Molecular Vision 2002; 8:205-220 <http://www.molvis.org/molvis/v8/a27/> Received 31 August 2001 | Accepted 14 December 2001 | Published 15 June 2002

Download Reprint

Expressed sequence tag analysis of human RPE/choroid for the NEIBank Project: Over 6000 non-redundant transcripts, novel genes and splice variants

Graeme Wistow,¹ Steven L. Bernstein,² M. Keith Wyatt,¹ Robert N. Fariss,¹ Amita Behal,¹ Jeffrey W. Touchman,³ Gerard Bouffard,³ Don Smith,¹ Katherine Peterson¹
 
 

¹Section on Molecular Structure and Function, National Eye Institute, National Institutes of Health, Bethesda, MD; ²Departments of Ophthalmology and Neurobiology & Genetics, University of Maryland School of Medicine, Baltimore, MD; ³NIH Intramural Sequencing Center, Gaithersburg, MD

Correspondence to: Graeme Wistow, Ph.D., Chief, Section on Molecular Structure and Function, National Eye Institute, Building 6, Room 331,National Institutes of Health, Bethesda, MD, 20892-2740; Phone: (301) 402-3452; FAX: (301) 496-0078; email: graeme@helix.nih.gov

Abstract

Purpose: The retinal pigment epithelium (RPE) and choroid comprise a functional unit of the eye that is essential to normal retinal health and function. Here we describe expressed sequence tag (EST) analysis of human RPE/choroid as part of a project for ocular bioinformatics.

Methods: A cDNA library (cs) was made from human RPE/choroid and sequenced. Data were analyzed and assembled using the program GRIST (GRouping and Identification of Sequence Tags). Complete sequencing, Northern and Western blots, RH mapping, peptide antibody synthesis and immunofluorescence (IF) have been used to examine expression patterns and genome location for selected transcripts and proteins.

Results: Ten thousand individual sequence reads yield over 6300 unique gene clusters of which almost half have no matches with named genes. One of the most abundant transcripts is from a gene (named "alpha") that maps to the BBS1 region of chromosome 11. A number of tissue preferred transcripts are common to both RPE/choroid and iris. These include oculoglycan/opticin, for which an alternative splice form is detected in RPE/choroid, and "oculospanin" (Ocsp), a novel tetraspanin that maps to chromosome 17q. Antiserum to Ocsp detects expression in RPE, iris, ciliary body, and retinal ganglion cells by IF. A newly identified gene for a zinc-finger protein (TIRC) maps to 19q13.4. Variant transcripts of several genes were also detected. Most notably, the predominant form of Bestrophin represented in cs contains a longer open reading frame as a result of splice junction skipping.

Conclusions: The unamplified cs library gives a view of the transcriptional repertoire of the adult RPE/choroid. A large number of potentially novel genes and splice forms and candidates for genetic diseases are revealed. Clones from this collection are being included in a large, nonredundant set for cDNA microarray construction.

Introduction

The retinal pigment epithelium (RPE) is a monolayer of neural ectoderm derived cells that is located between the photoreceptor layer of the retina and the blood supply of the choroid. The RPE fulfills several vital functions for the retina. It scavenges shed discs from the photoreceptor outer segments and recycles their components, particularly the retinoids that provide the visual pigment. It also transports nutrients from the copious blood supply of the choroid into the retina (and thereby to other parts of the eye) and transports waste in the opposite direction.

The RPE and choroid together thus play an essential part in the health and function of the eye. Inherited retinal degenerative diseases have been associated with mutations of genes with RPE expression [1-5] and damage to the RPE/choroid is also thought to be an early event in age related macular degeneration (AMD), the major cause of severe vision loss in many aging populations [6-8]. Consequently, there is considerable interest in further defining the molecular constituents of the RPE and developing tools, such as microarrays, for functional analysis.

Here we describe a cDNA library made from dissected human RPE/choroid from 75-80 year old donors, as part of the NEIBank project to improve genomics and bioinformatics resources for the eye [9]. Expressed sequence tag (EST) analysis of over 9000 clones has produced a large dataset of nonredundant gene clusters. These include large numbers of potentially novel genes as well as identification of many known genes expressed in these tissues. The complete dataset will be available through a web site. Additional sequence and continuing annotations will be added as the project continues. This paper describes the initial analysis of the data with some key examples. These include novel genes, possible alternative splice variants of genes involved in eye disease and candidates for other disease loci.

Methods

Tissue and RNA preparation

Human post-mortem eye tissues were obtained under University of Maryland School of Medicine IRB exemptions SB-019701 and SB-129901. Total RNA was extracted using RNAzol or TRIzol (Tel-Test Inc., Friendswood, TX). PolyA RNA was prepared using an oligo(dT) cellulose affinity column [10]. For RPE/choroid, two different donor eyes (75-80 years old) yielded approximately 600 mg of dissected tissue. This in turn yielded 340 μg of total RNA and 7 μg of mRNA.

cDNA library construction, sequence analysis, and clustering

A directionally cloned cDNA library in the pCMVSPORT6 vector was constructed at Life Technologies (Rockville, MD; now part of Invitrogen Corp), essentially following the protocols of the SuperScript Plasmid System (Invitrogen Corp.). The library code designation was cs. For this library, cDNA inserts were cloned into the NotI/MluI sites of the vector.

Sequencing was performed at the NIH Intramural Sequencing Center (NISC) and data were analyzed as described in detail elsewhere [11]. High quality sequences were clustered and identified using GRIST (GRouping and Identification of Sequence Tags) [12].

Sequences were also searched through genome resources at the National Center for Biotechnology Information, the Human Genome Project, and the Celera Genomics Group. Protein motifs were searched using GenomeNet and the Swiss Institute of Bioinformatics.

RH mapping

Radiation hybrid mapping for selected genes was performed at Research Genetics (Huntsville AL), using the Stanford G3 panel (Stanford Human Genome Center, Stanford, CA). PCR primers were designed from the 3' UTR of each sequence, a region unlikely to be interrupted by introns in the genome. PCR was used to amplify unique marker sequences from a total of 83 clones and two controls. An email server operated by the Stanford Human Genome Center was used to link the marker to more than 15000 framework markers. For clones cs22c04 and bx08d05 (oculospanin) primers oc9+: AAATCCTGGAGCTGACCCTC and oc9-: GGACCTTAGCTCCACCATCA were used.

Antiserum production and western analysis

Antiserum for a selected novel protein (Ocsp) was produced by selection of an antigenic peptide sequence with a natural N-terminal cysteine. The peptide was synthesized, linked to carrier, and used to produce antisera in rabbits by Quality Controlled Biochemicals (QCB, Hopkinton, MA). After testing, one bleed was selected for affinity purification at QCB and was then used for subsequent experiments.

Human lens, retina, RPE/choroid, iris, and ciliary body extracts were prepared essentially as described before [13,14]. Proteins were separated by SDS PAGE [14,15] using 18% gels in Tris-glycine SDS for 3 h at 120V and were transferred to nitrocellulose membranes (S&S, Keene, NH) using the Novex systems (Novex, San Diego CA). Western blots were performed as described before [16]. Membranes were incubated overnight with anti serum, diluted 1/1000, processed and visualized using the Vectastain Elite ABC (HRP) kit with DAB substrate for peroxidase (Vector Labs, Burlingame, CA), following manufacturer's instructions.

Immunofluorescence

These studies were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Adult rats were housed under fluorescent lights operating on a 12 h lights on / 12 h light off lighting cycle. Rats were killed by CO₂ asphyxiation in a closed chamber. Following nucleation, eyes were dissected and fixed by immersion for 2 h in freshly prepared 4% formaldehyde in isotonic phosphate buffered saline pH 7.3. Eyes were washed in chilled isotonic PBS (3 x 20 min). To cryoprotect tissues prior to freezing, tissues were incubated in 10% sucrose in PBS for 60 min and then transferred to 20% sucrose in PBS for 60 min. Tissues were embedded in O.C.T. embedding medium (Tissue-Tek, SakuraFinetek Inc.) and frozen in liquid nitrogen. Cryosections (12 μm) were cut and collected on Fisher Superfrost/Plus slides and air dried overnight.

Cryosections were blocked with 5% normal goat serum diluted in ICC buffer (PBS + 0.5% BSA + 0.2% Tween-20 + 0.1% sodium azide, pH 7.3) then incubated overnight at 4 °C in a rabbit polyclonal anti-peptide antibody to oculospanin diluted 1:400 in ICC buffer. Primary antibody was omitted from sections used as negative controls. Sections were washed repeatedly and incubated in the dark for 4 h with the nuclear dye DAPI (1 μg/ml) and a goat anti-rabbit Alexa568-conjugated secondary antibody (Molecular Probes) diluted in ICC buffer. Following repeated washing, sections were mounted in Gel Mount (Bio-Meda) and secured with a coverslip. Specimens were analyzed on a Leica SP2 laser scanning confocal microscope equipped with Nomarski optics. Immunolabeled and negative control sections were imaged under identical scanning conditions. Files were imported into Photoshop 5.5 (Adobe) and converted to psd format for analysis.

Polymerase chain reaction (PCR) methods

PCR was used to validate alternative splice forms, to obtain probes for hybridization and to complete sequences. For template, a sample of the complete cDNA library representing at least one million primary clones was amplified and plasmid was isolated using reagents from the QIAGEN Plasmid Kit (QIAGEN, Valencia, CA). Fragments were amplified using either Taq (Roche, Indianapolis, IN) or Elongase (Life Technologies/Invitrogen Corp.) polymerase systems, following manufacturers' protocols.

For 5' RACE [17], 1 μg of total RNA was transcribed with the appropriate primer using Superscript RT (Life Technologies/Invitrogen) following manufacturer's protocols. 10% of the resulting cDNA template was then used for PCR amplification with Taq, using 30 cycles of; 94 °C for 1 min, 60 °C for 1 min, and 72 °C for 1 min and a final 10 min extension at 72 °C. Sequence specific primers were used for first strand synthesis and cDNA was G-tailed using terminal transferase. Reagents for RACE were taken from the Life Technologies 5' RACE system kit, following manufacturer's instructions.

Northern blot

PCR was used to generate a 590 bp fragment of the alpha transcript from the RPE library. Primers used were: 5'-CCCGCTGCTATTAGAATGCATTGT and 5'-AACTGGAAGCTCCTTCTATAGTCT. The fragment was cloned into the pCR II dual promoter plasmid (Invitrogen, Carlsbad CA) and confirmed by sequencing. Nonradioactive riboprobes labelled with digoxigenin (DIG) were generated using reagents and protocols from the DIG RNA Labeling Kit (Roche). The antisense probe was generated by linearizing with BamHI and using T7 RNA polymerase in the presence of DIG labeled dNTPs. The sense probe was similarly generated by linearizing with XhoI and using Sp6 RNA polymerase. The entire reaction containing either probe was boiled for 5 min and cooled on ice for 1 min. The probes were added separately to 5 ml of DIG Easy Hyb (Roche, Indianapolis, IN), previously equilibrated to 68 °C in a rotating incubator, and applied to northern blots. The blots were hybridized overnight at 68 °C, briefly rinsed in 2X SSC at room temperature, and then washed two times in 0.2X SSC, 0.1% SDS for 15 min at 68 °C. Successful hybridization was detected using antibodies against DIG conjugated with alkaline phosphatase (Roche). Blots were exposed to Hyperfilm ECL film (Amersham Pharmacia Biotech, Little Chalfont, UK), typically for less than 20 s. Multi-tissue northerns were purchased from Clontech (Palo Alto, CA). Northerns for human and monkey (Macaca mulatta) eye tissue were prepared as described previously [18].

Results & Discussion

Library

The cs library contains 3.9x10⁷ primary recombinants with an average insert size of 1.5 kbp. Out of over 10,000 sequence reads, less than 1% contain no insert or bacterial contamination while 3% contain mitochondrial genome and 0.5% contain rRNA. 16% of the reads were rejected because of poor sequence quality according to the program Phred [19]. The average length of high quality sequence reads was 500 bp.

Compared with other cDNA libraries, particularly from lens (by), the RPE/choroid sequence data have a much flatter distribution with fewer classes of highly abundant ESTs and, consequently, a high content of single or double hits. After analysis through GRIST [12], 9326 unique clones produce 6337 clusters, potentially representing unique genes. Of these clusters, only 507 (8%) contain more than two members. Currently, approximately 45% of the clusters do not match named genes in GenBank, although many have matches with unidentified Unigene clusters or other ESTs in dbEST. These numbers are, of course, subject to change as more human genes are cloned and named. All sequences are available through NEIBank.

RPE markers

To validate any tissue specific library, it is useful to identify known marker genes. Numerous genes known to be associated with RPE, including pigment epithelium differentiation factor [20], Bestrophin [4], prominin [21], OTX2 [22], Kir7.1 [23], Cystatin C [24] and RPE65 [25] are present in the cs clone collection. Another RPE marker, RPE retinal G protein coupled receptor (RGR) [26] is represented by a single clone (cs86g02) that apparently contains intron sequence. This may be a novel splice form, but further work is needed for complete characterization. Contamination of the library with retina seems to be low, as judged by the appearance of only two ESTs for opsin, the major transcript found in our retina dataset [27], and a single clone for α-transducin.

It is important to remember that the results presented here are from combined RPE and choroid. Many transcripts must therefore be derived from ocular vasculature. However, it is striking how many of the most prominently expressed genes are known to be markers for RPE or for neural cells and how many are held in common with the iris, another tissue containing cells of neural origin.

Similarity between RPE/choroid and Iris

The most abundant cDNAs from cs are for the ubiquitous protein elongation factor 1α, which is highly abundant in libraries from many tissues. Other abundant transcripts are more interesting (Table 1). The second ranking transcript, (accounting for 0.5% of the total), is the protective enzyme glutathione peroxidase 3 (GSH-PO). Previous studies using quantitative electron microscopic immunocytochemistry did not detect GSH-PO protein in either normal or diseased human RPE cells [28]. However, levels of mRNA and protein are not always congruent. A similar discrepancy between abundant mRNA and undetectable protein is also seen for ferritin in the lens, possibly indicating post-transcriptional control of a protective protein [11,29].

It is possible that the GSH-PO transcripts are derived preferentially from the choroid, rather than the RPE itself. However, it is interesting that the high levels of GSH-PO in the cs library are mirrored in the unamplified human iris cDNA library (bx) [30]. This is just one of many similarities in EST content between these libraries that may reflect similarities between RPE and iris pigment epithelial (IPE) cells [31]. Other abundantly expressed genes that are shared between the cs and bx libraries include the enzyme prostaglandin D2 synthetase, TIMP3, the locus of Sorsby's fundus dystrophy [32] and myocilin, which is mutated in inherited open angle glaucoma and is thought to play a key role in the trabecular meshwork [33,34]. As might be expected, the two libraries additionally share a number of genes that are markers for pigmented cells, notably the highly abundant Pmel17 [35,36], whose homolog RPE1 was originally cloned from bovine RPE [37], and enzymes of melanin production, such as tyrosinase related protein 1 (TYRP1) and dopachrome tautomerase (TYRP2).

Glutamine synthetase

One of the most abundant transcripts in the RPE/choroid library is for the enzyme glutamine synthetase (GS) (Table 1). GS converts glutamate and ammonia, both potentially toxic compounds, into glutamine. Since glutamate is a major neurotransmitter of the retina, this role may be particularly important in the eye. In the mature neural retina, GS is localized to Muller cells [38]. In developing bovine retina, GS first localizes exclusively in RPE, but disappears as the RPE matures [39]. Glutamate is a major neurotransmitter of the retina, so the neuroprotective role of RPE expressed GS may be particularly important for maintaining normal RPE and retinal function. Increased levels of GS in adult retina and RPE may be associated with disease. The high levels seen in the cs library could therefore reflect the age or perhaps the undiagnosed disease status of a donor. GS expression in neural tissues is also regulated by glucocorticoids [38], so pre-morbid medication or stress induced corticosteroid expression could also be playing a role in elevated GS expression. These are issues upon which future microarray studies may be able to shed some light. GS is another transcript shared by iris, with two ESTs in the bx collection.

Oculoglycan/opticin: alternative splicing

Another highly abundant cDNA shared between RPE/choroid and iris is that for oculoglycan/opticin (Optc). Optc is a newly discovered member of the small leucine rich repeat proteoglycan (SLRP) family [40-42] whose gene maps to chromosome 1q31 [40] close to ARMD1, a locus associated with age related macular degeneration [43]. Alignment of the cluster of 15 ESTs for Optc from RPE/choroid reveals a novel alternative splice form. Eight partial sequences cover a region that contains a deletion in one of the clones (cs116h07).

The gene structure for Optc can now be deduced from available genome project sequences. The complete gene is contained in a single BAC clone (GenBank accession number AL391817). It is divided into eight exons, with the fairly unusual feature of an intron that interrupts the 3' UTR of the transcript (Figure 1). The first exon is noncoding while exons 2 and 3 encode the signal peptide and first domain of Optc. Exons 4-7 encode the leucine rich repeat (LRR) domain [40] with its seven LRR motifs and its class III SLRP cysteine rich signature sequence [44]. The repeat in protein sequence is not reflected in the exon/intron structure with one, three, two, and one LRR respectively in exons 4-7.

The deletion observed in EST cs116h07 is the result of an alternative splicing event in which an internal 5' splice site for exon 4 is used (Figure 1). This deletion of 66 bp is in frame and has an interesting effect. It neatly removes the class III signature sequence while leaving all the LRR motifs intact. The removal of a conserved motif is likely to have functional consequences for the protein product; unfortunately no role has yet been assigned to the signature sequence. PCR on DNA templates made from bx and cs libraries detected both major and minor splice forms in both iris and RPE/choroid. Other members of the SLRP family are also expressed in RPE/choroid and are described below in the section on "Matrix and matrix proteases".

Oculospanin

The similarity between RPE and iris is further illustrated by the discovery in the libraries from both tissues of cDNAs (cs22c04 and bx08d01) for a novel protein. The complete sequence of the longer clone (bx08d01) was determined (Figure 2; GenBank accession number AF325213). The insert is 1601bp in length and strikingly G/C rich (69%). It contains an open reading frame (ORF) of 1068 bp that predicts a protein of 355 amino acids, with a size of 36.4 kDa and a predicted pI of 5.7. Sequence comparisons show that this protein belongs to the large superfamily of tetraspan proteins or tetraspanins, integral membrane proteins with four transmembrane helices [45]. To reflect its discovery in eye, the protein was named oculospanin (Ocsp).

Tetraspanins are widespread, numerous and largely mysterious in function [45]. Two related tetraspan proteins of the retina, ROM1 and RDS/peripherin, are both involved in retinal degenerations [46]. To gain some insight into the distribution of Ocsp protein in the eye, antisera were raised against the specific peptide CIDPREDGASVNDQ (Figure 2). The affinity purified antibody, designated OCSPp1, was tested in western blot of human eye tissues. A single positive band of approximately 42 kDa was seen in iris, ciliary body, RPE, and retina but not in lens (Figure 3). The expected size of Ocsp predicted from cDNA sequence is 36 kDa. The larger size may be due to posttranslational modification, such as glycosylation, which is common in tetraspanins [45]. Indeed, the predicted amino acid sequence of Ocsp contains a potential N-linked glycosylation site in the probable extracellular loop between third and fourth transmembrane helices (Figure 2).

The OCSPp1 antiserum was also used in confocal microscopy immunofluorescence analysis of rat eye tissues (Figure 4). Prominent staining is seen in RPE (marked by an asterisk) and the inner plexiform layer of the neural retina, with weaker staining in regions of the outer plexiform and photoreceptor layers. Staining is also clear in the ciliary body and the anterior surface of the iris. Very weak signal is observed in corneal epithelium (not shown) and no staining is seen in lens.

The expression pattern suggests that the gene for Ocsp, like those for ROM1 and peripherin, is a candidate for involvement in inherited eye disease. When we began this work, there was no gene sequence for OCSP in the Human Genome Project or the public Celera Genomics Group databases, so radiation hybrid mapping was used to localize the OCSP gene to chromosome 17q25, between markers SHGC-10493 and SHGC-31983. A subsequent examination of the subscription version of the Celera human genome has allowed us to locate partial gene sequence for human OCSP. The gene has three exons and is located close to PDE6G on chromosome 17q. Efforts are underway to complete the gene sequence. We note that an inherited retinal degenerative disease, RP17, has been mapped to chromosome at 17q22 [47].

This is made difficult by the lack of any information on OCSP gene structure. Lacking any published human genome sequence for this gene, specific PCR primers were designed from the cDNA and used to screen available collections of human genomic clones. However, although the primers can successfully and specifically amplify an exon fragment of human OCSP from total human genomic DNA, no positive genomic clones have been identified in any of the human BAC, PAC, or P1 genomic libraries available through Incyte Genomics (St. Louis, MO). The OCSP gene seems to be located in one of the gaps in human genome sequence. As these gaps are filled in, more information on its gene structure should become available. In the meantime, attempts are underway to map the gene structure by direct PCR amplification of genomic DNA However, this too is proving to be a challenge, again probably due to the high G/C content of OCSP coding sequence.

An abundant mystery transcript: "Alpha" a gene from the BBS1 region

The third most abundant cDNA from the RPE/choroid library is for a transcription unit that was identified during the search for the MEN1 (multiple endocrine neoplasia type 1) gene in a 2.8 Mbp region of chromosome 11q13 [48]. One of the anonymous transcript regions discovered was designated "alpha". The cs library collection contains a remarkable cluster of 41 ESTs from the alpha gene region (Table 1). In contrast to the other similarities seen with iris, only one clone for the alpha gene is found in the unamplified iris collection. While ESTs for this gene, originating from kidney, brain, and other tissues can be found in dbEST, this gene does not currently have its own Unigene cluster. Due to several chimeric clones in the public databases, alpha ESTs are clustered with Unigenes for various other genes on different chromosomes. Intriguingly, the location of the alpha gene falls in the closely defined region for Bardet-Biedl Syndrome type 1 (BBS1) [49], a congenital disease with multiple clinical manifestations, including retinitis pigmentosa, renal failure, diabetes and mental retardation.

Surprisingly, assembly of our sequences together with others from dbEST produces an almost continuous distribution along the genomic region of the alpha gene (about 10 kbp) with no evidence of intron structure. In addition, exhaustive examination of the sequence contigs has uncovered no evidence for any significant ORF, although a number of short ORFs are of course present, as would be expected for any long sequence. While the majority of our directionally cloned ESTs share a common orientation, the data do not presently define a clear 3' end for the alpha gene transcript. The sequence is quite A/T rich and ESTs exhibit a wide range of what appear to be internal oligo(dT) priming sites at A-rich stretches in the gene sequences.

To confirm the orientation and size of the alpha gene transcript and to examine tissue distribution, a 590 bp fragment of the assembled cDNA sequence was amplified by PCR and used to generate single strand "sense" and "antisense" RNA riboprobes for Northern blots. On a human multi-tissue Northern, the antisense probe detects a single 8.5 kb band in all tissues (Figure 5A) while the sense probe gives no hybridization, confirming the orientation of the gene (not shown). The most intense hybridization is observed in kidney and heart with lower levels in several other tissues, which is generally consistent with the abundance of non-eye ESTs in dbEST. To confirm expression in eye, the antisense probe was also used on a total RNA blot of rhesus monkey RPE/choroid, fovea, and peripheral retina (Figure 5A). The same single 8.5 kb band is observed in all three tissues.

Since alpha has no likely ORF, no apparent introns, and no identified polyA tail, the function of the transcript is mysterious. Indeed, it may not be a protein coding mRNA and could very well be a novel functional or structural RNA. The transcript is apparently well conserved among species. The antisense probe hybridizes at high stringency with a single 8.5 kb band on a mouse multi-tissue Northern, a result which would not be expected for the UTR of a protein coding gene (Figure 5B). As in human, mouse kidney again shows high expression, with lower expression detected in brain and lung. We can conclude that the product of the alpha gene is a well conserved, widely expressed RNA, with some as yet unknown function. The location and expression pattern of the alpha gene make it an interesting candidate for BBS1, but the lack of identifiable protein coding exons makes it difficult to check for meaningful mutations at this time.

Splice variants: Bestrophin

EST analysis has some advantages over other methods, such as SAGE [50], for examining the transcript repertoire of tissues. In particular, EST sequences that cover regions of coding sequence can reveal variant transcripts and splice forms, many of which have functional significance. A good example of this from our data is provided by Bestrophin, a protein of unknown function that is localized to the basolateral plasma membrane of the retinal pigment epithelium [51]. Mutations in the Bestrophin gene cause Best's macular dystrophy, a condition producing accumulation of lipofuscin in the RPE [4]. The structure and function of Bestrophin are unknown, but the protein belongs to a family of putative transmembrane proteins (known as "nematode family 8") that were first described in C. elegans [52]. This family is predicted to contain four transmembrane helices, similar to the prediction for Bestrophin [53]. In the cs data, there is a cluster of seven ESTs for transcripts of this gene. Surprisingly, the majority of the ESTs derive from variant splice forms that could give rise to significantly modified Bestrophin proteins (Figure 6).

In one of these clones (cs55b03), the third intron of the Bestrophin gene is retained. Interestingly the ORF continues into the intron sequence for 57 codons, certainly long enough to encode a structural motif. If translated, such a transcript would produce a shortened version of Bestrophin ("Bestrophin-S") consisting of the first 82 residues of the canonical protein. Bestrophin-S would contain the first predicted transmembrane helix and extracellular loop attached to a novel C-terminal tail of 57 residues that is glycine rich (suggesting flexibility) and basic in charge.

Two other ESTs (cs60c05 and cs98c05, neither of which are full length at the 5' end) also contain a skipped splice junction. In canonical Bestrophin mRNA, the removal of intron 10 causes the ORF to terminate in exon 11 after only 5 more residues. In the two variant ESTs, the splice site at the 3' end of exon 10 is ignored and part of intron 10 is retained as the 3' end of the mRNA. As in the case of Bestrophin-S, the ORF continues into the intron for a significant length, in this case for 84 codons (Figure 6). Protein produced from these transcripts would have an essentially complete Bestrophin sequence with an additional C-terminal domain, giving a complete unmodified size of 76.2 kDa for a long version of Bestrophin ("Bestrophin-L"). While the putative novel domain has no strong sequence similarity with other proteins, it contains several histidine and cysteine residues in clusters that are suggestive of a possible metal-binding motif (Figure 6). These two clones for Bestrophin-L terminate with two different polyadenylation sites, located 948 bp and 1097 bp respectively into what would otherwise be intron 10.

Since these sequence variants were initially detected from 5' sequence reads of incomplete cDNA clones, longer clones in the collection, which have only been partially sequenced, could also contain the 3' variant sequence. Indeed, when the three remaining ESTs that are complete at the 5' end were checked for 3' sequence, only one (cs104h03) was found to have the canonical version of the Bestrophin ORF splicing to exon 11 while the other two (cs15a10 and cs29h02) contain the "long" version from read through into intron 10. Remarkably, it thus seems that the majority of Bestrophin cDNAs in this library derived from 80 year old RPE/choroid actually encode Bestrophin-L.

It is possible that the variant transcripts are "mistakes", produced by errors in the mRNA splicing machinery. If so, it raises the possibility that a significant fraction of Bestrophin transcripts may be aberrant in the samples from these old donors. Accumulation of aberrantly spliced Bestrophin forms during aging could contribute to the accumulation of lipofuscin in the aged RPE and retina. Alternatively, one or both of the variants may represent biologically significant alternative splice forms of Bestrophin, generating proteins with distinct functions. Indeed, a different alternative splice form of this transcript has been previously identified [4]. Skipping splice junctions is a known mechanism for generating protein diversity in other genes. For example, a skipped splice junction causes the insertion of 42 residues in the human lens protein Mp19 [11]. An intron run on splice variant, similar to the intron 10 run on in Bestrophin, has also been observed in the human UNC-119 homolog, a retina preferred gene [27,54].

Transcription factors: Pax6, Optx2

Not all alternative transcripts change protein sequence. Others define multiple gene promoters or give rise to mRNA species with altered sequence or structure that may be involved with post-transcriptional regulatory mechanisms. The cs data include a group of five ESTs for Pax6, a transcription factor with an essential role in normal eye development [55,56]. In quail and mouse, two promoters (P0 and P1) have been identified for this gene, each giving rise to alternative first exons [57,58]. Of four full length cDNAs from RPE/choroid, two arise from the proximal P1 promoter. A third contains a different first exon, located almost 7 kbp upstream in the gene sequence, that has sequence similarity with the first exon product of the mouse P0 promoter. This suggests that both Pax6 promoters are active in the adult RPE/choroid. A fourth clone contains most of the short first intron of the P1 transcript, producing an mRNA with a longer 5' UTR. Again, this could simply represent a failure in the splicing mechanism, but it could also have some significance for post-translational regulation [59,60].

In addition to Pax6, the cs collection contains representatives of many other transcription factors that may have significance for tissue preferred gene expression in RPE and choroid. This includes an interesting cluster of two ESTs from the transcription factor Optx2/Six6.This is unexpected, since Optx2/Six6 expression occurs early in chick eye development but ceases in the cells that form the pigment epithelium [61]. Furthermore, expression of recombinant Optx2/Six6 causes differentiation of chick RPE cells into neural retina and photoreceptors [61]. These clones may derive specifically from choroid or from retina contamination in the RPE dissection, although contamination seems unlikely since there are two different clones for this factor, which are not represented at high levels among cDNAs from the retina itself, while, for example, the abundant retina transcript NRL is not present in cs [27]. The specific origin of the Optx2/Six5 transcripts remains to be determined, but the possibility that the gene is expressed in human RPE or choroid is interesting.

A large number of other transcription factors and related proteins are represented by ESTs in the cs collection (Table 2). These include members of the Maf and Sox families, some of whose members have important roles in eye development [62-69]. Those found so far in RPE/choroid are MafG, Sox8, and Sox10, while one clone (cs96a10) contains sequence that closely matches mouse Sox5 (GenBank accession number NM_011444.1) and also a human BAC clone sequence (GenBank accession number AC087319), but no named human gene. Three members of the forkhead family, FOXF2, forkhead box OA1, and forkhead box OA3 are in the cs collection. Of these, FOXF2 has recently been shown to be associated with epithelio-mesenchymal interactions and to be localized in the RPE/choroid region in mouse embryos [70].

TIRC: a novel zinc-finger protein

The cs data contain many novel or unidentified transcripts. One cluster of three such transcripts (cs27e04, cs57e12, cs104f11) has been named TIRC (Three In RPE Choroid). Assembly of the three sequences reveals an ORF containing 11 consensus C2H2 type zinc finger (ZF) motifs, suggesting that it might be a transcription factor. BLAST searches identify a group of ESTs (Hs.22340) from brain, placenta, and a number of tumor cell cDNA libraries. The entire cDNA sequence is represented as an apparently intronless gene in a single BAC clone (GanBank accession number AC008813) that is localized on human chromosome 19q13.4. This is close to the mapped position of the retinal degeneration RP11, although this disease has recently been localized to the PRP31 gene [71].

Examination of the gene sequence for TIRC shows that the ORF extends in the 5' direction further than any of the ESTs from RPE or other tissues. The predicted amino acid sequence of this region is similar to the N-terminal regions of several large ZF proteins of unknown function. Including this predicted region, the largest unspliced ORF from the TIRC gene would contain 469 codons, coding for a protein of over 54 kDa (Figure 7). BLASTX of the complete TIRC gene sequence shows partial ORFs in the 3' UTR that are also similar to ZFs. These "ghostly fingers" may be remnants of ancestral coding sequence. Conceivably, TIRC arose from a processed pseudogene that subsequently acquired active expression and, presumably, a new role in RPE and other tissues. Whatever its history, TIRC is a member of a transcription factor family expressed in adult human RPE/choroid and may therefore have a role in control of gene expression in these tissues.

Growth factors

In both cs and bx libraries, a member of the insulin-like growth factor binding protein (IGFBP) family [72] is ranked in the twenty most abundant cDNAs. However, in contrast to their other similarities, the libraries differ in the identity of the particular IGFBP family member. In the iris collection, IGFBP7 is most abundant, representing 0.4% of the total number of cDNA clones [30], while IGFBP5 has similar overall representation in the RPE/choroid collection. IGFBPs were first identified as modulators of IGF activity, but it is now apparent that they also have independent growth-factor-like roles [72]. The high level of IGFBP5 in aged RPE/choroid is particularly intriguing, since work in other systems has linked elevated expression of IGFBP5 with programmed cell death [73]. A list of growth factor and growth factor related cDNAs identified so far in the RPE/choroid sequence collection is shown in Table 3. A total of five different IGFBPs are represented, along with the growth factor IGF-II and its receptor. Two members of the bone morphogenic protein (BMP) growth factor family are also detected; BMP-5 and BMP-7, a factor which is required for retinal pigmentation and for lens development [74].

Matrix and matrix proteases

In the healthy RPE, the monolayer of pigmented cells is strongly attached to Bruch's membrane [75]. This attachment is critically important for maintaining the differentiated state of the pigmented cells and in some states of injury or disease, RPE cells may detach and migrate, forming structures similar to contractile scars that distort the retina in a condition called proliferative vitreoretinopathy (PVR) [75]. Both for normal attachment and in PVR, extracellular matrix and the enzymes, such as metalloproteinases, that affect it, are of central importance. Table 4 lists cDNAs for matrix and related proteins and Table 5 lists the metalloproteinases and inhibitors that have been identified so far in the cs collection.

It is noticeable that several of the matrix related proteins identified in this collection are also associated with development and maintenance of bone tissue. For example, one of the most abundant transcripts is for SPARC/osteonectin, a protein that is normally associated with remodeling in bone and wound healing [76], but which has, nevertheless, been shown to be expressed in both chicken and monkey RPE [77,78]. Other proteins represented in the cs collection that were originally named for roles in bone include osteopontin and the related MEPE (matrix extracellular phosphoglycoprotein), which are both members of a family of bone-tooth mineral matrix phospho-glycoproteins [79] and osteoadherin and osteoglycin which are members of the SLRP family [44,80,81]. Three other members of the SLRP family, Optc, biglycan, and decorin are also present, with Optc by far the most abundant, as described above. The cs collection also includes many other transcripts related to extracellular matrix (Table 3) including numerous collagens and laminins. In addition, ESTs for two eye specific transcripts, interphotoreceptor matrix proteoglycans 1 and 200 [82,83] are present.

There are also representatives for several members of protein families that contain thrombospondin repeats (TSR) [84]. The founder member of this class is thrombospondin-1 (TSP-1), a component of the extracellular matrix that is widely expressed during development. TSP-1 interacts with many other proteins, including SLRP proteins, collagen, fibronectin, and TGFβ1 [84]. However, the common TSR domain is widespread and is found in proteins ranging from semaphorins (Table 4), to connective tissue growth factor (Table 2). Among the collection of TSP-1 related proteins is a single, partial clone (cs79e06) for a novel TSR containing protein, designated rpe-spondin. This clone has been completely sequenced (GenBank accession number AY040546) and contains an incomplete ORF of 298 amino acid residues (Figure 8). Search of the protein motif databases shows that the sequence contains both somatamedin B (vitronectin peptide) (Pfam accession number 01033) and thrombospondin type 1 domains (Pfam accession number 00090). Both of these are cysteine rich motifs frequently associated with matrix components. The gene for rpe-spondin is located on human chromosome 8. The dbEST database now contains some other tags for this transcript, mostly from tumor derived libraries, and these are grouped in Unigene cluster Hs.122544.

The TSR is also found in ADAMTS proteins, members of the ADAM (a disintegrase and a metalloproteinase) family [84,85] of soluble extracellular matrix proteases whose substrates are other extracellular matrix proteins. As shown in Table 5, the cs collection contains ADAMTS1 and ADAMTS10 in addition to three other ADAM proteins. Four other metalloproteinases are represented, two of which are functionally related. MMP14 (or MT1-MMP) is a membrane bound enzyme that activates the soluble enzyme MMP2, which in turn cleaves type IV collagen of the basement membrane [86]. BMP1 (Table 5) is also an enzyme that acts on procollagen [87]. TIMP2 is an inhibitor of MMP2 [86] and is represented by four ESTs in the cs collection; however, by far the most abundant inhibitor represented at the EST level is TIMP3, the locus of Sorsby's fundus dystrophy [32].

Sequences for all these factors are included in the nonredundant set of human eye transcripts to be used in future microarray experiments aimed at comparing expression profiles in age, disease and development.

Junctional proteins

Cell-cell adhesion and communication is another critical feature of RPE/choroid organization. In particular, formation of tight junctions in epithelial cells blocks diffusion of solutes between cells and creates a physical separation between apical and basolateral membranes [88]. This is key to the polarity of cell layers and essential to one of the major functions of the RPE, which is the directional transport of a wide variety of substances in and out of the retina. The RPE monolayer also creates the blood-retina barrier, another essential structure that separates the neural domain of the retina from the immune system, just as also occurs with the blood-brain barrier [89]. Table 6 lists cDNAs for proteins involved in cell adhesion and junctions. These include the tight junction protein 1 (ZO1) that is a key feature of tight junctions and is known to be required for the blood-retinal barrier in the RPE [90], as well as cerebral adhesion molecule, a factor associated with the blood-brain barrier [91].

Cadherins and protocadherins form another class of cell adhesion related proteins that define cell-cell interactions [92]. Two cadherins are found in the cs collection. The most abundant is cadherin related protein 23, which was recently shown to be the locus of Usher's syndrome 1D [93,94]. The other is OB-cadherin (cadherin 11), which, as for some of the matrix related proteins in RPE/choroid, has been previously associated mainly with osteoblasts [95]. More in keeping with the neural ectoderm lineage of the RPE, the cs collection also includes a number of protocadherins [96].

Another marker for RPE cell contacts and communication is the connexin Cx43, whose level of expression seems to be correlated with normal RPE function in a rat model [97]. Cx43 is the only connexin observed so far in the RPE/choroid cDNA collection. Also included in Table 6, are two members of the recently described junctophilin (JP) family. These are important components of the junctional complexes that form between plasma membrane and endoplasmic reticulum in excitable cells and are involved in coupling cellular responses (such as Ca²⁺ release) to external stimuli [98]. JP-3 is the major form expressed in neural tissue, while JP-2 is found principally in heart and muscle, suggesting that JP-3 arises from RPE itself while JP-2 may be derived from the vascular choroid component of the cDNA library.

Conclusions

The collection of over 9000 human RPE/choroid ESTs, potentially representing over 6000 nonredundant genes, gives a view of a significant fraction of the transcriptional repertoire of these tissues. While this initial analysis can only begin to scratch the surface of the resource that this cDNA library provides, some broad themes have emerged. The strategy for library construction, using RNA from dissected tissues with no amplification or normalization, was intended to give as close as possible a natural representation of the genes expressed and, as judged by the number of marker genes for RPE obtained, this seems to have been successful. The distributions of abundant (e.g. Optc, GSH-PO) and novel (e.g. Ocsp) transcripts also reveal interesting parallels between this library and that for the un-normalized human iris [30], which may reflect common developmental origins.

In addition to the cataloging of known genes and the discovery of novel transcripts, the sequence data reveal alternative splice forms. Of RPE expressed genes, the most notable example of this is found in the collection of cDNAs for Bestrophin. One splice variant in particular is predicted to give rise to a protein with a significant additional domain at the C-terminus. Another example is the minor splice variant of Optc that skips the conserved signature sequence but retains all the rest of the protein. Both of these are examples of variants that make structural sense from a protein point of view and could have important differences in function, cellular localization or stability. These and other variants will require further work to determine their significance.

Further work is also needed to analyze the potentially novel genes revealed by sequence tags. Several of these are currently under further investigation. Ocsp is a prominent example of a novel gene transcript that was discovered in the RPE/choroid and iris libraries. It illustrates some of the features that may be common to other genes that are "missing" from current builds of the human genome. It is highly G/C rich, making it a challenge both for reverse transcription and for growth in bacterial host cells. According to RH mapping, the gene for Ocsp also seems to be close to the telomeric region of chromosome 17q, and perhaps this also contributes to difficulties in cloning. Ocsp peptide is detectable in RPE and iris, consistent with the detection of ESTs in eye tissues. Immunoreactivity for Ocsp is also present in the inner plexiform layer (and more weakly in other regions) of the retina. No retinal ESTs for this gene have yet been identified, but this probably reflects a low abundance at the level of mRNA combined with the problems in cloning this "difficult" sequence. Again, since the NEIBank RPE/choroid and iris libraries are unamplified, they may have an improved representation of such sequences compared with other eye derived cDNA libraries.

Many other potentially novel genes are also represented by ESTs in the RPE/choroid collection. Several otherwise unidentified ESTs have significant similarities to known protein families at the level of predicted amino acid sequence or show exon/intron structure when compared with human genome sequence. All of these require individual investigation.

The clustered sequences from this study are available through NEIBank. Analysis of the data will continue, with corrections, additions and the results of comparisons with updated versions of GenBank and human genome databases. Thus the identification of specific clusters may change as new information becomes available. Efforts are also underway to provide systematic annotations for the clustered sequences. As this work shows, EST analyses, like other kinds of genome sequence projects, are valuable tools for gene discovery and for cataloguing genes expressed in particular tissues and cell types. While EST data alone cannot give definitive evidence of expression levels and the functional consequences of expression, they provide an essential resource for future studies in functional and structural genomics. Indeed, the large nonredundant set of clones from this and other NEIBank libraries is currently being used to build cDNA microarrays to allow simultaneous comparison of the expression profiles of several thousand gene transcripts in the eye.

Acknowledgements

SLB is supported by the V. Kann Rasmussen Foundation (Denmark) and is a Career Development Awardee of Research to Prevent Blindness (RPB). We thank Dr. Weinu Gan for cDNA sequencing and Ray Tabios for technical assistance.

References

1. Gu SM, Thompson DA, Srikumari CR, Lorenz B, Finckh U, Nicoletti A, Murthy KR, Rathmann M, Kumaramanickavel G, Denton MJ, Gal A. Mutations in RPE65 cause autosomal recessive childhood-onset severe retinal dystrophy. Nat Genet 1997; 17:194-7.

2. Marlhens F, Bareil C, Griffoin JM, Zrenner E, Amalric P, Eliaou C, Liu SY, Harris E, Redmond TM, Arnaud B, Claustres M, Hamel CP. Mutations in RPE65 cause Leber's congenital amaurosis. Nat Genet 1997; 17:139-41.

3. Morimura H, Fishman GA, Grover SA, Fulton AB, Berson EL, Dryja TP. Mutations in the RPE65 gene in patients with autosomal recessive retinitis pigmentosa or leber congenital amaurosis. Proc Natl Acad Sci U S A 1998; 95:3088-93.

4. Petrukhin K, Koisti MJ, Bakall B, Li W, Xie G, Marknell T, Sandgren O, Forsman K, Holmgren G, Andreasson S, Vujic M, Bergen AA, McGarty-Dugan V, Figueroa D, Austin CP, Metzker ML, Caskey CT, Wadelius C. Identification of the gene responsible for Best macular dystrophy. Nat Genet 1998; 19:241-7.

5. Morimura H, Saindelle-Ribeaudeau F, Berson EL, Dryja TP. Mutations in RGR, encoding a light-sensitive opsin homologue, in patients with retinitis pigmentosa. Nat Genet 1999; 23:393-4.

6. Cai J, Nelson KC, Wu M, Sternberg P Jr, Jones DP. Oxidative damage and protection of the RPE. Prog Retin Eye Res 2000; 19:205-21.

7. Lutty G, Grunwald J, Majji AB, Uyama M, Yoneya S. Changes in choriocapillaris and retinal pigment epithelium in age-related macular degeneration. Mol Vis 1999; 5:35 <http://www.molvis.org/molvis/v5/a35/>.

8. Zarbin MA. Age-related macular degeneration: review of pathogenesis. Eur J Ophthalmol 1998; 8:199-206.

9. Wistow G. A project for ocular bioinformatics: NEIBank. Mol Vis 2002; 8:161-3 <http://www.molvis.org/molvis/v8/a22/>.

10. Simms D. mRNA isolation for high quality cDNA. Focus 1995; 17:39-42.

11. Wistow G, Bernstein SL, Wyatt MK, Behal A, Touchman JW, Bouffard G, Smith D, Peterson K. Expressed sequence tag analysis of adult human lens for the NEIBank Project: Over 2000 non-redundant transcripts, novel genes and splice variants. Mol Vis 2002; 8:171-84 <http://www.molvis.org/molvis/v8/a24/>.

12. Wistow G, Bernstein SL, Touchman JW, Bouffard G, Wyatt MK, Peterson K, Gao J, Buchoff P, Smith D. Grouping and identification of sequence tags (GRIST): Bioinformatics tools for the NEIBank database. Mol Vis 2002; 8:164-70 <http://www.molvis.org/molvis/v8/a23/>.

13. Wistow G, Kim H. Lens protein expression in mammals: taxon-specificity and the recruitment of crystallins. J Mol Evol 1991; 32:262-9.

14. Wistow G. Lens crystallins: a model system for gene recruitment. In: Zimmer EA, White TJ, Cann RL, Wilson AC, editors. Molecular evolution: producing the biochemical data. Methods in enzymology, vol 224. San Diego: Academic Press; 1993. p. 563-75.

15. Laemmli UK. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 1970; 227:680-5.

16. Sinha D, Esumi N, Jaworski C, Kozak CA, Pierce E, Wistow G. Cloning and mapping the mouse Crygs gene and non-lens expression of [gamma]S-crystallin. Mol Vis 1998; 4:8 <http://www.molvis.org/molvis/v4/a8/>.

17. Frohman MA. RACE: rapid amplification of cDNA ends. In: Innis MA, Gelfand DH, Sninsky JJ, White TJ, editors. PCR protocols: a guide to methods and applications. San Diego (CA): Academic Press; 1990. p. 28-38.

18. Bernstein SL, Wong P. Regional expression of disease-related genes in human and monkey retina. Mol Vis 1998; 4:24 <http://www.molvis.org/molvis/v4/a24/>.

19. Ewing B, Green P. Base-calling of automated sequencer traces using phred. II. Error probabilities. Genome Res 1998; 8:186-94.

20. Tombran-Tink J, Shivaram SM, Chader GJ, Johnson LV, Bok D. Expression, secretion, and age-related downregulation of pigment epithelium-derived factor, a serpin with neurotrophic activity. J Neurosci 1995; 15:4992-5003.

21. Maw MA, Corbeil D, Koch J, Hellwig A, Wilson-Wheeler JC, Bridges RJ, Kumaramanickavel G, John S, Nancarrow D, Roper K, Weigmann A, Huttner WB, Denton MJ. A frameshift mutation in prominin (mouse)-like 1 causes human retinal degeneration. Hum Mol Genet 2000; 9:27-34.

22. Baas D, Bumsted KM, Martinez JA, Vaccarino FM, Wikler KC, Barnstable CJ. The subcellular localization of Otx2 is cell-type specific and developmentally regulated in the mouse retina. Brain Res Mol Brain Res 2000; 78:26-37.

23. Shimura M, Yuan Y, Chang JT, Zhang S, Campochiaro PA, Zack DJ, Hughes BA. Expression and permeation properties of the K(+) channel Kir7.1 in the retinal pigment epithelium. J Physiol 2001; 531:329-46.

24. Paraoan L, Grierson I, Maden BE. Analysis of expressed sequence tags of retinal pigment epithelium: cystatin C is an abundant transcript. Int J Biochem Cell Biol 2000; 32:417-26.

25. Hamel CP, Tsilou E, Pfeffer BA, Hooks JJ, Detrick B, Redmond TM. Molecular cloning and expression of RPE65, a novel retinal pigment epithelium-specific microsomal protein that is post-transcriptionally regulated in vitro. J Biol Chem 1993; 268:15751-7.

26. Pandey S, Blanks JC, Spee C, Jiang M, Fong HK. Cytoplasmic retinal localization of an evolutionary homolog of the visual pigments. Exp Eye Res 1994; 58:605-13.

27. Wistow G, Bernstein SL, Wyatt MK, Ray S, Behal A, Touchman JW, Bouffard G, Smith D, Peterson K. Expressed sequence tag analysis of human retina for the NEIBank Project: Retbindin, an abundant, novel retinal cDNA and alternative splicing of other retina-preferred gene transcripts. Mol Vis 2002; 8:196-204 <http://www.molvis.org/molvis/v8/a26/>.

28. Frank RN, Amin RH, Puklin JE. Antioxidant enzymes in the macular retinal pigment epithelium of eyes with neovascular age-related macular degeneration. Am J Ophthalmol 1999; 127:694-709.

29. Cheng Q, Gonzalez P, Zigler JS Jr. High level of ferritin light chain mRNA in lens. Biochem Biophys Res Commun 2000; 270:349-55.

30. Wistow G, Bernstein SL, Ray S, Wyatt MK, Behal A, Touchman JW, Bouffard G, Smith D, Peterson K. NEIBANK: expressed sequence tag analysis of human iris. Steroid-response factors and similarities with retinal pigment epithelium. Mol Vis. In press 2001.

31. Thumann G. Development and cellular functions of the iris pigment epithelium. Surv Ophthalmol 2001; 45:345-54.

32. Weber BH, Vogt G, Pruett RC, Stohr H, Felbor U. Mutations in the tissue inhibitor of metalloproteinases-3 (TIMP3) in patients with Sorsby's fundus dystrophy. Nat Genet 1994; 8:352-6.

33. Stone EM, Fingert JH, Alward WL, Nguyen TD, Polansky JR, Sunden SL, Nishimura D, Clark AF, Nystuen A, Nichols BE, Mackey DA, Ritch R, Kalenak JW, Craven ER, Sheffield VC. Identification of a gene that causes primary open angle glaucoma. Science 1997; 275:668-70.

34. Johnson DH. Myocilin and glaucoma: a TIGR by the tail? Arch Ophthalmol 2000; 118:974-8.

35. Adema GJ, de Boer AJ, Vogel AM, Loenen WA, Figdor CG. Molecular characterization of the melanocyte lineage-specific antigen gp100. J Biol Chem 1994; 269:20126-33.

36. Wagner SN, Wagner C, Schultewolter T, Goos M. Analysis of Pmel17/gp100 expression in primary human tissue specimens: implications for melanoma immuno- and gene-therapy. Cancer Immunol Immunother 1997; 44:239-47.

37. Kim RY, Wistow GJ. The cDNA RPE1 and monoclonal antibody HMB-50 define gene products preferentially expressed in retinal pigment epithelium. Exp Eye Res 1992; 55:657-62.

38. Vardimon L, Ben-Dror I, Havazelet N, Fox LE. Molecular control of glutamine synthetase expression in the developing retina tissue. Dev Dyn 1993; 196:276-82.

39. Frohlich E, Klessen C. Glutamine synthetase and marker enzymes of the blood-retina barrier in fetal bovine retinal pigment epithelial cells. Graefes Arch Clin Exp Ophthalmol 2000; 238:500-7.

40. Hobby P, Wyatt MK, Gan W, Bernstein S, Tomarev S, Slingsby C, Wistow G. Cloning, modeling, and chromosomal localization for a small leucine-rich repeat proteoglycan (SLRP) family member expressed in human eye. Mol Vis 2000; 6:72-8 <http://www.molvis.org/molvis/v6/a10/>.

41. Reardon AJ, Le Goff M, Briggs MD, McLeod D, Sheehan JK, Thornton DJ, Bishop PN. Identification in vitreous and molecular cloning of opticin, a novel member of the family of leucine-rich repeat proteins of the extracellular matrix. J Biol Chem 2000; 275:2123-9.

42. Friedman JS, Ducharme R, Raymond V, Walter MA. Isolation of a novel iris-specific and leucine-rich repeat protein (oculoglycan) using differential selection. Invest Ophthalmol Vis Sci 2000; 41:2059-66.

43. Klein ML, Schultz DW, Edwards A, Matise TC, Rust K, Berselli CB, Trzupek K, Weleber RG, Ott J, Wirtz MK, Acott TS. Age-related macular degeneration. Clinical features in a large family and linkage to chromosome 1q. Arch Ophthalmol 1998; 116:1082-8.

44. Iozzo RV. The biology of the small leucine-rich proteoglycans. Functional network of interactive proteins. J Biol Chem 1999; 274:18843-6.

45. Maecker HT, Todd SC, Levy S. The tetraspanin superfamily: molecular facilitators. FASEB J 1997; 11:428-42.

46. Bascom RA, Schappert K, McInnes RR. Cloning of the human and murine ROM1 genes: genomic organization and sequence conservation. Hum Mol Genet 1993; 2:385-91.

47. Bardien-Kruger S, Greenberg J, Tubb B, Bryan J, Queimado L, Lovett M, Ramesar RS. Refinement of the RP17 locus for autosomal dominant retinitis pigmentosa, construction of a YAC contig and investigation of the candidate gene retinal fascin. Eur J Hum Genet 1999; 7:332-8.

48. Guru SC, Agarwal SK, Manickam P, Olufemi SE, Crabtree JS, Weisemann JM, Kester MB, Kim YS, Wang Y, Emmert-Buck MR, Liotta LA, Spiegel AM, Boguski MS, Roe BA, Collins FS, Marx SJ, Burns L, Chandrasekharappa SC. A transcript map for the 2.8-Mb region containing the multiple endocrine neoplasia type 1 locus. Genome Res 1997; 7:725-35.

49. Katsanis N, Lewis RA, Stockton DW, Mai PM, Baird L, Beales PL, Leppert M, Lupski JR. Delineation of the critical interval of Bardet-Biedl syndrome 1 (BBS1) to a small region of 11q13, through linkage and haplotype analysis of 91 pedigrees. Am J Hum Genet 1999; 65:1672-9.

50. Velculescu VE, Vogelstein B, Kinzler KW. Analysing uncharted transcriptomes with SAGE. Trends Genet 2000; 16:423-5.

51. Marmorstein AD, Marmorstein LY, Rayborn M, Wang X, Hollyfield JG, Petrukhin K. Bestrophin, the product of the Best vitelliform macular dystrophy gene (VMD2), localizes to the basolateral plasma membrane of the retinal pigment epithelium. Proc Natl Acad Sci U S A 2000; 97:12758-63.

52. Sonnhammer EL, Durbin R. Analysis of protein domain families in Caenorhabditis elegans. Genomics 1997; 46:200-16.

53. Bakall B, Marknell T, Ingvast S, Koisti MJ, Sandgren O, Li W, Bergen AA, Andreasson S, Rosenberg T, Petrukhin K, Wadelius C. The mutation spectrum of the bestrophin protein--functional implications. Hum Genet 1999; 104:383-9.

54. Swanson DA, Chang JT, Campochiaro PA, Zack DJ, Valle D. Mammalian orthologs of C. elegans unc-119 highly expressed in photoreceptors. Invest Ophthalmol Vis Sci 1998; 39:2085-94.

55. Gehring WJ, Ikeo K. Pax 6: mastering eye morphogenesis and eye evolution. Trends Genet 1999; 15:371-7.

56. Wawersik S, Purcell P, Maas RL. Pax6 and the genetic control of early eye development. Results Probl Cell Differ 2000; 31:15-36.

57. Plaza S, Dozier C, Turque N, Saule S. Quail Pax-6 (Pax-QNR) mRNAs are expressed from two promoters used differentially during retina development and neuronal differentiation. Mol Cell Biol 1995; 15:3344-53.

58. Xu PX, Zhang X, Heaney S, Yoon A, Michelson AM, Maas RL. Regulation of Pax6 expression is conserved between mice and flies. Development 1999; 126:383-95.

59. van der Velden AW, Thomas AA. The role of the 5' untranslated region of an mRNA in translation regulation during development. Int J Biochem Cell Biol 1999; 31:87-106.

60. de Moor CH, Richter JD. Translational control in vertebrate development. Int Rev Cytol 2001; 203:567-608.

61. Toy J, Yang JM, Leppert GS, Sundin OH. The optx2 homeobox gene is expressed in early precursors of the eye and activates retina-specific genes. Proc Natl Acad Sci U S A 1998; 95:10643-8.

62. Rehemtulla A, Warwar R, Kumar R, Ji X, Zack DJ, Swaroop A. The basic motif-leucine zipper transcription factor Nrl can positively regulate rhodopsin gene expression. Proc Natl Acad Sci U S A 1996; 93:191-5.

63. Ogino H, Yasuda K. Induction of lens differentiation by activation of a bZIP transcription factor, L-Maf. Science 1998; 280:115-8.

64. Sharon-Friling R, Richardson J, Sperbeck S, Lee D, Rauchman M, Maas R, Swaroop A, Wistow G. Lens-specific gene recruitment of zeta-crystallin through Pax6, Nrl/Maf and brain suppressor sites. Mol Cell Biol 1998; 18:2067-76.

65. Kim JI, Li T, Ho IC, Grusby MJ, Glimcher LH. Requirement for the c-Maf transcription factor in crystallin gene regulation and lens development. Proc Natl Acad Sci U S A 1999; 96:3781-5.

66. Kawauchi S, Takahashi S, Nakajima O, Ogino H, Morita M, Nishizawa M, Yasuda K, Yamamoto M. Regulation of lens fiber cell differentiation by transcription factor c-Maf. J Biol Chem 1999; 274:19254-60.

67. Ring BZ, Cordes SP, Overbeek PA, Barsh GS. Regulation of mouse lens fiber cell development and differentiation by the Maf gene. Development 2000; 127:307-17.

68. Kamachi Y, Sockanathan S, Liu Q, Breitman M, Lovell-Badge R, Kondoh H. Involvement of SOX proteins in lens-specific activation of crystallin genes. EMBO J 1995; 14:3510-9.

69. Jay P, Sahly I, Goze C, Taviaux S, Poulat F, Couly G, Abitbol M, Berta P. SOX22 is a new member of the SOX gene family, mainly expressed in human nervous tissue. Hum Mol Genet 1997; 6:1069-77.

70. Aitola M, Carlsson P, Mahlapuu M, Enerback S, Pelto-Huikko M. Forkhead transcription factor FoxF2 is expressed in mesodermal tissues involved in epithelio-mesenchymal interactions. Dev Dyn 2000; 218:136-49.

71. Vithana EN, Abu-Safieh L, Allen MJ, Carey A, Papaioannou M, Chakarova C, Al-Maghtheh M, Ebenezer ND, Willis C, Moore AT, Bird AC, Hunt DM, Bhattacharya SS. A human homolog of yeast pre-mRNA splicing gene, PRP31, underlies autosomal dominant retinitis pigmentosa on chromosome 19q13.4 (RP11). Mol Cell 2001; 8:375-81.

72. Schneider MR, Lahm H, Wu M, Hoeflich A, Wolf E. Transgenic mouse models for studying the functions of insulin-like growth factor-binding proteins. FASEB J 2000; 14:629-40.

73. Tonner E, Allan G, Shkreta L, Webster J, Whitelaw CB, Flint DJ. Insulin-like growth factor binding protein-5 (IGFBP-5) potentially regulates programmed cell death and plasminogen activation in the mammary gland. Adv Exp Med Biol 2000; 480:45-53.

74. Jena N, Martin-Seisdedos C, McCue P, Croce CM. BMP7 null mutation in mice: developmental defects in skeleton, kidney, and eye. Exp Cell Res 1997; 230:28-37.

75. Hiscott P, Sheridan C, Magee RM, Grierson I. Matrix and the retinal pigment epithelium in proliferative retinal disease. Prog Retin Eye Res 1999; 18:167-90.

76. Brekken RA, Sage EH. SPARC, a matricellular protein: at the crossroads of cell-matrix. Matrix Biol 2000; 19:569-80.

77. Kim SY, Ondhia N, Vidgen D, Malaval L, Ringuette M, Kalnins VI. Spatiotemporal distribution of SPARC/osteonectin in developing and mature chicken retina. Exp Eye Res 1997; 65:681-9.

78. Rodriguez IR, Moreira EF, Bok D, Kantorow M. Osteonectin/SPARC secreted by RPE and localized to the outer plexiform layer of the monkey retina. Invest Ophthalmol Vis Sci 2000; 41:2438-44.

79. Rowe PS, de Zoysa PA, Dong R, Wang HR, White KE, Econs MJ, Oudet CL. MEPE, a new gene expressed in bone marrow and tumors causing osteomalacia. Genomics 2000; 67:54-68.

80. Henry SP, Takanosu M, Boyd TC, Mayne PM, Eberspaecher H, Zhou W, de Crombrugghe B, Hook M, Mayne R. Expression pattern and gene characterization of asporin, a newly discovered member of the leucine-rich repeat protein family. J Biol Chem 2001; 276:12212-21.

81. Lorenzo P, Aspberg A, Onnerfjord P, Bayliss MT, Neame PJ, Heinegard D. Identification and characterization of asporin, a novel member of the leucine-rich repeat protein family closely related to decorin and biglycan. J Biol Chem 2001; 276:12201-11.

82. Kuehn MH, Hageman GS. Expression and characterization of the IPM 150 gene (IMPG1) product, a novel human photoreceptor cell-associated chondroitin-sulfate proteoglycan. Matrix Biol 1999; 18:509-18.

83. Kuehn MH, Hageman GS. Molecular characterization and genomic mapping of human IPM 200, a second member of a novel family of proteoglycans. Mol Cell Biol Res Commun 1999; 2:103-10.

84. Adams JC, Tucker RP. The thrombospondin type 1 repeat (TSR) superfamily: diverse proteins with related roles in neuronal development. Dev Dyn 2000; 218:280-99.

85. Primakoff P, Myles DG. The ADAM gene family: surface proteins with adhesion and protease activity. Trends Genet 2000; 16:83-7.

86. Nagase H, Woessner JF Jr. Matrix metalloproteinases. J Biol Chem 1999; 274:21491-4.

87. Li SW, Sieron AL, Fertala A, Hojima Y, Arnold WV, Prockop DJ. The C-proteinase that processes procollagens to fibrillar collagens is identical to the protein previously identified as bone morphogenic protein-1. Proc Natl Acad Sci U S A 1996; 93:5127-30.

88. Tsukita S, Furuse M, Itoh M. Multifunctional strands in tight junctions. Nat Rev Mol Cell Biol 2001; 2:285-93.

89. Rizzolo LJ. Polarity and the development of the outer blood-retinal barrier. Histol Histopathol 1997; 12:1057-67.

90. Konari K, Sawada N, Zhong Y, Isomura H, Nakagawa T, Mori M. Development of the blood-retinal barrier in vitro: formation of tight junctions as revealed by occludin and ZO-1 correlates with the barrier function of chick retinal pigment epithelial cells. Exp Eye Res 1995; 61:99-108.

91. Starzyk RM, Rosenow C, Frye J, Leismann M, Rodzinski E, Putney S, Tuomanen EI. Cerebral cell adhesion molecule: a novel leukocyte adhesion determinant on blood-brain barrier capillary endothelium. J Infect Dis 2000; 181:181-7.

92. Angst BD, Marcozzi C, Magee AI. The cadherin superfamily: diversity in form and function. J Cell Sci 2001; 114:629-41.

93. Bork JM, Peters LM, Riazuddin S, Bernstein SL, Ahmed ZM, Ness SL, Polomeno R, Ramesh A, Schloss M, Srisailpathy CR, Wayne S, Bellman S, Desmukh D, Ahmed Z, Khan SN, Kaloustian VM, Li XC, Lalwani A, Riazuddin S, Bitner-Glindzicz M, Nance WE, Liu XZ, Wistow G, Smith RJ, Griffith AJ, Wilcox ER, Friedman TB, Morell RJ. Usher syndrome 1D and nonsyndromic autosomal recessive deafness DFNB12 are caused by allelic mutations of the novel cadherin-like gene CDH23. Am J Hum Genet 2001; 68:26-37.

94. Bolz H, von Brederlow B, Ramirez A, Bryda EC, Kutsche K, Nothwang HG, Seeliger M, del C-Salcedo Cabrera M, Vila MC, Molina OP, Gal A, Kubisch C. Mutation of CDH23, encoding a new member of the cadherin gene family, causes Usher syndrome type 1D. Nat Genet 2001; 27:108-12.

95. Okazaki M, Takeshita S, Kawai S, Kikuno R, Tsujimura A, Kudo A, Amann E. Molecular cloning and characterization of OB-cadherin, a new member of cadherin family expressed in osteoblasts. J Biol Chem 1994; 269:12092-8.

96. Wu Q, Maniatis T. A striking organization of a large family of human neural cadherin-like cell adhesion genes. Cell 1999; 97:779-90.

97. Malfait M, Gomez P, van Veen TA, Parys JB, De Smedt H, Vereecke J, Himpens B. Effects of hyperglycemia and protein kinase c on connexin43 expression in cultured rat retinal pigment epithelial cells. J Membr Biol 2001; 181:31-40.

98. Takeshima H, Komazaki S, Nishi M, Iino M, Kangawa K. Junctophilins: a novel family of junctional membrane complex proteins. Mol Cell 2000; 6:11-22.

Typographical corrections