Skip Navigation

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Smith, L.
Right arrow Articles by Greenfield, A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Smith, L.
Right arrow Articles by Greenfield, A.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

Human Molecular Genetics, 2003, Vol. 12, Review Issue 1 R1-R8
DOI: 10.1093/hmg/ddg053
© 2003 Oxford University Press

DNA microarrays and development

Lee Smith and Andy Greenfield*

MRC Mammalian Genetics Unit, Harwell, Didcot OX11 0RD, UK

Received November 28, 2002; Accepted December 23, 2002


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION: FUNCTIONAL...
 INVERTEBRATES SHOW THE WAY
 MICROARRAYS AND MAMMALIAN...
 THE FUTURE
 REFERENCES
 
Gene expression is a central concept in molecular biology: its control, frequently exquisite in terms of cell specificity and timing, forms part of our explanation of most biological processes. The importance of the control of gene expression for developmental biologists is made obvious by just considering the nature of their discipline. Development is the term we use to describe the coordination in time and space of numerous cellular activities such as mitosis, migration, differentiation and apoptosis. Understanding the role of genes in these processes thus necessitates the use of methods to determine patterns of transcription during development with a high degree of sensitivity and specificity, conventionally by in situ hybridization. However, there is a widespread conviction amongst biologists that the description of gene expression patterns is of no immediate functional relevance: definitive functional data are the exclusive prerogative of biochemistry and genetics. In this review of recent applications of DNA microarray technology by developmental biologists, we suggest that genome-wide expression profiling has met with some resistance owing to such preconceived ideas about the status of gene expression pattern descriptions and, in particular, the format in which these are delivered by microarrays.


   INTRODUCTION: FUNCTIONAL GENOMICS?
TOP
 ABSTRACT
 INTRODUCTION: FUNCTIONAL...
INVERTEBRATES SHOW THE WAY
 MICROARRAYS AND MAMMALIAN...
 THE FUTURE
 REFERENCES
 
The plight of contemporary genetics might be succinctly summarized as: many genes, few functions. The task of evening out this numerical discrepancy in some systematic way falls to functional genomics. Microarrays are a quintessential tool in functional genomics, but what is meant to be functional about the data they help to generate? Microarray data offer an insight into the transcriptional responses of a genome to a particular mutational event or environmental insult. How can such responses inform our understanding of causality in biology? In this review we address these questions in the context of developmental biology. We place reasonable emphasis on mammalian development, partly with a view to commenting on why this is such a (pardon the pun) embryonic discipline, as evidenced by the paucity of published microarray papers in the established development journals. We will assume familiarity with the mechanics of microarray technology, in both spotted and GeneChip platforms, as well as the use of relative gene expression measurements as a quantitative indicator of transcript levels in a given tissue. Several excellent reviews cover these topics, as well as the associated informatics required for productive analyses of microarray data (1–3). We will not discriminate between research based on the microarray platform employed, but rather focus the discussion on the biological application.

Before we turn to the literature, a few brief remarks about the term ‘function’ by way of scene-setting. Geneticists are characteristically concerned with one notion of function: roughly, the function of a gene is described by examining the abnormal phenotype(s) caused by its mutation(s). However, a good deal of functionally relevant data reside in the explanatory distance between the mutation and the basic phenotypic description. (To put it another way: why certain mutations give particular phenotypes isn't always clear.) Such data include the knock-on consequences of mutation on the expression of other genes. However, the use of ‘markers’ to examine the phenotype of a mutant is not usually seen as a way of examining the function of those marker genes. Rather, those markers are used to reveal some aspect of the function of the gene mutated. Traditionally, descriptive data, such as gene expression patterns, are so called because they do not impinge directly on the function of those genes whose expression is being studied. (One developmental biology journal even precludes ‘functional speculation’ in papers that contain no more than the description of gene expression patterns.) Such data are circumstantial, supporting or consistent with function, but not causative or definitive. One of the messages of this review is that a qualitative distinction between functional and descriptive data is becoming blurred by whole genome expression studies. Whilst distinctions still exist, they are perhaps best seen as residing within a continuum of functionally relevant data.


   INVERTEBRATES SHOW THE WAY
TOP
 ABSTRACT
 INTRODUCTION: FUNCTIONAL...
 INVERTEBRATES SHOW THE WAY
MICROARRAYS AND MAMMALIAN...
 THE FUTURE
 REFERENCES
 
The mechanistic basis of metazoan development represents one of the unsolved mysteries of biology: how does a single fertilized egg, through successive cell divisions and differentiation events, mature into an adult organism? The fruitfly Drosophila melanogaster has been a pioneering model organism for geneticists and developmental biologists for many decades. Drosophilogists have been quick to exploit the power of genome-wide expression profiling using DNA microarrays. One notable example is that of Arbeitman et al. (4). In this study the expression of 4028 genes was analysed in wild-type flies throughout Drosophila development during 66 sequential time periods. These included sampling RNA at fertilization, embryonic, larval and pupal periods as well as the first 30 days of adulthood. Each experimental sample was compared with a common reference sample, allowing the relative abundance of any transcript to be determined at every developmental stage. The analysis of such a huge amount of data conventionally proceeds by the use of algorithms that group or cluster genes according to similarity in their expression profiles (5). (The nature of such clusters is clearly determined by the experimental design and range of samples included for analysis, see below.) The analysis of the Drosophila dataset revealed that, despite the use of whole animals, it was possible to discern expression profiles in specific organs, as well as those associated with particular biological processes. For example, one cluster of 23 genes included eight known to be expressed in terminally differentiated muscle. The profile of this cluster has two peaks of expression, one coinciding with the larval stage and a second with adult muscle development. Initiation of larval muscle development is regulated by the basic helix–loop–helix (bHLH) transcription factor Twist, which induces expression of dMef 2, which itself encodes a MADS box transcription factor regulating the transcription of muscle differentiation genes. Crucially, this muscle-specific regulatory hierarchy was recapitulated in the microarray data: the peak of twist expression preceded that of dMef 2, which preceded transcription of genes in a muscle differentiation cluster. Moreover, 15 of the 23 genes in this latter cluster contained pairs of predicted dMEF2-binding sites. Similar clusters were identified revealing coordinate expression profiles associated with particular biochemical and cellular functions, including mitochondrial proteins, components of the 26S proteasome complex and cytoskeletal/neuronal factors.

Global transcriptional information during morphogenesis was also readily available: the vast majority of genes (>88%) that exhibit transcriptional modulation during the stages analysed are expressed during the first 20 h of development, before the end of embryogenesis. A total of 2103 changed during embryogenesis, 445 changed during larval life, 646 during the pupal stage and 118 during adult life. The transcript levels of only 16 genes changed significantly during the adult time period sampled. These data suggest a strong association between modulation of transcriptional activity and morphogenesis.

These results demonstrate how microarray data from appropriately designed experiments can shed light on tissue-specific gene regulatory hierarchies. A key question is: to what extent does membership of a given cluster reveal the function of previously uncharacterized or novel gene? Is it possible to undermine the ‘gene expression pattern alone does not determine function’ paradigm? Kim et al. (6) assembled data from 553 microarray experiments utilizing a variety of growth conditions, developmental stages and mutants of the nematode worm C. elegans. These experiments surveyed almost the complete transcriptome of the worm (between 12 000 and 18 000 genes). No single reference RNA target was common to all these experiments, so the authors employed an informatics approach aimed at displaying the similarity between the expression profile of each gene across all experiments in a manner that maximises their informativeness. Firstly, conventional two-dimensional clustering was performed in order to generate a matrix of similarity between genes and tissue samples (including organ type, developmental stage, genotype etc.). Secondly, for each gene, the similarity between it and its 20 nearest neighbours in the matrix was used to assign that gene a position on a two-dimensional scatter plot. The scatter plot was constructed in such a manner that genes with a high degree of correlation of profiles are placed near to each other. Finally, the two-dimensional display was transformed into a three-dimensional terrain map by including an indicator of gene density on the z-axis. The resulting ‘terrain’ is reminiscent of a model of a complex mountain range, with clusters appearing as individual mountains (totalling 44) of specific size and shape depending on the number of, and correlation between, genes in that cluster. The biological significance of the terrain was immediately apparent from the specific content of the expression mountains: some grouped together genes transcribed in the same tissue whilst others grouped genes with similar cellular functions. For example, genes highly enriched in sperm constituted ~69% of those found in mount 4. Genes required for sperm motility cluster together at one of its ends. This mount also contained a much higher than average number of protein kinases and phosphatases, suggesting that sperm may utilize protein phosphorylation to regulate protein activity in the absence of transcription and translation. Novel genes in mount 4 can be assigned a role in regulating sperm activity with a high degree of confidence. Overall, the authors were able to infer a potential function for 30 of the 44 mountains displayed, based on the presence in those mountains of sets of genes of known function. The presence of novel genes in these expression mountains prompted the authors to assert: ‘in addition to biochemistry and genetics, one could now infer gene functions with the use of gene expression data’.

It is worth considering again the significance of these data. Owing to the very large number of genes and conditions under which their expression was assayed in these experiments, a high level of discriminatory power is conferred upon the algorithms that group genes by similarity of profiles. This resolving power allows genome-wide functional relations to be discerned by similarity in a manner unthinkable before the advent of microarray technology. The subtlety and specificity of this functional resolution is a product of the sheer scale of the experiment. It would appear that passing a quantitative threshold in terms of the numbers of genes and conditions analysed results in a qualitative shift in the significance of such expression data, from descriptive to functional. This observation has obvious consequences for the design of future microarray experiments: a similar scale of analysis in a mammalian system will require incorporation of relevant controls and standards into each laboratory's protocols to allow appropriate inter-experiment comparisons to be made. With this in mind, the centralized provision of microarrays, such as that offered by the UK HGMP Resource Centre (http://www.hgmp.mrc.ac.uk/Research/Microarray/index.jsp), is to be welcomed, as is the standardization of annotation required in the description of an array-based experiment (7) and the development of databases for storing the output (8).

The pioneering experiments in invertebrates just described appear to suggest that the notion that gene expression profiles alone do not reveal biological function needs to be re-examined. Surveying gene expression under a wide range of conditions and tissues, in wild-type and mutant animals, seems to transform the significance of data that on a smaller scale would be considered descriptive. Of course, for any individual gene residing within a ‘functionally loaded’ cluster the task remains to determine the phenotypic consequences of its mutation. Yet perhaps such experiments should be seen as complementing our understanding of that gene's function developed by other means, rather than being exclusively definitive thereof: particularly given the high frequency with which no clear phenotype is observed after mutagenesis. To infer function from gene expression profiles is not mere speculation if the design of the experiment and the complexity of the dataset permit otherwise.


   MICROARRAYS AND MAMMALIAN DEVELOPMENT
TOP
 ABSTRACT
 INTRODUCTION: FUNCTIONAL...
 INVERTEBRATES SHOW THE WAY
 MICROARRAYS AND MAMMALIAN...
THE FUTURE
 REFERENCES
 
Why are there no studies attempting to profile gene expression in the whole mammalian embryo at a genome-wide level throughout its development, in a manner reminiscent of those just described for Drosophila and C. elegans? Given the widespread accessibility of microarray technology today, the answer is probably very complex, involving references to both technical and ‘cultural’ issues. The most common technical problem concerns the small amounts of RNA available from standard dissections of mammalian embryos. By ‘cultural’, we mean the familiarity that developmental biologists have with in situ hybridization (ISH), their relative lack of familiarity with microarrays and the common attitude that descriptions of gene expression patterns support only speculation about function. However, these remarks are equally applicable to developmental biologists using flies and worms as a model. The use of DNA microarrays to examine mammalian development is a small but rapidly growing field of study. It is currently dominated by the exploitation of arrays to perform screens for molecules involved in particular developmental processes. In the following sections we will review some of these applications and address in more detail some of the issues surrounding the use of microarrays in mammalian developmental biology.

Expression screens
Systematic genome-wide studies of mammalian development using microarrays stand out due to their rarity. Miki et al. (9) analysed the expression of 18 816 mouse genes in 49 different embryonic and adult tissues, permitting some clustering of genes pertinent to the development of specific tissues, such as the central nervous system. However, the limited number of embryonic samples, totalling 11, means that this study falls short of providing a transcriptional profile of mouse development. Perhaps due to the relative complexity of the mammalian embryo, more familiar are studies aimed at profiling expression at specific embryonic stages or in specific embryonic tissues, including (without attempting to be comprehensive): 12.5 days post coitum (dpc) mouse placenta (10), mouse retina (11), mouse lung (12), mouse mammary gland (13), preimplantation mouse embryos (14), mouse hippocampus (15), and mouse B cells (16). Developmental biologists have also been quick to adapt familiar techniques for the purposes of exploiting microarray technology, including the use of cell line models (17–24) and organ culture (25,26).

These experiments are frequently performed in order to screen for molecules regulating a specific biological process, such as the authors' use of microarrays to identify genes regulating the sexually dimorphic development of the mouse gonad and mesonephros (27) (Fig. 1). Most common is the use of microarrays to identify transcriptional targets, frequently employing the use of transfected or treated cell lines. For example, Connaci-Sorell et al. (28) used microarrays in an attempt to identify target genes induced by ß-catenin, a key component of the WNT signalling pathway. A human renal carcinoma cell line, which did not express catenins, was stably transfected with a ß-catenin construct. The most extensively elevated transcript in these stable transfectants, when compared to untransfected, was that of the cell adhesion molecule Nr-CAM. The authors then demonstrated the presence of LEF/TCF binding sites in the Nr-CAM promoter that mediate the induction by ß-catenin. Similar screens have been performed in order to identify the targets of WNT signalling (24), GDNF signalling (25), Mitf-dependent SCF/Kit signalling (29) and the targets of the transcription factors TBX2 (20), SIX5 (21), PAX3 (18), WHN (30), HOXA11 (17) and HOXA13 (26). These latter two studies are notable for their exploitation of mutants, an approach that has been employed in a most sophisticated way in Drosophila (31). Utilizing a familiar inducible expression system Valerius et al. (17) identified potential targets of the HOXA11 transcription factor. One gene, Integrin {alpha}8, was up-regulated 20-fold after induction of Hoxa11 expression. The authors then demonstrated that, as predicted, expression of Integrin {alpha}8 was dramatically reduced in embryos homozygous for null mutations in both Hoxa11 and Hoxd11. Zhao et al. (26) generated mice homozygous for a dominant mutant allele of Hoxa11, created by swapping the HOXA11 homeobox for that encoded by the contiguous gene Hoxa13. This swap resulted in a homeotic transformation in the reproductive tracts of female animals. When expression in the uterus of mutant females was compared with wild-type controls using DNA microarrays, over 30 genes were identified whose expression was normally associated with the cervix, consistent with the posteriorization observed at the tissue level. Such genes altered in expression in the mutant uterus are candidate downstream genes of HOXA13. However, the correlation between the genome-wide transcriptional response to the mutation and the observable phenotype hints at a significance for expression profiling of mutants which goes far beyond mere target gene identification. The gene expression signature associated with the transformed reproductive tract is reminiscent of those identified, in particular in tumours that have revolutionized tumour classification in oncology (32–34).



View larger version (62K):
[in this window]
[in a new window]
  		Figure 1. A screen for genes expressed in a sexually dimorphic fashion during mouse gonad and mesonephros development using DNA microarrays. The sexually dimorphic development of the mammalian gonad into either a testis or an ovary is controlled by the sex-determining gene SRY. We have used DNA microarrays to screen for molecules in the genetic pathways of ovarian and testiscular development by identifying genes expressed in a female- or male-specific fashion. Microarrays of cDNA (27) or oligonucleotide probes (spots) were hybridized with labelled complex CDNA target generated from gonadal (or mesonephric) total RNA at three stages of development (11.5–13.5 days post coitum). PCR-based amplification procedures, aimed at increasing the amount of available cDNA target, are employed prior to labelling (51). The rapid morphological divergence of the male and female gonad during this 48 h period is reflected in the growing number of differentially expressed (outlier) genes observed in the gonadal microarray data. These data are presented as false colour overlay images of the scanned microarrays and scatterplots of relative signal intensities. In these scatterplots, points removed from the diagonal represent spots deviating from the 1:1 ratio of equivalent expression levels. Microarray data from genes exhibiting the greatest deviation from a 1:1 ratio were validated independently by wholemount in situ hybridization (male gonads and mesonephroi to the right in each example). The cell-type specificity of transcription is determined by sectioning and functional studies are performed on specific genes. The circle is closed using microarrays to examine the effect of mutation of one of these genes on gonadal or mesonephric gene expression (molecular phenotyping).

 
Clustering of gene expression profiles is a two-dimensional exercise, allowing functional annotation of both genes and biological samples. As we have seen, cataloguing the genome-wide transcriptional changes that correlate with the overt morphological anomalies of a developmental mouse mutant is a potentially profitable exercise in shedding further light on the function of the mutated gene, but examples are rare. This is despite the fact that the phenotypes of numerous developmental mouse mutants have been characterized with the assistance of individual marker genes using wholemount in situ hybridization (WMISH), as any cursory glance at a development journal will testify. One other notable exception to the paucity of genome-wide ‘molecular phenotyping’ (or expression phenotyping) of mutants is provided by the work of Freiman et al. (35) with mice exhibiting defects in folliculogenesis after inactivation of the gene encoding the tissue-selective component of TFIID, TAFII105 (35). Despite detailed physiological, histological and immunohistochemical data the molecular mechanism underlying the ovarian phenotype was unclear. To address this the authors compared mutant and wild-type ovaries from age-matched and hormonally synchronized females using gene expression profiling of 11 000 genes. Only 1% of these were down-regulated by greater than 2-fold in the mutant ovary, including a number known to be important for female fertility. Notably, the expression of multiple components of the inhibin-activin-follistatin pathway was severely reduced. The role of inhibins and activins in modulating the release of pituitary follicle-stimulating hormone and regulating oestrogen synthesis within the ovary suggests a clear mechanism for the development of improper regulation of folliculogenesis. Data such as these suggest an important role for gene expression profiling in providing a mechanistic bridge between gene mutation and physiologically defined phenotypes.

Technical issues
Microarrays are now widely available, so what technical issues might account for the relatively rare use of microarrays in developmental biology when compared to the field of, for example, oncology? We can divide these issues into those associated with microarray availability, tissue availability and biological complexity.

In order to perform genome-wide expression profiling it is necessary to have genome-wide microarrays. Lack of array data from developmental models such as frog and chicken is no doubt best explained by the restricted availability of the relevant microarrays. In the case of the mouse, such whole genome arrays are on the horizon, although not yet available. This situation is likely to be resolved in the next 2–3 years, given the maturity of the genome and EST sequencing projects in the mouse and humans. Certainly the number of genes that can be investigated using microarrays is increasing all the time. It is likely that improvements in deposition or synthesis of DNA probes will allow a suitable density to be achieved, avoiding the requirement of hybridizing multiple arrays in order to achieve full genome coverage. The advent of spotted oligonucleotide arrays and related technologies also opens up the possibility of exon-specific probes for the detection of tissue-specifc (cell-specific?) splice variants.

Perhaps the most obvious impediment to the use of arrays by developmental biologists is the limiting amount of RNA available from standard embryonic dissections. For example, a whole 7.5 dpc mouse embryo yields ~0.5 µg of total RNA. Conventional labelling of RNA for microarray hybridization requires 5–100 µg of total RNA. It is clear that specific embryonic tissues or organs, especially from early stages, will yield only a fraction of the required target RNA. Embryo sorting techniques have been employed by Drosophila geneticists (31), but target amplification appears to be the most widely applicable solution. A number of amplification protocols have been described that allow the generation of sufficient target from small amounts of starting RNA, with varying degrees of success claimed. Included amongst these are RNA polymerase-based approaches (36–38) and PCR-based approaches (11,39). Recently it has been claimed that exponential PCR amplification preserves gene expression ratios through as high as 3x1011-fold amplification (40). It is not clear at the time of writing which method offers the ‘best’ solution to this problem of limiting RNA, partly because different scientists will have different criteria for the assessment of which is best, including simplicity, reliability and cost. However, perhaps the main reason is the lack of thorough testing of such methodologies. Many are assessed by their ability to identify the same ‘outliers’ or ‘differentials’ between two samples both with and without the use of the amplification method. However, it is unlikely that passing this test is sufficient to support the conclusion that genuine quantitative comparisons between independent amplifications will allow clustering algorithms to be applied with a degree of confidence comparable to conventional methods. This degree of confidence is required if developmental biologists are to establish the many subtle transcriptional differences that exist between different embryonic cells and tissues. The literature contains very few examples of quantitative expression profiling after the amplification of minute quantities (<25 ng) of total RNA. It may be that appropriate statistical methods will have to be devised to account for the inevitable bias and distortion introduced by amplification methodologies. However, RNA source limitation in itself is unlikely to be an insurmountable difficulty once amplification protocols become reliable and routine.

The last category of potential technical impediments is closely related to the last: does the innate complexity of the mammalian embryo, both transcriptionally and at the cellular level, necessitate a degree of sub-dissection and cell sorting that makes labelling and hybridization a practical impossibility? We are all familiar with the predictive lament: ‘using microarrays you probably will not find those incredibly rare transcripts that encode those critical regulators of development: not without resorting to single-cell approaches!’ It may well be true that a combination of variation in transcript abundance and cellular heterogeneity will make it difficult to perform expression profiling in standard tissue explants in a useful fashion. However, it is also certain that enabling technologies will be developed and improved. Single-cell approaches to studying developmental neurobiology have already been described (36) and will no doubt become easier to employ with time. Perhaps more daunting is the task of tracing cellullar heterogeneity in individual lineages embryo-wide. Promoter-driven reporter molecules such as GFP are of use in identifying and isolating certain cell-types, but we are then restricted to examining those lineages about which we already know. However, the point is to use microarrays to identify novel lineages and novel markers as a discovery tool in developmental biology. It is difficult to see how a technique such as laser capture microdissection (37) could be used to systematically isolate and characterize the expression profiles of individual cells in a way that was amenable to comparison from embryo to embryo. We need to look forward to future technological developments before we can countenance the genome-wide expression profiling of mammalian development on a cell-by-cell basis: but that can hardly be interpreted as a deterrent to those considering exploiting microarrays to perform less sophisticated experiments!

The availability of mutants
The importance of including tissue samples from mutants in expression profiling experiments cannot be overestimated. A wide range of mutant phenotypes is available to the Drosophila and Zebrafish developmental biologist, but similar resources are yet to be generated in mammalian models, primarily the mouse, despite the large number of gene knockouts performed. This situation is probably not helped by the financial and infrastructural resources required to generate and maintain large numbers of mouse mutants in any systematic way. However, the requirement for additional developmental phenotypes is now being addressed by genome-wide chemical mutagenesis efforts in the mouse aimed at supporting phenotype-driven screens (41–44). The phenotype-driven approach has the advantage of allowing the identification of additional allelic classes of mutation in addition to the null. Ultimately, the generation of distinct mutations that cause a common phenotype will offer the resources for identifying the molecular nature of developmental pathways.

Imaging gene expression in four dimensions
Development involves the fantastically complicated orchestration of cell differentiation, communication, death, migration and division in time and three-dimensional space. It is likely that the variety of gene expression patterns will reflect this dynamic four-dimensional complexity. (This is one reason why cell line models will never replace studies of whole embryos or organs.) Morphogen gradients, formed by secreted signalling molecules (such as members of the Hedgehog and WNT protein families) are thought to play an important role in creating the complex patterns of gene expression required for anatomical pattern to emerge (45). The transcriptional readout of such a gradient is known as interpretation, and this interpretation may vary from cell to cell due to differences in position within that gradient. Moreover, the transcriptional response may also be subtler than a simple binary ‘on–off’. Given the combined effect of multiple morphogen gradients and this degree of subtlety in the variation in transcriptional response across a group of embryonic cells, it is not surprising that developmental biologists are looking to more sophisticated methods for visualizing gene expression in time and space in a quantitative fashion (46). In the future such imaging techniques will allow the monitoring of dynamic patterns of gene expression in individual embryonic cells in a living embryo. From the perspective of this conception of a gene expression profile, the one offered by microarray data may seem rather static and coarse-grained in comparison. Perhaps this comparison is the cause of some reluctance on the part of mammalian developmental biologists to countenance gene expression profiling. However, techniques embedded in these differing conceptions of a gene expression profile should be viewed as complementary, not contradictory. On the one hand, microarrays excel in terms of throughput, quantitation and the ability to detect complex patterns of similarity and difference between expression profiles, from which function can be inferred. On the other, imaging techniques, including ISH, offer unique insights into the spatiotemporal dimension of gene expression and its cell specificity. Developing databases of embryonic gene expression in mammals, such as EMAGE (47), aim to develop close links with microarray expression databases precisely because of this complementarity. Such links make the development of standardized ontologies (anatomical, phenotypic etc.) a priority.


   THE FUTURE
TOP
 ABSTRACT
 INTRODUCTION: FUNCTIONAL...
 INVERTEBRATES SHOW THE WAY
 MICROARRAYS AND MAMMALIAN...
 THE FUTURE
REFERENCES
 
This review is not the first to draw attention to, and attempt to explain, the perceived lack of impact of DNA microarray technology on the field of mammalian developmental biology (48,49). We have chosen to highlight those reasons that may be persistent, independent of technological advances in the microarray field. One such example is a perceived deficiency in the nature of the expression data generated by use of arrays: for example, their lacking cellular or spatial specificity. We have suggested that this is the result of unwarranted direct comparison of complementary approaches to profiling gene expression: like comparing chalk with cheese. The other is a reluctance to accept the functional significance of microarray data, both for the genes and biological samples analysed. However, we look forward to genome-wide expression profiling being used as a routine tool in phenotyping of developmental mutants and to tackle some of the tougher problems in developmental genetics: what is the transcriptional basis of variable penetrance? Can transcriptomic responses to mutation explain the frequent absence of an expected phenotype? How homogeneous are the expression profiles of morphologically identical cells in a tissue? What is the functional relevance of the distribution of splice variants throughout the organism? One such problem, whether gene expression anomalies occur in cloned mice, has already been addressed by Humpherys et al. (50).

Biologists frequently amass reasons for not doing an experiment. Some of the reasons that developmental biologists think support their suspicion of genome-wide expression profiling using arrays have been reviewed here: limiting amounts of RNA, tissue or cellular complexity, rarity of key regulatory genes, or the subtlety of transcriptional changes of functional significance. The philosopher Ludwig Wittgenstein said famously: don't think, just look! Hopefully, this review has demonstrated that more and more developmental biologists are looking to see what gene expression profiling using microarrays can tell them about their favourite biological processes.


   ACKNOWLEDGEMENTS
 
The authors wish to thank Pam Siggers, Sam Cox, John Willan, Nick Van Hateren and Peter Underhill for unpublished images of ISH and microarray data and Ruth Arkell for useful comments on an earlier version of the manuscript.


   FOOTNOTES
 
* To whom correspondence should be addressed. Email: a.greenfield{at}har.mrc.ac.uk Back


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION: FUNCTIONAL...
 INVERTEBRATES SHOW THE WAY
 MICROARRAYS AND MAMMALIAN...
 THE FUTURE
 REFERENCES
 

  1. Eisen, M.B. and Brown, P.O. (1999) DNA arrays for analysis of gene expression. Meth. Enzymol., 303, 179–205.[ISI][Medline]

  2. Hegde, P., Qi, R., Abernathy, K., Gay, C., Dharap, S., Gaspard, R., Hughes, J.E., Snesrud, E., Lee, N. and Quackenbush, J. (2000) A concise guide to cDNA microarray analysis. Biotechniques, 29, 548–550, 552–554, 556.[ISI][Medline]

  3. Quackenbush, J. (2001) Computational analysis of microarray data. Nat. Rev. Genet., 2, 418–427.[CrossRef][ISI][Medline]

  4. Arbeitman, M.N., Furlong, E.E., Imam, F., Johnson, E., Null, B.H., Baker, B.S., Krasnow, M.A., Scott, M.P., Davis, R.W. and White, K.P. (2002) Gene expression during the life cycle of Drosophila melanogaster. Science, 297, 2270–2275.[Abstract/Free Full Text]

  5. Eisen, M.B., Spellman, P.T., Brown, P.O. and Botstein, D. (1998) Cluster analysis and display of genome-wide expression patterns. Proc. Natl Acad. Sci. USA, 95, 14863–14868.[Abstract/Free Full Text]

  6. Kim, S.K., Lund, J., Kiraly, M., Duke, K., Jiang, M., Stuart, J.M., Eizinger, A., Wylie, B.N. and Davidson, G.S. (2001) A gene expression map for Caenorhabditis elegans. Science, 293, 2087–2092.[Abstract/Free Full Text]

  7. Brazma, A., Hingamp, P., Quackenbush, J., Sherlock, G., Spellman, P., Stoeckert, C., Aach, J., Ansorge, W., Ball, C.A., Causton, H.C. et al. (2001) Minimum information about a microarray experiment (MIAME)—toward standards for microarray data. Nat. Genet., 29, 365–371.[CrossRef][ISI][Medline]

  8. Sherlock, G., Hernandez-Boussard, T., Kasarskis, A., Binkley, G., Matese, J.C., Dwight, S.S., Kaloper, M., Weng, S., Jin, H., Ball, C.A. et al. (2001) The Stanford Microarray Database. Nucl. Acids Res., 29, 152–155.[Abstract/Free Full Text]

  9. Miki, R., Kadota, K., Bono, H., Mizuno, Y., Tomaru, Y., Carninci, P., Itoh, M., Shibata, K., Kawai, J., Konno, H. et al. (2001) Delineating developmental and metabolic pathways in vivo by expression profiling using the RIKEN set of 18 816 full-length enriched mouse cDNA arrays. Proc. Natl Acad. Sci. USA, 98, 2199–2204.[Abstract/Free Full Text]

  10. Tanaka, T.S., Jaradat, S.A., Lim, M.K., Kargul, G.J., Wang, X., Grahovac, M.J., Pantano, S., Sano, Y., Piao, Y., Nagaraja, R. et al. (2000) Genome-wide expression profiling of mid-gestation placenta and embryo using a 15 000 mouse developmental cDNA microarray. Proc. Natl Acad. Sci. USA, 97, 9127–9132.[Abstract/Free Full Text]

  11. Livesey, F.J., Furukawa, T., Steffen, M.A., Church, G.M. and Cepko, C.L. (2000) Microarray analysis of the transcriptional network controlled by the photoreceptor homeobox gene Crx. Curr. Biol., 10, 301–310.[CrossRef][ISI][Medline]

  12. Mariani, T.J., Reed, J.J. and Shapiro, S.D. (2002) Expression profiling of the developing mouse lung: insights into the establishment of the extracellular matrix. Am. J. Respir. Cell Mol. Biol., 26, 541–548.[Abstract/Free Full Text]

  13. Master, S.R., Hartman, J.L., D'Cruz, C.M., Moody, S.E., Keiper, E.A., Ha, S.I., Cox, J.D., Belka, G.K. and Chodosh, L.A. (2002) Functional microarray analysis of mammary organogenesis reveals a developmental role in adaptive thermogenesis. Mol. Endocrinol., 16, 1185–1203.[Abstract/Free Full Text]

  14. Brambrink, T., Wabnitz, P., Halter, R., Klocke, R., Carnwath, J., Kues, W., Wrenzycki, C., Paul, D. and Niemann, H. (2002) Application of cDNA arrays to monitor mRNA profiles in single preimplantation mouse embryos. Biotechniques, 33, 376–378, 380, 382–385.[ISI][Medline]

  15. Mody, M., Cao, Y., Cui, Z., Tay, K.Y., Shyong, A., Shimizu, E., Pham, K., Schultz, P., Welsh, D. and Tsien, J.Z. (2001) Genome-wide gene expression profiles of the developing mouse hippocampus. Proc. Natl Acad. Sci. USA, 98, 8862–8867.[Abstract/Free Full Text]

  16. Hoffmann, R., Seidl, T., Neeb, M., Rolink, A. and Melchers, F. (2002) Changes in gene expression profiles in developing B cells of murine bone marrow. Genome Res., 12, 98–111.[Abstract/Free Full Text]

  17. Valerius, M.T., Patterson, L.T., Witte, D.P. and Potter, S.S. (2002) Microarray analysis of novel cell lines representing two stages of metanephric mesenchyme differentiation. Mech. Dev., 110, 151–164.[CrossRef][ISI][Medline]

  18. Mayanil, C.S., George, D., Freilich, L., Miljan, E.J., Mania-Farnell, B., McLone, D.G. and Bremer, E.G. (2001) Microarray analysis detects novel Pax3 downstream target genes. J. Biol. Chem., 276, 49299–49309.[Abstract/Free Full Text]

  19. Brunskill, E.W., Witte, D.P., Yutzey, K.E. and Potter, S.S. (2001) Novel cell lines promote the discovery of genes involved in early heart development. Dev. Biol., 235, 507–520.[CrossRef][ISI][Medline]

  20. Chen, J., Zhong, Q., Wang, J., Cameron, R.S., Borke, J.L., Isales, C.M. and Bollag, R.J. (2001) Microarray analysis of Tbx2-directed gene expression: a possible role in osteogenesis. Mol. Cell Endocrinol., 177, 43–54.[CrossRef][ISI][Medline]

  21. Sato, S., Nakamura, M., Cho, D.H., Tapscott, S.J., Ozaki, H. and Kawakami, K. (2002) Identification of transcriptional targets for Six5: implication for the pathogenesis of myotonic dystrophy type 1. Hum. Mol. Genet., 11, 1045–1058.[Abstract/Free Full Text]

  22. Rivolta, M.N., Halsall, A., Johnson, C.M., Tones, M.A. and Holley, M.C. (2002) Transcript profiling of functionally related groups of genes during conditional differentiation of a mammalian cochlear hair cell line. Genome Res., 12, 1091–1099.[Abstract/Free Full Text]

  23. Burton, G.R., Guan, Y., Nagarajan, R. and McGehee, R.E., Jr (2002) Microarray analysis of gene expression during early adipocyte differentiation. Gene, 293, 21–31.[CrossRef][ISI][Medline]

  24. Willert, J., Epping, M., Pollack, J.R., Brown, P.O. and Nusse, R. (2002) A transcriptional response to Wnt protein in human embryonic carcinoma cells. BMC Dev. Biol., 2, 8.[CrossRef][Medline]

  25. Linnarsson, S., Mikaels, A., Baudet, C. and Ernfors, P. (2001) Activation by GDNF of a transcriptional program repressing neurite growth in dorsal root ganglia. Proc. Natl Acad. Sci. USA, 98, 14681–14686.[Abstract/Free Full Text]

  26. Zhao, Y. and Potter, S.S. (2001) Functional specificity of the Hoxa13 homeobox. Development, 128, 3197–3207.

  27. Grimmond, S., Van Hateren, N., Siggers, P., Arkell, R., Larder, R., Soares, M.B., de Fatima Bonaldo, M., Smith, L., Tymowska-Lalanne, Z., Wells, C. et al. (2000) Sexually dimorphic expression of protease nexin-1 and vanin-1 in the developing mouse gonad prior to overt differentiation suggests a role in mammalian sexual development. Hum. Mol. Genet., 9, 1553–1560.[Abstract/Free Full Text]

  28. Conacci-Sorrell, M.E., Ben-Yendidia, T., Shtutman, M., Feinstein, E., Einat, P. and Ben-Ze'ev, A. (2002) Nr-CAM is a target gene of the b-catenin/LEF-1 pathway in melanoma and colon cancer and its expression enhances motility and confers tumorigenesis. Genes Dev., 16, 2058–2072.[Abstract/Free Full Text]

  29. McGill, G.G., Horstmann, M., Widlund, H.R., Du, J., Motyckova, G., Nishimura, E.K., Lin, Y.L., Ramaswamy, S., Avery, W., Ding, H.F. et al. (2002) Bcl2 regulation by the melanocyte master regulator Mitf modulates lineage survival and melanoma cell viability. Cell, 109, 707–718.[CrossRef][ISI][Medline]

  30. Bleul, C.C. and Boehm, T. (2001) Laser capture microdissection-based expression profiling identifies PD1-ligand as a target of the nude locus gene product. Eur. J. Immunol., 31, 2497–2503.[CrossRef][ISI][Medline]

  31. Furlong, E.E., Andersen, E.C., Null, B., White, K.P. and Scott, M.P. (2001) Patterns of gene expression during Drosophila mesoderm development. Science, 293, 1629–1633.[Abstract/Free Full Text]

  32. Golub, T.R., Slonim, D.K., Tamayo, P., Huard, C., Gaasenbeek, M., Mesirov, J.P., Coller, H., Loh, M.L., Downing, J.R., Caligiuri, M.A. et al. (1999) Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science, 286, 531–537.[Abstract/Free Full Text]

  33. Perou, C.M., Sorlie, T., Eisen, M.B., van de Rijn, M., Jeffrey, S.S., Rees, C.A., Pollack, J.R., Ross, D.T., Johnsen, H., Akslen, L.A. et al. (2000) Molecular portraits of human breast tumours. Nature, 406, 747–752.[CrossRef][Medline]

  34. Sorlie, T., Perou, C.M., Tibshirani, R., Aas, T., Geisler, S., Johnsen, H., Hastie, T., Eisen, M.B., van de Rijn, M., Jeffrey, S.S. et al. (2001) Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc. Natl Acad. Sci. USA, 98, 10869–10874.[Abstract/Free Full Text]

  35. Freiman, R.N., Albright, S.R., Zheng, S., Sha, W.C., Hammer, R.E. and Tjian, R. (2001) Requirement of tissue-selective TBP-associated factor TAFII105 in ovarian development. Science, 293, 2084–2087.[Abstract/Free Full Text]

  36. Kacharmina, J.E., Crino, P.B. and Eberwine, J. (1999) Preparation of cDNA from single cells and subcellular regions. Meth. Enzymol., 303, 3–18.[ISI][Medline]

  37. Luo, L., Salunga, R.C., Guo, H., Bittner, A., Joy, K.C., Galindo, J.E., Xiao, H., Rogers, K.E., Wan, J.S., Jackson, M.R. et al. (1999) Gene expression profiles of laser-captured adjacent neuronal subtypes. Nat. Med., 5, 117–122.[CrossRef][ISI][Medline]

  38. Wang, E., Miller, L.D., Ohnmacht, G.A., Liu, E.T. and Marincola, F.M. (2000) High-fidelity mRNA amplification for gene profiling. Nat. Biotechnol., 18, 457–459.[CrossRef][ISI][Medline]

  39. Dixon, A.K., Richardson, P.J., Lee, K., Carter, N.P. and Freeman, T.C. (1998) Expression profiling of single cells using 3 prime end amplification (TPEA) PCR. Nucl. Acids Res., 26, 4426–4431.[Abstract/Free Full Text]

  40. Iscove, N.N., Barbara, M., Gu, M., Gibson, M., Modi, C. and Winegarden, N. (2002) Representation is faithfully preserved in global cDNA amplified exponentially from sub-picogram quantities of mRNA. Nat. Biotechnol., 20, 940–943.[CrossRef][ISI][Medline]

  41. Schimenti, J. and Bucan, M. (1998) Functional genomics in the mouse: phenotype-based mutagenesis screens. Genome Res., 8, 698–710.[Abstract/Free Full Text]

  42. Kasarskis, A., Manova, K. and Anderson, K.V. (1998). A phenotype-based screen for embryonic lethal mutations in the mouse. Proc. Natl Acad. Sci. USA, 95, 7485–7490.[Abstract/Free Full Text]

  43. Hentges, K., Thompson, K. and Peterson, A. (1999) The flat-top gene is required for the expansion and regionalization of the telencephalic primordium. Development, 126, 1601–1609.[Abstract]

  44. Herron, B.J., Lu, W., Rao, C., Liu, S., Peters, H., Bronson, R.T., Justice, M.J., McDonald, J.D. and Beier, D.R. (2002) Efficient generation and mapping of recessive developmental mutations using ENU mutagenesis. Nat. Genet., 30, 185–189.[CrossRef][ISI][Medline]

  45. Gurdon, J.B. and Bourillot, P.Y. (2001) Morphogen gradient interpretation. Nature, 413, 797–803.[CrossRef][Medline]

  46. Dmochowski, I.J., Dmochowski, J.E., Oliveri, P., Davidson, E.H. and Fraser, S.E. (2002) Quantitative imaging of cis-regulatory reporters in living embryos. Proc. Natl Acad. Sci. USA, 99, 12895–12900.[Abstract/Free Full Text]

  47. Baldock, R., Bard, J., Brune, R., Hill, B., Kaufman, M., Opstad, K., Smith, D., Stark, M., Waterhouse, A., Yang, Y. and Davidson, D. (2001) The Edinburgh Mouse Atlas: using the CD. Brief. Bioinform., 2, 159–169.[Abstract/Free Full Text]

  48. Ko, M.S. (2001) Embryogenomics: developmental biology meets genomics. Trends Biotechnol., 19, 511–518.[CrossRef][ISI][Medline]

  49. Livesey, R. (2002) Have microarrays failed to deliver for developmental biology? Genome Biol., 3(8), comment2009.1–2009.5.

  50. Humpherys, D., Eggan, K., Akutsu, H., Friedman, A., Hochedlinger, K., Yanagimachi, R., Lander, E.S., Golub, T.R. and Jaenisch, R. (2002) Abnormal gene expression in cloned mice derived from embryonic stem cell and cumulus cell nuclei. Proc. Natl Acad. Sci. USA, 99, 12889–12894.[Abstract/Free Full Text]

  51. Smith, L., Underhill, P., Pritchard, C., Tymowska-Lalanne, Z., Abdul-Hussein, S., Hilton, H., Winchester, L., Williams, D., Freeman, T., Webb, S. and Greenfield, A. (2003) Single primer amplification (SPA) of cDNA for microarray expression analysis. Nucl. Acids Res., 31, e9.[Abstract/Free Full Text]


Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 Hum Mol GenetHome page
M. J. Abeyta, A. T. Clark, R. T. Rodriguez, M. S. Bodnar, R. A. R. Pera, and M. T. Firpo
Unique gene expression signatures of independently-derived human embryonic stem cell lines
Hum. Mol. Genet., March 15, 2004; 13(6): 601 - 608.
[Abstract] [Full Text] [PDF]

Home page
 J. Neurosci.Home page
J. Buchstaller, L. Sommer, M. Bodmer, R. Hoffmann, U. Suter, and N. Mantei
Efficient Isolation and Gene Expression Profiling of Small Numbers of Neural Crest Stem Cells and Developing Schwann Cells
J. Neurosci., March 10, 2004; 24(10): 2357 - 2365.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Smith, L.
Right arrow Articles by Greenfield, A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Smith, L.
Right arrow Articles by Greenfield, A.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?