Abstract
Central corneal thickness is a quantitative trait with important associations to human health. In a phenotype-driven approach studying corneal thickness of congenic derivatives of C57BLKS/J and SJL/J mice, the critical region for a quantitative trait locus influencing corneal thickness, Cctq1a, was delimited to a 10-gene interval. Exome sequencing, RNAseq, and studying independent mutations eliminated multiple candidate genes and confirmed one. Though the causative gene, Tyr, has no obvious direct function in the transparent cornea, studies with multiple alleles on matched genetic backgrounds, both in isolation and genetic complementation crosses, confirmed allelism of Tyr-Cctq1a; albino mice lacking Tyr function had thin corneas. Albino mice also had increased axial length. Because albinism exposes eyes to increased light, the effect of dark-rearing was tested and found to rescue central corneal thickness. In sum, the results point to an epiphenomenon; developmental light exposure interacts with genotype as an important determinate of adult corneal thickness.
Introduction
Central corneal thickness has important associations with ocular disease, but its natural determining factors remain predominately elusive. The cornea consists of three cellular layers (an outermost epithelium, middle stroma, and innermost endothelium) separated by two thinner basement membranes. The combined central thickness of these layers (central corneal thickness, CCT) increases rapidly through infancy and early childhood, reaches adult values in pre-teen ages, and remains relatively stable thereafter, with eventual modest age-related thinning1-3, 4. For largely unknown reasons, average CCT can vary by dozens of microns between ethnicities4,5. Thin CCT is associated with several corneal diseases, such as corneal dystrophy, brittle cornea syndrome, keratoconus, and cornea plana; diseases of connective tissue, such as Marfan syndrome, Ehlers-Danlos syndrome, Loeys-Dietz syndrome, and osteogenesis imperfecta; and at least two diseases in which the nature of, and/or reason for, the association is unclear, including myopia and primary open angle glaucoma6.
CCT is a highly heritable trait7,8, leading to many genetic studies. Variants influencing CCT have been identified from familial studies of Mendelian syndromes and GWAS of various large populations9-12. From these studies, some themes have begun to emerge. For example, some of the same CCT loci are identified in both multigenic and Mendelian disease studies10,13,14. It has also been common to identify variants related to collagen matrix integrity10,12,14,15. However, it is also clear that much remains unknown. Among the known associations, most known SNPs occur in non-coding regions, and the nearest genes typically have no obvious link to known structural components of the cornea15. It is also clear that many important genes remain to be discovered. Known SNPs give rise to a SNP-based heritability estimate of 42.5% and account for only 14.2% of the CCT variance9.
Here, an unique approach complementary to others is undertaken using inbred mouse strains to identify quantitative trait loci (QTL) that influence CCT16,17. Similar to humans, there is natural variation of CCT among inbred mouse strains18. Previous work using a quantitative approach with intercrosses between two such strains (C57BLKS/J [KS] mice with thin corneas and SJL/J [SJL] mice with thick corneas) identified the first CCT QTL, Central corneal thickness QTL 1 (Cctq1) on mouse chromosome 716. Here, Cctq1 was resolved into two closely linked regions, Cctq1a and Cctq1b, which each influence CCT. Through multiple genetic approaches, a mutation in the tyrosinase gene (Tyr) is identified as causative of the Cctq1a phenotype, which appears to influence CCT via an epiphenomenon dependent on developmental light exposure.
Results
Cctq1 contains two adjacent interacting QTL
The original 95% Bayesian credible interval of Cctq1 spanned a 38.3 cM region on chromosome 7 (34.1 cM – 72.4 cM)16. To reduce this interval, recombination mapping was used with 92 recombinant N4 intercross mice. However, the interval was originally recalcitrant to division— more than one interval conferred the increased CCT phenotype. This suggested that there was too much genetic heterogeneity for this trait at the N4 generation, potentially including the presence of more than one CCT-regulating gene within or near Cctq1. To address these possibilities, the original F2 dataset was subjected to additional evaluations while backcrossing of the congenic mice to the N10 generation was continued.
The original analysis of (KS Χ SJL) F2 mice was based upon a significance threshold determined empirically by stratified permutation testing with 1000 permutations16,19, and did not identify any loci that significantly interacted with Cctq1. Prompted by the recombination mapping, the dataset was re-examined by performing a pairwise scan of the markers on chromosome 7 using scantwo analysis with R/qtl19,20. Of the possible interactions, the markers that produced the highest LOD scores were D7Mit31 and rs13479535 (Figure 1A; Figure 1 – Source Data 1; full LOD score = 10.33, interactive LOD score = 5.05). D7Mit31 lies within the 95% Bayes credible interval of Cctq1; rs13479535 is 2 cM distal to the end of the interval at 74.3 cM. The putative linked loci were subsequently subjected to multiple regression analysis in which each locus and the interaction component were sequentially dropped from the 2-QTL model. This analysis indicated that both loci and their interaction had significant contributions to the model (Figure 1 – Source Data 2). The QTL at D7Mit31 was responsible for 31% of the phenotypic variability while the QTL at rs13479535 accounted for another 24% of the variability. These data indicate that both loci are true QTL. Cctq1 was thus resolved into two QTL, Cctq1a (95% Bayes credible interval: D7Mit318–D7Mit220, spanning 49.0 cM, peak at D7Mit31), and Cctq1b (95% Bayes credible interval: D7Mit105–rs13479545, spanning 74.3 cM, peak at rs13479535; Figure 1B).
To reduce genetic heterogeneity, N4 mice were further backcrossed onto the KS background to the N10 generation. Because the congenic interval was relatively large, a panel of six markers was used at each generation of backcrossing to keep the interval intact. At generation N10, congenic mice were intercrossed, recombination within the interval was allowed, and mice with all nine genetic combinations of Cctq1a and Cctq1b were phenotyped for CCT (Table 1; Table 1 – Source Data). Congenic control mice (KS.SJL-Cctq1aKS,Cctq1bKS) had a CCT indistinguishable from inbred KS mice (94.8± 2.4 μm vs. 94.5 ± 3.2 μm, respectively; one-way ANOVA with Tukey post-test; Table 1). Congenic mice with SJL genotypes at Cctq1a (i.e., KS.SJL-Cctq1aSJL) had significantly thinner corneas than inbred KS mice (87.9 ± 3.8 μm; n = 39; p < 0.001; Student’s two-tailed t-test) independent of the genotype at Cctq1b (n = 13 and p < 0.05 for each of three genotypes at Cctq1b; one-way ANOVA with Tukey post-test; Table 1; Supplementary Figure 1). The difference in thickness mediated by Cctq1a is predominantly due to the thickness of the stroma (Δ 9.6 μm, p < 0.001; Student’s two-tailed t-test), though there is a small (Δ 1.2 μm) and marginally significant (p = 0.022 Student’s two-tailed t-test) decrease in thickness of the epithelium as well. Congenic mice with KS genotypes at Cctq1a and SJL genotypes at Cctq1b (i.e., KS.SJL-Cctq1aKS,Cctq1bSJL) also had significantly thinner corneas (90.8 ± 2.6 μm; n = 13 mice; p < 0.05; one way ANOVA with Tukey post-test) than inbred KS mice (Table 1, Supplementary Figure 1). No other genetic combinations caused significant changes in CCT compared to inbred KS controls. As predicted from the analysis of the original F2 Dataset with R/qtl, these data with congenic mice independently support that both Cctq1a and Cctq1b are true QTL capable of altering the phenotypic variability of CCT on a uniform genetic background.
Mapping of Cctq1a using recombination mapping and sub-congenics
Because of its larger effect, initial efforts were focused on fine-scale mapping for Cctq1a. To identify the gene underlying Cctq1a, Cctq1a-recombinant N10F2 mice, with KS genotypes at Cctq1b, were used to narrow the critical region. From this recombination analysis, the gene underlying Cctq1a was deduced to be between SNP markers rs108403472 at 48.51 cM and rs6247100 at 50.26 cM. Simultaneously, sub-congenic mice were created by continued backcrossing of the KS.SJL-Cctq1aHET;Cctq1bKS N10 mice. Eyes of N10 congenic mice were overtly healthy, differing only in pigmentation between genotypes (Supplementary Figure 1). At N12, Cctq1a was physically reduced to a 15.9 cM region (KS.SJL-Cctq1a(15.9cM)) flanked by D7mit347 and D7mit321 and characterized by an association of SJL homozygosity with decreased CCT (87.6 ± 2.3 μm vs. 76.8 ± 2.4 μm, n = 15 mice per genotype, p < 0.001, Student’s two-tailed t-test). At N15, Cctq1a was physically reduced to a 9.9 cM region (KS.SJL-Cctq1a(9.9cM)) flanked by rs3672782 and D7mit321; again characterized by an association of SJL homozygosity with decreased CCT (91.6 ± 2.2 μm vs. 80.7 ± 2.9 μm, n = 7 mice per genotype, p < 0.001, Student’s two-tailed t-test). One N15F2 mouse harbored a recombination event within the minimal sub-congenic interval. The phenotype of this mouse indicated the gene underlying Cctq1a lies proximal to marker rs13479393 (Figure 2; Figure 2 – Source Data). Using this recombinant mouse in a progeny test, additional intercrossing to generation N15F7 confirmed the distal breakpoint proximal to marker rs13479393 at 49.65 cM, again characterized by an association of SJL homozygosity with decreased CCT (85.7 ± 1.6 μm vs. 94.4 ± 3.3 μm, n = 8–10 mice per genotype; p < 0.001, Student’s two-tailed t-test; Figure 3A; Figure 3 – Source data 1−3).
In sum, physical recombination mapping utilizing multiple generations of congenic and sub-congenic mice conclusively indicated that the gene underlying Cctq1a lies on chromosome 7 between markers rs108403472 at 48.51 cM and rs13479393 at 49.32 cM, a 0.81 cM region containing the entirety of eight RefSeq genes (Vmn2r78, Vmn2r79, Nox4, Tyr, Grm5, Ctsc, Rab38, and Tmem135), the 3’ portion of one gene (Vmn2r77), and the 5’ portion of one gene (Fzd4) (Figure 2).
Candidate identification and prioritization
To identify all the possible exonic variants within the Cctq1a critical region, whole exome sequencing was conducted on KS and SJL inbred mice. In the entire exome, 15,261 missense, frameshift, and splice-site mutations were found between KS and SJL (Supplementary File 1). There were six amino acid altering variants within the Cctq1a critical region, located within Vmn2r79 (A223T, L243M, T257I, I265V), Tyr (C103S), and Fzd4 (F27L). Vomeronasal receptor genes, such as Vmn2r79, have an increased rate of coding sequence variants 21 and the four altered residues between KS and SJL in Vmn2r79 are poorly conserved. The amino acid change in Tyr from cysteine to serine is the albinism-causing Tyrc allele conferred by the albino SJL strain. The Fzd4 amino acid variant is within the signal sequence of the protein; KS mice conferred the phenylalanine amino acid residue (the same residue as C57BL/6J mice) while SJL mice conferred the mammalian-conserved leucine residue.
Transcriptional profiling was additionally used to prioritize candidates. Using previously published microarray data comparing adult corneal RNA expression profiles in KS and SJL mice18, Nox4, Ctsc, Rab38, Tmem135, and Fzd4 were all present in the adult cornea; there was no evidence for adult corneal expression of Vmn2r77, Vmn2r78, Vmn2r79, Tyr, or Grm5. Of those expressed, Nox4, Ctsc, and Fzd4 showed differential expression between the two strains. Nox4 and Ctsc were both down-regulated 2.5-fold in SJL, while Fzd4 was up-regulated 1.6-fold in SJL.
Additionally, RNAseq was performed on corneas of 3-week-old KS.SJL-Cctq1a(15.9cM)SJL N12F3 mice. Analysis was focused on three comparisons: 1) KS.SJL-Cctq1a(15.9cM)SJL vs. KS (experimental, identifying genes with altered corneal expression in the congenic interval), 2) KS.SJL-Cctq1a(15.9cM)SJL vs. KS.SJL-Cctq1a(15.9cM)KS (experimental, also identifying genes with altered corneal expression in the congenic interval), and 3) KS.SJL-Cctq1a(15.9cM)KS vs. KS (control, identifying genes in the background of the congenic strain with altered corneal expression not associated with the congenic interval) (Supplementary File 2). In each comparison, genes were first filtered for those with a Q-value ≤ 0.001 and a FPKM ≥ 1 in at least one of the strains. Gene lists were subsequently compared to one another, identifying 87 genes consistently altered in both experimental comparisons but not in the control comparison (Supplementary File 2). Among these 87, only one was localized to the Cctq1a critical region, Ctsc, which was modestly (−0.8 log2fold) but consistently and significantly (p = 5 × 10−5; Q = 0.0009) down regulated in comparing the SJL allele to the KS allele. Web Gestalt22 was used for over-representation analysis comparing the list of 87 differentially expressed genes to a background list of all genes expressed in the cornea with a FPKM ≥ 1 in any strain. Results of the analysis indicate a strong signal for several collagen-related categories (fibrillar collagen trimer, abnormal cutaneous collagen fibril morphology, collagen biosynthesis and modifying enzymes, collagen degradation, etc.), extracellular-matrix-related categories (extracellular matrix component, ECM proteoglycans, degradation of the extracellular matrix, etc.) and ocular-related categories (decreased corneal stroma thickness, abnormal corneal epithelium morphology, abnormal cornea morphology, and abnormal eye morphology) (Supplementary Figure 2).
Functional tests of lead candidates
Based on candidate prioritization criteria, Fzd4 and Ctsc were initially considered the top candidates. To test the influence of Fzd4 on CCT, we tested a strain with a targeted mutation of Fzd4 (B6;129-Fzd4tm1Nat/J) on a segregating B6 and 129 background23. Fzd4tm1Nat homozygotes had a chocolate coat color not expected from either background. This observation is meaningful as the gene responsible for this phenotype, Rab38, is physically near Fzd4 and within the Cctq1 critical region, i.e., the strain is likely a double mutant for Fzd4 (genotype verified) and Rab38 (mutation unknown). However, there was no correlation between Fzd4 tm1Nat genotype and CCT (p = 0.819; one-way ANOVA comparing all three genotypes among littermates; Figure 3B; Supplementary Data 3–5). In genetic complementation crosses, Fzd4 tm1Nat mutation complemented the congenic interval ([Fzd4WT/Cctq1aSJL F1] vs. [Fzd4HET/Cctq1aSJL F1]; p = 0.528; Student’s two-tailed t-test; n = 4-6 per genotype; Figure 3C; Figure 3 – Source Data 1−3). The complementation cross also highlighted the coat color phenotype associated with the Fzd4tm1Nat mutation, with Fzd4WT/Cctq1aSJL F1 mice (TyrHET) having a black coat color with normally pigmented eyes and Fzd4HET/Cctq1aSJL F1 mice (TyrHET) having an unmistakable lightened (“light chocolate”) coat color and light brown irides instead of brown (Figure 3 – Source Data 5). Therefore, Fzd4 was ruled out as causative of the Cctq1a phenotype, and Rab38 further deprioritized as a candidate.
To test the influence of Ctsc on CCT, we imported a strain with a targeted mutation of Ctsc (B6.Cg-Ctsctm1Ley) on an N10 congenic B6 background24. Because the targeted mutation was generated on an albino 129 background, and Ctsc is physically linked to Tyr, Ctsctm1Ley homozygotes are albino, i.e., the strain is a double mutant for Ctsc and Tyr. This is meaningful because the SJL/J strain is albino and the Tyrc mutation is within the KS.SJL-Cctq1a congenic interval. Homozygotes had a statistically significant decrease in CCT compared to littermate controls (p < 0.002 for albino Ctsctm1Ley vs. pigmented CtscHET; p < 0.002 for albino Ctsctm1Ley vs. pigmented CtscWT; one-way ANOVA with Tukey post-test comparing all three genotypes among littermates; n=10–12 mice per genotype; Figure 3D; Figure 3 – Source Data 1−3). In genetic complementation crosses, Ctsctm1Ley failed to complement Cctq1aSJL (p < 0.001 for pigmented [Ctsctm1Ley/Cctq1aKS F1] vs. albino [Ctsctm1Ley/Cctq1aSJL F1]; Student’s two-tailed t-test; n = 10– 11 mice per genotype; Figure 3E; Figure 3 – Source Data 1−3). Therefore, Ctsc or Tyr were determined to be causative of the Cctq1a phenotype.
To differentiate Ctsc and Tyr as the causative mutation, independent alleles on a pure B6 background were assessed. To test Ctsc, four new mutations predicted to result in null protein were generated in B6 mice with CRISPR-Cas9 technology (Figure 3 – Source Data 4). To test Tyr, a well-known spontaneous mutation that was commercially available, Tyrc-2J, was analyzed25,26. There was no association between CtscKO genotype and CCT (p = 0.237; one-way ANOVA comparing all three genotypes among littermates; n = 9–17 mice per group; Figure 3F; Figure 3 – Source Data 1−4). In genetic complementation crosses with the congenic strain, the Ctsctm1Mga mutation (46bp-deletion in the coding sequence of exon 1 leading to no detectable CTSC protein; Figure 3 – Source Data 4) complemented the congenic phenotype (p = 0.696; Student’s two-tailed t-test; n = 8–12 mice per group; Figure 3G; Figure 3 – Source Data 1−4). In contrast, albino Tyrc-2J mice had decreased CCT relative to pigmented B6 (p < 0.001; Student’s two-tailed t-test; n = 9–10 mice per group; Figure 3H; Figure 3 – Source Data 1–3). In genetic complementation crosses with the congenic strain, the Tyrc-2J mutation failed to complement the congenic phenotype (p < 0.001 for pigmented [Cctq1aKS/Tyrc-2J F1] vs. albino [Cctq1aSJL/Tyrc-2J F1]; Student’s two-tailed t-test; n = 8 mice per genotype; Figure 3I; Figure 3 Source Data 1–3). Thus, only Tyr mutation was left as a feasible candidate for Cctq1a.
CRISPR-Cas9 technology was also used to generate new Tyr mutations on a C57BL/6J background (Figure 4 – Source Data 1). One allele was selected for propagation, Tyrtm4Mga, an albinism-causing 4 bp-deletion in the coding sequence of exon 1 that is predicted to cause a frameshift leading to a premature stop codon and RNA-mediated decay; i.e., a presumed null mutation. Tyrtm4Mga mice had a significantly thinner cornea than littermate controls (p < 0.002 for TyrWT vs. Tyrtm4Mga; p < 0.002 for TyrHET vs. Tyrtm4Mga; one-way ANOVA with Tukey post-test comparing all three genotypes among littermates; n = 6–10 mice per genotype; Figure 4; Figure 4 – Source Data 2−3). Thus, a presumed null allele of Tyr caused the same albinism and thinning of CCT as found for the c and c-2J alleles.
Mechanism of Tyr function on CCT
In considering the possible mechanism through which Tyr might influence CCT, three hypotheses were tested:
The first candidate mechanism centered on the role of DOPA, which is a cofactor in the oxidation of tyrosine by TYR, leading to melanin production27, and a substrate for tyrosine hydroxylase (TH), leading to dopamine synthesis. Dopamine is considered a key molecule in ocular growth28,29, and mice with a conditional knock-out of Th in the retina have previously been shown to have decreased CCT30. Rationalizing that some DOPA may normally escape from pigment producing cells to influence CCT via a TH-dependent mechanism, we tested whether providing supplemental DOPA in the drinking water of albino mice would rescue the decreased CCT associated with albinism. No statistically significant effect was observed (Figure 5; Figure 5 – Source Data 1−2), discounting the hypothesis that a DOPA deficit was rate-limiting for CCT determination in Tyr mutant mice.
The second candidate mechanism revolves around the possibility of a gene-environment interaction involving temperature and light. In this series of experiments, mice were reared in environmental control chambers in combinations of different temperatures and light cycles. Cohorts included pigmented B6, albino Tyrc-2J, and the temperature sensitive himalayan mutation (B6.Cg-Tyrc-h/J)31,32. At ambient temperatures, Tyrc-h homozygotes are only partially pigmented on the coolest parts of the body (such as the ears, nose, tail, and eyes) and CCT is intermediate between B6 and Tyrc-2J mice (Figure 6 – Source Data). For CCT of mice raised at ambient temperature, genetic complementation tests again confirmed the influence of Tyr-mediated albinism on CCT (p = 0.337 for [pigmented B6 x Tyrc-h F1] vs. [pigmented B6 x Tyrc-2J F1]; p < 0.002 for [pigmented B6 x Tyrc-h F1] vs. [albino Tyrc-h x Tyrc-2J F1]; and p < 0.002 for [pigmented B6 x Tyrc-2J F1] vs. [albino Tyrc-h x Tyrc-2J F1]; one-way ANOVA with Tukey post-test; n = 4–14 mice per group, (Figure 6 – Source Data)). At decreased temperatures Tyrc-h homozygotes can generate pigment more broadly, and at increased temperatures the albinism is accentuated (Figure 6).
Regarding temperature, comparison of B6.Tyrc-h mice, as well as B6 and B6.Tyrc-2J controls, raised at 10°C vs. 32°C in environmental control chambers with a standard light cycle, showed that CCT followed pigment status; the thin CCT of hypopigmented Tyrc-h mice reared at increased temperature (i.e., lower TYR activity) was rescued by rearing the Tyrc-h mice at decreased temperature (i.e., higher TYR activity; Tyrc-h at 32°C vs. Tyrc-h at 10°C; Δ 10.9 μm; p < 0.001; n = 13–15 mice per condition; one-way ANOVA with Sidak test; Figure 7; Figure 7 – Source Data 1−2). In testing the effect of temperature on corneal thickness in controls, there was a small (Δ 2.8 μm) and nominally significant (p = 0.03) decreased CCT in mice raised at 10ºC compared to mice raised at 32ºC when considering all C57BL/6J and Tyrc-2J mice across the experiment (n = 16 mice vs. n = 19 mice, Type II ANOVA) and there was no interaction between genotype and temperature. In sum, changing environmental temperature changed CCT of the Tyrc-h mice as predicted.
Regarding light, comparison of B6.Tyrc-h mice, as well as B6 and B6.Tyrc-2J controls, raised at 10°C vs. 32°C in environmental control chambers with dark-rearing of mice from conception to 10–15 weeks of age rescued the thin CCT phenotype associated with albinism (p < 0.001 for [Tyrc-h at 32°C standard light] vs. [Tyrc-h at 32°C dark rear]; p < 0.03 for [Tyrc-2J at 10°C standard light vs. Tyrc-2J at 10°C dark rear]; p < 0.001 for [Tyrc-2J at 32°C standard light vs. Tyrc-2J at 32°C dark rear]; n = 5–16 per condition; one-way ANOVA with Sidak test; Figure 7; Figure 7 – Source Data 1−2). Rearing pigmented B6 mice in constant light is known to lead to an increase in axial length33. OCT examinations of independent cohorts of C57BL/6J and Tyrc-2J mice for the purpose of measuring axial length show that Tyrc-2J mice have a 65.6 μm greater axial length on average compared to C57BL/6J mice (3.453 mm vs. 3.388 mm, respectively; p < 0.001, Student’s two-tailed t-test; n = 10 male and 10 female mice per strain; Figure 8; Figure 8 – Source Data 1−2).
Discussion
Using a phenotype-driven quantitative genetic analysis of CCT, physical mapping led to identification of a small critical region containing 10 genes, of which we ruled out three (Fzd4, Rab38, Ctsc) and found via an analysis of a Tyr allelic series (c, c-2J, c-h) that Tyr is the causative gene underlying the Cctq1a QTL. Tyr is by no means an unknown gene—it was in fact one of the first known mammalian genes whose initial discovery predates the word “gene”27,34,35. However, Tyr was a surprising gene to find linked to CCT. TYR is an oxidase whose only known biological role relates to melanin synthesis27,36.
Although there are small numbers of pigmented cells in the corneal limbus, the cornea is by and large not only non-pigmented, but transparent. Furthermore, previously published data indicate that Tyr is not even expressed in the adult cornea18, which is consistent with the current study which also found near-zero expression in the cornea at 3 weeks of age (RNA-Seq max expression ∼1.3 FPKM; Supplementary Data 7). Thus, there is no molecular rationale for proposing that TYR has a direct function in corneal cells. If not for the current experiments, there would also be sparse biological rationales for proposing that TYR might influence corneal anatomy through any mechanism. Regardless, the current agnostic QTL study led to the conclusion that Tyr contributes to the primary genetic influence on CCT, at least in the context of KS x SJL hybrids. Experiments using multiple alleles on matched genetic backgrounds, in isolation and genetic complementation crosses, conclusively confirmed allelism of Tyr-Cctq1a; albino mice lacking TYR function have thin corneas. Rationalizing that albinism would expose the developing eye to increased light, one of the mechanistic experiments performed here compared the effect of dark-rearing on the CCT of albino vs. pigmented C57BL/6J mice. The results showed that the thin CCT phenotype of albino B6.Tyrc-h mice raised at 32°C and B6.Tyrc-2J mice was rescued by dark-rearing. Thus, we are led to propose an epiphenomenon, whereby developmental light exposure interacts with genotype as an important determinate of corneal thickness (Figure 9).
All current CCT measurements were done with mice 10–15 weeks of age, which is slightly past the age at which the cornea of B6 mice reaches its final adult thickness (∼P55)37. Transcriptomic changes in the cornea related to Tyr genotype were detectable at 3 weeks of age. Thus, the timeframe for when Tyr can impact the cornea is presumably during anterior chamber development at some point preceding 3 weeks of age. Two mechanisms, which are both conjectural, might feasibly contribute to this early acting phenomenon. 1) Corneal development might be a component of refractive development. Emmetropization typically occurs in the first months following eyelid opening, with impacts from both the amount of light and its focus on the retina33,38. Thus, it is feasible that albinism could in effect cause blur (from light not being absorbed by melanin and reflecting within the eye), which induces relative myopia and CCT thinning with rearing in normal lights, but not in dark-rearing. 2) Corneal development might be influenced by central or corneal circadian outputs. Circadian outputs arise from both central and peripheral clocks, with the suprachiasmatic nucleus (SCN) being a central master pacemaker that receives light signals from retinal ganglion cells and subsequently coordinates phasing to peripheral tissues39,40. Notably, the SCN receives retinal input from the retinohypothalamic tract, which is known to be expanded in albinos41-43. Thus, one possibility is that factors increasing SCN output (such as albinism or cycling light conditions) might lead to decreased CCT and those decreasing the signaling (such as dark rearing) might lead to increased CCT. Refractive and circadian mechanisms could also be acting in an intertwined way44. Additional experiments are needed to distinguish these, and possibly other, mechanisms relevant to our current findings.
A leading candidate for contributing to the molecular mechanism causing thin CCT in albino mice was DOPA, which is a cofactor for TYR27,36, a substrate for TH leading to dopamine (reviewed in 28), and an endogenous ligand for the G-protein-coupled receptor GPR14345. DOPA can modulate refractive development27, 29, 40 and the circadian system46, as well development of multiple ocular tissues47,48. Notably, mice with a conditional knock-out of Th in the retina have decreased CCT30. The current experiments with Tyr mutant mice were not able to detect a role for DOPA in influencing CCT of these albino strains, though an important caveat to point out is that only a single dosing schedule for DOPA supplementation was currently tested.
It is unclear whether albinism or pigmentation influences CCT in humans, though our current experiments suggest this is likely. In humans, loss of function mutations in TYR cause oculocutaneous albinism type 1 (OCA 1)36,49. It’s unclear from the literature if humans with OCA 1 have decreased CCT; it may be difficult to ascertain because of relatively small patient populations and confounding variables such as eye rubbing50.
The current study has several implications with respect to mouse genetics and mouse models of disease: 1) The results highlight the potential for environmental influences on ocular development, which have been quantitated here for CCT, but may extend to other tissues as well. 2) The results indicate that albino mice should be tested for potentially being a naturally occurring model of myopia. 3) Because CCT is a complex trait, there is little reason to suspect that different inbred albino mouse strains would necessarily have the thinnest CCT in comparison to other inbred strains that are pigmented51, only that they would have thinner CCT as albino mice compared to pigmented mice within an inbred strain. However, for any experiment using such an albino strain, or cohorts in which an allele such as the common Tyrc allele is segregating, attention to the possibility of environmental influences is warranted. 4) Tyr has been linked with many ophthalmic traits in mice41-43,47,52-57; in some instances, a consideration of gene-environment interactions in the mechanism of these various models may be warranted. And finally, 5) our study uncovers a genetic peculiarity. Cctq1 was originally reported as a single QTL on chromosome 7, detected in an F2 intercross of KS and SJL inbred mice16. In studies of successive generations of congenic mice, N4F2 mice heterozygous for the Cctq1 alleles showed the original differential phenotype (over-dominant, increased CCT compared to littermate controls)16, whereas in N10–N15 intercrosses, mice homozygous for the SJL allele showed the differential phenotype (recessive, decreased CCT compared to littermate controls). As the sub-congenics were intercrossed and analyzed for CCT, KS.SJL-Cctq1aSJL;Cctq1bKS mice consistently had thinner corneas than KS.SJL-Cctq1aKS;Cctq1bKS mice of the same generation, but there were also fluctuations in absolute value related to generation and interval size. The likely explanation for these observations is that there was more than one CCT-modifying gene in the original Cctq1 interval, which was in fact found to be the case by the resolution of Cctq1 into Cctq1a and Cctq1b. This is consistent with the findings of human GWAS, indicating that there are likely hundreds of CCT-influencing genes dispersed throughout the genome, many of which will be physically close to one another. In mice, it is known that as a locus is narrowed using congenics, genes can be segregated away from nearby modifiers and the overall phenotype of the original QTL can be reduced, disappear, or reverse its apparent effect58-60. The current data seem to exemplify this phenomenon.
In summary, our phenotype-driven genetic study of CCT identified Tyr as a significant regulator of CCT in mice. The molecular findings of this study were unexpected. We propose that the results can be explained by an epiphenomenon whereby a gene:environment interaction; i.e., Tyr-mediated albinism allowing increased exposure of the eye to light has an important influence on corneal development.
Materials and Methods
Experimental animals
All animals were treated in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The majority of mice were housed and bred at the University of Iowa Research Animal Facility with approval for experimental protocols conferred by the Institutional Animal Care and Use Committee of the University of Iowa. Two cohorts of mice, the C57BL/6J (JAX Stock No. 000664) and B6.Cg-Tyrc-2J/J (JAX Stock No. 000058) used for in vivo axial length measurements (Figure 8) were purchased from The Jackson Laboratory at 10-weeks-old and subsequently housed at the University of California San Francisco until 12-weeks-old with approval for experimental protocols conferred by the Institutional Animal Care and use Committee at the University of California San Francisco. Mouse strains used in this study include: SJL/J (Stock No. 000686), C57BLKS/J (JAX Stock No. 000662), C57BL/6J (JAX Stock No. 000664), B6.Cg-Tyrc-2J/J (JAX Stock No. 000058), B6;129-Fzd4tm1Nat/J (JAX Stock No. 012823), B6.Cg-Tyrc-h/J (JAX Stock No. 000104; Imported from Dr. Brian Brooks at the NIH), B6.Cg-Ctsctm1Ley (Imported from Dr. Christine Pham at Washington University), KS.Cg-SJLCctq1, B6-Ctsctm1Mga, and B6-Tyrtm4Mga. All experiments included male and female mice.
Chromosome 7 QTL analysis
The chromosome 7 quantitative trait locus analysis was performed with R/qtl, using the two-dimensional genome-wide scan (scantwo). Significance thresholds were determined empirically by permutation testing, using 1000 permutations. The validity of a multiple QTL model was tested by performing a multiple regression analysis. Phenotypic variance was estimated and the full model was statistically compared to reduced models in which one QTL was dropped.
Constructing congenic mice
A ∼38.3cM genomic region (i.e., Cctq1) spanning from D7Mit318 (SSLP marker at 42.3 cM) to rs13479545 (SNP marker at 81.2 cM) was transferred from SJL/J mice (abbreviated throughout as SJL; thick cornea) onto the genetic background of C57BLKS/J mice (abbreviated throughout as KS; thin cornea) by reiterative backcrossing. Mice carrying the SJL alleles within the region (i.e., KS.SJL-Cctq1Het) were selected at each generation by using a panel of six markers that were tested and found to be polymorphic between KS and SJL mice. At the N10 generation, mice were intercrossed. At this point, Cctq1 was treated as a digenic locus, renamed to Cctq1a and Cctq1b (see Results and Figure 2). Cctq1a encompassed the region spanning from D7Mit318 (42.3 cM) to D7Mit220 (55.7 cM). Cctq1b spanned from D7Mit105 (70.3 cM) to rs13479545 (81.2 cM). All genotype combinations of Cctq1a and Cctq1b (9 possible combinations; i.e., homozygosity for KS alleles, heterozygous, or homozygosity for SJL alleles at each locus) were analyzed for their effect on CCT.
Sub-congenic mice harboring reduced Cctq1a intervals (KS alleles at Cctq1b) were also generated. At N12, Cctq1a was reduced to a 15.9 cM region spanning from D7mit347 to D7mit321. These N12 sub-congenic mice are referred to throughout as KS.SJL-Cctq1a(15.9cM). At N15, Cctq1a was reduced to a 9.9 cM region spanning from rs3672782 to D7mit321. These sub-congenic N15 mice are referred to throughout as KS.SJL-Cctq1a(9.9cM). Sub-congenic mice were intercrossed, and all genotypes were assessed for the CCT phenotype. The KS.SJL-Cctq1 line has been sperm cryopreserved.
CCT phenotyping
All measurements were recorded from adult mice. Cohorts were determined based on either 1) mice received from The Jackson Laboratory, or 2) genotypes of mice born in the first litters of a genetic cross. Mice were injected with a standard mixture of ketamine/xylazine (intraperitoneal injection of 100 mg ketamine + 10 mg xylazine / kg body weight). During induction of anesthesia, mice were provided supplemental indirect warmth by a heating pad. Immediately following anesthesia, eyes were hydrated with balanced salt solution (BSS) and corneal images were obtained with a Bioptigen optical coherence tomographer (SD-OCT; Bioptigen, Inc., USA). A 12mm telecentric bore was used to image the anterior segment of each eye. The bore was positioned such that the pupil of the eye was centered in the volume intensity projection. Scan parameters were as follows: radial volume scans 2.0 mm in diameter, 1000 A-scans/B-scan, 100 B-scans/volume, 1 frame/B-scan, and 1 volume. Central corneal thickness (CCT) was measured for each eye using vertical angle-locked B-scan calipers by a technician masked to genotype and/or treatment. Mice were included in the analysis if the difference between the right and left eyes was less than 7 μm and if both eyes were free from opacity. The average CCT with standard deviation for each genotype was statistically compared using Student’s two-tailed t-test for comparison of two cohorts or one-way ANOVA with a Tukey post-test for comparison of three or more cohorts.
Recombination mapping
For genetic mapping of the gene underlying Cctq1a, additional polymorphic markers were identified and tiled into the region. Intercrosses of N10 mice were continued and mice with informative recombination events were analyzed for the CCT phenotype. Based on the allelic effects of the QTL on the CCT phenotype (see Results), the genomic boundaries of the QTL (and hence, the region of the underlying gene) were deduced by comparing the phenotype of the recombinant mice with the location of the recombination event within the critical interval. The following is a complete list of all the polymorphic markers used for genotyping (listed in order from centromeric to telomeric): D7Mit318, rs13479346, rs13479362, D7Mit347, D7Mit62, rs6271685, rs108403472, Cctq1a-STR5, D7Mit31, rs3672782, rs32438580, rs3663323, rs13479392, rs13479393, rs6247100, rs13479395, D7Mit301, D7Mit321, D7Mit220, D7Mit238, D7Mit105, rs13479535, rs13479536, and rs13479545. All MIT marker primer sequences are available on the Mouse Genome Informatics website61. Cctq1a-STR5 is a short tandem repeat polymorphism with genotyping primers designed in-house: 1) GCAAATTAATCAGCACATTTC and 2) CATTTCCTCTAGGTGATTGC.
Exome sequence analysis
High quality genomic DNA was harvested from KS and SJL spleen tissue using a Qiagen DNeasy Blood and Tissue kit following the manufacturer’s instructions; an RNA digestion step was included. DNA samples were sent to BGI Americas for sequencing and passed their quality control standards. Libraries were constructed with an Agilent SureSelect 50Mb Mouse Exome Capture Kit and were sequenced with 50X coverage using an Illumina HiSeq2000. Standard bioinformatics analysis was conducted in which the data were filtered (by removing adaptor contamination and low-quality reads from raw reads), aligned, and SNPs were called using the GRChm38 mouse reference genome. The resulting VCF file was processed at the University of Iowa to annotate the consequence for each variant using a locally-developed pipeline.
RNASeq analysis
N12F3 KS.SJL-Cctq1a(15.9cM)KS, KS.SJL-Cctq1a(15.9cM)SJL, and KS inbred mice were euthanized at three weeks of age by cervical dislocation. Immediately upon death, mice were enucleated, and the eyes were placed in RNase free dishes containing RNAlater RNA stabilization reagent. Corneas from mice were dissected in RNAlater and pooled to make one sample (6 corneas per sample); three samples were collected per genotype. Cornea samples were either stored at -80°C in RNAlater or processed immediately. Corneas were transferred from RNAlater to 0.7 mL of lysis/binding buffer from the mirVana miRNA isolation kit (Ambion) and homogenized for 1 minute using a tissue tearer. The homogenate was then passed through a QIAshredder column (Qiagen) and the lysate was collected. For the remainder of the procedure, the samples were processed using the mirVana kit for total RNA according to the manufacturer’s instructions. The quality and concentration of the RNA was analyzed using a NanoDrop 2000 and the Agilent Model 2100 Bioanalyzer. All samples had RNA integrity numbers of 9.5 or greater, indicating high quality RNA with little degradation of the samples. Samples were barcoded and stranded libraries were prepared by the Genomics Division of the Iowa Institute of Human Genetics. The nine libraries were pooled together, split into two equal parts, and run on two lanes of an Illumina HiSeq to obtain 100 base pair, paired-end sequence reads.
Reads were mapped to the mm10 mouse genome build using Tophat2 (ver 2.0.1162). The ‘-r’ parameter was set to 135, and the ‘--no-coverage-search’ option was used. Transcript abundance was quantified using Cufflinks (ver. 2.163) for RefSeq transcript models from the Illumina iGenomes mm10 package. Ribosomal RNA and mitochondrial gene loci, obtained from UCSC Genome Table Viewer, were masked from the Cufflinks analysis, and the ‘--max-bundle-frags’ parameter was set to 20000000. Differential expression between genotypes was performed using Cuffdiff (ver. 2.1.164). Genes identified from each comparison were subsequently filtered for only those with a q-value ≤ 0.001 and a mean FPKM (Fragments Per Kilobase of transcript per Million mapped reads) value ≥ 1 in at least one of the strains. Functional enrichment analysis was performed with WebGestalt [32, 33] with analysis parameters detailed in Supplementary File 2 – Figure Supplement 1.
Constructing Ctsc null mice
B6-CtscKO mice were generated by the Genome Editing Facility at The University of Iowa on a pure C57BL/6J (JAX Stock No. 000664) background by targeting Ctsc Exon 1 with CRISPR/Cas9 using guide sequence: CGTGCGCTCCGACACTCCTGCC. Founders were crossed with C57BL/6J mice and offspring analyzed for germline transmission of Ctsc mutations. From four founders, we observed transmission of four separate Ctsc mutations, all predicted to be null based on Sanger sequencing results. Four Ctsc mutations were propagated in separate intercross mouse lines and were validated as null mutations by Western Blot using an antibody against CTSC (Catalog #AF1034; R&D Systems, Inc.; Minneapolis, MN; Supplementary Data 10). One of these mutations, Ctsctm1Mga, is a 46 bp exon 1 coding sequence deletion that was used for additional downstream analysis.
Constructing Tyr null mice
To generate mice with Tyr null mutations on a pure C57BL/6J (JAX Stock No. 000664) background, the Genome Editing Facility at the University of Iowa targeted Tyr Exon 1 with CRISPR/Cas9 using two guide sequences simultaneously: 1) CCATGGATGGGTGATGGGAG and 2) TTCAAAGGGGTGGATGACCG. Founders were crossed with C57BL/6J mice and offspring analyzed by Sanger sequencing for germline transmission of Tyr mutations. This experiment generated more alleles than we could reasonably work with (15 unique mutations identified via sequencing; Supplementary Data 12). For each unique mutation, we set up a complementation cross with B6.Tyrc-2J and screened progeny coat color for alleles conferring novel function. F1 progeny were screened for 14 alleles; the mice harboring the remaining allele did not produce F1 progeny. All 14 alleles produced a standard albino coat color in trans with the Tyrc-2J mutation, indicating failure to complement. Going forward, we chose to complete additional studies for one allele, Tyrtm4Mga, which is a 4 bp deletion in the coding sequence of exon 1, causes a frameshift and leads to a premature stop codon predicted to cause RNA-mediated decay (i.e., a presumed null mutation). Accordingly, homozygotes of this strain are albino.
L-DOPA supplementation studies
Breeder cages of C57BL/6J, B6.Cg-Tyrc-2J/J, and B6.Cg-Tyrc-h/J were provided with water bottles containing water only (control) or water with 200mg/L of L-DOPA, with 30mg/L of benserazide to minimize the conversion of L-DOPA to dopamine in the peripheral nervous system. Fresh water bottles were prepared and supplied every Monday, Wednesday, and Friday. Litters were weaned into cages where they continued to receive the same treatment of supplemented water until 13–16 weeks old. Cohorts were determined by the number of mice born under each treatment condition.
Environmental chamber
Breeder cages of C57BL/6J, B6.Cg-Tyrc-2J/J, and B6.Cg-Tyrc-h/J were housed in rodent environmental control chambers (Powers Scientific) either at 10°C or 32°C and with either a standard light cycle (12 h on/12 h off) or dark-rearing. Resulting pups were born and group-housed with age-matched mice regardless of genotype at the altered temperature and light cycle until adulthood, when CCT was analyzed by OCT at 10–15 weeks old. Cohorts were determined by the number of mice born under each treatment condition.
Axial length phenotyping
Envisu R4300 spectral-domain optical coherence tomography (SD-OCT, Leica/Bioptigen Inc., Research Triangle Park, NC, USA) was employed to measure the ocular axial length in adult (12-week-old) mice65. Twenty mice of each genotype were ordered from The Jackson Laboratory and all forty mice were phenotyped. Mice were anesthetized with ketamine/xylazine (100 mg/kg and 5mg/kg, respectively; intraperitoneal) and their eyes dilated before placing the animal in a cylindrical holder. The eye was hydrated with Genteal (Alcon, Fort Worth, TX, USA) and positioned in front of the OCT light source. Correct alignment of the eye was achieved by placing the Purkinje image in the center of the pupil. The images were acquired in rectangular volume and radial volume scans. The axial length was calculated by measuring the distance from the anterior corneal surface to the RPE/choroid interface for both the left and right eyes of a given mouse. Measurements from the left and right eye of each mouse were averaged to give a single measurement per animal. Measurements from all eyes were included in the analysis. To minimize the possible effect of body weight on ocular size, we ensured that body weight of littermates was within a narrow range in each of the comparative groups. The average axial length with standard deviation for each genotype was statistically compared using Student’s two-tailed t-test.
Statistics
Power calculations were performed using the Piface Java applet by Russell V. Lenth66 with an analysis for a two-sample t-test with an alpha = 0.05 or a one-way ANOVA with an alpha = 0.05. Assumptions entered for CCT phenotype for a two-sample t-test are sigma1/sigma2 = 2.5 (the standard deviation in μm of CCT measured previously in naïve adult mice), true difference of means = 8 (the measured CCT difference in μm between KS.SJL-Cctq1aSJL vs. KS.SJL-Cctq1aWT), and sample size = 4 mice per group, which results in a power of 96.18% to detect an 8 μm difference (7.8% difference in CCT on a B6 background) between groups. Assumptions entered for CCT phenotype for a one-way ANOVA testing three genotypes are SD[WT], SD[HET], SD[MUT] each = 2.5 (the standard deviation in μm of CCT measured previously in naïve adult mice), SD[Residual] = 5.2 (the standard deviation in μm of CCT measured previously between groups with a CCT difference of 10 μm), sample size = 5 mice per group, and a Tukey/HSD post-hoc analysis, which results in a power of 97.84% to detect an 8 μm difference between groups. Assumptions entered for the axial length phenotype for a two-sample t-test are sigma1/sigma2 = 0.25 (the standard deviation in mm of axial length measured previously in naïve adult mice), true difference of means = 0.33 (10% difference in axial length in mm), and sample size of 20 mice per group, which results in a power of 98.24% to detect a 0.33 mm difference between groups. If additional mice resulted from genetic crosses, phenotypes were collected and included in data analyses. A multiple regression analysis, in which each locus and the interaction component were sequentially dropped from the 2-QTL model, was used to analyze the presence of two interacting loci on chromosome 7. Student’s two-tailed t-test was used to evaluate the difference between two independent genotypes for CCT (t-values and degrees of freedom for each comparison are listed in Figure 3 – Source Data 3) and axial length (t-value = 5.415; degrees of freedom = 38). A one-way ANOVA with a Tukey post-test was used to evaluate the difference between three or more independent genotypes for CCT (f-values and degrees of freedom for each comparison are listed in Figure 3 – Source Data 3). A one-way ANOVA with Sidak test was used to evaluate four comparisons to determine the effect of two environmental conditions on CCT in B6.cg-Tyrc-h and B6.cg-Tyrc-2J mice (f-value and degrees of freedom are listed in Figure 7 – Source Data 2) because previous experiments identified significant differences between Tyr genotypes. No outliers were excluded.
AUTHOR CONTRIBUTIONS
K.J.M., D.R.L., and M.G.A. conceived and designed the experiments. K.J.M., D.R.L., C.J.V., A.H.B., L.M.D., N.P., H.E.M., M.N.A., and M.G.A. performed the experimental work. K.J.M., D.R.L., and M.G.A. performed experimental data analysis and interpreted the results with contributions from all authors. S.S.W. analyzed the RNA-seq data supervised by T.E.S. W.J.P. constructed the transgenic mice. S.K. performed the axial length experiment supervised by K.S.N. K.W. verified and performed statistical analyses. K.J.M., D.R.L. and M.G.A. wrote the manuscript with input from all authors. M.G.A. supervised the study and provided funding.
FIGURE SUPPLEMENT LEGEND
SOURCE DATA FILES
Figure 1 – Source Data 1. Table formatted for analysis with R/qtl containing genotypes and central corneal thickness measurements of KS x SJL F2 mice.
Figure 1 – Source Data 2. Multiple regression analysis for Cctq1a and Cctq1b comparing the full model to a model in which the indicated QTL or interaction is omitted.
Table 1 – Source Data. Table containing Mrdq1a and Mrdq1b interval genotypes for individual mice along with corresponding central corneal thickness measurements in KS.SJL-Cctq1 N10F2 mice.
Figure 2 – Source Data. Table containing genotypes from N10F2+ mice harboring a recombination event within the Mrdq1a interval.
Figure 3 – Source Data 1. Table containing central corneal thickness, corneal epithelium thickness, and corneal stroma thickness measurements for all individual mice. Genetic crosses are separated into individual tabs.
Figure 3 – Source Data 2. Graphs of corneal epithelium and stroma thickness for the same mice shown in figure 3.
Figure 3 – Source Data 3. Table containing CCT results and statistics for all experimental iterations.
Figure 3 – Source Data 4. CRISPR targeting strategy, sanger sequencing of novel mutations, and western blot for Ctsc.
Figure 3 – Source Data 5. Photographs and slit-lamp images of Fzd4tm1Nat/Cctq1a F1 mice.
Figure 4 – Source Data 1. CRISPR targeting strategy and sanger sequencing of novel mutations for Tyr.
Figure 4 – Source Data 2. Table containing central corneal thickness, corneal epithelium thickness, and corneal stroma thickness measurements for all individual mice.
Figure 4 – Source Data 3. Graphs of corneal epithelium and stroma thickness for the same mice shown in figure 4.
Figure 4 – Source Data 4. Table containing CCT results and statistics for all experimental iterations.
Figure 5 – Source Data 1. Table containing central corneal thickness measurements and treatment type for all individual mice. Data from C57BL/6J, B6.Tyrc-h, and B6.Tyrc-2J are presented in individual tabs.
Figure 5 – Source Data 2. Table containing CCT results and statistics for all experimental iterations.
Figure 6 – Source Data. Table containing CCT results and statistics for all experimental iterations.
Figure 7 – Source Data 1. Table containing central corneal thickness measurements for all individual mice, rearing temperature, rearing light cycle, and strain.
Figure 7 – Source Data 2. Table containing CCT results and statistics for all experimental iterations.
Figure 8 – Source Data 1. Table containing axial length measurements for all individual mice.
Figure 8 – Source Data 1. Table containing CCT results and statistics for all experimental iterations.
SUPPLEMENTARY FILES
Supplementary File 1. Exome results for SJL/J and C57BLKS/J.
Supplementary File 2. Results of cornea RNASeq from 3-week-old C57BLKS/J and KS.SJL-Cctq1a.
ACKNOWLEDGEMENTS
This work was supported in part by Merit Review Award (101RX001481) from the U.S. Departments of Veterans Affairs RR&D Service to MGA and by internal faculty development funds to MGA. DRL was supported in part by a NIH/NEI training grant, EY021436. AHB was supported in part by a NIH/NIDDK training grant, DK112751. Equipment maintenance was supported in part by the University of Iowa Multidisciplinary Investigations in Visual Science service cores (a resource supported by a P30 grant from the NIH/NEI to the University of Iowa, EY025580). We acknowledge use of the University of Iowa Genome Editing Core Facility (a core resource supported in part by grants from the NIH and from the Roy J. and Lucille A. Carver College of Medicine) directed by William Paradee, PhD, and we thank Norma Sinclair, Patricia Yarolem and Joanne Schwarting for their technical expertise in generating transgenic mice. The authors thank Karl Broman for analyzing chromosome 7 of our original F2 dataset with us. The contents do not represent the views of the U.S. Department of Veterans Affairs or the U.S. Government.
The authors declare no competing interests.