Keratin 12-Deficient Mice Have Fragile Corneal Epithelia

Winston W.—Y. Kao,* Chia-YangLiu,* Richard L. Converse,* Atsushi Shiraishi* Candace W.-C. Kao,* Masamichi Ishizaki* Thomas Doetschman,^ and John Duffy-f

Purpose. Expression of the K3-K12 keratin pair characterizes the corneal epithelial differentiation. To elucidate the role of keratin 12 in the maintenance of corneal epithelium integrity, the authors bred mice deficient in keratin 12 by gene-targeting techniques. Methods. One allele of murine Krtl.12 gene was ablated in the embryonic stem cell line, E14.1, by homologous recombination with a DNA construct in which the DNA element between intron 2 and exon 8 of the keratin 12 gene was replaced by a neo-gene. The homologous recombinant embryonic stem cells were injected to mouse blastocysts, and germ lines of chimeras were obtained. The corneas of heterozygous and homozygous mice were characterized by clinical observations using stereomicroscopy, histology with light and electron microscopy, Western immunoblot analysis, immunohistochemistry, in situ hybridization, and Northern hybridization. Results. The heterozygous mice (+/—) one allele of the Krtl.12 gene appear normal and do not develop any clinical manifestations (e.g., corneal epithelial defects). Homozygous mice (—/—) develop normally and suffer mild corneal epithelial erosion. Their corneal epithelia are fragile and can be removed by gentle rubbing of the eyes or brushing with a Microsponge. The corneal epithelium of the homozygote (—/—) does not express keratin 12 as judged by immunohistochemistry, Western immunoblot analysis with epitope-specific anti-keratin 12 antibodies, Nordiern hybridization with 32P-labeled keratin 12 cDNA, and in situ hybridization with an anti-sense keratin 12 riboprobe. Light and electron microscopy revealed subtle abnormalities in the corneal epithelia of —/— mice (i.e., a decrease in number of cell layers) and cytolysis of superficial cells, but the number of hemidesmosomes and desmosomes are normal in basal and suprabasal cells. The number of keratin intermediate filaments in basal and suprabasal corneal epithelial cells in — /— mice decreases, and they appear as dense bundles. This morphology is similar to that of keratin intermediate filaments in epidermal epithelial cells but differs from that of normal corneal epithelial cells in which the keratins form fine filamentous networks. The superficial epithelial cells are devoid of keratin intermediate filaments and often detach from the corneal surface of — /— mice. Conclusions. The presence of cornea-specific K3-K12 keratin pail's is essential for the maintenance of corneal epithelium integrity. Invest Ophthalmol Vis Sci. 1996;37:2572-2584.

IVeratins are a group of water-insoluble proteins that tive charge, immunoreactivity, and sequence homolo- form 10 nm intermediate filaments in epithelial gies to types I and II wool keratins, respectively.5'6 In cells.1"'1 Approximately 30 different keratin molecules vivo, a basic keratin usually is coexpressed and have been identified.1 They can be divided into acidic "paired" with a particular acidic keratin to form a and basic-neutral subfamilies according to their rela- heterodimer.6'7 The expression of various keratin pairs is tissue specific, differentiation dependent, and devel- opmentally regulated.2'58'9 From the Departments of*Ophthalmology and ^Molecular Genetics, University of The presence of specific keratin pairs is essential Cincinnati, Ohio. Supported by National Institutes of Health grants EY10556 and HL41496 and by for the maintenance of epithelium integrity. For ex- Ohio Lions Eye Research Foundation. ample, mutations in human K14-K510"12 and K10- Submitted for publication April 16, 1996; revised July 9, 1996; accepted July 25, Kl1314 genes have been linked to the human skin dis- 1996. Proprietary interest category: N. eases, epidermolysis bullosa simplex and epidermoly- Reprint requests: Winston W.-Y. Kao, Department of Ophthalmology, University of sis hyperkeratosis, respectively. Similar clinical mani- Cincinnati, Eden and Bethesda Avenues, P.O. Box 670527, Cincinnati, OH 45267-0527. festation of these diseases have been reproduced in

Downloaded from iovs.arvojournals.org on 09/29/2021 Keratin 12-Deficient Mice With Fragile Epithelia 2573

transgenic mice carrying dominant negative muta- Histologic Examination and tions of these keratin genes.15'16 In transgenic mice, Immunohistochemical Staining the dominant expression of mutated epidermal kera- The excised corneas were fixed in 4% paraformalde- tin 14 results in clinical features similar to those of hyde at 4°C overnight and embedded in paraffin as human epidermolysis bullosa simplex.1516 Recently, a previously described.20'22 Five-micrometer-thick sec- null mutation of K8 keratin by gene targeting has been tions were mounted on Super Frost slides (Fisher Sci- reported.17 Many of the homozygous mutant mice are entific, Pittsburgh, PA). Histologic examination was embryonic lethal at 10 days of gestation, when the K8- performed after Harris hematoxylin and eosin stain- K18 pair normally would be expressed during em- ing. Immunohistochemical staining was performed us- bryogenesis.17 These results demonstrate that keratin ing epitope-specific anti-keratin 1222 and anti-keratin intermediate filaments are vital for the integrity of 14 antibodies23 (a generous gift of Dr. D. R. Roop, epithelium. Baylor College of Medicine, Houston, TX) as pre- Expression of the K3-K12 keratin pair has been 20 22 viously described. ' found in human, bovine, guinea pig, rabbit, mouse, and chicken corneas and is regarded as a marker for In Situ Hybridization corneal-type epithelial differentiation. 1~3'5'718'19 We previously cloned the murine Krtl.12 gene and dem- The plasmid DNA containing a full-length keratin 12 onstrated that its expression is corneal epithelial cell cDNA insert was linearized with EcoR I and Kpn I. T- specific, differentiation dependent, and developmen- 7 and T-3 RNA polymerase were used to synthesize tally regulated.20'21 The cornea-specific nature of kera- antisense and sense riboprobes with digoxigenin using procedures recommended by the manufacturer (Boe- tin 12 gene expression signifies that keratin 12 plays hringer-Mannheim, Indianapolis, IN). Tissue sec- a unique role in maintaining normal corneal epithe- tions (5 fim) were incubated with riboprobes (1 //g/ lium functions. No hereditary human corneal epithe- ml using procedures recommended by Boehringer— lial disorder has been linked direcdy to a mutation of Mannheim). The nonspecific binding of riboprobes the keratin 12 gene, although such a human genetic was removed by RNase (20 //g/ml) digestion and a defect may exist. Nevertheless, the exact function of stringent wash in 0.2 X SSC at 65°C.20'22 Tissue sections keratin 12 remains unknown. To elucidate the func- were incubated with anti-digoxigenin antibody alka- tion of keratin 12, we have created knockout mice line phosphatase conjugate at 4°C overnight. The hy- lacking the Krtl.12 gene by gene targeting techniques. bridization signal was visualized with 5-nitroblue tetra- The homozygous mutant mice (Krtl.12, —/ — ) are zolium chloride as recommended by Boehringer— characterized by fragile corneal epithelium. Mannheim.

MATERIALS AND METHODS Western Blot Analysis Animal Experiments To isolate keratins, frozen tissues were first homoge- Animal experiments were performed in compliance nized in an extraction solution containing 0.1% Tri- with the ARVO Statement for the Use of Animals in ton X-100 and a mixture of protease inhibitors in Tris- Ophthalmic and Vision Research. Adult mice were saline (0.15 M NaCl, 20 mM Tris-HCl, pH 7.5) with anesthetized by intraperitoneal injections of 70 mg/ a Tissuemizer (Tekmar, Cincinnati, OH) as previously 22 24 kg sodium pentobarbital. Under a stereomicroscope, described. ' The homogenate was centrifuged at a partial epithelial defect was created in both eyes by 12,000 rpm for 15 minutes. The supernatant was dis- scraping the corneal surface with a number 69 Beaver carded, and the pellet was rehomogenized in extrac- blade (Becton-Dickinson, Franklin Lakes, NJ) or tion buffer containing 9 M urea in Tris-saline to ex- brushing with a wet Microsponge (Alcon, Dallas, TX). tract the keratin. After centrifugation, equal volumes Partial defects occupied approximately 60% to 70% of 2 X sodium dodecyl sulfate-polyacrylamide gel of the total corneal epithelium. Neomycin ointment electrophoresis sample buffer was added to the super- was applied on the eyes immediately after surgery. natant. The samples were boiled for 10 minutes and Eyes were examined using a stereomicroscope (Olym- were then subjected to Western immunoblot analysis pus, Melville, NY) every other day beginning on the in 7% or 8% acrylamide using epitope-specific anti- first day after wounding to evaluate reepithelialization keratin 12 antibodies, anti-K12n, and anti-K12c as pre- and to detect any signs of infection.22 The animals viously described.22'24 were killed in a CO2 chamber, and the corneas were removed. Corneas either were embedded in paraffin Northern Blot Analysis for histology, immunohistochemistry, or in situ hy- Total RNAs were isolated from mouse corneas with bridization, or they were prepared for Western blot TRI reagent (MRC, Cincinnati, OH), using proce- analysis as described below. dures recommended by the manufacturer. RNAs were

Downloaded from iovs.arvojournals.org on 09/29/2021 2574 Investigative Ophthalmology & Visual Science, December 1996, Vol. 37, No. 13

R R I R _L 8 neo J

+/+ +/- -/- +/+ +/- -/- M

23 kb 2.0 kb

386bp (TSHp)- -386bp(TSHp) 284bp(Neo) - •288bp(K12)

Neo PCR K12PCR FIGURE i. Genotypes of keratin 12-deficient mice produced through gene-targeting techniques. (A) Diagram indicating the recombination event leading to the production of the K] 2 knockout allele in transgenic mice. Possible recombination sites are indicated by crossed lines. Arrows indicate the primers used in polymerase chain reaction (PCR) analysis of the mice. Exons are indicated by black boxes. Introns are indicated by horizontal lines. R = EcoR I; Neo = neomycin-resistant gene; tk = thymidine kinase. (B) Southern hybridization of Eco Rl-digested genomic DNA from the tails of wild-type (+/+), heterozygous (+/—), and homozygous (-/—) mice. The probe used is a full-length 32P-labeled K12 cDNA. The knockout allele gives rise to two 4.3 kb fragments. The wild-type allele yields a ^9.^ kb band. Marker bands and sizes are indicated by the arrows to the right of the blot. (C) PCR analysis of genomic DNA from the tails of wild-type ( + /+), heterozygous ( + / — ), and homozygous {—/ —) mice. Neo-PCR is conducted using TSH-/? (internal control) and neo- primer pairs (left). K12 PCR is conducted using TSH-/3 (internal control) and K12 primer pairs (right). Wild-type mice give rise to a 386 bp TSH-/3 product and a 288 bp K12 product. Heterozygous mice give rise to a 386 bp TSH-/? product, a 288 bp K12 product, and a 284 bp neo-product. Homozygous mice give rise to a 386 bp TSH-/? product and a 284 bp neo- product. M" = size marker from \-phage Hind III digest.

subjected to Northern blot hybridization with 32P-la- was ligated to the BamH I site, and a 0.8 kb exon 8 beled cDNA probes of mouse keratin 12 as described fragment was ligated to Xho I site of MJK-KO vector previously.20 (a generous gift of Drs. S. Potter and M. J. Kern, Uni- versity of Cincinnati, OH). Thus, the DNA element Gene Targeting between intron 2 and exon 8 of the keratin 12 gene To prepare the targeting vector, a 5 kb 5'-end EcoR was replaced by the neomycin-resistant gene, and a I-BamH I genomic DNA fragment of the keratin 12 thymidine kinase gene flanked exon 8 of keratin 12

Downloaded from iovs.arvojournals.org on 09/29/2021 Keratin 12-Deficient Mice With Fragile Epithelia 2575

gene in the targeting construct (Fig. 1A). The tar- TSH-/3-2, 5'-GTAACTCACTCATGCAAAGT-3'31; geting vector was linearized by Not I digestion and neo-1, 5'-CTATGACTGGGCACAACAGACAATC- was electroporated into the embryonic stem cell line, 3'; El4.1.25 The embryonic stem cells were cultured in neo-2, 5'-TGATGGATACCTTTCTCGGCAGGAG- the presence of G418 and gancyclovir.26"28 The drug- 3'31; resistant cells were cloned by growing each individual keratin 12 (I4-R), 5'-CAATCTCTAGGTTCTGCA- colony in a well of 96-well tissue culture plate. To 3'; identify the homologous recombinants, genomic keratin 12(I4-D), 5'-GAGTGGATCTCACCAAGG- DNAs isolated from embryonic stem cells were sub- 3'.21 jected to Southern blot hybridization with the 32P-la- All these primers were synthesized at die DNA beled genomic DNA fragment, which flanks the kera- core facility, University of Cincinnati. tin 12 gene. The homologously recombinant embryonic stem cells were used in blastocyst injection. Electron Microscopic Study Creation of Transgenic Mice Lacking Krtl.12 For electron microscopy, the corneas excised from Gene experimental animals were cut into small pieces and fixed with cold 2.5% glutaraldehyde (Ted Pella, Redd- To produce chimeric mice carrying the ablated kera- ing, CA) in 0.1 M phosphate buffer, pH 7.4 overnight tin 12 gene, the modified embryonic stem cells were at 4°C. After washing in phosphate buffer, the tissues cloned and used in blastocyst injection.26'29 Chimeric were postfixed in 1% OsO4 in phosphate buffer for 2 mice then were crossed with NIH Swiss Black mice hours after dehydration with graded ethanol and were (Taconic, Germantown, NY), and mice heterozygous embedded in Epok 821 (Okon, Tokyo, Japan). Ultra- for the targeted allele were identified by Southern thin sections either were stained with the tannic acid hybridization, polymerase chain reaction, or both. method as described by Kajikawa et al32 or were doubly The heterozygous mice were bred to homozygosity. stained with uranyl acetate and lead citrate. All sec- Southern Hybridization tions were examined with a Hitachi (Tokyo, Japan) H-7000 electron microscope at an accelerating voltage Genomic DNA was isolated from cultured embryonic of 75 kV. stem cells or mouse tissues using proteinase K digestion in 0.2% sodium dodecyl sulfate at 65°C overnight as previously described.30 Restriction enzyme frag- RESULTS ments from genomic DNA were electrophoresed in Gross Phenotypic Changes of Mice Lacking 1% agarose and transferred to Magna Charge nylon Krtl.12 Gene membrane (MSI, Westboro, MA). Southern hybridization was carried out with 32P-labeled cDNA probes at To elucidate the role of keratin 12 in the maintenance 41°C in 50% formamide as previously described.21 of corneal epithelium integrity, we bred mice lacking the Krtl.12 gene through gene-targeting techniques. A Polymerase Chain Reaction targeting DNA construct was prepared by inserting a Approximately 200 ng genomic DNA obtained from neomycin-resistant gene between the BamH I site in experimental animals were used in a 50 /xl polymerase intron 2 and the Stu I site in exon 8, and a thymidine chain reaction (PCR) mixture containing 0.1 fiM kinase gene was ligated to the 3'-end of exon 8 of the primer, 0.2 mM dNTP, 0.1 U Taq polymerase (Pro- keratin 12 gene (as shown in Fig. 1A). The embryonic mega, Madison WI), 2.5 mM MgCl , 50 mM KC1, 1% stem cells, line E14.1, were electroporated with the tar- 2 geting construct and cultured in the presence of G418 Triton X-100, and 100 mM Tris-HCl, pH 9. In each 25 28 PCR, a pair of primers for the mouse thyrotropin (3- and gancyclovir as previously described. " Approxi- mately 500 colonies were picked and analyzed by South- subunit gene (TSH-/?) was included as a positive inter- 32 nal control, in addition to the presence of a specific ern hybridization with the P-labeled 5'- and 3'-genomic primer pair for keratin 12 and neomycin, respectively. DNA fragments that flank the keratin 12 gene. One The PCR cycles were set as follows: 5' at 94°C followed clone was identified as a homologous recombinant and by 35 cycles of 30 seconds at 94°C, 20 seconds at 53°C, was subsequently used in blastocyst injection to produce and 30 seconds at 72°C. This is then ended with a 4° chimeras (data not shown). Three animals were germ- soak. The expected sizes of the PCR products are 386 line chimeras and gave rise to heterozygous animals. The bp, 284 bp, and 288 bp for TSH 0, neo-gene and heterozygous mice were bred to homozygosity. Figure IB keratin 12, respectively. The primer sequences are as demonstrates a representative Southern blot hybridiza- follows: tion of EcoR I-digested tail DNA isolated from wild-type, heterozygous, and homozygous mice with 32P-labeled TSH-/M, 5'-TCCTCAAAGATGCTCATTAG-3'; keratin 12 cDNA. The wild type yields a 9.3 kb fragment

Downloaded from iovs.arvojournals.org on 09/29/2021 2576 Investigative Ophthalmology 8c Visual Science, December 1996, Vol. 37, No. 13

from authentic keratin 12 gene, the heterozygote gives the corneal epithelia of most — /— mice appear nor- rise to a 9.3 kb from the authentic keratin 12 gene and mal, and fluorescein staining occasionally reveals sub- two 4.3 kb fragments from the modified keratin 12 gene, tle epithelial defects (data not shown).. However, the and the homozygote yields two 4.3 kb fragments from corneal epithelium can be removed easily by gentle the modified keratin 12 gene. Subsequently, polymerase brushing with a Microsponge (Alcon) made wet with chain reactions with primers specific for the neo-gene phosphate-buffered saline (bottom right panel of Fig. and exons 4 and 5 of the keratin 12 gene were used to 2), whereas the corneal epithelia of the wild-type and identify the genotype of the targeted mice. A pair of heterozygous mice remain firmly attached to the cor- primers specific for mouse TSH-/? was used as a positive neal surface after similar treatment (top right and control for PCR (Fig. 1C). The wild type has a 288 bp middle right panels of Fig. 2). Histologic examination keratin 12 fragment but no 284 bp neo-gene fragment. confirms the presence of epithelial defects in homozy- The heterozygote has PCR products from both keratin gous mice after brushing as described below. 12 and neo-genes, whereas the homozygote only yields To examine whether the knockout mice might be a single neo-gene fragment. impaired in wound healing, partial epithelial defects (3 The homozygous mice (Krtl.12, —/—) reproduce mm in diameter) were created by scraping the corneal normally and do not have defects in skin or hair as surface with a#69 Beaver blade (Becton-Dickinson). The judged by gross examination. Without perturbation, injured corneas were allowed to heal for up to 7 days.

FIGURE 2. (oppositepage, top left) Corneal epithelial detects produced by brushing the corneal surfaces of knockout mice with a Microsponge. Experimental animals were anesthetized and brushed with a wet Microsponge as described in Materials and Methods. The eyes were stained with Fluo-I-Strip (Ayerst Laboratories, Philadelphia, PA) and examined with a stereomicroscope. (left) Before brushing, (right) After brushing, (top) Wild type; (middle) heterozygote; (bottom) homozygote. Brushing produces an epithelial erosion-stained green in the homozygous mouse. The staining is caused by the loss of superficial cells that form the barrier to the uptake of fluorescein by the cornea and often not to complete erosion. FIGURE 3. (opposite page, top right) Healing of epithelial defects in keratin 12-deficient mice. Mice were anesthetized, and a 3 mm epithelial defect was created in the center of each cornea with a #69 Beaver blade as described in Materials and Methods. Epithelial defect sizes were measured by the areas of corneal surfaces stained by the fluorescent dye, as shown in Figure 2. One day after injury, the percentage of the epithelial defect remaining in cornea was calculated by dividing the area of defect of day 1 by the area of defect of day 0. Two independent experiments were performed. All corneal epithelial defects healed 3 days after injury. Numbers in parentheses are the numbers of eyes used in each group of animals. There was no significant difference in healing between wild-type (+/ + ) and heterozygous (+/—) mice. The defects of homozygous (—/-) mice healed significantly more slowly than those of +/+ and +/— mice. FIGURE 4. (oppositepage, middle right)Wester n immunoblot analysis of keratin extracted from the eyes of keratin 12-deficient mice. Urea-soluble keratin fractions extracted from the whole eyes of transgenic mice were used in Western immunoblot analysis (8% acrylamide) with the epitope-specific anti-keratin antibodies, anti-K12n and anti-K12c. (A) Amido black staining of proteins in urea fractions. (B) Anti-K12n antibody staining. (C) Anti-K12c antibody staining. No 55 kDa or truncated keratin 12 molecule can be seen in — /— mice, whereas a 55 kDa keratin 12 molecule is detected in both +/+ and +/— mice, std = molecular weight standard. FIGURE 5. (opposite page, bottom) Immunostaining of corneas from keratin 12-deficient mice. Corneal sections (6 /zm) were incubated with rabbit anti-keratin antibodies at 1 fxg/m\ at 4°C overnight and then were incubated with goat anti-rabbit immunoglobulin G-peroxidase conjugates. The reactions were visualized with diamnobenzidine as described in Materials and Methods. Nonimmune rabbit immunoglobulin G was used in the control, and no positive reaction was seen (data not shown), (left) Wild-type mouse, (middle) Heterozygote mouse, (right) Homozygote mouse, (toproiu) Anti-K12n. (middle row) Anti-K12c. (bottom roiu) Anti-keratinl4. All cell layers of corneal epidielia from wild-type and heterozygous mice were labeled by the anti-keratin 12 antibodies, whereas those of homozygous mice were not labeled. The anti-keratin 14 antibodies reacted to basal and suprabasal corneal epithelial cells of all mice.

Downloaded from iovs.arvojournals.org on 09/29/2021 Keratin 12-Defident Mice With Fragile Epithelia 2577

Before After 1 P<0.2%

60 (13) 50 P>S%

30 (20) (16) 1 (11) 20 • I• 10 • 0 •_ • M r •1- +/+ -/ GeaotTpe FIGURES. +/•

MHCUi FIGURES. FIGURE 4.

K12n

K12c

K14

FIGURE 5.

Downloaded from iovs.arvojournals.org on 09/29/2021 2578 Investigative Ophthalmology 8c Visual Science, December 1996, Vol. 37, No. 13

lial cells of wild-type, heterozygous, and homozygous mice (bottom row of Fig. 5). To examine whether the modified keratin 12 gene is transcribed to produce stable mRNA, total RNA isolated from the eyes of ex- 2.4 kb perimental animals was subjected to Northern hybridization. Figure 6 demonstrates that no keratin 12 mRNA or its modified form can be detected in the corneas of homozygotes, whereas the keratin 12 mRNA is found in wild-type and heterozygous mice. Similarly, in situ hybridization demonstrates that positive signals are found in the corneal epithelia of normal and heterozygous mice (Figs. 7B, 7C), but not in 1.4 kb homozygous mice (Fig. 7D).

FIGURE 6. Northern blot hybridization of total RNAs isolated from keratin 12-defkient mice. A Northern blot of total RNA from the eyes of+/ + , +/—, and — /— mice was probed with a full-length MP-labeled K12 cDNA. Wild-type (+/ + ) and heterozygous knockout (+/—) mice yielded an «1.8 kb band. Homozygous {—/—) mice produced no detectable signal.

The areas of the epithelial defects were estimated by photographs of fluorescent staining with a stereomicroscope. Figure 3 demonstrates that 1 day after injury, a 20% epithelial defect remains in +/ + and +/- mice, whereas a 50% epithelial defect remains in — /— mice. Nevertheless, 3 days after injury, comeal epithelial defects healed in all experimental animals.

Expression of Keratin 12 in Targeted Mice Northern hybridization, in situ hybridization, Western immunoblot analysis, and immunostaining with epitope-specific anti-keratin 12 antibodies (anti-K12n and anti-K12c) were used to examine the expression of keratin 12 in wild-type, heterozygous, and homozygous mice. Urea-soluble fractions prepared from the whole eyes of experimental animals were subjected to Western blot analysis with epitope-specific anti-K12n and anti-K12c antibodies. Figure 4A shows the pattern of the amido black staining, indicating that equal amounts of proteins from each sample were subjected to sodium dodecyl sulfate-polyacrylamide gel electro- D phoresis. Figures 4B and 4C demonstrate that anti- FIGURE 7. In situ hybridization of corneal epithelia of keratin K12n and anti-K12c antibodies react to the 55-kDa 12-deficient mice. Corneal sections (5 ^,m) were incubated keratin 12 from normal and heterozygous mice. No with digoxigenin-labeled sense and antisense riboprobe of 55-kDa keratin molecule was detected in homozygous keratin 12 and were incubated further with anti-digoxigenin mice (—/— lanes of Figs. 4B and 4C). Similarly, immu- alkaline phosphatase conjugate. The reactions were visual- nostaining with the two epitope-specific antibodies did ized with 4-nitroblue tetrazolium chloride, as described in not label the corneal epidielia of homozygotes, Materials and Methods. (A) Sense riboprobe. (B toD) Anti- whereas positive staining was seen in the corneas of sense riboprobe. (A,B) Wild-type cornea. (C) Heterozygous cornea. (D) Homozygous cornea. Hybridization signals are normal and heterozygous mice (top and middle rows seen in all cell layers of corneal epithelia from wild-type and of Fig. 5). Anti-keratin 14 antibodies labeled the basal heterozygous mice but are not seen in those of homozygous and suprabasal, but not the superficial, corneal epithe- mice.

Downloaded from iovs.arvojournals.org on 09/29/2021 Keratin 12-Deficient Mice With Fragile Epithelia 2579

unTreatetf i' V*'

•h treated •> J FT.

20 Mm

FIGURE 8. Light microscopy of the corneas of keratin 12-deficient mice. Experimental mice were anesthetized, and the corneal surfaces were scraped with a wet Microsponge. Corneal sections (5 (J.m) were stained with Harris hematoxylin and eosin and were examined with a Nikon Diaphot microscope (Nikon, Melville, NY). (A) Wild-type cornea-treated, (B) heterozygote-treated, (C) homozygote-untreated, and (D) homozygote-treated epithelial cells were removed from the corneal surface by scraping.

Histologic Examination of Corneas of Targeted cells as shown in Figure 8D, and extensive brushing Mice (more than 10 times) eventually can produce a com- To examine the morphology of corneal epithelial pletely denuded corneal surface without basal epithe- cells, the corneas from normal and targeted mice were lial cells (Fig. 2D; histology data not shown). subjected to light and electron microscopy using pre- 20 22 Electron Microscopy of Corneal Epithelium of viously described procedures. ' Figures 8A and 8B the Targeted Mice show that the stratified corneal epithelium of +/+ and +/— mice is attached firmly to the cornea! surface Transmission election microscopy was performed to ex- after brushing with a Microsponge (Alcon). Figure 8C amine further the corneal epithelial cell morphology in shows the corneal epithelial cell morphology of a —/ targeted mice. Figure 9A shows an electron micrograph — mouse without brushing. There are approximately of a normal corneal epithelium (magnification, X2000) three to four fewer cell layers of the stratified epithe- in which 9 to 10 cell layers can be identified. Selected lium in the corneas of -/- mice than those of +/ + areas were examined at higher magnification (X IOJOOO) . and +/- mice (compare Figs. 8A, 8B, 8C). In -/- Numerous hemidesmosomes and desmosomes are iden- mice, superficial epithelial cells are swollen and do tified readily (Figs. 10A, 10B, IOC). The superficial epi- not firmly adhere to the underlying cell layers; occa- thelial cells of +/+ mice are attached closely to each sionally, detached superficial cells from the corneal other (Fig. 10A). Figure 9B shows a micrograph of —/ surface can be seen (Fig. 8C). It is of interest to note — corneal epithelium, in which only six to seven cell that no significant morphologic differences can be layers can be identified. The number of desmosomes detected in the corneal basal epithelial cells of — / — and hemidesmosomes appear normal in basal and su- mice compared to those of +/+ and +/— mice. Gen- prabasal epithelial cells of —/— mice when compared tle brushing of the corneal surface with a Microsponge to those of+/+ mice (Figs. 10E, 10F). Fewer superficial (Alcon) causes the removal of superficial epithelial epithelial cell layers and some cells undergoing cytolysis

Downloaded from iovs.arvojournals.org on 09/29/2021 2580 Investigative Ophthalmology & Visual Science, December 1996, Vol. 37, No. 13

FIGURE 9. Electron microscopy of corneal epithelia from keratin 12-deficient mice. Thin corneal sections from wild-type mice were examined by transmission electron microscopy as described in Materials and Methods. (A) Low magnification of corneal epithelium of a +/+ mouse, in which 9 to 10 cell layers can be identified. (B) Low magnification of corneal epithelium of a —/— mouse, in which 4 to 5 cell layers can be seen. Cells on the corneal surface underwent cytolysis, as indicated by the asterisks.

can be seen in —/- mice (Figs. 9B, 10D). In the normal sues. Keratin gene mutations have been linked to sev- corneal epithelium, keratin intermediate filaments exist eral hereditary skin diseases.10"14 The K3-K12 keratin as fine filamentous networks that differ from those dense pair is expressed specifically by corneal epithelial bundles of intermediate filaments seen in epidermal epi- cells.2021 To establish whether this cornea-specific ker- thelial cells.'*3 These fine keratin intermediate filaments atin 12 is essential for the health of corneal epithe- can be seen in all cell layers of the corneal epithelium of lium, we have bred three lines of mice lacking the +/+ mice (shown in Figs. 10A, 10B, IOC). In suprabasal Krtl.12 gene through gene-targeting techniques. All corneal epithelial cells of homozygous —/— mice, the mice of the same genotype show an identical pheno- keratin intermediate filaments resemble diose of epider- type. The homozygous keratin 12-deficient mice (—/ mal keratinocytes and appear in large, dense bundles —) are fertile and appear to have normal skin and (Fig. 10E). The number of keratin intermediate fila- hair as judged by gross examination. The heterozy- ments is reduced greatly in basal and suprabasal epithe- gous Krtl.12 +/— mice have an identical or similar lial cells of-/— mice (Figs. 10E, 10F). Interestingly, the phenotype to that of wild-type (+/ + ) mice (e.g., heal- superficial epithelial cells are devoid of keratin interme- ing of epithelial defects) (Figs. 2, 3). The homozygous diate filaments in —/— mice (Fig. 10D). Krtl.12 —/— mice have fewer cell layers in the corneal epithelium (Figs. 8, 9B). These epithelia are fragile and do not adhere firmly to the corneal surface (Figs. DISCUSSION 2, 8). Although the healing of epithelial defects is Keratin intermediate filaments play a pivotal role in impaired slightly (Fig. 3), the clinical manifestation is the maintenance of epithelium integrity in various tis- mild in —/— mice. For example, no corneal ulceration

Downloaded from iovs.arvojournals.org on 09/29/2021 Keratin 12-Deficient Mice With Fragile Epithelia 2581

FIGURE 10. Electron microscopy of corneal epithelia from keratin 12-deficient mice. (A to C) Corneal epithelium of a +/+ mouse; (D to F) Corneal epithelium of a -/- mouse. A and D superficial cells; B and E are suprabasal cells; C and F are basal cells. The corneal basal and suprabasal epithelial cells of both +/+ and —/— mice have numerous hemidesmosomes and desmosomes. In +/+ mice, keratin intermediate filaments are in fine filamentous networks in all epithelial cells and are difficult to identify. In -/— mice, the number of keratin intermediate filaments decreases in basal and suprabasal epithelial cells, and they appear as large, dense bundles of keratin filaments, a morpholog)' similar to that of epidermal epithelial cells. Cells on the corneal surface are devoid of keratin intermediate filaments and undergo cytolysis, as indicated by the asterisks.

or perforation has been observed in —/— mice. Be- Western and immunostaining analysis demon- cause corneal epithelial cells regenerate quickly, the strate the absence of keratin 12 molecules recognized loss of superficial epithelial cells in —/— mice is re- by the epitope-specific anti-K12n and anti-K12c anti- placed promptly; thus, severe epithelial defects do not bodies (Figs. 4, 5). No stable mRNA derived from the take place. This mild clinical manifestation of —/ — modified Krtl. 12 gene can be detected in the corneas mice provides an explanation for the lack of a report of +/— and — /— mice as determined by Northern of human hereditary corneal epithelial disease linked and in situ hybridization (Figs. 6, 7). It is possible that to the mutation of the keratin 12 gene. the transcript of the modified Krtl. 12 gene is unstable

Downloaded from iovs.arvojournals.org on 09/29/2021 2582 Investigative Ophthalmology & Visual Science, December 1996, Vol. 37, No. 13

and quickly degraded intracellularly, or that the inser- lial cells become devoid of keratin intermediate fila- tion of the neo-gene greatly reduced transcription of ments in — /— mice, and the absence of keratin inter- the modified Krtl.12 allele, or both. In wild-type mice, mediate filaments may account for the cytolysis of corneal epithelial cells express keratin 12 at a high superficial corneal epithelial cells (Figs. 9B, 10D). Sim- level.20 Although the regulatory ds-DNA elements es- ilarly, the decrease of keratin intermediate filaments sential and sufficient for cornea-specific keratin 12 ex- in the absence of the keratin 14 gene in humans,11'36 pression have not been identified, the lack of tran- or knockout mice bred through gene targeting,33 re- script from the modified Krtl.12 allele may be the sults in the cytolysis of basal epithelial cells in epider- result of loss of certain cornea-specific regulatory DNA mis and corneas. elements located between intron 2 and exon 8 of kera- It has been demonstrated that the expression of tin 12 gene. It has been reported that the enhancer mutated keratin genes produces dominant negative ef- elements of keratin genes may be located within in- fects on the structure of keratin intermediate fila- trons of the genes. For example, intron 1 of keratin ments.10"16 Because the interruption of the keratin 12 18 contains an enhancer element for its tissue-specific 34 gene does not produce truncated keratin 12 molecules expression in simple epithelia. Further studies are (Figs. 4 to 7), it is unlikely that the formation of dense needed to establish whether cornea-specific regulatory bundles of keratin intermediate filaments in suprabasal elements may be located in introns of the keratin 12 corneal epithelial cells of —/— mice results from the gene. Nevertheless, the absence of truncated mRNA truncated keratin 12 molecules. It is possible, however, is consistent with the notion that the interruption of that the presence of excessive keratin 3 or other part- keratin 12 gene by neo-gene prevents the production ner^) of keratin 12 may cause the abnormal polymeriza- of truncated keratin 12 molecules in the corneal epi- tion of K5-K14 pairs in the corneal epithelial cells of thelia of — /— mice. —/- mice. The mechanism by which different keratin Another interesting observation in —/— mice is types may interact with each other and may modulate the morphologic changes of the corneal epithelial the formation of intermediate filament remains un- cells. The basal corneal epithelial cells of —/— mice, known. For example, the K14-K5 keratin pair forms like those of +/ + and +/— mice, express keratin 14, dense bundles of intermediate filaments in epidermal and their integrity does not seem to be compromised epithelial cells,33 whereas in the presence of K12-K3 and in the absence of keratin 12. The formation of desmo- K14-K5 keratins, they form fine filamentous networks in somes and hemidesmosomes (Figs. 10E, 10F) is not corneal epithelial cells (Figs. 9, 10). The ablation of disturbed in the — /— mice, even though the number keratin 14 by gene targeting leads to the alteration of of keratin intermediate filaments in basal and su- intermediate filament morphology. For example, whisky prabasal epithelial cells is reduced substantially. The keratin intermediate filaments are found in the epider- keratin intermediate filaments that parallel the under- mal epithelial cells of K14 knockout mice.33 It is also not lying basement membrane can be identified readily known whether various keratin types can copolymerize in the — /— mice. This can be explained, in part, by and modulate the morphology of keratin intermediate the fact that the expression of keratin 14 is not per- filaments. It has been demonstrated that copolymeriza- turbed in the cornea of — /— mice (Fig. 5). The reduc- tion of various collagen types can control the diameter tion in the number of intermediate filaments is more of collagen fibrils found in the extracellular matrix.37'38 prominent in the suprabasal and superficial cell layers Therefore, our Krtl.12 —/— mice can serve as a model of —/— mice (Figs. 10D, 10E). The intermediate fila- for elucidating the mechanism by which the formation ments of —/— mice appear to form larger dense bun- of keratin intermediate filament is regulated in cells that dles compared to those of +/+ mice (Fig. 10). The coexpress several different keratin types. reason for the morphologic changes of keratin intermediate filaments in — /— mice is unknown. It is Key Words known, however, that in addition to the K3-K12 kera- corneal epithelial defect, differentiation, gene targeting, tin pair, the corneal basal epithelial cells also express keratin 12, mouse the K5-K14 keratin pair (Fig. 5) expressed by basal 1 33 35 cells of most stratified epithelia. ' ' The presence of References K5-K14 keratin intermediate filaments may be sufficient for the maintenance of basal epithelial cell integ- 1. Moll R, Franke WW, Schiller DL, Geiger B, Krepler R. The catalog of human cytokeratins: Patterns of ex- rity of — /— mice. When the epithelial cells migrate pression in normal epithelia, tumors and cultured upward and become differentiated, they reduce their 1>33>3D cells. Cell. 1982;31:11-24. expression of the K5-K14 keratin pair, and even- 2. Galvin S, Loomis C, Manabe M, Dhouailly D, Sun TT. tually they no longer synthesize this keratin pair when The major pathways of keratinocyte differentiation as they become the superficial cells in the corneal epithe- defined by keratin expression: An overview. Adv Der- lium (Fig. 5). Thereby, the superficial corneal epithe- matol. 1989; 4:277-299.

Downloaded from iovs.arvojournals.org on 09/29/2021 Keratin 12-Deficient Mice With Fragile Epithelia 2583

3. Cooper D, Schermer A, Sun TT. Classification of hu- adult corneal epithelial differentiation in the chick man epithelia and their neoplasms using monoclonal and in the rabbit. Cell Diff Dev. 1990; 32:97-108. antibodies to keratins: Strategies, applications, and 19. Tseng SC, Jarvinen MJ, Nelson WG, Huang JW, limitations. [Review.] Lab Invest. 1985;52:243-256. Woodcock-Mitchell J, Sun TT. Correlation of spe- 4. Blessing M, Jorcano JL, Franke WW. Enhancer ele- cific keratins with different types of epithelial differ- ments directing cell-type-specific expression of cyto- entiation: Monoclonal antibody studies. Cell. keratin genes and changes of the epithelial cytoskele- 1982;30:361-372. ton by transfections of hybrid cytokeratin genes. 20. Liu CY, Zhu G, Westerhausen-Larson A, et al. Cor- EMBOJ. 1989;8:117-126. nea-specific expression of K12 keratin during mouse 5. Quinlan RA, Schiller DL, Hatzfeld M, et al. Patterns development. Cwr Eye Res. 1993; 12:963-974. of expression and organization of cytokeratin interme- 21. Liu CY, Zhu G, Converse R, et al. Characterization diate filaments. [Review.] Ann NY Acad Sci. 1985; and chromosomal localization of the cornea-specific 455:282-306. murine keratin gene Krtl.12. / Biol Chem. 1994; 6. Steinert PM, Idler WW, Zhou XM, et al. Structural 269:24627-24636. and functional implications of amino acid sequences 22. Moyer PD, Kaufman AH, Zhang Z, Kao CW-C, Spauld- of keratin intermediate filament subunits. [Review.] ing AG, Kao WW. Conjunctival epithelial cells can Ann NY Acad Sci. 1985;455:451-461. resurface denuded cornea but does not transdifferen- 7. Cooper D, Sun TT. Monoclonal antibody analysis of tiate to express cornea-specific keratin 12 following bovine epithelial keratins: Specific pairs as defined by removal of limbal epithelium in mouse. Differentiation. coexpression. JBiol Chem. 1986; 261:4646-4654. 1996; 60:31-38. 8. Dlugosz AA, Yuspa SH. Coordinate changes in gene 23. Roop DR, Cheng CK, Toftgard R, Stanley JR, Steinert expression which mark the spinous to granular cell PM, Yuspa SH. The vise of cDNA clones and monospe- transition in epidermis are regulated by protein kinase cific antibodies as probes to monitor keratin gene ex- C.J Cell Biol. 1993;20:2l7-25. pression. Ann NY Acad Sci. 1985; 455:426-435. 9. O'Guin WM, Galvin S, Schermer A, Sun TT. Patterns 24. Zhu G, Ishizaki M, Haseba T, Wu RL, Sun TT, Kao of keratin expression define distinct pathways of epi- WW. Expression of K12 keratin in alkali-burned rabbit thelial development and differentiation. [Review.] corneas. Curr Eye Res. 1992; 11:875-887. Curr Topics Dev Biol. 1987; 22:97-125. 25. Hooper M, Hardy K, Handyside A, Hunter S, Monk 10. Coulombe PA, Hutton ME, Letai A, Hebert A, Paller M. HPRT-deficient (Lesch-Nyban) mouse embryos derived from germline colonization by cultured cells. AS, Fuchs E. Point mutations in human keratin 14 Nature. 1987; 326:292-295. genes of epidermolysis bullosa simplex patients: Ge- 26. Shull MM, Ormsby I, Kier AB, et al. Targeted disrup- netic and functional analyses. Cell. 1991; 66:1301 — 11. tion of the mouse transforming growth factor-beta 1 11. Rugg EL, McLean WH, Lane EB, et al. A functional gene results in multifocal inflammatory disease. Na- 'knockout' of human keratin 14. Genes Dev. ture. 1992; 359:693-699. 1994; 8:2563-2573. 27. Capecchi MR. Altering the genome by homologous 12. Lane EB, Rugg EL, Navsaria H, et al. A mutation in recombination. [Review.] Science. 1989; 244:1288- the conserved helix termination peptide of keratin 5 1292. in hereditary skin blistering. Nature. 1992;356:244- 28. Thomas KR, Capecchi MR. Targeted disruption of the 246. murine int-1 proto-oncogene resulting in severe ab- 13. Cheng J, Syder AJ, Yu QC, Letai A, Paller AS, Fuchs normalities in midbrain and cerebellar development. E. The genetic basis of epidermolytic hyperkeratosis: Nature. 1990; 346:847-850. A disorder of differentiation-specific epidermal kera- 29. Papaioannou V, Johnson R. Production of chimeras tin genes. Cell. 1992; 70:811-819. and genetically defined offspring from targeted ES 14. Chipev CC, Korge BP, Markova N, et al. A leucine- cells. In: Joyner AL, ed. Gene Targeting: A Practical Ap- proline mutation in the HI subdomain of keratin 1 proach. New York: IRL Press; 1993:104-146. causes epidermolytic hyperkeratosis. Cell. 1992; 30. Laird PW, Zijderveld A, Linders K, Rudnicki MA, Jae- 70:821-828. nisch R, Berns A. Simplified mammalian DNA isola- 15. Coulombe PA, Hutton ME, Vassar R, Fuchs E. A function procedure. Nucl Acids Res. 1991; 19:4293. tion for keratins and a common thread among differ- 31. Walter CA, Nasr-Schirf D, Luna VJ. Identification of ent types of epidermolysis bullosa simplex diseases. / transgenic mice carrying the CAT gene with PCR am- Cell Biol. 1991; 115:1661-1674. plification. Biotechniques. 1989; 7:1065-1070. 16. Vassar R, Coulombe PA, Degenstein L, Albers K, Fuchs 32. Kajikawa K, Yamaguchi T, Katsuda S, Miwa A. An im- E. Mutant keratin expression in transgenic mice proved electron stain for elastic fibers using tannic causes marked abnormalities resembling a human ge- add. J Electron Microsc. 1975; 24:287-289. netic skin disease. Cell. 1991; 64:365-380. 33. Lloyd C, Yu QC, Cheng J, et al. The basal keratin 17. Baribault H, Price J, Miyai K, Oshima RG. Mid-gesta- network of stratified squamous epithelia: Defining tional lethality in mice lacking keratin 8. Genes Dev. K15 function in the absence of K14. / Cell Biol. 1993;7:1191-202. 1995; 129:1329-1344. 18. Chaloin-Dufau C, Sun TT, Dhouailly D. Appearance 34. Oshima RG, Abrams L, Kulesh D. Activation of an of the keratin pair K3/K12 during embryonic and intron enhancer within the keratin 18 gene by ex-

Downloaded from iovs.arvojournals.org on 09/29/2021 2584 Investigative Ophthalmology 8c Visual Science, December 1996, Vol. 37, No. 13

pression of c-fos and c-jun in undifferentiated F9 for an intermediate filament protein. Genes Dev. embryonal carcinoma cells. Genes Dev. 1990;4:835- 1994;8:2574-2587. 848. 37. Doane KJ, Babiarz JP, Fitch JM, Linsenmayer TF, Birk 35. Stellmach V, Leask A, Fuchs E. Retinoid-mediated DE. Collagen fibril assembly by corneal fibroblasts in transcriptional regulation of keratin genes in human three-dimensional collagen gel cultures: Small-diame- epidermal and squamous cell carcinoma cells. Proceed- ter heterotypic fibrils are deposited in the absence of ings of the National Academy of Sciences of the United States keratan sulfate proteoglycan. Exp Cell Res. 1992; of America. 1991; 88:4582-4586. 202:113-124. 36. Chan Y, Anton-Lamprecht I, Yu QC, et al. A human 38. Rawe IM, Tuft SJ, Meek KM. Proteoglycan and colla- keratin 14 'knockout': The absence of K14 leads to gen morphology in superficially scarred rabbit cornea. severe epidermolysis bullosa simplex and a function HistochemJ. 1992; 24:311-318.

Downloaded from iovs.arvojournals.org on 09/29/2021