Reviews

DNA microarrays: translation of the genome from laboratory to clinic

Daniel H Geschwind
As the complete sequences of human and other mammalian genomes become available we are faced with the challenge of understanding how variation in sequence and gene expression contributes to neurological and psychiatric disorders. DNA microarrays, or DNA chips, provide the means to measure simultaneously where and when thousands of genes are expressed. Microarrays are changing the way that researchers approach work at the bench and have already yielded new insights into brain tumours, multiple sclerosis, acute neurological insults such as stroke and seizures, and schizophrenia. The study of disease-related changes in gene expression is the first step in the long process in translation of genome research to the clinic. Eventually, the changes observed in microarray studies will need to be independently confirmed and we wil need to understand how gene expression changes translate into functional effects at the cellular level in the nervous system. Progress in these studies will translate into arraybased disease classification schemes and help optimise therapy for individual patients based on gene expression patterns or their genetic background. Lancet Neurology 2003; 2: 27582

When put in context, the pace of progress in genetics has been astounding. William Bateson coined the term gene around the start of the 20th century, the discovery that chromosomes contained the genetic material of the cell occurred in the next decade, and the structure of DNA was published by Watson and Crick in 1953. The sequencing of the entire human and mouse genomes less than 50 years after the structure of DNA was first elucidated is even more astounding. Most of the success in the search for genetic links to human diseases has been in rare disorders with mendelian inheritance, which has led to major advances in our understanding of several common neurological diseases, such as Alzheimers disease,1,2 and those that are rarer, such as triplet repeat diseases.3,4 Although the identification of these genes has not yet led to major clinically applicable advances in therapy, the fulfilment of this promise is around the corner. Now we face the challenge of understanding how genes contribute to normal human development, individual variability, and common diseases.

human genome codes for expressed genes, via messenger RNA (mRNA), which is translated into protein. The rest of the genome is repeat sequences, regulatory regions, or unique non-coding sequences with unknown function. The first estimates suggested that there were about 30 000 genes in the human genometwice the number in Drosophila.5,6 Given the relatively modest increase in the number of genes expressed compared with increases in size and complexity from Drosophila to mouse to human being, the timing and levels of expression and alternative splicing of these genes clearly underlie the differential functioning of the human body and nervous system. The next stage in the quest to understand how genes contribute to normal neurological development and disease is to elucidate gene expression patterns and how they are affected in disease. To gain this understanding many genes will need to be studied at the translational (protein) and post-translational level (eg, protein phosphorylation or glycosylation). Several rapidly developing areas of technology make genome-level analysis of neurological and psychiatric diseases possible. These methods complement the traditional methods, such as northern blotting or in situ hybridisation, used to study one gene at a time. High throughput techniques similar to microarrays have been introduced into studies at the protein level (proteomics; figure 1). Each method of investigation has its limits and must be interpreted properly. For example, although three genes in a particular metabolic pathway may show changes at the RNA level, the actual function of that pathway may not be affected if none of these genes are the rate-limiting factor. Hence, these highthroughput genome-level technologies must be interpreted within the appropriate biological context. Complementary DNA (cDNA) microarray technology allows inexpensive study of all genes expressed in the genome (the transcriptome) in parallel. Although this technology is new, it has already led to significant research advances and its use is spreading rapidly (figure 2).7 In basic neuroscience, arrays have been used to explore differences in gene expression among mouse strains and areas of the brain,811 neural stem cell biology,1214 and the downstream effects of single-gene defects in mouse models of neurodevelopmental disorders.15,16 This review summarises the current state of
DHG is at the Program in Neurogenetics, Department of Neurology, and the Center for Neurobehavioral Genetics, Neuropsychiatric Institute, the David Geffen School of Medicine, University of California, Los Angeles, California, USA. Correspondence: Dr Daniel H Geschwind, Department of Neurology, University of California, 710 Westwood Plaza, Los Angeles, California, USA. Tel +1 310 206 6814; fax +1 310 267 2401; email dhg@ucla.edu

Gene expression
A genome is all of the genetic material (DNA) in an organism. Each cell within an organism has a copy of the same genome, identical apart from changes in individual cells that are caused by environmental factors. 12% of the
THE LANCET Neurology Vol 2 May 2003

http://neurology.thelancet.com

275

For personal use. Only reproduce with permission from The Lancet Publishing Group.

Review

DNA microarrays

AAAAA
AAA

AAAAAAA Genome Transcriptome Sequence mRNA concentrations Gene dosage mRNA localisation Chromatin structure Epigenetic modification Proteome Protein abundance Post-translational modification Proteinprotein interactions

Alzheimers diseaseeach labelled with a different fluorophore for detectionthe relative amounts of all the genes present on the array and expressed in the two samples can be measured. If a high-density array could be made of all of the expressed genes and expressed sequence tags (ESTs) in the brain, all of the important disease-causing genes could be identified.20
Array platforms

Figure 1. Levels of biological system understanding from the genome to the proteome. Disease can be studied at the genome, transcriptome, or proteome levels, all of which need to be connected to fully understand diseases with genetic contributions.

microarray technology and presents examples of its application to several common neurological and psychiatric disorders of human beings.

DNA microarrays
DNA microarraysalso called gene chipsconsist of DNA attached to a surface in an ordered, predetermined fashion at extremely high density.17,18 More than 20 000 genes (a half of expressed human genes) can be arrayed on a surface the size of a glass microscope slide. Microarrays allow the monitoring of gene expression for thousands or tens-of-thousands of genes simultaneously in one hybridisation experiment because of two basic principles of nucleic-acid hybridisation: DNA and RNA will specifically bind to their complementary sequence, and this binding happens in proportion to the abundance of a sequence in a mixture. The aim of a typical microarray experiment (figure 3)19 is to measure the amount of a given mRNA species (transcribed gene) in a tissue or cell type. To do this, the mRNA is normally transformed by reverse transcription into cDNA, which is more stable. mRNA or cDNA from a sample (eg, brain tissue from patients with Alzheimers disease) is applied to the array surface and allowed to hybridise. A direct measure of how much of each gene on the array is present in the tissue is obtained by the measurement of the amount of cDNA bound to each spot on the array. The more abundant a given mRNA (or cDNA) is in a sample, the more binding to its complement will occur. Thus, by comparison of the hybridisation of cDNA from a control individual with that from a patient with
100 90 80 70 60 50 40 30 20 10 0 1996 1997 1998 1999 2000 2001 2002 Year
Figure 2. PubMed citations containing the words brain and microarray have increased rapidly in the past few years, which shows acceptance of this technology by clinical and basic neuroscientists.

The two main forms of microarray are cDNA and oligonucleotide arrays.18,2123 cDNA arrays, which were originally developed in the Brown Laboratory at Stanford University,17,21 are composed of PCR-amplified cDNA clones arranged on a non-porous surface. A typical cDNA array is printed onto a 30 mm by 15 mm glass microscope slide by a computercontrolled robotic cantilever arm; each spot is about 50150 m in diameter. cDNA clone sets of 15 00020 000 mouse and 40 000 human genes are available for in-house arrays in addition to ready-made commercial chip sets. About 80% of all human genes are identified (the number increases daily) and can be queried on about two slides printed at high density. Other flexible and porous surfaces such as nylon membranes can be used as alternatives to glass.7,24,25 Arrays of short oligonucleotides (about 25 bp) made by photolithography in situ are also available commercially and form the basis of GeneChip technology (Affymetrix, Inc, Santa Clara, CA, USA).18 Arrays of longer oligonucleotides manufactured in situ by use of ink jet printing technology are available through Agilent Technologies (Palo Alto, CA, USA).26 Many investigators use longer (6070 bp) oligonucleotide arrays (eg, Codelink arrays; Amersham Biosciences, Piscataway, NJ, USA) that are made in a way similar to cDNA arrays. These longer oligonucleotide arrays have the advantage of not requiring the laborious process of PCR-amplification and clone-insert purification before arraying. Technical issues, such as optimisation of slide surface and oligonucleotide attachment, and the design of the most specific oligonucleotide probe(s) for each gene, are recurrent challenges. Cross-platform comparisons often yield significantly non-overlapping results, and the source of these discrepancies is unknown. Other techniques for the measurement of gene expression that involve large scale sequencing, such as serial analysis of gene expression (SAGE) and massively parallel signature sequencing (MPSS), can be more sensitive.2730 These methods involve highly efficient identification by sequencing of short fragments and counting of cDNA clones that represent the mRNA expressed in a particular cell or tissue and may have improved sensitivity. However, neither technique is as rapid as a microarray and, therefore, they are less likely to be useful in clinical settings.
Experimental design

Number of citations

The design and implementation of microarray experiments demand special attention, because small variations in conditions can induce significant changes in gene expression.31 For human tissue, these variables include tissue preservation

276

THE LANCET Neurology Vol 2 May 2003

http://neurology.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.

DNA microarrays

Review
Sample preparation False colour image representing greyscale values

methods, postmortem interval, dissection methods, and RNA quality.3133 Frozen tissue generally provides the best material, although ethanol fixation can be used. Differences in gene expression between samples may be due to differences in genetic background (ethnicity).8,34 Other general sources of variability include cDNA amplification methods, probe labelling, hybridisation conditions, and washing.7,31 However, when these issues are taken into account in the experimental design and proper technique is applied, microarray experiments provide reliable data on gene expression at a systems level.3438 The current limitations of microarray detection mean that some low abundance genes will not be detected in heterogeneous tissues such as the CNS.34,37,39 Methods such as single-cell PCR4042 and laser capture microdissection43,44 allow for the capture of cDNA from single cells and the study of gene expression in a single cell or group of cells. These methods could facilitate the study of epileptic foci, tuberous-sclerosis lesions,40 and other focal or cell-type specific lesions resulting from development and disease.42
Uses in neuropsychiatric disease

Wild-type mouse

Spotted mouse

B Data analysis

RNA extraction, labelling

cDNA Hybridisation

cDNA

Data mining/database

Image scanning Detectors Laser Slide

Studies of disease with cDNA microarray technology can be split into two main categories with interrelated goals: identification of key molecular changes in diseases and identification of biomarkers or molecular fingerprints that will aid in patient diagnosis and classification. Studies that identify molecular changes in disease will advance our understanding of disease pathophysiology, whereas studies that identify biomarkers will improve diagnostic accuracy and targeting of specific therapeutic interventions. In the discussion that follows, I do not summarise every paper that has used microarrays to study neurological disease but focus on a few topics that exemplify progress in the understanding of disease pathophysiology or identification of biomarkers for diagnostic and treatment purposes.

Figure 3. Experimental stages in a typical cDNA microarray experiment. RNA is extracted from the two comparison specimens (eg, spotted and non-spotted mice) and reverse transcribed to cDNA which is labelled with different colours. The two samples are cohybridised onto the array, the dyes are excited by a laser and the scattered light is collected and analysed. The images produced by each dye are registered and a false colour overlay is produced. In other platforms, such as Affymetrix or radioactively probed arrays, only one sample is hybridised onto each array and comparisons are done between arrays. Data visualisation and exploration algorithms such as clustering or principalcomponents analysis are applied. Reprinted with permission from Nature Reviews Neuroscience, Macmillan Press (www.nature.com/reviews).19

approach to standard positional cloning by assaying both the effects of genetic background and the environmental factors that lead to disease. In most common diseases, which are due to a combination of genetic and environmental susceptibility, these approaches may offer a method of identifying novel pathways involved in disease pathophysiology with the hope that these pathways will be good targets for therapy.
Multiple sclerosis

Expression changes relevant to disease pathophysiology: multiple sclerosis and schizophrenia

Complex or sporadic diseases are ideal targets for microarray analysis because other genetic approaches, although direct, require enormous coordinated effort from many investigators.4548 The use of arrays offers a complementary
THE LANCET Neurology Vol 2 May 2003

Whitney and colleagues24 did one of the earliest studies of a neurological disease in human beings with custom printed microarrays. By comparing gene expression in normal white matter with that in acute lesions in brain tissue from one patient with multiple sclerosis, the researchers identified many genes that were either upregulated or downregulated in plaques. Whether these changes are observed in other patients with multiple sclerosis and how these changes are functionally related to disease is unknown. This study, however, was a starting point for subsequent research.

http://neurology.thelancet.com

277

For personal use. Only reproduce with permission from The Lancet Publishing Group.

Review
A recent example of how knowledge of which genes have abnormal expression in patients with a complex disease (such as multiple sclerosis) can lead to the identification of treatment targets and validation of these targets in a model system, was recently published by Lock and colleagues.49 These researchers used Affymetrix GeneChip microarrays to study postmortem tissue from four women with multiple sclerosis in comparison with control samples from one man and one woman. About 90 genes with abnormal expression in sclerotic plaques were identified; changes specific to acute inflamed lesions and chronic scarred lesions were also found. On the basis of their microarray experiment, these investigators chose two targets and tested their validity in a mouse model of multiple sclerosisexperimental autoimmune encephalomyelitis. Acute disease was less severe, and chronic disease was absent in mice deficient in the immunoglobulin Fc-receptor; this protein was upregulated in the chronic lesions obtained from brain tissue of patients with multiple sclerosis. The second protein, granulocyte-colony stimulating factor, which is upregulated in acute multiple sclerosis lesions, decreased the severity of the early stages of experimental autoimmune encephalomyelitis in mice when it was given before disease onset.49 Lock and colleagues did not study many patients or do many replicates. However, it is a landmark paper since it is the first to link the assessment of gene expression with microarrays to candidate pathway validation in an animal model, effectively realising the potential of pharmacogenomics. The logical extension of this work would be to determine how the changes in gene expression identified by microarrays can be generalised to large numbers of patients with multiple sclerosis. Tissue microarrays, where several hundred samples can be studied in parallel, provide an efficient platform for preliminary target validation.50,51 The studies by Lock and colleagues49 and Whitney and colleagues24 provide a solid foundation for studies of gene expression to expand into larger population studies, animal models,52 and eventually clinical trials based on the identified targets.
Schizophrenia

DNA microarrays

Some of the most influential early research with microarrays in the study of human neuropsychiatric disease involved the assessment of gene expression in archived frozen postmortem tissue from community-dwelling patients with schizophrenia. Mirnics and colleagues53 extracted RNA from ten patients with schizophrenia and ten control individuals matched for age, sex, and postmortem interval. The investigators used cDNA arrays prepared by Incyte Genomics Inc (Palo Alto, CA, USA)37 to do pairwise comparisons of samples from the frontal lobes of patients with schizophrenia with samples from control individuals. Many changes were identified and the investigators were faced with the daunting task of making functional links between genes. Work on the functional links between individual genes is ongoing, but two common findings have emerged from the analysis so far. The first is that the expression of synapsin and genes involved in presynaptic function and those involved in postsynaptic responsiveness (eg, regulator of G-protein signalling 4 [RGS4]; which modulates G-protein coupled

neurotransmission) is downregulated.54 The second finding is a downregulation of five specific metabolic pathways that were previously not known to be involved in this disease.55 To ensure that the changes in people with schizophrenia were not related to drug effects, gene expression in frozen brain tissue from haloperidol-treated monkeys was compared with that in untreated control animals to identify changes in gene expression related to drug treatment.53 The work of Mirnics and colleagues has several strengths: many replicates; quality postmortem tissue from a large number of patients; confirmation of results by use of alternative methods, such as reverse transcriptase PCR and northern and western blotting; and attempts to interpret gene lists in the context of biological processes. By use of Affymetrix oligonucleotide arrays, Hakak and co-workers56 studied gene expression in dorsolateral prefrontal cortex (area 46) from the postmortem cerebral cortex of patients with chronic schizophrenia and control individuals. Age in both groups was similar, but the sex ratios differed, which was a potentially confounding factor. All patients had long-term neuroleptic exposure, although several of the patients were off treatment for various periods before the study.56 The investigators emphasise changes in oligodendrocyte-related proteins (a finding which was not observed in the study by Mirnics and colleagues) and highlight a possible defect in myelination in schizophrenia. The differences between the Hakak and co-workers study and the Mirnics and colleagues study may reflect differences in the arrays (eg, different sets of genes on each array) or in the patient populationsone of chronically institutionalised patients56 and the other of community-dwelling younger adults.53 These discrepancies highlight three important issues in studies of complex neurological or psychiatric diseases with microarrays. First, care must be taken not to overgeneralise on the basis of findings in a single group of patients because of genetic heterogeneity or differences in study design. Second, different microarrays contain different genes and different probes for a given gene, which confounds cross platform comparisons. Third, the study of different brain regions at different time points during the course of a disease will yield different answers. The first issue can be addressed by replication of the results with independent samples and methods. The second issue will diminish as cost-effective, whole genome arrays become more widely available or as investigators studying a particular disease agree to use the same platform across studies. The third issue will yield to time and experiencediscovery of which changes in which region are most relevant to the behavioural or cognitive disease phenotype in question will require years of follow-up work at the systems level. In addition, although disease-specific changes in gene expression in the brain can clearly be identified in complex neuropsychiatric diseases, identification of the initial causal changes in gene expression in neurodevelopment might not be possible in diseases such as autism, or perhaps even schizophrenia, since these changes occur quite early and tissue may not be available. All of these issues highlight the importance of careful experimental design and interpretation of results.

278

THE LANCET Neurology Vol 2 May 2003

http://neurology.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.

DNA microarrays

Review
workers62 studied primary motor cortex and an unspecified region of frontal cortex by comparison of three pools of cDNA from people with chronic alcoholism with three control pools on Unigem V2.0 cDNA arrays (Incyte Genomics). The case and control samples were carefully matched for age, gender, and time post mortem and cases with concomitant cirrhosis or drug abuse were omitted. Genes related to myelin and protein trafficking, as well as known alcohol-responsive genes, were among the most significant classes of genes with abnormal expression. No independent confirmation of the results by alternative methods, such as reverse-transcriptase PCR or northern blotting, was presented. Pooled sample design allows the use of fewer arrays; however, pooling five or six individuals could easily result in outliers owing to aberrations in one sample and this method somewhat mitigates the value of careful matching of cases with controls. Furthermore, the ability to manage the inherent heterogeneity in clinical response and identify subtypes of responses is lost in pooled studies. Additionally, which feature of alcoholism, such as susceptibility, withdrawal, seizures, or minor head trauma, these changes in gene expression are related to is also unknown. Lastly, although both studies identified changes in genes involved in the cell cycle, whether this effect represents increased proliferation of glia or inflammatory cells, abortive cell-cycle re-entry in dying neurons, or another process entirely, is unknown, because no follow up at the anatomical level is presented.61,62 Despite these limitations, these studies provide a starting point for future investigations.

Gene expression in autism and alcohol abuse: cause or sequelae?

Autism

Autism is a neurodevelopmental disorder characterised by deficits in language and social communication.57 Although there is a strong genetic risk for autism and related disorders, the genetics are complex with epigenetic influences.58 Therefore, microarrays represent an efficient manner in which to identify potential candidate pathways. Purcell and colleagues59 compared autopsy samples of the cerebellar cortex, the prefrontal cortex, and the caudate-putamen from patients with autism with those from healthy control individuals matched for age and sex, although the cause of death and other clinical features were not homogeneous. Genes were chosen on the basis of their mean expression ratios, rather than statistical criteria, and reverse-transcriptase PCR was used to confirm the results. Several genes related to neural transmission, including glutamate receptor AMPA1, the excitatory amino-acid transporter EAAT1, and the cytoskeletal protein glial fibrillary acidic protein, had similar increases in the brain samples from patients with autism on both array platforms; their identities were confirmed by reverse transcriptase PCR.59 Changes in mRNA were supported by changes at the protein level, and receptor autoradiography confirmed the changes in AMPA receptor density in the cerebellum of patients with autism; these changes were not observed in other brain regions. The consequence of the change in EAAT1 expression is unknown, but this could simply reflect seizures in a subset of the patients. Larger studies are need to take genetic heterogeneity into account,34 and need to be followed up at the cellular level in more brain regions to fully assess their biological meaning. Increased expression of glial fibrillary acidic protein could reflect inflammation and gliosis, which suggests an ongoing process rather than a monophasic early developmental injury. Similar observations were made in an analysis of Retts syndrome by the same investigators.60 One issue that potentially clouds the interpretation of gene expression studies of neurodevelopmental disorders is that such studies are commonly limited to tissue obtained late in the disease process; detection of the causal gene expression changes that occur early in the disease process may not be possible. Although animal models of complex neurobehavioural diseases such as autism are necessarily restricted, they are the only way to identify some early developmental changes that are hidden in human beings. A combination of gene expression studies in human beings and animals may be necessary to identify the early causal changes in gene expression in autism and related developmental disorders.
Alcoholism

Patient classification in acute neurological disease and cancer

Biomarkers of acute neurological disease

Two studies of gene expression in postmortem tissue from people with alcoholism identified downregulation of myelinrelated genes that could cause white-matter abnormalities associated with chronic alcohol abuse. Lewohl and colleagues61 studied postmortem tissue from the frontal lobes of people with alcoholism and identified changes in about 160 genes with both oligonucleotide and cDNA arrays. Mayfield and coTHE LANCET Neurology Vol 2 May 2003

To answer the critical question of whether the pattern of genes expressed in peripheral-blood leucocytes mirror acute CNS injury, Tang and colleagues63 studied gene expression patterns in rats subjected to stroke, hypoxia, hypoglycaemia, and seizures. The researchers were able to identify specific patterns of gene expression for each of the four injuries 24 h after the insult. These results provide proof of principle that complex patterns of peripheral-blood gene expression could accurately reflect a previous acute brain injury. Whether these patterns are useful in diagnosis of human patients in clinical settings is unknown. Such gene expression information might be used as additional data to improve diagnostic certainty in patients in an acute setting. Although the diagnosis of acute episodic or short-lived neurological dysfunction, such as transient ischaemic attack or seizure, may be challenging in certain cases, differentiation between chronic or subacute disorders in their early stages provides a different challenge. Gene expression in peripheral blood might allow differentiation between patients with Alzheimers disease, Lewy-body disease, Creutzfeldt-Jakob disease, or frontotemporal dementia, or allow for the early diagnosis of Parkinsons disease or multisystem atrophy. Whether a subset of the gene-expression abnormalities seen in the studies of schizophrenia or multiple sclerosis could be identified in peripheral-blood leucocytes or other peripheral cells from patients is an important clinical issue. Careful pilot

http://neurology.thelancet.com

279

For personal use. Only reproduce with permission from The Lancet Publishing Group.

Review
studies of gene expression in patients with chronic neurological disease are needed to determine whether this is worth pursuing.
Brain tumours

DNA microarrays

Microarrays were originally greeted with scepticism by many neuroscientists and were more readily adopted by oncologists. Thus, some of the clearest successes of microarrays in the classification of disease states on molecular bases have been in cancer studies. One obvious problem in neuro-oncology is that the response of brain tumours to chemotherapy is quite variable, even in tumours with apparently identical histological and molecular features. Some of this variability may be due to sampling artefacts that occur within biopsy specimens. However, it is likely that response and prognosis depend upon specific genetic pathways that are active in a particular tumour. Studies in various brain tumours have shown the ability of cDNA expression microarrays to distinguish tumour types.64,65 This technology makes possible the identification of geneexpression profiles associated with prognosis and specific treatment response. In addition, with the identification of novel patterns of gene expression associated with certain tumour behaviours, new therapeutic targets can be assessed and targeted therapies can be developed. Pomeroy and colleagues65 used oligonucleotide microarrays to differentiate gene expression patterns in tumour samples from 42 patients to identify molecular signatures for primitive neuroectodermal tumours, gliomas, and medulloblastomas; each of these tumours had a distinct

pattern of gene expression. Another group of tumours atypical teratoid and rhabdoid tumourswas similarly distinguished from the other tumour types. This finding, and the ability to separate subtypes of medulloblastoma that are histologically distinguishable, should not be entirely surprising because pathologists can currently distinguish these subtypes with less technology. Next, to determine whether gene expression could be used to classify prognostic groups among patients with medulloblastoma, the investigators studied 60 patients. Gene expression was shown to significantly support current predictors such as expression of a single growth-factor receptor or clinical stage. This striking success has led to great optimism; however, several issues need to be addressed. 5 year survival was the same in both the poor and the good outcome groups, so the usefulness of this scheme in assessment of individual patients is unclear. Furthermore, although the classification seemed powerful in this group, its predictive value needs to be validated in an independent cohort. These gene expression studies do provide a proof of principle. In addition, these results and those of other studies6466 have improved our understanding of the molecular pathology of these brain tumour types by the identification of many previously unknown changes in gene expression.

Gene dose and molecular karyotyping

The microarray application closest to widespread clinical use in diagnostic laboratories is probably the assessment of chromosome ploidy and gene dose (copy number). Arrays of segments of the genome at high density have been used to measure gene dose of specific DNA regions67,68 and across the genome.6972 Such arrays allow high density mapping of chromosome copy number A B (comparative genomic hybridisation) and, within chromosomes, a determination of Chromosomes Chromosomes specific regional copy number. This could 7 7 7 7q deletion allow detection of subtle or moderate microdeletions associated with developx y x x y mental disorders without having to assume a specific genotypephenotype relation. DNA Typically, cytogenetic studies are done with a DNA few probes at a time, which enables a few Cy5 Cy3 chromosomal regions to be assayed. With comprehensive array comparative genomic Cy3 Cy5 hybridisation, the entire genome can be assessed at a resolution of a few hundred Gene dose x x Duplicated region kilobases at one time in a single at each y y 17 region on hybridisation, a process that would take 17 7 7 each thousands of hybridisations with convenchromosome tional techniques for molecular cytogenetic Deleted region studies. Comparative genome hybridisation Figure 4. Measurement of gene dose with microarray. Left: the array is used to detect changes is likely to have a substantial effect on in chromosome number, such as that observed in Klinefelter syndrome (XXY) in males. The prenatal diagnosis and chromosomal studies relative hybridisation of samples from the normal individual (Cy3-green) is compared with those in other neurological settings (figure 4). from the aneuploidic patient (Cy5-red) on an array containing large genomic clones (eg, bacterial
artificial chromosomes [BACs]) from the sex chromosomes and autosomes and an increase in red signal on the X chromosome clones is detected. Right: a patient with a deletion is compared to a normal individual. The deleted region, which covers three of the BAC clones on the array, is identified by a decrease in signal (red) from the patient, which is seen as a relative increase in green signal. Such high-density arrays (500010 000 BAC clones) can detect small deletions that have previously gone undetected in most studies. This method is an efficient and inexpensive way of scanning the genome for changes in gene dose.

The promise of pharmacogenomics

The ability to optimise individual therapy for a particular neurological disease differs strategically from the use of pharmacogenomics in cancer therapy, in which

280

THE LANCET Neurology Vol 2 May 2003

http://neurology.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.

DNA microarrays

Review

Patients with Patients with Patients with microarrays are used to molecularly classify Alzheimers disease Alzheimers disease frontotemporal tumour tissue obtained from a patient. In dementia individualised therapy, the patients normal biological make-up or genetic background is assessed to identify potential biomarkers that favour one therapy over another, or to avoid adverse effects of medication (figure 5). Treatment Array technologies allow researchers to study common polymorphisms in the large numbers Responders WBC WBC Non-responders (red) of patients needed to find critical DNA (green) variants.73,74 For example, a similar kind of array technology to that used for gene expression can be used to sequence large portions of an individuals genome and thus identify single WBC WBC A molecular fingerprint with nucleotide polymorphisms that may underlie (red) (green) potential diagnostic value disease susceptibility or medication sensitivity. An arrayof either DNA fragments or oligonucleotidesthat covers the regions or genes of interest with some redundancy is made, which enables complete sequencing of these A molecular fingerprint with potential predictive value genes in one hybridisation step.75 In many cases Figure 5. The use of microarrays in disease classification. Left: microarrays can be used the hybridisation is coupled to a ligation or single base pair extension reaction to increase to detect single nucleotide polymorphisms (SNPs) or identify gene expression patterns that predict treatment outcome or adverse affects in patients. Oligonucleotide arrays fidelity. Currently, this would be limited to a that allow polymorphism detection in a number of candidate, treatment-modifying few genes at most, but high-throughput genes simultaneously, such as cytochrome p450, will soon be widely adopted. Right: single-nucleotide-polymorphism genotyping microarrays may also be used to support diagnostic protocols. In the speculative technology is changing rapidly, and many new example shown, mRNA from peripheral-blood leucocytes is hybridised onto a cDNA platforms are being developed.76,77 Similarly, once array and the pattern indicates whether the patient is more likely to have Alzheimers disease or frontotemporal dementia. Perhaps, more readily adopted into clinical single nucleotide polymorphisms related to practice will be the use of arrays to classify tumour samples from patients to help treatment response or adverse effects are diagnosis and to predict optimal treatment. WBC=white blood cells. identified, many individual polymorphisms in genes of interest can be screened to help avoid serious Conclusion adverse drug reactions. This kind of study could help The tools needed to exploit the genome in the treatment of identify the appropriate antidepressants for a patient, neurological and psychiatric diseases are available. When without having the current trial-and-error approach if properly applied, microarrays are powerful, and they have certain patterns of gene expression or polymorphisms were already provided new insights into disease. Study of diseaseassociated with clinical response. In cases where the therapy related abnormalities and changes in gene expression is the is toxic, such as immunotherapy or chemotherapy, specific first step in this process. Ultimately, we must independently polymorphisms that confer risk or classify treatment validate the changes observed in microarray studies and response could be used to identify patients who will respond understand how changes in gene expression translate into functional effects at the cellular level. Progress in these studies favourably. The assessment of gene expression within a particular will translate into array-based disease classification schemes patients brain-tumour specimen in combination with and help optimise therapy for individual patients on the basis knowledge of the relevant features of that patients genotype of gene expression or genetic background. The ability to study will eventually allow doctors to provide the ideal treatment the genetics of human variability on the scale afforded by this for each individual. The additional up-front cost of this technology carries with it enormous benefits for the diagnosis assessment will be more than offset by the avoidance of and treatment of neurological and psychiatric diseases. failed treatment trials, fewer adverse reactions, and the Linking of this data with changes at the protein level remains improved health of the patient. This example of customized largely uncharted territory. therapy is likely to be applicable to many clinical situations. Acknowledgments Thanks to Bonita Porch for her editorial contributions.
Conflict of interest

Search strategy and selection criteria

Articles were identified by searches of Medline with the words brain and microarray, of the authors files, and of references from these papers. Articles on the study of human neurological or psychiatric diseases and those with a proof of concept or principle in other areas were reviewed. Only papers published in English were included.

The are no conflicts of interest.

Role of funding source

Microarray research in the Geschwind Laboratory contributing to this review is supported by the National Institute of Mental Health, the National Institute of Neurological Disorders and Stroke, the National Institute on Aging, and the Waverly Smith Memorial fund. None of these funding sources had a role in the preparation of this review or in the decision to submit it for publication.

THE LANCET Neurology Vol 2 May 2003

http://neurology.thelancet.com

281

For personal use. Only reproduce with permission from The Lancet Publishing Group.

Review
References
1 2 3 4 5 6 7 8 Tanzi RE, Bertram L. New frontiers in Alzheimers disease genetics. Neuron 2001; 32: 18184. Hardy J. Genetic dissection of primary neurodegenerative diseases. Biochem Soc Symp 2001; 67: 5157. Rosenberg RN. DNA-triplet repeats and neurologic disease. N Engl J Med 1996; 335: 122224. Paulson HL, Fischbeck KH. Trinucleotide repeats in neurogenetic disorders. Annu Rev Neurosci 1996; 19: 79107. Venter JC, Adams MD, Myers EW, et al. The sequence of the human genome. Science 2001; 291: 130451. Nelson NJ. Draft human genome sequence yields several surprises. J Natl Cancer Inst 2001; 93: 493. Geschwind DH, Gregg J, eds. Microarrays for the Neurosciences. Cambridge: MIT Press, 2002. Sandberg R, Yasuda R, Pankratz DG, et al. Regional and strain-specific gene expression mapping in the adult mouse brain. Proc Natl Acad Sci USA 2000; 97: 1103843. Brown VM, Ossadtchi A, Khan AH, et al. Gene expression tomography. Physiol Genomics 2002; 8: 15967. Brown VM, Ossadtchi A, Khan AH, Cherry SR, Leahy RM, Smith DJ. High-throughput imaging of brain gene expression. Genome Res 2002; 12: 24454. Zirlinger M, Kreiman G, Anderson DJ. Amygdalaenriched genes identified by microarray technology are restricted to specific amygdaloid subnuclei. Proc Natl Acad Sci USA 2001; 98: 527075. Geschwind DH, Ou J, Easterday MC, et al. A genetic analysis of neural progenitor differentiation. Neuron 2001; 29: 32539. Terskikh AV, Easterday MC, Li L, et al. From hematopoiesis to neuropoiesis: evidence of overlapping genetic programs. Proc Natl Acad Sci USA 2001; 98: 793439. Phillips RL, Ernst RE, Brunk B, et al. The genetic program of hematopoietic stem cells. Science 2000; 288: 163540. DAgata V, Warren ST, Zhao W, Torre ER, Alkon DL, Cavallaro S. Gene expression profiles in a transgenic animal model of fragile X syndrome. Neurobiol Dis 2002; 10: 21118. Brown V, Jin P, Ceman S, et al. Microarray identification of FMRP-associated brain mRNAs and altered mRNA translational profiles in fragile X syndrome. Cell 2001; 107: 47787. Schena M, Shalon D, Davis RW, Brown PO. Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 1995; 270: 46770. Lipshutz RJ, Fodor SP, Gingeras TR, Lockhart DJ. High density synthetic oligonucleotide arrays. Nat Genet 1999; 21 (suppl 1): 2024. Geschwind DH. Sharing gene expression data: an array of options. Nat Rev Neurosci 2001; 2: 43538. Luo Z, Geschwind DH. Microarray applications in neuroscience. Neurobiol Dis 2001; 8: 18393. DeRisi J, Penland L, Brown PO, et al. Use of a cDNA microarray to analyse gene expression patterns in human cancer. Nat Genet 1996; 14: 45760. Shalon D, Smith SJ, Brown PO. A DNA microarray system for analyzing complex DNA samples using two- color fluorescent probe hybridization. Genome Res 1996; 6: 63945. Lockhart DJ, Winzeler EA. Genomics, gene expression and DNA arrays. Nature 2000; 405: 82736. Whitney LW, Becker KG, Tresser NJ, et al. Analysis of gene expression in multiple sclerosis lesions using cDNA microarrays. Ann Neurol 1999; 46: 42528. Chiang LW, Grenier JM, Ettwiller L, et al. An orchestrated gene expression component of neuronal programmed cell death revealed by cDNA array analysis. Proc Natl Acad Sci USA 2001; 98: 281419. Hughes TR, Mao M, Jones AR, et al. Expression profiling using microarrays fabricated by an ink-jet oligonucleotide synthesizer. Nat Biotechnol 2001; 19: 34247. Brenner S, Johnson M, Bridgham J, et al. Gene expression analysis by massively parallel signature sequencing (MPSS) on microbead arrays. Nat Biotechnol 2000; 18: 63034. Datson NA, van der Perk J, de Kloet ER, Vreugdenhil E. Expression profile of 30,000 genes in rat hippocampus using SAGE. Hippocampus 2001; 11: 43044. 29 Velculescu VE, Vogelstein B, Kinzler KW. Analysing uncharted transcriptomes with SAGE. Trends Genet 2000; 16: 42325. 30 Velculescu VE, Zhang L, Vogelstein B, Kinzler KW. Serial analysis of gene expression. Science 1995; 270: 48487. 31 Karsten SL, Geschwind DH. Gene expression analysis using cDNA microarrays. Curr Protoc Neurosci 2002: Unit 4.25. 32 VanDeerlin V, Ginsberg SD, Lee VM, Trojanowski JQ. The use of fixed human postmortem brain tissue to study mRNA expression in neurodegenerative diseases: applications of microdissection and amplification. In: Geschwind DH, Gregg J, eds. Microarrays for the neurosciences: an essential guide. Cambridge: MIT Press, 2002: 201236. 33 Karsten SL, Van Deerlin VM, Sabatti C, Gill LH, Geschwind DH. An evaluation of tyramide signal amplification and archived fixed and frozen tissue in microarray gene expression analysis. Nucleic Acids Res 2002; 30: E4. 34 Geschwind DH. Mice, microarrays, and the genetic diversity of the brain. Proc Natl Acad Sci USA 2000; 97: 1067678. 35 Nadon R, Shoemaker J. Statistical issues with microarrays: processing and analysis. Trends Genet 2002; 18: 26571. 36 Yuen T, Wurmbach E, Pfeffer RL, Ebersole BJ, Sealfon SC. Accuracy and calibration of commercial oligonucleotide and custom cDNA microarrays. Nucleic Acids Res 2002; 30: E48. 37 Yue H, Eastman PS, Wang BB, et al. An evaluation of the performance of cDNA microarrays for detecting changes in global mRNA expression. Nucleic Acids Res 2001; 29: E41-1. 38 Kohane IS, Kho A, Butte AJ. Microarrays for an integrative genomics. Cambridge: MIT Press, 2002. 39 Evans SJ, Datson NA, Kabbaj M, et al. Evaluation of Affymetrix Gene Chip sensitivity in rat hippocampal tissue using SAGE analysis: serial analysis of gene expression. Eur J Neurosci 2002; 16: 40913. 40 Crino PB, Trojanowski JQ, Dichter MA, Eberwine J. Embryonic neuronal markers in tuberous sclerosis: single-cell molecular pathology. Proc Natl Acad Sci USA 1996; 93: 1415257. 41 Eberwine J, Yeh H, Miyashiro K, et al. Analysis of gene expression in single live neurons. Proc Natl Acad Sci USA 1992; 89: 301014. 42 Ginsberg SD, Hemby SE, Lee VM, Eberwine JH, Trojanowski JQ. Expression profile of transcripts in Alzheimers disease tangle-bearing CA1 neurons. Ann Neurol 2000; 48: 7787. 43 Luo L, Salunga RC, Guo H, et al. Gene expression profiles of laser-captured adjacent neuronal subtypes. Nat Med 1999; 5: 11722. 44 Bonner RF, Emmert-Buck M, Cole K, et al. Laser capture microdissection: molecular analysis of tissue. Science 1997; 278: 148183. 45 Merikangas KR, Chakravarti A, Moldin SO, et al. Future of genetics of mood disorders research. Biol Psychiatry 2002; 52: 45777. 46 Risch N, Merikangas K. The future of genetic studies of complex human diseases. Science 1996; 273: 151617. 47 Nadeau JH, Frankel WN. The roads from phenotypic variation to gene discovery: mutagenesis versus QTLs. Nat Genet 2000; 25: 38184. 48 Frankel WN. Taking stock of complex trait genetics in mice. Trends Genet 1995; 11: 47177. 49 Lock C, Hermans G, Pedotti R, et al. Genemicroarray analysis of multiple sclerosis lesions yields new targets validated in autoimmune encephalomyelitis. Nat Med 2002; 8: 50008. 50 Hanzel DK, Trojanowski JQ, Johnston RF, Loring JF. High-throughput quantitative histological analysis of Alzheimers disease pathology using a confocal digital microscanner. Nat Biotechnol 1999; 17: 5357. 51 Sallinen SL, Sallinen PK, Haapasalo HK, et al. Identification of differentially expressed genes in human gliomas by DNA microarray and tissue chip techniques. Cancer Res 2000; 60: 661722. 52 Whitney LW, Ludwin SK, McFarland HF, Biddison WE. Microarray analysis of gene expression in multiple sclerosis and EAE identifies 5-lipoxygenase as a component of inflammatory lesions. J Neuroimmunol 2001; 121: 4048. 53 Mirnics K, Middleton FA, Marquez A, Lewis DA, Levitt P. Molecular characterization of schizophrenia viewed by microarray analysis of gene expression in prefrontal cortex. Neuron 2000; 28: 5367. 54 Mirnics K, Middleton FA, Stanwood GD, Lewis DA, Levitt P. Disease-specific changes in regulator of

DNA microarrays

55

56

57

58 59

60

9 10 11

61

62

12 13

63

64

14 15

65

66

16

67

17

68 69

18 19 20 21 22

70

71

23 24 25

72 73

74

26

75 76

27

28

77

G-protein signaling 4 (RGS4) expression in schizophrenia. Mol Psychiatry 2001; 6: 293301. Middleton FA, Mirnics K, Pierri JN, Lewis DA, Levitt P. Gene expression profiling reveals alterations of specific metabolic pathways in schizophrenia. J Neurosci 2002; 22: 271829. Hakak Y, Walker JR, Li C, et al. Genome-wide expression analysis reveals dysregulation of myelination-related genes in chronic schizophrenia. Proc Natl Acad Sci USA 2001; 98: 474651. Lord C, Volkmar F. Genetics of childhood disorders: XLII, autism, part 1: diagnosis and assessment in autistic spectrum disorders. J Am Acad Child Adolesc Psychiatry 2002; 41: 113436. Lord C, Cook EH, Leventhal BL, Amaral DG. Autism spectrum disorders. Neuron 2000; 28: 35563. Purcell AE, Jeon OH, Zimmerman AW, Blue ME, Pevsner J. Postmortem brain abnormalities of the glutamate neurotransmitter system in autism. Neurology 2001; 57: 161828. Colantuoni C, Jeon OH, Hyder K, et al. Gene expression profiling in postmortem Rett Syndrome brain: differential gene expression and patient classification. Neurobiol Dis 2001; 8: 84765. Lewohl JM, Wang L, Miles MF, Zhang L, Dodd PR, Harris RA. Gene expression in human alcoholism: microarray analysis of frontal cortex. Alcohol Clin Exp Res 2000; 24: 187382. Mayfield RD, Lewohl JM, Dodd PR, Herlihy A, Liu J, Harris RA. Patterns of gene expression are altered in the frontal and motor cortices of human alcoholics. J Neurochem 2002; 81: 80213. Tang Y, Lu A, Aronow BJ, Sharp FR. Blood genomic responses differ after stroke, seizures, hypoglycemia, and hypoxia: blood genomic fingerprints of disease. Ann Neurol 2001; 50: 699707. Rickman DS, Bobek MP, Misek DE, et al. Distinctive molecular profiles of high-grade and low-grade gliomas based on oligonucleotide microarray analysis. Cancer Res 2001; 61: 688591. Pomeroy SL, Tamayo P, Gaasenbeek M, et al. Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 2002; 415: 43642. Tanwar MK, Gilbert MR, Holland EC. Gene expression microarray analysis reveals YKL-40 to be a potential serum marker for malignant character in human glioma. Cancer Res 2002; 62: 436468. Geschwind DH, Gregg J, Boone K, et al. Klinefelters syndrome as a model of anomalous cerebral laterality: testing gene dosage in the X chromosome pseudoautosomal region using a DNA microarray. Dev Genet 1998; 23: 21529. Varis A, Wolf M, Monni O, et al. Targets of gene amplification and overexpression at 17q in gastric cancer. Cancer Res 2002; 62: 262529. Cai WW, Mao JH, Chow CW, Damani S, Balmain A, Bradley A. Genome-wide detection of chromosomal imbalances in tumors using BAC microarrays. Nat Biotechnol 2002; 20: 39396. OHagan RC, Chang S, Maser RS, et al. Telomere dysfunction provokes regional amplification and deletion in cancer genomes. Cancer Cell 2002; 2: 14955. Pollack JR, Sorlie T, Perou CM, et al. Microarray analysis reveals a major direct role of DNA copy number alteration in the transcriptional program of human breast tumors. Proc Natl Acad Sci USA 2002; 99: 1296368. Cheung VG, Nowak N, Jang W, et al. Integration of cytogenetic landmarks into the draft sequence of the human genome. Nature 2001; 409: 95358. Hacia JG, Fan JB, Ryder O, et al. Determination of ancestral alleles for human single-nucleotide polymorphisms using high-density oligonucleotide arrays. Nat Genet 1999; 22: 16467. Dong S, Wang E, Hsie L, Cao Y, Chen X, Gingeras TR. Flexible use of high-density oligonucleotide arrays for single-nucleotide polymorphism discovery and validation. Genome Res 2001; 11: 141824. Garcia CK, Mues G, Liao Y, et al. Sequence diversity in genes of lipid metabolism. Genome Res 2001; 11: 104352. Gilles PN, Wu DJ, Foster CB, Dillon PJ, Chanock SJ. Single nucleotide polymorphic discrimination by an electronic dot blot assay on semiconductor microchips. Nat Biotechnol 1999; 17: 36570. Santacroce R, Ratti A, Caroli F, et al. Analysis of clinically relevant single-nucleotide polymorphisms by use of microelectronic array technology. Clin Chem 2002; 48: 212430

282

THE LANCET Neurology Vol 2 May 2003

http://neurology.thelancet.com

For personal use. Only reproduce with permission from The Lancet Publishing Group.