Mutation of the EPHA2 Tyrosine-Kinase Domain Dysregulates Cell Pattern Formation and Cytoskeletal Gene Expression in the Lens

Genetic variations in ephrin type-A receptor 2 (EPHA2) have been associated with inherited and age-related forms of cataract in humans. Here, we have characterized the eye lens phenotype and transcript profile of germline Epha2 knock-in mutant mice homozygous for either a missense variant associated with age-related cataract in humans (Epha2-Q722) or a novel insertion-deletion mutation (Epha2-indel722) that were both located within the tyrosine-kinase domain of EPHA2. Confocal imaging of ex vivo lenses from Epha2-indel722 mice on a fluorescent reporter background revealed misalignment of epithelial-to-fiber cell meridional-rows at the lens equator and severe disturbance of Y-suture formation at the lens poles, whereas Epha2-Q722 lenses displayed mild disturbance of posterior sutures. Immunofluorescent labeling showed that EPHA2 was localized to radial columns of hexagonal fiber cell membranes in Epha2-Q722 lenses, whereas Epha2-indel722 lenses displayed disorganized radial cell columns and cytoplasmic retention of EPHA2. Immunoprecipitation/blotting studies indicated that EPHA2 formed strong complexes with Src kinase and was mostly serine phosphorylated in the lens. RNA sequencing analysis revealed differential expression of several cytoskeleton-associated genes in Epha2-mutant and Epha2-null lenses including shared downregulation of Lgsn and Clic5. Collectively, our data suggest that mutations within the tyrosine-kinase domain of EPHA2 result in lens cell patterning defects and dysregulated expression of several cytoskeleton-associated proteins.


Introduction
First identified as epithelial cell kinase (eck), ephrin type-A receptor 2 (EPHA2) belongs to the largest subfamily of receptor tyrosine kinases that were originally discovered in a human erythropoietin-producing-hepatoma (EPH) cell line [1,2]. EPH receptors and their membrane-bound EPH receptor interacting ligands, or ephrins, play key signaling roles in embryonic development including tissue patterning, neurogenesis and vasculogenesis, adult tissue physiology including bone homeostasis and insulin secretion along with various diseases including cancers and neurodegeneration [3][4][5]. The mammalian EPH/ephrin receptor subfamily comprises 14 receptors divided into type-A (EPHA1-8, 10) and type-B (EPHB1-5) that preferentially interact with ephrin type-A (EFNA1-5) and type-B (EFNB1-3) ligands, respectively, to elicit forward (receptor-driven) or reverse (ligand-driven) bidirectional signaling in neighboring cells. Like other receptor tyrosine kinases, EPHA2 shares a type-1 (single-pass) transmembrane glycoprotein topology with several functional domains including an extracellular (N-terminal) ligand binding domain and an intracellular (C-terminal) tyrosine kinase (TK) signaling domain and a sterile-alpha-motif (SAM) domain implicated in modulating kinase activity and receptor dimerization [6,7]. Canonical forward signaling by EFNA1-EPHA2 often promotes cell-cell repulsion accompanied by

Epha2-Mutant Mice and Lenses
Using CRISPR/Cas9 gene editing, we generated mice to model a missense variant (c. 2162G > A, rs116506614) in exon 13 of EPHA2 resulting in a conservative substitution of arginine-to-glutamine (p.Arg721Gln or p.R721Q) in the TK domain of EPHA2 that has been associated with age-related cortical cataract in humans [20]. Amino acid alignment revealed that the R721 codon (CGG) in human EPHA2 was phylogenetically conserved with the R722 codon (AGG) of mouse Epha2. A donor single-stranded oligonucleotide was designed to introduce a two base pair change (c.2164_2165delAG > CA) that converted R722 (AGG) to Q722 (CAG). Genomic PCR and Sanger sequencing confirmed the introduction of the correct A > C transversion and G > A transition in exon-13 of Epha2 in 9 of 47 (19%) founder (F0) mice. However, six of the nine F0 mice had acquired additional 'offtarget' sequence changes including insertions and/or deletions (data not shown) leaving three (6%) correctly targeted mice (one male homozygote and two female heterozygotes) that were crossed to generate homozygous F1 offspring. Sequencing confirmed germline transmission of the p.R722Q substitution ( Figure 1). Both heterozygous (p.R722Q) and homozygous (p.Q722Q) mutants were viable and fertile with no obvious signs of gross anatomical or behavioral abnormalities. In addition to the correctly targeted mutant mice, we found that one of the off-target F0 mice was heterozygous for a novel Epha2 mutation. At the DNA level this mutant allele comprised a single base insertion (c.2162_2163insT) causing a reading frame-shift that introduced three novel codons including the targeted 2-nucleotide change (GCA GGG TAT) followed by deletion of a 58-bp coding region before returning to the correct reading-frame 19 codons downstream (at codon R744). At the protein level, this compound mutation was predicted to cause an in-frame replacement of a 22-amino acid region (722-RGIASGMKYLANMNYVHRDLAA-743) from the TK domain of EPHA2 with three novel residues (722-AGY-724) resulting in an insertion-deletion mutation (p.R722_A743delinsAGY) that 'truncates' the wild type protein from 977 amino acids to 958 amino-acids ( Figure 1). As with correctly targeted mice, sequencing confirmed germline transmission of this insertiondeletion (indel722) allele ( Figure 1). Both heterozygous and homozygous F1 indel722 mutants were viable and fertile with no obvious anatomical or behavioral abnormalities.
Enucleated eyes from heterozygous and homozygous Epha2-mutant mice (Q722 and indel722) appeared indistinguishable from wild type and dissected lenses of heterozygous Epha2-mutants were grossly similar to wild type at 1-2 months of age (data not shown). By contrast, homozygous indel722 mutant lenses displayed translucent regions of internal refractive disturbance that were not observed in homozygous Q722 mutant or wild type lenses ( Figure 2). However, neither heterozygous nor homozygous Epha2-mutant lenses manifest cataract formation through 12 months of age (data not shown). All subsequent studies were performed on lenses from homozygous Epha2-mutants referred to as Epha2-Q722 and Epha2-indel722 throughout.
In addition to the correctly targeted mutant mice, we found that one of the off-target F0 mice was heterozygous for a novel Epha2 mutation. At the DNA level this mutant allele comprised a single base insertion (c.2162_2163insT) causing a reading frame-shift that introduced three novel codons including the targeted 2-nucleotide change (GCA GGG TAT) followed by deletion of a 58-bp coding region before returning to the correct readingframe 19 codons downstream (at codon R744). At the protein level, this compound mutation was predicted to cause an in-frame replacement of a 22-amino acid region (722-RGI-ASGMKYLANMNYVHRDLAA-743) from the TK domain of EPHA2 with three novel residues (722-AGY-724) resulting in an insertion-deletion mutation (p.R722_A743de-linsAGY) that 'truncates' the wild type protein from 977 amino acids to 958 amino-acids ( Figure 1). As with correctly targeted mice, sequencing confirmed germline transmission of this insertion-deletion (indel722) allele ( Figure 1). Both heterozygous and homozygous F1 indel722 mutants were viable and fertile with no obvious anatomical or behavioral abnormalities.
Enucleated eyes from heterozygous and homozygous Epha2-mutant mice (Q722 and indel722) appeared indistinguishable from wild type and dissected lenses of heterozygous Epha2-mutants were grossly similar to wild type at 1-2 months of age (data not shown). By contrast, homozygous indel722 mutant lenses displayed translucent regions of internal refractive disturbance that were not observed in homozygous Q722 mutant or wild type lenses ( Figure 2). However, neither heterozygous nor homozygous Epha2-mutant lenses manifest cataract formation through 12 months of age (data not shown). All subsequent studies were performed on lenses from homozygous Epha2-mutants referred to as Epha2-Q722 and Epha2-indel722 throughout.

Lens Cell Alignment and Suture Formation in Epha2-Mutant Mice
To visualize the global cellular organization of whole lenses ex vivo from Epha2-mutant mice, we generated Epha2-mutant and wild type littermates that constitutively express the red fluorescent protein tdT on cell membranes. First, we focused on the lens equator region where anterior epithelial cells begin terminal differentiation into highly-elongated fiber cells that form the refractive mass of the lens. In the wild type tdT lens (postnatal day 7, P7), equatorial imaging near the surface (10-20 µm depth) revealed the precise alignment of elongating, hexagonal-shaped fiber cells (in cross section) into meridional rows ( Figure 3A). Such meridional alignment occurs as elongating fiber cells start migrating their apical tips across the anterior epithelium toward the anterior pole and their basal tips across the posterior capsule toward the posterior pole. At intermediate equatorial depths (100-150 µm), wild type fiber cells were aligned parallel to the anterior-posterior polar (i.e., optical) axis of the lens ( Figure 3D). Imaging at greater equatorial depths (350-400 µm) in the wild type tdT lens revealed the 'fulcrum' (Figure 3G) where the apical tips of anterior epithelial cells pivot with the apical tips of elongating fiber cells [49]. Similar equatorial imaging of Epha2-Q722-tdT lenses revealed epithelial-to-fiber cell alignment including meridional rows and fulcrum formation along with pole-to-pole alignment of fiber cells resembling that found in wild type ( Figure 3B,E,H). By contrast, equatorial imaging of Epha2-indel722-tdT lenses revealed elongating fiber cells characterized by misaligned meridional rows, deviation from the polar axis particularly at the posterior pole, and less sharply defined fulcrum formation with abnormal epithelial cell gaps and clustering ( Figure 3C,F,I,J). We note that our attempts to image tdT-labelled lenses prior to P7 were hampered by their tendency to rupture during removal of the surrounding vasculature that is highly autofluorescent and interferes with imaging of these small lenses.
To visualize the global cellular organization of whole lenses ex vivo from Epha2-mutant mice, we generated Epha2-mutant and wild type littermates that constitutively express the red fluorescent protein tdT on cell membranes. First, we focused on the lens equator region where anterior epithelial cells begin terminal differentiation into highlyelongated fiber cells that form the refractive mass of the lens. In the wild type tdT lens (postnatal day 7, P7), equatorial imaging near the surface (10-20 µ m depth) revealed the precise alignment of elongating, hexagonal-shaped fiber cells (in cross section) into meridional rows ( Figure 3A). Such meridional alignment occurs as elongating fiber cells start migrating their apical tips across the anterior epithelium toward the anterior pole and their basal tips across the posterior capsule toward the posterior pole. At intermediate equatorial depths (100-150 µ m), wild type fiber cells were aligned parallel to the anteriorposterior polar (i.e., optical) axis of the lens ( Figure 3D). Imaging at greater equatorial depths (350-400 µ m) in the wild type tdT lens revealed the 'fulcrum' (Figure 3G) where the apical tips of anterior epithelial cells pivot with the apical tips of elongating fiber cells [49]. Similar equatorial imaging of Epha2-Q722-tdT lenses revealed epithelial-to-fiber cell alignment including meridional rows and fulcrum formation along with pole-to-pole alignment of fiber cells resembling that found in wild type ( Figure 3B,E,H). By contrast, equatorial imaging of Epha2-indel722-tdT lenses revealed elongating fiber cells characterized by misaligned meridional rows, deviation from the polar axis particularly at the posterior pole, and less sharply defined fulcrum formation with abnormal epithelial cell gaps and clustering ( Figure 3C,F,I,J). We note that our attempts to image tdT-labelled lenses prior to P7 were hampered by their tendency to rupture during removal of the surrounding vasculature that is highly autofluorescent and interferes with imaging of these small lenses. We next focused on the anterior and posterior pole regions of the lens where the tips of elongating fiber cells converge and overlap to form virtual, 3-branch (Y-shaped), suture lines centered on the optical axis [34,50,51]. In the wild type tdT lens (at P7 and P30), polar imaging revealed upright Y-shaped anterior suture lines and inverted Y-shaped posterior suture-lines that were rotationally spaced~120 • apart and centered on the optical axis  Figure 4A,D,G,J). Similarly, polar imaging of the Epha2-Q722-tdT lens, revealed anterior upright Y-sutures that resembled those of wild type ( Figure 4B,H). However, posterior sutures at P30 were more variable than wild type with a tendency to form longer and/or unequal suture branches ( Figure 4K). By contrast, polar imaging of the Epha2-indel722-tdT lens revealed variable disturbance of the anterior sutures often forming double-Y shapes ( Figure 4C,I), whereas the posterior sutures were severely disturbed with groups of locally aligned fiber cells veering away from the optical axis to form offset, irregular, cleft-like formations with no discernable pattern or orientation ( Figure 4F,L). Overall, these imaging data support a role for EPHA2 signaling in epithelial-to-fiber cell alignment (meridional row and fulcrum formation) at the lens equator and suture formation at the lens poles. We next focused on the anterior and posterior pole regions of the lens where the tips of elongating fiber cells converge and overlap to form virtual, 3-branch (Y-shaped), suture lines centered on the optical axis [34,50,51]. In the wild type tdT lens (at P7 and P30), polar imaging revealed upright Y-shaped anterior suture lines and inverted Y-shaped posterior suture-lines that were rotationally spaced ~120° apart and centered on the optical axis ( Figure 4A,D,G,J). Similarly, polar imaging of the Epha2-Q722-tdT lens, revealed anterior upright Y-sutures that resembled those of wild type ( Figure 4B,H). However, posterior sutures at P30 were more variable than wild type with a tendency to form longer and/or unequal suture branches ( Figure 4K). By contrast, polar imaging of the Epha2-indel722-tdT lens revealed variable disturbance of the anterior sutures often forming double-Y shapes ( Figure 4C,I), whereas the posterior sutures were severely disturbed with groups of locally aligned fiber cells veering away from the optical axis to form offset, irregular, cleft-like formations with no discernable pattern or orientation ( Figure 4F,L). Overall, these imaging data support a role for EPHA2 signaling in epithelial-to-fiber cell alignment (meridional row and fulcrum formation) at the lens equator and suture formation at the lens poles.

Expression and Distribution of EPHA2 Mutants in the Lens
To determine the effects of the Q722 and indel722 mutations on the expression and distribution of EPHA2 and other lens cell membrane proteins, we performed immunoblot analysis and immunofluorescence confocal microscopy. Immunoblotting revealed that the Q722 mutant was expressed at levels similar to wild type EPHA2 in the lens, whereas the indel722 mutant was present at reduced levels compared to the Q722 mutant and migrated with a molecular mass slightly lower (~2 kDa) than wild type EPHA2 ( Figure 5A). These data are consistent with the in-frame deletion of 19 amino acids from the TK domain of EPHA2 ( Figure 1) and suggest that the 'truncated' indel722 mutant protein and/or transcript may be relatively unstable compared to the full-length Q722 mutant in the lens. However, we cannot exclude reduced affinity and/or avidity of the EPHA2 antibody for the indel722 mutant versus the Q722 mutant on immunoblots.

Expression and Distribution of EPHA2 Mutants in the Lens
To determine the effects of the Q722 and indel722 mutations on the expression and distribution of EPHA2 and other lens cell membrane proteins, we performed immunoblot analysis and immunofluorescence confocal microscopy. Immunoblotting revealed that the Q722 mutant was expressed at levels similar to wild type EPHA2 in the lens, whereas the indel722 mutant was present at reduced levels compared to the Q722 mutant and migrated with a molecular mass slightly lower (~2 kDa) than wild type EPHA2 ( Figure 5A). These data are consistent with the in-frame deletion of 19 amino acids from the TK domain of EPHA2 ( Figure 1) and suggest that the 'truncated' indel722 mutant protein and/or transcript may be relatively unstable compared to the full-length Q722 mutant in the lens. However, we cannot exclude reduced affinity and/or avidity of the EPHA2 antibody for the indel722 mutant versus the Q722 mutant on immunoblots. In the wild type lens, immunofluorescent labeling revealed that EPHA2 was localized to fiber cell membranes highlighting the characteristic radial columns of flattened hexagonal cells of similar cross-sectional area serially aligned throughout the cortical region of the lens [50]-particularly along the short membrane faces ( Figure 5B). Similarly, in the Epha2-Q722 lens, immunolabeling revealed that EPHA2 was primarily localized to radial columns of hexagonal fiber cell membranes ( Figure 5C). By contrast, in the Epha2-indel722 lens, anti-EPHA2 labeling revealed that the radial columns of hexagonal fiber In the wild type lens, immunofluorescent labeling revealed that EPHA2 was localized to fiber cell membranes highlighting the characteristic radial columns of flattened hexagonal cells of similar cross-sectional area serially aligned throughout the cortical region of the lens [50]-particularly along the short membrane faces ( Figure 5B). Similarly, in the Epha2-Q722 lens, immunolabeling revealed that EPHA2 was primarily localized to radial columns of hexagonal fiber cell membranes ( Figure 5C). By contrast, in the Epha2-indel722 lens, anti-EPHA2 labeling revealed that the radial columns of hexagonal fiber cells were profoundly disorganized, particularly in the inner cortex ( Figure 5D). Instead of flattened hexagons, most Epha2-indel722 lens fiber cells exhibited an irregular cross-sectional size and shape including 4-or 5-sided cells that were randomly arranged throughout the cortex. Further, while EPHA2 was localized to fiber cell membranes there was reproducible evidence of diffuse membrane and perinuclear labeling suggesting that at least some of the indel722 mutant was retained in one or more cytoplasmic compartments-consistent with impaired targeting to the cell surface.
To visualize the effects of Epha2-mutants on the distribution of other lens cell membrane proteins we performed immunofluorescent labeling with antibodies to N-cadherin or cadherin-2 (CDH2), β-catenin or catenin β1 (CTNNB1), and connexin-46 or gap-junction alpha-3 protein (GJA3). In wild type lenses, both CDH2 and CTNNB1 labeling outlined the radial columns of flattened hexagonal fiber cells particularly along the narrow membrane faces (vertical plane), whereas GJA3 was localized to gap-junction plaques particularly on the broad faces of fiber cell membranes (horizontal plane) generating a 'ladder-like' appearance along the radial cell columns (Figure 6A,D,G). Immunolabeling of Epha2-Q722 lenses confirmed that CDH2, CTNNB1, and GJA3 were localized to fiber cell membranes aligned in radial columns similar to wild type ( Figure 6B,E,H). In Epha2-indel722 lenses, all three proteins were primarily localized to fiber cell membranes that were grossly disorganized lacking radial column alignment and, in places, deviating away from the optical plane of the lens section ( Figure 6C,F,I). Overall, these data suggest that while fiber cell membrane integrity was maintained in Epha2-mutant lenses, the radial column patterning of flattened, hexagonal, fiber cells was profoundly disturbed in Epha2-indel722 lenses. cells were profoundly disorganized, particularly in the inner cortex ( Figure 5D). Instead of flattened hexagons, most Epha2-indel722 lens fiber cells exhibited an irregular crosssectional size and shape including 4-or 5-sided cells that were randomly arranged throughout the cortex. Further, while EPHA2 was localized to fiber cell membranes there was reproducible evidence of diffuse membrane and perinuclear labeling suggesting that at least some of the indel722 mutant was retained in one or more cytoplasmic compartments-consistent with impaired targeting to the cell surface.
To visualize the effects of Epha2-mutants on the distribution of other lens cell membrane proteins we performed immunofluorescent labeling with antibodies to N-cadherin or cadherin-2 (CDH2), β-catenin or catenin β1 (CTNNB1), and connexin-46 or gap-junction alpha-3 protein (GJA3). In wild type lenses, both CDH2 and CTNNB1 labeling outlined the radial columns of flattened hexagonal fiber cells particularly along the narrow membrane faces (vertical plane), whereas GJA3 was localized to gap-junction plaques particularly on the broad faces of fiber cell membranes (horizontal plane) generating a 'ladder-like' appearance along the radial cell columns (Figure 6A,D,G). Immunolabeling of Epha2-Q722 lenses confirmed that CDH2, CTNNB1, and GJA3 were localized to fiber cell membranes aligned in radial columns similar to wild type ( Figure 6B,E,H). In Epha2-in-del722 lenses, all three proteins were primarily localized to fiber cell membranes that were grossly disorganized lacking radial column alignment and, in places, deviating away from the optical plane of the lens section ( Figure 6C,F,I). Overall, these data suggest that while fiber cell membrane integrity was maintained in Epha2-mutant lenses, the radial column patterning of flattened, hexagonal, fiber cells was profoundly disturbed in Epha2-indel722 lenses.

EPHA2 Complex Formation and Phosphorylation in the Lens
As EPHA2 has been reported to form complexes with CTNNB1 and CDH2 [52,53] and to elicit downstream signaling in the lens through the neuronal proto-oncogene tyrosine-protein kinase Src [32], we undertook immunoprecipitation and immunoblotting

EPHA2 Complex Formation and Phosphorylation in the Lens
As EPHA2 has been reported to form complexes with CTNNB1 and CDH2 [52,53] and to elicit downstream signaling in the lens through the neuronal proto-oncogene tyrosineprotein kinase Src [32], we undertook immunoprecipitation and immunoblotting techniques to detect lens EPHA2 interactions and phosphorylation status. Immunoprecipitation revealed that EPHA2 failed to form strong complexes with CDH2 and exhibited only trace complex formation with CTNNB1 relative to that between CTNNB1 and CDH2 ( Figure 7A,B). By contrast, EPHA2 formed strong complexes with Src kinase ( Figure 7C). Since the Q722 and indel722 mutations were located within the tyrosine-kinase domain of EPHA2 (amino acid residues 618-872), we attempted to compare the phosphorylation status of EPHA2 in wild type versus Epha2-mutant lenses. Phosphorylation of human EPHA2 has been reported at several tyrosine residues including Y588 (Y589 in mouse) and at several serine residues including S897 (S898 in mouse) in certain cancer cell lines [4,10,54]. EPHA2-S897/898 phosphorylation was readily detected by immunoblotting in both wild type and Epha2-Q722 lenses but not in Epha2-indel722 lenses-likely due to reduced levels of mutant protein in the latter ( Figure 7D). By contrast, immunoblotting failed to detect EPHA2-Y588/589 phosphorylation in wild-type and Epha2-mutant lenses (data not shown). However, we are cautious about interpreting such negative results as immunoblotting of whole lenses may lack the sensitivity required to detect tyrosine phosphorylation that may be restricted to a specific sub-region of the lens (e.g., anterior epithelium). Overall, these data suggest that EPHA2 forms complexes with Src and is abundantly serine phosphorylated in the lens.
techniques to detect lens EPHA2 interactions and phosphorylation status. Immunoprecipitation revealed that EPHA2 failed to form strong complexes with CDH2 and exhibited only trace complex formation with CTNNB1 relative to that between CTNNB1 and CDH2 ( Figure 7A,B). By contrast, EPHA2 formed strong complexes with Src kinase ( Figure 7C). Since the Q722 and indel722 mutations were located within the tyrosine-kinase domain of EPHA2 (amino acid residues 618-872), we attempted to compare the phosphorylation status of EPHA2 in wild type versus Epha2-mutant lenses. Phosphorylation of human EPHA2 has been reported at several tyrosine residues including Y588 (Y589 in mouse) and at several serine residues including S897 (S898 in mouse) in certain cancer cell lines [4,10,54]. EPHA2-S897/898 phosphorylation was readily detected by immunoblotting in both wild type and Epha2-Q722 lenses but not in Epha2-indel722 lenses-likely due to reduced levels of mutant protein in the latter ( Figure 7D). By contrast, immunoblotting failed to detect EPHA2-Y588/589 phosphorylation in wild-type and Epha2-mutant lenses (data not shown). However, we are cautious about interpreting such negative results as immunoblotting of whole lenses may lack the sensitivity required to detect tyrosine phosphorylation that may be restricted to a specific sub-region of the lens (e.g., anterior epithelium). Overall, these data suggest that EPHA2 forms complexes with Src and is abundantly serine phosphorylated in the lens.

Gene Expression Profiles of Epha2-Mutant and Epha2-Null Lenses
To determine the effects of the Epha2-Q722 and Epha2-indel722 mutations on lens gene expression, we performed RNA-seq analysis to compare global transcriptional changes in Epha2-mutant versus Epha2-null and wild type lenses at an early stage of phenotype development (P7). RNA-seq data files have been deposited in the Gene Expression

Gene Expression Profiles of Epha2-Mutant and Epha2-Null Lenses
To determine the effects of the Epha2-Q722 and Epha2-indel722 mutations on lens gene expression, we performed RNA-seq analysis to compare global transcriptional changes in Epha2-mutant versus Epha2-null and wild type lenses at an early stage of phenotype development (P7). RNA-seq data files have been deposited in the Gene Expression Omnibus (GEO) database under accession no. GSE181358. Biological triplicate samples were sequenced to a depth of >20 M reads and aligned to the mouse mm10 genome build with >98% alignment rate. Differential expression analysis using EdgeR (fold-change, FC ≥ 2, false discovery rate, FDR ≤ 0.05) was performed for each Epha2 genotype compared to wild-type (Figure 8). RNA-seq data showed high consistency between biological tripli-cates, with data from each Epha2 genotype clustering independent from other samples ( Figure S2A). We next analyzed the average fold-change (mutant/null vs. wild-type) within each Epha2 genotype of all significantly affected genes (FC ≥ 2, FDR ≤ 0.05) and found that they displayed a largely unique set of dysregulated genes compared to wild type ( Figure S2B). Overall, the total number of differentially expressed genes for each genotype was 86 in Epha2-Q722 lenses (48 up and 38 down, Table S3), 110 in Epha2-indel722 lenses (68 up and 42 down, Table S4), and 89 in Epha2-null lenses (68 up and 21 down, Table S5). We note that Epha2 was among the downregulated genes in Epha2-indel722 lenses ( Figure S2B, Table S4) suggesting that the in-frame insertion-deletion mutation (in exon-13) may decrease transcript stability. However, there were few other genes that displayed discordant expression between Epha2 genotypes ( Figure 8A and Figure S2B) and many genes up-or downregulated in one Epha2 genotype were generally not similarly affected in the other genotypes. Gene ontology (GO) enrichment analyses of each list of up or down regulated genes within an individual Epha2 genotype failed to yield any enriched categories or common pathways (data not shown). GO analysis of the combined upregulated gene set ( Figure 8A and Figure S2B) showed some enrichment for genes involved in extracellular matrix structure, chloride and gated channel activity, and calcium ion binding ( Figure S3), whereas the combined downregulated genes did not reveal any common pathways (data not shown). However, we note the shared downregulation of Lgsn and Clic5 ( Figure 8B, Tables S3-S5) both of which have been implicated in lens cytoskeletal differentiation [55,56].
Omnibus (GEO) database under accession no. GSE181358. Biological triplicate samples were sequenced to a depth of >20 M reads and aligned to the mouse mm10 genome build with >98% alignment rate. Differential expression analysis using EdgeR (fold-change, FC ≥ 2, false discovery rate, FDR ≤ 0.05) was performed for each Epha2 genotype compared to wild-type (Figure 8). RNA-seq data showed high consistency between biological triplicates, with data from each Epha2 genotype clustering independent from other samples ( Figure S2A). We next analyzed the average fold-change (mutant/null vs. wild-type) within each Epha2 genotype of all significantly affected genes (FC ≥ 2, FDR ≤ 0.05) and found that they displayed a largely unique set of dysregulated genes compared to wild type ( Figure S2B). Overall, the total number of differentially expressed genes for each genotype was 86 in Epha2-Q722 lenses (48 up and 38 down, Table S3), 110 in Epha2-indel722 lenses (68 up and 42 down, Table S4), and 89 in Epha2-null lenses (68 up and 21 down, Table S5). We note that Epha2 was among the downregulated genes in Epha2-indel722 lenses ( Figure S2B, Table S4) suggesting that the in-frame insertion-deletion mutation (in exon-13) may decrease transcript stability. However, there were few other genes that displayed discordant expression between Epha2 genotypes (Figures 8A and S2B) and many genes up-or downregulated in one Epha2 genotype were generally not similarly affected in the other genotypes. Gene ontology (GO) enrichment analyses of each list of up or down regulated genes within an individual Epha2 genotype failed to yield any enriched categories or common pathways (data not shown). GO analysis of the combined upregulated gene set (Figures 8A and S2B) showed some enrichment for genes involved in extracellular matrix structure, chloride and gated channel activity, and calcium ion binding ( Figure  S3), whereas the combined downregulated genes did not reveal any common pathways (data not shown). However, we note the shared downregulation of Lgsn and Clic5 ( Figure  8B, Tables S3-S5) both of which have been implicated in lens cytoskeletal differentiation [55,56].

Discussion
In this study, we have demonstrated that mice homozygous for mutations (Q722 or indel722) in the tyrosine kinase domain of EPHA2 underwent variable changes in lens cell organization and gene expression. Epha2-Q722 mice displayed clear lenses with mild defects in Y-suture branching at the posterior pole, whereas Epha2-indel722 mice presented clear lenses with translucent regions resulting from severe disturbance of (1) epithelial-tofiber cell alignment (meridional row and fulcrum formation) at the lens equator, (2) radial cell column formation throughout the lens cortex, and (3) Y-suture branching at the lens poles-similar to those described for Epha2-null lenses [35]. As meridional row and fulcrum formation were already disturbed at P7, it is conceivable that cell patterning defects may have arisen during earlier stages of lens development. EPHA2 was mainly localized to radial columns of hexagonal fiber cell membranes throughout the cortex of Epha2-Q722 lenses, whereas fiber cell columns were severely disorganized in Epha2-indel722 lenses along with cytoplasmic retention of EPHA2-consistent with failed targeting to the cell surface. EPHA2 formed strong immuno-complexes with Src kinase in vitro supporting a role for EPHA2/Src signaling during lens development [32]. However, we were unable to replicate strong EPHA2 complexes with CTNNB1 or CDH2 in the lens at wean-age (P21) similar to those reported in transfected (293T) cells and in the lens at an earlier stage of postnatal development (P10) [52,53]. EPHA2 was abundantly phosphorylated on serine-897/898 in wild type and Epha2-Q722 mutant lenses (P21), whereas EPHA2 tyrosine588/589 phosphorylation was not detected using similar immunoblot analysis of whole lenses. The relative abundance of serine-897/898 phosphorylation in the lens suggests that ephrinindependent or non-canonical EPHA2 signaling [57] may participate in lens cell migration. However, we cannot exclude a role for ephrin-dependent or canonical EPHA2 signaling since the hallmark tyrosine-588/589 phosphorylation may be restricted to specific subregions of the lens (e.g., specific lens epithelial cells) requiring more detailed studies. At the transcript level, several genes encoding cytoskeletal-associated proteins were differentially regulated including shared downregulation of Lgsn in both Epha2-mutant and Epha2-null lenses and Clic5 in Epha2-indel722 and Epha2-null lenses. Combined, our imaging and transcript data support a role for EPHA2 signaling-potentially via the cytoskeleton-in generating the precise cellular patterning underlying the refractive properties and optical quality of the crystalline lens.
Functional (over)expression studies in cultured (transfected) cell-lines have been used to predict diverse pathogenic mechanisms underlying EPHA2-related forms of human cataract. A non-coding risk allele for age-related cataract (rs6603883) located in a pairedbox-2 (PAX2) binding-site within the EPHA2 gene promoter suggested that it acts by down-regulating EPHA2 expression in cultured lens cells [58]. Several SAM domain mutations underlying early-onset cataract were reported to alter receptor stability, function and/or sub-cellular distribution [59][60][61]. Of three missense variants located within the TK domain of EPHA2 (amino acid residues 613-871), two (p.G668D, p.Q669H) have been associated with early-onset cataract and one (p.R721Q) with age-related cortical cataract in humans [20,62,63]. The p.G668D mutant has been associated with increased proteasome-mediated degradation, altered subcellular localization, and increased cell migration [63], whereas the p.R721Q mutant was associated with increased basal kinase activation in the absence of ligand, inhibition of clonal cell growth, and variable intracellular retention [20]. In our mouse model of the human EPHA2-p.R721Q variant (Epha2-Q722), homozygous expression of the equivalent variant protein at constitutive levels resulted in mild disturbance of the posterior Y-sutures but not in early-onset or age-related cataract (Figures 2 and 4). Similarly, homozygous expression of an in-frame TK domain mutant did not elicit cataract development in Epha2-indel722 lenses despite decreased levels and cytoplasmic retention of the mutant protein coupled with severe disorganization of lens fiber cells causing translucent regions of poor optical quality (Figure 2). While there was some mechanistic agreement between in vitro (overexpression) and in vivo (constitutive) expression studies of EPHA2 mutants (e.g., intracellular retention and altered cell growth/migration), we cannot account specifically for the lack of cataract penetrance in the Epha2-mutant mice reported here. Contributing factors include species differences in genetic background modifier effects, variable environmental risk factors (e.g., UV exposure in nocturnal mice versus diurnal humans), and morphological differences between the relatively small, almost spherical mouse lens with Y-suture branching versus the much larger, ellipsoidal human lens with more complex star-suture branching [51].
While we did not observe cataract formation in Epha2-mutant (Q722, indel722) or Epha2-null lenses [35], there were significant changes in lens gene expression at the transcript level between Epha2 genotypes as early as P7. Among the most upregulated genes (>4-fold) in both Epha2-Q722 and Epha2-indel722 mutant lenses were those for tubulin alpha 1C (TUBA1C) and alkaline ceramidase-2 (ACER2). TUBA1C serves as a prognostic biomarker for a variety of cancers [64] and ACER2 is a Golgi enzyme involved in regulating B1 integrin maturation and cell adhesion [65]. In Epha2-Q722 and Epha2-null lenses, the gene for steroidogenic acute regulatory protein-related lipid transfer (START) domaincontaining protein 9 (STARD9) was strongly upregulated, whereas that for doublecortin domain-containing 2a (DCDC2a) was strongly upregulated in Epha2-indel722 and Epha2null lenses. STARD9 functions as a centrosomal protein that regulates both interphase and mitotic spindle microtubules [66], whereas DCDC2a serves as a micro-tubule associated protein localized to hair cell kinocilia and supporting cell primary cilia that when mutated causes non-syndromic recessive deafness in humans [67]. The most consistently upregulated gene in both Epha2-mutant and Epha2-null lenses was that for WD-repeat and FYVE-domain-containing protein-1 (WDFY1), which serves as an adapter protein in tolllike receptor signaling [68]. Finally, the gene for dorsal inhibitory axon guidance protein (DRAXIN) was strongly upregulated in Epha2-indel722 lenses and that for actin, alpha 2, smooth muscle, aorta (ACTA2) was moderately upregulated in Epha2-null lenses. While ACTA2 serves as a marker for epithelial-mesenchymal transition during cataract formation [69] and several of the other upregulated genes share cytoskeletal-related or signaling functions, none have yet been associated with EPHA2 signaling or lens cell differentiation.
LGSN is a late marker for lens fiber cell terminal differentiation and has been shown to co-localize with actin and interact with the lens-specific intermediate filament protein, beaded filament structural protein-2 (BFSP2), also known as cytoskeletal protein 49 (CP49) or phakinin, suggesting that LGSN represents a recruited enzyme adapted to act as a cytoskeletal component or chaperone during remodeling of the lens cytoskeleton [55,70].
The most downregulated gene in Epha2-indel722 mutant lenses (<−1000-fold), and to a lesser extent in Epha2-null lenses (<−2-fold), was that for chloride intracellular channel 5 (CLIC5). Mutations in the human CLIC5 gene have been linked with progressive autosomal recessive, non-syndromic sensorineural hearing impairment with or without vestibular dysfunction and CLIC5 was found to be abundantly expressed in the fetal inner ear [71,72]. Similarly, in jitterbug (jbg) mice a spontaneous deletion mutation in Clic5 underlies hearing loss with vestibular and renal dysfunction and CLIC5 was localized to the base of hair cell stereocilia where it complexes with radixin, taperin, and myosin VI to stabilize cell membrane-actin cytoskeleton attachments [73]. Recently, CLIC5 been localized to cilia and/or centrosomes in the lens and Clic5-mutant (jtb) lenses were found to exhibit defective suture formation [56]. Further, EPHA2 has been shown to regulate Src/cortactin/F-actin complexes during epithelial-to-fiber cell morphogenesis (meridional row and fulcrum formation) at the lens equator [32]. Collectively, these observations point to a functional synergy between EPHA2 and several cytoskeletal proteins with LGSN and CLIC5 providing promising candidates for future studies of EPHA2 signaling in the lens.
In conclusion, our data suggest that EPHA2 signaling is required for lens cell pattern recognition and support a role for EPHA2 in cytoskeleton dynamics during lens cell differentiation.