Impaired GSH biosynthesis disrupts eye development, lens morphogenesis and PAX6 function

Abstract

Purpose:

The purpose of this study was to elucidate the role and molecular consequences of impaired glutathione (GSH) biosynthesis on eye development.

Methods:

GSH biosynthesis was impaired in surface ectoderm-derived ocular tissues by crossing Gclc^f/f mice with hemizygous Le-Cre transgenic mice to produce Gclc^f/f/Le-Cre^Tg/- (KO) mice. Control mice included Gclc^f/f and Gclc^wt/wt/Le-Cre^Tg/- mice (CRE). Eyes from all mice (at various stages of eye development) were subjected to histological, immunohistochemical, Western blot, RT-qPCR, RNA-seq, and subsequent Gene Ontology, Ingenuity Pathway Analysis and TRANSFAC analyses. PAX6 transactivation activity was studied using a luciferase reporter assay in HEK293T cells depleted of GSH using buthionine sulfoximine (BSO).

Results:

Deletion of Gclc diminished GSH levels, increased reactive oxygen species (ROS), and caused an overt microphthalmia phenotype characterized by malformation of the cornea, iris, lens, and retina that is distinct from and much more profound than the one observed in CRE mice. In addition, only the lenses of KO mice displayed reduced crystallin (α, β), PITX3 and Foxe3 expression. RNA-seq analyses at post-natal day 1 revealed 1,552 differentially expressed genes (DEGs) in the lenses of KO mice relative to those from Gclc^f/f mice, with Crystallin and lens fiber cell identity genes being downregulated while lens epithelial cell identity and immune response genes were upregulated. Bioinformatic analysis of the DEGs implicated PAX6 as a key upstream regulator. PAX6 transactivation activity was impaired in BSO-treated HEK293T cells.

Conclusions:

These data suggest that impaired ocular GSH biosynthesis may disrupt eye development and PAX6 function.

Graphical Abstract

Introduction

Microphthalmia is a developmental disease that affects around 1 in 7000 births and results when one or both eyes fail to properly develop^1–3. Vision impairment can occur in the most severe cases of microphthalmia; as such, it is a leading cause of childhood blindness¹. Approximately 80% of microphthalmia cases can be explained by mutations in SOX2, OTX2, PAX6, VSX2, RAX, FOXE3, STRA6, ALDH1A3, or RARB^3,4. The remaining 20% of microphthalmia cases may be explained by exposure to environmental factors, such as pharmaceuticals (e.g., thalidomide⁵), ionizing radiation^6,7, gestationally-acquired infections⁸, or other factors (e.g., maternal vitamin A deficiency)^9,10. Currently, microphthalmia is an incurable disease and existing treatments improve the associated cosmetic abnormalities but not the associated vision loss¹¹. The development of an in vivo therapy capable of restoring eye development and visual function postnatally (such as that being developed in a Pax6 mutant mouse model) provides new hope that the vision loss associated with microphthalmia can be treated^12,13. Importantly, this research demonstrates that the eye retains developmental plasticity into early adulthood, opening the opportunity for the development of additional postnatal therapies to treat microphthalmia. A more complete understanding of the mechanisms regulating eye development are expected to accelerate the discovery of preventative strategies and novel therapeutic treatments for microphthalmia.

Epidemiological studies have identified associations between environmental factors and microphthalmia, including rubella infection, thalidomide and maternal malnutrition^3,5–10. Given that these factors have the potential to disrupt redox homeostasis^14,15, it is conceivable that changes in this biological process may contribute to abnormal eye and lens development. Indeed, the finding that the deletion of a transporter of glutathione (GSH) conjugates, RLIP76, from the developing lens impairs its development and results in microphthalmia¹⁶, suggests a role for redox homeostasis in regulating lens development. The redox theory of development suggests that GSH, the most abundant intracellular antioxidant, protects the embryonic (or intracellular) microenvironment and thereby maintains normal embryogenesis¹⁷. The lens maintains millimolar concentrations of GSH^18,19 through biosynthesis, and direct cellular uptake of GSH^20–23. The high concentrations of GSH in the lens are known to be required for maintenance of lens optical clarity^20,24,25. However, no information is available about the influence of GSH and redox status on early lens development or in the genesis of eye pathologies, such as microphthalmia.

To investigate the role of GSH in early eye and lens development, we have developed a novel mouse model in which GSH synthesis is ablated from ocular surface-ectoderm-derived tissues (i.e., eyelid, corneal epithelium, conjunctiva and lens) from as early as embryonic day 9 (E9). Targeted deletion of Gclc, the gene controlling expression of the glutamate-cysteine ligase catalytic subunit, resulted in a strong reduction in intracellular GSH content, and development of an overt microphthalmia phenotype with impaired development of multiple ocular structures, including the lens. To investigate the mechanism by which Gclc deletion impaired lens development, the present study investigated changes in gene expression by RNA sequencing of lenses collected at postnatal day (P) 1. These experiments demonstrated that Gclc KO caused upregulation of genes involved in the immune response and downregulation of genes related to lens development. Further analysis of gene expression altered by Gclc ablation revealed impairment of PAX6 transcriptional regulation activity, which was confirmed by inhibiting GCLC activity in cultured HEK293T cells. When taken together, it is apparent that reduced GSH influences eye development through a mechanism that may involve alterations in transcriptional regulation activity of PAX6.

Methods

Generation of Surface-Ectoderm Specific Gclc Knock-out Mice

In order to delete Gclc from surface ectoderm-derived tissues, Gclc floxed (Gclc^f/f) mice (C57BL/6J background) were bred with Le-Cre (FVB/N background) transgenic mice (Le-Cre^Tg/-). The resultant Gclc^f/f/Le-Cre^Tg/- (KO) mice have ocular Cre recombinase expression restricted to only the eyelid, corneal epithelium, conjunctiva and lens from age E9²⁶. The generation of the Le-Cre (Tg(Pax6-cre,GFP)1Pgr) transgenic and Gclc^f/f mice have been previously described^26,27. Le-Cre transgenic mice were originally produced by Dr. Ruth Ashery-Padan (Department of Human Molecular Genetics and Biochemistry, Sackler Faculty of Medicine, Tel Aviv University)²⁶, and were obtained from Dr. David Beebe (Department of Ophthalmology and Visual Sciences, Washington University School of Medicine). Gclc^wt/f/Le-Cre^Tg/- (HET) mice were continuously crossed with Gclc^f/f mice to generate the three experimental mouse genotypes: Gclc^f/f/Le-Cre^−/− (CON), Gclc^wt/f/Le-Cre^Tg/- (HET), and Gclc^f/f/Le-Cre^Tg/- (KO). All data presented in this paper were generated from B6/FVB mixed genetic background mice. Mice of both sexes were group-housed (no more than 5 mice/cage) and maintained on a 12-hour light-dark cycle, with food and water available ad libitum. All experiments were performed in strict accordance with the National Institutes of Health guidelines, with protocols approved by the Yale University Institutional Animal Care and Use Committee.

Generation of Le-Cre Transgene Control Mice

To generate Le-Cre transgene control mice, we bred Gclc^f/f/Le-Cre^−/− (CON) mice with Gclc^wt/f/Le-Cre^Tg/- (HET) mice. The F1 mice (Gclc^wt/f/Le-Cre^Tg/- and Gclc^wt/f/Le-Cre^−/−) were then interbred to produce F2 mice of the Gclc^wt/wt/Le-Cre^Tg/- (CRE) genotype. These CRE mice served as Le-Cre transgene control mice.

Genotyping

Genomic DNA (obtained from a 2 mm ear punch collected at post-natal day (P) 14) was extracted using DirectPCR Lysis Reagent (Tail) (Viagen Biotech, Los Angeles, CA) and PCR Grade Proteinase K (Sigma-Aldrich) according to the manufacturer’s protocol. Briefly, the ear punch sample was placed in 100 μL DirectPCR Lysis Reagent (Tail) (Viagen Biotech, Los Angeles, CA) containing 1 μL PCR Grade Proteinase K. The sample was incubated overnight at 55°C, and subsequently incubated at 85°C for 45 min. Two separate PCR reactions were needed to properly describe each individual animal: one detected the Gclc(wt) and Cre(Tg) alleles (Fig. 1A; left lane for each mouse), and the other detected the Gclc(f) allele (Fig. 1A; right lane for each mouse). Gclc alleles were determined using 5^′-CTATAATGTCCTGCACTGGG and 5^′-TAGTGAACGCTGTTAAAGG and 5′-CGGGTGTTGGGTCGTTTGT and the presence of the Cre transgene was identified using 5^′-GCGGTCTGGCAGTAAAAACTATC and 5^′-GTGAAACAGCATTGCCACTT, as previously described²⁷.

Figure 1. Surface-ectoderm derived tissue-specific deletion of Gclc reduces ocular GSH content.

(A) Agarose gel (2%) separation of genomic DNA obtained from ear tissue of P14-aged Gclc^f/f/Le-Cre^−/− (CON), Gclc^wt/f/Le-Cre^Tg/- (HET), and Gclc^f/f/Le-Cre^Tg/- (KO) mice by PCR amplification of the Gclc wild-type (Gclc(wt)) or Gclc floxed (Gclc (f)) alleles, or the Le-Cre transgene (Cre(Tg)). Two PCR reactions were needed to genotype mice to detect the presence of Gclc wt and Cre transgene alleles (left lane for each mouse) and the Gclc f allele (right lane for each mouse). (B) Western blot analysis of GCLC and α-crystallin proteins in the lens and retina of P1-aged CON, HET and KO mice. (C) Immunohistochemical staining of GCLC in CON and KO mice aged E14.5. Lens germinative zone and newly differentiated fiber cells (black box) are shown at higher magnification (bottom row). Black arrow indicates high expression of GCLC in the germinative zone. (D) GSH levels in the whole eyes of P21 CON (white bar) and KO (black bar) mice. Data are presented as the mean and associated s.e.m. from 3–4 mice. * P < 0.05, two-tailed Mann-Whitney test. (E) ROS levels in the lenses of P21 CON (black bar), KO (grey bar) and CRE (white bar). Data are presented as the mean and associated s.d. from 3 mice. **** P < 0.0001, one-way ANOVA, with Tukey’s correction, compared to other group (as indicated). (F) Representative photographs (A’-D’) and slit-lamp microscopy images (E’-F’) of eyes from P21-aged CON and KO mice. The images are representative of results obtained in all 3 mice per genotype.

Quantification of GSH

Postnatal day (P) 21 KO, CON and CRE mice were anesthetized by the open drop isoflurane method and euthanized by cervical dislocation. The eyes were enucleated and either immediately assayed for GSH or the lens was removed and then assayed for GSH. To measure GSH levels, the whole eye tissue was rinsed with phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 8 mM Na₂HPO₄, and 2 mM KH₂PO₄), resuspended in ice-cold Mammalian Lysis Buffer (Abcam, Cambridge, MA) and homogenized using a Tissuelyser (QIAGEN, Venlo, Netherlands) at a frequency of 30 Hz for 2 min at 4°C. The homogenate was centrifuged at 10,000 g for 30 min at 4°C. GSH levels were measured in supernatant samples using the one-step fluorometric reaction of samples using the GSH Ratio Detection Assay Kit (Fluorometric-Green) (Abcam, Cambridge, MA) according to the manufacturer’s protocol. Samples were incubated with kit buffer and probes (i.e., Thiol Green Indicator, GSSG Probe, Assay Buffer) for 30 min while protected from light. Thereafter, fluorescence intensity was monitored at Ex/Em of 490/520 nm on a microplate reader (Spectramax M3, Molecular Devices). Data are presented as the mean and associated standard error of the mean. Differences were determined using Student’s unpaired t-test (GraphPad Prism version 7.0a for Mac OS X, GraphPad Software, La Jolla California, USA), with P < 0.05 being considered significant.

Quantification of ROS

Postnatal day (P) 21 CON, KO and CRE mice were anesthetized by the open drop isoflurane method and euthanized by cervical dislocation. The eyes were enucleated, and lenses removed. The ROS levels were assayed in the lenses as previously described²⁸ with some modifications. Briefly, the freshly isolated lenses from one animal (3 animals per genotype) were placed into a single 96-well plate in 200 µL of Medium 199 (Sigma-Aldrich) at 4°C. 7.5 µM Dihydrorhodamine 123 (DHR) (Invitrogen, Waltham, MA), a colorless stain that easily passes through membranes and is oxidized by ROS into rhodamine 123, was added to each well and incubated at 4°C for 30 minutes. Stained lenses were washed 3 times in chilled PBS and then resuspended in 200 µL chilled PBS for imaging. Fluorescence intensity was measured at Ex/Em of 507/529 nm on a microplate reader (Spectramax M3, Molecular Devices). The lenses were then removed from the PBS, dried on a Kimwipe and weighed on a XPE105 (Mettler Toledo, Columbus, OH) scale. Data are presented as the mean relative fluorescence (and associated standard deviation) normalized to tissue weight. Differences in ROS were determined using one-way ANOVA with Tukey’s correction (GraphPad Prism version 9.1.1 for PC, GraphPad Software, La Jolla California, USA), with P < 0.05 being considered significant.

Histopathological Examination and Immunohistochemical Staining

E14.5 embryos, and eyes from P1, P3, P20, or P50 mice were collected, mice genotyped, and fixed overnight in 10% neutral buffered formalin (Sigma-Aldrich). Embryo heads were processed, embedded in paraffin, sectioned (5 μm thickness), and mounted onto glass slides by Dr. Mark Petrash’s group (Department of Ophthalmology, University of Colorado Anschutz School of Medicine). The resultant sections were either stained with hematoxylin and eosin (H&E) or subjected to immunohistochemical analysis. Immunostaining was conducted using the TSA Plus Biotin Kit (Perkin Elmer, Waltham, MA) according to the manufacturer’s protocol. Briefly, tissues were blocked with TNB Blocking Buffer (Perkin Elmer, Waltham, MA) for 30 min at room temperature. Subsequently, slides were incubated overnight with primary antibody in TNB Blocking Buffer (Perkin Elmer, Waltham, MA) at 4°C. After primary antibody exposure, the tissues were washed 3 times in 5 min with TNT buffer solution (Santa Cruz Biotechnology, Santa Cruz, CA), and subsequently incubated with horseradish peroxidase (HRP)-conjugated secondary antibody for 30 min at room temperature. Tissues were then washed 3 times for 5 min with PBST (0.05% Tween-20, 1 X Phosphate Buffered Saline) at room temperature. Thereafter, the tissues were incubated with TSA Plus working solution (Perkin Elmer, Waltham, MA) for 10 min at room temperature, and subsequently incubated in streptavidin-conjugated (SA)-HRP (diluted 1:100 in TNB) for 30 min at room temperature. Harris hematoxylin counterstain was then applied, and coverslip mounted for imaging. The primary antibodies in these studies targeted the following proteins: PAX6 (Abcam, ab195045), GCLC (Abcam, ab53179). All antibodies were used at a concentration of 1:200 unless otherwise stated.

Cell Culture

Single cell suspensions of human embryonic kidney (HEK293T) cells were seeded into 24 well cell culture plates (Corning, Corning, NY) containing Dulbecco’s Modified Eagle Medium (DMEM) (Thermo Fisher Scientific Inc., MA) supplemented with 10% fetal bovine serum (FBS) (Sigma-Aldrich, St. Louis, MO), 1% antibiotic-antimycotic (Sigma-Aldrich, St. Louis, MO) and 1% MEM non-essential amino acids (Sigma-Aldrich, St. Louis, MO) at 37°C in a humidified 5% CO₂ atmosphere. The cells were then used for Western blot analyses and Luciferase reporter assays.

Western Blot Analysis

Tissue and cell culture samples were prepared for Western blot analysis. For tissue samples, total protein lysates were generated in RIPA buffer (1% Nonidet P40, 0.5% sodium deoxycholate, 0.1% SDS in PBS) by homogenization with a Tissuelyser (QIAGEN, Venlo, Netherlands) at a frequency of 30 Hz for 2 min at 4°C. For cell culture samples, twenty four h after seeding, buthionine sulfoximine (BSO) (Sigma-Aldrich, St. Louis, MO, B2515–5G) was added to the HEK293T cell culture medium (at the desired concentration) and the cells were cultured for a further 48 h. Cells were harvested by treatment with 0.05% trypsin and total protein lysates were collected in RIPA buffer (1% Nonidet P40, 0.5% sodium deoxycholate, 0.1% SDS in PBS) by vortexing for 1 min and subsequent incubation on ice for 10 min. Cellular debris were removed by centrifuging the total cell lysates at 14,000 rpm for 10 min and the supernatant was collected. Protein concentrations from tissue or cell culture samples in the supernatant were quantified using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Ten (tissue samples) or fifteen (cell culture) μg of supernatant protein was resolved on 4–20% SDS-PAGE gradient gel (Bio-Rad, Hercules, CA) and transferred to a 0.2 μm nitrocellulose blot (Bio-Rad, Hercules, CA). Primary antibodies (1:1000) directed against GCLC (Abcam; ab190685), α-crystallin (Santa Cruz; sc-28306), γ-crystallin (Santa Cruz; sc-514201), PITX3 (Abcam; ab106827), PAX6 (Abcam, ab195045) or GAPDH (Abcam, ab9485) were used for immunoblotting. HRP-conjugated goat anti-rabbit or goat anti-mouse secondary antibodies (1:5000, CST; 7074P2) were used to visualize immunolabeled proteins. Quantitation of band densities was performed using the NIH Image J software²⁹. Target protein expression was normalized to the corresponding GAPDH expression or total protein as determined by Ponceau S stain (Sigma-Aldrich, St. Louis, MO) per the manufacturer’s instructions. Data are presented as the mean density (and associated standard deviation) of the normalized protein. Differences between protein expression were determined using ANOVA with post-hoc Dunnett’s or Tukey’s test correction (GraphPad Prism version 9.1.1 for PC, GraphPad Software, La Jolla California, USA), with P < 0.05 being considered significant.

RNA isolation and RT-qPCR analysis

RNA was collected from the lenses of CON, KO and CRE mice aged P21. Mice were anesthetized by open drop isoflurane method and euthanized by cervical dislocation. Their eyes were enucleated, placed into ice-cold PBS, and cleared of extraneous non-ocular tissues using forceps. The eyes were then transferred to 1 mL ice-cold PBS, non-lenticular tissues were carefully removed with forceps, and the resultant lenses were stored in 200 µL ice-cold RNAlater solution (Invitrogen, Waltham, MA). Total RNA was isolated from the lenses of three animals per genotype using the RNeasy Micro Kit (QIAGEN, Venlo, Netherlands) according to the manufacturer’s instructions. 75 ng of total RNA was reverse transcribed using the iScript cDNa Synthesis Kit (Bio-Rad) per the manufacturer’s instructions. 3.5 ng of cDNA was then used to estimate the abundance of specific mRNA transcripts with the iTaq Universal SYBR Green Supermix (Bio-Rad) on a CFX96 Real-Time PCR System (Bio-Rad). Relative Foxe3 mRNA transcript (5′-CCACAACCTCACCCTCAAC and 5′-AGCTACCGTTGTCGAACATG) abundance was estimated using the ΔΔCt method³⁰ with the housekeeping gene, GAPDH (5^′-TTGATGGCAACAATCTCCAC and 5^′- CGTCCCGTAACAAAATGGT), used as an internal normalization control for each sample. Data are presented as the mean relative gene expression (and associated standard deviation). Differences between gene expression were determined using ANOVA with Dunnett’s test correction (GraphPad Prism version 9.1.1 for PC, GraphPad Software, La Jolla California, USA), with P < 0.05 being considered significant.

Luciferase Reporter Assay

Single cell suspensions of HEK293T cells were transiently transfected with 0.50 µg of the Cignal PAX6 Luciferase Reporter (QIAGEN, CCS-3042L) using FuGENE HD transfection reagent (Promega, Madison, WI) at a ratio of 3:1 (FuGENE HD: DNA) at the time of seeding. HEK293T cells were treated with buthionine sulfoximine (BSO) (Sigma-Aldrich, St. Louis, MO) to reduce intracellular GSH levels^31,32 thusly, twenty-four h after seeding, the medium was replaced with medium containing 0 (control), 500 or 1,000 µM buthionine sulfoximine (BSO) (Sigma-Aldrich, St. Louis, MO). Cells were then incubated for a further 48 h after which luciferase reporter activity was measured using the Dual-Glo Luciferase Activity System (Promega, Madison, WI, E2920) according to the manufacturer’s instructions. Luminescence was detected using a SpectraMax M3 multimode microplate reader (Molecular Devices, San Jose, CA). The ratios of Firefly Luciferase/Renilla Luciferase luminescence were calculated for each sample. Activity in control and BSO-treated cells were compared using ANOVA with Dunnett’s correction (GraphPad Prism version 9.0, GraphPad Software, La Jolla, CA). P < 0.05 was considered significant.

Macroscopic Assessment of Microphthalmic Eyes

P21 mice were anesthetized by the open drop isoflurane method and euthanized by cervical dislocation. Eyes were imaged using a digital camera (OnePlus 4T smartphone, Shenzhen, China). Magnified images were collected using a DC-4 Digital Photographic Slit Lamp (Topcon, Oakland, NJ), as previously described³³.

RNA Sequencing (RNA-Seq) Library Preparation and Sequencing

RNA was collected from the lenses of CON and KO mice aged P1. Mice were anesthetized by open drop isoflurane method and euthanized by cervical dislocation. Their eyes were enucleated, placed into ice-cold PBS, and cleared of extraneous non-ocular tissues using forceps. The eyes were then transferred to 1 mL ice-cold PBS, non-lenticular tissues were carefully removed with forceps, and the resultant lenses were stored in 200 µL ice-cold RNAlater solution (Invitrogen, Waltham, MA). Six lenses from three mice of each genotype were pooled into three biological replicates per genotype. Total RNA was isolated using the RNeasy Micro Kit (QIAGEN, Venlo, Netherlands) according to the manufacturer’s instructions. Total RNA was analyzed using the Agilent 2100 Bioanalyzer RNA 6000 Pico assay to determine the RNA integrity number (RIN). Samples with a RIN ≥ 8.0 were used to prepare cDNA libraries for sequencing. Polyadenylated RNA was purified from the total RNA using the NEBNext® Single Cell/Low Input RNA Library Prep Kit for Illumina® (New England BioLabs). Sequencing was conducted on an Illumina NovaSeq 6000 machine using an S4 flow cell generating pairwise 100bp reads by the Yale Center for Genome Analysis. All RNA-seq data related to this project, including raw and processed RNA-seq data files, have been deposited in NCBI Gene Expression Omnibus (GEO) (accession number GSE175394).

RNA-seq Analysis

Raw RNA-seq data (FASTq files) was uploaded into the Galaxy web platform³⁴ for all data analyses [accessed at usegalaxy.org]. All settings for the tools used within the platform were set to default (unless otherwise stated). The FASTq files were evaluated for sequence quality using the FastQC tool (v0.72+galaxy1). Sequences were trimmed for quality using a sliding window (phred ≥ 20), and ambiguous bases (N) and any contaminating sequencing adapters were removed using the Trimmomatic tool³⁵. The trimmed reads were then mapped to the Mus musculus reference genome (GRCm38/mm10) by HISAT2 (v2.1.0+galaxy5)³⁶. Mapped reads were counted using the FeatureCounts tool (v1.6.4+galaxy1)³⁷. Differential gene expression tests and principle component analysis (PCA) were conducted using DESeq2 (v2.11.0.6)³⁸. Genes with cutoff values of ≥ ±1.0 log₂ fold-change (log2FC, KO/CON) and adjusted P < 0.05 (Benjamini-Hochberg method³⁹) were considered differentially-expressed and used for downstream bioinformatics analyses. Processed fragments per kilobase per million (FPKM) values for the Le-Cre RNA-seq data were retrieved from the Gene Expression Omnibus: GSE125108⁴⁰, and the log2FC was calculated. Processed Pax6 KO (NMRI/FVB mixed background) microarray datasets were retrieved from a previous publication⁴¹. Processed microarray datasets for a model of lens epithelial-mesenchymal transition (EMT) were retrieved from a previous publication⁴². All differentially-expressed genes (DEGs) in the lenses of P0 Le- Cre mice (lenses from Le-Cre mice compared with lenses from wildtype FVB/N mice)⁴⁰ were compared with the expression in Gclc KO lens RNA-seq data regardless of significance or fold- change in the Gclc KO lens (KO/CON). All significant DEGs in the Pax6 KO lens⁴¹ were compared with significantly DEGs in the Gclc KO lens regardless of the genes fold-change (KO/CON). Differentially expressed lens identity genes^43,44 were compared between Pax6 KO⁴¹ and Gclc KO lenses. Significantly DEGs were compared between a model of lens EMT⁴² and Gclc KO lenses. Pearson correlation coefficient of determination (R²) values were used to derive P values with the following equation t = [r √(n-2)]/ √ (1-r²) and determining the corresponding two-tailed P value⁴⁵. An R² ≥ 0.50 and P < 0.05 were considered significant. Odds ratio was used to determine the strength of association between the transcriptomes of Gclc KO, Le-Cre, and Pax6 KO lenses; P < 0.05 was considered significant.

Bioinformatics Analyses

Gene ontology (GO) functional annotation analysis was performed on DEGs using the Database for Annotation, Visualization and Integrated Discovery (DAVID) bioinformatics resource⁴⁶ based on Fisher’s exact test P value or Fisher’s exact test with Benjamini-Hochberg corrected P value. GO functional annotation analysis was used to identify the biological pathway terms overrepresented among downregulated and upregulated DEGs in KO mice (relative to CON mice). Ingenuity Pathway Analysis (IPA) (Version 52912811, Ingenuity Systems, QIAGEN) was used to identify ‘canonical pathways’ in the DEGs based on the Fisher’s exact test with Benjamini-Hochberg corrected P value. Large scale enrichment of transcription factor binding motifs in the DEGs was determined by F-match analysis from the TRANSFAC® database (genexplain, Germany)⁴⁷. Overrepresented transcription factor binding sites in the regulatory regions of interest (‘Yes’ sequences) were compared with the mouse genome background sequences (‘No’ sequences) to provide an overrepresentation score and multiple tests were corrected using the Benjamini-Hochberg method.

Results

Lens-specific Gclc deletion reduces whole eye GSH content and causes microphthalmia.

We successfully generated three genotypes, CON, HET and KO, by crossing Gclc^f/f mice with Le-Cre^Tg/- (Fig. 1A) (abbreviations defined in Table 1). As expected from previous results²⁷, the Le-Cre transgene-driven deletion of Gclc did not cause embryonic lethality, i.e., all three genotypes were born in the expected Mendelian frequency (data not shown). Western blot analysis of P1- aged mice confirmed the absence of GCLC expression from the lenses of KO mice. The retinae of KO mice also showed decreased GCLC expression (Fig. 1B); this result is expected given the ectopic expression of CRE in Le-Cre transgenic mice⁴⁰. Immunohistochemical staining was also used to detect changes in GCLC in the developing eye. In E14.5-aged mice, GCLC expression was abundant in the developing eye, but absent from the surface ectoderm-derived tissues (and the lens) of KO mice (Fig. 1C). At E14.5, the lens epithelial cells in the germinative zone of CON mice had high GCLC expression (Fig. 1C, black arrow). Given that GCLC catalyzes GSH synthesis, it is not surprising that whole eye levels of GSH were markedly reduced in KO mice (Fig. 1D). Additionally, the deletion of Gclc resulted in an increase in ROS in the lenses from only KO mice (Fig. 1E). In KO mice, Gclc deletion from surface ectoderm-derived tissues resulted in an overt microphthalmia phenotype that occurred in all KO mice (Fig. 1F).

Table 1.

Abbreviations used for mouse genotypes.

Abbreviation	Genotype
CON	Gclc f/f /Le-Cre −/−
HET	Gclc wt/f /Le-Cre Tg/−
KO	Gclc f/f /Le-Cre Tg/−
CRE	Gclc wt/wt /Le-Cre Tg/−

Suppression of Gclc impairs normal eye development.

To characterize structural changes in the eyes of KO mice, gross morphological and histological analyses were conducted in the eyes of mice at embryonic (i.e., E14.5) early post-natal (i.e., P1), juvenile (i.e., P20), and adult (i.e., P50) stages of development (Fig. 2). The eyes of CON mice did not show ocular abnormalities at any of the ages examined (Fig. 2A–L). In contrast, age-dependent pathological changes in the lens, cornea, iris, and retina occurred in KO mice (Fig. 2A’-P’). At E14.5, KO mouse eyes had developed all ocular tissues; however, several abnormalities had started to emerge. Specifically, there was a reduction in the size of the anterior chamber (Fig. 2A’) and the density of blood vessels in the posterior chamber was increased (Fig. 2D’, black arrowhead). At P1, the changes in the eyes of KO mice were even more drastic, i.e., the anterior chamber was markedly decreased in size (Fig. 2E’). In addition, severe lens fiber cell vacuolation (Fig. 2E’, F’, G’) occurred, and the increased number of blood vessels in the posterior chamber persisted (Fig. 2F’, black arrowhead). The KO eye abnormalities worsened with age such that by P20, the iris had formed a thick layer that remained attached to the cornea (Fig. 2I’, J’), the anterior chamber was absent, the lenses were smaller in size and the lens fiber cells remained highly vacuolated with malformed epithelial cells (Fig. 2I’, J’, yellow arrowhead). The corneas had thinner anterior epithelial cell layers that lacked proper differentiation (Fig. 2K’). In addition, retinal cellular layers had abnormalities with the outer and inner plexiform layers appearing similar in thickness (compared with those in CON mice) and the presence of blood vessels in the posterior chamber (Fig. 2I’, L’). At P50, the KO eyes had the same abnormalities as seen at P20, i.e., thickened layer iris that was attached to the cornea (Fig. 2N’, O’) and absence of the anterior chamber (Fig. 2N’), thinner anterior corneal epithelial cell layer that lacked proper differentiation (Fig. 2O’), small lens with vacuolated fiber cells and malformed epithelial cells (Fig. 2M’, N’, yellow arrowhead), and severe abnormalities in the retina including in-folding (Fig. 2P’).

Figure 2. Gclc deletion causes severe ocular malformations.

(A-P’) Representative images of hematoxylin & eosin-stained eyes from CON (A-L) and KO (A’- P’) mice aged E14.5 (A-C, A’-D’), P1 (D-F, E’-H’), P20 (G-I, I’-L’) or P50 (J-L, M’-P’). The images are reflective of results obtained in all 3 mice per genotype. Lens (black box) in A, B’, D, E’, I’ and M’ are shown at higher magnification in B, B’, E, G’, J’, and N’, respectively. Retina (dashed black box) in A, A’, F’, I’ and M’ are shown at higher magnification in C, D’, H’, L’ and P’, respectively. Cornea (yellow box) in G, J’ and M’ is shown at higher magnification in H, L’, and P’. Blood vessels are indicated by a black arrowhead (C, D’, F, F’) and malformed lens epithelium is indicated by a yellow arrowhead (J’, N’). Abbreviations: AC, anterior chamber; C, cornea; I, Iris; L, lens; R, retina; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; ONL, outer nuclear layer.

The microphthalmia phenotype in Le-Cre hemizygous mice is exacerbated by Gclc deletion.

Le-Cre transgenic mice can display a microphthalmia phenotype, regardless of floxed alleles^40,48–50. Therefore, the extent to which hemizygosity for the Le-Cre caused microphthalmia was investigated in our B6/FVB mixed background mice (Supp. Fig. 1.) To do so, CON (Gclc^f/f/Le-Cre^−/−) mice were bred with HET (Gclc^wt/f/Le-Cre^Tg/-) mice for at least three generations to produce Gclc^wt/wt/Le-Cre^Tg/- (WT/Cre) mice, thus isolating the Le-Cre transgene on our B6/FVB mixed background (Supp Fig. 1A). At age P21, the combined eye weights of WT/Cre mice and KO mice were less than those in CON mice and the eye weights of KO mice were less than those in WT/Cre mice (P < 0.05) (Supp. Fig. 1B). Eye weights were measured at P21 because major ocular deformities are observed at this age (Fig. 2). Analysis of GSH levels in the lens revealed that only the lenses of KO mice had a significant reduction in GSH (Supp. Fig. 1C). Histological analysis of the eyes from CRE animals (CRE eyes) at P50 revealed minor pathological changes in the lens, iris, cornea and retina (Supp. Fig. 1D). P50 was chosen for this analysis as we observed that the ocular phenotypes in KO mouse eyes progress with age; thus, we would expect any phenotype observed at age P50 to be more severe than it would be at P21. When compared with the changes in KO mice (Fig. 2), the morphological changes observed at P50 in the CRE eyes were markedly less severe. The eyes from all KO mice displayed severe morphological changes, whereas only some CRE eyes displayed morphological changes (Supp. Fig. 1D). By P50, CRE animals displayed smaller lenses (Supp. Fig. 1A’, B’, C’) with some containing vacuolated cells (Supp. Fig. 1B’, C’). A reduction in the anterior chamber was observed in the eyes of some CRE mice (Supp. Fig. 1A’). Iris hyperplasia (Supp. Fig. 1D’, E’) and altered anterior corneal epithelial differentiation were observed in the eyes of CRE animals (Supp. Fig. 1D’, E’), including thinning of the corneal epithelial cell layer and a loss of corneal epithelial cell stratification with hyperproliferation and corneal stromal cells with slightly altered histology (Supp. Fig. 1D’, E’). Thickening of the retinal inner plexiform layer was observed, as well as slightly increased numbers of ganglion cells in the eyes of only some CRE animals (data not shown).

Gclc deletion in the lenses of newborn mice disrupts the transcriptomic signature.

RNA-seq was used to assay changes in the newborn (P1-aged) lens transcriptome induced by the Gclc gene deletion. The biological replicates for KO and CON produced between ≈15 million and ≈44 million 100 base pair sequence reads. Principle component analysis (PCA) of the CON and KO gene expression data revealed strong clustering of biological replicates (Supp. Fig. 2). Differential gene expression analysis revealed that the deletion of Gclc altered the expression of 1,552 genes, with 530 genes being downregulated and 1,022 being upregulated (Fig. 3A, Supp. Excel Sheet 1). The RNA-seq data was validated using qRT-PCR on a selection of nine genes representative of downregulated or upregulated genes, as well as genes that were not differentially expressed. A strong correlation between RNA-seq and was qRT-PCR determined in the lenses of P1-aged KO and CON mice (R² = 0.88, P < 0.001) (Supp. Fig. 3, Supp. Table 1). Given the mild microphthalmia phenotype observed in the lens of Le-Cre transgenic mice (Gclc^wt/wt;Le-Cre^Tg/-) (Supp. Fig. 1), it was important to identify the changes in the transcriptome of lenses from KO mice that may be attributed to the presence of the Le-Cre transgenes. Comparison of the KO lens transcriptome with the recently published lens transcriptome from P0-aged Le-Cre hemizygous mice⁴⁰ revealed that the DEGs present in both genotypes are not correlated (R² = 0.17, P = 0.39) (Supp. Fig. 4, Supp. Table 2); of the 16 DEGs previously identified in the lenses of Le-Cre hemizygous mice (when compared with FVB/N controls)⁴⁰, only six were also differentially expressed (irrespective of log₂FC magnitude) in the lenses of KO mice (when compared with CON). This result indicates that the dysregulation of gene expression in KO lenses is primarily driven by the deletion of Gclc.

Figure 3. Transcriptomic differences between lenses of P1-aged CON and KO mice.

(A) Volcano plot illustrating differentially expressed genes (DEGs) in the lenses of P1-aged CON and KO mice. Expression levels in the lenses of KO mice are shown as log₂fold change (log2FC) and -log10Padj relative to lenses of CON mice (pooled samples from 3 mice). DEGs were classified as: |log₂FC| ≥ 1; Benjamini-Hochberg adjusted P (Padj) < 0.05. Colors: Blue, expression in KO mice significantly downregulated; Grey, no significant change in expression; Red, expression in KO mice significantly upregulated. (B-C) Gene ontology of upregulated (B) and downregulated (C) genes as determined by David Bioinformatics Resources. The ten most highly significant biological process (BP) terms and the number of genes in each term are shown. P values were calculated by Benjamini-Hochberg method (B) or Fischer’s exact test (C). Fischer’s exact test was used because the Benjamini-Hochberg method limited our ability to detect BP terms in the downregulated genes.

Gene Ontology (GO) biological process (BP) terms enriched among the DEGs (Fig. 3A) when divided into two categories, upregulated and downregulated, were analyzed (Fig. 3B, C). Among the upregulated genes in the lenses of KO mice, many GO terms associated with the activation of the immune system were found, such as “immune system process”, “inflammatory process”, “neutrophil chemotaxis”, “chemokine-mediated signaling pathway”, “positive regulation of inflammatory response” and “immune response” (Fig. 3B). Analysis of the ten most upregulated genes in KO lenses revealed several genes involved with the immune system and inflammation (i.e., Arg1, Ccl7, Ptgs2, Rbp4) (Table 2). Among the downregulated genes, many GO terms associated with eye and lens development were found, such as “lens development in camera-type eye”, “eye development”, “camera-type eye development”, “lens fiber cell development” and “lens fiber cell differentiation” (Fig. 3C). Analysis of the ten most downregulated genes revealed many genes involved with eye functions in the retina (i.e., Drd4, Mtnr1a, S1pr1, Th) and lens development (Pitx3) (Table 2). Consistent with the general enrichment of immune function in upregulated genes, genes that are known to repress immune function were among the ten most downregulated genes (i.e., Avp, Drd4) (Table 2).

Table 2. The ten most downregulated or upregulated genes in the lens of KO mice.

Differentially expressed genes are presented as log₂ fold change (log2FC) of transcripts from the lenses of KO mice compared with CON mice. P values were calculated by Benjamini-Hochberg method (adj. P value).

Downregulated genes			Upregulated genes
Gene name	log2FC	adj. P value	Gene name	log2FC	adj. P value
Drd4	−3.95	6.13E-15	Arg1	7.41	2.10E-56
Scand1	−3.64	2.16E-08	Ptgs2	7.24	2.57E-38
6430562O15Rik	−3.25	2.80E-08	3300005D01Rik	6.34	1.10E-25
Th	−3.22	5.70E-05	Cdsn	6.03	7.87E-22
Sox1ot	−3.13	2.03E-11	Slc14a1	6.01	2.95E-22
Pitx3	−2.84	1.07E-4	Ccl7	5.99	8.79E-83
Avp	−2.76	1.38E-08	Rbp4	5.70	3.70E-28
Mtnr1a	−2.70	1.85E-09	Cdkn1a	5.68	3.39E-57
Dyrk3	−2.70	7.18E-04	Dglucy	5.62	1.75E-18
S1pr5	−2.67	5.38E-04	Arr3	5.59	7.37E-22

Expression of lens cell identity genes are deregulated in Gclc KO lenses

Because of the overrepresentation of lens development GO terms found among the downregulated genes, the expression of lens cell identity genes in the lenses from KO mice were investigated (Fig. 4). Optical clarity is maintained in the lens, in large part, through the expression and accumulation of water-soluble structural proteins called crystallins. A disruption in crystallin expression is indicative of disturbed lens development and function⁵¹. Gene expression of paralogs from three classes of crystallins (i.e., α-, β-, γ- crystallin), identified as downregulated in KO lenses by RNA-seq (Fig. 4A), were also found to be downregulated by Western blot analysis (Supp. Fig. 5). Lens identity genes, as identified in two recent studies^43,44, were further investigated (Figs. 4B, C). Interestingly, the majority (326 out of 384) of these genes were upregulated in the lenses of KO mice (Fig. 4B), which was further confirmed by calculating the odds ratio of overrepresentation of upregulated lens epithelial identity genes in the lens of KO mice (odds ratio = 2.37, P < 0.0001). In contrast, the vast majority (287 out of 325) of the lens fiber cell identity genes were downregulated in the lens of KO mice (Fig. 4C), which was further confirmed by calculating the odds ratio of overrepresentation of upregulated lens epithelial identity genes in the lenses of KO mice (odds ratio = 9.07, P < 0.0001). Analysis of crystallin and PITX3 (transcription factor critical for lens development) protein expression in the lenses of newborn CON, CRE and KO mice revealed that lens identity protein expression was only decreased in the lenses of KO mice (Supp. Fig. 5B). Analysis of Foxe3 (transcription factor critical for lens development) gene expression in the lenses of CON, CRE, KO aged P21 mice revealed that Foxe3 expression was only decreased in the lenses of KO mice (Supp. Fig. 5C). In summary, the finding that crystallin, PITX3 and Foxe3 expression is unchanged in the lenses of CRE mice suggests that the impaired lens development and microphthalmia phenotype in KO mice is unique from that in CRE mice.

Figure 4. Altered expression of crystallin and other lens cell identity genes in KO lenses of P1-aged KO mice.

(A-C) RNA-seq box plot and heatmaps indicating the expression of key lens cell identity genes in the lens at P1-aged KO mice (pooled samples from 3 mice). (A) RNA-seq box plots indicating variance stabilizing transformed (VST) normalized count data for differentially-expressed crystallin genes in the lens of KO mice. VST count data are shown as mean (thin horizontal bar) ± standard deviation (error bar). P values were calculated by Benjamini-Hochberg method. Colors: red, CON; blue, KO. (B-C) RNA-seq expression heatmap showing differentially-expressed lens epithelial cell (B) and fiber cell (C) identity genes in the lenses of KO mice. Data are shown as VST count data and z-score (standard deviations from mean). Gclc control sample is indicated by the prefix CON and each Gclc knockout sample is indicated by the prefix KO. Dot represents one pooled sample. (D) Pearson correlation (coefficient of determination) of lens gene expression levels (with gene expression shown as log₂fold change (log2FC) in the lenses of KO mice relative to the lenses of CON mice) between Pax6 microarray⁴¹ and Gclc KO RNA-seq datasets restricted to DEGs in Pax6 KO lens (upper left corner) or restricted to only differentially-expressed lens identity genes (orange points; grey box). Each point represents one gene. Blue line indicates linear trendline.

Lens development and the expression of lens identity genes are coordinated by transcription factors^52–54. Therefore, TRANSFAC analysis was performed to identify transcription factors that may be responsible for the transcriptomic changes in the lenses of Gclc KO mice. This analysis revealed the PAX family transcription factor binding motif to be the most overrepresented among the promoters of DEGs in the Gclc KO lenses (Table 3). To further explore the possibility that PAX6 transcriptional regulation was altered in the KO lenses, the transcriptome of the lenses from KO mice were compared with the transcriptome of lenses from Pax6 KO mice⁴¹. Notably, a strong overlap was found between genes differentially expressed in Pax6 KO and Gclc KO datasets (odds ratio = 9.55, P < 0.0001), with 31% (310/1013) of Pax6 KO DEGs also showing differential expression in the Gclc KO. Among these genes, a strong positive correlation between log₂fold changes (log2FC) between these two KO datasets (R² = 0.58, P = 3.37E-60; Fig. 4D) was observed. Since PAX6 directs lens development through the regulation of gene expression⁵⁵, a focus was placed on lens identity genes differentially expressed in both Gclc KO and Pax6 KO lenses (Table 4, Fig. 4D, orange dots). This revealed a stronger correlation between these datasets (R² = 0.93, P = 1.60E-06; Fig. 4D). Taken together, these data suggest that the dysregulation of lens cell identity expression in the Gclc KO may be due to changes to PAX6 transcriptional regulation.

Table 3. Top overrepresented transcription factor binding sites in the lenses of Gclc KO mice as determined by TRANSFAC analysis.

P values were calculated by Benjamini-Hochberg method.

Factor name	Matrix	Overrepresentation score	P value
PAX factors	V$PAX_Q6	18.11	3.63E-05
FOX factors	V$HNF3B_Q6	8.05	6.29E-04
HIC1	V$HIC1_08	5.36	2.13E-03
MAZ group	V$MAZR_01	3.37	8.55E-17
PBX	V$PBX_Q3	2.55	2.13E-03

Table 4. Lens identity genes differentially expressed in the lens of both Gclc KO and Pax6 KO mice.

Differentially expressed genes are presented as log2 fold change (log2FC) of transcripts from the lenses of Gclc KO mice or Pax6 KO mice compared with the lenses of respective control mice.

Gene name	Gclc KO (log2FC)	Pax6 KO (log2FC)
Crybb3	−1.34	−1.15
Foxp2	1.78	0.85
Lhx2	1.26	1.37
Lsamp	−1.16	−0.61
Mxd1	−1.41	0.58
Six6	1.88	1.84
Slc24o4	−1.29	−0.58
Sox11	1.26	1.20
Sox9	2.02	1.29
Vsx2	2.53	2.14
Zic2	1.66	1.53

PAX6 transcriptional regulation is altered by changes in intracellular GSH content

By focusing on altered expression of PAX6-regulated genes (as identified by a previous report⁵⁶), we discovered that the PAX6 gene regulatory network was altered despite no decrease in Pax6 expression (Fig. 5A, Supp. Fig. 5A). To further analyze this network, we utilized IPA to leverage our full set of DEGs and predict genes whose function may have been impaired in Gclc KO mice. Of those identified, we found that PAX6 activity was predicted to be impaired (Fig. 5B). This prediction involves increased expression of genes repressed by PAX6 (i.e., Snca, Vsx2) and decreased expression of genes activated by PAX6 (i.e., MAF, CRYAA/CRYAA2) (Figs. 5A, B). IPA also predicted that the changes in the PAX6 gene regulatory network would lead to lens disorders (Fig. 5B), which was observed in Gclc KO mice. PAX6 transcriptional regulation can be altered through several mechanisms independent of PAX6 expression, including subcellular localization and post-translational modification (PTM)^55,57,58. Immunohistochemical staining of PAX6 early in lens development (E14.5) and in postnatal mice (P3) revealed PAX6 cellular and subcellular localization to be unchanged in the lenses of KO mice (Figs. 5C, D). We then assessed the impact of low GSH on PAX6 transactivation activity in cultured HEK293T cells using BSO, a specific inhibitor of GCLC. BSO-induced depletion of GSH from HEK293T cells resulted in a concentration-dependent decrease in PAX6 transactivation activity (Fig. 5E). This observed decrease in PAX6 transactivation activity was not due to a reduction in PAX6 expression, which remained constant even at the highest levels of BSO treatment (Supp. Fig. 6B). We would note, however, that while BSO treatment reduced firefly luciferase activity, it also increased renilla luciferase activity (Supp. Table 3).

Figure 5. Impaired PAX6 function is predicted to cause a lens disorder and is influenced by intracellular GSH concentration.

(A) Volcano plot indicating the expression of PAX6-regulated genes that were differentially expressed in the lenses of Gclc KO mice relative to CON mice. PAX6-regulated genes are highlighted. Differentially expressed genes were classified as |log₂FC| ≥ 1; Benjamini-Hochberg adjusted P (Padj) < 0.05. Colors: blue, expression downregulated; grey, no significant change in expression; red; expression upregulated. (B) Changes in PAX6-regulated genes predicts impaired PAX6 function. Colors: blue, predicted reduced activity; green, reduced gene expression; orange, predicted lens disorder; red, increased gene expression. Shapes: described by key located to the right of the figure. Diagram created using Ingenuity Pathway Analysis software (Qiagen, Hilgen, Germany). (C-D) Representative images of immunohistochemical staining of ocular PAX6 expression in E14.5 (C) and P3 (D) -aged CON and KO mice. Whole eye (top row) and lens (middle and bottom rows) expression of PAX6 are shown as brown stain. Images are representative of results obtained in all 3 mice per genotype. Regions contained in boxes are shown at higher magnification, 400X (bottom row). Abbreviations: L, lens; R, retina. (E) Impact of BSO treatment on PAX6 transactivation activity in HEK293T cells transfected with a PAX6 luciferase reporter construct. Data are presented as the mean of the ratio of measured firefly luciferase (FLuc) to renilla luciferase (RLuc) and associated s.d. (n ≥ 6 wells from at least three independent plates). * P < 0.05, one-way ANOVA with post-hoc Dunnett’s test, compared to 0 µM BSO.

Discussion

In this study, the conditional deletion of Gclc from surface ectoderm-derived ocular tissues from early in eye development was demonstrated to cause an overt microphthalmia phenotype. Given the marked lens deformities observed in KO mice, the molecular analysis was limited to this tissue. RNA-sequencing of the lenses from KO mice revealed several unexpected findings: i) induction of an inflammatory response (the lens has long been thought of as immune privileged⁵⁹), ii) disruption in the expression of lens identity genes (i.e., upregulation in the expression of lens epithelial genes and downregulation in the expression of lens fiber cell genes), and iii) dysregulation of PAX6-regulated genes in a manner that may impair lens development. The importance of GSH for regulating PAX6 transcriptional regulation was then tested by in vitro experiments in which impaired GSH biosynthesis caused a decrease in PAX6 transactivation activity. Collectively, these results reveal unexpected insights about lens biology, including the requirement of GSH for eye and lens morphogenesis and normal PAX6 function.

The importance of GSH and cellular redox status for maintenance of lens health (specifically, optical clarity) has been known for many decades^19,60. Recently, much work has leveraged the Cre-Lox system to reduce GSH levels from only the lens through the conditional deletion of Gclc, a mouse model termed the LEGSKO model²². LEGSKO mice display reduced lens GSH content, elevated levels of lens reactive oxygen species (ROS), and cataract formation beginning as early as 4 months of age. Analysis of the LEGSKO lens transcriptome has revealed that GSH deficiency induces expression of detoxifying genes and activation of epithelial-mesenchymal transition (EMT) signaling⁶¹. Proteomic analysis of LEGSKO lenses further confirmed the activation of EMT signaling, changes in stress response proteins, and revealed a loss of lens-specific markers⁶². Our Gclc KO mice show induction of EMT (i.e., downregulation of Cdh1 and Spock2 expression and upregulation of Ier3, Tgfbi, Tnc (Suppl. Spreadsheet)) in a manner consistent with a similar GSH depletion-induced EMT⁶¹. In addition, the transcriptome of lenses from Gclc KO mice displays a strong positive correlation with the transcriptome of lenses from an EMT mouse model (Supp. Fig. 7). Interestingly, analysis of the LEGSKO mouse model revealed that this EMT phenotype is promoted by the activation of Wnt signaling⁶³. However, in the lenses of our KO mice, we failed to find evidence which would indicate that Wnt signaling is activated, namely we failed to find upregulation of Wnt10a expression, as previously reported⁶¹. Further differences between our Gclc KO mice and the LEGSKO mice were revealed by the observation that the LEGSKO mice fail to display a microphthalmia phenotype, whereas our model described herein does. This discrepancy may be explained by the observation that LEGSKO mice and our KO mice have Gclc deletion at different developmental times and in different tissues. The LEGSKO mouse model relies on the MLR10 Cre transgene to drive gene deletion in lens-specific epithelial and fiber cells at ≈E10.5^20,64 whereas our KO mouse model employs the Le-Cre transgene to drive gene deletion in surface ectoderm-derived tissues at ≈E9²⁶. It is conceivable that the mechanisms that maintain redox homeostasis are different in lens cells between E9 and E10.5, and this difference may allow E10.5 lens cells to better maintain redox homeostasis and prevent the oxidative stress-induced impairment of PAX6 transcriptional regulation. While experiments are needed to support this hypothesis in the lens, much evidence indicates that early eye development is more susceptible to malformations caused by exposure to oxidative stress-inducing teratogens than later stages of eye development⁶⁵. CRE expression in MLR10 Cre transgenic mice is restricted to only the lens whereas Le-Cre transgenic mice express CRE in surface ectoderm-derived tissues⁴⁰. The more promiscuous expression of CRE by the Le-Cre transgene may further drive differences in the phenotype between our KO mice and the LEGSKO mice. Furthermore, it is possible that the phenotypic response in these other structures (i.e., in the eyes of KO mice) could be due to a disruption in the exchange of GSH between them^23,66.

Mice that have at least one Le-Cre transgene allele may develop microphthalmia regardless of any floxed alleles^40,48–50. Mice homozygous and hemizygous for the Le-Cre transgene display variable penetrance microphthalmia in a strain-dependent manner^48,49. FVB/N inbred mice served as the original background for the Le-Cre transgene²⁶ and Le-Cre hemizygous transgenic mice maintained on this background exhibit normal eye development⁴⁰. In contrast, a microphthalmia phenotype appears in Le-Cre hemizygous transgenic mice as the genetic contribution of FVB/N is reduced⁴⁸. Given that the Le-Cre transgene utilizes the Pax-P0 promoter, some have postulated that it may deplete cofactors required for endogenous Pax6 expression⁴⁸. However, a recent study was unable to confirm this hypothesis⁴⁰. Instead, it was found that the microphthalmia phenotype in Le-Cre homozygous mice was caused by changes in the expression of genes involved in the negative regulation of cell proliferation and cell growth, and positive regulation of apoptosis pathways⁴⁰. Regardless of its cause, it is critical that any study utilizing the Le-Cre transgene control for the effect of the Le-Cre alone. Isolating the Le-Cre transgene on the mixed B6/FVB background (as used throughout this study) confirmed that the Le-Cre transgene alone causes microphthalmia but that Gclc deletion during early eye development exacerbates this microphthalmia phenotype. Critically, we failed to find evidence that the transcriptional changes observed in the lenses of KO mice were due to the Le-Cre transgene; GSH was not reduced in the lenses of CRE control mice, no correlation existed between the transcriptomes of lenses from CRE and KO mice, and reductions in the expression of lens identity proteins was only observed in the lenses of KO mice. Furthermore, our data supports the previous suggestion⁴⁰ that the Le-Cre microphthalmia phenotype is not influenced by changes in PAX6 expression. Finally, our cell culture work was conducted in a system that was independent of any CRE protein. Collectively, these data suggest that the mechanism proposed here is independent from the Le-Cre transgene.

Hydrogen peroxide (H₂O₂) is an abundant ROS molecule in the developing lens equatorial epithelial cells⁶⁷ and it serves as the primary redox-regulated signaling molecule⁶⁸. The presence of high amounts of H₂O₂⁶⁷ and Gclc expression in equatorial epithelial cells may indicate that redox signaling is important for the differentiation process of lens epithelial cells into fiber cells. Indeed, many of the signaling pathways critical for lens development⁵⁴ produce H₂O₂⁶⁸. Further, lens epithelial cells are known to be particularly susceptible to damage by oxidative stress^69,70. It is likely that the disrupted lens development observed in KO mice results from oxidative stress-induced impairment of the normal epithelial cell to fiber cell differentiation process. This proposal is supported by the observation that ROS is increased in the lenses of KO mice and that key lens fiber cell genes, such as crystallins, transcription factors (e.g., Maf, Pitx3) and structural proteins (e.g., Bfsp2), were downregulated, which is further reflected by the gene ontology analysis which revealed that terms, such as ‘lens development in camera-type eye’, ‘eye development’, ‘lens fiber cell differentiation’, were enriched. A recent report indicates a role for HSF4 and PLAAT-family phospholipases in the degradation of lens organelles during lens fiber cell differentiation⁷¹. Indeed, we observed a failure of proper lens fiber cell differentiation and a downregulation in the expression of both Hsf4 and Plaat3 (Suppl. Spreadsheet). However, the present results provide compelling evidence that PAX6 function is impaired in a manner that would impair lens development.

PAX6, the “master eye regulator”, is a member of the paired box (PAX) gene family⁷². PAX6 regulates the expression of target genes (gene regulatory network) and, ultimately, eye development through DNA-binding at gene promoter regions⁷³. Pax6 expression is required for lens epithelial cells to exit the cell cycle and differentiate into lens fiber cells⁷⁴. A failure of lens epithelial cells to properly differentiate into lens fiber cells is similarly observed in our KO mice, further suggesting that PAX6 transcriptional regulation was impaired in KO mice.

We propose that PAX6 transcriptional activity is redox-regulated; evidence is building to support this. An investigation into the regulation of the aldose C gene expression in the rat brain suggested that PAX6 DNA-binding is redox-regulated by the oxidation of cysteine residues⁷⁵. Additionally, studies on another PAX family member, PAX8⁷³, have identified redox regulation of PAX8 by cysteine oxidation^76–78. The mechanism by which redox state regulates PAX8 is through glutathionylation of a cysteine found within the paired domain⁷⁸, which sterically hinders DNA-binding capabilities⁷⁶. The susceptible cysteine residues in PAX8 are conserved in the same location in PAX5 and PAX6 which suggests that they may be similarly redox-regulated by glutathionylation⁷⁶. PAX5 transactivation activity and DNA binding is redox-regulated⁷⁹, which further suggests that PAX6 would be redox-regulated. In addition to glutathionylation, susceptible cysteine residues may be subject to multiple forms of oxidation including sulfenic, sulfinic and sulfonic acids⁸⁰, disulfide formation, S-nitrosylation and several others⁸¹. Any of these forms of cysteine oxidation may be capable of impairing PAX6 transcriptional regulation upon conditions of low GSH. Unfortunately, our data is unable to determine which oxidation may be occurring on PAX6 to impair its transcriptional regulation. However, it is clear from the previous reports and our data that PAX6 is redox-regulated, likely by the oxidation of cysteine residues within the PAX6 paired domain.

It is unlikely that oxidative stress is producing the observed gene expression changes and microphthalmia through mechanisms independent of PAX6. While our TRANSFAC analysis revealed the overrepresentation of transcription factors (other than PAX6) that are required for normal eye and lens development, namely FOX and PBX, it is unlikely that these other transcription factors are the primary cause of the impaired lens development in the Gclc KO mice for several reasons. First, FOXE3, the FOX family member that is critical for lens development, is downstream of PAX6⁵³. Second, while, MEIS-PBX is required for the expression of PAX6 in the pancreas⁸², PAX6 expression was not decreased in the lens of the Gclc KO mice used here. This indicates that the change in the PAX6 gene regulatory network is independent of MEIS-PBX. Lastly, while our RNA-seq data revealed a decrease in RNA coding for proteins critical for intracellular communication and equilibration (e.g., connexins, adherins) and an inflammatory response, it is unlikely that this would be largely responsible for the impaired lens development observed in Gclc KO mice. While the genes encoding for connexin (CX) 37 (Gja4), CX46 (Gja3) and CX50 (Gja8) are downregulated in the lenses of Gclc KO mice and mice deficient for CX50 have impaired lens development and reduced lens volume⁸³, the phenotype in CX50-deficient mice is distinct from that in Gclc KO mice⁸⁴, i.e., malformations in only nuclear fiber cells vs. malformations in both cortical and nuclear fiber cells. Moreover, we suggest that potential activation of the immune system is secondary to the dysgenesis of the lens in Gclc KO mice, as was reported in a N-cadherin KO mouse model⁸⁵.

Microphthalmia cases in human have been associated with several environmental exposures^5–11. Despite this, little is understood regarding the mechanism(s) by which these exposures may cause microphthalmia. It is known that these exposures can disturb the cellular redox state in other systems; thus, it is likely that can similarly disturb the cellular redox state in developing lens cells. In the present study, we describe the first mouse model (Gclc KO) of microphthalmia that is induced by depleted GSH and contains increased ROS. Given the high homology of PAX6 between mouse and human⁷² and our data from HEK293T cells, we anticipate more reports of redox-regulation of PAX6 in humans. Finally, we expect that this mouse model will be a useful tool for further elucidating how redox status regulates lens development and microphthalmia and will enable the development of pharmacological agents that can prevent or reverse microphthalmia by preserving or restoring cellular redox status in the lens.

Supplementary Material

1. Supplemental Figure 1. Gclc deletion potentiates the microphthalmia phenotype of the Le-Cre transgene.

(A) Agarose gel (2%) separation of genomic DNA obtained from ear tissue of P14 Gclc^f/f/Le-Cre^−/− (CON), Gclc^f/f/Le-Cre^Tg/-(KO), and Gclc^wt/wt/Le-Cre^Tg/-(CRE) mice by amplification of the Gclc wild- type (Gclc(wt)) or Gclc floxed (Gclc(f)) alleles, or the Le-Cre transgene (Cre(Tg)). Two PCR reactions were needed to genotype mice to detect: i) the presence of Gclc wt and Cre transgene alleles, ii) the Gclc f allele. (B) Combined (left and right) eye weights from P21-aged CON, KO and CRE mice. Each point represents results from an individual mouse. Horizontal lines indicate the mean and error bars indicate s.d.. * P < 0.05, ANOVA with Tukey correction, compared to other group (as indicated). (C) GSH levels in the lenses of P21-aged CON (black bar), KO (grey bar), and CRE (white bar) mice. Data are presented as the mean and associated s.d. from 3 mice. * P < 0.05, ANOVA, with Tukey correction, compared to other group (as indicated). (D) Representative images of hematoxylin & eosin-stained eyes from one CRE animal aged P50 (A’-F’). Lens (black box) in B’ is shown at higher magnification in C’. Cornea (black box) in D’ is shown at higher magnification in E’. Abbreviations; AC, anterior chamber; C, cornea; I, iris; L, lens; R, retina. Images are representative of results obtained in all 3 mice of each group.

3. Supplemental Figure 2. Principle component analysis of differentially expressed genes in lenses from P1-aged CON and KO mice.

Principle component analysis was conducted in the lenses of three P1-aged CON (red symbols) and KO (blue symbols) mice by analyzing the 500 most variable genes.

4. Supplemental Figure 3. Correlation between RNA-seq and qRT-PCR data in lenses of P1-aged KO mice.

The relationship between lens gene expression levels (shown as log₂ fold-change (log2FC) of expression in CON mice) of select genes identified by RNA-seq and those measured by qRT-PCR data were explored using Pearson correlation. Each point represents one gene. The coefficient of determination (R²), P value and trendline are shown.

5. Supplemental Figure 4. Correlation between KO and Le-Cre RNA-seq dataset.

The relationship between lens gene expression levels (shown as log₂fold change (log2FC) of expression in wildtype FVB/N mice (for Le-Cre) or Gclc^f/f;Le-Cre^Tg/- mice (for KO)) identified by RNA-seq datasets for only genes differentially expressed in both KO and Le-Cre lenses were explored using Pearson correlation. Le-Cre data retrieved from ⁴⁰. Each point represents one gene. The coefficient of determination (R²), P value and trendline are shown.

6. Supplemental Figure 5. Gclc deletion causes loss of lens cell identity markers.

(A) Upper image: Western blot analysis of α-crystallin, γ-crystallin, PITX3 expression in total protein lysates collected from the lenses of P1 aged CON, CRE and KO mice. Each lane represents the protein collected from left and right eye lenses of one animal. A total of 3 mice were used per genotype. Lower image: Protein, detected by Ponceau S stain, was used as a loading control. (B) Quantitative expression levels of lens α-crystallin, γ-crystallin and PITX3 expressed as relative to CON. Data are presented as the mean and associated s.d. from 3 mice. # P < 0.1, * P < 0.05, *** P < 0.001, one-way ANOVA with Tukey correction, compared to other group indicated. (C) RT-qPCR analysis of Foxe3 expression in the lenses of P21 aged CON, CRE and KO mice. Data represent the mean and associated s.d. from 3 mice. ** P < 0.01, one-way ANOVA with Tukey correction, compared to other group indicated.

7. Supplemental Figure 6. Expression of PAX6 in GSH-deficient lenses and HEK293T cells.

Western blot analysis and quantification by densitometric analysis of PAX6 protein expression in the (A) lenses of P1 CON, CRE and KO mice, and (B) HEK293T cells treated with 0 (control (CON)), 500 or 1,000 µM BSO for 48 h. PAX6 expression was normalized to the expression of an internal control gene, GAPDH. Quantitative expression levels of PAX6 expressed as relative to control indicated. Data are presented as the mean and associated s.d. from 3 mice or replicate wells. # P < 0.1, * < 0.05 one-way ANOVA with Tukey correction, compared to other group indicated.

8. Supplemental Figure 7. Correlation between KO and a RNA-seq dataset.

The relationship between lens gene expression levels (shown as log₂fold change (log2FC) in expression) in epithelial-mesenchymal transition model (EMTm) mice and KO mice. RNA-seq datasets were used to identify genes differentially expressed in the lenses of both KO and EMTm mice which were then evaluated using Pearson correlation. EMTm data was retrieved from ⁴². Each point represents one gene. The coefficient of determination (R²), P value and trendline are shown.

9

Supplemental Table 1. RNA-seq expression values of select genes that were validated by qRT- PCR. Data shown as log₂fold change (log₂FC) from the lenses of KO mice compared with CON mice.

Supplemental Table 2. Gene expression values of genes significantly differentially regulated in both KO and Le-Cre lenses (retrieved from a previous publication²¹). Data shown as log₂ fold change (log₂FC) from the lenses of KO or Le-Cre mice compared with respective control mice.

Supplemental Table 3. Table detailing the effect of BSO on Firefly and Renilla luciferase activities. Values calculated from the Firefly and Renilla luciferase activities are presented.