Mice were maintained in accordance with an approved Institutional Animal Care and Use Committee (IACUC) protocols (#21-010 and #24-002) and the Guide for the Care and Use of Laboratory Animals by the National Institutes of Health (NIH). Generation of Epha2^–/– and Efna5 ^–/– mice was previously described (28, 29). All mice were maintained in C57BL/6J backgrounds with wild-type beaded filament structural protein 2 (Bfsp2; CP49) genotypes, as Bfsp2^–/– mice exhibit disruptions in the lens fiber cytoskeleton, gap junctions, and ionic homeostasis (3, 30). Bfsp2 mutations occur spontaneously across various inbred strains, including 129/SvJ, 129/OLa, and FVB/N mice, and have been linked to cataracts in both humans and mice (31–34). Therefore, the wild-type Bfsp2 gene was validated in these mice prior to investigating cataract mechanisms. Male and female mice ranging from 6-9 weeks (young adult; Y) and 7 months (middle-aged; M) were used for these studies. Littermates were used for comparison between wild-type and KO samples.

RNA was isolated from epithelium and fiber cell mass fractions using our previous protocol (35). Briefly, samples were collected from at least three different mice of each genotype and each age for RNA isolation. For Efna5^–/– mice, the lenses can present with obvious anterior cataracts. Efna5^–/– mice were excluded if the lenses had obvious cellular defects to prevent detection of gene expression changes that are downstream of EMT in the KO lens epithelial cells (3, 5). Briefly, lenses were carefully dissected from freshly enucleated eyes. To isolate lens epithelium, the collagenous basement membrane around the lens, the capsule, was gently peeled from the lens posterior. The epithelial cells are well-adhered to the lens capsule (3, 35, 36). The remaining fiber mass was not further separated as the organelle-free nuclear fibers are presumed to contain little to no mRNA. Autologous pairs of capsules with attached epithelial cells or fiber cell bulk masses were pooled and homogenized into 400 µL of cold TRIzol (Invitrogen; Waltham, MA, USA; Cat # 15596026). Samples were incubated for 30 minutes at room temperature. For phase separation, 200 µL of chloroform (Alfa Aesar; Ward Hill, MA, USA; Cat # 22920) was added and the samples were shaken vigorously for 15 seconds by hand. Samples were incubated at room temperature for 15 minutes and centrifuged at 14,000g for 15 minutes at 4°C. The aqueous phase was transferred to RNase-free tubes (USA Scientific; Ocala, FL; Cat # 1615-5500), and 1 equivalent volume of 200 proof ethanol (Fisher Scientific; Waltham, MA, USA; Cat # BP2818500) was added. Samples were transferred to ‘RNA Clean and Concentrator-5’ kit columns (Zymo Research; Tustin, CA, USA; Cat # ZR1013), and RNA was isolated using manufacturer instructions, with the omission of RNA Binding Buffer. RNA concentrations were measured using a NanoDrop One™ (Thermo Scientific; Waltham, MA, USA; Cat # ND-ONE-W).

RNA samples were reverse transcribed to cDNA using a SuperScript™ IV VILO™ polymerase (Invitrogen; Cat # 11756010), according to manufacturer instructions. RNA transcripts were reverse transcribed using the commercial kit primers containing a mix of random and oligo(dT) primers (37). Briefly, 2000 ng of RNA was added to 4 µL of 5X VILO™ reaction mix and brought to a total of 20 µL with molecular grade water (Fisher Scientific; Cat # BP2819-1). The samples were thermocycled for 1 step of annealing at 25°C for 10 minutes, 1 step of reverse transcription at 50°C for 10 minutes, followed by 1 step of enzyme inactivation at 85°C for 5 minutes. These experiments were performed using a MiniAmp Thermal Cycler (Applied Biosystems; Waltham, MA, USA; Cat # A37834). cDNA content was assumed to be a 1:1 conversion from starting RNA. cDNA samples were divided into lots so that no sample would exceed 5 freeze-thaw cycles. cDNA lots were stored at -80°C until use.

For each reaction, 5 ng of cDNA was amplified using standard TaqMan probe conditions. Briefly, forward and reverse primers (900 nM each), a TaqMan probe (250 nM), and TaqMan Fast Advanced Master Mix (ThermoFisher Scientific; Cat # 4444557) were mixed to a final reaction volume of 10 µL. TaqMan primers and probes details are listed in Supplementary Table S1 . Plates were sealed using MicroAmp™ optical adhesive film (Applied Biosystems; Cat # 4360954). PCR samples were thermocycled for 1 step of uracil-N-glycosylase (UNG) inactivation at 50°C for 2 minutes, 1 step of denaturation at 95°C for 2 minutes, followed by 45 cycles of denaturation at 95°C for 1 second and annealing at 60°C for 20 seconds. Samples were cycled using a QuantStudio 3 Real-Time PCR System (96-well, 0.1 mL format; Applied Biosystems; Cat # A28567). Custom TaqMan probes and primers were designed using Primer Express 3.0.1 (Applied Biosystems; Cat # 4363991), spanning exon-exon junctions where possible. Quantification cycle (C_q) values were compared to the reference gene, peptidylprolyl isomerase a (Ppia). Ppia was chosen from a screen of 32 potential reference genes supplied in a TaqMan Array Mouse Endogenous Control Plate (Applied Biosystems; Cat # 4426696).

Epha7 has 6 variants, none of which have any regions unique to a given variant within the coding sequence. Therefore, a combinatorial approach was taken to selectively amplify each variant ( Supplementary Figure S1 ). Exons 8 and 9 are conserved across all 6 variants, so a single probe was designed to target a region within exon 8 (Variant 1; 1868–1882; NM_010141). A 15-nucleotide sequence at the 5’ end of exon 7 is present in variants 1-3, but not in 4-6 (Variant 1; 1838-1852). Exon-spanning forward primers (referred to as forward A and B) were targeted towards this site to differentiate variants into two sets. Reverse primers were designed at the exon 9-10 junction, as exon 10 had variations that clustered the 6 variants into 3 pairs (reverse primers 1-3). Using the 6 permutations of forward primers A and B with reverse primers 1-3, each of the 6 variants could be specifically targeted ( Supplementary Table S1 ). Custom primers were validated on mouse brain samples (data not shown). Further validation was performed only on EphA7 due to every sample yielding signal indistinguishable from noise. Other custom probes were presumed effective based on effective amplification, PrimerExpress 3.0.1 (Applied Biosystems) scoring, and Primer BLAST comparison.

For each reaction, 20 ng of cDNA was amplified using Quick-Load Taq 2X Master Mix (New England Biolabs; Ipswich, MA, USA; Cat # M0271) and 200 nM of each primer, following the manufacturer’s protocol (25). Primers and amplicon information are listed in Supplementary Table S1 . Reactions were performed using the manufacturer-recommended mix composition at a final reaction volume of 25 µL. PCR samples were thermocycled for 1 step of initial denaturation at 95°C for 30 seconds, followed by 45 cycles of denaturation at 95°C for 30 seconds, annealing at 53°C or 54°C for 30 seconds, and elongation at 68°C for 60 seconds per kilobase. A final extension step at 68°C for 5 minutes followed. Samples were separated using a 1% agarose gel supplemented with GelGreen Nucleic Acid Gel Stain (Biotium; Fremont, CA, USA; Cat # 41005). Bands were cut and isolated using the Qiaquick Gel Extraction Kit (Qiagen; Hilden, Germany; Cat # 28704) following manufacturer instructions. The isolated amplified products were sent to Quintara Biosciences (Cambridge, MA, USA) for Sangar sequencing following Quintara sample preparation guidelines.

QuantStudio 3 readings were interpreted using QuantStudio Design & Analysis Software 2.6.0 (Thermo Fisher Scientific). A signal threshold of 0.3 was used to determine quantification cycle (C_q) values (31, 32). ΔC_q values were determined using a custom R script and analyzed using GraphPad Prism version 9.0.0 (121) (GraphPad Software, LLC; Boston, MA, USA). Two-way ANOVA followed by multiple comparison correction by controlling for false discovery rate was implemented. A two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli was used with a Q value (false discovery rate) of 0.05. A noise threshold of ΔC_q = 11.9 was determined using Epha2 TaqMan probe readings from Epha2 ^–/– mice, taking the lowest (strictest) ΔC_q value obtained from this data set. Data sets with more than 2 biological test groups exhibiting a mean above this threshold were considered non-specific amplification and were not included in statistical analysis. Epha2 and Efna5 readings in their respective knockouts were tested using multiple unpaired t-tests followed by a two-stage step-up method of Benjamini, Krieger, and Yekutieli with a Q value of 0.05.

To appropriately control for well-to-well variation, an internal reference gene was used for each sample. Thirty-two potential reference genes were screened and assessed based on quantification cycle (C_q) and standard deviation ( Supplementary Figure S2 ). Probes and primers for the target genes were pre-aliquoted into a pre-configured 96-well plate containing 16 commonly used reference genes and 16 mouse orthologs of human genes shown to be constitutively expressed in mice (Applied Biosystems; Cat # 4226694). As this screen was in search of endogenous controls, these samples did not include an internal reference gene and were quality controlled based on the passive reference dye, ROX. Therefore, results are shown and analyzed as C_q values rather than ΔC_q.

Of the targets screened, the 18s ribosomal RNA was the highest expressing (lowest C_q) reference gene with a mean C_q (M_Cq) = 10.4 (SD = 0.79). However, 18s was not chosen as the internal control to avoid depletion of reaction reagents due to its rapid amplification. The next cluster of moderately-expressed targets amplified with C_q values between 19.0-22.0. This group consisted of ß-actin (Actb), mitochondrially encoded ATP synthase 6 (ATP6), glyceraldehyde-3-phosphate dehydrogenase (Gapdh), and peptidylprolyl isomerase A (Ppia). Previous studies showed that disruption of EphA2 can dysregulate actin distribution and alter cytoskeletal morphology (20). Thus, Actb was eliminated as a candidate control despite being a commonly used control. Between ATP6 (M_Cq = 21.5, SD = 1.07), Gapdh (M_Cq = 20.0, SD = 0.61), and Ppia (M_Cq = 20.5, SD = 0.39), Ppia exhibited the smallest standard deviation. The ribosomal protein L37a (Rpl37a; M_Cq= 25.8, SD = 0.40) also had a comparable standard deviation but crossed the detection threshold at a later amplification cycle. Ultimately, Ppia was chosen as the endogenous control due to its earlier amplification and tighter standard deviation. The distribution of Ppia across all tested samples are compiled in Supplementary Figure S3 .

The nature of the Epha2 and Efna5 knockout mice results in successfully disrupted protein products; however, portions of the gene remain in the mRNA transcripts. The Epha2 gene is disrupted with the insertion of a neomycin resistance (neo) cassette at a HindIII restriction site located in exon 5 at positions 1410–1415 (NM_010139.3), previously reported at position 1372 (NM_010139), and a downstream XbaI site. This is predicted to produce a non-functional protein truncated after Arg426, between the two extracellular fibronectin domains (28). The Efna5 knockout uses a similar neo insertion technique, targeting a BamHI site at positions 249-254 in exon 1 and an EcoRI restriction site at positions 691-696 in exon 2 (NM_207654.3), replacing amino acids 1–128 (29).

Custom Epha2 probes were designed to amplify a 65-base-pair (bp) amplicon from positions 1405–1469 (NM_010139.3). The forward primer bound at positions 1405–1422 ( Supplementary Table S1 ), spanning the HindIII restriction site. This Epha2 primer set should not amplify the template if the neo cassette is present, and the TaqMan probe was designed to bind 3’ of the restriction site in a presumably deleted portion. However, the custom Epha2 probe yielded a positive signal in 2 of the 12 Epha2^–/– samples. These readings were used to determine a noise threshold of ΔC_q = 11.9, denoting any higher quantification cycles as non-specific signal.

Custom Efna5 probes and primers were designed to amplify a 70 bp amplicon from positions 831-900 (NM_207654.3). This TaqMan probe spanned exons 4 and 5, binding from positions 860-876. Although this portion of the gene was 3’ of the neo insertion, this position was targeted to differentiate Efna5 variant 2 from variants 1 and 3, which are identical within the coding sequence and only differ in the 5’ untranslated region (UTR). Surprisingly, this probe produced a positive signal from all Efna5^–/– samples, so the knockout was further investigated using Sangar sequencing.

Sequencing results from Efna5^–/– isolated lens fiber samples revealed that exon 2 of the Efna5 gene was excised, resulting in an out-of-frame deletion from bases 432–724 (amino acids 43–139) (data not shown). This was not expected based on the information available from the original paper describing the construction of the knockout (29). Importantly, exon 2 contains almost the entirety of the receptor binding ectodomain (amino acids 30-159) that is conserved through ephrin-A ligands (Conserved Domain Database: cd10425) (38). While the protein products are non-functional, amplifiable transcripts are still produced and thus, can be detected using RT-qPCR. These knockout readings were not included in our analyses.

To investigate gene expression differences related to age and Epha2 or Efna5 knockouts, 4 sample groups were used per strain: young-adult wild-type (Y-WT), middle-aged wild-type (M-WT), young-adult knockouts (Y-KO), and middle-aged knockouts (Y-KO). Age-related changes were determined by Y-WT vs. M-WT and Y-KO vs. M-KO comparisons. Genotype-related changes were determined by Y-WT vs. Y-KO and M-WT vs. M-KO. Transcript measures were not compared between tissue compartments nor between strains. Y-WT vs M-KO and M-WT vs Y-KO were not considered useful comparisons due to the confounding effects of both age and genotype. Eph and ephrin targets were chosen based on previous qualitative PCR results (27). Targets and variants that were absent from the lens from the initial screen were not quantitated in this current study.

Expression levels were further divided into three categories of high, moderate, and low expression based on their ΔC_q values. While the noise threshold was determined via the Epha2 knockout, the lowest ΔC_q value in the dataset was used to determine a maximum expression value. This lowest ΔC_q value also belonged to an Epha2 reading from Efna5^–/– fibers, at ΔC_q = 2.9. Using a ΔC_q range of 2.9 – 11.9, the readings were binned into approximate categories using a step size of 3. Using Epha2 in the fibers (M_ΔCq = 3.9) and Efna5^V1/3 in the epithelium (M_ΔCq = 4.4) as high-expression landmarks, samples were categorized into high expression (M_ΔCq < 6.0), moderate expression (6.0 ≤ M_ΔCq < 9.0), low expression (9.0 ≤ M_ΔCq < 11.9), and noise (M_ΔCq ≥ 11.9) categories. These thresholds were based on previous observations that EphA2 is predominantly expressed in the fibers while ephrin-A5 is predominantly expressed in epithelial cells (3).

Within the lens epithelium, Epha1, Epha5^V3 , and Epha5^V12 changed significantly with age regardless of genotype ( Figures 1A, C ). Epha1 increased with age regardless of genotype in both strains (Epha2 and Efna5; Y-WT vs. M-WT; Y-KO vs. M-KO). A genotype-dependent change was seen in Epha5^V3 , where expression levels increase with age only in the Efna5^–/– samples (Efna5; Y-KO vs. M-KO). In contrast, a strain difference is observed with Epha5^V12 as expression decreased with age in both wild-type and Efna5^–/– lens epithelium (Efna5; Y-WT vs M-WT; Y-KO vs. M-KO), but not between the matched samples in the Epha2 strain. Epha2 did not show appreciable change in either strain ( Table 1 ). Epha3, Epha4, Epha5^V9 , Epha5^V14 , Epha6, Epha7^V1-6 , and Epha8 were all either in the noise range or did not produce readable signal due to low transcript levels.

While no changes were observed in any of the quantitated Ephb genes in the lens epithelium ( Figures 1B, D ), Ephb2, Ephb3, and Ephb4^V1 were moderately expressed, while Ephb4^V2 and Ephb6 were highly expressed ( Table 2 ). Although both mouse strains are backcrossed to the C57BL/6J background, the two strains are maintained separately and not intermixed. Due to the separate maintenance of these inbred colonies, some genetic drift, and therefore some endogenous strain-related changes are expected. This highlights the necessity of using littermate controls for experiments. Overall, the strains are similar, but strain-dependent differences are evident and discussed below. Of the 14 characterized Ephs in the mouse, only 3 showed notable changes in the epithelium.

Ephrin-A and ephrin-B ligands are encoded by Efna and Efnb genes, respectively. Within the lens epithelium, only three ephrin ligands showed significant changes in transcript levels. Efna2, Efna3^IsoA/C (isoforms A and C), and Efna5^V1/3 changed significantly with expression levels in the valid range ( Figures 2A, C , Table 1 ). Efna2 exhibits an age-dependent decrease in both Epha2^–/– and Efna5^–/– epithelium (Epha2 and Efna5; Y-KO vs M-KO). This age-related decrease is also seen in Efna5^+/+ samples (Efna5; Y-WT vs. M-WT) but was not significant in Epha2^+/+ mice ( Figure 2A ).

Efna3 has 7 variants characterized to date that produce 4 unique protein isoforms. Variant 1 codes for isoform A, the longest of all the isoforms, which has an additional 81-amino-acid extension on the N-terminus compared to isoforms C and D. Variant 2 encodes for isoform B and also has the 81 amino acid extension; however, it lacks exon 4 (bases 591-668 in variant 1, NM_010108.1; bases 242-319 in variants 3 (NM_001377116.1), 4 (NM_001377117.1), and 5 (NM_001377118.1). Variants 3–5 encode for isoform C and only differ in the 5’ untranslated region (UTR), sharing an identical coding region. Isoform C lacks the 81 N-terminal residues but retains exon 4. Isoform D is encoded by Efna3 variants 6 and 7, which also only differ in the 5’ UTR. Isoform D lacks the 81 amino acid N-terminal extension like isoform C and also lacks exon 4, making this the shortest isoform. An age-matched genotype difference is observed where Efna3^IsoA/C expression increases between young-adult Epha2^+/+ and Epha2^–/– mice (Epha2; Y-WT vs. Y-KO), but not in middle-aged mice. However, it should be noted that Efna3^IsoA/C was very low expressing with a mean ΔC_q of 11.84, compared to the noise threshold of ΔC_q = 11.9.

Although ephrin-A5 is not an exclusive binding partner of EphA2, the knockout of Epha2 resulted in an increased expression of Efna5 variants 1 and 3 compared to wild-type in young adult mice (Epha2; Y-WT vs Y-KO). This elevated expression diminishes with age (Epha2; Y-KO vs. M-KO), becoming indistinguishable from wild-type at middle-age (Epha2; M-WT vs. M-KO, ns). Efna1^V2 and Efna3^IsoB/D were both in the noise range (ΔC_q ≥ 11.9) and thus were not considered meaningful readings. None of the Efnb genes showed significant changes between test groups ( Figures 2B, D ). Although the following genes did not change with age or genotype, Efna4, Efna5^V2 , Efnb2, and Efnb3 were moderately expressed, and Efna1^V1 , Efna5^V1&3 , and Efnb1 were highly expressed ( Table 2 ). Overall, 3 out of 8 ephrins showed significantly altered transcript levels in the lens epithelium ( Table 1 ).

In isolated lens fibers, Epha2, Epha5^V3 , and all the Ephb subtypes (Ephb1-4, Ephb6) showed altered gene expression either with age or genotype ( Figure 3 , Table 3 ). Epha2 levels increased with age in Epha2 ^+/+ fibers (Epha2; Y-WT vs. M-WT) but showed no changes in Efna5 mice ( Figure 3C ). Epha5^V3 showed both age- and genotype-dependent increases in expression. Epha5^V3 transcript levels increased in aging mice regardless of genotype (Epha2; Y-WT vs. M-WT; Y-KO vs. M-KO). Moreover, genotype differences were observed where Epha5^V3 transcript levels were increased in Epha2^–/– lens fibers compared to wild-type mice at both age groups (Epha2; Y-WT vs. Y-KO; M-WT vs. M-KO). In contrast, a strain difference is observed where Epha5^V3 only increased in Efna5^–/– fibers with aging (Efna5; Y-KO vs. M-KO) but not in aging Efna5^+/+ mice (Efna5; Y-WT vs. M-WT, ns) or in response to Efna5 knockout (Efna5; Y-WT vs. Y-KO, ns; M-WT vs. M-KO, ns).

All the tested Ephb genes exhibited a significant change in the lens fibers ( Figures 3B, D ); however strain-dependent changes are evident here. Ephb1 expression did not change in Epha2 ^+/+ WT lens fibers (Epha2; Y-WT vs. M-WT, ns), but showed a marked increase with age in Epha2^–/– fibers (Epha2; Y-KO vs. M-KO). Ephb1 undergoes a strain-dependent increase in aging wild-type lenses (Efna5; Y-WT vs. M-WT) of Efna5^+/+ mice, unlike the Epha2 strain. A further genotype difference is seen where Ephb1 increases with the knockout of Efna5 in young-adult mice (Efna5; Y-WT vs. Y-KO). In the Epha2 strain, Ephb2 transcript levels decrease between young-adult and middle-aged wild-type mice (Epha2; Y-WT vs. M-WT) but remains steady in middle-aged Epha2^–/– mice (Epha2; Y-KO vs. M-KO). The level of Ephb2 expression is significantly higher in middle-aged Epha2^–/– fibers compared to middle-aged wild-type samples (Epha2; M-WT vs. M-KO). Ephb2 is the only instance of decreased expression with age observed within lens fibers in this screen. Ephb3 only showed an age-related increase in Epha2^–/– fibers (Epha2; Y-KO vs. M-KO).

Ephb4^V1 increased with age in both strains (Epha2 and Efna5; Y-WT vs. M-WT; Y-KO vs. M-KO), but also showed an increased expression in middle-aged Epha2^–/– samples (Epha2; M-WT vs. M-KO). In contrast, Ephb4^V2 increased with age only in the Efna5 strain (Efna5; Y-WT vs. M-WT; Y-KO vs. M-KO). Ephb6 showed a significant increase in the middle-aged Epha2^–/– fibers compared to its wild-type counterpart (Epha2; M-WT vs. M-KO) and its young-adult knockout control (Epha2; Y-KO vs. M-KO).

Of all the Eph changes in aging lens fibers, Ephb1 was the only target that changed concomitantly with the knockout of Efna5 ( Figure 3D ). Epha5^V3 , Ephb2, Ephb4^V1, and Ephb6 all showed increases in expression in Epha2^–/– samples compared to their age-matched Epha2^+/+ control ( Table 3 ). Epha3, Epha4, Epha^V9 , Epha5^V14 , Epha6, and all Epha7 variants (1–6) were in the noise range or did not produce readable signal due to low transcript levels. Epha8 was above the threshold in the Efna5 strain and was low expressing in the Epha2 strain. Epha5^V12 , Epha8, Ephb1, and Ephb4^V1 were all low expressing, and Epha1, Epha5^V3 , Ephb2, Ephb3, Ephb4^V2 , and Ephb6 were all moderately expressing ( Table 4 ). Out of all the Eph receptor genes in the fibers, Epha2 was the only highly expressed target with a mean ΔC_q of 3.9. Overall, changes in Eph gene expression are more numerous in the lens fibers compared to the epithelium, with 7 of 14 characterized Ephs exhibiting a significant change.

In the isolated lens fibers, Efna1^V1 , Efna5^V2 , Efnb1, and Efnb2 showed significant changes with age or genotype ( Figure 4 , Table 4 ). Efna1^V1 increased with age across both strains in wild-type and knockout lens fibers (Epha2 and Efna5; Y-WT vs. M-WT; Y-KO vs. M-KO). A genotype difference is observed as young adult Epha2^–/– fibers showed significantly higher expression of Efna1^V1 than the wild-type controls (Epha2; Y-KO vs. M-KO). Efna5^V2 increased with age in wild-type Efna5 mice (Efna5; Y-WT vs. M-WT), but not in the Epha2 strain ( Figures 4A, C ). In contrast, Efna5^V1/3 did not change in the fibers. Targets in the noise range included Efna1^V2 , Efna2, Efna3^IsoA/C , Efna3^IsoB/D , and Efnb3. These targets either returned low transcript levels indistinguishable from noise or failed to produce readable signals. Efnb1 and Efnb2 increased with age regardless of genotype across both strains (Epha2 and Efna5; Y-WT vs. M-WT; Y-KO vs. M-KO). However, Efnb1 also increased in Epha2 knockouts compared to their wild-type counterparts (Epha2; Y-WT vs. Y-KO; M-WT vs. M-KO). In the lens fibers, Efna1 and Efna4 were lowly expressed while Efna5^V1&3 , Efna5^V2 , Efnb1, and Efnb2 were moderately expressed. There were no ephrins considered highly expressed in the lens fibers ( Table 3 ).

In summary, within the lens epithelium, Epha1, Epha5^V3 , Epha5^V12 , Efna2, Efna3^IsoA/C , and Efna5^V1&3 , showed significant changes with age. Efna3^IsoA/C, and Efna5 were the only targets that changed with genotype, increasing in middle-aged Efna5^–/– and young adult Epha2^–/– epithelium, respectively, compared to their age-matched wild-type samples. In the lens fibers, Epha2, Epha5^V3 , all the Ephb genes (Ephb1-Ephb6), Efna1^V1 , Efna5^V2 , Efnb1, and Efnb2 demonstrated age-related changes, some of which were strain-dependent. Epha5^V3 , Ephb2, Ephb4^V1 , Ephb6, and Efna4 exhibited expression changes concomitant with genotype, notably all in Epha2^–/– samples. Ephb1 was the only target that increased in Efna5 knockout fibers compared to the wild-type control ( Figure 4D ). Moreover, Ephb2 was the only target that decreased with age in the lens fibers.