NEWS AND VIEWS
The transcellular railway: insights into
leukocyte diapedesis
Elisabetta Dejana
Leukocyte recruitment from blood to areas of infection is a key step in both innate and adaptive immunological responses. The
predominant model has been that leukocytes transmigrate through the junctions between adjacent endothelial cells. However,
leukocytes can also migrate through the endothelial cells, and new insights suggest that both caveolae and intermediate filaments
are important for this.
To reach the site of infection, leukocytes need
to adhere to the vessel wall, migrate across the
endothelium and infiltrate into the underlying
tissues. Most of the adhesion molecules that
mediate leukocyte rolling and arrest on the
vascular surface have been identified and stud-
ied in detail. In contrast, leukocyte diapedesis
through the endothelium is still poorly defined
and remains a matter of debate. The most
accepted view is that leukocytes find their way
through endothelial cell–cell junctions which,
acting as sliding doors, would open to allow
their passage and then rapidly close to maintain
the integrity of the endothelium1–4.
However, some studies have challenged this
model and support the idea of an alternative
pathway whereby leukocytes may migrate
through single endothelial cells (reviewed in
ref. 5). This possibility is supported by a series
of morphological studies in vivo. Already in the
1960s, Marchesi and Florey6 described leuko-
cyte penetration through the endothelial cyto-
plasm, and Williamson and Grisham7 observed
leukocytes present in endothelial cytoplasmic
vacuolar structures. Feng et al.8 described the
transcellular passage of neutrophils through
the microvasculature of the skin in response
to chemotactic agents8. However, these papers
were then contradicted by other morphologi-
cal observations showing leukocytes migrating
through inter-endothelial junctions (reviewed
Elisabetta Dejana is in the Department of
Biomolecular Sciences and Biotechnology, Faculty of
Sciences, University of Milan and FIRC Institute of
Molecular Oncology, Milan Italy.
e-mail: elisabetta.dejana@ifom-ieo-campus.it
NATURE CELL BIOLOGY VOLUME 8 | NUMBER 2 | FEBRUARY 2006
PECAM
Blood JAMs
ESAM
..........
Endothelial cells
Basal membrane
Figure 1 Diapedesis of leukocytes through endothelial cell junctions. Leukocytes can cross the
endothelium by interacting with adhesive proteins at junctions. Some have been identified, including
the PECAM (platelet endothelial cell adhesion molecule) and JAM (junctional adhesion molecules)
family of adhesive proteins, as well as other factors such as CD99. For further details, see refs 1–4.
in ref. 5). Thus, the controversy of whether adhesion molecule), which bind beta2 and
migration in vivo uses the ‘para-’ or ‘trans-’ beta1 integrins on leukocytes1–4. These pro-
cellular pathway remains unresolved. Now two teins concentrate in an actin-rich, cup-like
papers in this issue — Millàn et al.9 on page 113 structure formed by endothelial projections
and Nieminen et al.10 on page 156 — analyse that embrace the leukocytes. This ‘docking’
the possible mechanisms of leukocyte transen- structure is enriched not only in adhesive
dothelial migrations further, and bring new proteins but also in cytoskeletal compo-
strength to the idea that the transcellular and nents12,13. Carman and Springer14 previously
paracellular pathways may co-exist. showed that this cup-like structure — the
Once the molecular organization of endothe- transmigratory cup — is important not only
lial junctions became clear11, it became possible for leukocyte arrest but also for their migra-
to develop new tools with which to modulate tion through individual endothelial cells.
leukocyte diapedesis. For example, block- In the first of the new studies, Millan
ing endothelial junctional proteins such as et al.9show that the clustering of ICAM-1
the JAMs (junctional adhesion molecules) or (and to a lesser extent that of VCAM-1) by
PECAM (platelet endothelial cell adhesion specific antibodies induces their redistribu-
molecule) significantly impaired leukocyte tion into caveolin-1-rich regions and their
extravasation, and supported the concept that subsequent internalization. Upon internali-
leukocytes traverse the junctions by binding to zation, ICAM-1 is delivered to the basal side
these proteins1–4 (Fig. 1). of the endothelial cell membrane where it
Firm adhesion of leukocytes to endothelial remains in vesicular structures that have the
cells is mediated by the endothelial adhesion characteristics of caveolae.
molecules ICAM-1 (intercellular adhesion The use of clustering antibodies is a strat-
molecule) and VCAM-1 (vascular cell egy that mimicks leukocyte binding to these
105
©2006 Nature Publishing Group
NEWS AND VIEWS
a

Blood Fusion of VVO?

Endothelial cell

ICAM-1
Basal membrane Caveolin

b
Adhesion/transmigratory cup
Blood

Endothelial cell

ICAM-1
VCAM-1
Basal membrane vimentin
actin?

Figure 2 Leukocytes can cross the endothelium by penetrating the cell
cytoplasm. (a) Leukocytes may actively penetrate the endothelial cell
cytoplasm by elongating pseudopods inside vesicles containing caveolin
and ICAM-1. These vesicles can fuse with vesiculo-vacuolar organelles
(VVOs), forming a channel that allows leukocyte migration through the
endothelial monolayer. (b) When leukocytes adhere to the endothelial
surface, an adhesion/transmigration cup is formed. This docking structure
contains microvilli that elongate from both endothelial cells and leukocytes.
The microvilli contain adhesion molecules (such as ICAM-1 and VCAM-1)
and cytoskeletal proteins (such as vimentin and actin). The adhesion/
transmigratory cup may mediate leukocyte phagocytosis and movement
towards the basal membrane of endothelial cells.

proteins. The investigators, therefore, went on
to study the effect of ICAM-1 engagement by
leukocytes. They found that, after adhering to
endothelial cells, lymphoblasts extend protru-
sions that penetrate into the endothelial cell
cytoplasm. ICAM-1 and caveolin co-cluster in a
‘ring’ around the penetrating lymphoblast pseu-
dopod and follow the passage of the lymphoblast
by moving towards the basal side of the mem-
brane (Fig. 2a). The overall picture is that leuko-
cyte adhesion induces ICAM-1 clustering and
recruitment to caveolae. These, in turn, may fuse
into multivesicular structures that form a sort of
channel through which leukocytes squeeze and
cross the endothelial cell body. These observa-
tions fit with a previous publication8 showing
that leukocytes migrate through the endothe-
lial cell cytoplasm in vivo through multivesicular
structures called vesiculo-vacuolar organelles.
Interestingly, junctional proteins such as
PECAM or VE-cadherin are absent in areas
of ICAM-1 and caveolin clustering, suggesting
that the para- and trans-cellular pathways are
regulated by different molecular mechanisms.
This is further supported by the evidence that
silencing caveolin with short-interfering RNA
(siRNA) only inhibits the transcellular, not the
paracellular, pathway of leukocyte migration.
All the described changes require remod-
elling of the cytoskeleton. Actin seems to be
implicated and it is likely that ERM proteins
(ezrin/radizin and moesin) are also impor-
tant for linking ICAM-1 at, and controlling its
movement on, the cell surface.
In the second study, Nieminen et al.10 put
forward the possibility that intermediate fila-
ments, and in particular vimentin, might also
be important for leukocyte diapedesis. These
authors describe a structure very similar to
the docking/transmigration cup discussed by
Carman and Springer14 (Fig. 2b). They show
that after leukocyte adhesion, both endothe-
lial cells and leukocytes form microvilli that
elongate and embrace the cells. This process
is followed by the penetration of leukocytes
through the endothelial cell cytoplasm. They
found that the microvilli, both at endothelial
and leukocyte sites of contacts, are enriched in
vimentin, but not in actin or tubulin. In addi-
tion, lymphocyte homing at peripheral lymph
nodes and spleen is reduced in vimentin−/−
(vim−/− mice, suggesting that this protein is
necessary for lymphocyte trafficking. This
may be explained by the observation that in
vim−/− endothelial cells, ICAM-1 and VCAM-1
expression and clustering are strongly reduced,
and vim−/− lymphocytes are unable to correctly
polarize and maintain directional movement.
Taken together, these observations point to
a cup-like structure on the endothelium that
embraces leukocytes and directs their passage
through the cell body (Fig. 2b). This process
is reminiscent of cell phagocytosis of foreign
organisms but, whereas phagocytosis is usually
followed by death of the pathogen inside the
cells, leukocytes can traverse the endothelium
without apparent functional alterations.
It is probable that the structures described by
Niemen et al.10 are identical to those reported by
Millan et al.9, but more work on their molecular
organization and functional interaction with the
cytoskeleton is needed to clarify this point.
Several questions remain. It is not clear what
the trigger is that induces the formation of the
transmigratory cup. It is likely that engagement
of ICAM-1 or VCAM-1 by adhering leukocytes
induces their clustering, which in turn medi-
ates re-shaping of the cytoskeleton and forma-
tion of the cup; however, which intracellular
signals are responsible for these changes is not
known. In addition, it remains to be clarified
which cytoskeletal proteins are required and
how they might contribute. Actin, tubulin and
vimentin all seem to be involved in the process.

106 NATURE CELL BIOLOGY VOLUME 8 | NUMBER 2 | FEBRUARY 2006

©2006 Nature Publishing Group


It could be that they each have separate and spe-
cific roles, or because they are interconnected
the disorganization of one type of cytoskeleton
also affects the others.
Another important question is whether
both para- and trans-cellular pathways co-
exist. Only a small proportion (approximately
10–30%) of leukocytes follow the transcellular
pathways; the majority cross junctions9. Millàn
et al.9 show that there are differences when
they compare endothelial cells of different
origin, and that the percentage of leukocytes
migrating through the transcellular pathway
increases using microvascular endothelium.
This suggests that, depending on the region
of the vascular tree, leukocytes may predomi-
nantly use one pathway or the other. It is likely
that where endothelial cell–cell junctions are
particularly tight and well organized (such
as in the brain microcirculation), leukocytes
might preferentially cross the endothelium
through the transcellular pathway. By contrast,
in post-capillary venules (where junctions are
V-ATPase: a potential pH sensor
Chiara Recchi and Philippe Chavrier
An interaction between V-ATPase, a multi-subunit complex responsible for endosome acidification, and ARNO, the GDP/GTP
exchange factor for ARF1 and ARF6, indicates that V-ATPase is the long-sought pH-sensor that regulates trafficking in the
endocytic pathway.
Endocytosis is responsible for internalizing
extracellular fluid and macromolecules, as well
as plasma membrane proteins, receptors and
their bound ligands. While most internalized
surface molecules are recycled, many endocy-
tosed ligands dissociate from their receptors
because of the low endosomal pH and travel
through endosomal compartments for degrada-
tion1. Endosomal acidification is accomplished
by vacuolar (H+)-ATPases (V-ATPases), which
translocate protons across endosomal mem-
branes2. In addition to triggering ligand–recep-
tor dissociation, intra-endosomal acidification
is also known to have an impact on transport
along the endocytic pathway. A pH-sensor that
Chiara Recchi and Philippe Chavrier are in the
Membrane and Cytoskeleton Dynamics Group,
UMR144 CNRS/Institut Curie, Paris, France.
e-mail: philippe.chavrier@curie.fr
NATURE CELL BIOLOGY VOLUME 8 | NUMBER 2 | FEBRUARY 2006
poorly organized), the paracellular pathway
would be favoured.
Niemenen et al.10 show that the choice of
the migratory pathway may also be cell spe-
cific. In their system, only lymphocytes use the
transcellular route, whereas neutrophils cross
intercellular junctions. It is possible that the
conditions used (for example the activation
of endothelial cells or leukocytes), the type of
stimuli or the exposure to flow, influence and
direct the leukocyte’s passage through one or
another pathway.
Together, these two studies provide new tools
and approaches with which to probe this path-
way. But numerous other issues remain to be
investigated. For example, it will be important
to identify experimental methods to selectively
inhibit one, but not the other, pathway. One
possibility, however, is that the two pathways
are interconnected, so that blocking one will
also interfere with the other. On another note,
which is the biological significance of these two
pathways and why do we need both? Could
may couple intra-endosomal pH to the forma-
tion of endocytic transport vesicles has been
postulated, but this mechanism has remained
poorly defined at the molecular level. On page
124, Hurtado-Lorenzo et al. demonstrate that
the V-ATPase interacts directly with the endo-
cytic transport machinery. In addition, through
its pH-dependent interactions with the small
GTP-binding protein ARF6 and its GDP/GTP
exchange factor ARNO, the V-ATPase regu-
lates transport from early to late endosomes,
thus meeting the criteria expected of the pro-
posed pH sensor3.
V-ATPases are large multi-protein complexes
that translocate protons from the cytosol to the
luminal compartment; they consist of two sub-
domains, a peripheral V1 complex that cataly-
ses ATP hydrolysis and a Vo integral complex
composed of five subunits (a, d, c, c′ and c′′)
©2006 Nature Publishing Group
NEWS AND VIEWS
the leukocyte be activated and instructed by
endothelial cells in a different way during pas-
sage through each pathway? By addressing
these issues, the eventual hope is that we might
be able to modulate leukocyte trafficking dur-
ing inflammation and homing.
1. Muller, W. A. Trends Immunol. 24, 326–333 (2003).
2. Johnson-Leger, C., Aurrand-Lions, M. & Imhof, B. A. J.
Cell Sci. 113, 921–933 (2000).
3. Vestweber, D. Nature 421,703–705 (2003).
4. van Buul, J. D. & Hordijk, P. L. Arterioscler. Thromb.
Vasc. Biol. 24, 824–833 (2004).
5. Engelhardt, B. & Wolburg, H. Eur. J. Immunol. 34,
2955–2963 (2004).
6. Marchesi, V. T. & Florey, H. W. Q. J. Exp. Physiol. Cogn.
Med. Sci. 45, 343–348 (1960).
7. Williamson, J. R. and Grisham, J. W. Am. J. Pathol 39,
239–256 (1961).
8. Feng, D. et al. J. Exp. Med 187, 903–915 (1998).
9. Millàn, J. et al. Nature Cell Biol. 8, 113–123 (2006).
10. Nieminen, M. et al. Nature Cell Biol. 156–162
(2006).
11. Dejana, E. Naure. Rev. Mol. Cell. Biol. 5, 261–270
(2004).
12. Barreiro, O. et al. J. Cell Biol. 157,1233–1245
(2002).
13. Yang, L. et al. Blood 106, 584–592 (2005).
14. Carman, C. V. & Springer, T. A. et al. J. Cell Biol. 167,
377–388 (2004).
responsible for H+ translocation2. The activity
of V-ATPases in different compartments of the
endocytic pathway result in a pH gradient that
decreases from pH ~6.0 in early endosomes to
pH 5.0–5.5 in lysosomes (Fig. 1). Intra-endo-
somal acidification is required for the dissocia-
tion of ligands bound to internalized receptors
in early endosomes, the release of lysosomal
enzymes from mannose 6-phosphate receptors
in late endosomes, and the enzymatic activity
of hydrolytic enzymes2. Specific inhibitors of
the V-ATPase proton pump, including bafilo-
mycin, prevent endosomal acidification and
hence interfere with endocytic functions.
Moreover, inhibiting intra-endosomal
acidification also inhibits transport along the
endocytic pathway. Indeed, several laboratories
have previously observed that in bafilomycin-
treated cells, transport of endocytic markers
107