Molecular Cloning and Tissue Distribution of Keratocan BOVINE CORNEAL KERATAN SULFATE PROTEOGLYCAN 37A*

Previous studies showed that the keratan sulfate-con- taining proteoglycans of bovine corneal stroma contain three unique core proteins designated 37A, 37B, and 25 (Funderburgh, J. L., Funderburgh, M. L., Mann, M. M., and Conrad, G. W. (1991) J. Biol. Chem. 266, 14226– 14231). Degenerate oligonucleotides designed from amino acid sequences of the 37A protein were used to screen a cDNA expression library from cultured bovine keratocytes. A cDNA clone coding for keratocan, a 37A protein, was isolated and sequenced. The deduced keratocan amino acid sequence is unique but related to two other keratan sulfate-containing proteins, lumican (the 37B core protein) and fibromodulin. These three pro- teins share approximately 35% amino acid identity and a number of conserved structural features. Northern hy- bridization and immunoblotting of tissue extracts found keratocan distribution to be more limited than that of lumican or fibromodulin. Keratocan is abundant in cornea and sclera and detected in much lesser amounts in skin, ligament, cartilage, artery, and striated muscles. Only in cornea was keratocan found to contain large, sulfated keratan sulfate chains. Keratocan, like lumican, is a core protein of a major corneal proteoglycan but is present in non-corneal tissues primarily as a non-sulfated glycoprotein. Proteoglycans of the corneal stroma transparency of Oligonucleotides were designed by reverse translation of amino acid sequences of tryptic peptides from the 37A protein (6) inserting inosine to reduce the degeneracy of the probes. The oligonucleotide 5 (cid:57) -TGG- TAYYTITAYYTIGARAAYAAYYT-3 (cid:57) (degeneracy 256) was produced from the sequence of the peptide designated IWYL, and the oligonu- cleotide 5 (cid:57) -YTIGAYYTICARCAYAAYAAYAA-3 (cid:57) (degeneracy 128) was produced from the peptide designated FSNL. Initial screening was done at 20–30 (cid:51) 10 3 plaques/150-mm plate. Phage DNA from plaques was transferred to Magna Lift nylon membranes (Micron Separations); de- natured in 0.5 m M NaOH, 1.5 M NaCl; neutralized in 0.5 M Tris-HCl, pH 7.6, 1.5 M NaCl; and fixed with 120 mJ/cm 2 UV light. The membranes were prehybridized overnight at 42 °C in prehybridization solution containing 6 (cid:51) SSC (SSC is 0.15 M NaCl, 0.015 M sodium citrate, pH 7.0), 0.010 M sodium phosphate, pH 6.8, 5 (cid:51) Denhardt’s solution (9), 100 (cid:109) g/ml denatured herring sperm DNA, 1 m M EDTA, and 0.5% SDS. The oligonucleotide probe FSNL, end-labeled with [ (cid:103) - 32 P]ATP to a specific activity of (cid:46) 2 (cid:51) 10 8 cpm/ (cid:109) g, was diluted to 1 (cid:51) 10 6 cpm/ml of prehy- bridization solution, and the membranes were hybridized at tempera-tures decreasing from 60 to 37 °C over a period of 2–2.5 days. The membranes were washed in 6 (cid:51) SSC, 0.05% SDS at room temperature for 1 h with three changes of wash solution

Proteoglycans of the corneal stroma are important in maintaining the transparency of the cornea. These complex molecules are responsible for the hydrophilic character of the tissue, providing tissue hydration that is essential for transparency (1). It is not surprising, therefore, that the corneal stroma has a markedly different proteoglycan composition than that of other fibrous connective tissues such as skin and sclera. The unique character of corneal proteoglycans was recognized almost 60 years ago with the initial description of keratan sulfate, the most abundant glycosaminoglycan in the cornea (2). Corneal keratan sulfate is a highly sulfated, linear polymer of N-acetyllactosamine, linked to asparagine residues in the KSPG 1 core proteins via a mannose-containing oligosaccharide (3,4). Because the keratan sulfate glycosaminoglycans are posttranslational modifications of the KSPG proteins, determination of the structure and tissue-specific expression of these core proteins are essential to understanding the biological roles of KSPG. Research from our laboratory has demonstrated the existence of three KSPG proteins in bovine cornea. These proteins (designated 37A, 37B, and 25) have unique primary structures and differ in glycosylation, with protein 37A containing three keratan sulfate chains and the other two proteins containing one keratan sulfate chain each (5). Complementary DNA coding for protein 37B (lumican) has been cloned, and the deduced amino acid sequence revealed homology to three other proteoglycan proteins, fibromodulin, decorin, and biglycan (6,7). Lumican is present in several tissues other than cornea in a non-sulfated form (6,7).
In this paper, we report the sequence of cDNA encoding a KSPG 37A protein (keratocan). This DNA was isolated by screening a cDNA library with degenerate oligonucleotides representing amino acid sequences of tryptic peptides from bovine corneal KSPG protein 37A. The deduced amino acid sequence, size of mRNA, and tissue distribution of keratocan differentiate it from lumican.
Total RNA was isolated from cultured keratocytes and from various bovine tissues after pulverization in liquid nitrogen and homogenization in a denaturing solution containing 4 M guanidine isothiocyanate, 0.1 M 2-mercaptoethanol, 0.5% Sarkosyl, and 25 mM sodium citrate, pH 7.0 (8). The homogenate was mixed with 0.1 ϫ volume 2 M sodium acetate, pH 4.0, and then extracted with an equal volume of phenol/ chloroform/isoamyl alcohol (25:24:1). RNA was ethanol-precipitated from the aqueous phase, and carbohydrates were extracted with 4 M LiCl. After reprecipitation with isopropyl alcohol, RNA was dissolved in 10 mM Tris, pH 7.6, 1 mM EDTA, and stored at Ϫ70°C.
Construction and Screening of a Bovine Corneal cDNA Library-A cDNA library was constructed using the Uni-ZAP XR kit (Stratagene, LaJolla, CA). Keratocyte poly(A) ϩ RNA (6 g) was reverse-transcribed using an oligo(dT) primer-linker containing an XhoI restriction site. The cDNA was made double-stranded and then ligated to EcoRI adaptors. The cDNA was digested with XhoI, and then molecules greater than 600 base pairs in length were selected by Sephacryl S-400 spin columns and agarose gel electrophoresis. The cDNA was inserted unidirectionally into the EcoRI/XhoI-digested vector, and the product was packaged into phage particles and transfected into Escherichia coli XL1-Blue MRFЈ. The library contained more than 5 ϫ 10 6 primary plaque-forming units with Ͼ95% recombinants and was amplified once on NZY agar (9) before use. * This work was supported by American Heart Association Kansas Affiliate Grant-in-aid KS-94-GS-11 (to J. L. F.), National Institutes of Health Grants EY09368 (to J. L. F.) and EY00952 (to G. W. C.), and NASA-NAGW 2368 (to G. W. C.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) U48360.
Oligonucleotides were designed by reverse translation of amino acid sequences of tryptic peptides from the 37A protein (6) inserting inosine to reduce the degeneracy of the probes. The oligonucleotide 5Ј-TGG-TAYYTITAYYTIGARAAYAAYYT-3Ј (degeneracy 256) was produced from the sequence of the peptide designated IWYL, and the oligonucleotide 5Ј-YTIGAYYTICARCAYAAYAAYAA-3Ј (degeneracy 128) was produced from the peptide designated FSNL. Initial screening was done at 20 -30 ϫ 10 3 plaques/150-mm plate. Phage DNA from plaques was transferred to Magna Lift nylon membranes (Micron Separations); denatured in 0.5 mM NaOH, 1.5 M NaCl; neutralized in 0.5 M Tris-HCl, pH 7.6, 1.5 M NaCl; and fixed with 120 mJ/cm 2 UV light. The membranes were prehybridized overnight at 42°C in prehybridization solution containing 6 ϫ SSC (SSC is 0.15 M NaCl, 0.015 M sodium citrate, pH 7.0), 0.010 M sodium phosphate, pH 6.8, 5 ϫ Denhardt's solution (9), 100 g/ml denatured herring sperm DNA, 1 mM EDTA, and 0.5% SDS. The oligonucleotide probe FSNL, end-labeled with [␥-32 P]ATP to a specific activity of Ͼ2 ϫ 10 8 cpm/g, was diluted to 1 ϫ 10 6 cpm/ml of prehybridization solution, and the membranes were hybridized at temperatures decreasing from 60 to 37°C over a period of 2-2.5 days. The membranes were washed in 6 ϫ SSC, 0.05% SDS at room temperature for 1 h with three changes of wash solution and then washed again under the same conditions except that the temperature was raised to 37°C. Positive clones were rescreened with probe IWYL using the same methods at a density of 200 -300 plaques/100-mm plate to ensure isolation of pure plaques.
The pBluescript plasmid containing the cDNA insert was released from the Uni-Zap XR bacteriophage by coinfecting E. coli XL1-Blue cells with purified phage and ExAssist helper phage. The inserts were excised from purified plasmid DNA by digestion with EcoRI and XhoI, and their sizes were checked by agarose gel electrophoresis. The sequence of the cDNA insert was determined using an Applied Biosystems sequencer and the walking primer method (9). The sequence was read at least two times in each direction.
Northern Hybridization-Total RNA (10 g) from keratocyte cultures and various bovine tissues was subjected to electrophoresis through a 1.5% agarose gel containing 2.2 M formaldehyde (9) and blotted onto nylon membrane using a Turboblotter apparatus (Schleicher & Schuell). Northern blot filters were prehybridized overnight at 42°C in a solution containing 50% formamide, 6 ϫ SSC, 50 mM NaH 2 PO 4 , pH 6.8, 5 ϫ Denhardt's solution, 0.5% SDS, 100 g/ml denatured herring or salmon sperm DNA, and 10% dextran sulfate. Filters were hybridized in the same solution at 42°C for [21][22] h. An 872-base pair cDNA insert probe was generated by polymerase chain reaction from clone 3710, ␣-32 P-labeled to a specific activity of Ͼ1 ϫ 10 9 cpm/g DNA, and used at 1 ϫ 10 6 cpm/ml of the above solution. Filters were washed three times with 1 ϫ SSC and 0.1% SDS for 20 min at 50°C each wash, then twice with 0.1 ϫ SSC and 0.1% SDS using the same conditions. They were exposed to Hyperfilm-MP (Amersham Life Science, Inc.) at Ϫ70°C for 8 -24 h. To assay a second mRNA on the same blot, membranes were stripped of labeled probe in 0.1% SDS at 100°C, and the solution was allowed to cool to room temperature. Complete stripping of labeled probe was confirmed by exposure to film for 2 days before rehybridization.
Antibody Against the 37A Protein-Five mg of a synthetic peptide (RSVRQVYEASDPEDWTMHC) corresponding to the deduced 18 Nterminal amino acids of mature 37A protein with an added C-terminal cysteine was dissolved in 6 M guanidine HCl, 0.05 M Tris-HCl, 0.005 M EDTA, pH 8.5, and reduced with 0.1 M dithiothreitol for 1 h at 37°C. The peptide was separated from the dithiothreitol by gel filtration and immediately reacted with 2 ml of iodoacetyl-derivatized agarose gel (SulfoLink Gel, Pierce) for 16 h at room temperature. The excess binding sites were quenched with 0.05 M cysteine in the same buffer. For affinity purification, 1 ml of a ␥-globulin fraction of antiserum against bovine KSPG was allowed to bind to a 2-ml column of the peptide affinity gel for 1 h at room temperature. Unbound proteins were removed by washing with 20 ml of phosphate-buffered saline, and the specific antibody was eluted with 3.5 M KSCN, 0.02 M Tris-HCl, pH 8. Protein-containing fractions were dialyzed to phosphate-buffered saline and subjected to a second round of purification on the same gel. The purified antibody was stored at Ϫ70°C in the presence of 1 mg/ml bovine serum albumin.
Immunoblotting-Tissues from slaughter-aged steers were minced and extracted in 10 volumes of 4 M guanidine HCl in the presence of protease inhibitors (10). The extracts were dialyzed to 6 M urea containing 0.1 M Tris phosphate, pH 6.8, and stored frozen at Ϫ20°C. Before electrophoresis, the extracts were treated overnight at 4°C with 0.015 unit/ml affinity-purified endo-␤-galactosidase (11). Proteins were separated by electrophoresis in a 7.5-15% gradient SDS-polyacrylamide gel and transferred to nitrocellulose as described previously (12). Total proteins on the nitrocellulose were stained with 0.1% Ponceau S in 0.1% acetic acid and photographed, and then the membrane was incubated in 10% nonfat powdered milk in phosphate-buffered saline for 30 -60 min. Further procedures were carried out in 10 mM Tris-HCl, pH 7.4, 0.01% thimerosal, 0.05% Tween 20, and 0.15 M NaCl. The membrane was incubated 14 -18 h at 4°C with 1% bovine serum albumin and diluted primary antibody. After rinsing 3 times for 10 min each, membranes were incubated for 2-4 h at room temperature with affinity-purified, peroxidase-labeled goat anti-rabbit IgG (Sigma). The membranes were rinsed as before, and peroxidase activity was detected using a luminescent peroxidase substrate (ECL Reagent, Amersham Life Science, Inc.) and exposure to x-ray film.

Identification of a cDNA Clone and Deduced Amino Acid
Sequence of Protein 37A-We used tryptic peptide sequences from the bovine KSPG core protein 37A to design degenerate oligonucleotide probes, two of which hybridized to clones in a cDNA library constructed from cultured bovine keratocytes. Thirteen clones hybridized to FSNL, a 128-fold degenerate, 20-base oligonucleotide probe; and two of these hybridized to IWYL, a 256-fold degenerate, 26-base oligonucleotide probe. The two clones (designated 3709 and 3710) that hybridized to both probes were found to be identical on initial sequencing, and clone 3710 was chosen for complete sequence analysis.
The nucleotide sequence of the clone 3710 insert is shown in Fig. 1. The IWYL and FSNL probe sequences are located starting at nucleotides 445 and 671, respectively. The longest open reading frame is shown as a 352-amino acid protein. This predicted protein contains the amino acid sequences (underlined in Fig. 1) of all three tryptic peptides (IWYL, FSNL, and NVXV) previously identified in bovine corneal 37A protein (5). The 20 most N-terminal amino acids have characteristics of a signal peptide (13), predicting a mature protein of 332 amino acids with a molecular mass of 38,047 Da. This size corresponds closely to the experimentally determined molecular mass of the 37A core protein (5).
Comparison of 37A and Lumican Sequences-Alignment of the cDNA sequences for lumican and 37A (not shown) found no significant regions of identity, indicating that 37A and lumican do not arise as a result of alternate splicing of a single RNA transcript. Alignment of the deduced amino acid sequences of 37A and lumican (Fig. 2) revealed a 37% amino acid identity between the two bovine core proteins. These identities are dispersed throughout the sequence and not in contiguous groups. The 37A protein contains structural features similar to those in lumican. These include four closely spaced cysteines (37A residues 42, 46, 48, and 58) at the N-terminal region and 11 leucine-rich repeats with consensus sequence of LXXLX-LXXNXL (where X is any amino acid and L is leucine or another hydrophobic amino acid). Two conserved cysteine residues flank the 11th leucine-rich repeat at the C-terminal region (residues 303 and 343). The 37A protein contains a tyrosine followed by an acidic amino acid (Glu) near the N terminus (residue 27), a consensus sequence for tyrosine sulfation (14) also found in lumican. Five consensus sites for N-linked glycosylation are present in the 37A sequence (residues 93, 167, 222, 260, and 298); three of these (93, 167, and 260) match the locations of those in lumican.
Relationship to Other Proteins-A BLAST search of protein sequence data bases revealed four bovine proteins closely related to the deduced sequence of 37A. Fig. 3 presents a graphic comparison of amino acid sequence similarity among these five related bovine proteins. The 37A protein sequence is most similar to that of lumican and fibromodulin, both of which bear keratan sulfate chains. The dermatan sulfate proteoglycans, decorin and biglycan, were less closely related to the 37A protein, involving only 10-15% identity.
Northern Blot Analyses of 37A and Lumican mRNA-Electroporetically separated total RNA from cultured corneal keratocytes and various bovine tissues was transferred to nylon membranes and probed with 37A cDNA. The blot was stripped and reprobed with lumican cDNA to determine the size and relative abundance of the mRNA transcripts of these two genes. Fig. 4 shows that the 37A cDNA hybridized to a 2.5-kilobase transcript (Fig. 4A), whereas the lumican cDNA hybridized to a 2.4-kilobase (Fig. 4B) transcript. Relative intensity of labeling differed between the two probes, indicating differing abundance of the mRNAs for these proteins in different tissues. The highest levels of 37A hybridization were detected with RNA from the cornea and sclera. Lower levels of this transcript were detected, in decreasing order, in cultured keratocytes, skin, ligamentum nuchae, and skeletal muscle. Longer exposure of the films also detected 37A mRNA in nasal cartilage and cardiac muscle. The lumican transcript (Fig. 4B) was abundant in cornea and cultured keratocytes and somewhat stronger in sclera. This transcript also was found, in decreasing order, in ligamentum nuchae, skin, lung, aorta, cardiac muscle, kidney, and skeletal muscle. In extended exposures, lumican mRNA was also detected in cartilage, coronary artery, and intestine. Lumican and 37A transcripts showed different relative abundance in bovine keratocyte cultures compared with those in corneal RNA. In corneal RNA, lumican and 37A transcripts hybridized with equal intensity, whereas lumican mRNA appeared to be more abundant than 37A mRNA in cultured keratocytes.
Immunoblotting-Detection of 37A proteins was carried out with antibody to KSPG that was affinity-purified using a synthetic peptide representing the predicted 20 N-terminal amino acids of the mature protein. The specificity of the affinitypurified antibody was examined (Fig. 5) by Western blotting with purified KSPG core proteins. Lanes 1 and 3 each contain a fraction with core proteins 37A and 25, whereas lanes 2 and 4 each contain 37B, lumican. The unfractionated antiserum (Fig. 5, lanes 1 and 2) contained antibodies that recognized all three KSPG proteins. The 37A affinity-purified antibody (Fig.  5, lanes 3 and 4) bound only the 37A protein, demonstrating its specificity. In Fig. 6, the anti-37A antibody was used to stain immunoblots of unpurified, endo-␤-galactosidase-treated tissue extracts from a range of bovine tissues. The 37A protein was detected as broad and occasionally heterogeneous bands in the range of 48 -52 kDa. Cornea and sclera contained the most intensely stained 37A proteins. Skin, cartilage, skeletal muscle, tendon, and aorta all showed lesser amounts of the 37A protein. The size of the fully glycosylated 37A glycoprotein was FIG. 2. Alignment of KSPG protein sequences. The deduced amino acid sequences of bovine 37A and lumican (6) are aligned inserting gaps to maximize amino acid identities. Leucine-rich repeats are underlined, and the six conserved cysteines are designated by (ϩ). The three consensus sites for N-linked glycosylation shared between the two proteins are shown by inverted triangles. Two other such sites in the 37A sequence are designated by asterisks. Sites of potential tyrosine sulfation are boxed. The sequences are shown from the known or predicted N terminus of the mature protein (residues 17 for lumican and 21 for 37A) (21). Residues are numbered from the beginning of the deduced amino acid sequence.
FIG. 3. Amino acid sequence relationships among bovine proteoglycan proteins. Bovine lumican, 37A, fibromodulin, decorin, and biglycan were aligned using a PAM 250 matrix (22), and the relatedness of the five sequences was determined using the unweighted pair group method with arithmetic mean (23). Similarity of amino acid sequences between any two proteins is inversely proportional to the line lengths connecting the two. examined in the experiment shown in Fig. 7, in which intact, partially purified 37A proteins were separated by electrophoresis, blotted onto nitrocellulose, and then treated with endo-␤galactosidase to unmask the 37A epitope. Under these conditions, the 37A from cornea exhibited a broad size heterogeneity from 60 to Ͼ200 kDa, confirming the presence of high molecular weight keratan sulfate chains on the native 37A protein in cornea (5). The 37A from skin, sclera, and aorta, on the other hand, appeared as broad bands in the 60 -70-kDa range, indicating a clear difference in the glycosylation of the 37A protein between corneal and non-corneal tissues.

DISCUSSION
Our previous studies have demonstrated that bovine corneal KSPG is a mixture of at least three unique core proteins. Our current data confirm that two of these three are structurally similar proteins that arise from unique messenger RNAs. The lack of significant regions of identity between lumican and 37A mRNAs indicates that they do not arise as a result of alternate splicing. The most reasonable interpretation of these data is that the 37A and lumican proteins are products of different genes. This interpretation is confirmed by recent studies that found lumican (15) and 37A 2 genes to map at different chromosomal loci in the mouse. We propose the name "keratocan" for the 37A gene because only in the cornea does the protein assume the form of a proteoglycan.
The keratocan gene appears to be the newest member of the family of proteoglycans containing leucine-rich repeat motifs. These proteins each contain an N-terminal hypervariable region, six highly conserved cysteines, and 11 leucine-rich repeats (6). There are now five bovine proteins in this group for which cDNA sequences have been determined. Amino acid sequence similarities suggest two subgroups within this family of proteins. Keratocan, fibromodulin, and lumican (all keratan sulfate proteoglycans) are in one group, and decorin and biglycan (dermatan sulfate proteoglycans) are in the other (Fig. 3). This grouping also appears to be reflected at the genomic level. The genes for decorin and biglycan contain 8 exons (16), whereas those for fibromodulin and lumican contain 3 exons (17,18). Another feature that differentiates these groups is the presence of consensus sites for tyrosine sulfation. Keratocan, lumican, and fibromodulin each contain at least one tyrosine followed by an acidic amino acid (Asp or Glu) in the N-terminal (hypervariable) portion of the molecule. In the case of fibromodulin, these tyrosines have been shown to be sulfated (14).  4. Detection of lumican and 37A transcripts. Northern blotting of 10 g of total RNA purified from several bovine tissues was carried out using a labeled probe for the 37A mRNA as described under "Experimental Procedures" (A). After exposure, the blot was stripped of 37A probe, assessed for completeness of stripping, and reprobed with a probe for lumican (B) as described under "Experimental Procedures." Migration of RNA standards (in kilobases) is shown on the left.
FIG. 5. Specificity of anti-37A antibody. Bovine corneal KSPG core proteins were separated into two fractions by DEAE-Sephacel chromatography as described previously (5). The fractions were then separated by SDS-polyacrylamide gel electrophoresis and subjected to immunoblotting with antibody against KSPG (lanes 1 and 2) or with antibody affinity-purified with a synthetic peptide from the deduced 37A sequence (lanes 3 and 4). Size of marker proteins (in kDa) is shown on the left. Crude tissue extracts were fractionated by selective alcohol precipitation, then analyzed by SDS-polyacrylamide gel electrophoresis, and transferred to nitrocellulose as described under "Experimental Procedures." The 37A antigens were unmasked by treatment of the membrane in 0.1 M Tris phosphate, pH 6.8, 1 mg/ml bovine serum albumin with endo-␤-galactosidase, 0.00015 unit/ml, overnight at room temperature. The antigens were detected with anti-37A antibody as in Fig. 6. Size of marker proteins (in kDa) is given on the right.
In keratocan, a single potentially sulfated tyrosine is found at amino acid residue 27 (Fig. 2). Similar tyrosine residues are absent in biglycan and decorin. The presence of potentially sulfated tyrosines in all three keratan sulfate-containing proteins and their absence in the dermatan sulfate proteoglycans suggest the possibility that sulfated tyrosine may provide a signal for posttranslational addition of keratan sulfate chains.
This family of related proteoglycans currently has five known members, but it is highly likely that more exist. A BLAST search of GenBank TM sequences similar to keratocan identified an unpublished sequence (accession number U29089) from a human cartilage cDNA library that codes for a protein with an amino acid sequence very similar to those for the three keratan sulfate-containing members of the family (not shown). In addition, amino acid sequences from tryptic fragments of the third corneal KSPG core protein (the 25-kDa protein) indicate that it may also be a member of this group (5).
Lumican and keratocan both have a characteristic unlike other known proteoglycans. In the cornea, these proteins are core proteins of corneal keratan sulfate proteoglycan and bear long, highly sulfated keratan sulfate chains. In the other tissues in which they are found, however, these proteins occur as poorly sulfated or non-sulfated glycoproteins. Lumican from the artery contains oligomeric N-acetyllactosamine, essentially a non-sulfated form of keratan sulfate (19). Our data (Fig. 7) suggest that keratocan, like lumican, contains short, non-sulfated poly(N-acetyllactosamine) chains in tissues other than cornea. The different abundance and tissue localization of these two proteins suggest that, outside of the cornea, they may have different functions. In the cornea, however, both proteins have been recruited into the unique KSPG molecules thought to be essential for corneal transparency.
The specialized nature of the KSPG form of these two proteins may be inferred from their pattern of glycosylation. Recent x-ray crystallographic studies of ribonuclease inhibitor, another leucine-rich repeat-containing protein, show that the repeats in this protein form ␤-sheets in the hydrophobic core of the molecule, whereas the intervening sequences form ␣ helices on the exterior of the molecule (20). Lumican and keratocan have three conserved consensus sites for N-glycosylation (Fig.  2). Each of these occurs between leucine-rich repeats, therefore making these sites candidates for the addition of keratan sulfate chains. Our previous work showed that corneal lumican contains one keratan sulfate chain and that keratocan contains three (5). This pattern of post-translational addition of keratan sulfate indicates a high level of selectivity in terms of both the individual proteins involved and of choice of sites to which keratan sulfate is added. Such a level of targeting suggests the presence in keratocytes of a unique glycosylation system involved in biosynthesis of corneal KSPG. Understanding the nature of this system and the signals that initiate its action are important goals in understanding the role of KSPG in the cornea and the biochemical mechanisms by which this role is maintained.