Altered Cell Clusters and Upregulated Aqp1 in Connexin 50 Knockout Lens Epithelium

Abstract

Purpose: To characterize the heterogeneity and cell clusters of postnatal lens epithelial cells (LECs) and to investigate the downstream targets of connexin 50 (Cx50) in the regulation of lens homeostasis and lens growth. To determine differentially expressed genes (DEGs) in the connexin 50 knockout (Cx50KO) lens epithelial cells that shed light on novel mechanism underlying the cataract and small size of the Cx50KO lenses.

Methods: Single-cell RNA sequencing (scRNA-seq) of lens epithelial cells isolated from one-month-old Cx50KO and wild-type (WT) mice were performed. Differentially expressed genes were identified, and selected DEGs were further studied by quantitative real-time PCR (RT-qPCR) analysis and Western blot analysis.

Results: The expression profiles of several thousand genes were identified by scRNA-seq data analysis. In comparison to the WT control, many DEGs were identified in the Cx50KO lens epithelial cells, including growth regulating transcriptional factors and genes encoding water channels. Significantly upregulated aquaporin 1 (Aqp1) gene expression was confirmed by RT-qPCR, and upregulated AQP1 protein expression was confirmed by Western blot analysis and immunostaining both in vivo and in vitro.

Conclusions: Lens epithelial cells exhibit an intrinsic heterogeneity of different cell clusters in regulating lens homeostasis and lens growth. Upregulated Aqp1 in Cx50KO lens epithelial cells suggests that both connexin 50 and AQP1 likely play important roles in regulating water homeostasis in lens epithelial cells.

The avascular lens is composed of two types of cells: lens epithelial cells (LECs) and fiber cells. The epithelial cells form a monolayer covering the anterior hemisphere of a bulk of elongated fibers.^1,2 Lens growth relies on epithelial cells undergoing proliferation, migration, and differentiation to produce new elongating fibers at the lens equator throughout life.³ Therefore, the proliferation, differentiation, and homeostasis of lens anterior monolayer epithelium are critical for regulating lens growth, lens size, and transparency, which are required for focusing and transmitting a clear light image onto the retina.² Lens epithelium is traditionally divided into three anatomic zones: the central zone (CZ) around the anterior pole region, the pre-germinative zone (PGZ), and the germinative zone (GZ) near the lens equator.² The maximum mouse lens growth occurs during the first postnatal week when mitotic cells appear in all zones of lens epithelium and also when mouse eyes remain closed; mouse eyelids open at the age of around two weeks, and the mitotic cells are restricted to the GZ after the age of three weeks.^4,5 The epithelial cells in the CZ and PGZ zones are quiescent and presumably regulate the homeostasis of the whole lens to maintain its transparency.⁶ The gene network for both signal transduction pathways and transcriptional factors that control the proliferation and differentiation of lens epithelium have been extensively studied,^7–15 but less is known about the heterogeneous properties of lens epithelial cells from the anterior pole to the equator. The gene network underlying the heterogeneity of lens epithelial cells is not well defined at molecular and cellular levels. It is also poorly understood how lens epithelium provides multiple functions to regulate lens growth and maintain homeostasis of both the epithelial cells and underlying fiber cells in response to various external stimuli over the course of postnatal lens development and aging as well as during accommodation. 

Intercellular gap junction channels play essential roles in controlling lens size and maintaining the lens homeostasis by providing pathways for transporting second messengers, metabolites, ions, and fluid between lens epithelial cells and fiber cells.^4,6,16–19 Disrupted gap junction channels by genetic mutations cause cataract formation and/or smaller lens size in both humans and mouse models.^18,20–24 The rodent and human lenses mainly use three isoforms of connexin: Cx43 (Gja1), primarily in the lens epithelium; Cx46 (Gja3), mainly in the lens fiber cells; and Cx50 (Gja8), in both lens epithelial cells and fibers.^25–27 Both Cx46 knockout and Cx50 knockout mice develop cataracts, whereas only Cx50 knockout mice display smaller lens size.^18,23,24 The heterozygous Cx46 knockout and Cx50 knockout mice develop transparent lenses. Compared to the wild-type control, about half the amount of Cx50 proteins was detected in the heterozygous lens,²⁴ and the protein level and distribution of both Cx46 and Cx43 in Cx50 knockout lenses are comparable to those in wild-type lenses.²³ Cx50 is also the most highly expressed connexin isoform in the lens²⁸; dysfunctional connexin 50 affects proper lens growth and fiber cell differentiation.^23,24 The Cx50KO mice exhibit microphthalmia with 40% smaller lenses caused by reduced LEC proliferation.^4,23,24,29 Previous study has suggested that Cx50 plays an essential role in lens growth by regulating the LEC proliferation and differentiation through the PI3K signal transduction pathway.³⁰ However, the mechanism for how Cx50-mediated gap junction communication determines the properties of lens epithelium remains elusive. For example, the role of gap junctional communication in the regulation of ion flow or water homeostasis is unclear. 

A microcirculation flow model has been proposed for the delivery of ions, nutrients, and water from the lens epithelial cells into deeper fiber cells through gap junction coupling, as well as for the removal of ions, metabolic waste, and water out of the lens fibers through the ATPase pumps on the lens epithelial cells.³¹ Aquaporins (AQP) are transmembrane proteins that function as water channels for water transport.^32,33 The lens mainly uses three aquaporins—AQP0, AQP1, and AQP5. These aquaporins mediate the passive diffusion of water across lens cell membranes,³⁴ and proper water transport is required for maintaining lens transparency and refractive properties.³⁵ 

To gain insight into the heterogeneity and functional properties of lens epithelium and to discover novel regulatory mechanisms of lens growth and homeostasis, single-cell RNA sequencing (scRNA-seq) was performed on lens epithelial cells isolated from one-month-old Cx50KO and age-matched C57BL/6J wild-type (WT) mice. The scRNA-seq data reveal seven unique cell clusters in the WT lens epithelium and eight distinct cell clusters in the Cx50KO epithelium. Differentially expressed genes (DEGs) for both upregulated and downregulated genes were characterized in the Cx50KO epithelial cells, including growth regulating transcriptional factors in the small Cx50KO lenses. Moreover, upregulated aquaporin 1 (Aqp1) was further characterized as a specific DEG in Cx50KO lens epithelial cells in vivo and in vitro whereas both Aqp5 and Gja3 remain unchanged. This work suggests a direct association between Cx50 function and AQP1 in the regulation of water homeostasis in lens epithelial cells. 

Material and Methods

Mice

All experimental procedures were approved by the Animal Care and Use Committee at the University of California, Berkeley and were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Age-matched connexin 50 knockout mice²⁴ and C57BL/6J wild-type mice were used for all experiments. 

Lens Epithelial Cell Isolation and Single-Cell RNA Sequencing

The lens epithelial cells were isolated based on a previously published procedure.³⁶ Mice at one month of age were used for lens primary epithelial cell culture. Briefly, lenses were dissected from enucleated eyeballs of CO₂-euthanized mice in Advanced DMEM/F-12 media (cat. no. 12634-010; Gibco, Thermo Fisher Scientific, Waltham, MA, USA). The lenses were cleaned by removing the surrounding tissues and were incubated in 0.05% Trypsin-EDTA (cat. no. 25300-054; Gibco, Thermo Fisher Scientific) at 37°C for 10 minutes. Lens capsules were peeled off and transferred into 2 U/mL dispase (cat. no. D4693; Sigma-Aldrich Corp., St. Louis, MO, USA) for five minutes at 37°C. Then equal volumes of 10×TrypLE (cat. no. A12177-02; Gibco, Thermo Fisher Scientific) were added for 10 minutes. The suspended epithelial cells from lens capsules were spun down at 1000 rpm for five minutes, and the cell pellet was resuspended in PBS. Collected lens epithelial cells pooled from three WT littermates and four Cx50KO littermates were used for single-cell RNA sequencing in the Functional Genomics Laboratory at UC Berkeley. Single-cell cDNA libraries were prepared from the isolated cells using the Chromium Single Cell 3′ System according to the manufacturer's instructions. The protocol from 10× used v3.1 chemistry with the following modification: the cell suspension was directly added to the reverse transcription master mix, along with the appropriate volume of water to achieve the approximate cell capture target. Completed libraries were sequenced on Illumina 100PE HiSeq4000 to produce 10× recommended reads. 

Data Analysis of ScRNA-seq

Sequencing data was processed through the 10× Genomics Cell Ranger (Cell Ranger Version 2.2.0) Analysis Pipelines. Reads were aligned to the mouse reference genome GRCm38.3 (mm10) and were quantified using the mouse reference index provided by 10× genomics. The cloupe files for WT and Cx50KO were generated, and WT-KO-aggreg-cloup file were generated using the cellranger aggr function. The sequencing data in cloupe files were then visualized and analyzed using the desktop application 10× Loupe Cell Browser (Loupe Browser 6.1.0). Graph-based cell clustering was presented in the result. The violin plots in Figures 3 and 4 were generated using the Loupe Browser 6.1.0. 

Integrated Data Analysis of WT and KO ScRNA-seq Datasets

Doublets were removed from WT and CX50KO samples using DoubletFinder v2.0.4. The filtered cell matrix was analyzed with Seurat V.5.0.1 using command line interface, R version 4.2.3 and R Studio Version 2023.12.0. Cells with more than 250 genes, fewer than 5000 genes, and less than 15% mitochondrial reads were retained. Filtered WT and KO single-cell datasets were integrated and normalized by applying the IntegrateData function. The integrated dataset was scaled and subjected to PCA, UMAP, determined using Seurat FindClusters and Louvian algorithm, with the resolution parameter set to 0.3 to avoid over-clustering. The integrated dataset was divided into seven subpopulations. Cluster markers were selected using FindAllMarkers function, and pct1 and pct2 difference was used as criteria for selecting cluster markers. The DEGs of each cluster between KO and WT were identified using FindMarkers and the Wilcoxon Rank Sum test. Log2-fold changes of average expression of DEGs between two genotypes were calculated. P values were adjusted by Bonferroni correction using all features in the dataset. DEGs, which have adjusted P values < 0.05, were considered statistically significant. 

Quantitative Real-Time PCR Analysis

Given the smaller lens size and cataracts in KO mice, we hypothesized that some of the downstream candidate genes identified from the DEG list between WT and KO must be involved in the regulations of cell proliferation or lens homeostasis. The comparative results from large scRNA-seq datasets may indicate false-positive gene candidates in the DEG list. Therefore, multiple testing correction was performed using the false discovery rate to adjust the P values, and the Benjamini-Hochberg method was used to implement the false discovery rate. We selected some gene candidates based on two criteria: (1) have adjusted P values < 0.05 and (2) are known to be related to the regulation of cell proliferation or lens homeostasis, as reported in the literature, to examine for significant changes in their expression levels using qPCR assays. For quantitative real-time PCR (RT-qPCR) analysis, the following primers were used. Aqp1: forward 5′-CTG GCG ATT GAC TAC ACT GG and reverse 5′-AAG TCA TAG ATG AGC ACT GCC; Mip: forward 5′- CAG CTG TCC GAG GAA ACC and reverse 5′- CCT CTC GTC GTA TGT AGC AAA G; Aqp5: forward 5′- CTC CCC AGC CTT ATC CAT TG and reverse 5′- ACC CAG AAG ACC CAG TGA G; Gja1: forward 5′- TTC CTT TGA CTT CAG CCT CC and reverse 5′- CGT GGA GTA GGC TTG GAC; Gja3: forward 5′- GGA CCC GGT TAG CTG TTG and reverse 5′- CGC CCC TAA CAC CAG AAT G; Lim2: forward 5′- CAA TCC ACC TTC ACC CGT C and reverse 5′- AAG CGC CAG TCT CCA AAG; Pdpn: forward 5′- GGA GGG CTT AAT GAA TCT ACT GG and reverse 5′- GGT TGT ACT CTC GTG TTC TCT G; Btg1: forward 5’- GTC CTT CAT CTC CAA GTT CCT C and reverse 5′- AAC GGT AAC CTG ATC CCT TG; Btg2: forward 5′- TCA AGG TTT TCA GTA GGG CG and reverse 5′- TGT GGT TGA TAC GGA TAC AGC; Tob1: forward 5′- GCC TCA ACG AGA AGT CTT TTG and reverse 5′- GGG CCG ATG CAT TTT AAC TTG; Socs3: forward 5′- CCT ATG AGA AAG TGA CCC AGC and reverse 5′- TTT GTG CTT GTG CCA TGT G; and GAPDH: forward 5′- TGA AGG TCG GTG TGA ACG GAT TTG GC and reverse 5′- CAT GTA GGC CAT GAG GTC CAC CA. Lens epithelial cells, collected from four lenses of two littermates of Cx50KO or age-matched WT mice, were used for isolating total RNA samples; cDNA samples were generated by using the iScript Reverse Transcription Supermix (Bio-Rad Life Science, Hercules, CA, USA). Quantitative gene expression profiles of given genes were determined using the SsoAdvanced Universal SYBR Green Supermix (Bio-Rad Life Science) and the CFX96 Real-Time System (Bio-Rad Life Science) with their specific primers. Three independent cDNAs samples (n = 3), generated from three separated mouse littermates for both WT and KO genotypes, were used for the RT-qPCR assay of selected genes, and triplicate wells for each individual cDNA sample on a PCR plate were used for a given gene analysis. The gene expression (ΔΔCq) was normalized to the amount of GAPDH, and a two-sample Student t-test was used for statistical analysis. Representative sets of data were shown in the result. 

Western Blotting and Immunostaining

Due to the low protein amount of epithelial cells attached to the capsule of a single lens, each in vivo LEC sample was pooled from four lens capsules of two littermates and homogenized in the lysis buffer, containing 20 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, and proteinase inhibitors (cat. no. 11836153001; Roche, Basel, Switzerland), for Western blot analysis. Primary cultured lens epithelial cells were prepared from both WT and Cx50KO lenses based on our previous published methods.³⁶ Cultured lens epithelial cells were harvested in the lysis buffer and lysed in a Minilys (Bertin Technologies, Montigny-le-Bretonneux, France). Homogenized samples were spun in a centrifuge at 14,000 rpm for five minutes, and the supernatant was collected and measured for protein concentration (Pierce Coomassie Protein Assay Kit; Thermo Fisher Scientific). These epithelial cell protein samples were mixed with an equal volume of 2×loading buffer (120 mM Tris-PO4 (pH 6.8), 4% SDS, 20% glycerol, 10% 2 mercaptoethanol, and 0.002% bromophenol blue), and equal amounts (∼30 µg protein) of the samples were loaded onto the 4% to 12% Bis-Tris gel (NuPAGE; Invitrogen, Carlsbad, CA) for separation, and the gel was transferred to the PVDF membrane (Bio-Rad Laboratories, Hercules, CA). The membranes were blocked with 5% non-fat milk and incubated with antibodies at 4°C overnight. After three 5-minute washes in TBST buffer (0.15 M NaCl, 0.05 M Tris, pH 7.4, 0.1% Tween 20), the membranes were incubated in HRP-conjugated secondary antibodies for one hour at room temperature. After three washes with TBST, blots were developed with chemiluminescence (SuperSignal West Pico Chemiluminescent substrate kit; Thermo Fisher Scientific), and protein bands were visualized by an Azure Biosystems c600 imager (Azure Biosystems, Inc., Dublin, CA, USA). 

For immunostaining, cultured lens epithelial cells in glass bottom dishes were washed with PBS and fixed in 4% paraformaldehyde in PBS for 10 minutes. Fixed cells were washed with PBS for three times and incubated in the blocking solution (3% BSA, 3% goat serum, and 0.3% Triton X-100 in PBS) for one hour at room temperature, followed by overnight incubation in specific primary antibodies at 4°C. The cells were washed with PBS for three times and were then incubated with fluorescently labeled secondary antibodies for two hours at room temperature; after washing with PBS for three times, antifade mounting medium with DAPI (H-1200; Vector Laboratories, Burlingame, CA, USA) was applied before imaging acquisition. Fluorescent images were collected with a Zeiss LSM 700 confocal microscope (Zeiss, Oberkochen, Germany). 

Specific antibodies were used for Western blots and immunostaining: a rabbit polyclonal antibody against the C-terminal region of Cx50 connexin (a gift provided by Dr. M. J. Wolosin, Mount Sinai School of Medicine, New York, NY, USA); rabbit polyclonal MP26 antibody (a gift provided by Joseph Horwitz at Jules Stein Eye Institute, Los Angeles, CA, USA); rabbit monoclonal Aquaporin 1 antibody (cat. no. ab168387; Abcam, Waltham, MA, USA); rabbit polyclonal Aquaporin 5 antibody (cat. no. PA5-97290; Invitrogen, ThermoFisher Scientific); Golden Syrian Hamster monoclonal antibody for Podoplanin (cat. no. NB600-1015; Novus Biologicals, Centennial, CO, USA); and β-actin mouse monoclonal antibody (Sigma-Aldrich, St. Louis, MO). Antibodies were diluted 1:100 for immunostaining and 1:1000 for Western blots. 

Results

Single-Cell RNA Sequencing of Lens Epithelial Cells

We have confirmed that Cx50KO mice at the C57BL/6J background develop very mild nuclear cataracts with smaller lens size as previously reported (Fig. 1A).^23,24 Single-cell RNA sequencing analyses were performed on lens epithelial cells isolated from one-month-old Cx50KO mice and age-matched WT mice. Sequencing data were processed through the 10× Genomics Cell Ranger Analysis Pipelines. A high-quality scRNA-seq dataset was obtained from a total of 2394 WT lens epithelial cells, which displayed 15,942 detected genes, about 1572 median genes per cell, and revealed seven distinctive cell clusters (Fig. 1B) with t-SNE graph-based clustering.³⁷ In the Cx50KO, a high-quality scRNA-seq dataset of 2463 lens epithelial cells detected a total of 16,172 genes, about 1571 median genes per cell, and identified eight distinctive cell clusters (Fig. 1B). 

Figure 1.

 

(A) The lens photos of one-month-old Cx50KO (KO) and age-matched C57BL/6J WT show the reduced size and mild cataract in the KO lens. Scale bar: 1 mm. (B) The tSNE plots of scRNA-seq data identify distinct cell clusters among lens epithelial cells of ∼one-month-old WT and Cx50KO mice. Seven cell clusters are identified in 2394 WT cells whereas eight clusters appear in 2463 KO cells. The percentage of cells for each cluster is indicated. (C) UMAP plots of integrated WT and KO scRNA-seq datasets display seven cell clusters, analyzed by Seurat FindClusters and Louvian algorithm (resolution = 0.3). (D) Proportional cell numbers of WT and KO in integrated cell clusters are shown.

To further examine the alteration of specific cell clusters in the Cx50KO lens epithelium, integrated data analysis of WT and Cx50KO scRNA-seq datasets were performed. Filtered datasets of 2071 WT and 2181 KO cells were integrated, and UMAPs were determined using Seurat FindClusters and the Louvian algorithm, with the resolution parameter set to 0.3. The integrated dataset of WT and KO cells display seven clusters (Fig. 1C). As shown by the proportional cell numbers of WT and KO in integrated cell clusters (Fig. 1D), compared to the cell numbers of the same cell clusters in WT epithelial cells, the cell numbers of clusters 0, 1 and 6 in KO increased by 92.3%, 76.8% and 66.5%, respectively, whereas the cell numbers of clusters 2, 3, 4, and 5 in KO decreased by 70.6%, 75.2%, 57.5%, and 42.9%, respectively. Thus, it is evident that integrated cell cluster numbers are altered by Cx50KO. The top 25 marker genes for each integrated cell cluster (Supplementary Table S1) and the differentially expressed genes between KO and WT in each cell cluster (Supplementary File S1) are shown as supplementary data. 

Integrated data analysis may introduce bias by forcing WT and KO datasets into a specific number of cell clusters (seven in this case), potentially resulting in the separation of biologically similar cells into different clusters or the merging of distinct cell types. Therefore nonintegrated datasets were used for further analysis in this article. 

DEGs Between KO and WT Lens Epithelial Cells

To identify specific gene candidates affected in Cx50KO cell clusters, heatmaps were generated and analyzed for DEGs in each cell clusters between WT (Fig. 2A) and KO (Fig. 2B) lens epithelial cells (genes in the heat maps were listed in the Supplementary Tables S2 and S3). More than 100 genes were listed to show various expression levels in different cell clusters of WT and KO samples. Several well-known genes, including Cryaa, Cryg, Gja1, Mip, and Lim2, which encode alphaA-crystallin, gamma-crystallin, connexin 43, aquaporin 0 (AQP0) and MP19, respectively, were marked in the heatmaps (Figs. 2A, 2B). The differential expression analysis was performed on the WT-KO-aggreg-cloup file using Loupe Brower 6.1.0. The heatmaps of DEGs between KO and WT datasets confirmed the loss of Gja8 transcripts in the Cx50KO lens epithelial cells (Fig. 2C, Supplementary Tables S4 and S5). In comparison to WT, many DEGS were identified in Cx50KO LECs. After excluding genes with low average counts, KO LECs display a total of 164 upregulated genes with a Log2(fold change) > 0.5849 and 29 downregulated genes with a Log2(fold change) < −0.5849 (Supplementary File S2). Pathway enrichment GO analysis was performed on the top 50 upregulated genes and the top 32 downregulated genes using the ShinyGO 0.80 program.³⁸ The enriched GO terms in Cx50KO compared to WT include genes involved in gap junction-mediated intercellular transport, cell proliferation, lens development and other functions (Supplementary Figures S1 and S2). Genes with upregulated expression in Cx50KO include Socs3, Tob1, Btg1, Btg2 and Aqp1, which encode the suppressor of cytokine signaling 3, Transducer of ERBB2, antiproliferative protein BTG1 and BTG2, and the water channel protein aquaporin-1 (AQP1), respectively. Genes with downregulated expression in Cx50KO includes Pdpn, Maf, Mip, Lim2, Grifin, S100a6, S100a11, and several crystallin genes. Because Btg1, Btg2, Tob1 and Socs3 are known to mediate antiproliferative activity in cell growth, they likely are specific downstream targets that inhibit lens epithelial cell proliferation in the Cx50KO lens. Moreover, Aqp1 expression was unexpectedly upregulated in Cx50KO lens epithelial cells. Therefore, some of these DEGs were further investigated with violin plots and RT-PCR assays. 

Figure 2.

 

Heat maps show differentially expressed genes between Cx50KO (KO) and WT lens epithelial cells and among different cell clusters of WT or KO cells. (A) Heat maps show genes (names on X-axis) expressed in each cell clusters (Y-axis) of WT. (B) Heat maps show genes expressed in each cell clusters of Cx50KO lens. (C) Heat maps reveal the topmost differentially expressed genes between KO and WT lens epithelial cells. Several well-characterized lens genes are marked. Grid cells are colored by a gene's Log2 fold change in its cluster row, compared to the other clusters (A, B) or compared between WT and KO (C).

Violin Plots and Real-Time PCR Quantification of Selected DEGs

Violin plots were generated to visualize the differential expression distributions of selected genes among various cell clusters of WT or KO, and between the KO and WT samples. Quantitative real-time PCR (RT-qPCR) was performed on selected DEGs to examine and confirm their gene expression changes. All selected genes have an adjusted P value of < 0.05 after multiple testing correction, with Log2FC > 0.5849 for upregulated genes and < -0.5849 for downregulated genes (Supplementary Table S6). Three lens connexin genes, Gja1, Gja3 and Gja8 which encode for Cx43, Cx46 and Cx50 respectively, were first examined (Figs. 3A–C). The WT violin plot shows the expression of Gja8 in all seven cell clusters with higher expression in clusters 1, 2, 3, and 4 (Fig. 3C); in the KO cells, no Gja8 expression is observed in any cell clusters. The violin plots of Gja1 (Fig. 3A) and Gja3 (Fig. 3B) reveal their expression distribution in the KO and WT cell clusters. In the WT, Gja1 is mainly expressed in two clusters (clusters 2 and 4) and has no expression in the other five clusters (Fig. 3A); in the KO, two clusters (clusters 1 and 4) have high Gja1 expression, and two other clusters (clusters 3 and 6) show low expression (Fig. 3A). The violin plots (Fig. 3A, the upper left plot) indicate increased Gja1 expression in the KO compared to the WT; however, quantitative RT-qPCR result (Fig. 3A, the upper right graph) shows the transcription level of Gja1 is not significantly changed between WT and KO (P > 0.05). The Gja3 violin plots (Fig. 3B) reveal its expression in five clusters of WT cells (clusters 1, 2, 3, 4, and 7), and in seven clusters of KO cells (clusters 1, 2, 3, 4, 5, 6, and 8); the Gja3 expression in the KO seems to increase in the violin plots (Fig. 3B, the upper left plot); however, RT-qPCR results show no significant Gja3 expression change between KO and WT (P > 0.05). Thus, in the Cx50 knockout lens epithelial cells, both Gja1 and Gja3 expressions are not significantly changed. 

Figure 3.

 

Violin plots and quantitative RT-qPCR of connexin genes, aquaporin genes and Lim2. The X-axis indicates the clusters, and the Y-axis is Log2 gene expression in the violin plots. Violin plots display the expression distributions of three lens connexin genes – Gja1 (A), Gja3 (B) and Gja8 (C), three lens aquaporin genes—Mip (D), Aqp1 (E) and Aqp5 (F), and Lim2 (G). The Y-axis displays Log2 gene expression, vertical lines represent maximum expression, the shape of each violin represents all cell results, and the width of each violin represents the frequency of the respective expression level. (A) The violin plots show the expression of Gja1 among cell clusters of both WT and KO cells; the upper left violin plot shows the relative Gja1 expression between WT and KO cells. Quantitative RT-qPCR data is shown in the upper right graph, Gja1 expression is not significantly changed in the KO cells comparing to the WT (P = 0.50, mean ± SD, n = 3, Student's t-test). (B) The violin plots show the expression distributions of Gja3 among clusters of WT and KO cells (the lower plots), between WT and KO cells (the upper left). Quantitative RT-qPCR graph shows no significant change between WT and KO (P = 0.75, mean ± SD, n = 3, Student's t-test). (C) The violin plots of Gja8 show its absence in KO cells. (D) The violin plots display the expression of Mip among cell clusters of WT and KO (the lower plots), between KO and WT (the upper left plot). Quantitative RT-qPCR result (the upper right graph) shows no significant Mip gene expression change between KO and WT (P = 0.24, mean ± SD, n = 3, Student's t-test). (E) The violin plots of Aqp1 reveal its expression mostly in one cell cluster (cluster 1) in the WT cells, but extended to five clusters in the KO cells (the lower plots); the KO cells seem to have more Aqp1 expression comparing to the WT control, RT-qPCR reveals significantly increased Aqp1 expression in the KO cells comparing to the WT (the upper right graph; P = 0.0060, mean ± SD, n = 3, Student's t-test). (F) The violin plots show the expression distribution of Aqp5 among cell clusters of WT and KO (the lower plots), between KO and WT (the upper left plot), RT-qPCR shows no significant change of Aqp5 expression between KO and WT (the upper right graph; P = 0.83, mean ± SD, n = 3, Student's t-test). (G) The violin plots display Lim2 expression distributions among cell clusters of WT and KO (the lower plots), between WT and KO (the upper left panel), and RT-qPCR quantification (the upper right panel) indicates the decreased Lim2 expression in the KO cells is not statistically significant (P = 0.32, mean ± SD, n = 3, Student's t-test).

Three lens aquaporin genes, Aqp0 (also known as Mip), Aqp1 and Aqp5, which encode water channel proteins aquaporin 0 (AQP0), aquaporin 1 (AQP1) and aquaporin 5 (AQP5) respectively, were examined by violin plots (Figs. 3D–F). The violin plots show the expression of Mip in all cell clusters of WT and KO cells (Fig. 3D, the lower plots); quantitative RT-qPCR reveals no significant change of Mip transcription level in KO cells compared to WT cells (P > 0.05). The violin plots of Aqp1 (Fig. 3E) show its expression only in one cluster of WT cells (cluster 1), whereas its expression is extended to five clusters (clusters 1, 2, 3, 4, and 6) in KO cells; quantitative RT-qPCR (Fig. 3E, the right upper graph) confirms the significantly increased Aqp1 expression in the KO cells comparing to the WT cells (P < 0.01). The violin plots of Aqp5 (Fig. 3F) display its restricted expression in clusters 3 and 7 in the WT cells, and cluster 8 in KO cells; quantitative RT-qPCR reveals no significant expression level change of Aqp5 between KO and WT cells (P > 0.05). Thus, among three aquaporin genes, only Aqp1 expression is statistically significantly increased in the KO cells when compared to the WT cells, whereas no significant changes were detected in Aqp0 and Aqp5 expressions. 

Violin plots were also graphed for Lim2, which encodes lens intrinsic membrane protein 2, a lens-specific protein found at the junctions of lens fibers cell.³⁹ The violin plots (Fig. 3G) reveal the expression of Lim2 in all cell clusters in both WT and KO cells; RT-qPCR result (Fig. 3G, the upper right bar graph) shows that the Lim2 expression is not statistically significant different between KO and WT (P > 0.05). 

Violin plots were studied for several DEGs with antiproliferative properties (Fig. 4). Btg1 and Btg2, encoding protein BTG antiproliferation factor 1 and 2, are members of an antiproliferative gene family that regulates cell growth and differentiation.⁴⁰ The violin plots of Btg1 (Fig. 4A) show its extended expression in seven KO cell clusters (clusters 1, 2, 3, 4, 6, 7, and 8) and only in four WT cell clusters (clusters 1, 2, 4, and 6); although the violin plots (Fig. 4A, the upper left graph) indicate increased Btg1 expression in KO cells, quantitative RT-qPCR comparison reveals no statistically significant change (P > 0.05). The violin plots of Btg2 (Fig. 4B) reveal its expression in five WT cell clusters (clusters 1, 2, 3, 4, and 6) and in all clusters of KO cells; RT-qPCR results (Fig. 4B, the upper right bar graph) confirm statistically significantly increased Btg2 expression in the KO cells comparing to the WT control (P < 0.05). Tob1 encodes a member of the transducer of erbB-2/B-cell translocation gene protein family, which are antiproliferative factors that have the potential to regulate cell growth.⁴¹ The violin plots of Tob1 (Fig. 4C) show its expression in three WT cell clusters (clusters 1, 2, and 4), and its expression extends to seven KO cell clusters (clusters 1, 2, 3, 4, 5, 6, and 8); although the violin plots suggest increased overall Tob1 expression in the KO comparing to WT (Fig. 4C, the upper left plot), the RT-qPCR results (Fig. 4C, the upper right bar graph) reveal the increase is not statistically significant (P > 0.05). The upregulated DEG list includes Socs3, which encodes a member of the STAT-induced STAT inhibitor, known as suppressor of cytokine signaling (SOCS) family.⁴² Violin plots (Fig. 4D) reveal that Socs3 is only expressed in two cell clusters of WT cells (clusters 1 and 4), while its expression is extended into all eight KO cell clusters; the RT-qPCR results (Fig. 4D, the upper right bar graph) confirm the significantly increased Socs3 expression in the KO cells (P < 0.05). Violin plots were also graphed for Pdpn, a down-regulated DEG suggested by scRNA-seq data. Pdpn encodes Podoplanin (PDPN), a type-I intergral membrane glycoprotein with diverse distribution and unknown specific functions.^43–45 The violin plots (Fig. 4E) show in the WT cells, Pdpn is mainly expressed in three cell clusters (clusters 1, 2 and 4), with two more clusters also have some expression (clusters 3 and 6). In the KO cells, Pdpn is expressed in six clusters (clusters 1, 2, 3, 4, 6, and 8); although the violin plot (Fig. 4E, the upper left plot) indicates reduced expression of Pdpn in KO, RT-qPCR results (Fig. 4E, the upper right bar graph) reveal no statistically significant change between KO and WT (P > 0.05). 

Figure 4.

 

Violin plots and RT-qPCR quantification of Btg1, Btg2, Tob1, Socs3 and Pdpn. The X-axis indicates the clusters, and the Y-axis is Log2 gene expression in the violin plots. (A) The violin plots display Btg1 expression distributions among various cell clusters of WT and KO (the lower plots), between WT and KO (the upper left plot). RT-qPCR result (the upper right graph) reveals the increased Btg1 expression in KO is not statistically significant (P = 0.19; mean ± SD, n = 3, Student's t-test). (B) The violin plots show Btg2 expression among various cell clusters of WT and KO (the lower plots), and between WT and KO (the upper left plots). KO cells show significantly increased Btg2 expression by RT-qPCR (the upper right graph; P = 0.045, mean ± SD, n = 3, Student's t-test). (C) The violin plots show Tob1 expression distributions among cell clusters of WT and KO (the lower plots), and between WT and KO (the upper left panel). RT-qPCR quantification indicates the increased Tob1 expression in KO cells is not statistically significant (the upper right graph, P = 0.17, mean ± SD, n = 3, Student's t-test). (D) The violin plots show the expression distribution of Socs3 among various cell clusters of WT and KO (the lower plots), and between WT and KO (the upper left plot). Quantitative RT-qPCR (the upper right graph) reveals significantly increased Socs3 expression in KO (P = 0.047, mean ± SD, n = 3, Student's t-test). (E) The violin plots show Pdpn expression distributions among various cell clusters of WT and KO (the lower plots), and between WT and KO (the upper left plot). RT-qPCR result (the upper right graph) reveals the decreased Pdpn expression in KO is not statistically significant (P = 0.21, mean ± SD, n = 3, Student's t-test).

Western Blot Analysis and Immunostaining of Selected DEG Encoded Proteins

The upregulation of Aqp1 expression in Cx50KO lens epithelial cells suggests a novel regulatory mechanism of water homeostasis between Cx50 and AQP1. AQP1 protein expression was further examined by Western blotting and immunostaining. In vivo protein expression of aquaporins in lens epithelial cells was examined with lens capsule protein homogenates made from one-month-old WT and Cx50KO mice (Fig. 5). To prepare enough amount of protein samples for Western blotting, four lens capsules from two littermate mice were combined as one sample. Each comparison experiment was performed with at least three samples pooled from six littermate mice, both female and male mice were used, and experiments were repeated at least three times with samples from three different litters. The representative Western blot images were shown (Fig. 5). The Cx50 protein was not detected in the knockout samples as expected. The Western blot showed similar AQP0 protein expression level between WT and KO; because AQP0 protein is normally expressed in the lens fibers, its expression in the lens capsule preparation is most likely caused by attached fibers to the capsule. AQP5 was expressed in the lens capsule preparation with two bands, the non-glycosylated form at ∼27 kDa, and the additional upper glycosylated band.^28,46–48 No significant AQP5 expression level change was observed between KO and WT. Consistent with the transcriptional level increase revealed by RT-qPCR, the AQP1 protein expression was significantly increased in the KO samples compared to the WT control (P < 0.05), showing about a 2.5-fold increase of AQP1 protein level in the Cx50KO lens epithelial cell homogenates compared to wild-type control samples; moreover, AQP1 protein displayed two bands in the western blot, the lower major band at ∼28 kDa and an additional band at ∼34 kDa, which is the glycosylated protein.⁴⁹ Both bands showed significantly increased expression. 

Figure 5.

 

Western blot analysis reveals increased in vivo AQP1 protein expression. Peeled lens capsule homogenates were prepared from KO or WT mice at three weeks of age. The representative images with molecular mass markers are shown. The Western images show the absence of Cx50 protein expression in the KO samples; the AQP0 and AQP5 expression levels are comparable between Cx50KO and WT control; however, the AQP1 expression level is obviously increased in the KO samples comparing to the WT control. The protein band intensity was quantified and normalized by the β-actin band, the data reveal significantly increased AQP1 protein expression level in the KO (P = 0.016, mean ± SD, n = 3, Student's t-test), whereas AQP0 and AQP5 both show insignificant protein expression changes between KO and WT.

We further prepared primary cultured lens epithelial cells for Western blot analysis and immunostaining.³⁶ The primary cultured cells grew confluent within ∼10 days and were lysed for Western blot or fixed for immunostaining. For the cell culture, cells isolated from two lenses of each mouse were plated onto one plate. For each experiment, at least three littermates of WT or KO mice were used, both female and male mice were used, and experiments were repeated at least three times with three different litters of mice. In cultured primary lens epithelial cells, AQP1 protein expression was obviously upregulated (Fig. 6A), including non-glycosylated lower band and glycosylated upper band of protein; statistical analysis revealed significantly upregulated AQP1 protein expression in cultured KO lens epithelial cells compared to the WT control (P < 0.01), showing ∼3-fold increase in cultured Cx50KO lens epithelial cells compared to the WT control. When the fixed cells were immunostained with an anti-AQP1 antibody (Fig. 6B), enriched membrane staining was observed. The AQP1 protein staining was only observed in sparsely restricted WT cells (Fig. 6B, the upper panels), whereas most of the KO cells showed intense membrane AQP1 expression (Fig. 6B, the lower panels). The staining intensity of AQP1 was quantified using ImageJ (Fig. 6B, the lower bar graph), and the bar graph of staining intensity revealed that KO cells had increased AQP1 expression compared with the WT cells (P < 0.01). 

Figure 6.

 

Increased AQP1 protein expression in cultured primary lens epithelial cells. (A) Western blot analysis shows increased AQP1 protein in cultured KO lens epithelial cell homogenates compared to the WT. The quantification of band intensity (the lower band) reveals significantly increased AQP1 expression in the KO cells after normalized with β-actin expression (P = 0.004, mean ± SD, n = 3, Student's t-test). (B) AQP1 protein expression is sparsely detected in the WT cells (the upper images), while intense membrane AQP1 staining (green signals, co-stained with DAPI in blue) is seen in clusters of cells. AQP1 staining intensity quantification by ImageJ reveals significantly increased AQP1 expression in the KO compared to the WT (P = 0.004, mean ± SD, n = 3, Student's t-test).

Western blots were also attempted with Socs3, Btg1, Btg2, and Tob1 antibodies, but the results were inconclusive with antibodies purchased from different vendors (data not shown). Decreased PDPN protein level in the KO was detected by a PDPN polyclonal antibody compared to the WT (P < 0.05) (Fig. 7A). Pdpn is a downregulated DEG according to our scRNA-seq analysis between KO and WT. In the cultured primary lens epithelial cells, the PDPN protein expression was also significantly reduced compared to the WT (P < 0.01) (Fig. 7B). Moreover, PDPN immunostaining showed enriched membrane signals in cultured WT cells; in contrast, cytosolic PDPN signals appeared in cultured KO cells (Fig. 7C). Therefore, PDPN protein expression is significantly decreased both in vivo and in vitro; moreover, its staining distribution is altered in the Cx50KO cells. 

Figure 7.

 

Reduced PDPN protein expression in Cx50 knockout lens epithelial cells. (A) Western blot analysis of peeled lens capsule homogenates shows reduced PDPN protein expression in the KO compared to the WT, the molecular mass markers are indicated; the quantification of band intensity reveals significantly decreased PDPN expression in the KO cells after normalized with β-actin expression (P = 0.033, mean ± SD, n = 3, Student's t-test). (B) Western blot of cultured lens epithelial cell homogenates. The PDPN protein expression is significantly reduced in the KO (P < 0.01, mean ± SD, n = 4, Student's t-test). (C) Altered PDPN staining in cultured primary lens epithelial cells of Cx50 knockout. Cells were stained with an anti-PDPN antibody (green signals, co-stained with DAPI in blue). WT cells show PDPN staining signals in the cell membrane; whereas in KO cells, the cell membrane staining of PDPN disappeared, and the staining signals were mainly detected in the cell cytosol. Quantification of the staining intensity also revealed a significant reduction of PDPN expression in the KO cultured cells comparing to the WT contro=l (P = 0.039, mean ± SD, n = 3, Student's t-test).

Discussion

The scRNA-seq results reveal the cell heterogeneity of lens monolayer epithelium that can be classified through distinctive cell clusters on their gene expression profiles. Seven distinctive cell clusters were identified in the WT lens epithelial cells while eight distinctive cell clusters were identified in the Cx50 knockout cells. Different cell clusters between WT and Cx50KO lens epithelial cells suggest that Cx50-mediated gap junction communication or signaling pathways are important for regulating distinct cell clusters (or heterogeneity) in lens epithelial cells. 

Integrated analysis of forced WT and Cx50KO datasets into seven cell clusters reveals altered cell numbers in specific clusters by Cx50KO. Therefore the cell clustering results from both nonintegrated and integrated data analyses support the notion of altered cell clusters between WT and Cx50KO lens epithelial cells. The violin plots of the integrated datasets display differential expression distributions of selected genes (Supplementary Figures S3 and S4), reflecting the up- or down-expression distributions of specific DEGs across various cell clusters in the nonintegrated datasets. 

Based on the DEGs from the scRNA-seq analysis of distinct cell clusters, we have further identified Cx50-mediated specific downstream factors, including antiproliferative transcription factors Btg2 and Socs3, as well as PDPN protein, that likely contribute to Cx50KO lens phenotypes. Moreover, RT-qPCR, Western blot, and immunostaining data have further confirmed the unexpectedly upregulated Aqp1 in Cx50KO lens epithelial cells in vivo and in vitro whereas expressions of both Cx46 and Cx43 remain unchanged. Therefore, this work indicates that water homeostasis in lens epithelium is maintained by a compensatory mechanism between Cx50 and AQP1. 

Heterogeneity of Lens Epithelial Cells

A recent study of scRNA-seq data from mouse postnatal day 2 (P2) lenses reports that lens epithelial cell can be divided into seven cell clusters, and the heterogeneity of lens epithelial cells is associated with distinct transcriptional states, variations in signaling molecules and pathways, and differences in the expression of crystallin, cell-cycle, and metabolic genes among these subclusters (Giannone et al., 2023 IOVS).⁵⁰ Similar to P2, scRNA-seq results from one-month-old WT mice also show seven distinct cell clusters, reflecting the diverse roles of the lens epithelium in regulating proliferation, signaling, and homeostasis of the lens. 

The heterogeneity of lens epithelial cells in WT lens is demonstrated in seven distinct cell clusters based on the scRNA-seq data analysis. Using t-SNE graph-based visualization,³⁷ seven distinctive cell clusters of WT lens epithelial cells have been classified, with each cluster having different cell numbers, ranging from 9.9% (cluster 7) to 21.8% (cluster 1) of total cells (Fig. 1). The heat maps of WT clusters indicate different expression levels of some DEGs among these cell clusters. Moreover, WT lens epithelial cells display heterogeneity in connexin gene expression pattern. The violin plots of WT cells show that Gja1 is restrictively expressed in clusters 2 and 4, and Gja3 is expressed in clusters 1, 2, 3, 4, and 7, while Gja8 is expressed in all cell clusters. In other words, clusters 2 and 4 express all three genes–Gja1, Gja3 and Gja8; clusters 1, 3 and 7 express Gja3 and Gja8; and clusters 5 and 6 express only Gja8. Cx50KO lens epithelial cells display eight clusters and show altered connexin gene expression distribution. In the KO cells, Gja1 expression is detected in clusters 1, 3, 4, and 6; the Gja3 expression is expanded into clusters 1, 2, 3, 4, 5, 6, and 8; neither Gja1 nor Gja3 is expressed in cluster 7. These results suggest that Cx50 plays an important role, at least in some distinct cell clusters, probably by controlling the expression of specific downstream genes that regulate proliferation, differentiation, or homeostasis of lens epithelial cells. Therefore, a deletion of Cx50 in the Cx50KO mice alters distinct cell cluster(s) and DEGs to impair lens growth or homeostasis, leading to lens pathological outcomes including smaller lens and cataracts. 

Cx50 and AQP1 Regulate Water Homeostasis in Lens Epithelial Cells

The regulation of water homeostasis in the lens epithelial cells is critical for maintaining the lens hydrostatic state and transparency. At least five aquaporins, including AQP0, AQP1, AQP5, AQP7 and AQP8, are utilized in the lens.³⁴ This work confirms three aquaporin genes, Mip, Aqp1 and Aqp5, expressed in lens epithelial cells. Previous studies report that AQP1 is restrictively expressed in the lens epithelial cells,^51,52 AQP0 is predominantly used in the lens fibers,^53–56 and AQP5 is expressed in both lens epithelial cells and fiber cells.^28,47,57 AQP0 is a poor water channel and also functions as adhesion molecules between lens fibers, and both Cx50 and AQP0 provide cell-cell adhesion function in the lens fiber cells.^48,58–60 AQP5 has been shown to act as a regulated water channel with a relatively high water permeability.⁵⁸ Mutations of Aqp0 and Aqp5 cause cataracts in mouse and humans.^61–67 AQP5 expression remain unchanged in Cx50KO lens epithelial cells but AQP1, a constitutively active water channel with high water permeability,^52,68,69 is upregulated. Moreover, expression of Cx43 and Cx46 remain unchanged in Cx50KO lens epithelial cells. 

The discovery of upregulated Aqp1 in Cx50KO lens epithelial cells suggests that Cx50 likely plays a critical role in water homeostasis in lens epithelial cells. A compensatory regulatory mechanism between Cx50 and AQP1 may be necessary to maintain the water homeostasis in normal lens epithelial cells. This likely explains why a functional loss of Aqp1 alone is not adequate to cause lens phenotypes in mice. A deletion of Aqp1 did not directly cause lens cataract phenotype in the knockout mice, although epithelial cells isolated from Aqp1 knockout lenses exhibited a threefold reduction in water permeability⁷⁰; moreover, when lenses of Aqp1 knockout mice were incubated in high glucose solution in vitro, lens opacities were observed.⁷⁰ Increased AQP1 membrane protein expression in the lens epithelium of cataract patients was reported, although no change was observed at the mRNA level.⁷¹ Thus, AQP1 plays essential function in maintaining lens transparency, and AQP1 is required in lens epithelial cells to promote water influx and efflux across the epithelium and maintain lens transparency, especially following exposure to stress conditions such as hyperglycemia and osmotic imbalance. Upregulated AQP1 in Cx50KO lens epithelial cells indicates a compensatory mechanism for water transport, thus both Cx50 and AQP1 play important roles in regulating lens water homeostasis. At this current stage, mechanism for how a loss of Cx50 leads to the upregulation of Aqp1 is unknown. Additional studies will be needed to address both transcriptional and translational regulations of Aqp1 in lens epithelial cells. 

Increased Btg2 and Socs3 and Decreased Pdpn Expression in Cx50KO Lens Epithelial Cells

This work reveals that Cx50 mediated regulation of lens epithelial cell proliferation is associated with upregulated expression of antiproliferation genes, including Btg1, Btg2, and Tob1. BTG1, BTG2, and TOB1 belong to the BTG/Tob family, which share a key conserved antiproliferative domain.⁷² BTG1 and BTG2 likely play redundant functions in many tissues; they are very similar except that the C-terminal region of BTG1 is slightly longer.⁷³ Moreover, upregulation of the Socs3 expression in Cx50KO lens epithelial cells suggests that Cx50-mediated signaling inhibits the expression of Socs3 in wild-type lens epithelial cells. The Socs3 gene encodes the suppression of cytokine signaling 3 (SOCS3), a member of the STAT-induced STAT inhibitor, also known as suppressor of cytokine signaling (SOCS) family.⁷⁴ The STAT-induced STAT inhibitor family members are cytokine-inducible negative regulators of cytokine signaling. The SOCS3 protein can bind JAK2 kinase to inhibit its kinase activity. Therefore, this work indicates Cx50 functions as a negative regulator of antiproliferative factors such as BTG2 and SOCS3 to control the proliferation of lens epithelial cells during lens development. 

Changes in Pdpn gene expression and its protein distribution are related to a loss of Cx50 in lens epithelial cells. The Pdpn gene encodes the protein Podoplanin or PDPN, which has a single transmembrane domain, a glycosylated extracellular domain, and a short (nine amino acids) cytoplasmic tail.⁷⁵ PDPN protein displays diverse distribution, and its expression is concentrated in actin-rich plasma membrane protrusion structures, where it directly bind ezrin and moesin to regulate cell motility and adhesion.^75–77 However, PDPN function is largely unknown, and the underlying mechanism for the involvement of PDPN in Cx50KO lens pathology needs further investigation. 

In conclusion, this work suggests that Cx50 and AQP1 play important roles in water homeostasis in lens epithelial cells. The Cx50-mediated signal transduction pathway controls antiproliferative factors such as Btg2 and Socs3 to regulate lens growth rate and size during development. However, additional mechanistic studies are needed to clarify whether Cx50-mediated hemichannel function and/or Cx50 gap junction communication are responsible for these DEGs in lens epithelial cells. 

Acknowledgments

The authors thank Haiping Hao for his help with the scRNA-seq data analysis, and Jessica Wang, Chenxi Ou, and Jonathan Gong for their assistance in the experiments. 

Supported by grant EY031253 from the National Eye Institute (X. Gong). 

Disclosure: C.-H. Xia, None; W. Lin, None; R. Li, None; X. Xing, None; G.J. Shang, None; H. Zhang, None; X. Gong, None