Serial Analysis of Gene Expression (SAGE) in the Rat Limbal and Central Corneal Epithelium

Abstract

purpose. To identify genes preferentially expressed in the stem-cell–rich limbal epithelium of the rat cornea.

methods. The limbal and central corneal epithelial cells of 6-week-old rats were isolated by microdissection. Serial analysis of gene expression (SAGE) libraries were constructed and analyzed, and in situ hybridization, reverse transcription–polymerase chain reaction (RT-PCR) and cDNA cloning were conducted by conventional procedures.

results. The rat limbal and central corneal epithelial SAGE libraries consisted of 41,894 and 40,691 tags, respectively. After annotation, this was reduced to 759 transcripts specific for the limbal library and 844 transcripts specific for the central corneal library; 2292 transcripts overlapped. Transcripts encoding proteins with metabolic functions comprised the major functional category in both libraries. In situ hybridization and/or RT-PCR results of 12 of the most abundant, highly enriched transcripts in the limbal epithelium were in general agreement with the SAGE data and showed that these proteins are also expressed in the conjunctival epithelium. Interesting limbal-enriched transcripts encode WDNM1-like protein (similar to WDNM1/Expi, a putative secreted proteinase and inhibitor of metastasis), mesothelin (a cancer marker), marapsin (a trypsin-like serine protease that may control cell growth and migration), K4 and K15 (both cytokeratins), and membrane-spanning four-domain subfamily A member 8B. WDNM1-like protein was cloned and confirmed as a member of the four-disulfide core family.

conclusions. The SAGE results extend the database of genes expressed in the rodent cornea and suggest an association between several genes preferentially expressed in the limbal epithelium with cellular proliferation and migration.

The transparent cornea of terrestrial vertebrates comprises a stratified anterior epithelium, a relatively thick extracellular collagenous stroma littered with keratocytes and a monolayer of posterior endothelial cells. ¹ After division, the basal epithelial cells move anteriorly and are sloughed at the surface. ^{2 3 4} The differentiation of corneal basal-to-superficial surface epithelial cells is associated with numerous temporal and spatial changes in gene expression. ^{5 6} The slow-cycling basal limbal stem cells migrate centripetally and join the basal corneal epithelial cells, which initiate the stratification process. ^{7 8 9 10 11 12 13 14 15 16}  

The limbal stem cells are clinically important. ¹⁴ They provide a potential method for reconstruction of severely damaged corneas with limbal deficiencies ^{17 18 19 20 21} and are also favorable candidates for gene therapy of the cornea. ²² Moreover, the tumorigenicity of stem cells ^{23 24} makes the limbal epithelial cells the predominant source of corneal cancers. ²⁵ Corneal wound healing also depends on limbal stem cells dividing, migrating to the basal region of the cornea, and stratifying to regenerate a normal epithelium. ²⁶ Thus, characterization of the stem cell-enriched limbal epithelium is likely to provide new opportunities for medical intervention of diseased or damaged corneas as well as basic knowledge for understanding the maintenance of healthy corneas. 

There are a few proteins that are known to be characteristically high in the stratified corneal epithelium but poorly represented or absent from the basal limbal cells in different species. ^{7 12 27 28 29 30 31 32 33 34} Identification of proteins that characterize the limbal stem cells has been more difficult. The surface ABCG2 transporter, a member of the ATP-binding cassette transporters and characteristic of stem cells in general, ³⁵ is present in conjunctival and limbal but not central corneal basal epithelial cells of humans and rabbits. ^{12 36 37} A few other proteins, including integrin α9 and K19, appear to be selectively localized in the corneal limbal stem cells. ^{34 38 39 40} α-Enolase is progressively sequestered in the limbal cells when stratification occurs, ⁴¹ but this enzyme is also present in the stratified corneal epithelium. ³⁴ p63 as well was originally thought to be restricted to the corneal limbal basal epithelial cells, ^{42 43} but subsequently was found in the basal epithelial cells of the central cornea. ^{6 44 45 46 47}  

In the present study, we have used SAGE ⁴⁸ to search for differences in gene expression between the limbal and central corneal epithelial cells of the rat. SAGE is a powerful technique used to identify the profile of expressed genes in different tissues. ⁴⁹ Although our earlier SAGE analysis of the developing cornea was performed on mice, ⁵ we chose to examine the larger rats here due to the need to separate the relatively narrow limbus from the central cornea for these analyses. The results indicate promising differences in the patterns of gene expression between the stem cell-rich limbal and central corneal epithelial cells of the adult rat. 

Materials and Methods

Sample Preparation

Central corneal and limbal epithelial cells were obtained from 16 eyes and 48 eyes of 6-week-old male Wister rats, respectively. All animal procedures were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. After animals were asphyxiated with carbon dioxide, a central region of the cornea was marked with a 2.5-mm diameter trephine (Storz, St. Louis, MO), and the corneal epithelium in this area was collected by scraping with a miniature scalpel (Roboz, Rockville, MD), quick-frozen in liquid nitrogen and stored at −80°C. Approximately 1 mm of the limbal epithelium was scraped and the cells frozen and stored as just described. Figure 1shows the scraped regions from the central cornea and limbal regions. Note that the limbal epithelium that was used in these studies is adjacent to the end of Bowman’s membrane and overlies the cornea–iris angle. 

SAGE Library Construction and Analysis

Total RNA was extracted from the frozen epithelial cells (SNAP Total RNA Isolation Kit; Invitrogen, Carlsbad, CA), according to the manufacturer’s instructions. SAGE libraries were constructed from 13 μg of total RNA from the central corneal epithelial cells and from 7 μg of RNA from the limbal epithelial cells (I-SAGE Kit; Invitrogen). SAGE tags were sequenced (model CEQ 2000; Beckman, Fullerton, CA). The sequence data were analyzed by the SAGE2000 software. ⁴⁸ Ambiguous bases, abnormally short tags (<10 bps) and abnormally long ditags (>24 bps) were omitted automatically. Tags were identified by using the reliable rat tag database from SAGEmap (Rattus norvegicus; Unigene Build 146; http://www.ncbi.nlm.nih.gov/SAGE/ and http://www.ncbi.nlm.nih.gov/UniGene/ provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD). 

In Situ Hybridization and RT-PCR

Tissues were prepared and in situ hybridizations were performed as described in Norman et al. ⁵ The riboprobes for K15 (277 bp), WDNM1-like protein (223 bp) and marapsin (360 bp) were made by in vitro transcription of the cDNA fragments cloned in the pSPT 19 vector. An EcoRI site in the forward primers and HindIII site in the reverse primers were introduced to facilitate cloning of the PCR products in the pSPT 19 vector. The inserted fragments were generated by PCR amplification using the following specific primers: K15 (F [forward]: 5′-GCCAAACCTGAATTCCACATCAGCGT-3′; R [reverse] 5′-CACTAAG-CTTGGAACAGCCACCTACA-3′); WDNM1-like protein (F: 5′-GGTCACCGAATTCTTTTAGGCACCTGT-3′; R: 5′-CCACGAAGCTTAAACACC-GCCTTGTC-3′); and marapsin (F: 5′-CTTGAGGAATTCTGGAGGCCCATTGGTA-3′; R: 5′-AGGAGCTGAAGCTTAAGATAGGGGGACA-3′). Trans-formants (One Shot TOP 10 Electrocomp Escherichia coli; Invitrogen) were identified by PCR and confirmed by nucleotide sequencing. The purified plasmids were linearized with EcoRI (for antisense probe) and with HindIII (for sense probe), and treated with proteinase K and served as templates for antisense and sense digoxigenin-labeled riboprobe synthesis using an RNA labeling kit (DIG Ki; SP6/T7; Roche Molecular Biochemicals, Indianapolis, IN). 

For RT-PCR, total RNAs were extracted from central corneal, limbal and conjunctival epithelium (SNAP Total RNA Isolation Kit; Invitrogen). Conjunctival epithelium was scraped at the region far from the limbus to avoid contamination. RT-PCR was performed using 12 ng total RNA, gene-specific primers, and Taq polymerase (SuperScript One-Step RT-PCR with Platinum Taq; Invitrogen). 

The primers used for these reactions were as follows: K15 (F: 5′-AGCAATTTCCACATCAGCGT-3′; R: 5′-TGAAGGAACAGCCACCTACA-3′); WDNM1-like protein (F: 5′-TGCAGCTTTTAGGCACCTGT-3′; R: 5′-AAAAGAAACACCGCCTTGTC-3′); chemokine (C-C motif) ligand 6: (F: 5′-TCAGAGAGTTCAGACGTGCA-3′; R: 5′-TGGGGGTCACTTACTGCTCT-3′); membrane-spanning 4-domain, subfamily A, member 8B (F: 5′-AACCACCCAACCCAATACCA-3′; R: 5′-TATCCCCAAAGCTCCTGTAG-3′); mesothelin (F: 5′-AGATACGTTAGCCCTGAGGA-3′; R: 5′-TTTGGCAATGACCCCTGGTA-3′); CD74 antigen (invariant polypeptide of major histocompatibility class II antigen associated) (F: 5′-AGCCCTTCTTACACTCCCTA-3′; R: 5′-TCACGTGAACCATGGTCCTA-3′); marapsin (F: 5′-ACTCTGGAGGCCCATTGGTA-3′; R: 5′-GGATGTAAGATAG-GGGGACA-3′); secretory (zymogen) granule membrane glycoprotein GP2 (F: 5′-GTCCTCTTGAATAGGGGTGA-3′; R: 5′-TTCTGCCCACTATCAGGACA-3′); EST (expressed sequence tag; Rn.30331) (F: 5′-AA-GTGCCTTGTACTGCCTGA-3′; R: 5′-GCAGACCATAAGACCTCCGT-3′); EST (Rn.9974) (F: 5′-TGTCTGACTCTGGCAGCACT-3′; R: 5′-CAGTA-TCTTGCTCTCACAGCT-3′); peptidyl arginine deiminase, type 1 (F: 5′-GCTGTTGGTCCAGATGTGAA-3′ R: 5′-AGAAGACTTGCTCCCATGCT-3′); and EST (Rn.96234) (F: 5′-AGTCCAAACTGAGCAGCACA-3′; R: 5′-CGGTATCTTAGCCCTCTCTA-3′). 

The amplified PCR products were subjected to electrophoresis on a 1.5% agarose gel containing ethidium bromide. 

Obtaining and Sequencing WDNM1-like Protein cDNA

A full-length cDNA for WDNM1-like protein was constructed with a cDNA synthesis kit (SMART PCR cDNA; BD-Clontech, Palo Alto, CA). Total RNA (50 ng) from the limbal epithelium were reverse transcribed with a modified oligo (dT) primer according to the manufacturer’s instructions. The amplified double-stranded DNA was used as a template to amplify by PCR the full-length cDNA using a modified oligonucleotide primer (5′-TGGTATCAACGCAGAGTACGC-3′) and a specific primer (5′-AAAAGAAACACCGCCTTGTC-3′). The cDNA was sequenced (CEQ 2000 sequencer; Beckman, Fullerton, CA). 

Results

Gene Expression in the Rat Limbal and Corneal Epithelial Cells

SAGE libraries were constructed using total RNA from the epithelial cells scraped from the limbal and central corneal epithelium, as described in the Materials and Methods section. Figure 1shows the corneal and limbal regions of the rat eye from which the epithelial cells were taken. Although the corneal cells represent a clean preparation with negligible contamination of other tissues, the limbal cells must unavoidably be contaminated with some conjunctival cells. 

The SAGE libraries have been submitted to the NCBI SAGEmap database (http://www.ncbi.nlm.nih.gov/SAGE; series accession number GSE4121). The limbal and corneal epithelial cell libraries contained 41,894 and 40,691 tags, respectively. The tags distributed much like those in other SAGE libraries, including those from total mouse corneas. ⁵ Approximately 4.5% of the 41,894 limbal tags and 4.8% of the 40,691 central corneal tags occurred only once. 

The compositions of the two libraries were compared. We considered only tags present in two or more copies and expressed the data in terms of transcripts. There were instances that a given transcript was represented by a major (∼90%) and one or more minor (∼10%) tags. There were 3051 different transcripts represented in the limbal library and 3136 different transcripts represented in the central corneal library. Of these, 2292 transcripts overlapped in both the epithelial and limbal libraries. The two libraries contained 617 distinct tags that did not match with a known gene in the database and 1080 ESTs (nonannotated ESTs) in 3895 different transcripts. The proportions of annotated to total transcripts was 38% for limbal-specific, 48% for cornel-specific and 66% for overlapping transcripts. All transcripts are presented in Supplementary Table S1. 

The DAVID annotation tool (http://apps1.niaid.nih.gov/david/ provided in the public domain by the National Institute of Allergy and Infectious Diseases, Bethesda, MD) was used to classify the functional roles of the transcripts based on the GO annotation database (www.geneontology.org). Approximately 70% of the transcripts from each library mapped to a functional group consistent with the incomplete status of rat genome annotations. The functional groups of the transcripts as determined by these computer-assisted tools are shown in Figure 2 . The functions of the transcripts and the number of tags attributed to each function were generally similar in the two libraries. The larger number of cytoskeletal transcripts in the corneal library is due to the high expression of K12 in the central corneal region. ^{7 50 51 52} The transcripts encoding metabolic proteins were the most frequent in the two libraries. 

The 50 most abundant transcripts (eliminating transcripts for ribosomal proteins) represented in the limbal and corneal epithelial SAGE libraries are shown in Table 1 . All are present in both libraries, except for the one that we have named WDNM1-like protein (Rn.41321) which has 169 SAGE tags in the limbal library and none in the central corneal library. The abundant transcripts that have orthologues or homologues that contain >10 tags in the 6-week-old mouse corneal SAGE library ⁵ are shaded in Table 1 . 

Table 2lists the most abundant transcripts found in each library that (1) are represented by >30 tags, (2) are at least five times more prevalent in their respective library than in any other reported rat SAGE library, and (3) have a limbal-corneal or corneal-limbal ratio no less than 3. Of these, WDNM1-like protein, membrane-spanning 4-domain subfamily A member 8B, and mesothelin are present exclusively in the limbus and 1 EST transcript (Rn.105851) is present only in the cornea. Although Aldh3a1 is highly expressed in differentiated cornea, ^{29 53 54 55} we found it sufficiently represented in the limbal library to be excluded from Table 2 . However, the characteristically abundant K12 ^{7 50 51 52} and transketolase ⁵⁶ corneal transcripts are represented as selectively enriched in the central cornea in Table 2 . 

Finally, Table 3lists all the transcripts that were annotated to encode transcription factors and were present in at least five copies in one of the two libraries. Most are appreciably enriched in either the limbal or corneal libraries and deserve further attention. The transcription factors that are shaded in the table come from protein families also expressed in the mouse cornea. ⁵  

In Situ Hybridization and RT-PCR

To test whether our SAGE data reliably represent the spatial distributions of the transcripts, we complemented the SAGE results with RNA in situ hybridization and RT-PCR experiments, using selected probes that were enriched in the limbal region (Table 2) . The in situ hybridization results using probes from these transcripts were generally consistent with the SAGE data. The WDNM1-like protein transcript was in the limbal conjunctival epithelia (Fig. 3A) , but not in the central corneal epithelium (Fig. 3B) . The in situ hybridization signal for the K15 transcript was also confined to the limbal and conjunctival regions (Figs. 3C 3D) . The marapsin transcript gave an in situ hybridization signal in the limbus and conjunctiva (Fig. 3E)and, as expected (Table 2) , in the central cornea (Fig. 3F) . Note that all the hybridization signals are situated in the basal cells and not in the suprabasal cells. The sense strands for marapsin (Figs. 3G 3H)and other probes (data not shown) did not hybridize. 

The results of RT-PCR tests were also in agreement with the SAGE and in situ hybridization data (Figs. 4A 4B) . RT-PCR products derived from the WDNM1-like protein and K15 transcripts were obtained from limbal and conjunctival epithelial RNAs but not from central corneal epithelial RNAs. The marapsin RT-PCR product was obtained mostly from the limbal and conjunctival epithelial RNA, but was also made in lesser amounts from RNA obtained from the central corneal epithelial cells. Additional transcripts listed in Figure 4B , enriched in the limbal region (Table 2) , were examined by RT-PCR. The results consistently showed more transcripts in the limbal than the corneal region. However, in all but four cases, some RT-PCR product was detected with RNAs from the central corneal epithelium, indicating that the absence of a specific SAGE tag does not necessarily signify the total absence of transcript for that gene. Indeed, the number of unique tags that we obtained is insufficient for saturating the total number of tags present in the library (see Norman et al. ⁵ for further discussion). Figure 4Balso indicates that these transcripts were present in the conjunctival epithelium (taken from regions well separated from the limbus) as well as the limbal epithelium. We conclude that the SAGE data approximate the relative amounts of individual transcripts in the central corneal and limbal epithelia. However, the data do not distinguish between transcripts that may be present in limbal and conjunctival epithelia. 

Isolation and Characterization of Full-Length WDNM1-like Protein cDNA

A cDNA for the WDNM1-like protein transcript was cloned and sequenced because it comprises the most abundant SAGE tag present exclusively in our limbal library. The full-length cDNA has an open reading frame of 72 amino acids (Fig. 5) . The closest match to this sequence was a mouse protein that contains 63 amino acids (NP_899072; hypothetical protein LOC66107). ⁵⁷ Amino acids 10-72 encoded in the rat WDNM1-like protein cDNA showed 85% identity and 93% similarity with the mouse hypothetical protein LOC66107. These two proteins showed similarity to WDNM1 from the rat (P14730; also called extracellular proteinase inhibitor, or Expi), the mouse (CAA63605), and the bovine (AAF76524). 

Two conserved domains have been identified in these similar proteins (see Fig. 5 ). Use of PROSITE ⁵⁸ reveals a leucine zipper pattern in amino acids 17-38 of rat WDNM1-like protein. The carboxyl terminal half of the rat limbal protein is also similar to the “four-disulfide core” family of proteins that includes whey acidic proteins and extracellular proteinase inhibitors (www.protonet.cs.huji.ac.il). ^{59 60} Computer analysis using PSORT ⁶¹ predicts that the rat WDNM1-like protein is secreted, and this has been experimentally demonstrated for mouse WDNM1. ⁶²  

Discussion

The present SAGE experiments augment the database of gene expression in the cornea of the rat, which lags well behind that of the mouse. ⁵ Although many of the highly expressed genes in the rat cornea are also highly expressed in the mouse cornea, it is interesting that some variations have occurred. For example, glutathione S-transferase α4 and ω1 are prevalent in the mouse corneal epithelial cells, whereas glutaredoxin 1 is highly expressed in the rat corneal epithelial cells. In addition, our data comparing gene expression in the central stratified corneal epithelial cells and the epithelial stem cell-rich limbus ^{7 8 10 11 12 13 14} suggest potentially interesting genes with respect to limbal stem cells and corneal biology. 

Not surprisingly, our SAGE libraries indicate that many of the same genes are expressed in the limbal and central corneal epithelial cells of 6-week-old rats. Indeed, a minimum of 75% and 73% of the transcripts overlap in the limbal and central corneal libraries, respectively. Some of the overlap may arise from the fact that the basal cells of the central corneal epithelium retain stem cell characteristics during early maturation of the cornea. ^{41 63} Mammalian corneal basal epithelial cells lose potential to proliferate over several months after birth. ^{64 65} This loss is accompanied by a redistribution of p63, a regulatory protein associated with proliferation and differentiation of epithelial cells in general. ^{66 67 68 69} p63 is expressed throughout the basal epithelial cells of the cornea after birth and becomes gradually restricted to the peripheral and limbal regions in mice ^{6 44 45} rats ⁴⁶ and humans, ⁴⁷ where it may have a role in stratification of the epithelium. High proliferative potential and stem-like characteristics of the basal epithelial cells of the central cornea to maintain homeostasis with desquamation at the anterior surface shortly after birth are consistent with the fact that the progeny of the limbal stem cells do not begin migrating into the corneal basal epithelial cells until the fifth postnatal week in mice. ¹⁵ Thus, it is likely that the differences in gene expression between limbal and central corneal epithelial cells in rats >6 weeks old exceed that reported herein. 

Gene expression differences between the stem-cell–enriched limbal epithelium and the central corneal epithelium may also be underrepresented in the present libraries due to the relatively low percentage of stem cells in the limbal epithelium. The slow-cycling stem cells comprise only a fraction of the basal epithelial cells and are absent from the numerous suprabasal cells of the limbus. ^{7 8 12 52 70 71} It has been estimated that there are approximately 100 stem cells in the limbus of young mice, and this small number declines with age. ¹⁵ In addition, as suggested in our in situ hybridization and RT-PCR tests, the dissected preparations appear to contain conjunctival cells as well as corneal limbal cells, although our SAGE data do not discriminate between coexpression of genes in conjunctival and limbal epithelial cells from gene expression confined to contaminating conjunctival cells. It would be interesting to compare a limbal SAGE library with a library made from a side population of stemlike cells created by purification on the basis of the ABCG2 transporter, a stem cell marker. ^{12 36} It is of interest in this connection, however, that we found no evidence for ABCG2 transporter gene expression in our SAGE libraries. The absence of ABCG2 transporter tags may be due to an annotation difficulty with the rat genome or to very low expression of this gene in the rat corneal cells. Comparison of a conjunctival SAGE library with the limbal library may also facilitate the identification of genes expressed specifically in the limbus. However, it has recently been shown that there are clusters of K12-containing cells with gene expression profiles similar to corneal cells in the human conjunctiva, ⁷² and this complicates identifying conjunctiva-specific gene expression by SAGE analysis. 

It is possible that dendritic cells (often immune-related) comprising macrophages, lymphocytes, antigen-presenting cells, and melanocytes are a source of SAGE tags in the present investigation. That the limbal epithelial cells may contain a higher proportion of these cells than the central corneal epithelial cells ⁷³ raises the possibility that a portion of the transcripts specific to the limbus are derived from dendritic cells. However, immune-related cells are also present, although at lower levels, in the central cornea. ^{74 75} Moreover, marker transcripts for immune-related cells (CD11 for macrophages, CD11c for dendritic cells, and CD3 and CD8 for T cells) were not found in either SAGE library. In addition, only two SAGE tags for CD4 (a T-cell marker) were found in the limbal library, and none were present in the central cornea library. Finally, the probable rat homologue (Melan-A) to MART1, a marker of human melanocytes found in the corneal limbus, ³⁹ was not present in our corneal libraries. Thus, it is unlikely that gene expression in immune-related cells makes a substantial contribution to our SAGE data. 

Transcripts that are enriched in the limbus and are promising candidates for further study include K15, WDNM1-like protein, mesothelin, and membrane-spanning 4-domain subfamily A member 8B. Membrane-spanning 4-domain subfamily A member 8B may provide a useful surface antigen for purifying limbal cells. The limbal localization of the K15 transcript is consistent with this protein being a stem cell marker. K15 is a marker for human hair follicle stem cells in the bulge region and its expression may be one of the earliest signs of the transition from stem to transit-amplifying (TA) cells. ^{76 77 78} Thus, it is possible that K15 becomes a useful stem cell or early differentiation marker for the corneal epithelial cells. 

In addition to being new candidate markers for limbal stem cells, WDNM1-like protein and mesothelin each have interesting, albeit speculative, connections to cancer. WDNM1 may be an early or premalignant marker for initiation of mouse mammary adenocarcinomas induced by Neu and Ras ⁶² as well as being a preneoplastic marker in premalignant p53-null mammary cells and mammary outgrowth lines. ⁷⁹ WDNM1 is upregulated during mammary gland differentiation, ⁸⁰ but this upregulation is diminished when lactogenic differentiation is suppressed and tumorigenesis is stimulated by overexpression of Wnt1. ⁸¹ WDNM1 is downregulated in metastatic cells. ^{82 83} Secretion of WDNM1 has been demonstrated experimentally, ⁶² suggesting that this protein may suppress extracellular proteases necessary for metastases. Therefore, although the function of the rat WDNM1-like protein is not known, it may have some connection with proliferation and/or cell migration, both characteristics of limbal stem cell biology. It is also noteworthy that corneal cancers occur in the limbus, where the WDNM1-like protein transcript is present, rather than in the central region of the cornea, where this transcript is absent. ²⁵  

Mesothelin gene expression, restricted in the present study to the corneal limbus, has also been associated with cancer. Mesothelin is a cell surface glycoprotein that may play a role in cellular adhesion and is named because of its expression in mesothelial cells. ⁸⁴ Mesothelin is diagnostic of numerous tumors and thus is considered to be a marker for various cancers. ^{85 86 87} Attempts to connect WDNM1 and mesothelin gene expression to stem cell characteristics and cancer are, of course, premature at present, but may warrant further consideration. 

Finally, the limbal-enriched marapsin is a member of the large family of serine proteases. ⁸⁸ These proteins are versatile regulators of many epithelial cell processes including cancer, cell proliferation, and signal transduction. ^{89 90 91 92} Serine proteases coupled with metalloproteinases may play an important role in corneal epithelialization and wound healing ^{93 94 95} via activation of hepatocyte growth factor (HGF). ^{96 97} Serine protease activity has also been implicated in controlling the invasive phenotype of cervical squamous cell carcinoma cells, most likely by stimulating metalloproteinase-9 production. ⁹⁸ Given these relationships, the characterization of rat marapsin may open new avenues of investigation for limbal–corneal biology. 

 Supported by the National Eye Institute intramural program.

 Submitted for publication February 28, 2006; revised April 6, 2006; accepted June 22, 2006.

 Disclosure: W. Adachi, None; H. Ulanovsky, None; Y. Li, None; B. Norman, None; J. Davis, None; J. Piatigorsk y, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Photomicrographs of the 2.5-mm region of central corneal (A, B) and ∼1-mm region of limbal epithelia (C, D) before (A, C) and after (B, D) scraping. The epithelia between the arrowheads were scraped. Note the absence of stromal contamination in the scraped regions. CC, central cornea; PC, peripheral cornea; LR, limbal region; CJ, conjunctiva. Original magnification: (A, B) ×13; (C, D) ×8.

Figure 2.

 

Functional classification of transcripts expressed in the limbal epithelium and the central corneal epithelium.

Table 1.

 

Top 50 Limbal and Corneal Epithelium Transcripts

 

Table 2.

 

Most Abundant Transcripts in the Limbal and Corneal Epithelium

	Tag	Limbus	Cornea	Unigene	Description
Limbus
1	GGAGGGGTGT	623	17	Rn.100688	Keratin Kb4 type II
2	TTCTTCCTAG	278	2	Rn.153568	Keratin KA15 type I
3	GGGAGGGACA	169	0	Rn.41321	WDNM1-like protein
4	GATGAGGATG	149	45	Rn.21104	Calmodulin 4
5	GCTCCGTGGC	67	18	Rn.15842	Arhgdib Rho, GDP dissociation inhibitor (GDI) beta
6	CAGTTTTGCT	65	18	Rn.21284	Marapsin
7	CCTGGGTCTC	60	16	Rn.73434	Annexin A9 (Annexin 31)
8	GGTCCTGCTG	43	12	Rn.137175	EST
9	CATCTGTATG	33	0	Rn.136564	Membrane-spanning 4-domain subfamily A member 8B
10	TGCCCCCTTG	32	0	Rn.18607	Mesothelin
11	GGATATGGGG	32	6	Rn.48232	Mal, T-cell differentiation protein-like
12	TGACCAGCTC	32	8	Rn.3605	Membrane-associated protein 17
Cornea
1	GCAATGAAAA	187	2282	Rn.66890	Keratin K12
2	GGATGCCGGG	134	496	Rn.5950	Transketolase
3	GCCATTTGGC	63	471	Rn.19695	LY6/PLAUR domain containing 2
4	CTTGGCTCTG	14	336	Rn.20329	ARS component B precursor
5	TTACCACTCA	83	275	Rn.145079	Stratifin
6	GCAATGAAGT	31	247	Rn.34217	EST
7	CTGCGCACAG	13	187	Rn.10066	Aquaporin 5
8	GCTTCCACAC	24	89	Rn.40392	Procollagen, type XVII, alpha 1
9	GGTTAAATGT	3	65	Rn.1294	Cathepsin L
10	TCTTAATAGT	0	59	Rn.105851	EST
11	TGCTTTAAAA	7	57	Rn.54711	Desmoplakin
12	TAGCAAGCCA	11	52	Rn.38823	EST
13	TACTTTCTGT	11	51	Rn.3841	Heat shock 27kDa protein 1
14	CTTGTCACTG	8	44	Rn.9230	Gap junction membrane channel protein beta 2
15	CCGCTGTTCC	9	41	—	Unidentified tag
16	TCAGCATTCA	6	31	Rn.18560	Secretory leukocyte peptidase inhibitor
17	TGTGAATGTT	9	31	—	Unidentified tag

 If a transcript was represented by more than one tag, the counts of all tags were added to obtain a count for the transcript. A transcript is abundant if the pooled number of tags is >30, the main tag is at least five times more than in any other rat SAGE library (January 2006, 17 other libraries; normalized to 1 million tags), and if the limbus-cornea ratio or the cornea-limbus ratio is three or more.

Table 3.

 

Transcripts with GO Transcription Factor Activity Annotation and Tag Count ≥5 in at Least One of the Libraries

 

Figure 3.

 

In situ hybridization of WDNM1-like protein RNA (A, B), K15 (C, D), and marapsin (E, F). Sense probe controls for marapsin (G, H). CJ, conjunctiva; LR, limbal region; CC, central corneal region. AS, antisense; S, sense.

Figure 4.

 

RT-PCR analysis of conjunctival, limbal, and central corneal RNAs. (A) K15, WDNM1-like protein, and marapsin RNAs in the limbal and central corneal epithelia. (B) Approximate levels of transcripts in the limbal, central corneal, and conjunctival epithelia as judged by RT-PCR and electrophoresis.

Figure 5.

 

Deduced amino acid sequence of the rat WDNM1-like protein cDNA. The following proteins were aligned: R_WDNM1-like (rat WDNM1-like protein, BAC81506); M_LOC66107 (Mouse hypothetical protein LOC66107, NP_899072); R_WDNM1 (rat extracellular peptidase inhibitor precursor (Protein WDNM1), P14730); M_WDNM1 (mouse WDNM1 protein, CAA63605); B_WDNM1 (bovine WDNM1, AAF76524). Color code: lavender, amino acids identical with the rat WDNM1-like protein amino acids; pale blue, similar amino acids (BLOSUM62 ≥ 0). The conserved cysteine residues are indicated in bold. Alignments determined with ClustalW, a multiple sequence alignment program. (European Bioinformatics Institute, European Molecular Biology Laboratory, Heidelberg, Germany; available at http://www.ebi.ac.uk/clustalw/).

Supplementary Materials

Supplementary Table S1 - (.xls) Annotated transcripts of the limbal and corneal epithelium.