Keratocan-deficient Mice Display Alterations in Corneal Structure*

Keratocan (Kera) is a cornea-specific keratan sulfate proteoglycan (KSPG) in the adult vertebrate eye. It be-longs to the small leucine-rich proteoglycan (SLRP) gene family and is one of the major components of ex-tracellular KSPG in the vertebrate corneal stroma. The Kera gene is expressed in ocular surface tissues including cornea and eyelids during morphogenesis. Corneal KSPGs play a pivotal role in matrix assembly, which is accountable for corneal transparency. In humans, mutations of the KERA gene are associated with cornea plana (CNA2) that manifests decreases in vision acuity due to the flattened forward convex curvature of cornea. To investigate the biological role of the Kera gene and to establish an animal model for corneal plana, we generated Kera knockout mice via gene targeting. Northern and Western blotting and immunohistochem-ical analysis showed that no Kera mRNA or keratocan protein was detected in the Kera (cid:1) / (cid:1) cornea. The expression

The vertebrate cornea is a tough and transparent tissue that provides greater than 60% of refractive power to the incoming light to cast a focused image on retina. To exact its proper functions, the cornea has to maintain transparent, appropriate curvature for refraction and toughness for protection. Genetic and epigenetic factors that cause the change of corneal curvature, reduction of transparency, and weakness in structure can lead to the impairment of visual acuity.
The corneal stroma consists of uniformly small collagen fibrils with an average diameter of 25 nm that are arranged in orthogonal lamellae. Stromal keratocytes are responsible for the formation and maintenance of a unique collagenous matrix that is essential for proper corneal curvature and transparency (1,2). The mechanisms that govern the assembly of different levels of stromal architecture are not well understood; however, proteoglycan-collagen and collagen-collagen interactions have been implicated. It has been suggested that the stoichiometry and interaction of different collagen types play an important role in modulating collagen fibril diameter (1,(3)(4)(5). The proteoglycans in the stroma are members of the small leucine-rich proteoglycan (SLRP) 1 gene family and are thought to regulate collagenous matrix assembly in connective tissue because of their bifunctional character: the protein moiety that binds collagen fibrils and the highly charged hydrophilic glycosaminoglycans that regulate interfibrillar spacing (1, 2, 6 -11). In addition to interactions with collagen fibrils, corneal stromal proteoglycans also play a role in corneal hydration due to the highly negative charge of their sulfated carbohydrate moieties and the glycosaminoglycan chains.
Keratocan, lumican, and mimecan/osteoglycin are the major keratan sulfate-containing proteoglycans in vertebrate corneal stroma. Their core proteins consist of 6 -10 tandem repeats of 24 amino acids with hydrophobic residues in conserved positions. We and others have recently shown that mice lacking lumican revealed an age-dependent corneal opacity and a high proportion of abnormally thick collagen fibrils in the corneal stroma (12)(13)(14). In humans, the first SLRP gene that is directly linked to a disease is the keratocan gene (KERA). Mutations of the human KERA gene are associated with cornea plana (CNA2) in which the forward convex curvature is flattened, leading to a decrease in light refraction (15). The clinical synopsis of CNA2 includes reduced visual activity, extreme hyperopia, hazy corneal limbus, corneal parenchymal opacities, thin corneal stroma, and indistinct sclerocorneal boundary. Moreover CNA2 with extremely flat cornea is often associated with angle closure glaucoma.
We have found previously that, unlike lumican or mimecan/ osteoglycin, which is ubiquitously expressed by most interstitial connective tissues, keratocan expression is specific to the adult corneal stroma in the mouse. During embryonic development, Kera gene expression tracks ocular surface tissues mor- phogenesis including cornea and eyelids (16). To investigate the biological role of the Kera in vivo and to generate an animal model for corneal plana, we created Kera knockout mice via gene targeting.

EXPERIMENTAL PROCEDURES
Western Blotting-To characterize keratocan in wild-type mice, two corneas were isolated from both eyes of normal FVB/N postnatal day 1, 7, and 14 mice. The corneas were minced and extracted in 1 ml of 4 M guanidine-HCl containing 10 mM sodium acetate, 10 mM sodium EDTA, 5 mM p-aminobenzamidine, and 0.1 M ⑀-amino-n-caproic acid at 4°C overnight. The extracts were dialyzed exhaustively in distilled water at room temperature, and the water-insoluble fraction was dissolved in 0.1 M Tris acetate solution (pH 6.0) containing 6 M urea. The protein concentration was measured by spectrophotometer at A 280 nm . Aliquots (100 g of protein) were incubated with or without endo-␤-galactosidase (0.1 unit/ml) at 37°C overnight. An equal volume of 2ϫ SDS sample buffer was added into digested and undigested samples, boiled for 5 min, electrophoresed on an SDS-PAGE gradient (4 -15%) or regular 10% gel, and transferred to polyvinylidene difluoride membrane. Affinity-purified antibodies against mouse keratocan peptide (VRQAYEIQDPEDWDVHDDFYC) were raised in rabbit and used to detect the keratocan in corneal samples. Alkaline phosphataseconjugated goat anti-rabbit IgG (heavy and light) was used as secondary antibody, and the immunocomplex was visualized with a Western Blue TM kit (Promega, Madison, WI). To detect decorin and fibromodulin of the mouse corneas, affinity-purified antibodies against mouse decorin peptide (IIPYDPDNPLISMC (17)) or mouse fibromodulin (CDK-VGRKVFSKLRHLER and CDPYDPYPYEPSEPYPYGVEE) (a generous gift from Dr. Larry Fisher, National Institutes of Health) 2 were raised in rabbit and used as primary antibody. Horseradish peroxidaseconjugated goat anti-rabbit IgG (heavy and light) was used as secondary antibody, and the immunocomplex was visualized with ECL TM Western blotting detection substrate (Pierce).
Generation of Kera Knockout Mice by Gene Targeting-The mouse keratocan genomic 6.6-kb BamHI-EcoRI fragment, which contains 3.1-kb promoter, exon 1, intron 1, exon 2, and 1.1-kb intron 2, was first cloned into pBluescript SK vector (Stratagene, La Jolla, CA), and the resulting plasmid was named pKera-BR6.6 (16). A 0.6-kb XbaI fragment was then deleted from the 5Ј end of the pKera-BR6.6, and the plasmid was self-ligated. Next a 2.6-kb XhoI fragment containing 1.0-kb intron 1, entire exon 2, and 1.1-kb intron 2 was deleted from the 3Ј end and replaced with a positive selection marker, neomycin resistance gene (pgkpr-neopA) cassette in the antisense orientation with respect to the Kera gene. The resulting plasmid was designated as pKera5Ј-pgk-Neo. A 1.6-kb KpnI fragment, which contains 0.6-kb exon 2 and 1.0-kb intron 2, was excised from the pKera-BR6.6 and inserted into pKera5Ј-pgkNeo digested with KpnI. Finally a negative selection marker gene, diphtheria toxin A fragment (pgkpr-Dta) cassette, was placed on the 3Ј end of the targeting vector. Therefore, the Kera gene targeting vector, pKera-TV, contains a 3.1-and a 1.6-kb homologous sequence on 5Ј and 3Ј ends, respectively. The pKera-TV (80 g) was linearized and transfected into an embryonic stem cell line (Duffy) via electroporation (250 microfarads and 800 V/cm). After 12 days of selection with Geneticin ® (G418, 500 g/ml), the G418-resistant colonies were individually picked and expanded. The homologous recombinant embryonic stem clones were identified by PCR-based analysis. PCR was carried out with the Expand Long Template PCR System (Roche Diagnostics). The primers used for 5Ј PCR were as follows: kera-1, 5Ј-gatggcctagtcggccatcactgcaaagag-3Ј; kera-2, 5Ј-cacccagagtattaaacagtttggggttgc-3Ј; neo(ϩ), 5Јcgcttcctcgtgctttacggtatcgccgctc-3Ј. The primers used for 3Ј PCR were as follows: kera-3, 5Ј-gccacccagctgagatggatc-3Ј; kera-4, 5Ј-ggcttatgctgagaatgaagg-3Ј; pgkpr(Ϫ), 5Ј-cgagactagtgagacgtgct-3Ј. Amplification of the 5Ј PCR product was performed by 35 cycles of 94°C for 1 min, 62°C for 1 min, and 72°C for 7 min. Amplification of 3Ј PCR product was identical except the extension time at 72°C was 3 min. The blastocyst microinjection and chimeric mice production were performed by the gene targeting core facility at the University of Cincinnati.
Immunohistochemistry-To examine keratocan and lumican expression, 5-m paraffin sections were deparaffinized, blocked, and processed for indirect immunostaining using epitope-specific rabbit antibodies against synthetic mouse lumican (8 g/ml) (14) and mouse keratocan (40 ng/ml) peptide as primary antibodies, and peroxidase-conjugated goat anti-rabbit IgG antibodies were used as secondary antibodies (Southern Biotechnologies, Birmingham, AL) as reported previously (14). Antibody reaction was visualized with 3,3Ј-diaminobenzidine. The specimens were counterstained with hematoxylin, dehydrated through a graded ethanol series, mounted in balsam, and observed by light microscopy.
In Situ Hybridization-The nonradioactive digoxigenin-riboprobe labeling mixtures as well as detection reagents were purchased from Roche Diagnostics. The full-length mouse keratocan cDNA insert (2.1 kb) (16) and lumican cDNA (2.0 kb) (18) in pBluescript SK(ϩ/Ϫ) plasmid vector (Stratagene) were linearized at either end of the insert with restriction enzyme EcoRI or XhoI. Single-stranded antisense and sense digoxigenin-labeled riboprobes of keratocan mRNA were generated by in vitro transcription using appropriate RNA polymerases (T7 for antisense and T3 for sense keratocan probes) at 37°C for 2 h. The DNA templates were completely degraded by RQI RNase-free DNase I (Promega) at 37°C for 15 min. The in vitro synthesized riboprobes were then partially degraded into 200 -400 bases by incubating the specimens in 0.3 M NaHCO 3 , pH10.2, at 60°C for 30 min. Finally the riboprobes were quantified by spectrophotometer, and 5 g/ml was used for the hybridization. The mouse eyes and embryos were fixed with 4% paraformaldehyde at 4°C and embedded in paraffin. The in situ hybridization was performed on paraffin sections (5 m) mounted on Superfrost/Plus microscope slides (Fisher Scientific) as described previously (19). To remove nonspecifically bound probes, slides were subjected to a stringent wash with 0.5ϫ SSC at 65°C and treated with 20 g/ml RNase (Sigma) at room temperature for 1 h followed by washing with 0.2ϫ SSC at 65°C. The hybridization signals were visualized with anti-digoxigenin antibody-alkaline phosphatase conjugates using procedures recommended by Roche Diagnostics.
Transmission Electron Microscopy-Corneas from keratocan-null and age-matched wild-type controls were fixed in 4% paraformaldehyde, 2.5% glutaraldehyde, 0.1 M sodium cacodylate, pH 7.4, with 8.0 mM CaCl 2 for 2 h on ice. The corneas were dissected and postfixed with 1% osmium tetroxide for 1 h. After dehydration in a graded ethanol series followed by propylene oxide, the corneas were infiltrated and imbedded in a mixture of Embed 812, nadic methyl anhydride, dodecenylsuccinic anhydride, and DMP-30 (EM Sciences, Fort Washington, PA). Thick sections (1 m) were cut and stained with methylene blueazur blue for examination and selection of specific regions for further analysis. Thin sections were prepared using a Reichert UCT ultramicrotome and a diamond knife. Staining was with 2% aqueous uranyl acetate followed by 1% phosphotungstic acid, pH 3.2. Sections were examined and photographed at 80 kV using a Hitachi 7000 transmission electron microscope. The microscope was calibrated using a line grating.
Fibril Diameter Measurements-The cornea stroma was divided into two regions for analysis: anterior and posterior stroma. The anterior stroma was defined as the 10 m subjacent to the epithelium, and the posterior stroma was defined as the 10 m adjacent to Descemet's layer. Micrographs of appropriate regions were taken at 48,700ϫ magnification. Calibrated micrographs from each region were randomly chosen in a masked manner. The micrographs were digitized, and diameters were measured using an R&M Biometrics-Bioquant (Memphis, TN) image analysis system.
Statistical Methods-In the initial analyses of fibril diameter, the fibril diameter measurements from two mutant animals were analyzed separately, and no significant differences were observed between the two. Therefore, in the final analyses data from different animals were combined. The field-to-field variability and correlation among measurements from the same field were accounted for by fitting a linear mixed effect model. For all but 2 of 24 fields, the assumption of normal distribution was appropriate after excluding 4.5% of the outliers. We fit the mixed effect model twice, with and without the two fields in question, and obtained similar estimates and consistent p values. Therefore, the analysis based on the full data (excluding outliers) is reported. The number of fibrils measured was 348 and 352 for the anterior regions and 783 and 897 for the posterior regions. All analyses were performed using SAS, version 8.2, software (SAS Institute Inc., Cary, NC).

Accumulation of Sulfated KS-keratocan in Cornea during the 1st and 2nd
Week after Birth-To characterize keratocan protein expression in the wild-type mice, corneas were obtained from day 1, 7, and 14 mice. The keratocan core protein was detected by SDS-PAGE Western blotting analyses with antibodies against an N-terminal peptide of mouse keratocan just after the signal peptide (Fig. 1). Keratocan migrated as 70 -80-kDa bands in undigested samples (Fig. 1, U). Digestion of the corneal extract with endo-␤-galactosidase to remove the keratan sulfate chains reduced keratocan to its core protein of ϳ50 kDa (Fig. 1, D). There is very little KS-keratocan present in cornea of mice at postnatal day 1 and a marked increase in both the amount and size of FIG. 2. Generation of Kera knockout mice via gene targeting. A, the targeting vector, kera-TV, contains a 3.1-and a 1.5-kb homologous sequence on 5Ј and 3Ј ends, respectively. A 1.4-kb fragment from the XhoI site in intron 1 to a KpnI site in exon 2 was deleted and replaced with a positive selection marker, neomycin resistance gene (pgkpr-neopA) cassette in the antisense orientation with respect to the Kera gene. A negative selection marker gene, diphtheria toxin A fragment (pgkpr-Dta) cassette, was placed on the 3Ј end of the targeting vector. B, schematic diagram showing the organization and the restriction enzyme map of the mouse Kera gene. The three exons are represented as boxes and indicated by numbers. C, the predicted structure of a targeted mutant allele (KO) after homologous recombination. The primers used for identification of allele-specific recombination in PCR are also shown. D, the table shows the expected size of a 5Ј PCR and a 3Ј PCR product amplified by various primer pairs to distinguish wild-type and targeted mutant allele. E, PCR-based genotyping analysis using tail DNA. PCR was carried out with the Expand Long Template PCR system (Roche Diagnostics). The primers used for 5Ј PCR were as follows: kera-1, 5Ј-gatggcctagtcggccatcactgcaaagag-3Ј; kera-2, 5Ј-cacccagagtattaaacagtttggggttgc-3Ј; neo(ϩ), 5Ј-cgcttcctcgtgctttacggtatcgccgctc-3Ј. The primers used for 3Ј PCR were as follows: kera-3, 5Ј-gccacccagctgagatggatc-3Ј; kera-4, 5Ј-ggcttatgctgagaatgaagg-3Ј; pgkpr(Ϫ), 5Ј-cgagactagtgagacgtgct-3Ј. Amplification of the 5Ј PCR product was performed by 35 cycles of 94°C for 1 min, 62°C for 1 min, and 72°C for 7 min. Amplification of 3Ј PCR product was identical except that the extension time at 72°C was 3 min.  1-6) or lumican (lanes 7-12). It was clear that, in the Kera Ϫ/Ϫ cornea, no keratocan was detected (lanes 5 and 6), but lumican expression was not changed (lanes 7-12). U, undigested sample; D, endo-␤-galactosidasedigested sample.
KS-keratocan in corneas of mice between postnatal day 7 and day 14, a period in which the eye opens in mice.
Generation of Keratocan Gene-targeted Mutation in Mice-To examine the biological function of keratocan in vivo, we generated keratocan knockout (Kera-KO) mice via standard gene targeting. To ensure Kera-KO mice reveal keratocan-null background, we deleted the majority of exon 2, which contains the translation initiation codon, and replaced it with pgkpr-neopA cassette in the antisense orientation with respect to the Kera gene (Fig. 2, A-C). The genotypes of the Kera-KO mice were determined by PCR with specific primer sets designed for differentiating the wild-type from the mutant allele at both 5Ј and 3Ј ends (Fig. 2, D-E). Among 324 G418-resistant embryonic stem cell clones, three independently targeted heterozygous embryonic stem cell clones were identified and then injected into B6 blastocysts. Germ line transmission from each clone was obtained, and no differences were observed among different lines.
Analyses of Keratocan Knockout Mice-The Kera-KO mice are viable and fertile without any apparent abnormal phenotypes such as hazy cornea at the age of 4 weeks. In situ hybridization analyses showed that keratocan mRNA was completely undetectable in the Kera-KO cornea, whereas the lumican mRNA expression level showed no difference between the wildtype and the Kera-KO corneas (Fig. 3). Likewise the keratocan protein was completely missing as judged by an immunohistochemical staining (data not shown). As shown by the light microscopic picture (Fig. 3), the cornea-iris angle of the anterior segment was 30 -40% narrower, and the corneal stroma was 40% thinner in the Kera Ϫ/Ϫ as compared with the wild-type littermates.
Western blotting analyses of the corneal extracts from 1-month-old animals also demonstrated that no keratocan protein could be detected in the Kera-KO mice indicating Kera-KO revealed a keratocan-null background (Fig. 4). The expression of lumican, fibromodulin, and decorin in the Kera-KO cornea as determined by Western blotting analyses showed no significant alterations between the wild-type and the Kera-KO corneas (Fig. 5).
Ultrastructural Analysis-The corneal stromata from keratocan-deficient and wild-type 7.5-month-old mice were analyzed using transmission electron microscopy. Since the lumican-deficient mouse had a spatially restricted phenotype in the corneal stroma (13), the anterior and posterior stromata were analyzed separately. Qualitatively there were consistent differences between the keratocan-deficient and wild-type stromata (Fig. 6). First, the collagen fibrils were larger in keratocandeficient stromata. Second, there was less regular fibril packing in the deficient stromata. This was true throughout the stroma with no apparent differences between anterior and posterior stroma. The collagen fibrils in the keratocan-deficient stromata had regular, circular profiles characteristic of normal fibrils (Fig. 6). This was true for fibrils in both the anterior and posterior stroma. However, fibril diameters in both regions were larger in the mutant animals. Examination of the frequency distributions of collagen fibril diameters from wild-type and homozygous mutant mice indicates that mutant mice have significantly larger diameters in both the anterior and posterior regions of the cornea than the wild-type mice (Fig. 7). Both the wild-type and mutant animals had a normal distribution of fibril diameters. There were no obvious populations of fibrils that were significantly different from the wild-type controls in either anterior or posterior regions of the mutant corneas. However, the distributions in both anterior and posterior regions of the mutant stromata showed a significant shift to larger diameters compared with the wild-type controls. The anterior fibrils had a mean diameter of 26.6 nm in the wildtype versus 30.7 nm in the keratocan-deficient stromata. This difference of 4.1 nm (95% confidence interval: 1.7, 6.7) was significant (p ϭ 0.01). Likewise in the posterior stroma, fibrils had a mean diameter of 26.9 nm in the wild-type versus 32.0 nm in the mutant stromata. This difference of 5.1 nm (95% confidence interval: 3.8, 6.3) also was significant (p Ͻ 0.001). There was no statistical difference between anterior and posterior regions in the normal stromata; however, in the keratocan-deficient stromata the difference between anterior and posterior (1.2 nm 95% confidence interval: 0.3, 2.2) is significant (p ϭ 0.02), suggesting that the posterior region may be affected to a greater degree by the keratocan deficiency. DISCUSSION In mice, eyelids are closed at birth and open in the 2nd week of postnatal age. We observed that the increase of protein level of keratocan during the 2nd postnatal week of the mouse coincided with the opening of the eye at this stage (Fig. 1). However, the maximum accumulation of keratocan protein in the cornea does not correspond with the peak of keratocan mRNA (16). It is likely that the synthesis of keratocan may be regulated at translational and/or post-translational levels. This phenomenon has also been found in lumican (18) and collagen biosynthesis (20).
Neural crest cells are important to the morphogeneses of various tissues during embryonic development. In ocular tissues, neural crest cells contribute to corneal stroma, iris, and eyelid mesenchyme. We have showed previously that keratocan expression is restricted to the keratocytes after embryonic day 16.5 (E16.5), but it is also transiently expressed by different sublineages of neural crest cells in several non-ocular tissues such as snout, ear, limb, and diaphragm during E13.5-E16.5. Keratocan knockout mice are viable and fertile, indicating that keratocan is dispensable in tissues that transiently express Kera during embryonic development. The expression levels of other SLRP members including lumican, decorin, and fibromodulin were not altered in the Kera Ϫ/Ϫ cornea as compared with that of the wild-type littermates, suggesting that the phenotypic changes observed in the Kera Ϫ/Ϫ mice resulted exclusively from the loss of function of Kera. Genetic evidence has shown that mice lacking a transcription factor, Lmx1b, decrease keratocan mRNA synthesis and exhibit abnormal corneal collagen fiber morphology at E15.5 (21). However, we did not find any significant alteration in collagenous matrix of the Kera Ϫ/Ϫ cornea at this stage (data not shown). The discrepancy of our data from the Lmx1b knockout mice is not known, but it is likely that ablation of Lmx1b actually may perturb the expression of genes other than Kera that are involved in collagen fibrillogenesis.
The corneas among keratocan, lumican, and mimecan/osteoglycin knockout mice show different but overlapping phenotypes. Lumican knockout mice have the most severe phenotypes with age-related corneal opacity, disorganized collagen fibrils, and thinner corneal stroma (12)(13)(14). Mimecan/osteoglycin knockout mice have slightly larger but uniform collagen fibrils with transparent cornea (22). Keratocan knockout mice also reveal larger diameters of corneal collagen fibers but less organized packing. However, like lumican-deficient mice, keratocan-null mice have a thinner corneal stroma and a narrower cornea-iris angle. Although the keratocan expression level is comparable to lumican in the corneal stroma, keratocan does not complement lumican functions and alleviate the corneal phenotype observed in the lumican-null mice. This result suggested that keratocan and lumican might bind to different sites of collagen fibril. The reduced thickness of corneal stroma seen in the keratocan-null mice was likely due to reduced hydration caused by a decrease of keratan sulfate proteoglycan content in Kera Ϫ/Ϫ cornea. Mimecan/osteoglycin expression level is the lowest in the cornea as compared with keratocan and lumican. Although lumican, keratocan, and mimecan/osteoglycin have been known to be corneal KS-SLRPs in various species including human, bovine, and chick, it has been shown that, unlike lumican and keratocan, mimecan has no KS chain in mouse cornea. This may explain the subtle phenotypes observed in the cornea of mimecan knockout mice.
Human corneal plana (CNA2) is an autosomal recessive inherited disease that is directly linked to the mutation of KERA. So far at least three mutations have been identified. Forty-six Finnish patients are homozygous for a founder having AAC to AGC transition at codon 247 predicting an amino acid change from Asn to Ser. This amino acid substitution affects the Asn residue in the tandem leucine-rich repeat consensus motif (LXXLXLXXNXL) (15). Four affected Bangladesh patients are homozygous of sequence change from ACA to AAA at codon 215, which results in an amino acid change from Thr to Lys in the linker region between the sixth and seventh tandem leucine-rich repeat motif (23). In an American CNA2 patient of Chinese origin, a homozygous CAG to TAG transversion was found in exon 2 that causes a change of Gln to a stop codon, predicting a truncated protein of 173 amino acids (15). It is likely that all these mutations will yield a mutant keratocan gene product that may be quickly degraded in the cornea because they do not exhibit a dominant negative effect in a heterozygous mutant. To test this hypothesis, transgenic mice harboring recombinant human KERA cDNA (wild type or mutants) bred into a mouse with Kera-null background would become an ideal model to investigate the structure-function relationship of keratocan in vivo. FIG. 7. Morphometric analysis of corneal collagen fibrils. The distributions of collagen fibril diameters were analyzed in the anterior (A and C) and posterior (B and D) cornea of 7.5-month-old wild-type Kera ϩ/ϩ (A and B) and homozygous mutant Kera Ϫ/Ϫ (C and D) littermates. The mean diameter values for the Kera Ϫ/Ϫ and the wild-type controls are indicated. The mean fibril diameter and the 95% confidence interval (CI) are indicated in each histogram.