given that these cells express wild-type AID, it is possible that the mutation or mutations in these patients resides in genes that encode proteins required in the same reaction in which the putative CSR-specific AID cofactors participate. Therefore, the identification of proteins that interact with the C-terminus of AID and/or the cloning of the gene(s) that contribute to this immunodeficiency syndrome should shed light on the potential CSR-specific roles of AID downstream of deamination. Similarly, it could also be postulated that there will be mutations that could affect SHM but leave CSR intact: for example, mutations that affect the interaction of AID with factors, such as RPA, that target non-R-loop forming structures.
A plethora of unresolved issues
ability of AID to mediate CSR on artificial substrates, when it is ectopically expressed in fibroblasts, might be viewed as arguing against IgH-isotype-specific CSR factors40, the efficiency of CSR in these assays is low compared with endogenous efficiency. Therefore, it will be important to determine whether putative isotypespecific factors or other AID-specific cofactors exist and how they might function. Another important question centres on the current evidence indicating that CSR requires the participation of at least three different DNA-repair pathways: namely, BER, MMR and NHEJ. In this regard, the way in which pathways that use distinct sets of proteins are coordinated to mediate CSR is a mystery. Also, it is unknown how the BER pathway downstream of UNG and APE1 would be subverted from repairing an S-region nick before it is converted into a DSB. One possibility would be the introduction of numerous S-region lesions that overwhelm the BER machinery. Another possibility would be that BER activities downstream of UNG and APE1 might somehow be downregulated in cells undergoing CSR or at S lesions.
Perspective
Since the discovery of AID, our understanding of CSR has advanced considerably. We can now envision a model that mechanistically links DNA deamination by AID to long-standing observations on the unique primary structure of S regions and the requirement for germline transcription. However, in addition to the mechanism of S-region synapsis and the other outstanding questions outlined earlier, several additional key issues remain unresolved. It remains to be determined whether AID is specifically targeted to S regions or is passively recruited to any ssDNA. Studies using extrachromosomal CSR substrates indicate that AID targeting to S regions might be influenced by CSR-specific factors101,102. Whereas the
The discovery of AID and the subsequent delineation of its enzymatic activity have provided us with a model by which the initiation of CSR is effected. However, as discussed here, several aspects of CSR both upstream and downstream of AID remain unresolved, including the identity of potential factors that impart specificity to AID, the process that converts an initial AID-generated DNA lesion into DNA breaks and the mechanism that synapses broken S regions to allow completion of the CSR reaction. A combination of new in vivo-genetic manipulations and in vitro-biochemical studies will probably be required to address these issues.
1.
Bassing, C. H., Swat, W. & Alt, F. W. The mechanism and regulation of chromosomal V(D)J recombination. Cell 109, S45S55 (2002). 2. Jung, D. & Alt, F. W. Unraveling V(D)J recombination: insights into gene regulation. Cell 116, 299311 (2004). 3. Papavasiliou, F. N. & Schatz, D. G. Somatic hypermutation of immunoglobulin genes: merging mechanisms for genetic diversity. Cell 109, S35S44 (2002). 4. Martin, A. & Scharff, M. D. AID and mismatch repair in antibody diversification. Nature Rev. Immunol. 2, 605614 (2002). 5. Manis, J. P., Tian, M. & Alt, F. W. Mechanism and control of class-switch recombination. Trends Immunol. 23, 3139 (2002). 6. Honjo, T., Kinoshita, K. & Muramatsu, M. Molecular mechanism of class switch recombination: linkage with somatic hypermutation. Annu. Rev. Immunol. 20, 165196 (2002). 7. Storb, U. et al. Cis-acting sequences that affect somatic hypermutation of Ig genes. Immunol. Rev. 162, 153160 (1998). 8. Lansford, R., Manis, J. P., Sonoda, E., Rajewsky, K. & Alt, F. W. Ig heavy chain class switching in Rag-deficient mice. Int. Immunol. 10, 325332 (1998). 9. Dunnick, W., Hertz, G. Z., Scappino, L. & Gritzmacher, C. DNA sequences at immunoglobulin switch region recombination sites. Nucleic Acids Res. 21, 365372 (1993). 10. Shinkura, R. et al. The influence of transcriptional orientation on endogenous switch region function. Nature Immunol. 4, 435441 (2003). Provides the first evidence in vivo that the transcriptional orientation of S regions is important in CSR. In addition, the authors show that a
11.
12.
13. 14.
15. 16.
17.
18.
19.
synthetic G-rich sequence that lacks S-region motifs can replace an endogenous S region during CSR in vivo. Chaudhuri, J. et al. Transcription-targeted DNA deamination by the AID antibody diversification enzyme. Nature 422, 726730 (2003). Demonstrates biochemically that AID is a ssDNA deaminase. The authors also show a direct link between transcription and DNA deamination. Neuberger, M. S., Harris, R. S., Di Noia, J. & PetersenMahrt, S. K. Immunity through DNA deamination. Trends Biochem. Sci. 28, 305312 (2003). Stavnezer, J. Immunoglobulin class switching. Curr. Opin. Immunol. 8, 199205 (1996). Luby, T. M., Schrader, C. E., Stavnezer, J. & Selsing, E. The switch region tandem repeats are important, but not required, for antibody class switch recombination. J. Exp. Med. 193, 159168 (2001). Stavnezer, J. Antibody class switching. Adv. Immunol. 61, 79146 (1996). Stavnezer, J. Molecular processes that regulate class switching. Curr. Top. Microbiol. Immunol. 245, 127168 (2000). Yancopoulos, G. D. et al. Secondary genomic rearrangement events in pre-B cells: VHDJH replacement by a LINE-1 sequence and directed class switching. EMBO J. 5, 32593266 (1986). Stavnezer-Nordgren, J. & Sirlin, S. Specificity of immunoglobulin heavy chain switch correlates with activity of germline heavy chain genes prior to switching. EMBO J. 5, 95102 (1986). Reaban, M. E. & Griffin, J. A. Induction of RNA-stabilized DNA conformers by transcription of an immunoglobulin switch region. Nature 348, 342344 (1990).
20. Reaban, M. E., Lebowitz, J. & Griffin, J. A. Transcription induces the formation of a stable RNA. DNA hybrid in the immunoglobulin switch region. J. Biol. Chem. 269, 2185021857 (1994). 21. Daniels, G. A. & Lieber, M. R. Strand specificity in the transcriptional targeting of recombination at immunoglobulin switch sequences. Proc. Natl Acad. Sci. USA 92, 56255629 (1995). 22. Tian, M. & Alt, F. W. Transcription-induced cleavage of immunoglobulin switch regions by nucleotide excision repair nucleases in vitro. J. Biol. Chem. 275, 2416324172 (2000). 23. Daniels, G. A. & Lieber, M. R. RNA:DNA complex formation upon transcription of immunoglobulin switch regions: implications for the mechanism and regulation of class switch recombination. Nucleic Acids Res. 23, 50065011 (1995). 24. Mizuta, R. et al. Molecular visualization of immunoglobulin switch region RNA/DNA complex by atomic force microscope. J. Biol. Chem. 278, 44314434 (2003). 25. Yu, K., Chedin, F., Hsieh, C. L., Wilson, T. E. & Lieber, M. R. R-loops at immunoglobulin class switch regions in the chromosomes of stimulated B cells. Nature Immunol. 4, 442451 (2003). Uses chemical probes to demonstrate that S regions form R loops in vivo in B cells undergoing CSR. 26. Tashiro, J., Kinoshita, K. & Honjo, T. Palindromic but not G-rich sequences are targets of class switch recombination. Int. Immunol. 13, 495505 (2001). 27. Sen, D. & Gilbert, W. Formation of parallel fourstranded complexes by guanine-rich motifs in DNA and its implications for meiosis. Nature 334, 364366 (1988).
550
REVIEWS
28. Dempsey, L. A., Sun, H., Hanakahi, L. A. & Maizels, N. G4 DNA binding by LR1 and its subunits, nucleolin and hnRNP D, a role for GG pairing in immunoglobulin switch recombination. J. Biol. Chem. 274, 10661071 (1999). 29. Ratmeyer, L., Vinayak, R., Zhong, Y. Y., Zon, G. & Wilson, W. D. Sequence specific thermodynamic and structural properties for DNA.RNA duplexes. Biochemistry 33, 52985304 (1994). 30. Kinoshita, K., Tashiro, J., Tomita, S., Lee, C. G. & Honjo, T. Target specificity of immunoglobulin class switch recombination is not determined by nucleotide sequences of S regions. Immunity 9, 849858 (1998). 31. Mussmann, R., Courtet, M., Schwager, J. & Du Pasquier, L. Microsites for immunoglobulin switch recombination breakpoints from Xenopus to mammals. Eur. J. Immunol. 27, 26102619 (1997). 32. Ramiro, A. R., Stavropoulos, P., Jankovic, M. & Nussenzweig, M. C. Transcription enhances AID-mediated cytidine deamination by exposing single-stranded DNA on the nontemplate strand. Nature Immunol. 4, 452456 (2003). Shows that AID targets the non-template strand of DNA when transcription is carried out in bacteria. 33. Sohail, A., Klapacz, J., Samaranayake, M., Ullah, A. & Bhagwat, A. S. Human activation-induced cytidine deaminase causes transcription-dependent, strand-biased C to U deaminations. Nucleic Acids Res. 31, 29902994 (2003). 34. Pham, P., Bransteitter, R., Petruska, J. & Goodman, M. F. Processive AID-catalysed cytosine deamination on singlestranded DNA simulates somatic hypermutation. Nature 424, 103107 (2003). Shows that AID preferably deaminates cytidine residues within SHM motifs in an ordered fashion. 35. Muramatsu, M. et al. Specific expression of activationinduced cytidine deaminase (AID), a novel member of the RNA-editing deaminase family in germinal center B cells. J. Biol. Chem. 274, 1847018476 (1999). 36. Muramatsu, M. et al. Class switch recombination and hypermutation require activation-induced cytidine deaminase (AID), a potential RNA editing enzyme. Cell 102, 553563 (2000). 37. Revy, P. et al. Activation-induced cytidine deaminase (AID) deficiency causes the autosomal recessive form of the Hyper-IgM syndrome (HIGM2). Cell 102, 565575 (2000). 38. Yoshikawa, K. et al. AID enzyme-induced hypermutation in an actively transcribed gene in fibroblasts. Science 296, 20332036 (2002). 39. Martin, A. & Scharff, M. D. Somatic hypermutation of the AID transgene in B and non-B cells. Proc. Natl Acad. Sci. USA 99, 1230412308 (2002). 40. Okazaki, I. M., Kinoshita, K., Muramatsu, M., Yoshikawa, K. & Honjo, T. The AID enzyme induces class switch recombination in fibroblasts. Nature 416, 340345 (2002). References 38 to 40 provide evidence that AID is probably the only activated B-cell factor required for SHM and CSR. 41. Doi, T., Kinoshita, K., Ikegawa, M., Muramatsu, M. & Honjo, T. De novo protein synthesis is required for the activationinduced cytidine deaminase function in class-switch recombination. Proc. Natl Acad. Sci. USA 100, 26342638 (2003). 42. Harris, R. S., Petersen-Mahrt, S. K. & Neuberger, M. S. RNA editing enzyme APOBEC1 and some of its homologs can act as DNA mutators. Mol. Cell 10, 12471253 (2002). Shows that other APOBEC1-family members have DNA-deamination activity, indicating that the RNAediting enzyme APOBEC1 is probably an outlier in this family of proteins. 43. Petersen-Mahrt, S. K., Harris, R. S. & Neuberger, M. S. AID mutates E. coli suggesting a DNA deamination mechanism for antibody diversification. Nature 418, 99103 (2002). The first paper to present the DNA-deamination model for CSR and SHM. In addition, it provides evidence that the target of AID is not RNA but DNA. 44. Bransteitter, R., Pham, P., Scharff, M. D. & Goodman, M. F. Activation-induced cytidine deaminase deaminates deoxycytidine on single-stranded DNA but requires the action of RNase. Proc. Natl Acad. Sci. USA 100, 41024107 (2003). 45. Dickerson, S. K., Market, E., Besmer, E. & Papavasiliou, F. N. AID mediates hypermutation by deaminating single stranded DNA. J. Exp. Med. 197, 12911296 (2003). 46. Nambu, Y. et al. Transcription-coupled events associating with immunoglobulin switch region chromatin. Science 302, 21372140 (2003). Shows that AID binds to chromatin that is associated with transcribed S regions, thereby providing further evidence that AID acts directly on DNA. 47. Wuerffel, R. A., Du, J., Thompson, R. J. & Kenter, A. L. Ig S3 DNA-specific double strand breaks are induced in mitogen-activated B cells and are implicated in switch recombination. J. Immunol. 159, 41394144 (1997). 48. Catalan, N. et al. The block in immunoglobulin class switch recombination caused by activation-induced cytidine deaminase deficiency occurs prior to the generation of DNA double strand breaks in switch region. J. Immunol. 171, 25042509 (2003). 49. Rogakou, E. P., Pilch, D. R., Orr, A. H., Ivanova, V. S. & Bonner, W. M. DNA double-stranded breaks induce histone H2AX phosphorylation on serine 139. J. Biol. Chem. 273, 58585868 (1998). 50. Petersen, S. et al. AID is required to initiate Nbs1/-H2AX focus formation and mutations at sites of class switching. Nature 414, 660665 (2001). Shows that H2AX has a role in CSR and that the role of AID in CSR is probably upstream of S-region breaks. 51. Bassing, C. H. et al. Histone H2AX: a dosage-dependent suppressor of oncogenic translocations and tumors. Cell 114, 359370 (2003). Reports that haploinsufficiency of H2AX can lead to transclocations that involve S regions. 52. Manis, J. P. et al. 53BP1 links DNA damage response to immunoglobulin class switch recombination. Nature Immunol. 5, 481487 (2004). Describes the effect of 53BP1 mutation in CSR and SHM. 53. Casellas, R. et al. Ku80 is required for immunoglobulin isotype switching. EMBO J. 17, 24042411 (1998). 54. Manis, J. P. et al. Ku70 is required for late B cell development and immunoglobulin heavy chain class switching. J. Exp. Med. 187, 20812089 (1998). 55. Manis, J. P., Dudley, D., Kaylor, L. & Alt, F. W. IgH class switch recombination to IgG1 in DNA-PKcs-deficient B Cells. Immunity 16, 607617 (2002). Together with references 93 and 94, this paper indicates that the role of DNA-PKcs in CSR might extend beyond the role it has in NHEJ. 56. Bross, L., Muramatsu, M., Kinoshita, K., Honjo, T. & Jacobs, H. DNA double-strand breaks: prior to but not sufficient in targeting hypermutation. J. Exp. Med. 195, 11871192 (2002). 57. Papavasiliou, F. et al. V(D)J recombination in mature B cells: a mechanism for altering antibody responses. Science 278, 298301 (1997). 58. Faili, A. et al. AID-dependent somatic hypermutation occurs as a DNA single-strand event in the BL2 cell line. Nature Immunol. 3, 815821 (2002). 59. Woo, C. J., Martin, A. & Scharff, M. D. Induction of somatic hypermutation is associated with modifications in immunoglobulin variable region chromatin. Immunity 19, 479489 (2003). 60. Rada, C. et al. Immunoglobulin isotype switching is inhibited and somatic hypermutation perturbed in UNG-deficient mice. Curr. Biol. 12, 17481755 (2002). 61. Imai, K. et al. Human uracil-DNA glycosylase deficiency associated with profoundly impaired immunoglobulin classswitch recombination. Nature Immunol. 4, 10231028 (2003). References 60 and 61 show that UNG has a crucial role in CSR and thereby provide evidence for a DNAdeamination step in CSR. 62. Vora, K. A. et al. Severe attenuation of the B cell immune response in Msh2-deficient mice. J. Exp. Med. 189, 471482 (1999). 63. Ehrenstein, M. R. & Neuberger, M. S. Deficiency in Msh2 affects the efficiency and local sequence specificity of immunoglobulin class-switch recombination: parallels with somatic hypermutation. EMBO J. 18, 34843490 (1999). 64. Schrader, C. E., Edelmann, W., Kucherlapati, R. & Stavnezer, J. Reduced isotype switching in splenic B cells from mice deficient in mismatch repair enzymes. J. Exp. Med. 190, 323330 (1999). 65. Yu, K. & Lieber, M. R. Nucleic acid structures and enzymes in the immunoglobulin class switch recombination mechanism. DNA Repair (Amst.) 2, 11631174 (2003). 66. Rush, J. S., Fugmann, S. D. & Schatz, D. G. Staggered AIDdependent DNA double strand breaks are the predominant DNA lesions targeted to S in immunoglobulin class switch recombination. Int. Immunol. 16, 549557 (2004). 67. Faili, A. et al. DNA Polymerase is involved in hypermutation occurring during immunoglobulin class switch recombination. J. Exp. Med. 199, 265270 (2004). 68. Schrader, C. E., Vardo, J. & Stavnezer, J. Role for mismatch repair proteins Msh2, Mlh1, and Pms2 in immunoglobulin class switching shown by sequence analysis of recombination junctions. J. Exp. Med. 195, 367373 (2002). 69. Min, I. M. et al. The S tandem repeat region is critical for Ig isotype switching in the absence of Msh2. Immunity 19, 515524 (2003). 70. Paques, F. & Haber, J. E. Multiple pathways of recombination induced by double-strand breaks in Saccharomyces cerevisiae. Microbiol. Mol. Biol. Rev. 63, 349404 (1999). 71. Tian, M., Jones, A. M., Smith, M., Shinkura, R. & Alt, F. W. Deficiency in nuclease activity of xeroderma pigmentosum G in mice leads to hypersensitivity to UV radiation. Mol. Cell. Biol. 24, 22372242 (2004). 72. Tian, M., Shinkura, R., Shinkura, N. & Alt, F. W. Growth retardation, early death, and DNA repair defects in mice deficient for the nucleotide excision repair enzyme XPF. Mol. Cell. Biol. 24, 12001205 (2004). 73. Hodgkin, P. D., Lee, J. H. & Lyons, A. B. B cell differentiation and isotype switching is related to division cycle number. J. Exp. Med. 184, 277281 (1996). 74. Kuzminov, A. Single-strand interruptions in replicating chromosomes cause double-strand breaks. Proc. Natl Acad. Sci. USA 98, 82418246 (2001). 75. Dudley, D. D. et al. Internal IgH class switch region deletions are position-independent and enhanced by AID expression. Proc. Natl Acad. Sci. USA 99, 99849989 (2002). Shows that S-region internal deletions are AID dependent and provides evidence that synapsis between two S regions is not required for targeting AID. 76. Alt, F. W., Rosenberg, N., Casanova, R. J., Thomas, E. & Baltimore, D. Immunoglobulin heavy-chain expression and class switching in a murine leukaemia cell line. Nature 296, 325331 (1982). 77. Gu, H., Zou, Y. R. & Rajewsky, K. Independent control of immunoglobulin switch recombination at individual switch regions evidenced through CreloxP-mediated gene targeting. Cell 73, 11551164 (1993). 78. Schrader, C. E. et al. Mutations occur in the Ig S region but rarely in S regions prior to class switch recombination. EMBO J. 22, 58935903 (2003). 79. Reina-San-Martin, B. et al. H2AX is required for recombination between immunoglobulin switch regions but not for intra-switch region recombination or somatic hypermutation. J. Exp. Med. 197, 17671778 (2003). Indicates that H2AX might have a role in the synapsis of two S regions during CSR. 80. Bassing, C. H. & Alt, F. W. H2AX may function as an anchor to hold broken chromosomal DNA ends in close proximity. Cell Cycle 3, 149153 (2004). 81. Shiloh, Y. Ataxia-telangiectasia and the Nijmegen breakage syndrome: related disorders but genes apart. Annu. Rev. Genet. 31, 635662 (1997). 82. Pan, Q. et al. Alternative end joining during switch recombination in patients with ataxia-telangiectasia. Eur. J. Immunol. 32, 13001308 (2002). 83. Pan-Hammarstrom, Q. et al. ATM is not required in somatic hypermutation of VH, but is involved in the introduction of mutations in the switch region. J. Immunol. 170, 37073716 (2003). 84. Burma, S., Chen, B. P., Murphy, M., Kurimasa, A. & Chen, D. J. ATM phosphorylates histone H2AX in response to DNA double-strand breaks. J. Biol. Chem. 276, 4246242467 (2001). 85. Stiff, T. et al. ATM and DNA-PK function redundantly to phosphorylate H2AX after exposure to ionizing radiation. Cancer Res. 64, 23902396 (2004). 86. Ward, I. M. & Chen, J. Histone H2AX is phosphorylated in an ATR-dependent manner in response to replicational stress. J. Biol. Chem. 276, 4775947762 (2001). 87. Xu, Y. et al. Targeted disruption of ATM leads to growth retardation, chromosomal fragmentation during meiosis, immune defects, and thymic lymphoma. Genes Dev. 10, 24112422 (1996). 88. DeFazio, L. G., Stansel, R. M., Griffith, J. D. & Chu, G. Synapsis of DNA ends by DNA-dependent protein kinase. EMBO J. 21, 31923200 (2002). 89. Ehrenstein, M. R., Rada, C., Jones, A. M., Milstein, C. & Neuberger, M. S. Switch junction sequences in PMS2deficient mice reveal a microhomology-mediated mechanism of Ig class switch recombination. Proc. Natl Acad. Sci. USA 98, 1455314558 (2001). 90. Hall, M. C., Wang, H., Erie, D. A. & Kunkel, T. A. High affinity cooperative DNA binding by the yeast Mlh1Pms1 heterodimer. J. Mol. Biol. 312, 637647 (2001). 91. Schrader, C. E., Vardo, J. & Stavnezer, J. Mlh1 can function in antibody class switch recombination independently of Msh2. J. Exp. Med. 197, 13771383 (2003). 92. Brown, K. D. et al. The mismatch repair system is required for S-phase checkpoint activation. Nature Genet. 33, 8084 (2003). 93. Bosma, G. C. et al. DNA-dependent protein kinase activity is not required for immunoglobulin class switching. J. Exp. Med. 196, 14831495 (2002).
VOLUME 4 | JULY 2004 | 5 5 1
REVIEWS
94. Cook, A. J. et al. Reduced switching in SCID B cells is associated with altered somatic mutation of recombined S regions. J. Immunol. 171, 65566564 (2003). 95. Ma, Y., Pannicke, U., Schwarz, K. & Lieber, M. R. Hairpin opening and overhang processing by an Artemis/DNAdependent protein kinase complex in nonhomologous end joining and V(D)J recombination. Cell 108, 781794 (2002). 96. Sekiguchi, J. et al. Genetic interactions between ATM and the nonhomologous end-joining factors in genomic stability and development. Proc. Natl Acad. Sci. USA 98, 32433248 (2001). 97. Ta, V. T. et al. AID mutant analyses indicate requirement for class-switch-specific cofactors. Nature Immunol. 4, 843848 (2003). 98. Barreto, V., Reina-San-Martin, B., Ramiro, A. R., McBride, K. M. & Nussenzweig, M. C. C-terminal deletion of AID uncouples class switch recombination from somatic hypermutation and gene conversion. Mol. Cell 12, 501508 (2003). References 97 and 98 provide evidence that the carboxyl terminus of AID could have specific roles in CSR that might be downstream of DNA deamination. 99. Bemark, M. et al. Somatic Hypermutation in the absence of DNA-dependent protein kinase catalytic subunit (DNAPKcs) or recombination-activating gene (RAG) 1 activity. J. Exp. Med. 192, 15091514 (2000). Imai, K. et al. Hyper-IgM syndrome type 4 with a B lymphocyte-intrinsic selective deficiency in Ig classswitch recombination. J. Clin. Invest. 112, 136142 (2003). Reports on a newly characterized human immunodeficiency, in which the defect in CSR is probably downstream of AID-mediated DNA deamination. Shanmugam, A., Shi, M. J., Yauch, L., Stavnezer, J. & Kenter, A. L. Evidence for class-specific factors in immunoglobulin isotype switching. J. Exp. Med. 191, 13651380 (2000). Ma, L., Wortis, H. H. & Kenter, A. L. Two new isotypespecific switching activities detected for Ig class switching. J. Immunol. 168, 28352846 (2002). Ito, S. et al. Activation-induced cytidine deaminase shuttles between nucleus and cytoplasm like apolipoprotein B mRNA editing catalytic polypeptide 1. Proc. Natl Acad. Sci. USA 101, 19751980 (2004). Martin, A. et al. Msh2 ATPase activity is essential for somatic hypermutation at AT basepairs and for efficient class switch recombination. J. Exp. Med. 198, 11711178 (2003). 105. Bardwell, P. D. et al. Altered somatic hypermutation and reduced class-switch recombination in exonuclease 1-mutant mice. Nature Immunol. 5, 224229 (2004).
100.
Acknowledgements
We thank John Manis, Ming Tian, Ali Zarrin, Sheila Ranganath, Craig Bassing and Chan Khuong and other members of the laboratory for careful evaluation of this manuscript. Work on CSR in the Alt laboratory is supported in part by the National Institutes of Health. F.W.A. is an Investigator of the Howard Hughes Medical Institute.
101.

102.
Online links
DATABASES The following terms in this article are linked online to: Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene 53BP1 | AID | APE1 | APOBEC1 | ARTEMIS | ATM | DNA Ligase IV | DNA-PKcs | ERCC1 | H2AX | MLH1 | MSH2 | PMS2 | RAG1 | RAG2 | UNG | XPF | XPG | XRCC4 Access to this interactive links box is free online.
103.
104.
552
REVIEWS
NEONATAL ADAPTIVE IMMUNITY COMES OF AGE

Becky Adkins*, Claude Leclerc and Stuart Marshall-Clarke
The past decade has brought great strides in our understanding of adaptive immunity in neonatal mice. Although poor immune responses are commonly observed, it is now clear that mature function can be achieved by all arms of the adaptive immune system. An everincreasing body of evidence indicates that the neonatal period of life is a unique developmental stage in which responses are highly plastic and dependent on the conditions of antigen exposure. This review focuses on our new understanding in mice and, where it is clear that related phenomena occur, in humans.
Neonatal life is characterized by heightened sensitivity to infectious agents. One particularly clear example of this is infection with HIV. In the absence of retroviral therapy, there is a more rapid progression to disease in paediatric infection and this is associated with severely reduced HIV-specific cellular and humoral immune responses (reviewed in REFS 1,2). The sensitivity of newborns to infectious diseases might be partly due to the lack of pre-existing immunological memory in newborns. Another important contributing factor might be the small number of immune cells that are present in peripheral lymphoid tissues in early life, especially in mice. However, aside from these quantitative differences, many studies in both humans and mice have shown that newborn immune cells are qualitatively distinct from adult cells. Subsets of cells are present in different proportions in neonates and adults and, among cells of the same subtype, phenotypic differences have been described. In addition, many in vivo and in vitro studies have described deficiencies or immune deviation among T cells, B cells and antigen-presenting cells (APCs) in neonates. Historically, the function of neonatal adaptive immune cells has been considered to be immature. However, in 1996, three seminal papers35 changed our previous understanding by showing that neonates are competent, under certain circumstances, to mount adult-level T-cell responses in vivo (FIG. 1). Since then, it has become increasingly clear that neonatal adaptive responses show a great deal of variability, ranging from non-responsiveness to fully mature function. As we discuss, these findings are mainly in mice, but there are some indications that a similar situation is true in humans in early life, although this needs to be investigated further.
T-cell immunity
TH1/TH2-CELL POPULATIONS
At least two distinct subsets of activated CD4+ T cells have been described. T helper 1 (TH1) cells produce interferon-, lymphotoxin and tumournecrosis factor, and support cellmediated immunity. TH2 cells produce interleukin-4 (IL-4), IL-5 and IL-13, support humoral immunity and downregulate TH1-cell responses.
*Department of Microbiology and Immunology, University of Miami Medical School, Miami, Florida 33136, USA. Unit of Biology of Immune Regulation, Institut Pasteur, Paris 75015, France. Department of Human Anatomy and Cell Biology, University of Liverpool, Liverpool L69 3GE, UK. Correspondence to B.A. e-mail: radkins@med.miami.edu
doi:10.1038/nri1394
CD4+ T helper 1 (TH1)- versus TH2-cell polarization. Until about a decade ago, neonates were referred to as immunodeficient. This concept was due, in part, to limited interleukin-2 (IL-2) production and proliferation by neonatal T cells. However, with the discovery of T 1/T 2-CELL POPULATIONS, it became clear that T-cell responses in NEONATAL MICE were not deficient but were biased to the TH2-cell lineage. Notably, mice initially immunized as neonates mounted TH2-type-dominant memory responses when re-exposed to the same antigens as adults. However, with the identification of agents that promote strong TH1-cell responses, such as DNA vaccines or oligonucleotides containing CPG MOTIFS, many of these substances were shown to elicit adultlike TH1-cell responses in neonates. The many successes in promoting mature neonatal CD4+ TH1-type primary and memory responses in mice have been described and reviewed in detail elsewhere68. It is important to emphasize that the clear skewing towards TH2-type responses that can be observed in mice is not readily apparent in humans. In human neonates, TH-cell responses are often diminished in magnitude, but this
H H
VOLUME 4 | JULY 2004 | 5 5 3
REVIEWS
can affect TH1- or TH2-cell function or both. For example, in malaria infection, both the TH1- and TH2-type antigen-specific responses of children less than 5 years old were reduced, compared with infected adults9,10. Nonetheless, human and mouse neonates are similar in that adult-level TH1-cell responses can also develop in human infants, as demonstrated by responses to Mycobacterium bovis bacillus CalmetteGuerin (BCG)11,12 (see later). So, under some circumstances, human neonates also seem to be competent to develop mature TH-cell responses. Although it is clear that adult-like TH1-cell function can be achieved in situ, TH1-cell responses are commonly limited in magnitude or partially biased to TH2-cell function68. What are the potential mechanisms that contribute to these variant responses? Immature responses might be at least partially due to quantitative differences in T cells. In neonatal mice, the number of T cells is reduced by a factor of 12 logs, compared with adult animals4,6. When challenged with the same dose of an infectious agent, neonates might have a more limited protective response simply due to the small numbers of immune cells. Also, in the case of acute infection, the quality of the response might be affected by antigen load. This is illustrated by the studies of Sarzotti et al.3, in which decreasing the dose of a mouse leukaemia virus (CasBr-M) led to a TH2 to TH1 deviation and protection from disease (FIG. 1b). Differences between neonatal and adult responses might also be contributed to by differences in the composition of the T-cell compartment (FIG. 2a). In neonatal mice, there is an overlap in the generation of T cells by fetal and adult haematopoietic precursors1315. Consequently, peripheral T cells in the first few weeks of life are a mixture of fetal- and adult-derived T cells. It has recently been shown that mouse CD4+ T cells of fetal origin mount TH2-skewed responses in an antigen-dose-dependent manner16. Interestingly, there also seems to be a fetal influence in human neonatal responses, as Holt and colleagues1719 have shown that responses that develop in utero to common environmental allergens are universally skewed towards TH2-type. So, the fetal origin of some T cells in neonates might have an important role in T-cell responses in early life (FIG. 2a). In intact neonates, it is not possible to distinguish whether variant responses are due to T-cell-intrinsic properties or to regulation by non-T cells and/or the endogenous cytokine milieu. In mice, this can be tested experimentally by comparing neonatal and adult T-cell function in adoptive transfer models, in which the background environment is constant for both cell types. We recently reported that neonatal (day 7) CD4+ lymph-node cells retain their propensity to develop efficient antigen-specific TH2-type responses and are highly deficient in the development of TH1-type responses when transferred to adult hosts20. By contrast, Garvy and colleagues21 reported that, although 10-day-old mice are highly sensitive to Pneumocystis carinii-induced pneumonia, splenic T cells from these mice transferred to an adult environment developed
a Injection of donor cells
T cells
Monocytes Tolerance to male antigen
B cells
Dendritic cells No CTLs
Injection of donor total adult male spleen cells
Female neonatal mouse
Priming to male antigen (adult-like response)
Injection of donor adult male dendritic cells Female neonatal mouse Mature CTLs
b Infection with leukaemia virus
TH2 cells Neonatal mouse Adult-level dose Good TH2-cell response Poor CTL response No protection from disease
TH1 cells Reduced dose Neonatal mouse
CTLs
Adult-like response Mature TH1-cell response Protective CTL response
c Immunization with antigen

Classic tolerance protocol: antigen in IFA
TH2-cell-polarized adult-like response TH2 cells TH1-cell-polarized adult-like response TH1 cells
Neonatal mouse
TH1-cell-promoting conditions: antigen in CFA
Figure 1 | Pivotal demonstrations that neonatal mice are competent to mount mature adaptive immune responses in vivo. An important turning point in the study of neonatal immunology occurred in 1996, when three laboratories reported mature immune responses in newborn mice. a | As opposed to the tolerance seen with injection of total male spleen cells, Matzinger and colleagues4 achieved mature cytotoxic T lymphocyte (CTL) responses against H-Y antigen in female neonatal mice by injection of adult male dendritic cells. b | Sarzotti and colleagues3 reduced the infecting dose of virus, leading to protection from disease and accompanying mature T helper 1 (TH1)-cell and CTL responses. c | Lehmann and colleagues5 showed that, when immunized with protein in incomplete Freunds adjuvant (IFA), neonatal mice mounted TH2-cell responses only. By contrast, the strong TH1-cell-promoting adjuvant complete Freunds adjuvant (CFA) elicited TH1-cell-polarized responses that were similar to those of adult mice.
554
REVIEWS
mature TH1-type functions that were sufficient to resolve the pneumonia. One interpretation of these seemingly disparate results is that whether neonatal T cells behave differently from adult T cells in vivo might depend on the local environment (such as lung versus peripheral lymphoid tissue). A recent study by Zaghouani and colleagues22 took the opposite approach to examine the TH-cell responses of neonatal (day 1) and adult T-cell receptor (TCR)-transgenic CD4+ cells in adoptive neonatal hosts. They found that neonatal, but not adult, TH1 cells underwent
a Contribution of fetus-derived cells
Peripheral T cells in early life
apoptosis in response to re-exposure to antigen (FIG. 2b). As neonatal T H2 cells were not similarly affected, these results provide a potential mechanism for the TH2-biased secondary responses that are often seen in neonatal mice. In summary, both quantitative and qualitative properties of neonatal CD4+ T cells are probably factors in the plasticity of neonatal TH1/TH2-cell responses. Given the central role of CD4+ TH cells in influencing the responses of many other immune cell types, variation in the TH-cell responses of neonates might contribute substantially to variability in the responses of other cell populations (see later). CD8+ cytotoxic T lymphocytes. As for CD4+ T-cell responses, neonates were long thought to be deficient in CD8+ CTL function. For example, tolerance of mouse neonates to semi-allogeneic (F1) cells is characterized by the development of alloantigen-specific TH2 cells and the failure of development of specific CTL activity2329. However, concomitant treatment of neonates with a neutralizing antibody specific for IL-4 (REFS 23,30) or with the T H1-cell-promoting agents CD40-specific antibody31 or IL-12 (REF. 32) prevents tolerance induction. These treatment regimens led to the development of marked cytotoxicity, decreased TH2-cell activity and increased T H1-cell function. Strong primary and/or memory CTL functions have also been elicited in neonatal mice by administering live replicating vaccines3,3335, adult dendritic cells (DCs)4 or other strong TH1-cell-promoting agents, such as DNA vaccines3640 or oligonucleotides containing CpG motifs41,42. In humans, strong CTL function has been detected in infants congenitally infected with cytomegalovirus43 or Trypanosoma cruzi44, indicating that, even as early as fetal life, developing humans are competent to develop potent CTL activity under some conditions. More recently, CTL function in neonatal mice has been examined in greater detail to address the important issue of timing. In earlier studies, immune responses were often measured many weeks or months after the initial antigen exposure. As a result, it was not clear whether the observed responses were due to priming during the actual neonatal period. In neonatal mice, two recent studies have analysed CTL responses within two weeks of antigen exposure. Using DNA vaccines, Hassett and colleagues45 found that antigenspecific CD8+ CTLs developing within this early period of life were present at frequencies comparable to those in adults, and had similar levels of effector function and avidity for antigen. Similar observations were reported by Siegrist and colleagues46 in response to a non-replicative virus-like particle. Moreover, it was recently shown that human cord blood cells adoptively transferred to SEVERE COMBINED IMMUNODEFICIENT (SCID) hosts mounted adult-level allospecific cytotoxic responses within one week of antigen exposure47. So, CD8+ T cells in both human and mouse neonates can develop mature CTL functions within the neonatal period of life in certain circumstances.
Fetal mouse
Fetal T cells
Neonatal T cells
Neonatal mouse
Potential source of TH2-cell function in early life
b Contribution of selective apoptosis

Antigen re-exposure Proliferation TH2 cell Primary effector-cell development Naive neonatal T cell TH1 cell Antigen re-exposure
IL-4R- or IL-13Rspecific antibodies Apoptosis
c Contribution of homeostatic proliferation

c cytokines (such as IL-7 and IL-15)
CD4+ TH cell Effector function in the absence of exogenous antigen?
Naive neonatal T cell in lymphopenic environment
Homeostatic proliferation ?
CD8+ cytotoxic T cell
Figure 2 | Unique properties of the neonatal T-cell compartment might contribute to plasticity of responses in early life. a | In early life, some peripheral T cells are derived from fetal thymic precursors. These fetus-derived cells might make important contributions to responses in early life, as fetus-derived cells preferentially mount T helper 2 (TH2)-cell responses16,17. b | Transfer of neonatal T-cell receptor (TCR)-transgenic CD4+ T cells into adoptive neonatal hosts leads to the development of both TH1- and TH2-cell primary effector function after immunization. However, after re-exposure to antigen in vivo, neonatal TH1, but not TH2, cells undergo marked apoptosis. This apoptosis can be inhibited by the in vivo administration of interleukin-4 receptor (IL-4R)- or IL-13Rspecific blocking antibodies22. c | There is increasing evidence that neonatal T cells undergo homeostatic proliferation in situ due to the lymphopenic environment of the neonate53,54,56. This might be due, in part, to their sensitivity to common -chain (c)-containing cytokines in the absence of antigenic stimulation. This process might lead to the development of effector function in neonates, in the absence of exogenous antigen administration.
VOLUME 4 | JULY 2004 | 5 5 5
REVIEWS
neonatal, than adult, T cells that are cycling spontaneously. Interestingly, it has recently been shown in humans that, unlike naive adult T cells, cord-blood CD4+ and CD8+ T cells undergo marked proliferation and acquire TH1/TH2-type cytokine secretion in vitro in response to common -chain cytokines, such as IL-7 and IL-15, alone5963 (FIG. 2c). So, perhaps due to higher levels of expression of cytokine receptors57,59,62,63, neonatal T cells might be especially sensitive to homeostatic signals, leading to clonal expansion and possible effectorcell development in the absence of specific antigen exposure. How this process might contribute to the variability of neonatal T-cell responses is an important question in the field. For example, if neonatal T cells, through a homeostatic proliferation process, do indeed develop into interferon- (IFN-)-secreting effectors, why are TH1-cell responses often reduced in neonates? The answer remains to be found, but possibilities could include a heightened susceptibility of both mouse and human neonatal T cells59,64, and TH1 cells in particular22, to apoptosis.
B-cell immunity
Box 1 | Maternal antibody effects In humans and mice, antibodies that are transferred to the neonate from its mother through the placenta and through milk provide protection during the postnatal maturation of the immune system. The effect of these maternal antibodies on early life responses has been widely studied and reviewed elsewhere66,134. These effects are not a complicating factor in mouse studies because maternal antibody levels for standard immunogens are negligible. They are, however, an important consideration when designing human immunization strategies. Although the data are contradictory, probably because of variable maternal antibody levels, most reports indicate that the presence of maternal antibodies further limits the intrinsic weakness of infant primary antibody responses to both natural infections and vaccination, most probably by antigen removal135,136. By contrast, T-cell responses seem to be unaffected by maternal antibodies in both humans137 and mice138. In addition, although maternal antibodies affect primary responses, infants immunized in the presence of high levels of passive antibody were nonetheless effectively primed for recall responses to a subsequent challenge, possibly reflecting the generation of memory T cells on first contact with antigen136,139.
NEONATAL MICE
Newborn mice are defined experimentally as those ranging in age from 1 to 10 days. Although there are major changes in lymphoid-organ development during the first week of life, at least some adaptive immune responses are similar in 1-day-old and 7-dayold mice. Nonetheless, mice are immunologically less mature at birth than humans and it has been proposed that 7-day-old mice are most comparable to human newborns.
CPG MOTIFS
DNA oligodeoxynucleotide sequences that include a cytosineguanosine sequence and certain flanking nucleotides, which have been found to induce innate immune responses through interaction with Toll-like receptor 9.
SEVERE COMBINED IMMUNODEFICIENT
(SCID). Mice of this phenotype lack functional T and B cells, owing to a spontaneous mutation in the Prkdc gene (protein kinase, DNA activated, catalytic polypeptide) located on chromosome 16. These mice can be reconstituted with T-cell subsets and used to study T-cell functions in vivo.
REGULATORY T CELLS
Regulatory T cells. In adults, the importance of REGULAcells) is becoming increasingly apparent Reg in all aspects of immunity. At present, we are lacking any information on the comparative properties of neonatal and adult TReg cells. Nonetheless, it is important to point out that TReg cells also seem to have central functions in early life. For example, it has long been known that neonatal thymectomy in mice leads to an increased incidence of autoimmune disease48. This autoimmunity is prevented by the infusion of adult CD4+CD25+ TReg cells49. In mice, cells with a CD4+CD25+ TReg-cell phenotype have been identified in peripheral lymph nodes within several days of birth49. It seems probable that neonatal thymectomy results in insufficient numbers of these cells and, therefore, the subsequent development of autoimmunity. CD4+CD25+ TReg cells have also been implicated in the development of neonatal tolerance to transplantation antigens. Field and colleagues50 showed that CD4+CD25+ T cells from neonatally tolerized mice suppressed the development of donor-specific CD8+ T-cell responses in vitro. In addition to CD4+ TReg cells, recent reports implicate key roles for CD8+ TReg cells in neonatal life. CD8+ TReg cells have been found to downregulate TH2-cell-mediated pathology in a neonatal transplantation model in mice51, as well as TH2-cell-mediated autoimmunity in neonatal rats52. So, emerging evidence indicates that both CD4+ and CD8+ TReg cells have key roles in shaping the developing immune system.
TORY T CELLS (T
(TReg cells). Naturally occurring TReg cells contribute to the maintenance of immunological self-tolerance and modulate the reactivity of other T cells. They constitute a network of heterogeneous CD4+ T-cell subsets, CD8+ T cells and other minor T-cell populations.
Homeostatic T-cell proliferation in neonates. Due to the small numbers of lymphoid cells in mouse neonates, the environment can be characterized as lymphopenic. Indeed, recent studies5355 have shown that both CD4+ and CD8+ adult T cells transferred to 1-day-old neonatal mice proliferate and acquire TH1-effector-cell function (FIG. 2c), similar to the homeostatic proliferation observed in lymphopenic adult hosts. Le Campion et al.56 showed that a fraction of endogenous neonatal T cells in mice are actively proliferating in vivo during the first week of life. This observation is supported by additional findings in humans57 and mice58 that there are proportionally more
In both mice and humans, many studies have shown that the primary T-CELL-DEPENDENT ANTIBODY RESPONSES induced in the neonatal period differ from adult responses (reviewed in REFS 65,66). Neonatal antibody responses are delayed in onset, reach lower peak levels, are of shorter duration, differ in the distribution of IgG isotypes (with neonates showing lower IgG2a (mice) and IgG2 (humans) than adults), and are of lower average affinity and reduced heterogeneity. T-CELL-INDEPENDENT ANTIBODY RESPONSES to TI-2 type antigens (including bacterial polysaccharides) are also deficient in both species. Reduced antibody responses, especially in human infants, might be partially caused by the presence of maternal antibodies (BOX 1). Nonetheless, there are some data in mice showing that, under certain circumstances, high-affinity antibodies67 and antibodies with somatically mutated variable (V) regions68 can be generated after neonatal immunization. The deficiencies that are normally encountered in neonatal antibody production could result from underlying differences between neonatal B cells and those of adults or be a consequence of immaturity in TH cells (see previous section). In mice, phenotypic and functional differences between neonatal and adult B cells have long been reported. Immature B cells (IgM+IgDlow/) comprise the main B-cell population in the neonatal spleen. In contrast to mature, adult B cells, immature B cells are negatively signalled by ligation of the B-cell receptor (BCR) and fail to upregulate costimulatory molecules (such as CD86) and MHC class II molecules that are essential for effective interaction with T cells6971. However, there is reason to think that examination of splenic B cells in neonatal mice, although revealing interesting differences, might not give us the whole picture. The neonatal spleen is a haematopoietic environment, where, during the first weeks of life, production of B cells from the fetal wave of B-cell lymphopoiesis crosses over with their generation
556
REVIEWS
Adult mice
Mature IgD+ B cell B-cell homing
a
V(D)J recombination Immature IgM+ B cell
b
Lymphoid follicle Marginalzone B cell
d
Memory B cell
e
Plasma cell Bone marrow Neonatal mice Spleen
Germinal centre
Lymph node
a
V(D)J recombination only after 1 week Mature IgD+ B cell
a
Central arteriole
V(D)J recombination (fetal-wave B cells)
? ?
B-cell homing
e
Inefficient homing to bone marrow Bone marrow
Spleen
Lymph node
T-CELL-DEPENDENT ANTIBODY RESPONSES
Antibody responses to protein antigens that require recognition of the antigen by T helper cells and cooperation between antigen-specific B and T cells.
T-CELL-INDEPENDENT ANTIBODY RESPONSES
Antibody responses to polymeric antigens, such as polysaccharides and lipids, that do not require T-cell help.
B1-CELL LINEAGE
Figure 3 | Differences in B-cell development and differentiation during ontogeny. a | In adult mice and humans, B cells are generated in the bone marrow and exit from there as immature (IgM+IgD) B cells. Adult-type progenitors appear in the bone marrow in the first week of life (mice); prior to this time, B cells are generated in the fetal liver and spleen from fetal-type precursors. These remain active in the developing spleen until at least 4 weeks after birth, generating fetal-wave B cells. b | Immature B cells undergo maturation in the adult spleen to become mature IgMlowIgDhi B cells. Some cells are driven (possibly by antigen) to become marginal-zone B cells that reside in this specialized location (the marginal zone). Mature B cells are essentially absent in neonatal mice until after the first week of life, when they can be found in discrete clusters close to the central arteriole. The marginal zone and its populating B cells are not found until 12 weeks after birth. c | Mature B cells can circulate to other lymphoid organs, where entry is controlled by expression of homing molecules, such as L-selectin. In both humans and mice, at least a proportion of neonatal B cells either lack L-selectin or express it at low levels. In the neonatal period, B cells do gain entry to lymph nodes, possibly by the use of other molecular cues. It is not clear whether fetal-wave B cells can gain entry. d | Mature B cells can enter B-cell follicles, which are crucial structures for T-cellB-cell interactions, where activation by antigen drives B cells to differentiate. Activation leads to the formation of germinal centres, which provide the microenvironment to support isotype class switching and further differentiation to long-lived plasma cells or memory B cells. Organized B-cell follicles are first found 710 days after birth and germinal cetres are absent until week 3 in mice. e | Both plasma cells and memory B cells can home to the bone marrow, where the plasma cells are responsible for long-term antibody production. This homing process is less efficient in neonates.
IgM IgD Mac1 B220 CD23 cells that are dominant in the peritoneal and pleural cavities. Their precursors develop in the fetal liver and omentum, and in adult mice, the size of the B1-cell population is kept constant owing to the ability of these cells to self-renew. B1 cells recognize self components, as well as common bacterial antigens, and secrete antibodies that tend to have low affinity and broad specificity.
hi
low
low
from adult progenitors72,73. The splenic B-cell population contains proportionally more of the B1-CELL LINEAGE in neonates74. Because of their different subset compositions, comparisons between adult and neonatal splenic B-cell populations in mice might give misleading information. In humans, neonatal B-cell function has been studied among cells in the peripheral circulation. Interestingly, these cells show few of the defects that are usually associated with neonatal splenic B cells in mice75. Even in mice, the B cells in neonatal lymph nodes are phenotypically and functionally more mature than those in the spleen (B.A. and S.M.-C., unpublished observations). Mouse
neonatal hapten-specific IgG1 and IgG2a responses have been shown to be largely unaffected by splenectomy76, indicating that splenic B cells might not be the major contributors to antibody responses in early life (at least to certain antigens). The more mature phenotype of lymph-node B cells might also account for the adult-like responses to a bacterial vaccine conjugate administered to neonatal mice by the mucosal route77. Taken together and viewed at the level of the whole organism, it seems that intrinsic differences between neonatal and adult B cells are unlikely by themselves to provide an insurmountable obstacle to strong responses.
VOLUME 4 | JULY 2004 | 5 5 7
REVIEWS
Why then are neonatal antibody responses often immature (FIG. 3)? For T-cell-independent TI-2 antigens, the answer may be straightforward. A major subset that responds to these antigens is the marginal-zone B cells78. Marginal-zone B cells are only found in the spleen and are present in low numbers at birth. Their appearance during ontogeny coincides with the initiation of the ability to mount responses to polysaccharide, at 12 weeks in mice and 12 years in humans7982. A parallel situation exists for so-called IgM-expressing memory B cells in humans83. The reliance of these responses on a B-cell subset that only appears late in ontogeny indicates that manipulations (of antigen dose or adjuvant) that might improve other responses are unlikely to be effective for this group of antigens. The ability of polysaccharides to activate complement by the alternative pathway also seems to be important for their immunogenicity. Low expression of complement receptor 2 (CR2; CD21) by human neonatal B cells75 and the relatively low levels of complement component C3 in neonatal serum84 might therefore contribute to the marked deficiency in responses to this type of antigen. These factors have provided the rationale for the development of polysaccharide conjugate vaccines that effectively turn polysaccharides into T-cell-dependent antigens. For B-cell responses to T-cell-dependent antigens, location might be a key factor (FIG. 3). Full antibody responses require three crucial microarchitectural structures: lymphoid follicles, follicular-dendritic-cell (FDC) networks and germinal centres. These structures are absent at birth in mice and develop in the first few days or weeks of life by a mechanism regulated by interaction between stromal cells and B cells, and involving chemokines85, the cytokines tumour-necrosis factor (TNF) and lymphotoxin- (LT)86 and their receptors. The functional immaturity of neonatal B cells could result in an inability to initiate the development of these structures. However, neonatal B cells have been shown to express LT1287, and the transfer of spleen cells from 2-day-old mice into adult SCID mice allowed the development of segregated B-cell areas the first step in the process of generating lymphoid follicles88. By contrast, the formation of an FDC network could not be induced in neonatal mouse spleen or lymph nodes, even by transfer of adult B cells88. This was attributed to unresponsiveness at the level of FDC precursors. As a mature FDC network is essential for germinal-centre formation, there is a fundamental developmentally regulated control on this process. The importance of this is illustrated by the temporal coincidence of the appearance of germinal centres at around 3 weeks of age in mice and the ability to mount adult-like antibody responses88. In humans, full development of the lymphoid microarchitecture also occurs postnatally, with germinal centres only becoming apparent at around 4 months after birth89. It seems probable that these developmental processes, rather than intrinsic differences at the B-cell level, limit neonatal antibody responses. Can this developmental process be altered to allow earlier germinalcentre formation and improved antibody responses? Remarkably, it might be possible. In one report, oral administration of the polyamine spermine, which is involved in postnatal maturation of the gut, accelerated the development of splenic architecture and B-cell and macrophage populations in 12-day-old mice to resemble the level of 510-week-old mice90. It will be interesting to see if this treatment can modify neonatal B-cell immunity in mice and be extended to humans. The final phase of B-cell development is differentiation to immunoglobulin-secreting, long-lived plasma cells or memory cells (FIG. 3). As mentioned in BOX 1, priming for improved recall antibody responses has been observed in human neonates despite poor primary antibody responses, indicating the generation of B-cell memory. However, there is little direct evidence for the production of antigen-specific B cells with the phenotypic characteristics of memory B cells in the period immediately following immunization, leaving open the possibility that improved secondary responses result from the generation of memory B cells that are driven by retained antigen later in ontogeny after the development of germinal centres. Once generated, plasma cells (and memory B cells) relocate from secondary organs to the bone marrow, the environment of which might be important for their longevity and continued function. Antibody production by plasma cells in the bone marrow is required for the sustained nature of adult responses. Pihlgren et al.91 showed that the process of colonization of the bone marrow by plasma cells was markedly compromised in neonatal mice. This finding might explain the brevity of neonatal antibody responses in mice. The mechanisms underlying this failure to migrate could relate to differences between the neonatal and adult bone-marrow environments. A better understanding of these differences and whether they also apply to humans might offer further opportunity to improve neonatal responses.
Dendritic cells
Early studies9294 reported that mouse neonatal APC populations were immature in terms of both number and function. More recent studies have largely focused on development of the DC compartment. Indeed, DCs are now well recognized as probably the most important APCs for stimulating naive T cells. DCs have a dual function, owing to their ability to stimulate both CD4+ and CD8+ T-cell responses. CD4+ T cells recognize peptides derived from antigens that enter the endocytic route of DCs and are displayed in association with MHC class II molecules. After recognition of such complexes and interaction with co-stimulatory molecules, these T cells are activated and provide help to other immune cells, particularly B cells and CD8+ T cells. Moreover, various DC-derived factors can induce the polarization of CD4+ TH cells, such as IL-12, which induces TH1-cell function. There is universal agreement that the absolute number of DCs in neonatal mice is reduced by several logs, compared with adults9598. Nonetheless, an adultlike T cell to DC ratio in the spleen is reached by day 7 (REFS 96,98) , when the different areas of the spleen are becoming organized 96. However, there is some
558
REVIEWS
controversy about the functional capacity of mouse neonatal DCs. This controversy might be partially due to the different methods of isolation of DCs, as well as the complexity of the DC compartment. Stein and colleagues97 isolated mouse DCs using a combination of density centrifugation and adherence methods. They reported that neonatal DCs were deficient in upregulation of MHC class II and co-stimulatory molecule expression and had limited capacity to promote antigenspecific T-cell proliferation. Several other groups have isolated CD11c+ DCs by magnetic-bead sorting to compare the subpopulations present in neonatal and adult mice. PLASMACYTOID DCs, which are characterized as CD11clow (REFS 99,100), made up 3040% of the total number of neonatal DCs at day 7 of life, compared with 1015% of all adult DCs96,98. IFN- production in response to CpG oligonucleotides, which is a characteristic of plasmacytoid DCs, was similar for neonatal and adult mouse CD11c+ cells. Non-plasmacytoid DCs, defined as CD11chi cells, are usually subdivided into different subsets according to the expression of CD4 and CD8. The distribution of
a Neonatal environment
Weak inflammatory signals Low expression of co-stimulatory molecules Low IL-12 production Poor APC function
these subsets was found to be different in neonatal mice, with a higher percentage of splenic CD4CD8+ DCs and a lower percentage of CD4+CD8 DCs96,98. Whereas purified neonatal CD4CD8+ DCs produced modestly (~2.5 fold) lower amounts of IL-12p70, compared with adult CD4CD8+ DCs, total CD11c+ cells from neonatal and adult mouse spleens produced similar amounts of IL-12p70 in response to CpG oligonucleotides98. So, although DCs are reduced in absolute number in neonatal mice and have different subset distributions than in adult animals, they can produce cytokines that have key roles in immunity. Analyses of in vivo function in neonatal mice have produced mixed results. A recent study101 indicated that the small numbers of DCs in early life might be limiting, as treatment of neonatal mice with FMS-like tyrosine kinase 3 ligand (FLT3L) induced an increase in the number of DCs and enhanced their resistance to intracellular pathogens. There is also some evidence that those DCs that are present in neonates might be functionally impaired. Mice aged 37 days have been shown to have a poorly formed network of Langerhans cells in
MHC class II Immature neonatal DC Strong inflammatory signals
CD80/CD86 Expression of co-stimulatory molecules IL-12 production Mature APC function TH1 or TH2 CD4+ T cells Low to high CTL activity IL-12
Mature DC
b In vitro/adult environment
IL-12, IFN- and IFN- In vitro activation CD8+ CTLs CD4+ TH1 cells
PLASMACYTOID DCs
Upregulation of MHC and co-stimulatory molecules CD8+ CTLs Purified CD11c+ mouse neonatal DC CD4+ TH1 cells Adult mouse recipient
A subset of dendritic cells (DCs) that were first described in humans and termed plasmacytoid because of their microscopic appearance, which is similar to plasmablasts. In humans, these DCs can be derived from lineage-negative stem cells in peripheral blood and are the main producers of type I interferon (IFN) in response to virus infections. Recent studies have identified a subset of type I IFN-producing DCs in mice, which are characterized by expression of B220 and Ly6C/G.
In vivo injection
Figure 4 | Neonatal dendritic-cell function is regulated by environmental signals. a | In situ and under low inflammatory conditions, neonatal dendritic cells (DCs) are phenotypically and functionally immature. However, exposure of neonates to strong inflammatory signals, such as CpG oligonucleotides, leads to the upregulation of expression of crucial co-stimulatory molecules and the probable acquisition of mature function in situ. b | Purified CD11c+ mouse neonatal DCs can be efficiently activated in vitro, leading to the production of mature levels of interleukin-12 (IL-12), as well as type I and II interferons (IFNs). When injected into adult mouse recipients, purified neonatal DCs induce strong cytotoxic T lymphocyte (CTL) responses. Efficient IL-12 production as well as the in vitro induction of CTL responses are also observed with human DCs derived from highly purified CD14+ cordblood monocytes. So, as for neonatal T cells, neonatal DC function ranges from immature to fully mature, depending on the environmental signals. APC, antigen-presenting cell; TH, T helper.
VOLUME 4 | JULY 2004 | 5 5 9
REVIEWS
the skin with a lower cell density and reduced MHC class II expression on dendritic-cell processes, compared with 14-day-old mice, and application of a fluorescent antigen to 3-day-old mice revealed that the Langerhans cells were inefficient in transporting antigen to the draining lymph nodes102. Examination of draining lymph-node DCs in vivo after antigenic challenge of the skin has also revealed neonatal mouse DCs to be deficient in the expression of co-stimulatory molecules, CD40 signalling and the capacity to promote antigenspecific T-cell proliferation103. However, other studies have shown that neonatal mouse DCs have functionally mature properties under certain conditions. Injection of neonatal mice with CpG oligonucleotides resulted in the in situ upregulation of MHC class II, CD25, CD40 and CD86 expression by CD11c+ DCs96, and purified neonatal CD11c+ cells promoted strong CTL responses to lymphocytic choriomeningitis virus (LCMV)-derived peptide in adoptive adult hosts95 (FIG. 4). Overall, neonatal APCs, and DCs in particular, are reduced in absolute number in mice and have different subset distributions than in adult animals. Neonatal mouse DCs seem to be functionally immature in some settings. Nonetheless, as for neonatal T cells, it is clear that these cells are capable of mature responses under certain circumstances. Studies of human neonatal APC function have largely focused on IL-12 production. Whole cordblood mononuclear cells or DCs derived from cordblood mononuclear cells by cytokine culturing have been reported to produce reduced levels of IL-12p70 protein and IL-12p35 mRNA, compared with adult DCs, after activation with various stimuli104108. For the cytokine-derived DCs, the addition of IFN- enhanced the production of IL-12p35 mRNA, as well as the bioactive dimeric form of IL-12p70, to adult levels105. However, neonatal monocyte-derived macrophages, purified by adherence, could not be fully activated to kill microorganisms with IFN- a phenomenon that was linked to a marked deficiency in the phosphorylation of the IFN- receptor-associated signal transducer and activator of transcription 1 (STAT1)109. In addition, after overnight incubation of whole blood with CpG DNA, neonatal cord-blood plasmacytoid DCs showed decreased upregulation of expression of co-stimulatory molecules and a marked reduction in IFN- production110. Although the studies described above have generally shown impaired IL-12 production by human neonatal APCs, others have reported levels of IL-12 synthesis that are comparable to adult APCs. Karlsson et al.111 purified neonatal CD14+ monocytes by magnetic-bead sorting and found adult-like levels of IL-12 production in response to stimulation with bacterial strains from normal intestinal flora. Similarly, DCs from adults and neonates, derived from CD14+ monocytes purified by magnetic-bead sorting, showed similar IL-12p70 production in response to microbial stimulation106. Moreover, Salio et al.112 have recently shown that mature human DCs generated from purified cord-blood CD14+ monocytes efficiently prime antigen-specific CTLs in vitro, leading to cytolytic activity and IFN- production. The detection of mature function in these cases might be due to the specific activating agents used and/or purification of monocytes on the basis of CD14 expression, as opposed to adherence procedures, which also select some non-monocytic cells. So, these studies indicate that the impaired capacity for IL-12 synthesis and APC function is not an intrinsic property of neonatal APCs. In summary, for both human and mouse DCs, although deficiencies in function are evident under some circumstances, these cells are competent to function at adult levels (FIG. 4). Mature function is most commonly observed when DCs are highly purified, based on specific marker expression. So, it is possible that some of the discrepancies in observed function might be due to regulatory functions of other cell types that are present in unfractionated cell populations.
The next generation of challenges
EPIGENETIC
Refers to the heritable, but potentially reversible, states of gene activity that are imposed by the structure of chromatin or covalent modifications of DNA and histones.
Although it is clear that neonatal and adult immune responses in both mice and humans differ under most physiological circumstances, the molecular regulation of these responses is poorly understood. In humans, sites within and adjacent to the IFN- promoter in cordblood CD4+ T cells are hypermethylated relative to those in naive adult CD4+ T cells113, whereas cordblood-derived DCs show a deficit in the nucleosome remodelling of an important cis-acting element in the IL-12p35 promoter114. Genes encoding TH2-cell products or other anti-inflammatory cytokines, such as IL-10, have not been similarly analysed. Nonetheless, one attractive hypothesis is that EPIGENETIC differences underlie the differential responsiveness of neonatal and adult immune cells. The possibility that differences in the expression of transcription factors have a role has also been examined, although the results are controversial. For example, ONeill and Reen115 have found similar levels of the transcription factor NFAT1 (nuclear factor of activated T cells) in naive human neonatal and adult T cells. However, others116118 have described decreased levels of this transcription factor in cord-blood, compared with adult, T cells. One possible explanation for these discordant results lies in the specific comparisons made. Whereas ONeill and Reen compared cordblood T cells with purified naive adult T cells, the others compared cord-blood T cells with total adult peripheral-blood T cells. As adult peripheral blood, but not cord-blood, contains a marked and variable number of memory T cells, it is more difficult to interpret results from the latter approach. This cautionary note is, in fact, applicable to all studies that compare regulatory pathways in human neonates and adults. In addition to potential differences at the epigenetic or transcription-factor level, neonatal and adult immune cells might respond differently to the same environmental cues. Certainly, the marked sensitivity of naive cord-blood T cells to common -chain cytokines5963 is consistent with this idea. Higher levels of IL-7 receptor expression by human cord-blood
560
REVIEWS
CD4+ and CD8+ T cells, compared with their counterparts in adult blood, have been described57,59,62,63. Together, these observations support the idea that differences in cytokine receptor expression levels lead to differences in intracellular signalling pathways, which, in turn, lead to different functional outcomes. Overall, the differences between neonatal and adult immune cells might be regulated by a combination of cell-surface events, intracellular signalling and epigenetic modifications. Research over the past decade has improved our understanding of the cellular biology of neonatal immune cells, mainly in mice but also in humans. Armed with this knowledge, we are now ready to dissect the molecular mechanisms that control neonatal immunity.
Implications for paediatric vaccination
Many childhood diseases, including diptheria, tetanus, pertussis, Haemophilus influenzae disease, polio, pneumococcal infection and measles, are now effectively controlled at a population level by vaccines. However, most infant vaccines that are available at present require several doses to elicit protection when administered before 6 months of age66, and there is a persistent risk of pertussis119 or H. influenzae disease120 in partly vaccinated young infants. It would clearly be advantageous to achieve protective immunity earlier in life by enhancing the vaccine responses of newborns. In addition, there are other infectious diseases, such as malaria, HIV, tuberculosis, herpes simplex virus and respiratory syncytial virus, that are often acquired in early life66; the development of effective paediatric vaccines for such agents would be of tremendous benefit.
Box 2 | Synthesis of emerging data: a hypothesis It is increasingly clear that Bacterial products Environmental antigens fully mature adaptive Viruses (low doses) Peripheral self-antigens immune responses can be Strong adjuvants Conventional vaccines generated in neonatal mice and humans; so, why arent Danger mature responses produced all of the time? We think that this question might be partly Quantitatively Deviant, blunted explained by the uniqueness and qualitatively responses of neonatal life. In early mature responses development, the organism is exposed to a large number of antigens, both of extrinsic and intrinsic origin. To respond vigorously to all of these antigens would lead to a state of chronic hyperinflammation. This could be detrimental during early life for two important reasons. First, in the neonate, immune cells must establish a tolerant state to common environmental antigens as well as to newly encountered peripheral self-antigens. High levels of pro-inflammatory signals would be likely to interfere with this process. Second, postnatal development of some organs occurs in both mice and humans. Inflammatory responses can be inherently dangerous to postnatally developing tissues, such as the lungs. Overall, we think that it might be advantageous under most circumstances for neonates to keep inflammatory reactions controlled, while retaining the capacity to fully mobilize the adaptive immune system when faced with lifethreatening or highly damaging infectious agents. So, the flexibility of neonatal immunity might have evolved to achieve low responsiveness under non-threatening conditions but mature responses when the level of danger is high.
Vaccine-induced protection against extracellular organisms relies on the induction of sufficient neutralizing antibodies, whereas protection against intracellular pathogens generally requires both antibodies and T-cell-mediated direct (cytotoxicity) and/or indirect (cytokine release) functions66. Decades of studies have shown that human infants that are vaccinated within the first year of life produce generally poor antibody responses, as described earlier. There are, nonetheless, indications that some adjuvants and/or delivery systems might be able to increase neonatal antibody responses. This area has been reviewed in detail elsewhere66. However, the exploration of TH-cell and CTL responses to infant vaccination is a relatively new field. Recent reports indicate that infant responses to viral vaccines (hepatitis B, polio, measles and mumps) have reduced initial TH1-cell function, compared with immunized adults, as well as reduced memory responses after booster immunizations121123. The emerging evidence from mouse models makes it tempting to speculate that the environment could be manipulated to generate mature vaccine responses in human infants. Indeed, there are several cases in which mature responses have been observed after infant vaccination. Immunization of newborns with BCG results in a potent TH1-cell response that is quantitatively similar to that seen in immunized adults11,12,124. Both polio and hepatitis B virus vaccines, administered at birth and several additional times before 4 months of age, promote poor TH1-cell responses, but induce strong antibody responses to levels similar to those found in immune adults122,123. The potent immuno-enhancing effects of Mycobacterium on DC function125,126 might contribute to the development of mature TH1-cell responses to BCG vaccination; the mechanisms leading to mature antibody responses to some viruses are unknown but might also be contributed to by enhanced activation of components of the innate immune system. Regardless of the mechanisms, the observation that mature responses can be elicited under some circumstances indicates that it might be possible to enhance infant responses to other vaccines, possibly to adult levels. Indeed, the strong TH1-cell-promoting effects of BCG vaccination are being explored as possible enhancers of other vaccine responses concomitant BCG vaccination has been shown to increase antibody, TH1- and TH2-cell responses to hepatitis B virus vaccination, and antibody responses to polio virus vaccine in human newborns127. Although it is an appealing goal, the prospect of enhancing newborn vaccine responses should be considered carefully. In general, the agents used in neonatal mice to obtain mature vaccine responses, including DNA vaccines, the addition of CpG-containing oligonucleotides, replicating viruses or strong TH1-cell-promoting agents such as complete Freunds adjuvant (CFA) or IL-12 (REF. 66), are potent immunomodulators that could potentially lead to a state of heightened inflammation. Such a state might be deleterious to normal development or even survival. This possibility is well illustrated in neonatal mice, for which IL-12 treatment enhances
VOLUME 4 | JULY 2004 | 5 6 1
REVIEWS
TH1-cell responses32,128130 but causes marked toxicity, as assessed by suboptimal weight gain or death at higher doses131. Immediate toxicity is not the only concern. The newborn period is characterized by exposure to a large array of antigens and by the establishment of tolerance to many self-antigens in peripheral tissues. Enhancement of responses to one antigen in early life could lead to bystander effects on unrelated antigens. In fact, it is clear that bystander effects can be obtained in both neonatal mice132,133 and humans127. The effects of eliciting a strong response to the administered vaccine antigens on the responses to the many other antigens that are encountered simultaneously during the neonatal period are unknown. It is also unclear whether the generation of potent responses to vaccine antigens might affect the developmental acquisition of tolerance to self-antigens. This is potentially of concern as many of the delivery systems under consideration are strong promoters of TH1-cell development and many autoimmune diseases are TH1-cell mediated. However, the extrapolation of mouse studies to humans must be carefully considered, as millions of human infants safely receive the potent TH1-cell-promoting vaccine BCG. Therefore, human infants, who are immunologically more mature than neonatal mice, might tolerate strong TH1-cell-promoting agents more readily. It is also possible that the specific interaction of BCG antigens with selected cellular receptors results in a unique combination of efficacy and safety in infants. Learning how key receptors for microbial products are regulated during human development might allow the formulation of vaccines targeting elements of the inflammatory response that can be safely triggered during neonatal life.
Concluding remarks
Our understanding of neonatal immunity has developed in a step-wise manner. Initial assumptions were that neonates were immature or immunodeficient. This supposition was replaced by the idea that neonates are immunodeviant. Subsequently, demonstrations that neonates are competent to mount mature responses raised the possibility that simply smaller size or immune-cell numbers could account fully for the differences between neonates and adults. However, it is becoming increasingly clear that both human and mouse neonates mount various responses, ranging from deficient or deviant to fully mature, depending on the conditions of antigen exposure. This flexibility of responsiveness might have important functions in protecting the developing organism from potentially dangerous inflammatory situations, at the same time as providing protection against potentially life-threatening infections (BOX 2). The challenges for the future include discovering the molecular mechanisms that underlie the plasticity of neonatal immune function and methods for safely harnessing neonatal malleability to improve the prevention and treatment of childhood disease.
Goulder, P. J., Jeena, P., Tudor-Williams, G. & Burchett, S. Paediatric HIV infection: correlates of protective immunity and global perspectives in prevention and management. Br. Med. Bull. 58, 89108 (2001). 2. Luzuriaga, K. & Sullivan, J. L. Pediatric HIV-1 infection: advances and remaining challenges. AIDS Rev. 4, 2126 (2002). 3. Sarzotti, M., Robbins, D. S. & Hoffman, P. M. Induction of protective CTL responses in newborn mice by a murine retrovirus. Science. 271, 17261728 (1996). 4. Ridge, J. P., Fuchs, E. J. & Matzinger, P. Neonatal tolerance revisited: turning on newborn T cells with dendritic cells. Science 271, 17231726 (1996). 5. Forsthuber, T., Yip, H. C. & Lehmann, P. V. Induction of TH1 and TH2 immunity in neonatal mice. Science 217, 17281730 (1996). References 3, 4 and 5 were fundamental in bringing a new era to neonatal immunology, by demonstrating that mature TH1-cell/CTL responses could be elicited in newborn mice. 6. Garcia, A. M., Fadel, S. A., Cao, S. & Sarzotti, M. T cell immunity in neonates. Immunol. Res. 22, 177190 (2000). 7. Siegrist, C. A. Vaccination in the neonatal period and early infancy. Int. Rev. Immunol. 19, 195219 (2000). 8. Adkins, B. Development of neonatal TH1/TH2 function. Int. Rev. Immunol. 19, 157171 (2000). 9. Winkler, S. et al. Frequency of cytokine-producing T cells in patients of different age groups with Plasmodium falciparum malaria. J. Infect. Dis. 179, 209216 (1999). 10. Xainli, J. et al. Age-dependent cellular immune responses to Plasmodium vivax Duffy binding protein in humans. J. Immunol. 169, 32003207 (2002). 11. Hussey, G. D. et al. Neonatal mycobacterial specific cytotoxic T-lymphocyte and cytokine profiles in response to distinct BCG vaccination strategies. Immunology 105, 314324 (2002). 12. Vekemans, J. et al. Neonatal bacillus CalmetteGuerin vaccination induces adult-like IFN- production by CD4+ T lymphocytes. Eur. J. Immunol. 31, 15311535 (2001). References 11 and 12 show mature TH1-cell responses after vaccination of human infants with the BCG vaccine.
1.
13. Jotereau, F., Heuze, F., Salomon-Vie, V. & Gascan, H. Cell kinetics in the fetal mouse thymus: precursor cell input, proliferation, and emigration. J. Immunol. 138, 10261030 (1987). 14. Douagi, I. I., Andre, I. I., Ferraz, J. C. & Cumano, A. Characterization of T cell precursor activity in the murine fetal thymus: evidence for an input of T cell precursors between days 12 and 14 of gestation. Eur. J. Immunol. 30, 22012210 (2000). 15. Foss, D. L., Donskoy, E. & Goldschneider, I. The importation of hematogenous precursors by the thymus is a gated phenomenon in normal adult mice. J. Exp. Med. 193, 365374 (2001). 16. Adkins, B. Peripheral CD4+ lymphocytes derived from fetal versus adult thymic precursors differ phenotypically and functionally. J. Immunol. 171, 51575164 (2003). 17. Prescott, S. L. et al. Transplacental priming of the human immune system to environmental allergens: universal skewing of initial T cell responses toward the TH2 cytokine profile. J. Immunol. 160, 47304737 (1998). References 16 and 17 show that fetal cells might be an important source of TH2-type cytokines in early life in mice and humans, respectively. 18. Prescott, S. L. et al. Developing patterns of T cell memory to environmental allergens in the first two years of life. Int. Arch. Allergy. Immunol. 113, 7579 (1997). 19. Prescott, S. L. et al. Development of allergen-specific T-cell memory in atopic and normal children. Lancet 353, 196200 (1999). 20. Adkins, B., Bu, Y. & Guevara, P. Murine neonatal CD4+ lymph node cells are highly deficient in the development of antigen-specific TH1 function in adoptive adult hosts. J. Immunol. 169, 49985004 (2002). 21. Qureshi, M. H. & Garvy, B. A. Neonatal T cells in an adult lung environment are competent to resolve Pneumocystis carinii pneumonia. J. Immunol. 166, 57045711 (2001). 22. Li, L. et al. IL-4 utilizes an alternative receptor to drive apoptosis of TH1 cells and skews neonatal immunity toward TH2. Immunity 20, 429440 (2004). This paper shows that mouse neonatal TH1 cells undergo IL-4-dependent apoptosis after re-exposure to antigen in vivo. This work provides a potential mechanism for the typically poor TH1-cell function
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
observed after secondary immunization of mouse neonates. Schurmans, S. et al. In vivo effects of anti-IL-4 monoclonal antibody on neonatal induction of tolerance and on an associated autoimmune syndrome. J. Immunol. 145, 24652473 (1990). Powell, T. J., Jr & Streilein, J. W. Neonatal tolerance induction by class II alloantigens activates IL-4- secreting, tolerogenresponsive T cells. J. Immunol. 144, 854859 (1990). Abramowicz, D. et al. Persistence of anti-donor allohelper T cells after neonatal induction of allotolerance in mice. Eur. J. Immunol. 20, 16471653 (1990). Nossal, G. J. & Pike, B. L. Functional clonal deletion in immunological tolerance to major histocompatibility complex antigens. Proc. Natl Acad. Sci. USA 78, 38443847 (1981). Powell, T. J. & Streilein, J. W. In vitro suppression of cytotoxic T cell generation by lymphocytes from mice rendered neonatally tolerant of class II MHC alloantigens. Transplantation 52, 383386 (1991). Streilein, J. W., Mohler, K. & Wood, P. J. Mechanisms of neonatal transplantation tolerance. Transplant. Proc. 19, 8287 (1987). Wood, P. J., Socarras, S. & Streilein, J. W. Modification of the cytotoxic T cell repertoire in neonatal tolerance. Evidence for preferential survival of cells with low avidity for tolerogen. J. Immunol. 139, 32363244 (1987). Gao, Q., Chen, N., Rouse, T. M. & Field, E. H. The role of interleukin-4 in the induction phase of allogeneic neonatal tolerance. Transplantation 62, 18471854 (1996). Flamand, V. et al. CD40 ligation prevents neonatal induction of transplantation tolerance. J. Immunol. 160, 46664669 (1998). Donckier, V. et al. IL-12 prevents neonatal induction of transplantation tolerance in mice. Eur. J. Immunol. 28, 14261430 (1998). Franchini, M. et al. Protective T-cell-based immunity induced in neonatal mice by a single replicative cycle of herpes simplex virus. J. Virol. 75, 8389 (2001). Kovarik, J. et al. Induction of adult-like antibody, TH1, and CTL responses to measles hemagglutinin by early life murine immunization with an attenuated vaccinia-derived NYVAC(K1L) viral vector. Virology 285, 1220 (2001).
562
REVIEWS
35. Fadel, S. A., Ozaki, D. A. & Sarzotti, M. Enhanced type 1 immunity after secondary viral challenge in mice primed as neonates. J. Immunol. 169, 32933300 (2002). The authors show that when infected as neonates with low doses of leukaemia virus, mice can mount strong CTL memory responses after virus rechallenge. 36. Bot, A., Bot, S. & Bona, C. Enhanced protection against influenza virus of mice immunized as newborns with a mixture of plasmids expressing hemagglutinin and nucleoprotein. Vaccine 16, 16751682 (1998). 37. Hassett, D. E., Zhang, J. & Whitton, J. L. Neonatal DNA immunization with a plasmid encoding an internal viral protein is effective in the presence of maternal antibodies and protects against subsequent viral challenge. J. Virol. 71, 78817888 (1997). 38. Hassett, D. E., Zhang, J., Slifka, M. & Whitton, J. L. Immune responses following neonatal DNA vaccination are longlived, abundant, and qualitatively similar to those induced by conventional immunization. J. Virol. 74, 26202627 (2000). 39. Martinez, X. et al. DNA immunization circumvents deficient induction of T helper type 1 and cytotoxic T lymphocyte responses in neonates and during early life. Proc. Natl Acad. Sci. USA 94, 87268731 (1997). 40. Sarzotti, M. et al. Induction of cytotoxic T cell responses in newborn mice by DNA immunization. Vaccine 15, 795797 (1997). 41. Brazolot Millan, C. L. et al. CpG DNA can induce strong TH1 humoral and cell-mediated immune responses against hepatitis B surface antigen in young mice. Proc. Natl Acad. Sci. USA 95, 1555315558 (1998). 42. Kovarik, J. et al. CpG oligodeoxynucleotides can circumvent the TH2 polarization of neonatal responses to vaccines but may fail to fully redirect TH2 responses established by neonatal priming. J. Immunol. 162, 16111617 (1999). 43. Marchant, A. et al. Mature CD8+ T lymphocyte response to viral infection during fetal life. J. Clin. Invest. 111, 17471755 (2003). This report indicates that infection with human cytomegalovirus (HCMV) in utero leads to clonal expansion and differentiation of mature HCMVspecific CD8+ T cells with adult-like properties. Therefore, humans can develop mature CTL function as early as during fetal life. 44. Hermann, E. et al. Human fetuses are able to mount an adult-like CD8 T-cell response. Blood 100, 21532158 (2002). 45. Zhang, J., Silvestri, N., Whitton, J. L. & Hassett, D. E. Neonates mount robust and protective adult-like CD8+ T-cell responses to DNA vaccines. J. Virol. 76, 1191111919 (2002). 46. Martinez, X. et al. CD4-independent protective cytotoxic T cells induced in early life by a non-replicative delivery system based on virus-like particles. Virology 305, 428435 (2003). 47. Marcus, H. et al. T cells from newborn humans are fully capable of developing into cytotoxic T lymphocyte effector cells in adoptive hosts. Transplantation 73, 803810 (2002). 48. Nishizuka, Y. & Sakakura, T. Thymus and reproduction: sex-linked dysgenesia of the gonad after neonatal thymectomy in mice. Science 166, 753755 (1969). 49. Suri-Payer, E., Amar, A. Z., Thornton, A. M. & Shevach, E. M. CD4+CD25+ T cells inhibit both the induction and effector function of autoreactive T cells and represent a unique lineage of immunoregulatory cells. J. Immunol. 160, 12121218 (1998). 50. Gao, Q., Rouse, T. M., Kazmerzak, K. & Field, E. H. CD4+CD25+ cells regulate CD8 cell anergy in neonatal tolerant mice. Transplantation 68, 18911897 (1999). 51. Adams, B. et al. CD8+ T lymphocytes regulating TH2 pathology escape neonatal tolerization. J. Immunol. 171, 50715076 (2003). 52. Field, A. C., Caccavelli, L., Bloch, M. F. & Bellon, B. Regulatory CD8+ T cells control neonatal tolerance to a TH2-mediated autoimmunity. J. Immunol. 170, 25082515 (2003). 53. Min, B. et al. Neonates support lymphopenia-induced proliferation. Immunity 18, 131140 (2003). 54. Schuler, T., Hammerling, G. J. & Arnold, B. Cutting edge: IL-7-dependent homeostatic proliferation of CD8+ T cells in neonatal mice allows the generation of long-lived natural memory T cells. J. Immunol. 172, 1519 (2004). References 53 and 54 show that the mouse neonatal environment, similar to that found in lymphopenic adult hosts, promotes the homeostatic proliferation and/or development of TH1-cell function by adult CD4+ or CD8+ T cells. 55. Ichii, H. et al. Role for Bcl-6 in the generation and maintenance of memory CD8+ T cells. Nature Immunol. 3, 558563 (2002). 56. Le Campion, A. et al. Naive T cells proliferate strongly in neonatal mice in response to self- peptide/self-MHC complexes. Proc. Natl Acad. Sci. USA 99, 45384543 (2002). 57. Schonland, S. O. et al. Homeostatic control of T-cell generation in neonates. Blood 102, 14281434 (2003). 58. Adkins, B., Williamson, T., Guevara, P. & Bu, Y. Murine neonatal lymphocytes show rapid early cell cycle entry and cell division. J. Immunol. 170, 45484556 (2003). 59. Hassan, J. & Reen, D. J. Human recent thymic emigrants identification, expansion, and survival characteristics. J. Immunol. 167, 19701976 (2001). 60. Hassan, J. & Reen, D. J. IL-7 promotes the survival and maturation but not differentiation of human post-thymic CD4+ T cells. Eur. J. Immunol. 28, 30573065 (1998). This report shows that human neonatal, but not adult, naive phenotype CD4+ T cells undergo marked proliferation and acquire IFN- secretion in response to the common -chain cytokine IL-7. 61. Cookson, S. & Reen, D. IL-15 drives neonatal T cells to acquire CD56 and become activated effector cells. Blood 102, 21952197 (2003). 62. Fukui, T. et al. IL-7 induces proliferation, variable cytokineproducing ability and IL-2 responsiveness in naive CD4+ T-cells from human cord blood. Immunol. Lett. 59, 2128 (1997). 63. Dardalhon, V. et al. IL-7 differentially regulates cell cycle progression and HIV-1-based vector infection in neonatal and adult CD4+ T cells. Proc. Natl Acad. Sci. USA 98, 92779282 (2001). 64. Adkins, B., Chun, K., Hamilton, K. & Nassiri, M. Naive murine neonatal T cells undergo apoptosis in response to primary stimulation. J. Immunol. 157, 13431349 (1996). 65. Marshall-Clarke, S., Reen, D., Tasker, L. & Hassan, J. Neonatal immunity: how well has it grown up? Immunol. Today 21, 3541 (2000). 66. Siegrist, C. A. Neonatal and early life vaccinology. Vaccine 19, 33313346 (2001). 67. Schallert, N. et al. Generation of adult-like antibody avidity profiles after early-life immunization with protein vaccines. Eur. J. Immunol. 32, 752760 (2002). 68. Giorgetti, C. A. & Press, J. L. Somatic mutation in the neonatal mouse. J. Immunol. 161, 60936104 (1998). 69. King, L. B., Norvell, A. & Monroe, J. G. Antigen receptorinduced signal transduction imbalances associated with the negative selection of immature B cells. J. Immunol. 162, 26552662 (1999). 70. Marshall-Clarke, S., Tasker, L. & Parkhouse, R. M. Immature B lymphocytes from adult bone marrow exhibit a selective defect in induced hyperexpression of major histocompatibility complex class II and fail to show B7.2 induction. Immunology 100, 141151 (2000). 71. Benschop, R. J., Brandl, E., Chan, A. C. & Cambier, J. C. Unique signaling properties of B cell antigen receptor in mature and immature B cells: implications for tolerance and activation. J. Immunol. 167, 41724179 (2001). 72. Lu, L. S. et al. Identification of a germ-line pro-B cell subset that distinguishes the fetal/neonatal from the adult B cell development pathway. Proc. Natl Acad. Sci. USA 99, 30073012 (2002). 73. Kincade, P. W. et al. Nature or nurture? Steady-state lymphocyte formation in adults does not recapitulate ontogeny. Immunol. Rev. 187, 116125 (2002). 74. Hardy, R. R. & Hayakawa, K. B cell development pathways. Annu. Rev. Immunol. 19, 595621 (2001). 75. Tasker, L. & Marshall-Clarke, S. Functional responses of human neonatal B lymphocytes to antigen receptor crosslinking and CpG DNA. Clin. Exp. Immunol. 134, 409419 (2003). 76. Adkins, B., Bu, Y., Vincek, V. & Guevara, P. The primary responses of murine neonatal lymph node CD4+ cells are TH2-skewed and are sufficient for the development of TH2-biased memory. Clin. Dev. Immunol. 10, 4351 (2003). 77. Jakobsen, H. et al. Intranasal immunization with pneumococcal conjugate vaccines with LT-K63, a nontoxic mutant of heat-labile enterotoxin, as adjuvant rapidly induces protective immunity against lethal pneumococcal infections in neonatal mice. Infect. Immun. 70, 14431452 (2002). Location matters. Neonatal responses to a pneumococcal polysaccharide conjugate vaccine in mice are shown to be markedly enhanced by intranasal, rather than parenteral, delivery with a mucosal adjuvant and to generate protective immunity. 78. Martin, F., Oliver, A. M. & Kearney, J. F. Marginal zone and B1 B cells unite in the early response against T-independent blood-borne particulate antigens. Immunity 14, 617629 (2001). 79. Balogh, P., Aydar, Y., Tew, J. G. & Szakal, A. K. Ontogeny of the follicular dendritic cell phenotype and function in the postnatal murine spleen. Cell. Immunol. 214, 4553 (2001). 80. Mosier, D. E. & Johnson, B. M. Ontogeny of mouse lymphocyte function. II. Development of the ability to produce antibody is modulated by T lymphocytes. J. Exp. Med. 141, 216226 (1975). 81. Timens, W., Boes, A., Rozeboom-Uiterwijk, T. & Poppema, S. Immaturity of the human splenic marginal zone in infancy. Possible contribution to the deficient infant immune response. J. Immunol. 143, 32003206 (1989). 82. Rijkers, G. T., Sanders, E. A., Breukels, M. A. & Zegers, B. J. Infant B cell responses to polysaccharide determinants. Vaccine 16, 13961400 (1998). 83. Kruetzmann, S. et al. Human immunoglobulin M memory B cells controlling Streptococcus pneumoniae infections are generated in the spleen. J. Exp. Med. 197, 939945 (2003). 84. Davis, C. A., Vallota, E. H. & Forristal, J. Serum complement levels in infancy: age related changes. Pediatr. Res. 13, 10431046 (1979). 85. Ansel, K. M. et al. A chemokine-driven positive feedback loop organizes lymphoid follicles. Nature 406, 309314 (2000). 86. Fu, Y. X. & Chaplin, D. D. Development and maturation of secondary lymphoid tissues. Annu. Rev. Immunol. 17, 399433 (1999). 87. Ngo, V. N., Cornall, R. J. & Cyster, J. G. Splenic T zone development is B cell dependent. J. Exp. Med. 194, 16491660 (2001). 88. Pihlgren, M. et al. Unresponsiveness to lymphoid-mediated signals at the neonatal follicular dendritic cell precursor level contributes to delayed germinal center induction and limitations of neonatal antibody responses to T-dependent antigens. J. Immunol. 170, 28242832 (2003). This paper shows that the ontogenetic development of lymphoid follicles, which are crucial structures for effective antibody responses, is regulated by the responsiveness of FDC precursors to B-cell signals. It, therefore, indicates that factors which enhance FDC development might improve neonatal responsiveness. 89. Timens, W., Rozeboom, T. & Poppema, S. Fetal and neonatal development of human spleen: an immunohistological study. Immunology 60, 603609 (1987). 90. Jolois, O. et al. Spermine induces precocious development of the spleen in mice. Exp. Physiol. 87, 6975 (2002). 91. Pihlgren, M. et al. Delayed and deficient establishment of the long-term bone marrow plasma cell pool during early life. Eur. J. Immunol. 31, 939946 (2001). 92. Lu, C. Y., Calamai, E. G. & Unanue, E. R. A defect in the antigen-presenting function of macrophages from neonatal mice. Nature 282, 327329 (1979). 93. Lu, C. Y., Beller, D. I. & Unanue, E. R. During ontogeny, Ia-bearing accessory cells are found early in the thymus but late in the spleen. Proc. Natl Acad. Sci. USA 77, 15971601 (1980). 94. Levin, D. & Gershon, H. Antigen presentation by neonatal murine spleen cells. Cell. Immunol. 120, 132144 (1989). 95. Dadaglio, G. et al. Efficient in vivo priming of specific cytotoxic T cell responses by neonatal dendritic cells. J. Immunol. 168, 22192224 (2002). 96. Sun, C. M. et al. Ontogeny and innate properties of neonatal dendritic cells. Blood 102, 585591 (2003). Papers 95 and 96 were the first detailed analysis of mouse DC ontogeny and show that the innate properties of highly purified neonatal and adult DCs are comparable. 97. Muthukkumar, S., Goldstein, J. & Stein, K. E. The ability of B cells and dendritic cells to present antigen increases during ontogeny. J. Immunol. 165, 48034813 (2000). 98. Dakic, A. et al. Development of the dendritic cell system during mouse ontogeny. J. Immunol. 172, 10181027 (2004). 99. Asselin-Paturel, C. et al. Mouse type I IFN-producing cells are immature APCs with plasmacytoid morphology. Nature Immunol. 2, 11441150 (2001). 100. Nakano, H., Yanagita, M. & Gunn, M. D. CD11c+B220+Gr-1+ cells in mouse lymph nodes and spleen display characteristics of plasmacytoid dendritic cells. J. Exp. Med. 194, 11711178 (2001). 101. Vollstedt, S. et al. Flt3 ligand-treated neonatal mice have increased innate immunity against intracellular pathogens and efficiently control virus infections. J. Exp. Med. 197, 575584 (2003). 102. Dewar, A. L. et al. Acquisition of immune function during the development of the Langerhans cell network in neonatal mice. Immunology 103, 6169 (2001). 103. Simpson, C. C., Woods, G. M. & Muller, H. K. Impaired CD40-signalling in Langerhans cells from murine neonatal draining lymph nodes: implications for neonatally induced cutaneous tolerance. Clin. Exp. Immunol. 132, 201208 (2003).
VOLUME 4 | JULY 2004 | 5 6 3
REVIEWS
104. Joyner, J. L. et al. Effects of group B streptococci on cord and adult mononuclear cell interleukin-12 and interferon- mRNA accumulation and protein secretion. J. Infect. Dis. 182, 974977 (2000). 105. Goriely, S. et al. Deficient IL-12(p35) gene expression by dendritic cells derived from neonatal monocytes. J. Immunol. 166, 21412146 (2001). 106. Upham, J. W. et al. Development of interleukin-12producing capacity throughout childhood. Infect. Immun. 70, 65836588 (2002). 107. Stefanovic, V., Golubovic, E., Vlahovic, P. & Mitic-Zlatkovic, M. Age-related changes in IL-12 production by peripheral blood mononuclear cells (PBMC). J. Intern. Med. 243, 8384 (1998). 108. Langrish, C. L., Buddle, J. C., Thrasher, A. J. & Goldblatt, D. Neonatal dendritic cells are intrinsically biased against TH1 immune responses. Clin. Exp. Immunol. 128, 118123 (2002). 109. Marodi, L., Goda, K., Palicz, A. & Szabo, G. Cytokine receptor signalling in neonatal macrophages: defective STAT-1 phosphorylation in response to stimulation with IFN-. Clin. Exp. Immunol. 126, 456460 (2001). 110. De Wit, D. et al. Blood plasmacytoid dendritic cell responses to CpG oligodeoxynucleotides are impaired in human newborns. Blood 103, 10301032 (2004). 111. Karlsson, H., Hessle, C. & Rudin, A. Innate immune responses of human neonatal cells to bacteria from the normal gastrointestinal flora. Infect. Immun. 70, 66886696 (2002). 112. Salio, M. et al. Efficient priming of antigen-specific cytotoxic T lymphocytes by human cord blood dendritic cells. Int. Immunol. 15, 12651273 (2003). This paper shows that purified newborn human DCs are highly efficient in inducing CTL responses in vitro. 113. White, G. P., Watt, P. M., Holt, B. J. & Holt, P. G. Differential patterns of methylation of the IFN- promoter at CpG and non-CpG sites underlie differences in IFN- gene expression between human neonatal and adult CD45RO T cells. J. Immunol. 168, 28202827 (2002). 114. Goriely, S. et al. A defect in nucleosome remodeling prevents IL-12(p35) gene transcription in neonatal dendritic cells. J. Exp. Med. 199, 10111016 (2004). This study is the first demonstration that the regulation of neonatal IL-12 production in humans can take place at the chromatin level. 115. ONeill, R. M. & Reen, D. J. Equivalent functional nuclear factor of activated T cell 1 mRNA and protein expression in cord blood and adult T cells. Transplantation 76, 15261528 (2003). 116. Miller, R. E. et al. Reduced CTLA-4 protein and messenger RNA expression in umbilical cord blood T lymphocytes. Exp. Hematol. 30, 738744 (2002). 117. Kadereit, S. et al. Reduced NFAT1 protein expression in human umbilical cord blood T lymphocytes. Blood 94, 31013107 (1999). 118. Kaminski, B. A. et al. Reduced expression of NFATassociated genes in UCB versus adult CD4+ T lymphocytes during primary stimulation. Blood 102, 46084617 (2003). 119. Center for Disease Control and Prevention, United States. Pertussis United States January 1992June 1995. Morbid Mortal Wkly Rep. 44, 525 (1995). 120. Bisgard, K. M. et al. Haemophilus influenzae invasive disease in the United States, 19941995: near disappearance of a vaccine-preventable childhood disease. Emerg. Infect. Dis. 4, 229237 (1998). 121. Gans, H. et al. Immune responses to measles and mumps vaccination of infants at 6, 9, and 12 months. J. Infect. Dis. 184, 817826 (2001). 122. Vekemans, J. et al. T cell responses to vaccines in infants: defective IFN- production after oral polio vaccination. Clin. Exp. Immunol. 127, 495498 (2002). 123. Ota, M. O. et al. Hepatitis B immunisation induces higher antibody and memory TH2 responses in new-borns than in adults. Vaccine 22, 511519 (2004). 124. Marchant, A. et al. Newborns develop a TH1-type immune response to Mycobacterium bovis bacillus CalmetteGuerin vaccination. J. Immunol. 163, 22492255 (1999). 125. Inaba, K., Inaba, M., Naito, M. & Steinman, R. M. Dendritic cell progenitors phagocytose particulates, including bacillus CalmetteGuerin organisms, and sensitize mice to mycobacterial antigens in vivo. J. Exp. Med. 178, 479488 (1993). 126. Henderson, R. A., Watkins, S. C. & Flynn, J. L. Activation of human dendritic cells following infection with Mycobacterium tuberculosis. J. Immunol. 159, 635643 (1997). 127. Ota, M. O. et al. Influence of Mycobacterium bovis bacillus CalmetteGuerin on antibody and cytokine responses to human neonatal vaccination. J. Immunol. 168, 919925 (2002). 128. Arulanandam, B. P., Van Cleave, V. H. & Metzger, D. W. IL-12 is a potent neonatal vaccine adjuvant. Eur. J. Immunol. 29, 256264 (1999). 129. Urban, J. F., Jr. et al. IL12 protects immunocompetent and immunodeficient neonatal mice against infection with Cryptosporidium parvuum. J. Immunol. 156, 263 (1996). 130. Mancuso, G. et al. Role of interleukin 12 in experimental neonatal sepsis caused by group B streptococci. Infect. Immun. 65, 37313735 (1997). 131. Kovarik, J. et al. Limitations of in vivo IL-12 supplementation strategies to induce TH1 early life responses to model viral and bacterial vaccine antigens. Virology 268, 122131 (2000). 132. Pack, C. D. et al. Neonatal exposure to antigen primes the immune system to develop responses in various lymphoid organs and promotes bystander regulation of diverse T cell specificities. J. Immunol. 167, 41874195 (2001). 133. Walzl, G. et al. Influenza virus lung infection protects from respiratory syncytial virus-induced immunopathology. J. Exp. Med. 192, 13171326 (2000). 134. Glezen, W. P. Effect of maternal antibodies on the infant immune response. Vaccine 21, 33893392 (2003). 135. Albrecht, P., Ennis, F. A., Saltzman, E. J. & Krugman, S. Persistence of maternal antibody in infants beyond 12 months: mechanism of measles vaccine failure. J. Pediatr. 91, 715718 (1977). 136. Bjorkholm, B. et al. Influence of high titers of maternal antibody on the serologic response of infants to diphtheria vaccination at three, five and twelve months of age. Pediatr. Infect. Dis. J. 14, 846850 (1995). 137. Gans, H. A. et al. IL-12, IFN-, and T cell proliferation to measles in immunized infants. J. Immunol. 162, 55695575 (1999). 138. Siegrist, C. A. et al. Influence of maternal antibodies on vaccine responses: inhibition of antibody but not T cell responses allows successful early primeboost strategies in mice. Eur. J. Immunol. 28, 41384148 (1998). 139. Santosham, M. et al. Safety and antibody persistence following Haemophilus influenzae type b conjugate or pneumococcal polysaccharide vaccines given before pregnancy in women of childbearing age and their infants. Pediatr. Infect. Dis. J. 20, 931940 (2001).

Online links
DATABASES The following terms in this article are linked online to: Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene CD11c | CD14 | CD21 | FLT3L | IFN- | IL-2 | IL-7 | IL-12p35 | IL-15 | NFAT1 | STAT1 Access to this interactive links box is free online.
564
PERSPECTIVES
SCIENCE AND SOCIETY
Breastfeeding: maintaining an irreplaceable immunological resource

Miriam H. Labbok, David Clark and Armond S. Goldman
Breastfeeding the main source of active and passive immunity in the vulnerable early months and years of life is considered to be the most effective preventive means of reducing the death rate of children under five. Given this, one must wonder why it has slipped quietly down the priorities of the global health and development agendas. In this era of publicprivate partnerships, can its role as an irreplaceable immunological resource help keep it at the top of global agendas?
Of the more than 10 million children who die each year in the developing world, about 60% of these deaths are preventable. The authors of a major collaborative study published in 2003 (REF. 1) proposed that a single preventive measure increasing optimal breastfeeding behaviours (defined in this study as exclusive breastfeeding for 6 months, with continued breastfeeding for at least 1 year) could prevent 13% of deaths. That is, it could save 1.3 million children annually (TABLE 1). When compared with the other
interventions noted in the report such as supplementation with zinc (which could prevent 4% of deaths), vaccination against infection with Haemophilus influenzae type b (4% of deaths), or vaccination against tetanus (2% of deaths) or measles (1% of deaths) it becomes even clearer that meaningful intervention to support increased breastfeeding should be a priority. Also, the current level of breastfeeding cannot be allowed to diminish. By some estimates, current advances in breastfeeding practice are already saving millions of lives annually. In addition to the huge impact on mortality readily seen in developing countries, breastfeeding has measurable and profound impacts in all settings. Breastfeeding also makes important contributions to growth2, and to cognitive and psychosocial development3, and therefore leads to a healthier population, capable of greater achievement and enjoying longer life. Compared with buying breast-milk substitutes, breastfeeding
Table 1 | Potential impact of preventive measures on mortality of children under 5: how many children could we save?
Preventive intervention Breastfeeding Insecticide-treated materials Complementary feeding Clean delivery (efforts to ensure that childbirth is free of unnecessary contamination) Haemophilus influenzae type b vaccination Zinc supplementation Clean water, sanitation, hygiene Antenatal steroids Vitamin A supplementation Tetanus toxoid vaccination Nevirapine and replacement feeding Measles vaccination Antimalarial intermittent preventive treatment in pregnancy Newborn temperature management Antibiotics for premature rupture of amniotic membranes Estimated deaths prevented* (thousands) 1,301 691 587 411 403 351 326 264 176 161 150 103 22 0 0 Estimated deaths prevented* (% of all deaths) 13 7 6 4 4 4 3 3 2 2 2 1 1 0 0
Based on data from REF. 1. *Determined using data from the 42 countries that had 90% of the child deaths worldwide in 2000, by estimating the impact of universal coverage with each individual intervention.
VOLUME 4 | JULY 2004 | 5 6 5
PERSPECTIVES
provides economic benefits to families, which then translate into national economic gains. Clearly, this crucial practice contributes considerably to the health and life of each child and to the development of the population in all settings. Optimal practices include early initiation, exclusive breastfeeding for 6 months and continued breastfeeding (together with appropriate complementary feeding) for 2 years or more. Yet breastfeeding seems to be given little importance in todays global health and development action agendas. Despite its phenomenal effects, interventions proven to increase breastfeeding rates are not being highlighted for support by important donors. Globally, it seems that donor resources specified for this crucial work have declined. Why is such a clear winner being left at the post in the race to prevent untimely infant and child mortality? Is this a result of concerns over the transmission of HIV or of pressure from the infant-feeding industry? Or is breastfeeding merely no longer fashionable among those working in international development? This perspective article examines the immunological basis for the positive impact of breastfeeding on health and survival, and it reviews the support accorded to breastfeeding in international public-health and development agendas. We also discuss why these messages seem to be absent today and how focusing on the irreplaceable functions of human milk which are provided mainly by its immunological components might be a crucial way to re-establish and maintain global attention on support for breastfeeding.
Immunology of breastfeeding
acquired immunodeficiency that increases the risk of infections and other diseases. The antimicrobial, anti-inflammatory and immunomodulatory agents that are present in human milk48 are multifunctional, function synergistically and compensate for developmental delays in the infant (BOX 1). The antimicrobial factors include the following: growth promoters of protective enteric bacteria agents; that can interfere with bacterial metabolism; enzymes that lyse bacteria; antibodies adapted to mucosal sites; oligosaccharides; glycosylated proteins; antiviral lipids created by the digestion of neutral fats in milk; and leukocytes7,8. Some of the main examples of these are given here. Growth promoters. The proliferation of commensal Bifidobacterium bifidum and lactobacilli in the lower intestinal tract9 is promoted by certain glycopeptides and glycoproteins, including caseins in human milk1012. These bacteria then produce organic acids that retard the growth of enteric pathogens13. Inhibitors of metabolism. Lactoferrin, a ferric iron-binding glycoprotein, is largely unsaturated14 and is present in high concentrations in human milk8. Lactoferrin inhibits the growth of certain pathogens by competing with bacteria for ferric iron. Its amino-terminal peptide, lactoferricin, also kills Streptococcus mutans and Vibrio cholerae by a chelation-independent mechanism15. Lactoferrin resists digestion by gastrointestinal proteolytic enzymes 8. Consequently, much of the ingested lactoferrin from human milk survives throughout the gastrointestinal tract of infants8,16. Some of this lactoferrin and its fragments are absorbed and excreted in the urine of recipient infants16, where they might also protect against urinary-tract infections. Lytic enzymes. Lysozyme cleaves peptidoglycans in the cell walls of bacterial pathogens. Its concentration is 300-fold greater in human milk than in milk from cows8. The
high levels of lysozyme in human milk8 compensate for the low-level mucosal production of the protein during infancy17. Normal intraluminal concentrations of lysozyme in the tracheobronchial tree and the intestinal tract of infants are therefore dependent on breastfeeding. Immunoglobulins. The main immunoglobulin in human milk is secretory IgA (sIgA). The antibodies of this isotype that are present are specific for numerous enteric and respiratory pathogens8. The immunoglobulin portion of the molecule is produced by plasma cells in the mammary gland, which originate from B cells in the small intestine (entero-mammary pathway)8 and the bronchial tree (bronchomammary pathway)5. In the mammary glands, sIgA molecules are assembled as IgA dimers bound to polymeric immunoglobulin receptors on the basolateral membranes of mammary epithelial cells8. Secretory IgA protects against mucosal pathogens by immobilizing them, by preventing their adherence to epithelial surfaces, or by neutralizing toxins or virulence factors. The provision of sIgA by breastfeeding is important in early infancy, when little sIgA is produced18. Secretory IgA is more resistant than other immunoglobulins to tryptic or chymotryptic digestion in the gut19, and the concentration of IgA2 (the IgA subclass that resists bacterial IgA proteases) is higher in human milk than in blood20. Furthermore, sIgA specific for bacterial IgA proteases is often found in human milk21. Consequently, the concentration of sIgA in stools is much higher in infants that are fed human milk than those fed cows milk8. Oligosaccharides. The oligosaccharides that are present in human milk are produced constitutively. Many function as receptor analogues that inhibit the binding of enteric or respiratory bacterial pathogens, or their toxins, to epithelial cells8. The chemical structure of these oligosaccharides determines the
The relationships between the immune functions of the mammary gland and the immune status of the infant49 are summarized in TABLE 2. These relationships demonstrate the unique immunological adaptations of the mammary gland for the benefit of the infant. So, mammalian milk is species specific, and non-breastfed human infants experience an
Table 2 | Relationships between the immune system of the nursing infant and the immune function of breast milk
Immune effects Compensate directly for developmental delays in the infant immune system Compensate indirectly for developmental delays in the infant immune system Enhance reduced functions, such as specific-antibody production Adapt the gastrointestinal tract to extra-uterine life Prevent inflammation Enhance survival of defence agents Establish commensal bacterial flora Agents in milk Secretory IgA, lactoferrin, lysozyme, IFN- and PAF-acetylhydrolase Oligosaccharides and nucleotides Cytokines and anti-idiotypic antibodies Epithelial growth factors PAF-acetylhydrolase, antioxidants, IL-10 and TGF- Secretory IgA, lactoferrin and lysozyme Growth factors for commensal bacteria
Adapted from REFS 79. IFN, interferon; IL, interleukin; PAF, platelet-activating factor; TGF, transforming growth factor.
566
PERSPECTIVES
and CCL5), TH1-type-response promoting agents (IFN- and IL-12) and TH2-typeresponse promoting agents (IL-6 and IL-10). These immunomodulators have been shown to prime for the production of IFN- in response to infection with respiratory syncytial virus37, to raise the blood levels of fibronectin38, to enhance the systemic39 and mucosal40 production of antibodies and to reduce the risk of necrotizing enterocolitis in infants41.
Role in health and development
Box 1 | Immunological developmental delays compensated by human milk
Antimicrobial
Lactoferrin | lysozyme | sIgA | memory T cells | antibodies specific for T-cell-independent antigens
Anti-inflammatory
Lactoferrin | lysozyme | sIgA | IL-10 | platelet-activating factor-acetylhydrolase
Immunomodulatory
sIgA | interferon- | IL-8 | IL-10
Based on data from REFS 48. IL, interleukin; sIgA, secretory IgA.
specificity of binding to adherence structures of bacteria or their toxins. For example, GM1 gangliosides are receptor analogues for toxins produced by V. cholerae and Escherichia coli 22, whereas globotriaosylceramide binds to the B-subunits of Shiga toxin23. Glycosylated proteins. Certain glycosylated proteins, particularly heavily glycosylated milk mucins 24, interfere with bacterial or viral adherence. Mucin-1 (MUC1), the most abundant mucin in human milk, is present mainly in the membranes of milkfat globules24. MUC1 inhibits the binding of S-fimbriated E. coli to human epithelial cells 25. Another component of milk-fat globules, lactadhedrin26, protects against infection with rotavirus, the most common cause of infectious enteritis in human infants. Antiviral lipids. Other antiviral agents are produced by the digestion of substrates that are present in human milk. These include free fatty acids and monoglycerides, which are generated by the enzymatic digestion of triglycerides in human milk. These small lipids then disrupt enveloped viruses27. Leukocytes. Live activated leukocytes are also present in human milk8. In contrast to B cells, which differentiate into plasma cells that remain sessile in the mammary gland, other leukocytes attracted to this site traverse the mammary epithelium and enter milk secretions. The leukocytes that are present in human milk include neutrophils (4065% of the total leukocyte population), macrophages (3555%) and lymphocytes (510%). The lymphocyte population of human milk mainly consists of activated T cells8. Both CD4+ and CD8+ T cells are found in human milk28, and there is an increased proportion of CD8+ T cells in human milk compared with human blood 28. Following stimulation, milk T cells produce interferon- (IFN-) and chemokines that inhibit macrophage migration and promote the chemotaxis of monocytes8.
In mammals such as sheep, mononuclear leukocytes from milk enter the tissues of the recipients intestinal tract and mesenteric lymph nodes6. Although this has not been demonstrated in humans, cellular immunity seems to be transferred from mother to infant during breastfeeding29. Anti-inflammatory agents. The anti-inflammatory agents that are present in human milk30 include the following: antioxidants (such as ascorbic acid, uric acid, -tocopherol and -carotene); protease inhibitors; enzymes (such as lysozyme, which binds elastin)31; anti-inflammatory cytokines (interleukin 10, IL-10 (REF. 32) and transforming growth factor-, TGF-33); soluble receptors that bind to pro-inflammatory cytokines such as tumour-necrosis factor (TNF)34; enzymes that degrade inflammatory mediators30; epithelial growth factors30; and cytoprotective agents30. The growth factors that are specific for epithelia include epithelial growth factor, lactoferrin, cortisol and polyamines 30. Other hormones and growth factors might also affect the growth and differentiation of epithelial cells, thereby limiting the penetration of antigens and pathogens and affecting other intestinal barrier functions. In that respect, the maturation of the biophysical and/or biochemical organization and function of mucosal barriers in neonates35 is accelerated by the ingestion of human milk36. Immunomodulatory agents. The immunomodulatory agents that are found in human milk include prolactin, anti-idiotypic antibodies, nucleotides that enhance the activity of natural killer cells, macrophages and T helper (TH) cells, and cytokines. These cytokines7,8,32,33 include pro-inflammatory factors (TNF, IL-1 and IL-6), anti-inflammatory factors (TGF- and IL-10), growth promoters (erythropoietin, granulocyte colony-stimulating factor and macrophage colony-stimulating factor), chemokines (IL-8, also known as CXCL8;
Because of these immune factors in human milk and their presumed health benefits, over the past 15 years there has been increased research on breastfeeding and increased evaluation of breastfeeding programmes. The findings of these studies confirm the positive impact of breastfeeding on survival, and growth and development. An analysis prepared to support the Innocenti Declaration on the Protection, Promotion and Support of Breastfeeding (1990)42 indicated that improved breastfeeding practices could save 12 million lives annually43. At that time, the noted benefits of breastfeeding included improved nutrition, enhanced growth and development, reduced severity of illnesses, and social and economic benefits, as well as health benefits for women, such as reduced breast and ovarian cancer, and contribution to birth spacing42. Today, we not only have a better understanding of the importance of these benefits, but we also have further information about the additive benefits of optimal breastfeeding. New studies (both prospective studies and survey research) have further established the importance of breastfeeding, especially optimal breastfeeding, for child survival. Studies continue to show that breastfed infants have considerably fewer episodes of illness4447 and less than half of the mortality rate attributed to common infections (such as diarrhoea or respiratory infections) than do non-breastfed infants48,49. Two different research groups used well-documented data to model the risks of mortality (in 3 and 29 developing countries, respectively), and both groups found independently that the absence of breastfeeding increased of early mortality by nearly sixfold, at 6 months by two- to threefold, and in the second year of life by approximately 20% (REFS 50,80). Two other large meta-analyses underline the importance of early nutrition, especially breastfeeding. Pelletiers work51 demonstrated that approximately 3 million children (of more than 6 months of age) die each year because of malnutrition, much of which could be prevented
VOLUME 4 | JULY 2004 | 5 6 7
PERSPECTIVES
recognized that breastfeeding was a key aspect of self-reliance, primary health care and current development approaches at the joint WHO/United Nations Childrens Fund (UNICEF) meeting on infant and young child feeding. The meeting also recommended that the marketing of breast-milk substitutes and weaning foods should be designed not to discourage breastfeeding. This led to the adoption of the International Code of Marketing of Breast-milk Substitutes by the World Health Assembly in 1981 (REF. 61). One important tenet of the Code (as it is known) is that industry must not market directly to women; this remains a necessary component to reduce bias in the decision to breastfeed62. Although many manufacturers of infant formula report that they support breastfeeding, marketing practices are still observed that might instil doubt among women that they can successfully breastfeed, by noting potential problems or creating the false assumption that every woman will need to use formula at some stage. In the late 1980s and the 1990s, the importance of breastfeeding began to feature in international policy documents (BOX 2). A policy for programme action was issued on 1 August 1990, when the health leaders of 40 countries attended a meeting convened by the WHO, UNICEF, the United States Agency for International Development and the Swedish International Development Cooperation Agency, and signed the Innocenti Declaration on the Protection, Promotion and Support of Breastfeeding42. This declaration calls for the implementation of ten steps to improve hospital practices, which are the central steps of the Baby-friendly Hospital Initiative, and of the International Code of Marketing of Breast-milk Substitutes. The Innocenti Declaration also called on all countries to develop a national policy and a national authority to provide oversight, and to develop innovative legislation to support women in breastfeeding. In 2002, both the WHO and UNICEF supported and endorsed a new strategic plan the Global Strategy for Infant and Young Child Feeding which calls for the revitalization of the Innocenti goals of 1990, as well as additional support for developing a comprehensive policy on infant and young child feeding, in the context of national policies and programmes for nutrition, child and reproductive health, and poverty reduction. This strategy calls upon the health sector and other relevant sectors to protect, promote and support exclusive breastfeeding. It further notes that women should have access to the support they require in the family, community and work place
Box 2 | Major global policies and declarations supporting breastfeeding
World Health Organization (WHO)/United Nations Childrens Fund (UNICEF) International Code of Marketing of Breast-milk Substitutes (1981)61
Limited the influence of commercial marketing on infant-feeding decisions made by women
The Convention on the Rights of the Child (1989)74

Adopted by the General Assembly of the UN Obliged governments to take appropriate measures to ensure that the advantages of breastfeeding can be put into practice
The Innocenti Declaration on the Protection, Promotion and Support of Breastfeeding (1990)42
Pronounced that all women should be enabled to practise exclusive and continued breastfeeding
The Plan of Action adopted at the World Summit for Children (1990)75
Advocated the empowerment of all women to breastfeed their children exclusively for four to six months and to continue breast-feeding, along with complementary food, well into the second year
The World Declaration and Plan of Action for Nutrition (1992)76

Pledged to reduce substantially within this decade social and other impediments to optimal breast-feeding
The International Conference on Population and Development in Cairo (1994)77

Recognized breastfeeding as an important strategy for child survival
The Beijing Platform for Action (1995)78

Called for the promotion of breastfeeding, the implementation of the International Code (1981) and the facilitation of breastfeeding by working women
The World Fit for Children report from the UN General Assembly Special Session on Children (2000)64
Agreed to protect, promote, and support exclusive breastfeeding of infants for six months and continued breastfeeding with safe, appropriate and adequate complementary feeding for up to two years of age or beyond
The WHO/UNICEF Global Strategy for Infant and Young Child Feeding (2002)79
Added emphasis on the need for a comprehensive response (that is, a coordinated, multi-sectoral national response by all concerned parties to the numerous challenges of infant and young child feeding) and added a degree of urgency for implementation
by improving breastfeeding and/or complementary feeding practices. The collaborative report published in The Lancet in 2003 (REF. 1) confirmed earlier work that was carried out in preparation for the Innocenti Declaration42: improving breastfeeding practices could save approximately 1.3 million lives annually (that is, prevent 13% of all child deaths), and improved complementary feeding (during continued breastfeeding) could prevent approximately 600,000 deaths1. From the perspective of growth and development, perhaps the most important findings have arisen from the numerous studies that confirm that breastfeeding is associated with good growth rates and, in fact, can help overcome low birth weights5256. Also, new interest in psychosocial and cognitive development has sparked much research on the relationship between these parameters and breastfeeding. Breastfeeding is associated with measurably improved cognitive development57. A recent meta-analysis, which included studies with
outcomes measured from an early age through to an age of 15 years, concluded that even after adjustment for important cofactors, breastfeeding is associated with significantly higher scores in tests of cognitive development57. There have also been many studies of the costs associated with breastfeeding success58. In addition to the savings generated by reducing morbidity and mortality, studies show considerable savings to families, in terms of food budget59, and to hospitals, in terms of the cost of maternity/newborn care60.
Efforts to support breastfeeding
Breastfeeding rates reached their lowest point in the middle decades of the twentieth century. This was probably because of a combination of effective commercial advertising of cows-milk formula for infants, ill-informed training of health workers, and real or perceived social and work pressures to not breastfeed. In 1979, the international community
568
PERSPECTIVES
to achieve this goal. While continuing support for breastfeeding, this strategy also supports the following: promoting timely, adequate, safe and appropriate complementary feeding (during continued breastfeeding); providing guidance on feeding infants and young children in exceptionally difficult circumstances; and providing guidance on the related support required by mothers, families and other carers. The plan also emphasizes the consideration of legislation, or other new measures that might be required as part of a comprehensive policy on infant and young child feeding, to institute the principles and the aim of the International Code of Marketing of Breastmilk Substitutes, as well as subsequent relevant Health Assembly resolutions. The achievement of these operational goals will demand sustainable commitment and resources to support the operational targets and related activities (BOX 3). In parallel with the evolution of breastfeeding policy, increases in and restructuring of related action programming occurred. In the 1990s, the synergistic combination of the International Code of Marketing of Breastmilk Substitutes and the Baby-friendly Hospital Initiative supported an increase in exclusive breastfeeding. Indeed, exclusive breastfeeding during that period increased in urban areas from 30% to 42% a 40% increase. However, in recent years, the action programming needed for these increases to continue has not been as evident.
Will policy and trends be maintained?
Although previous policies by international organizations remain in effect, it does not seem that the message has been well translated and anchored in the action planning of international development agencies. The UN Millennium Development Goals of 2000 (REF. 63) called for reducing the mortality rate
Box 3 | Activities to improve breastfeeding
Develop national and programme policies and action plan

Develop, implement, monitor and evaluate a comprehensive policy on infant and young child feeding, in the context of both an overall national development strategy and framework, and national policies and programmes for nutrition, child and maternal reproductive health, and poverty reduction.
Design and implement legislation

Support the International Code of Marketing of Breast-milk Substitutes and subsequent relevant World Health Organization (WHO) resolutions (known as the Code) with adequate and well-enforced monitoring and regulation. Also support the recommendations (no. 183 and 191) of the International Labour Organization Maternity Protection Convention (2000). Provide additional support, as needed, to protect the rights of breastfeeding mothers and breastfed children, in all settings.
Create a multi-sectoral coordinator and/or committee

Appoint people with appropriate authority and designated resources to provide oversight and monitoring of the action plan.
Institute baby-friendly practices in all health facilities and training institutions

Enable all maternity services to fully practise the Ten Steps to Successful Breastfeeding (the Baby-friendly Hospital Initiative). All services that have contact with mothers and/or children of 03 years should monitor feeding and support women in achieving optimal practices. Train all health workers in skills to support optimal feeding.
Ensure multi-sectoral and multi-level support for optimal breastfeeding

Build on existing strategies in all health, work place, and civil-society sectors to provide consistent messages in support of optimal breastfeeding and to increase womens and families acceptance, skills, and knowledge of where to seek care if needed.
Support proper complementary foods and feeding

Include information in programmes on the preparation and content of age-appropriate complementary foods, as well as active, frequent and responsive feeding needs, while continuing active support for breastfeeding.
Provide guidance on feeding infants and young children in exceptionally difficult circumstances
Include guidance in national policies on infant and young child feeding in emergencies. Institute guidance and support for infant feeding with respect to HIV/AIDS. Include guidance on the related support required by mothers, families and other carers.
Based on the operational targets of the WHO and United Nations Childrens Fund (UNICEF) Global Strategy.
of children under 5 years of age by twothirds between 2000 and 2015; however, there was no mention of the role breastfeeding would have in achieving that goal. In addition, although the World Fit for Children outcome document (from the UN General Assembly Special Session in 2002) (REF. 64) does state the need for action to protect, promote and support breastfeeding, this is not directly reflected in the major development policy documents of many countries and organizations. Unfortunately, attention to breastfeeding, and the concomitant significant increases in breastfeeding rates are thought to have been diminishing in recent years, perhaps because of misinformation regarding the transmission of HIV and breastfeeding. The extraordinary need to address the problem of HIV, and to prevent every possible case, has eclipsed support for breastfeeding by highlighting the fact that some HIV-positive women pass the virus to their children through breastfeeding. Discouraging breastfeeding was considered by many to provide an easy way to prevent some cases of HIV. However, this limited perspective might have led to an imbalance in the public-health messages on these two issues preventing HIV transmission by reducing breastfeeding versus promoting breastfeeding to reduce overall mortality. Today, it is clear that, although HIV will be passed to some breastfeeding infants, most cases of mother to child transmission occur before birth. Furthermore, ongoing studies conducted in Africa by local and global networks of researchers are beginning to confirm what has been postulated by some publichealth workers: the risk of HIV transmission can be reduced considerably by a combination of interventions, including antiretroviral drugs, maintenance of maternal health and healthier breastfeeding behaviours. For example, exclusive breastfeeding has been associated with reduced transmission through breastfeeding, and maintenance of breast health is associated with lower levels of transmission. Furthermore, in many developing countries, the risk of not breastfeeding might outweigh the risk of HIV transmission, even among known HIV-positive women65. In these situations, the need to continue breastfeeding support for the population is confirmed by the birth spacing provided by the contraceptive effects of lactation66 and the fact that, in developing countries, human milk contributes one-third to two-thirds of the energy requirements of children of more than 6 months of age, which would not easily or feasibly be replaced67.
VOLUME 4 | JULY 2004 | 5 6 9
PERSPECTIVES
Regardless of the presents consensus that breastfeeding needs to continue to be supported despite the HIV/AIDS pandemic, the years of confusion have resulted in decreased momentum and decreased support for optimal breastfeeding. There can be little doubt that commercial interests have had a role in diverting attention from the importance of breastfeeding as a necessary public-health intervention, by focusing on the potential risks of HIV transmission through breastfeeding, without noting the health and financial benefits of breastfeeding that are lost for most children whose mothers are not infected with HIV68. The potential impact of reduced breastfeeding rates on survival, growth, development and population growth is a mounting concern. In developing countries, if breastfeeding rates decline from current levels to the levels of 1990, there would be profound adverse impacts on birth rates and on infant survival, morbidity and growth. Because of the loss of immunological protection, at least a million more children would die in the next decade possibly several million. The increased incidence of diarrhoea and respiratory illness alone would increase family expenses for treatment by approximately twofold. In some countries, birth rates would increase considerably, but many of these children would not survive childhood. Moreover, approximately 50% of the nutritional needs in infancy would be lost, with little available to adequately replace them.
Why has global momentum waned?
which has resulted in high levels of optimal breastfeeding and concomitant returns in good family health. In the United States and other countries, conflicting pressures might lead to a certain level of cognitive dissonance: that is, difficulty in accepting that alternative approaches (such as sufficient paid maternity leave) might be better than the existing ones (early return of mothers to the work place). Commercial marketing is a powerful tool. Because there is no funding available from the public sector to match that of the commercial sector, most marketing remains in support of the use of formula milk, despite the compelling scientific evidence in support of breastfeeding. More than 20 years ago, in 1981, the World Health Assembly recognized this issue and noted that, given the vulnerability of infants in the early months of life and the risks involved in the unnecessary and improper use of breastmilk substitutes, the marketing of breastmilk substitutes requires special treatment. The International Code of Marketing that resulted remains crucial to help protect against these commercial biases. Message burnout. There are many reports of burnout that has resulted from the duration of a concern or from conflicting messages. For example, Ronald Valdiseeri (Deputy Director of the CDCs National Center for HIV, Sexually Transmitted Diseases and Tuberculosis Prevention) commented recently on the resurgence of sexually transmitted diseases in the United States: Some of these people have been hearing these [public-health] messages for three decades now and are suffering from prevention burnout and scepticism. 81 Regarding womens health, it has been noted that, Conflicting public-health messages on health attitudes and preventive practices are evidenced by the contradictory guidelines on mammography screening and cervical cancer screening seen in many countries.82 Message burnout is a real possibility in support for breastfeeding. The Code is nearly 25 years old, and the Baby-friendly concept is more than a decade old. So, it is conceivable that there is donor and public burnout in reaction to breastfeeding-support messages. Competing issues. In addition to message burnout, competing issues especially the HIV/AIDS pandemic and the increased availability of treatments to reduce the transmission of HIV have eclipsed the focus on breastfeeding. There is no doubt that the initial reaction to HIV and its transmission through breastfeeding was one of reasonable
Given the available scientific rationale and the policy documents, we are left with the question of why breastfeeding does not seem to be as central to the international agenda as it merits. At least three possibilities need to be considered: social and commercial pressures; public-health message burnout; and increases in competing issues, such as the misunderstandings about HIV and infant feeding. Social and commercial pressures. Worldwide, many families address daily economic pressures that might conflict with social and health priorities. As a result, there is considerable pressure for increased family income. An early return of mothers to the work place might have negative impacts on the best health choices, which might not be evident on a day-to-day basis or in small populations. Some countries, such as Norway, have addressed the apparent conflict between optimal breastfeeding and family-income generation by providing sufficient maternity leave,
fear, leading to a global reduction in support for breastfeeding69. However, there is now clear agreement across UN agencies that this issue must be viewed in perspective and that the continued promotion of breastfeeding (particularly exclusive breastfeeding for the first 6 months) is a crucial, timely public-health priority70. The challenge is to ensure that, in achieving the desired public-health results, the role of breastfeeding remains central and is given proper consideration, at the same time as related issues gain increased support. Breastfeeding can still be supported in HIV/AIDS programming approaches and in development plans being drawn up by lessdeveloped countries and donors, with the provision of treatment modalities for the many illnesses that are also prevented by breastfeeding. Support for exclusive breastfeeding will increase child survival, even where HIV is endemic, by increasing the survival among HIV-negative infants, by reducing the rate of complicating infections among HIV-infected children71,72 and, if current findings are confirmed73, by decreasing the transmission of HIV from mother to child. The last point is particularly important because most HIVinfected women in developing countries are unaware of their HIV status. Plans and strategies for development in less-developed settings would benefit from the inclusion of breastfeeding support. With its concomitant improvements in survival, and growth and development, breastfeeding improves the overall population potential for development. This might be a new message to build on, to ensure that breastfeeding is more actively included in current development planning mechanisms. The increased availability of medical treatments is, of course, a welcome improvement in public health. However, this sometimes diverts attention from prevention. Since oral-rehydration programming has offered a new route to survive diarrhoeal disease, the importance of the impact of breastfeeding on diarrhoeal disease is not being addressed. Similarly, strategies to tackle respiratory illnesses (for example, 6% of deaths could be prevented by antibiotic treatment1) often fail to mention breastfeeding as an important preventive measure. Scientists and programme planners who address epidemics often see immune-system complements (such as antibiotics) as the first line of defence. Consequently, and in part because of the misuse of such treatments, the incidence of antibiotic-resistant microbial pathogens has increased rapidly. Therefore, this is an
570
PERSPECTIVES
opportune time to emphasize that breastfeeding provides immunological agents that are our best defence against a host of common infections and that infectious agents are less likely to become rapidly resistant to these natural-defence agents. Given the realities of prevention burnout and competing demands, this could be a crucial time to revitalize interest in the importance of breastfeeding, by calling attention to its uniqueness and its irreplaceable immunological role in child health and survival.
Conclusions
There is no doubt that breastfeeding is a costeffective strategy in reducing mortality in preschool children and that attention needs to be paid to its protection, promotion and support. Indeed, the member states of the World Health Assembly and the UNICEF Executive Board both recently confirmed support for the Global Strategy for Infant and Young Child Feeding. This calls on all parties to strengthen activities and to develop new approaches to protect, promote and support exclusive breastfeeding for 6 months (as a global public-health recommendation) and to provide safe and appropriate complementary foods with continued breastfeeding up to 2 years of age or more. This Global Strategy provides guidance on what needs to be done to achieve this (BOX 3). Meanwhile, the HIV and Infant Feeding-Framework for Priority Action states clearly that breastfeeding remains important and that support for it should not be reduced as a result of the HIV pandemic. But breastfeeding has no important commercial advocate. In todays world of publicprivate partnerships, there is nowhere that we can seek unbiased private-sector support for each mother to succeed in breastfeeding. Therefore, we must look to professional, government and non-governmental organizations to send the message and to support breastfeeding programmes. The biologically irreplaceable immunological and immunomodulatory components of human milk might be the key to stimulating support for breastfeeding. If we do not provide support, breastfeeding might diminish because of commercial bias, competing interests, misunderstandings and societal pressures on women. We will have lost crucial, naturally developed, immunological protection for us all, which can extend much further than the period of breastfeeding. Can we afford this loss? Scientists immunologists, epidemiologists and other researchers in the field of human milk and lactation need to continue
to present their work in public forums so that policy leaders and other decision makers become convinced of the superiority of human milk and breastfeeding for the survival, health and nurturance of future generations. When this is recognized, support for breastfeeding can be more firmly anchored in international health and development agendas. If we succeed in presenting the message that there are immunological factors and other crucial factors that are found only in human milk, and not in other infant foods, the number of programmes and resources to actively promote, initiate and support breastfeeding might increase again, thereby saving many lives, reducing suffering in infants, children and their families, and improving the general state of society in all countries.
Miriam H. Labbok and David Clark are at UNICEF/PD/Nutrition, UNICEF House, 3 UN Plaza, East 44th Street, New York, New York 10017, USA. Armond S. Goldman is at the Division of Immunology/Allergy/Rheumatology, Department of Pediatrics, University of Texas Medical Branch, Galveston, Texas 77555, USA. Correspondence to M.H.L. e-mail: mlabbok@unicef.org
doi:10.1038/nri1393
1. 2. Jones, G. et al. How many child deaths can we prevent this year? Lancet 362, 6571 (2003). Dewey, K. G. Cross-cultural patterns of growth and nutritional status of breast-fed infants. Am. J. Clin. Nutr. 67, 1017 (1998). Anderson, J. W., Johnstone, B. M. & Remley, D. T. Breast-feeding and cognitive development: a metaanalysis. Am. J. Clin. Nutr. 70, 525535 (1990). Goldman, A. S., Chheda, S. & Garofalo, R. Evolution of immunological functions of the mammary gland and the postnatal development of immunity. Pediatr. Res. 43, 155162 (1998). Goldman, A. S. Modulation of the gastrointestinal tract of infants by human milk. Interfaces and interactions. An evolutionary perspective. J. Nutr. 130, 426S431S (2000). Goldman, A. S. Evolution of the mammary gland defense system and ontogeny of the immune system. J. Mammary Gland Biol. Neoplasia 7, 277289 (2002). Hanson, L. et al. The immunological role of breast feeding. Pediatr. Allergy Immunol. 12, 1519 (2001). Goldman, A., Chheda, S. & Keeney, S. E. in Fetal and Neonatal Physiology 3rd edn Vol. 2 Ch. 184 (eds Polin, R. A., Fox, W. W. & Abman, S. H.) 20222032 (Elsevier, Philadelphia, 2002). Smith, H. W. & Crabb, W. E. The fecal bacterial flora of animals and man: its development in the young. J. Pathol. Bacteriol. 82, 5366 (1961). Gyrgy, P., Jeanloz, R. W., von Nicolai, H. & Zilliken, F. Undialyzable growth factors for Lactobacillus bifidus var. Pennsylvanicus. Protective effect of sialic acid bound to glycoproteins and oligosaccharides against bacterial degradation. Eur. J. Biochem. 43, 2933 (1974). Bezkorovainy, A. & Topouzian, N. Bifidobacterium bifidus var. Pennsylvanicus growth promoting activity of human milk casein and its derivatives. Int. J. Biochem. 13, 585590 (1981). Liepke, C. et al. Human milk provides peptides highly stimulating the growth of bifidobacteria. Eur. J. Biochem. 269, 712718 (2002). Norton, R. C. & Shohl, A. T. The hydrogen ion concentration of the stools of newborn infants. Am. J. Dis. Child. 32, 183191 (1926). Fransson, G. B. & Lonnerdal, B. Iron in human milk. J. Pediatr. 96, 380384 (1980). Yamauchi, K., Tomita, M., Giehl, T. J. & Ellison, R. T. Antibacterial activity of lactoferrin and a pepsin-derived
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
3.
29.
4.
30.
5.
31.
6.
32. 33.
7. 8.
34.
9.
35.
10.
36.
11.
37.
12.
38.
13.
39.
14. 15.
40.
lactoferrin peptide fragment. Infect. Immun. 61, 719728 (1993). Goldman, A. S., Garza, C., Schanler, R. J. & Goldblum, R. M. Molecular forms of lactoferrin in stool and urine from infants fed human milk. Pediatr. Res. 27, 252255 (1990). Boat, T. F., Kleinerman, J. I., Fanaroff, A. A. & Stern, R. C. Human tracheobronchial secretions: development of mucous glycoprotein and lysozyme-secreting systems. Pediatr. Res. 11, 977980 (1977). Hanson, L. A. et al. Neonatal colonization with Escherichia coli and the ontogeny of the antibody response. Prog. Allergy 33, 4052 (1983). Lindh, E. Increased resistance of immunoglobulin dimers to proteolytic degradation after binding of secretory component. J. Immunol. 114, 284286 (1975). Goldman, A. S. & Goldblum, R. M. in Protein and Nonprotein Nitrogen in Human Milk (eds Atkinson, S. A. & Lonnerdal, B.) 4351 (CRC, Boca Raton, 1989). Gilbert, J. V., Plaut, A. G., Longmaid, B. & Lamm, M. E. Inhibition of bacterial IgA proteases by human secretory IgA and serum. Ann. NY Acad. Sci. 409, 625636 (1983). Laegreid, A. & Kolsto Otnaess, A. B. Trace amounts of ganglioside GM1 in human milk inhibit enterotoxins from Vibrio cholerae and Escherichia coli. Life Sci. 40, 5562 (1987). Newburg, D. S., Ashkenazi, S. & Cleary, T. G. Human milk contains the Shiga toxin and Shiga-like toxin receptor glycolipid Gb3. J. Infect. Dis. 166, 832836 (1992). Schroten, H. in Advances in Nutritional Research Vol. 10 (eds Woodward, B. & Draper, H. H.) 231245 (Plenum, New York, 2001). Schroten, J. et al. Inhibition of adhesion of S-fimbriated Escherichia coli to buccal epithelial cells by human milk fat globule membrane components: a novel aspect of the protective function of mucins in the nonimmunoglobulin fraction. Infect. Immun. 60, 28932899 (1992). Newburg, D. et al. Role of human-milk lactadhedrin in protection against symptomatic rotavirus infection. Lancet 351, 11601164 (1998). Issacs, C. E. in Advances in Nutritional Research Vol. 10 (eds Woodward, B. & Draper, H. H.) 271285 (Plenum, New York, 2001). Wirt, D. P., Adkins, L. T., Palkowetz, K. H., Schmalstieg, F. C. & Goldman, A. S. Activated and memory T lymphocytes in human milk. Cytometry 13, 282290 (1992). Pabst, H. F., Godel, J., Grace, M., Cho, H. & Spady, D. W. Effect of breast-feeding on immune response to BCG vaccination. Lancet 1, 295297 (1989). Goldman, A. S., Goldblum, R. M. & Hanson, L. A. in Antioxidant Nutrients and the Immune Response (eds Bendich, A., Phillips, M. & Tengerdy, R.) 6976 (Plenum, New York, 1989). Park, P. W., Biedermann, K., Mecham, L., Bissett, D. L. & Mecham, R. P. Lysozyme binds to elastin and protects elastin from elastase-mediated degradation. J. Invest. Dermatol. 106, 10751080 (1996). Garofalo, R. et al. Interleukin-10 in human milk. Pediatr. Res. 37, 444449 (1995). Saito, S., Yoshida, M., Ichijo, M., Ishizaka, S. & Tsujii, T. Transforming growth factor- (TGF-) in human milk. Clin. Exp. Immunol. 94, 220224 (1993). Buescher, E. S. & McWilliams-Koeppen, P. Soluble tumor necrosis factor- (TNF-) receptors in human colostrum and milk bind to TNF- and neutralize TNF- bioactivity. Pediatr. Res. 44, 3742 (1998). Biol, M. C., Pintori, S., Mathian, B. & Louisot, P. Dietary regulation of intestinal glycosyl-transferase activities: relation between developmental changes and weaning in rats. J. Nutr. 121, 114125 (1991). Shulman, R. J. et al. Early feeding, antenatal glucocorticoids, and human milk decrease intestinal permeability in preterm infants. Pediatr. Res. 44, 519534 (1998). Chiba, Y. et al. Effect of breast feeding on responses of systemic interferon and virus-specific lymphocyte transformation in infants with respiratory syncytial virus infection. J. Med. Virol. 21, 714 (1987). Friss, H. E., Rubin, L.G., Carsons, S., Baranowski, J. & Lipsitz, P. J. Plasma fibronectin concentrations in breast fed and formula fed neonates. Arch. Dis. Child. 63, 528532 (1988). Stephens, S., Kennedy, C. R., Lakhani, P. K. & Brenner, M. K. In vivo immune responses of breast- and bottle-fed infants to tetanus toxoid antigen and to normal gut flora. Acta Paediatr. Scand. 73, 426432 (1984). Prentice, A. et al. The nutritional role of breast-milk IgA and lactoferrin. Acta Paediatr. Scand. 76, 592598 (1987).
VOLUME 4 | JULY 2004 | 5 7 1
PERSPECTIVES
41. Lucas, A. & Cole, T. J. Breast milk and neonatal necrotising enterocolitis. Lancet 336, 15191523 (1990). 42. World Health Organization/United Nations Childrens Fund. Innocenti Declaration on the Protection, Promotion and Support of Breastfeeding. UNICEF [online], http://www.unicef.org/programme/breastfeeding/innoce nti.htm (1990). 43. Huffman, S., Yeager, B., Levine, R., Shelton, J. & Labbok, M. Breastfeeding Saves Lives: An Estimate of the Impact of Breastfeeding on Infant Mortality in Developing Countries. (NURTURE, Washington DC, 1991). 44. Cushing, A. H. et al. Breastfeeding reduces risk of respiratory illness in infants. Am. J. Epidemiol. 147, 863870 (1998). 45. Perera, B. J., Ganesan, S., Jayarasa, J. & Ranaweera, S.The impact of breastfeeding practices on respiratory and diarrhoeal disease in infancy: a study from Sri Lanka. J. Trop. Pediatr. 45, 115118 (1999). 46. Clemens, J. et al. Early initiation of breastfeeding and the risk of infant diarrhea in rural Egypt. Pediatrics [online], 104, e3 (1999). 47. Kramer, M. S. et al. Promotion of breastfeeding intervention trial (PROBIT): a randomized trial in the Republic of Belarus. JAMA 285, 413420 (2001). 48. Butz, W. P., Habicht, J. P. & DaVanzo, J. Environmental factors in the relationship between breastfeeding and infant mortality: the role of sanitation and water in Malaysia. Am. J. Epidemiol. 119, 516525 (1984). 49. Arifeen, S. et al. Exclusive breastfeeding reduces acute respiratory infection and diarrhea deaths among infants in Dhaka slums. Pediatrics [online], 108, e67 (2001). 50. WHO Collaborative Study Team on the Role of Breastfeeding on the Prevention of Infant Mortality. Effect of breastfeeding on infant and child mortality due to infectious diseases in less developed countries: a pooled analysis. Lancet 355, 451455 (2000). 51. Pelletier, D. & Frongillo, E. Changes in child survival are strongly associated with changes in malnutrition in developing countries. J. Nutr. 133, 107119 (2003). 52. Piwoz, E. G., Creed de Kanashiro, H., Lopez de Romana, G. L., Black, R. E. & Brown, K. H. Feeding practices and growth among low-income Peruvian infants: a comparison of internationally-recommended definitions. Int. J. Epidemiol. 25, 103114 (1996). 53. Villalpando, S. & Lopez-Alarcon, M. Growth faltering is prevented by breastfeeding in underprivileged infants in Mexico City. J. Nutr. 130, 546552 (2000). 54. Froozani, M. D., Permehzadeh, K., Motlagh, A. R. & Golestan, B. Effect of breastfeeding education on the feeding pattern and health of infants in their first 4 months in the Islamic Republic of Iran. Bull. World Health Organ. 77, 381385 (1999). 55. Arifeen, S. E., Black, R. E., Caulfield, L. E., Antelman, G. & Baqui, A.H. Determinants of infant growth in the slums of Dhaka: size and maturity at birth, breastfeeding and morbidity. Eur. J. Clin. Nutr. 55, 167178 (2001). 56. Kramer, M. S. et al. Breastfeeding and infant growth: biology or bias? Pediatrics 110, 343347 (2002). 57. Anderson, J. W., Johnstone, B. M. & Remley, D. T. Breast-feeding and cognitive development: a metaanalysis. Am. J. Clin. Nutr. 70, 525535 (1999). 58. Ball, T. M. & Bennett, D. M. The economic impact of breastfeeding. Pediatr. Clin. North Am. 48, 253262 (2001). 59. Bhatnagar, S., Jain, N. P. & Tiwari, V. K. Cost of infant feeding in exclusive and partially breastfed infants. Indian Pediatr. 33, 655658 (1996). 60. Valdes, V. et al. The impact of a hospital and clinicbased breastfeeding promotion programme in a middle class urban environment. J. Trop. Pediatr. 39, 142151 (1993). 61. World Health Organization. International Code of Marketing of Breast-milk Substitutes. WHO [online], http://www.who.int/nut/documents/code_english.PDF (1981). 62. Allain, A. & Chetley, A. Protecting Infant Health. A Health Workers Guide to the International Code of Marketing of Breastmilk Substitutes 10th edn (International Baby-Food Action Network/International Child Development Centre, Penang, 2002). 63. United Nations. Resolution adopted by the General Assembly. United Nations Millennium Declaration. UN A/RES/55/2. UNICEF [online], http://www.un.org/millennium/declaration/ares552e.pdf (2000). 64. United Nations. Resolution adopted by the General Assembly. A World Fit for Children. UN A/RES/S-27/2, article 37.5. UNICEF [online], http://www.unicef.org/ specialsession/docs_new/documents/A-RES-S27-2E.pdf (2002). 65. Ross, J. & Labbok, M. Modeling the effects of different infant feeding strategies on young child survival and mother-to-child transmission of HIV. Am. J. Public Health (in the press). 66. Becker, S., Rutstein, S. & Labbok, M. H. Estimation of births averted due to breastfeeding and increases in levels of contraception needed to substitute for breastfeeding. J. Biosoc. Sci. 35, 559574 (2003). 67. World Health Organization, United Nations Childrens Fund, Institut Franais de Recherche Scientifique pour le Developpement en Cooperation & University of California at Davis. Complementary Feeding of Young Children in Developing Countries: A Review of Current Scientific Knowledge (World Health Organization, Geneva, 1998). 68. Richter, J. Holding Corporations Accountable: Corporate Conduct, International Codes, and Citizen Action (Zed, London, 2001). 69. Latham, M. & Kisanga, P. Four Country Study of Impact of HIV on Infant Feeding. UNICEF consultants report. UNICEF [online], http://www.unsystem.org/scn/ Publications/AnnualMeeting/SCN28/28breastfeeding.htm 70. World Health Organization, Joint United Nations (UN) Programme on HIV/AIDS, UN Childrens Fund, Food and Agriculture Organization of the UN, UN High Commissioner for Refugees, UN World Food Programme, World Bank, UN Population Fund & International Atomic Energy Agency. HIV and Infant Feeding: Framework for Priority Action (World Health Organization, Geneva, 2003). 71. Coutsoudis, A. et al. Morbidity in children born to women infected with human immunodeficiency virus in South Africa: does mode of feeding matter? Acta Paediatr. 92, 890895 (2003). 72. Phadke, M. A. et al. Replacement-fed infants born to HIV-infected mothers in India have a high early postpartum rate of hospitalization. J. Nutr. 133, 31533157 (2003). 73. Coutsoudis, A. et al. Method of feeding and transmission of HIV-1 from mothers to children by 15 months of age: prospective cohort study from Durban, South Africa. AIDS 15, 379387 (2001). 74. United Nations. Resolution adopted by the General Assembly. Convention on the Rights of the Child. UN A/RES/44/25, annex 44 UN GAOR Supp. No. 49 at 167, UN Doc. A/44/49. UNHCR [online], http://www.unhchr.ch/html/menu3/b/k2crc.htm (1989). 75. United Nations Childrens Fund. Plan of Action for Implementing the World Declaration on the Survival, Protection and Development of Children in the 1990s (United Nations Childrens Fund, New York, 2000). 76. Food and Agriculture Organization of the United Nations. World Declaration and Plan of Action for Nutrition. FAO [online], http://www.fao.org/docrep/U9920t/u9920t0a. htm (1992). 77. United Nations Population Information Network. Report of the International Conference on Population and Development. UN A/CONF.171/13. UN [online], http://www.un.org/popin/icpd/conference/offeng/poa. html (1994). 78. United Nations. Beijing Declaration and Platform for Action. Report of the Fourth World Conference on Women. UN A/CONF.177/20 and A/CONF.177/20/Add.1. UN [online], http://www.un.org/documents/ga/conf177/ aconf177-20en.htm, http://www.un.org/documents/ ga/conf177/aconf177-20add1en.htm (1995). 79. World Health Organization & United Nations Childrens Fund. Global Strategy for Infant and Young Child Feeding (World Health Organization, Geneva, 2003). 80. Rutstein, S. O. in DHS Analytical Studies (ORC Macro, Calverton, Maryland, in the press). 81. Guthrie, P. Georgia leads nation as syphilis increases. The Atlanta Journal-Constitution (Atlanta) (21 Nov 2003). 82. Holroyd, E., Tam, S. M., Marie, L. A. & Li, F. Editorial. Gender Research Centre Newsletter (Apr 2001). Gender Research Centre [online], http://www.cuhk.edu.hk/ hkiaps/grc/activity/pub_news.html
Disclaimer
The statements in this publication are the views of the authors and do not necessarily reflect the policies or views of UNICEF or the University of Texas, United States.

Online links
DATABASES The following terms in this article are linked online to: Entrez Gene: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=gene CCL5 | CD4 | CD8 | CXCL8 | erythropoietin | granulocyte colonystimulating factor | IFN- | IFN- | IL-1 | IL-6 | IL-10 | IL-12 | macrophage colony-stimulating factor | MUC1 | TGF- | TNF Access to this interactive links box is free online.
572

Nri V04 N07 2004

Caricato da

Informazioni sul documento

Descrizione originale:

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Nri V04 N07 2004

Caricato da

Copyright:

Formati disponibili

FOREWORD

TOLL-LIKE RECEPTORS AND THEIR PLACE IN IMMUNOLOGY

| JULY 2004 | VOLUME 4

FOCUS ON TLR SIGNALLING

NATURE REVIEWS | IMMUNOLOGY

VOLUME 4 | JULY 2004 | 4 9 9

Box 1 Box 2 TIR domain

TAB2 TAK1 TAB1

IKK- IKK- IKK- MAP kinases

| JULY 2004 | VOLUME 4

FOCUS ON TLR SIGNALLING

Table 1 | Toll-like receptors and their ligands

TLR8 TLR9 TLR10 TLR11

TLR/IL-1R-superfamily signalling cascade

NATURE REVIEWS | IMMUNOLOGY

VOLUME 4 | JULY 2004 | 5 0 1

SMALL INTERFERING RNAS

| JULY 2004 | VOLUME 4

FOCUS ON TLR SIGNALLING

p65 Late-phase NF-B

Inflammatory cytokines MyD88-dependent response

IFN-, IFN-inducible gene products MyD88-independent response

NATURE REVIEWS | IMMUNOLOGY

VOLUME 4 | JULY 2004 | 5 0 3

IKK- IRAK1 TBK1

IKK- IKK- IKK-

IB p50 p65 Nuclear membrane

| JULY 2004 | VOLUME 4

FOCUS ON TLR SIGNALLING

NATURE REVIEWS | IMMUNOLOGY

VOLUME 4 | JULY 2004 | 5 0 5

| JULY 2004 | VOLUME 4

FOCUS ON TLR SIGNALLING

RIP1 TRIF TBK1 TRAF6

IB p50 p65 IRF3

NATURE REVIEWS | IMMUNOLOGY

VOLUME 4 | JULY 2004 | 5 0 7

ST2 CD14 TLR4 MD2 TIRAP MyD88s MyD88

IKK- IKK- IKK-

| JULY 2004 | VOLUME 4

FOCUS ON TLR SIGNALLING

Negative regulation of TLR signalling

NATURE REVIEWS | IMMUNOLOGY

VOLUME 4 | JULY 2004 | 5 0 9

| JULY 2004 | VOLUME 4

FOCUS ON TLR SIGNALLING

Competing interests statement

NATURE REVIEWS | IMMUNOLOGY

VOLUME 4 | JULY 2004 | 5 1 1

THERAPEUTICS TARGETING THE INNATE IMMUNE SYSTEM

| JULY 2004 | VOLUME 4

FOCUS ON TLR SIGNALLING

Microbial-clearance and killing pathways

Apoptotic and necrotic cell-death pathways

Innate immune system

Microbial-recognition pathways TLR Exogenous ligands NOD protein

Sterile-inflammation pathways TLR Endogenous ligands

NATURE REVIEWS | IMMUNOLOGY

VOLUME 4 | JULY 2004 | 5 1 3

| JULY 2004 | VOLUME 4

FOCUS ON TLR SIGNALLING

IRAK1 RIP1 TRAF6 TRAF6