Molecular Vision 2004; 10:637-649 <http://www.molvis.org/molvis/v10/a76/> Received 17 November 2003 | Accepted 6 August 2004 | Published 31 August 2004

Download Reprint

Comparative gene expression analysis of murine retina and brain

Abigail S. Hackam,¹ Jiang Qian,¹ Dongmei Liu,² Tushara Gunatilaka,¹ Ronald H. Farkas,¹ Itay Chowers,¹ Masaaki Kageyama,³ Giovanni Parmigiani,^4,5,6 Donald J. Zack^1,6,7,8
 
 

¹Guerrieri Center for Genetic Engineering and Molecular Ophthalmology at the Wilmer Eye Institute, the Departments of ²Biostatistics, ⁴Oncology, ⁵Pathology, ⁶Molecular Biology and Genetics, ⁷Neuroscience, and the ⁸McKusick-Nathans Institute of Genetic Medicine, Johns Hopkins University School of Medicine, Baltimore, MD; ³Santen Pharmaceuticals, Inc., Baltimore, MD

Correspondence to: Donald J. Zack, 809 Maumenee Building, Johns Hopkins University School of Medicine, 600 North Wolfe Street Baltimore, MD, 21287; Phone: (410) 502-5230; FAX: (410) 502-5382; email: dzack@bs.jhmi.edu
 
Dr. Hackam is now at the Bascom Palmer Eye Institute, University of Miami School of Medicine, Miami, FL.

Abstract

Purpose: Several high-throughput studies have described gene expression in the central nervous system (CNS), and recently there has been increasing interest in analyzing how gene expression compares in different regions of the CNS. As the retina is often used as a model system to study CNS development and function, we compared retina and brain gene expression using microarray analyses.

Methods: Mouse retina, brain and liver RNA was hybridized to a custom cDNA microarray containing 5,376 genes and ESTs, and the data from the quantified scanned images were analyzed using Bioconductor and SAM. Preferential retina expression was confirmed by real-time PCR. The cellular distribution of genes newly identified as retina enriched genes was determined by imunohistochemistry.

Results: Using stringent statistical analyses we identified 733 genes that were preferentially expressed in retina and 389 in brain. The retina-liver hybridizations identified an additional 837 retina enriched genes. The cellular distribution in the retina was determined for two genes that had not previously been reported to be expressed in the retina, the transcription regulatory proteins EWS and PCPB1. Both proteins were found primarily in the inner nuclear layer. Finally, a comparison of the microarray data to publicly available SAGE and EST library databases demonstrated only limited overlap of the sets of retina enriched genes identified by the different methodologies. The preferential retinal expression of a subset of genes from the microarray, which were not identified as differentially expressed by other methods, was confirmed by quantitative PCR.

Conclusions: The finding of differences in the groups of identified retina enriched genes from the various profiling techniques supports the use of multiple approaches to obtain a more complete description of retinal gene expression. Characterization of gene expression profiles of retina and brain may facilitate the understanding of the processes that underlie differences between the retina and other parts of the central nervous system.

Introduction

Various high-throughput methodologies are increasingly being used to describe the RNA profile of the retina under normal and disease conditions [1,2]. Microarray analyses of mouse [3-5], chick [6] and human [7-9] retinae have successfully identified numerous novel retina enriched genes and ESTs. Serial Analysis of Gene Expression (SAGE) has also been used to identify genes expressed in rod photoreceptors in the mouse [10,11] and human [12]. Sequencing and annotation of genes from eye and retina cDNA libraries from different developmental periods have been a rich source of transcripts expressed in retina [13,14] and innovative computation tools have been designed to query public EST databases for retina expressed genes and sequences [15,16]. Due to the various strengths and limitations of these techniques, the gene groups identified by the different studies overlap but are not identical [7]. The combination of the above studies has resulted in significant progress towards characterizing the complete transcriptome of the retina.

Similar approaches have also been applied to investigate gene expression in the brain, generating an extensive RNA profile of various brain regions under different experimental conditions [17-19]. Although the retina is extensively used as a model tissue for exploring general neuronal processes, few studies have compared gene expression between retina and brain. We recently reported studies that identified 345 genes that were differentially expressed in retina compared with brain in human [9], and 272 genes with higher expression in retina than brain in the embryonic chick [6]. Also, as part of a larger study of retina expressed genes in the mouse, SAGE techniques were used to compare gene expression in retina and one brain region, the hypothalamus [10,11]. The limited overlap of genes identified in expression profiles of individual tissues suggests that combining the results of multiple gene profiling techniques will lead to greater understanding of the molecular pathways contributing to the differences between retina and brain.

We previously described the use of a custom murine retinal microarray in an investigation of gene expression changes during photoreceptor degeneration in the rd1 mutant mouse [4]. The microarray was constructed by using all available ESTs from murine retina and brain, including clones selected from a mouse eye library, and clones specifically chosen based on function and molecular pathway. In the present study, we have used the custom retina microarray in order to generate an extensive comparison of gene expression in the retina, brain and liver tissue. Comparisons of the microarray results with published retina and brain SAGE and EST libraries revealed that many of the identified retina enriched genes were not previously described.

Methods

Custom retina array preparation

The array contains 5376 genes and was assembled from clones from a mouse eye cDNA library and from IMAGE consortium clones selected by their function (Research Genetics, Huntsville, AL), regardless of whether retinal expression was known [4]. The IMAGE clones were selected by searching literature resources and online databases (OMIM, PubMed, and Unigene EST databases at NCBI) [20] to include all genes with roles in normal retinal function and retinal diseases. Clones were also included by searching the Unigene database using keywords representing different functional classes. Finally, to include genes that may not have been present in the searches above, we also obtained all murine ESTs expressed in retina and brain that were available in Unigene. To further expand the microarray gene set, clones were also chosen from a partially sequenced adult mouse eye cDNA library, generously provided by Drs. Jeremy Nathans and Amir Rattner (Johns Hopkins University, Baltimore, MD). The clones were identified by BLAST searches using the NCBI public database, Unigene numbers were determined and duplicate clones were eliminated from the array by comparing corresponding Unigene numbers. Clones with poor sequence and those that did not have matches in the database were resequenced. Duplicates in the clone sets were removed to produce a non-redundant clone set.

The functional annotation of the datasets was performed using the SOURCE and GO databases [21,22]. Additionally, we used manual annotation from NCBI PubMed and LocusLink databases to confirm or provide missing annotations. The list of functions for each category is included in Appendix 1. The major functions of each gene were determined. In many cases more than one function was assigned.

The clones were rearrayed into 96-well plates and grown overnight in LB/10% glycerol (for bacterial clones) or SM buffer (for phage). The clone inserts were PCR amplified from the bacterial or phage stocks in duplicate 100 μl reactions using primers from flanking vector sequences. The PCR products were purified using Millipore Multiscreen PCR filters (Millipore, Bedford, MA) and eluted in TE (10 mM Tris-HCl pH 7.5, 0.1 mM EDTA) or water. Amplification efficiency and purification of each clone was verified by analyzing an aliquot of purified product on a 2% agarose gel. The PCR products were suspended in a final concentration of 50% DMSO and arrayed in duplicate onto SuperAmine silylated slides (Telechem, Sunnyvale, CA) using the Microgrid II arrayer with 100 μm-tip quill pins (Biorobotics, Cambridge, UK). Each gene was printed in duplicate on each array but was considered separately in the analyses. The printing environment was maintained at approximately 55% humidity and 22 °C. Spot size, morphology and quality were verified by sybr-green II staining (Molecular Probes, Eugene, OR) on several slides from each print.

RNA isolation and probe labeling

All procedures involving mice were carried out in accordance with the statement by the Association for Research in Vision and Ophthalmology for the Use of Animals in Ophthalmic and Vision Research and were approved by the Animal Care and Use Committee at Johns Hopkins University. Retina, brain and liver tissue were dissected, flash-frozen and stored at -80 °C. Tissue was pooled from animals of mixed ages and genotypes to minimize biological sources of variability and to obtain a wider profile of gene expression. Retinas were obtained from approximately 30 mice; brains were obtained from 8 animals and livers from 3 mice. Total RNA was isolated using phenol-chloroform extraction (Trizol reagent, Invitrogen, Carlsbad, CA). A reference sample was made by combining retina (30%) and brain (70%) RNA, derived from a mix of strains at various ages. Gel electrophoresis, A260/A280 absorbance ratios and a Bioanalyzer (Agilent, Palo Alto, CA) were used to assess RNA integrity.

Hybridization and data analysis

The custom microrrays were optimized and validated as described in Hackam et al. [4] and Chowers et al. [9], and microarray hybridization was based on the protocols described by Hedge et al [23]. A reference design experiment was used, in which each tissue was compared to a common reference sample. Four replicates were performed for each comparison, and included reciprocal "dye-swaps" to reduce dye-bias. Briefly, 20 μg of total RNA were treated with 2 U amplification grade DNase I (Ambion, Austin, TX), purified through RNeasy columns (Qiagen, Chatsworth, CA) then first-strand cDNA was synthesized by SuperScript II reverse transcriptase (Invitrogen) in reaction mix (6 μg random hexamers [Invitrogen], 10 mM DTT 1.25 mM dATP, dCTP and dGTP, 1 mM dTTP [Invitrogen] and 0.25 mM aminoallyl dUTP [Sigma, St. Louis, MO]), at 42 °C for 16 h. The cDNA was purified using Tris-free buffers, dried then resuspended in 4.5 μl 0.1 Na₂CO₃ buffer, pH 9.0, mixed with 4.5 μl monoreactive NHS-cy3 or NHS-cy5 dye (Amersham Biosciences, Piscataway, NJ) resuspended in DMSO, and incubated for 2 h at room temperature. The dye coupled samples were purified using QIAquick PCR purification columns (Qiagen), dried and resuspended in hybridization solution (50% formamide, 5X SSC, 0.1% SDS).

The microarray slides were hybridized at 42 °C overnight using 10 μg poly(dA) DNA (Pharmacia, New York, NY) and 10 μg C_ot = 1 DNA fraction mouse DNA (Invitrogen) to reduce non-specific hybridization. The arrays were washed in 1X SSC/0.2% SDS (at 42 °C), then 0.1X SSC/0.2%SDS and 0.1X SSC (both at room temperature). The slides were scanned using a ScanArray 5000 scanner and and ScanArray software version 4.1 (Perkin Elmer, Boston, MA). Image analysis and quantification was performed using Imagene software version 5.6 (Biodiscovery, Inc., Marina del Rey, CA). A linear combination method was used to calculate the adjusted intensity measurements for arrays with multiple scans, as described in [7]. Loess normalization was performed based on MVA (multivariate analysis) plots using local background subtracted data with the software program Bioconductor [24]. Low intensity spots were floored to a threshold defined by the average intensity of all low intensity spots within the same array. Spots that were floored for both cy3 and cy5 channels were not included in the normalization. Statistical analysis of the normalized and transformed data was performed using the Significance Analysis of Microarrays (SAM) program (version 1.21) [25].

Comparisons to SAGE and EST libraries were performed using a t test based algorithm. T statistics were calculated for the SAGE comparison using the gene expression values from 4 retina libraries and 8 non-retina libraries, and for the EST library comparison using values from 3 retina and 5 non-retina libraries. The SAGE datasets were from Blackshaw et al. [10], and the non-normal adult retina EST libraries used were 20165|NIH_BMAP_Ret2, 20632|NIH_MGC_94, and 20167|NIH_BMAP_Ret3. Comparisons to SAGE and EST libraries were performed using a two sample t-test with the assumption of equal variance. The genes were sorted by significance and the genes with the highest significance (using the same number of genes as in the microarray group) were compared with the microarray dataset.

CNS regional expression of retina enriched genes was performed on the 13 brain regions using the published Affymetrix data from Su, et al. [26] (Gene Expression Atlas) and visualized by assigning colors to the log of the average signal intensity values.

Quantitative PCR

Quantitative real-time PCR (QPCR) was performed using the Lightcycler-FastStart DNA Master SYBR Green I kit (Roche Diagnostics, Nutley, NJ), according to the manufacturer's instructions. cDNA was synthesized from 1 μg of total RNA from retina, brain or liver tissue with Thermoscript reverse transcriptase (Invitrogen). Primers were chosen from exons separated by large introns, and the PCR quality and specificity was verified by melting curve dissociation analysis and gel electrophoresis of the amplified product. Primer sequences are listed in Table 1. Relative transcript levels of each gene in each tissue were calculated using the second derivative maximum values from the linear regression of cycle number versus log concentration of the amplified gene. Amplification of the control gene ARP (acidic ribosomal phosphoprotein P0 (ARP)) [27] was used for normalization.

Immunohistochemistry

Adult C57/Bl6 mouse eyes were fixed in 4% paraformaldehyde/0.1 M phosphate buffer, equilibrated in 5% sucrose at 4 °C overnight, then incubated in a series of increasing sucrose concentrations (5-20%). The tissues were then embedded in a 2:1 (v/v) mixture of OCT (Sakura Finetechnical Co., Ltd., Tokyo, Japan) and 20% sucrose in phosphate buffer. Cryosections (10 μm thick) were collected onto Superfrost Plus microscope slides (Fisher Scientific, Atlanta, GA). The slides were washed in PBS to remove the OCT then blocked in 10% normal goat serum (Sigma)/PBS with 0.3% Triton X-100 at room temperature. The slides were then incubated in primary antibody against Ewing sarcoma protein (EWS) [28] or poly(rC) binding protein 1 (PCPB1) [29], diluted in blocking solution for 16 h at 4 °C. After several washes in PBS the slides were incubated with Alexa fluor 488 goat anti-rabbit antibody (Molecular Probes), which was diluted in blocking solution containing the nuclear stain DAPI, for 1 h at room temperature, washed in PBS and covered with a cover-slip. The sections were viewed with a Nikon microscope and digitally captured for visualization.

Results

Comparison of retina and brain gene expression using a custom microarray

The expression profile of mouse retina, GEO entry (pending), was compared to whole brain using a custom cDNA microarray. Each tissue was hybridized with a common reference sample, and a dye-swap design was incorporated to eliminate the influence of dye bias effects (see Methods). The Significance Analysis of Microarrays (SAM) algorithm [25] was used to identify statistically significant expression differences. This algorithm incorporates the variability of replicate arrays to calculate the likelihood that observed expression level changes reflect "real" expression level differences rather than experimental artifacts. A false discovery rate (FDR) was calculated by comparing the expression ratios from the microarrays to ratios of randomly permuted control and experimental groups (in this case, the different tissues being compared). The choice of FDR is a balance between the number of differentially expressed genes identified and the number of false positives. Using SAM with a high statistical stringency (FDR=4.5%, meaning that 4.5% of the genes were likely to be improperly determined to be differentially expressed), we identified 733 genes that were preferentially expressed in retina and 389 genes that were preferentially expressed in the brain (Appendix 1).

The functional classification of a subset of retina enriched and brain enriched genes is shown in Figure 1. Many genes important to neuronal activity were differentially expressed, such as ion channels (transport class) and cytoskeletal proteins (structural class). Genes involved in general processes, such as energy generation (metabolism class), also showed retina or brain enriched expression. Compared to the total number of genes with preferential expression, we found higher frequencies of proliferation and cell death genes differentially expressed in the brain than retina, and nucleic acid processing genes in retina than brain. As expected, known photoreceptor genes were exclusively expressed in retina, demonstrating the specificity of the hybridization. Over a third of the differentially expressed clones were ESTs, representing potentially novel genes that are enriched or even specific to retina or brain.

Comparison of retina and brain with non-neuronal tissue

To identify neuronal enriched genes, we compared the gene profiles of retina and brain to liver. Increasing the stringency by using a lower FDR (1.4%) resulted in the identification of 837 genes that were preferentially expressed in retina and 487 that were preferentially expressed in the liver (Appendix 1). Similarly, the brain-liver comparison showed 776 genes with higher expression in the brain and 421 in the liver (FDR 1.2%; Appendix 1). The higher number of genes identified in the retina than brain likely reflects the selection bias towards retinal genes used to construct the array. In total, the microarray identified 1134 unique retina enriched genes in the retina-brain and retina-liver comparisons (Appendix 1). Of these, 431 genes were common to both comparisons. The functional classification of the top retina enriched 30 genes identified by both comparisons is shown in Table 2.

Quantitative real-time PCR (QPCR) was used to confirm the differential expression of nine genes and ESTs. To independently validate the microarray results, RNA samples extracted from mouse retinas not used on the microarrays were tested by QPCR. Two of the top retina enriched genes in Table 2 were tested by QPCR: TPA regulated locus and RIRD-1 (IMAGE 4505626). The other genes that were tested represent a broad range of expression differences in order to avoid the potential bias of confirming the most differentially expressed genes. The QPCR analysis on each of the selected genes confirmed the retina enriched expression, although the magnitude of expression was frequently underestimated in the microarray compared with the QPCR for some of the genes and tissues. This independent confirmation of the microarray results indicates that the custom arrays accurately reflect biological tissue differences (Table 3).

Distribution of retina enriched genes in the CNS

To further characterize the retina enriched genes found in the retina-brain comparisons, particularly those not known to be involved in vision, we determined the regional distribution and expression levels of these genes in subregions of the brain and nervous system. The retina enriched gene set was compared to publicly available Affymetrix oligonucleotide microarray data from 13 different neuronal regions, including whole eye [26] (Gene Expression Atlas). We determined the tissue distribution of a group of retinal differentially expressed genes (Figure 2A) and ESTs (Figure 2B) that had high SAM scores (excluding photoreceptor genes). The majority of genes and ESTs were expressed in several brain regions at varying levels. Several genes, for example Hif-1, showed higher expression in eye than other brain tissues. Other genes had more restricted expression, such as B-cell translocation gene 1, which was expressed in eye, olfactory bulb and cerebellum. Several of the differentially expressed genes and ESTs did not show preferential expression in eye, possibly due to a lower proportion of the corresponding transcripts in the whole eye compared with retina, or to differences in hybridization kinetics between Affymetrix arrays and our cDNA microarrays.

Retinal expression pattern of two differentially expressed genes, Ewing sarcoma protein and poly(C)-binding protein 1

We next determined the cellular distribution in the retina of two genes that were found to be differentially expressed on the microarray. Retinal expression for these genes was a novel finding from this study. Both genes are known to be involved in transcriptional regulation in other cell types. EWS contains an N-terminal transcriptional activation domain and a C-terminal RNA-binding domain [30,31]. Chromosomal translocations associated with Ewing sarcoma and primitive neuroectodermal tumor result in the fusion of EWS to several different transcription factor genes, forming proteins that have oncogenic transcriptional activities. A brain-specific splice variant of EWS may be involved in neural differentiation, and the spliced 18 nucleotide exon has been suggested to impede the oncogenic properties of EWS fusion transcripts [28]. Although the aberrant fusion proteins have been studied in detail, the activity and tissue distribution of native EWS are not well described. QPCR demonstrated that EWS expression was seven fold higher in the retina than brain (data not shown). As shown in Figure 3, EWS was located in the cytoplasm in retinal cells. The cellular distribution of EWS in the retina was striking. There was prominent staining in the INL and in cellular extensions reaching into the ganglion cell layer. The positions of the stained cells suggested that EWS was expressed most highly in horizontal cells and Müller glia processes (Figure 3, top and middle). There was also marked staining in a subset of photoreceptor inner segments (Figure 3, top). Other retinal cells also expressed EWS but at lower levels. There was no signal detected in the no primary antibody negative control (data not shown). A similar staining pattern was observed in bovine retina (data not shown).

The second gene analyzed, PCPB1, is a 40 kDa RNA binding protein that regulates gene expression post-transcriptionally by stabilizing specific RNAs [29,32,33]. PCPB1-dependent regulation has been best described for viral genes, and there is increasing evidence that it regulates expression of mammalian genes such as erythropoietin [32]. A recent study reported the interaction of the family member poly(C)-binding protein 2 (PCPB2) with over a hundred human genes, including transcription factors, proto-oncogenes and genes involved in cell signaling [34]. The role of poly(C)-binding proteins in the retina have not been explored. The QPCR results demonstrated that PCPB1 expression was two fold higher in the retina than brain (data not shown).

We found that PCPB1 was present in most cells in the retina, although the staining intensity varied widely (Figure 3, bottom). Intense staining was observed in cells at the edge of the inner plexiform layer, consistent with the horizontal cell layer, in photoreceptor inner segments and in the ganglion cell layer. PCPB1 appeared to be nuclear in ganglion cells. No signal was evident in the no primary antibody negative control (data not shown). This analysis illustrates that the microarray in combination with immunochemistry is a powerful mechanism for characterizing expression patterns of retina enriched genes.

Identification and chromosomal localization of novel retina genes

Many of the retina genes identified by the arrays, SAGE and library databases are known photoreceptor genes, and a number of these have been associated with human retinal degenerations. We recently reported the localization of a group of putative photoreceptor expressed genes identified in the analysis of the rd1 mouse to retinal disease regions [4]. To determine whether any additional genes could be considered candidate molecules in causing hereditary retinal disease, the chromosomal localization of the human homologues of the retina enriched genes and ESTs found in this study were compared with the chromosomal intervals linked to retinal diseases (defined by the RetNet database). Known genes with causative mutations, and their corresponding chromosomal regions, were omitted. In total, 205 of the retina enriched genes were mapped to retinal disease regions (Table 4 and Appendix 1). In addition to several intriguing candidate genes of known function, numerous novel genes and ESTs were localized within the critical regions.

Comparison of retina enriched genes to SAGE and library databases

As part of ongoing research community efforts to define the mouse retina transcriptome, we further analyzed the retina enriched genes identified in the microarray comparisons above. The majority of clones on the custom mouse microarray used in this study were selected based on their function, whether or not information on their retina expression was known. Therefore, it is possible that there will be differences between the groups of genes that were identified as retina enriched in this study and non-array studies. To determine the amount of overlap between the techniques, the genes identified by the microarray were compared to the publicly available SAGE dataset of retina expressed genes [10]. Out of 1134 retina enriched genes identified by the custom microarrays (from the retina-brain and retina-liver comparisons), a total of 914 genes were also found in the SAGE dataset.

Next we wanted to determine how many genes in the total SAGE dataset were preferentially expressed in retina. A t test based algorithm (see Methods) was designed to compare retina-derived SAGE tags [10] to tags from assorted extra-ocular tissues, allowing the construction of a "SAGE reference dataset" of retina enriched genes. This dataset was then sorted by significance, in which genes that were found more frequently in retina than in other tissues were distinguished from genes that had similar expression levels in retina as in other tissues (unpublished data). The 1134 retina enriched genes from the retina-brain and retina-liver microarray hybridizations were compared to this SAGE reference dataset. The majority of the 914 genes found in both the microarray and SAGE studies were defined as retina enriched only on the microarray. Only 119 (11%) genes overlapped, in that they were determined as retina enriched by both SAGE and microarray (Appendix 1). These overlapping genes included many known photoreceptor genes, confirming that both methods are effective at identifying retina-specific genes.

A "library reference dataset" was also created by comparing databases of retina EST libraries to libraries of other tissues (Gene Expression Omnibus, see Methods) to generate a dataset of library-derived retina enriched genes. The 1134 retina enriched genes from the microarray hybridizations were compared to this dataset, as above. There was low overlap with the retina enriched genes identified on the microarray. We found that 102 (9%) of the retina enriched genes from the microarray were shared with the library reference dataset (Appendix 1). Twenty-eight of these genes were also identified as retina enriched by SAGE. A Venn diagram summarizing these comparisons is shown in Figure 4. Notably, we confirmed by quantitative PCR (Table 3) the retinal-enrichment of eight genes that were differentially expressed in retina on the microarrays but were not indicated as retina enriched in the SAGE or library datasets.

Discussion

The retina is a highly specialized sensory region of the brain. Retina and brain have common embryonic origins and contain many of the same cell types. Indeed, numerous functional and pathological studies have used retinal neurons as a substitute for neurons found elsewhere in the CNS. In this study we utilized a custom microarray to characterize molecular differences between murine retina and brain. We identified 733 genes that were preferentially expressed in retina and 389 genes that were preferentially expressed in brain. Retina and liver comparisons identified an additional 837 primarily neuronal-specific retina expressed genes.

Many of the differentially expressed genes in the retina are not obviously related to visual function. One prominent class was genes that are involved in nucleic acid metabolism. Two RNA binding proteins, EWS and PCPB1, were identified as more highly expressed in the retina than brain. Immunohistochemistry showed that EWS was widely distributed in the retina, with more intense staining in the inner nuclear layer, and was found in a pattern consistent with Müller glia processes. EWS has been proposed to mediate RNA stability, synthesis and processing [35-37]. Although most studies have demonstrated nuclear localization of EWS in various cell lines, including our own analysis on HEK 293 cells (unpublished data), our results in retina demonstrated a cytoplasmic localization. Cytoplasmic EWS has previously been implicated in G-protein coupled receptor cell signaling [38]. Determining the protein binding partners of EWS and identifying which genes are regulated by EWS will be important to understanding its role in the retina.

PCPB1 was most highly expressed in a pattern that was consistent with horizontal cells and ganglion cells. The function of PCPB1 in normal retina or neuronal tissue has not been reported. Previous studies have shown that hypoxic cortical neuron cultures and rat cerebral cortex have increased expression of PCPB1 and decreased expression of its related isoform, PCPB2 [39]. We found that PCPB1 was highly expressed in the inner nuclear layer, and was also observed at lower levels in the choroid. Future studies will investigate whether PCPB1 related gene regulation is a component of the cellular response to altered oxygen conditions in the retina.

Comparisons between the microarray results and other studies demonstrated that a large number of the genes found in this analysis were not previously recognized as being retina enriched by SAGE and library databases. However, quantitative PCR validation experiments did confirm retina enriched expression for a group of genes identified in the microarray analysis. This difference among gene profiling methodologies is consistent with observations in the human retina [9]. Retina enriched genes identified by human retinal cDNA microarray had limited overlap with genes reported as retina enriched by EST data-mining (28 genes in common) [40], human retina SAGE (11 genes) [12] and mouse retina SAGE (8 genes) [10]. Similar comparisons between methodologies have been quantified in other tissues. A study of human corneal endothelial cells found low (29%) overlap of genes identified by Affymetrix microarray and SAGE [41]. Additionally, 41% of a set of 1000 rat hippocampal transcripts identified by SAGE were reliably detected by Affymetrix microarrays [42]. Expression level was a good (but not absolute) predictor for detection efficiency: high abundance transcripts (SAGE tags greater than 50) were more reliably detected by microarrays than medium or low abundance transcripts [42].

The profiling techniques described above use vastly different experimental designs to identify expressed genes. Although these are obviously powerful techniques, each has the potential for under-representation of transcripts. Issues of probe synthesis and hybridization can alter the efficiency of microarray hybridization, sequence dependent cloning difficulties can limit the appearance of some transcripts in EST libraries, and low efficiency tag anchoring, sequencing failures and inappropriate mapping of tag to genomic sequences can affect SAGE results. Since each of the techniques has limitations of transcript sensitivity, many more retinal expressed genes may yet be discovered. Combining gene profiling with microdissection, single cell analysis and RNA amplification will help identify genes expressed in less frequent cell types that are diluted out when analyzing the whole tissue. Increasing the sensitivity of existing techniques, the development of new methods, and the refinement of statistical tools to accurately analyze expression data, will be crucial for reliably identifying low abundance and rare transcripts. The differences in the groups of identified retina enriched genes support the use of multiple approaches to obtain a more complete description of the retinal transcriptome.

Acknowledgements

The anti-EWS antibody was a generous gift from Dr. O. Delattre (INSERM U 509, Pathologie Moléculairedes Cancers, Paris, France) and anti-PCPB1 antibody was a generous gift from Dr. R. Andino (Department of Microbiology and Immunology, University of California, San Francisco). The authors thank Dr. Celia Cingolani (Johns Hopkins University) for retinal dissections tissue, Drs. Jeremy Nathans and Amir Rattner (Johns Hopkins University) for providing the sequenced arrayed mouse eye cDNA library used to construct the microarrays, and Mr. Daniel Bondroff for assistance with the figures.

This work was supported by grants from the National Eye Institute and Macula Vision Foundation, and by generous gifts from Mr. and Mrs. Marshall and Stevie Wishnack and from Mr. and Mrs. Robert and Clarice Smith. Dr. Abigail S. Hackam was supported by a Canadian Institutes of Health Research Senior Research Fellowship. Dr. Donald J. Zack is the Guerrieri Professor of Genetic Engineering and Molecular Ophthalmology, and a recipient of a Senior Investigator Award from Research to Prevent Blindness, Inc.

Funding for this study was also provided by Santen Pharmaceuticals, Inc. Under a licensing agreement between Santen Pharmaceuticals, Inc. and the Johns Hopkins University, Drs. Hackam, Zack, Chowers, Kageyama, and Farkas are entitled to a share of royalty received by the University on sales of potential products described in this article. The terms of this arrangement are being managed by Johns Hopkins University in accordance with its conflict of interest policies.

References

1. Swaroop A, Zack DJ. Transcriptome analysis of the retina. Genome Biol 2002; 3:REVIEWS1022.

2. Wilson AS, Hobbs BG, Speed TP, Rakoczy PE. The microarray: potential applications for ophthalmic research. Mol Vis 2002; 8:259-70 <http://www.molvis.org/molvis/v8/a33/>.

3. Farjo R, Yu J, Othman MI, Yoshida S, Sheth S, Glaser T, Baehr W, Swaroop A. Mouse eye gene microarrays for investigating ocular development and disease. Vision Res 2002; 42:463-70.

4. Hackam AS, Strom R, Liu D, Qian J, Wang C, Otteson D, Gunatilaka T, Farkas RH, Chowers I, Kageyama M, Leveillard T, Sahel JA, Campochiaro PA, Parmigiani G, Zack DJ. Identification of Gene Expression Changes Associated with the Progression of Retinal Degeneration in the rd1 Mouse. Invest Ophthalmol Vis Sci 2004; 45:2929-42.

5. Diaz E, Yang YH, Ferreira T, Loh KC, Okazaki Y, Hayashizaki Y, Tessier-Lavigne M, Speed TP, Ngai J. Analysis of gene expression in the developing mouse retina. Proc Natl Acad Sci U S A 2003; 100:5491-6.

6. Hackam AS, Bradford RL, Bakhru RN, Shah RM, Farkas R, Zack DJ, Adler R. Gene discovery in the embryonic chick retina. Mol Vis 2003; 9:262-76 <http://www.molvis.org/molvis/v9/a38/>.

7. Chowers I, Liu D, Farkas RH, Gunatilaka TL, Hackam AS, Bernstein SL, Campochiaro PA, Parmigiani G, Zack DJ. Gene expression variation in the adult human retina. Hum Mol Genet 2003; 12:2881-93.

8. Yoshida S, Yashar BM, Hiriyanna S, Swaroop A. Microarray analysis of gene expression in the aging human retina. Invest Ophthalmol Vis Sci 2002; 43:2554-60.

9. Chowers I, Gunatilaka TL, Farkas RH, Qian J, Hackam AS, Duh E, Kageyama M, Wang C, Vora A, Campochiaro PA, Zack DJ. Identification of novel genes preferentially expressed in the retina using a custom human retina cDNA microarray. Invest Ophthalmol Vis Sci 2003; 44:3732-41.

10. Blackshaw S, Fraioli RE, Furukawa T, Cepko CL. Comprehensive analysis of photoreceptor gene expression and the identification of candidate retinal disease genes. Cell 2001; 107:579-89.

11. Blackshaw S, Kuo WP, Park PJ, Tsujikawa M, Gunnersen JM, Scott HS, Boon WM, Tan SS, Cepko CL. MicroSAGE is highly representative and reproducible but reveals major differences in gene expression among samples obtained from similar tissues. Genome Biol 2003; 4:R17.

12. Sharon D, Blackshaw S, Cepko CL, Dryja TP. Profile of the genes expressed in the human peripheral retina, macula, and retinal pigment epithelium determined through serial analysis of gene expression (SAGE). Proc Natl Acad Sci U S A 2002; 99:315-20.

13. Yu J, Farjo R, MacNee SP, Baehr W, Stambolian DE, Swaroop A. Annotation and analysis of 10,000 expressed sequence tags from developing mouse eye and adult retina. Genome Biol 2003; 4:R65.

14. Mu X, Zhao S, Pershad R, Hsieh TF, Scarpa A, Wang SW, White RA, Beremand PD, Thomas TL, Gan L, Klein WH. Gene expression in the developing mouse retina by EST sequencing and microarray analysis. Nucleic Acids Res 2001; 29:4983-93.

15. Katsanis N, Worley KC, Gonzalez G, Ansley SJ, Lupski JR. A computational/functional genomics approach for the enrichment of the retinal transcriptome and the identification of positional candidate retinopathy genes. Proc Natl Acad Sci U S A 2002; 99:14326-31.

16. Bortoluzzi S, d'Alessi F, Danieli GA. A novel resource for the study of genes expressed in the adult human retina. Invest Ophthalmol Vis Sci 2000; 41:3305-8.

17. Sandberg R, Yasuda R, Pankratz DG, Carter TA, Del Rio JA, Wodicka L, Mayford M, Lockhart DJ, Barlow C. Regional and strain-specific gene expression mapping in the adult mouse brain. Proc Natl Acad Sci U S A 2000; 97:11038-43.

18. Gunnersen JM, Augustine C, Spirkoska V, Kim M, Brown M, Tan SS. Global analysis of gene expression patterns in developing mouse neocortex using serial analysis of gene expression. Mol Cell Neurosci 2002; 19:560-73.

19. Gustincich S, Batalov S, Beisel KW, Bono H, Carninci P, Fletcher CF, Grimmond S, Hirokawa N, Jarvis ED, Jegla T, Kawasawa Y, LeMieux J, Miki H, Raviola E, Teasdale RD, Tominaga N, Yagi K, Zimmer A, Hayashizaki Y, Okazaki Y, RIKEN GER Group, GSL Members. Analysis of the mouse transcriptome for genes involved in the function of the nervous system. Genome Res 2003; 13:1395-401.

20. Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. J Mol Biol 1990; 215:403-10.

21. Diehn M, Sherlock G, Binkley G, Jin H, Matese JC, Hernandez-Boussard T, Rees CA, Cherry JM, Botstein D, Brown PO, Alizadeh AA. SOURCE: a unified genomic resource of functional annotations, ontologies, and gene expression data. Nucleic Acids Res 2003; 31:219-23.

22. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis AP, Dolinski K, Dwight SS, Eppig JT, Harris MA, Hill DP, Issel-Tarver L, Kasarskis A, Lewis S, Matese JC, Richardson JE, Ringwald M, Rubin GM, Sherlock G. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 2000; 25:25-9.

23. Hegde P, Qi R, Abernathy K, Gay C, Dharap S, Gaspard R, Hughes JE, Snesrud E, Lee N, Quackenbush J. A concise guide to cDNA microarray analysis. Biotechniques 2000; 29:548-50,552-4,556.

24. Dudoit S, Yang JYH. Bioconductor R packages for exploratory analysis and normalization of cDNA microarray data. In: Parmigiani G, Garrett ES, Irizarry RA, Zeger SL, editors. The analysis of gene expression data: methods and software. New York: Springer Verlag; 2003. p. 45-65.

25. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001; 98:5116-21. Erratum in: Proc Natl Acad Sci U S A 2001; 98:10515.

26. Su AI, Cooke MP, Ching KA, Hakak Y, Walker JR, Wiltshire T, Orth AP, Vega RG, Sapinoso LM, Moqrich A, Patapoutian A, Hampton GM, Schultz PG, Hogenesch JB. Large-scale analysis of the human and mouse transcriptomes. Proc Natl Acad Sci U S A 2002; 99:4465-70.

27. Simpson DA, Feeney S, Boyle C, Stitt AW. Retinal VEGF mRNA measured by SYBR green I fluorescence: A versatile approach to quantitative PCR. Mol Vis 2000; 6:178-83 <http://www.molvis.org/molvis/v6/a24/>.

28. Melot T, Dauphinot L, Sevenet N, Radvanyi F, Delattre O. Characterization of a new brain-specific isoform of the EWS oncoprotein. Eur J Biochem 2001; 268:3483-9.

29. Gamarnik AV, Andino R. Two functional complexes formed by KH domain containing proteins with the 5' noncoding region of poliovirus RNA. RNA 1997; 3:882-92.

30. Ohno T, Rao VN, Reddy ES. EWS/Fli-1 chimeric protein is a transcriptional activator. Cancer Res 1993; 53:5859-63.

31. Ohno T, Ouchida M, Lee L, Gatalica Z, Rao VN, Reddy ES. The EWS gene, involved in Ewing family of tumors, malignant melanoma of soft parts and desmoplastic small round cell tumors, codes for an RNA binding protein with novel regulatory domains. Oncogene 1994; 9:3087-97.

32. Czyzyk-Krzeska MF, Bendixen AC. Identification of the poly(C) binding protein in the complex associated with the 3' untranslated region of erythropoietin messenger RNA. Blood 1999; 93:2111-20.

33. Silvera D, Gamarnik AV, Andino R. The N-terminal K homology domain of the poly(rC)-binding protein is a major determinant for binding to the poliovirus 5'-untranslated region and acts as an inhibitor of viral translation. J Biol Chem 1999; 274:38163-70.

34. Waggoner SA, Liebhaber SA. Identification of mRNAs associated with alphaCP2-containing RNP complexes. Mol Cell Biol 2003; 23:7055-67.

35. Belyanskaya LL, Delattre O, Gehring H. Expression and subcellular localization of Ewing sarcoma (EWS) protein is affected by the methylation process. Exp Cell Res 2003; 288:374-81.

36. Bertolotti A, Melot T, Acker J, Vigneron M, Delattre O, Tora L. EWS, but not EWS-FLI-1, is associated with both TFIID and RNA polymerase II: interactions between two members of the TET family, EWS and hTAFII68, and subunits of TFIID and RNA polymerase II complexes. Mol Cell Biol 1998; 18:1489-97.

37. Araya N, Hirota K, Shimamoto Y, Miyagishi M, Yoshida E, Ishida J, Kaneko S, Kaneko M, Nakajima T, Fukamizu A. Cooperative interaction of EWS with CREB-binding protein selectively activates hepatocyte nuclear factor 4-mediated transcription. J Biol Chem 2003; 278:5427-32.

38. Felsch JS, Lane WS, Peralta EG. Tyrosine kinase Pyk2 mediates G-protein-coupled receptor regulation of the Ewing sarcoma RNA-binding protein EWS. Curr Biol 1999; 9:485-8.

39. Zhu Y, Sun Y, Mao XO, Jin KL, Greenberg DA. Expression of poly(C)-binding proteins is differentially regulated by hypoxia and ischemia in cortical neurons. Neuroscience 2002; 110:191-8.

40. Stohr H, Mah N, Schulz HL, Gehrig A, Frohlich S, Weber BH. EST mining of the UniGene dataset to identify retina-specific genes. Cytogenet Cell Genet 2000; 91:267-77.

41. Gottsch JD, Seitzman GD, Margulies EH, Bowers AL, Michels AJ, Saha S, Jun AS, Stark WJ, Liu SH. Gene expression in donor corneal endothelium. Arch Ophthalmol 2003; 121:252-8.

42. Evans SJ, Datson NA, Kabbaj M, Thompson RC, Vreugdenhil E, De Kloet ER, Watson SJ, Akil H. Evaluation of Affymetrix Gene Chip sensitivity in rat hippocampal tissue using SAGE analysis. Serial Analysis of Gene Expression. Eur J Neurosci 2002; 16:409-13.