Double heterozygous Mafg ^+/-:Mafk ^+/- germline knockout (KO) mice, which were generated in a previous study (Shavit et al., 1998; Onodera et al., 2000), were used to derive double knockout (Mafg ^−/−:Mafk ^−/−) and compound KO (Mafg ^−/−:Mafk ^+/-) mouse strains used in this study. Animals were housed in the Office of Laboratory Animal Medicine (OLAM) at the University of Delaware and experiments involving animals adhered to the Association of Research in Vision and Ophthalmology (ARVO) statement for the use of animals in ophthalmic and vision research and were approved by the Institutional Animal Care and Use Committee (IACUC) (IACUC Protocol No. 1226). Unless otherwise mentioned, mice were maintained on an ICR background. Animals were housed at a temperature range of 20–23°C in 12:12-h light-dark cycles with free access to water and food. Mafg ^+/-:Mafk ^+/- double heterozygous mice were crossed to Mafg ^+/-:Mafk ^−/− compound KO to generate various KO allele combinations. Genotyping was performed as previously described (Onodera et al., 2000; Agrawal et al., 2015). Briefly, genotyping was performed on tail-DNA prepared from post-natal or embryonic tissue using a commercial DNA-extraction kit (Qiagen, Catalog#158908 (Cell lysis solution), 158,912 (Protein precipitation solution), 158,916 (DNA hydration solution)) and by using the following primers: Mafg WT: Forward 5′-GCA​TGA​CTC​GCC​AGG​AAC​AG-3′, Mafg WT: Reverse- 5′-CCC​AAG​CCC​AGC​CTC​TCT​AC-3′, Mafk WT: Forward 5′-CCT​ACC​GTT​TCT​GTC​TTT​CCA​G-3′, Mafk WT: Reverse 5′-AAT​TCC​TGA​GGA​CAA​AGC​TGA​C-3′, and LacZ: 5′-CCT​GTA​GCC​AGC​TTT​CAT​CAA​C-3’.

Pregnant female mice collected from crosses between Mafg ^+/-:Mafk ^+/- and Mafg ^+/-:Mafk ^−/− mice were harvested for obtaining embryonic tissue at different stages. Observation of the vaginal plug was considered embryonic day (E) 0.5, and tissues was collected at E12.5, E14.5, and E16.5. Mouse embryonic head tissues were embedded without fixation in Optimal Cutting Temperature (OCT) (Fisher Scientific, Catalog# 14-373-65) and stored at -80°C until downstream applications. Tissue was sectioned to obtain coronal sections of the eye using a Leica CM3050 cryostat (Leica Microsystems, Buffalo Grove, IL, United States). Tissue sections were collected on Colorfrost Plus slides (Fisher Scientific, Hampton, NH, United States, Catalog #12-500-18) at chamber temperature of -18°C, at 12 μm thickness for E12.5 and 14 μm thickness for E14.5 and E16.5. For immunostaining, after thawing and air-drying the slides with sections, they were immediately fixed in 4% Paraformaldehyde (PFA) (Fisher Scientific, Catalog# AC416780010) in 1x phosphate buffer saline (PBS) (Corning, Catalog#21-031-CV) for 20 min at room temperature, which was followed by three washes in 1x PBS (5 min each wash). The slides were blocked for 1.5–2 h in blocking buffer (Table 1) followed by overnight incubation at 4°C with the specific primary antibody at the appropriate concentration in blocking or dilution buffer. The following day, section slides were subjected to three washes in 1x PBS (10 min each wash) and incubated for 1.5 h with secondary antibody at 1:200 dilution in blocking or dilution buffer (Table 1) along with the nuclear counterstain 4’,6-diamidino-2-phenylindole (DAPI, 1:1,000 dilution) (Thermo Fisher Scientific; Catalog# 62,248). The specific conditions for different primary antibodies are listed in Table 1. Slides were washed three times in 1x PBS for 10 min each after the secondary antibody incubation followed in mounting media and sealing with cover slips. Slides were stored at −20°C until they were imaged using a Zeiss LSM 880 confocal configured with Argon/Krypton laser (358 and 561 nm excitation wavelengths) (Carl Zeiss Inc., Göttingen, Germany).

In situ hybridization was performed as previously described (Lachke et al., 2012b). Briefly, mouse embryonic tissue was isolated and fixed in 4% para-formaldehyde, overnight. The tissue freezing media OCT (Tissue Tek, Torrance, CA) was used to embed and freeze the tissue in an orientation to yield coronal sections (16 µm). Oligomers that included T7 promoter sequence upstream of the gene-specific region were used to amplify cDNA that was used as template in an in vitro transcription reaction to prepare an antisense digoxygenin-labeled RNA probe. Frozen tissue was thawed and subjected to the previously described in situ protocol and images were imaged using light microscope. For hematoxylin and eosin (H&E) staining, mouse embryonic head tissue were fixed in Pen-Fix (Richard Allan Scientific, Kalamazoo, MI) overnight, followed by dehydration using ethanol, and embedding in paraffin. Sagittal paraffin sections (5 µm) were stained with H&E as previously described (Siddam et al., 2018) and visualized using light microscopy.

Fiji ImageJ software (v1.52P, NIH, Bethesda, MD) was used to quantify the differences in the mean fluorescence signal intensity between control (Mafg ^+/-:Mafk ^+/-) and Mafg ^−/−:Mafk ^−/− KO lens sections. Images were split into a single channel to measure and quantify the mean fluorescence intensity (Ki67, p57^Kip2, Phalloidin), counting the average cell number, or counting the average number of nuclei (Ki67, p57^Kip2) as previously described (Shihan et al., 2021). The background subtraction was performed for normalization after the Threshold application, and the fluorescence intensity was measured in the red channel depending on the criteria for each section in at least three biological replicates for control and Mafg ^−/−:Mafk ^−/− KO lenses. Finally, all statistics were assessed using either Student’s two sample t-test (correct for multiple comparisons using the Holm-Sídák method) or one-way ANOVA. Data are presented as mean ± SE (SEM) and differences were considered significant at p ≤ 0.05.

For RNA isolation for downstream assays namely, RNA-sequencing (RNA-Seq) and/or RT-qPCR, embryonic lenses at E16.5 (n = 8 lenses per biological replicate; total three biological replicates) were collected from control (Mafg ^+/-:Mafk ^+/-), compound (Mafg ^−/−:Mafk ^+/-) and double KO (Mafg ^−/−:Mafk ^−/−) mice. Total RNA isolation was performed using the mirVana™ RNA isolation kit (Thermo Fisher Scientific, Catalog#AM1560), followed by removal of small molecular weight RNA according to the manufacturer’s instructions. RNA quality was analyzed using fragment analyzer (Advanced Analytical Technologies, AATI FEMTO Pulse) and samples with RNA quality number (RQN) greater than 7.2 were used for library preparation and RNA-seq.

Lens RNA at E16.5 from control (Mafg ^+/-:Mafk ^+/-), compound (Mafg ^−/−:Mafk ^+/-) and double KO (Mafg ^−/−:Mafk ^−/−) mice was used to generate strand-specific, paired-end 101 bp-length libraries which were sequenced by DNA Link, United States (901 Morena Blvd. Ste 730 San Diego, CA 92117, United States) on a NovaSeq 6000 (San Diego, CA, United States). The processed reads were aligned against Mus musculus reference genome (mm39) using HISAT2 (Kim et al., 2015). The aligned reads were assembled using StringTie (Pertea et al., 2015) to obtain transcript-level expression counts. The count file was imported in edgeR package using R-statistical environment (Robinson et al., 2010) to analyze differential gene expression in control, compound and double KO datasets. In edgeR, reads that were lowly expressed, i.e. < 1 count per million in less than two samples were filtered out. The RNA-seq data is deposited in GEO and the accession number is: GSE207853.

The Database for Annotation, Visualization and Integrated Discovery (DAVID, v6 .8) was used for functional annotation by Gene Ontology (GO) categories (Huang et al., 2009). The pathways and GO categories identified were prioritized based on Benjamini corrected significant p-values. All GO comparisons were made against the 14 October 2020 release of the Gene Ontology Consortium (GOC) database (Ashburner et al., 2000), specifically KEGG (Kanehisa and Goto, 2000).

The isolated mouse embryonic lens RNA was used for cDNA synthesis using iScript cDNA Synthesis Kit (Bio-Rad, Catalog#1708890) followed by quantitative PCR (RT-qPCR). iScript reaction [5x iScript reaction mixture (4 µL), iScript reverse transcriptase (1 µL), Total RNA Template (300 ng), Nuclease-free H₂O (to make to total volume 20 µL)] was performed using a custom program [25°C (5 min), 42°C (30 min), 85°C (5 min), 4°C (hold)]. RT-qPCR was performed using cDNA-specific primer sets (Table 2) and a Power SYBR Green kit (Fisher Scientific, Catalog # A25742). The RT-PCR reaction [PowerUp SYBR master mix (12.5 µL), Forward Primer (0.5 µL of 10 µM), Reverse Primer (0.5 µL of 10 µM), cDNA (1 µL of 400 ng), Nuclease-free H₂O (10.5 µL to bring volume to total 20 µL volume)] using a custom program [95°C (2 min), followed by 40 cycles of 95°C (5 s) followed by 58°C (20 s), and terminal cycle of 95°C (15 s), 60°C (1 min), 95°C (15s) and hold at room temperature] was analyzed on a 7500 Fast PCR system (Applied Biosystems, Foster City, California). Three biological replicates (each biological replicate having two technical replicates) were used for control and KO samples. Fold-change was calculated using the ΔΔCT-method using Gapdh as a house-keeping gene and statistical significance was determined using a two-sample Student’s t-test.

iSyTE (integrated Systems Tool for Eye gene discovery) expression analysis has indicated that Mafg and Mafk are significantly expressed in mouse lens embryonic development (Supplementary Figure S1A). The expression of Mafg in mouse embryonic lens development is confirmed by in situ hybridization (Supplementary Figure S1B). Further, examination of previous generated RNA-seq data from isolated epithelium and fiber cells at E14.5, E16.5, E18.5 and P0 (Zhao et al., 2018) shows that Mafg and Mafk are both expressed in the epithelium and fiber cells (Supplementary Figure S1C). This analysis shows that as relative expression of Mafg decreases, that of Mafk increases, in progressive stages of development. Together, these data suggesting a role for Mafg and Mafk in lens embryonic development. Therefore, to examine the impact of Mafg and Mafk deficiency on lens development, we generated Mafg ^−/−:Mafk ^−/− double KO mice and first characterized the lens tissue with marker analysis. Immunostaining for the epithelial marker, E-cadherin (also known as Cdh1), demonstrated profound abnormalities in the appearance of the epithelium in Mafg ^−/−:Mafk ^−/− lenses. At E16.5, control lenses exhibit uniform E-cadherin protein expression and localization in the epithelium, which appears monolayered, as also suggested by nuclei stained with DAPI (Figure 1). In contrast, the Mafg ^−/−:Mafk ^−/− lenses exhibit nuclear staining that indicates multilayered epithelium wherein the E-cadherin protein expression pattern also appears abnormal. E-cadherin staining reveals that Mafg ^−/−:Mafk ^−/− epithelial cells appear irregularly shaped and aggregated compared to control. While this defect is observed in the vast majority of Mafg ^−/−:Mafk ^−/− lenses (92%, n = 12), the extent of the multilayered epithelium cellular abnormalities and cell shape changes varies between individual embryos (Figure 1). In Mafg ^−/−:Mafk ^−/− lenses that exhibit a comparatively milder defect, the abnormal multilayered cellular region is flanked by normal-appearing mononucleated epithelium. To identify the onset of these defects, we examined embryos at earlier stages. While less severe when compared to the defects observed at E16.5, the E-cadherin protein staining pattern appears abnormal in Mafg ^−/−:Mafk ^−/− epithelium at E12.5 (Figure 2) and E14.5 (Supplementary Figure S2). Interestingly, the E12.5 Mafg ^−/−:Mafk ^−/− epithelium exhibits early indications of multilayer formation of cells (Figure 2), suggesting that the lens defects may initiate at this stage and become progressively severe in development. Additionally, when compared to control lenses, the population of E-cadherin expressing cells appears to extend further toward the posterior region in Mafg ^−/−:Mafk ^−/− lenses (Figures 1, 2). Care was taken to analyze centrally located sections, avoiding peripheral sections. However, it should be noted that not all Mafg ^−/−:Mafk ^−/− lenses had an abnormal expansion of epithelial cells in the posterior region of the lens. It should also be noted that no overt changes were observed between control and Mafg ^−/−:Mafk ^−/− lenses with respect to E-cadherin’s localization within the cells. In contrast to Mafg ^−/−:Mafk ^−/− animals, Mafg ^−/−:Mafk ^+/- compound mice did not exhibit such severe lens defects at E16.5 (Supplementary Figure S3), suggesting that the absence of both alleles of Mafg and Mafk is necessary to cause these severe lens defects.

We next analyzed Mafg ^−/−:Mafk ^−/− mice for potential defects in lens fiber cells. Near the equator of the lens, at a region termed the transition zone, cells of the epithelium exit the cell cycle and begin differentiation into fiber cells. At the initial stage of differentiation, as fiber cells begin to migrate toward the interior of the lens, they undergo a sharp re-orientation in their position relative to the epithelium. At the lens equator location termed “the fulcrum” or “modiolus”, the apical regions of epithelial cells constrict to form an anchor point prior to elongation during their differentiation into fiber cells (Zampighi et al., 2000; Sugiyama et al., 2009). As a result, at this location, early fiber cells first begin to get positionally re-oriented so that their apical regions face the apical regions of cells of the epithelium. We sought to examine this region in Mafg ^−/−:Mafk ^−/− mouse lens. Staining for phalloidin (which stains F-actin) showed that E16.5 Mafg ^−/−:Mafk ^−/− mice exhibit abnormal F-actin distribution in cells near the fulcrum region of the lens (Figure 3). While all E16.5 Mafg ^−/−:Mafk ^−/− mice that were examined showed this defect on at least one side of the lens, in about one-third of the animals it was observed on both sides. Further, while in E16.5 control lenses, F-actin appears uniformly distributed at the apical junctions of epithelial and fiber cells, it appears reduced in the region anterior to the fulcrum in Mafg ^−/−:Mafk ^−/− lenses (Figure 3). This abnormal F-actin staining pattern was also observed earlier in development, at E14.5, in Mafg ^−/−:Mafk ^−/− lenses (Figure 4A). In control lenses, F-actin staining intensity is highest at the initial junction region of epithelial and differentiating fiber cells. In contrast, in Mafg ^−/−:Mafk ^−/− lenses, the highest staining intensity of F-actin is observed anterior to this region (Figures 4B–E). It should be noted that the quantitative analysis serves to demonstrate that F-actin levels are abnormally distributed in the area measured and are not necessarily a reflection of change in the total actin levels in the lens. Further, histological analysis shows that while in control, the fulcrum region appears normal, in E16.5 Mafg ^−/−:Mafk ^−/− mice, the fiber cell organization in this region, and beyond, is abnormal, suggestive of suboptimal interactions with the overlying epithelium (Supplementary Figure S4A). Moreover, in control lens, fiber cells appear to “curve” in the same direction as the overall structure of the lens (Supplementary Figure S4B). In contrast, fiber cells do not follow this natural curvature and their appearance is somewhat sigmoidal in Mafg ^−/−:Mafk ^−/− mice (Supplementary Figure S4B,C). Together, these data indicate that Mafg and Mafk deficiency results in abnormal abundance of F-actin in distinct regions of early differentiating fiber cells, which in turn impacts their organization in the lens.

Because E-cadherin is known to interact with beta-catenin, and both are implicated in cell-cell adhesion, we next sought to examine whether beta-catenin was altered in Mafg ^−/−:Mafk ^−/− lenses. Immunostaining for beta-catenin demonstrated no change in the levels of its expression in Mafg ^−/−:Mafk ^−/− lenses at E14.5 (Supplementary Figure S5A) or E16.5 (Supplementary Figure S5B). However, because beta-catenin is localized to the membrane, these data offer independent validation of these defects and their impact on the lens through the view of the cell membrane in Mafg ^−/−:Mafk ^−/− mice. These data corroborate that while in control lenses, epithelium and fiber cell organization appears uniform, in Mafg ^−/−:Mafk ^−/− lenses, cells in these regions appear disorganized at E14.5 (Supplementary Figure S5A’,A’’) and E16.5 (Supplementary Figure S5B’,B’’). At E16.5, the cortical fiber cell nuclei appear different between control and Mafg ^−/−:Mafk ^−/− lenses. Together with the findings on E-cadherin, these data suggest a general loss of normal epithelium architecture in Mafg ^−/−:Mafk ^−/− lenses. Furthermore, expression of the fiber cell marker, gamma crystallin, was unaltered in Mafg ^−/−:Mafk ^−/− lenses at E16.5 or earlier stages (Supplementary Figure S6A,B). Together, these data suggest that while the morphology of lens cells is abnormal, certain aspects of fiber cell gene expression are unaltered in Mafg ^−/−:Mafk ^−/− lenses.

To identify the specific RNA changes resulting from Mafg and Mafk deficiency, we next took an unbiased genome-wide approach and performed high-throughput RNA-sequencing (RNA-seq) of E16.5 Mafg ^−/−:Mafk ^−/− lenses. Principle component analysis showed that control and Mafg ^−/−:Mafk ^−/− lens replicates clustered away from each other (Figure 5). Further, as expected, compared to control, RNA-seq identified Mafg and Mafk to be severely reduced in Mafg ^−/−:Mafk ^−/− lenses. Comparative analysis showed that Mafg ^−/−:Mafk ^−/− exhibit 241 differentially expressed genes (DEGs) (±≥1.5-fold, p ≤ 0.05) with 144 being elevated and 97 being reduced in the absence of Mafg and Mafk (Supplementary Table S1).

Next, to further prioritize candidate genes relevant to lens biology so as to uncover the underlying molecular changes associated with Mafg ^−/−:Mafk ^−/− lens defects, we applied various selection criteria to the list of DEGs identified by RNA-seq. These include gene ontology (GO) and pathway analysis as well as iSyTE analysis, which informs on the expression of candidate genes in normal lens development and its altered expression in various gene perturbation mouse models with lens defects/cataract, and has been shown to be effective in prioritizing key genes and pathways in the lens (Lachke et al., 2011, 2012a, 2012b; Wolf et al., 2013; Manthey et al., 2014; Agrawal et al., 2015; Anand et al., 2015, 2018b, 2021; Audette et al., 2016; Cavalheiro et al., 2017; Patel et al., 2017; Kakrana et al., 2018; Krall et al., 2018; Siddam et al., 2018; Padula et al., 2019; Aryal et al., 2020; Barnum et al., 2020; Choquet et al., 2021). We also considered the potential relevance of the candidate genes to lens biology based on their function described in the published literature. In particular, we prioritized candidates that were starkly altered in Mafg ^−/−:Mafk ^−/− double KO lenses as opposed to in Mafg ^−/−:Mafk ^+/- compound KO lenses. GO analysis of Mafg ^−/−:Mafk ^−/− lens DEGs identified several potentially important categories such as “ephrin receptor signaling pathway”, “extracellular matrix”, “cell projection”, “cell junction”, “response to oxidative stress”, “cyclin-dependent protein serine/threonine kinase activity”, “actin filament binding” and “positive regulation of gene expression”, among others (Figure 5). Based on their differential expression in Mafg ^−/−:Mafk ^−/− lenses and potential significance to lens development, several candidates from these categories were selected for further analysis by iSyTE and independent validation by RT-qPCR (Table 3). iSyTE analysis showed that majority of these candidates exhibit significant expression or “enriched” expression in normal lens development (Supplementary Figure S7). Further, several of these candidates were found to also be misexpressed in specific gene perturbation mouse models with lens defects or cataract (Supplementary Figure S8). RT-qPCR analysis showed over 10-fold reduction of Mafg and Mafk transcripts in Mafg ^−/−:Mafk ^−/− lenses. As reported previously for Mafg ^−/−:Mafk ^+/- compound KO lenses (Agrawal et al., 2015), RT-qPCR confirmed elevated expression of Hmox1 and reduced expression of Trex1 in Mafg ^−/−:Mafk ^−/− double KO lenses, suggesting that some genes were similarly differentially expressed regardless of whether one or both copies of Mafk were absent in the context of homozygous deletion of Mafg (Figure 6). Importantly, we sought to confirm the differential expression of those genes that were starkly altered only in Mafg ^−/−:Mafk ^−/− double KO lenses compared to control and Mafg ^−/−:Mafk ^+/- compound lenses, as these would likely contribute to the early onset lens defects observed only in the double KO animals. Among these candidate genes, the peroxidase enzyme PXDN—secreted in the extracellular matrix, identified as lens-enriched by iSyTE and shown to linked to congenital cataract and other ocular defects in human or animal models—was found to be significantly overexpressed in Mafg ^−/−:Mafk ^−/− lenses (Figure 7). Furthermore, the ephrin signaling receptor, Epha5, was found to be significantly reduced in Mafg ^−/−:Mafk ^−/− double KO lenses. This is interesting because iSyTE analysis shows that Epha5 exhibits enriched expression in the lens, second only to the other ephrin receptor, Epha2, the perturbation of which is linked to cataract in humans and animal models (Figure 8). Moreover, Epha5 is known to be the receptor for the ligand Efna5 (ephrin-A5), the perturbation of which is linked to lens defects (Cheng and Gong, 2011; Cheng et al., 2017) resembling those observed in Mafg ^−/−:Mafk ^−/− lenses. Additionally, several other genes that may contribute to the lens defects were found to be differentially expressed in Mafg ^−/−:Mafk ^−/− lenses. These include cyclin-dependent kinase Cdk1, leucyl t-RNA synthase Lars2, mitogen-activated protein kinase Map3k12 and signal-induced proliferation protein Sipal1l, all exhibiting reduced expression in Mafg ^−/−:Mafk ^−/− lenses (Figure 6). In contrast, the genes that were found to be sharply elevated in Mafg ^−/−:Mafk ^−/− lenses were calmodulin-regulated spectrin associated protein Camsap1, collagen Col3a1 and cyclin-dependent kinase inhibitor Cdkn1c (p57^Kip2) (Figure 6). Further, compared to control, E16.5 Mafg ^−/−:Mafk ^−/− lenses exhibit elevated levels of Cdkn1c (p57^Kip2) protein in the transition zone and in differentiating fiber cells (Figures 9A,B). Cdkn1c (p57^Kip2) protein levels were also elevated in fiber cells located deeper in the Mafg ^−/−:Mafk ^−/− lens tissue, compared to control. Together, misexpression of these genes likely contribute to the lens defects in Mafg ^−/−:Mafk ^−/− mice. To examine whether increased cell numbers contributed to the multilayered epithelium defects observed in Mafg ^−/−:Mafk ^−/− lenses, we next performed staining with the established cell proliferation marker Ki67 (Gerdes et al., 1984). Compared to control, Mafg ^−/−:Mafk ^−/− lenses exhibit elevated number of cells with Ki67 staining and these were restricted to the epithelium region of the lens (Supplementary Figure S9). However, no stark difference was observed in the percent subset of cells that were Ki67 positive between control and Mafg ^−/−:Mafk ^−/− lenses at E16.5 (data not shown), suggesting that any change in rate of proliferation had ceased by this stage.