Generation of mouse model of TGFBI-R124C corneal dystrophy using CRISPR/Cas9-mediated homology-directed repair

Mutations in transforming growth factor-beta-induced (TGFBI) gene cause clinically distinct types of corneal dystrophies. To delineate the mechanisms driving these dystrophies, we focused on the R124C mutation in TGFBI that causes lattice corneal dystrophy type1 (LCD1) and generated novel transgenic mice harbouring a single amino acid substitution of arginine 124 with cysteine in TGFBI via ssODN-mediated base-pair substitution using CRISPR/Cas9 technology. Eighty percent of homozygous and 9.1% of heterozygous TGFBI-R124C mice developed a corneal opacity at 40 weeks of age. Hematoxylin and eosin and Masson trichrome staining showed eosinophilic deposits in subepithelial corneal stroma that stained negative for Congo-red. Although amyloid deposition was not observed in TGFBI-R124C mice, irregular amorphous deposits were clearly observed via transmission electron microscopy near the basement membrane. Interestingly, we found that the corneal deposition of TGFBI protein (TGFBIp) was significantly increased in homozygous TGFBI-R124C mice, suggesting a pathogenic role for the mutant protein accumulation. Furthermore, as observed in the LCD1 patients, corneal epithelial wound healing was significantly delayed in TGFBI-R124C mice. In conclusion, our novel mouse model of TGFBI-R124C corneal dystrophy reproduces features of the human disease. This mouse model will help delineate the pathogenic mechanisms of human corneal dystrophy.

of the human disease. Therefore, it is imperative to develop physiologically relevant animal models to decipher the mechanisms driving TGFBI corneal dystrophy.
The popularity of clustered regularly interspaced short palindromic repeats/CRISPR-associated protein 9 (CRISPR/Cas9) gene-editing technology has increased dramatically because it enables rapid and precise genetic manipulation of various cell types [9][10][11] . Moreover, CRISPR/Cas9 is a powerful one-step strategy for establishing knockout animal models because it allows single-nucleotide conversion via single-stranded donor oligodeoxynucleotide (ssODN) template-mediated homology-directed repair (HDR) 12,13 . Single-nucleotide polymorphism (SNP)-based animal models of human disease are a useful and physiologically relevant approach that reflects point mutations or substitutions in the human genome.
To understand the effect of TGFBI mutation in corneal dystrophy, we generated mutant mice with the Tgfbi R124C mutation, which is the representative mutation causing LCD1, using the CRISPR/Cas9 approach. We then examined how the expression of this phenotype affected corneal integrity and wound healing in this mouse model.

Results
Generation of mouse model of TGFBI-R124C corneal dystrophy using ssODN-mediated basepair substitution introduced via CRISPR/Cas9. We used CRISPR/Cas9-induced HDR-mediated nucleotide conversion technology to develop a clinically relevant mouse model of TGFBI corneal dystrophy 14 . We targeted the arginine residue at position 124 of TGFBI; this amino-acid residue is important in the development of TGFBI corneal dystrophy. To generate Tgfbi R124C knock-in mice without inducing off-target cleavage of CRISPR/Cas9, we took advantage of Cas9 nickase (Cas9 D10A ), a mutated Cas9 that makes single-strand breaks, which allows us to generate region targeted knock-in mice without off-target mutations 13,15,16 . Two single-guide RNAs (sgRNAs) were designed to target exon4 of Tgfbi (Fig. 1a). To convert arginine 124 into cysteine, we designed a 107 nt ssODN containing the BsrGI restriction enzyme site (Fig. 1b). The BsrGI restriction enzyme site enables easy genotyping of TGFBI-R124C mice by PCR-RFLP analysis, without affecting the amino-acid sequence of the Tgfbi R124C protein. A mixture of Cas9 D10A mRNA, sgRNAs, and ssODN was injected into zygotes, as reported previously 13 . Then, two-cell embryos were transferred into pseudopregnant recipient female mice. The genotype of the resultant pups was examined by PCR-RFLP analysis. We PCR amplified 475 bp fragment around the target site and digested the PCR product with BsrGI restriction enzyme. As the PCR product including a cleavage site for BsrGI, PCR-RFLP digestion using BsrGI produced an undigested 475 bp fragment in wild-type mice, and fragments of 253 and 222 bp in homozygotic TGFBI-R124C mice (Fig. 1c). The genotype of the pups was further examined using DNA sequencing, and 5 in 19 pups had the TGFBI-R124C mutant allele. DNA sequencing of Tgfbi R124C/R124C mice confirmed that the mutation had been successfully induced (Fig. 1d). One of the mutant mice was used for breeding, and the descendants were used for the experiments. Both Tgfbi R124C/R124C and Tgfbi R124C/wt mice were viable and fertile, showing normal body weight and life span.
TGFBI-R124C mice develop corneal opacity with high frequency. We used a stereomicroscope to analyse the effects of the TGFBI-R124C mutation on the development of corneal opacity in mice. In TGFBI-R124C homozygous mice, corneal opacity was first observed at 12 weeks of age and tended to gradually worsen until 24 weeks of age. While wild-type littermates (n = 9) did not show any substantial changes in the corneas (Fig. 2a), all corneal opacity was observed in the central cornea of both TGFBI-R124C homozygous and heterozygous mice (Fig. 2b,c). In TGFBI-R124C homozygous mice, corneal opacity was observed in 51 of 71 mice (71.8%) at 20 weeks of age and in 57 mice (80.3%) at 40 weeks of age (Fig. 2d). On the other hand, corneal opacity was not observed in any TGFBI-R124C heterozygous mice below 24 weeks of age, and only 1 of the11 mice (9.1%) developed bilateral corneal opacity until 40 weeks of age (Fig. 2d). Over 71% of the diseased TGFBI-R124C homozygous mice showed bilateral corneal opacity at 12 weeks of age, and the percentage gradually increased to 91% at 40 weeks of age, whereas 5 mice developed unilateral corneal opacity. In summary, total number of the eye showed corneal opacity at 40 weeks of age are 105 (73.9%) and 2 (9.1%) in the TGFBI-R124C homozygous and heterozygous mice, respectively (Table 1).
Abnormal deposits in the subepithelial stroma of TGFBI-R124C mice. We used histological analysis to evaluate corneal opacity in TGFBI-R124C mice. As demonstrated in Fig. 3a,b, several parts of the subepithelial stroma showed degeneration and fibrosis. No inflammatory cell infiltration was observed. These areas were positive for Masson's trichrome staining, which indicates an increase in collagen fibres (Fig. 3c,d). All corneal tissues were negative for Congo-red staining (Fig. 3e,f). Transmission electron microscopy (TEM) imaging of the corneas showed pathognomonic, irregular, and amorphous deposits in the subepithelial stroma (Fig. 3g,h).
TGFBIp deposits in the subepithelial stroma of TGFBI-R124C mice. To assess whether Tgfbi mutations result in deposits of TGFBIp in the corneas of TGFBI-R124C mice, we analyzed the expression levels of the Tgfbi gene and the TGFBIp protein. First, we performed immunohistochemical analysis to determine the localization of TGFBIp. In the corneas of wild-type mice, TGFBIp immunoreactivity was mainly observed in the corneal epithelium (Fig. 4a). In the corneas of TGFBI-R124C homozygous mice, TGFBIp expression was observed in all the corneal layers (Fig. 4a), whereas no significant staining was observed in control sections.
Expression of Tgfbi mRNA in the corneas of wild-type mice and homozygous mutant mice was examined via real-time reverse transcription-polymerase chain reaction (RT-PCR). Corneal expression of Tgfbi mRNA did not significantly differ between wild-type and TGFBI-R124C mice ( Fig. 4b; p = 0.31). Conversely, expression of TGFBIp in the cornea of homozygous mice was significantly increased compared with that of wild-type mice (

Discussion
The mechanisms driving the formation of corneal opacity in TGFBI corneal dystrophy are largely unknown 5 . Establishing clinically relevant mouse models of TGFBI corneal dystrophy would be very helpful to understand the mechanisms or pathophysiology underlying corneal opacity. Two animal models of TGFBI corneal dystrophy have been established via introduction of human mutant TGFBI genes into mice 6,8 . However, these models were generated by artificial induction of human genetic mutations, and the frequency of corneal opacity observed was insufficient to explain the patient phenotype. Therefore, these models are inadequate for studying the pathogenesis of TGFBI corneal dystrophy.
In this study, we established a mouse model of TGFBI corneal dystrophy using HDR-mediated base substitution introduced via CRISPR/Cas9. To the best of our knowledge, this study is the first to show that a base substitution introduced via CRISPR/Cas9 into Tgfbi can induce the development of corneal dystrophy.
In Tgfbi R124C/R124C mice, more than 70% of eyes developed corneal opacity by 40 weeks of age ( Table 1). The frequency of corneal opacity in our mutant mice was nearly four times greater than that in previously established mouse models 8 . On the other hand, no corneal opacity was observed in Tgfbi R124C/WT mice up to 20 weeks of age, and the frequency of corneal opacity by 40 weeks was lower than that in Tgfbi R124C/R124C mice. In cases of TGFBI corneal opacity, homozygous patients have earlier onset than heterozygous patients 18 ; our finding is consistent with that. Among the mutant mice produced by Yamazoe et al., 8% of heterozygous mutant mice were found to have corneal opacity by 40 weeks of age 8 ; the same result was obtained for the TGFBI-R124C mice that we studied.
Although corneal opacity was mostly bilateral in the Tgfbi R124C/R124C mice, some mice developed unilateral corneal opacity (Fig. 2d). Among the TGFBI-R124C mice with unilateral corneal opacity, some mice developed bilateral corneal opacity. In cases of TGFBI corneal dystrophy, some cases with unilateral corneal opacity developed corneal opacity in the other eye several years later 19 . It is possible that even in the mice that developed unilateral corneal opacity during our study, bilateral corneal opacity could occur later.
In patients with TGFBI corneal dystrophy (such as lattice and granular dystrophy), deposits and opacities occur in the centre of the cornea 3,20 . Consistent with the patterns observed in human corneal dystrophy, we The representative pictures of wild-type at 28 weeks of age mouse and TGFBI-R124C mice eyes at 32 weeks of age. Corneal opacity is not observed in wild-type mice (a). Corneal opacity is observed in the centre of the cornea in heterozygous (b) and homozygous mice (c). Dashed line represents corneal opacity. (d) Frequency of corneal opacity by genotype. In Tgfbi R124C/R124C mice, corneal opacity occurred in 78.9% at 24 weeks of age. Then the frequency of corneal opacity gradually increased to 80.3% of mice at 40 weeks of age. Over 91% of the diseased TGFBI-R124C homozygous mice showed bilateral corneal opacity at 40 weeks of age. In eleven Tgfbi R124C/WT mice (Het), bilateral corneal opacity was observed one mouse (9.1%) before 24 weeks of age. No corneal opacity was observed in wild-type mice (Wt). www.nature.com/scientificreports www.nature.com/scientificreports/ detected a well-defined mass in the central cornea of the TGFBI-R124C mice (Fig. 2b,c). Corneal deposits in TGFBI-R124C homozygous mice appeared more frequently and developed earlier than in TGFBI-R124C heterozygous mice. Human patients bearing TGFBI-R124H homozygous mutations present with more severe corneal www.nature.com/scientificreports www.nature.com/scientificreports/ opacity than patients harbouring heterozygous mutations 18,21 . Taken together, our mice model presented very similar symptoms to human corneal dystrophy.

Genotypes Total number of eyes analyzed
Consistent with patterns of TGFBI corneal degeneration in humans 2 , the deposits in TGFBI-R124C homozygous mice showed eosinophilic staining (Fig. 3b). The deposits caused by the TGFBI-R124C mutation consist of amyloids 3 , which are positively stained by Congo-red 22 . However, positive Congo-red staining was not observed in TGFBI-R124C mice (Fig. 3f). Conversely, the deposits were stained faintly positive by Masson trichrome staining, revealing the presence of hyaline substance (Fig. 3d). These discrepancies may result from differences in degradation of TGFBIp between humans and mice. Corneal opacity manifests as LCD when arginine-124 is replaced with cysteine, as GCD1 when it is replaced with leucine, and as GCD2 when it is replaced with histidine. Korvatska et al. speculated that these phenotypic variations in the degradation process are due to this mutation 23 . There are differences in the fragment size of keratoepithelin in the cornea, for each of these diseases 23 .
Conversely, some cases of TGFBI-R124C mutation do not result in amyloid deposits 24,25 . The pathway of TGFBIp degradation is still unclear. Nonetheless, the absence of amyloid deposits in our study may be because TGFBIp in the TGFBI-R124C mice was processed similarly to how it is processed in GCD1 in humans. In humans, GCD1 is caused by the R124L mutation and does not involve amyloid deposits.
The observation periods that we used maybe another reason why we did not observe amyloid deposit formation in TGFBI-R124C mice. In mouse models of amyloid-associated diseases, the formation of amyloid plaques occurs via accumulation of mutant protein and requires a certain length of time. For example, Tg2576 mice, which are used in the well-known model of Alzheimer's disease, are generated by overexpressing a mutant form of amyloid precursor protein 26 ; however, 11-13 months are required to develop sufficient deposits of amyloid www.nature.com/scientificreports www.nature.com/scientificreports/ plaque 26,27 . Because we did not use artificial overexpression of mutant genes to establish our mouse model of TGFBI corneal dystrophy, our mice likely required more time to form amyloid plaques. Furthermore, because the lifespan of mice is shorter than that of humans, it is possible that there was not sufficient time to form amyloid plaques in our mice.
Under TEM, hyaline in GCD1 or GCD2 appears as unstructured club-shaped or trapezoidal deposits 3 . In TGFBI-R124C mice, hyaline appeared as an unstructured amorphous deposit, unlike the hyaline observed in human TGFBI corneal dystrophy. Although the reason for this discrepancy is unclear, deposits in R124H transgenic mice also presented an amorphous appearance, similar to that in our TGFBI-R124C mice 8 . Different mechanisms may drive hyaline formation in human and mouse corneas.
Corneal deposits in TGFBI corneal dystrophy may be caused by abnormal accumulation of TGFBIp resulting from mutated TGFBI 4 . Our results support this notion because sites with aggregations of TGFBIp also showed eosinophilic deposits in our study (Fig. 4a). TGFBIp was expressed at higher levels in the corneas of TGFBI-R124C mice than in those of wild-type mice (Fig. 4d). Nevertheless, the expression levels of Tgfbi mRNA did not differ significantly between wild-type mice and TGFBI-R124C mice (Fig. 4b). Therefore, accumulations of TGFBIp in the corneas of TGFBI-R124C mice may be caused by abnormal TGFBIp degradation and not enhanced the production of Tgfbi. Choi et al. previously reported that TGFBIp is usually digested via autophagy, but mutant TGFBIp inhibits the fusion of autophagosomes and lysosomes, which results in TGFBIp accumulation 28 . These mechanisms will be examined further in our future studies of TGFBI-R124C mice.
Patients with TGFBI corneal dystrophy often present with recurrent erosion of the corneal epithelium 29 . In LCD involving R124C mutations, regeneration of corneal epithelium after corneal erosion or phototherapeutic keratectomy requires an extended length of time 17,30 . Our results indicate that regeneration of the corneal epithelium after induction of corneal abrasion required more time in TGFBI-R124C mice than in wild-type mice. The domain containing R124 in fascilin1-1 (FAS1-1) is used for binding to integrin 2 . TGFBIp mediates cell adhesion and spreading via integrin 31 . A study using keratocytes has shown that TGFBI is involved in the proliferation, adhesion, and migration of corneal epithelium 32 . Mutated TGFBIp may have influenced the repair of epithelial wounds in our mice. Also, we assume that mutant TGFBIp affected the basement membrane function of epithelial cells because TGFBIp aggregated near the basement membrane in our mouse model. Taken together, our findings suggest some insights into the kinds of mechanisms underlying delayed wound healing in human corneal dystrophy.
In this study, we established the TGFBI-R124C mutant mouse model. The mice showed corneal opacity similar to that observed in human TGFBI corneal dystrophy. The TGFBI-R124C mutant mice might not be reliable as a model of LCD1, given that amyloid deposition was not detected in these mice. Nonetheless, our findings reveal that the corneal opacity was caused by the accumulation of mutant TGFBIp, similar to that observed in TGFBI corneal dystrophy in humans. Epithelial wound healing was delayed in TGFBI-R124C mice compared with wild-type mice. This delayed wound healing in mutant mice was similar to that observed in human patients with R124C LCD. Characterising corneal dystrophy via our TGFBI-R124C mouse model will help to uncover the mechanisms driving the pathogenesis of human TGFBI corneal dystrophies.

Material and Methods
Ethics statement. All animal care and experiments involving animals were conducted in full compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The study was approved by the Animal Research Committee of The University of Tokyo.
Establishment of TGFBI-R124C mouse model. In the 124th amino acid of mouse Tgfbi exon 4, arginine (CGC) was replaced with cysteine (TGT) by offset-nicking of CRISPR/Cas9 system 13 with using the two sgRNAs. The ssODN was designed as follows: TA TG AG AC CA TG GG AG TT GT GG GA TC GA CC AC CACA CA GC TG TA TA CA GA TG TACAGAA  AAGCTGAGGCCTGAGATGG AGGGACCCGGAAGCTTCACCATCTTTGC-3′ Mutant mice were generated as reported previously 13 . Briefly, 100 ng/μl Cas9 D10A mRNA, 10 ng/μl of each sgRNAs, and 200 ng/μl ssODN were injected into the zygotes of C57BL/6NCr mice. The two-cell embryos were transferred into the oviductal ampullae of 8-week-old female mice from the Institute of Cancer Research (ICR) mated with vasectomized ICR males.
After birth, genomic DNA was extracted from the tail tip using Nucleospin Kit (TaKaRa Bio Inc., Shiga, Japan). The obtained sample was amplified via the following primers using a KOD FX kit (Toyobo Co., Ltd., Osaka, Japan): R124C-Forward: 5′-CATCTGACTCCTGCGGTTCC-3′ R124C-Reverse: 5′-CTTTCACTTTCCTTGGGGCTG-3′ PCR products were then sequenced to ensure that the mutation had been introduced. RFLP assay. RFLP assay was performed to assess the genotype of mice used in this study. DNA was extracted from the mice, and the mutation site was amplified via PCR as described in the previous section. The obtained PCR products were treated with the restriction enzyme BsrGI (New England Biolab, Ipswich, MA) and electrophoresed on a 2% agarose gel.
Determination of corneal opacity. Eighteen eyes from 9 wild-type mice, 22 eyes from 11 heterozygous mice, and 142 eyes from 71 homozygous mice were examined using a slit lamp. The examination was performed under general anaesthesia induced via intramuscular injection of a xylazine (5 mg/ml) and ketamine (50 mg/ml) mixture. Examinations were conducted every month from the age of 12 weeks to 40 weeks. Mice were divided into those with bilateral or unilateral corneal opacity. The age of the mice and the presence of corneal opacity was recorded.
Histology. Eyes were removed from 22-week-old mice under general anaesthesia induced as mentioned previously. The eyes were fixed in 4% paraformaldehyde, after which the corneal buttons were excised, subjected to standard processing, embedded in paraffin, and sectioned at 5-7 μm. These sections were then deparaffinised using xylene, dehydrated using graded ethanol, and stained using H&E, Congo-red to detect the presence of amyloids and Masson's trichrome to detect hyaline, as per the manufacturers' instructions.
The deparaffinised sections were also evaluated using immunohistochemistry. The sections were washed with PBS, blocked with 5% skim milk for 30 min at room temperature, and incubated with a rabbit anti-TGFBI antibody (Proteintech Group Inc., Rosemont, IL;#10188-1-AP, 1:1000) for 60 min at room temperature. Adjacent sections were processed without primary antibody used as negative controls. Peroxidase-labelled anti-rabbit polyclonal antibody (Nichirei Biosciences Inc., Tokyo, Japan, # 714341) was used as the secondary antibody and the sections were incubated for 30 min at room temperature. Immunoreactivity was visualized using the Dako Liquid DAB + Substrate Chromogen System (Agilent Technologies, Inc., Santa Clara, CA, # K3467). Corneal wound healing model. Corneal epithelial debridement was performed using a previously reported method with several modifications 34 . A circle of filter paper (diameter 2 mm) was soaked with 70% EtOH and placed onto the central cornea for 30 s; then, the corneal epithelium was debrided with a #11 scalpel. The cornea was evaluated using fluorescein staining at 0, 24, 48, and 72 h after debridement. The wound area was calculated using ImageJ (https://imagej.nih.gov/ij/).

Statistical analysis.
Comparisons between groups were performed using the Wilcoxon rank-sum test. P < 0.05 was considered statistically significant. All statistical analyses were performed using JMP Pro 14 (SAS Institute, Cary, NC).

Data availability
The datasets analyzed during the current study available from the corresponding author on reasonable request.