Congenital microcoria deletion in mouse links Sox21 dysregulation to disease and suggests a role for TGFB2 in glaucoma and myopia

Summary

Congenital microcoria (MCOR) is a rare hereditary developmental defect of the iris dilator muscle frequently associated with high axial myopia and high intraocular pressure (IOP) glaucoma. The condition is caused by submicroscopic rearrangements of chromosome 13q32.1. However, the mechanisms underlying the failure of iris development and the origin of associated features remain elusive. Here, we present a 3D architecture model of the 13q32.1 region, demonstrating that MCOR-related deletions consistently disrupt the boundary between two topologically associating domains (TADs). Deleting the critical MCOR-causing region in mice reveals ectopic Sox21 expression precisely aligning with Dct, each located in one of the two neighbor TADs. This observation is consistent with the TADs’ boundary alteration and adoption of Dct regulatory elements by the Sox21 promoter. Additionally, we identify Tgfb2 as a target gene of SOX21 and show TGFΒ2 accumulation in the aqueous humor of an MCOR-affected subject. Accumulation of TGFB2 is recognized for its role in glaucoma and potential impact on axial myopia. Our results highlight the importance of SOX21-TGFB2 signaling in iris development and control of eye growth and IOP. Insights from MCOR studies may provide therapeutic avenues for this condition but also for glaucoma and high myopia conditions, affecting millions of people.

Graphical abstract

Congenital microcoria (MCOR) is a rare ocular developmental disease we attribute to pathological chromatin reorganization, ectopic Sox21 expression in the iris, and TGFB2 accumulation in the aqueous humor. This accumulation, known to trigger glaucoma and myopia, highlights novel pathways for treating these conditions, common in both MCOR and the general population.

Introduction

The iris is a remarkable ocular structure, serving a dual purpose by finely tuning the amount of light that reaches the retina and regulating intraocular pressure (IOP). It comprises a stroma, a double epithelium layer, and two dynamic muscles, the dilator and the sphincter, that work together to adjust the size of the pupil and hence the light intensity for optimal vision. The iris is rooted to both the corneal-scleral junction and the ciliary body (CB), whose stroma and epithelial layers are continuous with those of the iris. This structural arrangement creates an open space known as the irido-corneal angle where the aqueous humor (AH) produced by the CB to nourish ocular tissues is drained out, playing a vital role in maintaining IOP.¹ An increase in IOP is a prominent risk factor for optic nerve damage and glaucoma (GLC).²

Congenital microcoria (MCOR; MIM: 156600) is an extremely rare autosomal-dominant condition affecting both of these functions; our recent review identified 160 affected individuals from 49 families since the disease was first mentioned in 1862.³ It is characterized by the partial or complete absence of the iris dilator muscle, which normally originates from the anterior part of the epithelial layer (AEL) and extends along its stroma. The sphincter muscle near the pupil’s edge retains full functionality. The absence of dilator muscle fibers is evident in the presence of pinhole-sized pupils (<2 mm) that exhibit minimal or no dilation, even when mydriatic drugs are used. This also leads to the iris thinning and atrophy of the stroma, particularly in the dilator muscle region, which permits light to pass through, inducing iris transillumination.³ In the vast majority of the cases, this abnormal development of the iris does not affect the chamber angle that is wide-open although prominent iris processes and a higher insertion of the iris root in the angle have been typically described.³ Alongside dilation problems, individuals with MCOR are at a substantially increased risk of developing high axial myopia, affecting 70% of cases.³ Among those with high myopia, half also experiences juvenile-onset high IOP GLC, an unusually high prevalence compared to the general population, which stands at 34% of all MCOR cases.³

The disease was ascribed to submicroscopic rearrangements of chromosome 13q32.1, particularly deletions and a reciprocal duplication, suggesting that the disease is linked to an alteration of the regulatory landscape of the region.^4,5,6,7,8

Here, we provide a 3D architecture model of the 13q32.1 region that suggests that MCOR-related deletions consistently alter the boundary between two interacting topologically associating domains (TADs). Deleting the critical 35 kb MCOR-causing region in mice leads to aberrant Sox21 expression in developing and adult iris and CB, aligning precisely with Dct, which is located in the adjacent TAD. This suggests ectopic Sox21 expression may stem from Dct regulatory element adoption, supporting the alteration of TAD boundaries.

Additionally, we identify the transforming growth factor beta-2, Tgfb2, as a SOX21 target and observe TGFB2 accumulation in the AH of an MCOR-affected subject. This accumulation, known for its role in GLC^9,10,11,12 and potential impact on axial myopia,^13,14 underscores the significance of SOX21-TGFB2 signaling in these ocular anomalies and iris development. Insights from MCOR studies may offer therapeutic avenues for individuals with this condition, as well as those with primary open-angle glaucoma (POAG) and high myopia, affecting millions.

Material and methods

In silico analysis of 13q32.1 Hi-C maps

Analysis of experimentally derived Hi-C chromatin structure maps

Publicly available high-resolution chromosome conformation capture (Hi-C) datasets for murine and human neural progenitor cells were downloaded from Gene Expression Omnibus (GEO): GSE96107.^15,16 Hi-C reads underwent trimming and quality control assessment using the Trim Galore package¹⁷ (TrimGalore v0.6.5, Cutadapt v2.6, and Fastqc v0.11.9). Subsequently, reads were mapped to hg19 or mm10 genomes and filtered for common Hi-C artifacts using Hi-C Unifying Pipeline (HiCUP v0.7.2, Bowtie2 v2.3.5, and R v3.6.0_3.9).¹⁸ Analysis of Hi-C files was conducted using Juicer and associated Juicer Tools¹⁹ (v1.22.01). Contact maps were generated using Juicer with the parameter: -s DpnII. Map resolution was determined using Juicer’s “calculate_map_resolution.sh” script. All observed matrices were computed using the Hypergeometric Optimization of Motif EnRichment (HOMER) suite,²⁰ factoring in both linear distance and sequencing depth between loci. Hi-C contact maps and annotations were visualized using Juicebox.¹⁹ Subsequently, interaction matrices were converted from hic to cool format using hicConvertFormat to facilitate analysis with the HiCExplorer tool suite.^21,22,23 For clarity in Hi-C heatmaps, observed interaction values were plotted, enabling better visualization of TAD structures across large-scale windows.

Analysis of simulated Hi-C maps

We utilized the Modeling DNA Loop Extrusion (MoDLE) computational tool for rapidly and stochastically modeling molecular contacts from DNA loop extrusion, capable of simulating realistic contact patterns genome wide within minutes,²⁴ to generate Hi-C maps. In essence, insulation scores were computed using the MoDLE tool as described by the Paulsen group,²⁴ leveraging CTCF chromatin immunoprecipitation followed by sequencing (ChIP-seq) data from both human and mouse neural progenitor cells downloaded from GEO (GEO: GSE96107)¹⁵ for mice and from Rajarajan et al. for human data.¹⁶ The mm10 mouse genome and hg19 human genome were employed for predicting CTCF coordinates. MCOR deletion extrusion files were created by removing CTCF sites encompassing the deletion in the extrusion file, and predictions were re-run using these updated extrusion files.

Generation of an MCOR-mouse model

The B6.cΔMCOR strain utilized in this study was generated through CRISPR/Cas9 methodology at the Transgenesis Platform of the Animal Facility (LEAT) at the Imagine Institute (Paris). Animal procedures received approval from the French Ministry of Research and were conducted in compliance with the French Animal Care and Use Committee from the University of Paris Cité (APAFIS #14311-201801151627355). Two guide RNAs specific to the 5′ and 3′ boundaries of the MCOR locus (Table S1) were designed using the CRISPR/Cas9 sgRNA and Talen Optimized Resource (CRISPOR) web tool (http://crispor.tefor.net/).²⁵ Deleterious alleles were induced via CRISPR/Cas9 ribonucleoprotein complex microinjection into C57BL/6 mouse zygote pronuclei, as described previously.²⁶ Deletion of the MCOR locus was detected in offsprings genomic DNA through PCR amplification of the intervening segment using primers flanking the locus (Table S1). To eliminate potential off-target mutations, transgenic animals underwent multiple generations of backcrossing with C57BL/6 mice. SW.cΔMCOR/+ animals were obtained by crossing B6.cΔMCOR/+ with Swiss albino mice over several generations.

Pupillometry analysis

Two-month-old mice (7 B6.WT and 8 B6.cΔMCOR/+) underwent dark adaptation overnight, and pupillary responses were recorded following established protocols.²⁷ To avoid the potential effects of anesthesia, animals were not anesthetized. The baseline pupil diameter was defined as the mean pupil diameter during the 500 ms before light onset. Subsequently, all pupil sizes were normalized relative to this baseline value. The light stimulus sequence comprised 50 ms exposures to 0.01, 0.1, and 3.16 W/m2 white light and a 20 s exposure to 1 W/m² blue light. Pupil diameter was automatically determined using Neuroptics A2000, Inc. software. One-way/two-way analysis of variance (ANOVA) analyses were employed to identify significant differences.

RNA extraction

Mice aged from 0 days postnatal (P0) to 12 months, were euthanized, and their eyes were enucleated. On ice, the eyes were dissected to collect the iris and CB. Each eye was halved following the ora serrata, and after lens removal, the iris and CB were extracted from the anterior segment. Tissues from both eyes were pooled for total RNA extraction using the RNAeasy MiniKit (Qiagen, Hilden, Germany), following the manufacturer’s instructions, post-mechanical tissue disruption. RNA quality and concentration were assessed through optical density measurements (Nanodrop, Thermo Fisher Scientific, Waltham, MA, USA), and RNA integrity was analyzed using capillary electrophoresis with a Tape Station (Agilent, Santa Clara, CA, USA) and High Sensitivity RNA Screen Tape.

RNA-sequencing analysis

The Ovation Mouse RNA-Seq System from NuGEN (Tecan Genomics, Leek, NL) was employed to prepare RNA sequencing (RNA-seq) libraries using 25 ng of total RNA, following the manufacturer’s recommendations. This kit enables strand-specific RNA-seq library construction using 10 to 100 ng of total RNA and utilizes the Insert Dependent Adaptor Cleavage technology to eliminate ribosomal RNA transcripts. An equimolar pool of the final indexed RNA-seq libraries was sequenced on an Illumina HiSeq 2500, targeting a sequencing depth of approximately 30 million paired-end reads per library. On average, 28 million reads were obtained. FASTQ files were mapped to the ENSEMBL Mouse (GRCm38/mm10) reference genome using hierarchical indexing for spliced transcript alignment 2 (HISAT2),²⁸ and gene counts were obtained using the featureCounts function from the Subread R package.²⁹ Read count normalization and group comparisons were conducted using three independent and complementary statistical methods: DESeq2,³⁰ edgeR,³¹ and Limma-voom.³² Flags were generated from counts normalized to the mean coverage, where counts <20 were considered as background (flag 0) and counts ≥20 were considered as signal (flag = 1). P50 lists, utilized for statistical analysis, grouped genes showing flag = 1 in at least half of the compared samples. Results from the three methods were filtered at p value ≤ 0.05 and fold changes of 1.2/1.5/2 and grouped by Venn diagram. Cluster analysis employed hierarchical clustering with the Spearman correlation similarity measure and ward linkage algorithm. Heat maps were generated using the R package ctc: Cluster and Tree Conversion and visualized with Java Treeview software.³³ Functional analyses were conducted using Ingenuity Pathway Analysis (IPA; Qiagen).

Real-time qPCR analysis

mRNA from mouse ocular tissue and human U-251 glioma cells were reverse transcribed from total RNA using the cDNA Verso reverse transcription kit (Thermo Fisher Scientific). The abundance of Sox21 and Dct in mouse tissues as well as SOX21 and TGFΒ2 in U-251 glioma cells was evaluated by real-time qPCR relative to the reference genes B2m, Tbp, Gusb, and Hprt1, and GAPDH, GUSB, and HPRT1, respectively. Reverse transcribed mRNAs were amplified using specific primers (Table S1) and the Sso Advanced Universal SybrGreen Supermix (Bio-Rad, Hercules, CA, USA) in a Realplex2 Mastercycler (Eppendorf, Montesson, France).

For each cDNA sample, the mean of quantification cycle (Cq) values was calculated from triplicates (standard deviation [SD] < 0.3 Cq). Target gene expression levels were normalized to the “normalization factor” obtained from the geNorm software for Microsoft Excel,³⁴ which uses the most stable reference genes. No reverse transcriptase (non-RT) and no template control (NTC) reactions were used as negative controls in each run (Cq values NTC = undetermined, non-RT >40).

In mouse ocular tissue, Sox21 and Dct data represent the mean ± standard error of the mean (SEM) of three independent experiments, presented as mRNA value ratios. Statistical significance was assessed using Fisher’s protected least significant difference (PLSD) after analysis of variance (Statview Software, version 5; SAS Institute, Cary, NC). In human glioma cells, SOX21 and TGFΒ2 data represent the mean ± SEM of five independent experiments, normalized to the non-target control small interfering RNA (siRNA), with statistical significance evaluated using an unpaired t test.

Western blot analysis

The western blots were performed using 40 μg of total iris and ciliary body (IrCB) protein lysates. The following antibodies were used in this study: SOX21 (AF3538, R&D Systems, Minneapolis, MN, USA) and beta-Actin (ab8227, Abcam, Cambridge, UK). Secondary antibodies, goat IgG or mouse IgG HRP (Thermo Fisher Scientific) were used at a 1:10,000 dilution. Western blot membranes were developed using the Clarity Western ECL substrate (Bio-Rad), and the signal was detected with a Chemidoc MP Imaging System (Bio-Rad).

ChIP-seq analysis

The IrCB of 24 newborns B6.cΔMCOR/+ and 24 B6.WT mice (P0) were collected for chromatin immunoprecipitation sequencing (ChIP-seq) analysis. Three samples were prepared for B6.cΔMCOR/+ mice and three for B6.WT mice, with each sample comprising 16 irises. The ChIP assay was performed on all samples using 2 μg of chromatin.

Cells were crosslinked with 1% formaldehyde for 10 min at room temperature. ChIP was carried out using the iDeal ChIP-seq kit for Transcription Factors (C01010055, Diagenode, Liege, Belgium) following the provided protocol. Immunoprecipitation utilized an anti-SOX21 antibody (AF3538, R&D Systems). Sequencing was performed on an Illumina HiSeq 4000, running HiSeq Control Software HD version 3.4.0.38. Quality control of sequencing reads was conducted using FastQC. Reads were aligned to the reference genome (GRCm38/mm10) from the University of California Santa Cruz (UCSC) genome browser³⁵ using Burrows-Wheeler aligner (BWA) software v.0.7.5a.³⁶ All samples were filtered using a list of genomic regions created from the two input samples. Subsequently, samples were deduplicated using Sequence Alignment/Map (SAMtools) version 1.3.1.³⁷ Alignment coordinates were converted to Browser Extensible Data (BED) format using BEDTools v.2.17,³⁸ and peak calling was performed using Model-based Analysis of ChIP-Seq version 2 (MACS2) software.³⁹

siRNA-mediated expression knockdown of SOX21

U-251 glioma cells were cultured in a 37°C, 5% CO2 incubator. One day before transfection, the cells were trypsinized using 0.05% trypsin-EDTA (Thermo Fisher Scientific) and seeded in 12-well culture plates at a concentration of 100,000 cells per well. A mixture containing siRNAs targeting SOX21 (FlexiTube GeneSolution GS11166, Qiagen; siRNA sequences listed in Table S1) or a non-target control siRNA (AllStars Neg. Control siRNA, Qiagen) was transfected at a concentration of 50 nM using Lipofectamine 2000 reagent (Invitrogen, Illkirch, France) following the manufacturer’s protocol. Cells were harvested 48 h after transfection and subjected to RNA extraction. A total of 5 replicates were performed. RNA extraction was carried out using the PureLink RNA Mini Kit (Invitrogen) according to the manufacturer’s instructions. Reverse transcription was then performed using the Verso cDNA Synthesis Kit (Thermo Fisher Scientific), including a DNase treatment step to eliminate any remaining genomic DNA.

ELISA assay

Mice were euthanized at 5, 6, and 12 months old, and AH was collected from both eyes using a 27G-syringe inserted at the ora serrata (around 5 μL per eye). The AH from both eyes was pooled, and TGFΒ2 from the samples were activated using one volume of HCl and two volumes of NaOH prior to the quantification of total inactive TGFΒ2 using the Mouse TGF-beta 2 Quantikine ELISA Kit (R&D Systems) following the manufacturer’s instructions. Mean TGFB2 concentrations ± SEM were determined, and statistical significance was assessed by comparing mean concentrations using Fisher's PLSD test following ANOVA with StatView software (version 5).

Human AH was collected from eleven subjects undergoing cataract surgery at Clinique Jules Verne, Nantes, France. The mean age of these controls individuals was 76.7 ± 5 years old, while the MCOR individual was the youngest at 44 years old. All samples were collected with the written and informed consent of the subjects in accordance with the principles outlined in the Declaration of Helsinki. The study was approved under authorization CPP:2015-03-03/DC2014-2272 by the institutional review board Comité de Protection des Personnes Ile-de-France II. TGFΒ2 dosage in each eye (one eye per control subject and both eyes for the MCOR case) was conducted with the Human TGF-beta 2 Quantikine ELISA kit (R&D Systems) following the provided instructions. Mean TGFB2 concentrations ± SEM were determined.

Immunofluorescence and RNAscope analysis

Pregnant mice were euthanized to collect embryos at various stages, from E10.5 to P0. Accurate embryonic stages were assigned post-cutting by consulting established atlases due to variations in early embryo development within littermates. Embryo heads were fixed with 4% paraformaldehyde (PFA), and fixed heads as well as fresh adult eyes were embedded in optimal cutting temperature (OCT) compound and frozen in liquid nitrogen for rapid tissue freeze. OCT blocks were stored at −80°C and brought up to −20°C prior to cutting (CM3050S, Leica, Wetzlar, Germany). Frozen tissue sections (14 μm thickness) were collected at the level of the pupillary aperture, air-dried, and stored at −80°C.

Frozen sections were rinsed in PBS to remove surrounding OCT, and antigen retrieval was performed using Citrate Buffer (pH 6.0) in a 95°C water bath for 30 min. Slides were then cooled for 30 min at room temperature. Tissues were blocked for 1 h in PBS with 1% bovine serum albumin (BSA), labeled with primary antibodies anti-SOX21 (AF3538, R&D Systems), anti-DCT (ab74073, Abcam), and Cy3-coupled anti-alpha Smooth Muscle Actin (C6198, Sigma-Aldrich, Saint-Louis, MO, USA) followed by appropriate secondary antibodies. Slides were counterstained with 4',6-Diamidino-2-Phenylindole (DAPI) and mounted with Fluoromount medium (Sigma-Aldrich).

For mRNA detection, slides were processed using the RNAscope Multiplex Fluorescent v2 Assay, and subsequent immunohistochemistry was performed. All images were acquired with a Spinning Disk Confocal microscope (Carl Zeiss, Oberkochen, Germany), capturing whole sections with a 10× objective and specific areas of interest (iris, CB, and margins of the optic cup) with a 40× oil objective. Raw images were converted to TIFF for further analysis using Fiji (v1.53t).⁴⁰

Statistical analysis

Fisher's PLSD test following ANOVA (Statview Software, version 5) or unpaired two-tailed Student’s t test was used for two-group analyses. two-way ANOVA analysis was used for multigroup analyses (GraphPad Prism, ver. 10.0.3). Data are presented as the mean ± SEM; p < 0.05 was considered significant.

Results

Deletions causing MCOR are predicted to modify the boundary between two adjacent topologically associating domains on chromosome 13q32.1

The MCOR locus, defined by its most distant deletion boundaries, spans 99.8 kb pairs on chromosome 13q32.1 (UCSC Genome Browser: GCRCh37/hg19_chr13: 95,209,609–95,309,380).³ Based on Hi-C chromatin structure maps accessible on the 3D Genome Browser,⁴¹ this locus was initially located within a one mega-base (Mb) TAD, covering the genomic region from DCT (MIM: 191275) to UGGT2 (MIM: 605898). To refine the structural organization of the region, we examined higher-resolution Hi-C maps and insulation scores in human induced pluripotent stem cell-derived neural progenitor cells (hNPCs) and adult retina cells. These cell types were selected from publicly available options for their relevance to the disease, which involves the iris epithelia developing from the optic cup concurrently with the retina. We identified three interacting TADs: T1, T2, and T3 of 1.2 Mb, 0.25 Mb, and 0.5 Mb, respectively (Figures S1A and 1A). Remarkably, all documented deletions associated with MCOR³ pass through the boundary between T2 and T3 (Figure 1B), underscoring the significance of this boundary and emphasizing the intricate interplay within these TADs in the context of MCOR. Interestingly, this chromatin architecture appears to be consistent across various tissue and cell types, with the boundary conserved in 5 out of 9 primary and tumoral human cell lines encompassing skin, connective tissue, epithelial, endothelial, and blood-derived cells for which Hi-C data were publicly available (Figure S1). To assess the impact of the critical MCOR deletion (UCSC Genome Browser: GCRCh37/hg19_chr13: 95,241,606–95,276,905) on 3D structure, we utilized MoDLE, which integrates CTCF DNA binding motifs with ChIP-seq data to predict insulation scores and generate interaction heatmaps (https://doi.org/10.1186/s13059-022-02815-7).²⁴ Analyzing data from hNPCs, MoDLE produced predictive interaction heatmaps that closely matched experimentally generated heatmaps (Figures 1A and S1A). This agreement prompted us to employ the software to generate a predictive Hi-C map of the region deleted from the critical MCOR region, which encompasses two CTCF sites at the boundary between T2 and T3. The resulting map shows T2 extending on the right to include the SRY-related HMG-box 21 transcription factor SOX21 (MIM: 604974), originally located in T3 about 50 kb downstream of the MCOR locus’s right boundary in T2 (Figures 1A and 1B). This led us to analyze chromatin interactions under wild-type (WT) and MCOR conditions by focusing on the SOX21 promoter. In wild-type hNPC interaction maps, SOX21 promoter interactions were limited to the T3 domain (hNPC and WT Predict; Figure 1C). However, under MCOR conditions, the SOX21 promoter showed increased interactions within the T2 domain, particularly at its left boundary (Figure 1C; MCOR Predict). This region includes DCT, encoding the dopachrome tautomerase, located approximately 75 kb upstream of the left boundary of the MCOR locus. The coordinates of human 13q32.1 TADs (T) and syntenic mouse 14E4 TADs (t) are provided in Table S2.

Figure 1.

Chromatin organization of the human and mouse genomes in wild-type and MCOR conditions

(A) Chromatin interaction heatmaps on human chromosome 13q32.1 encompassing the MCOR locus with positions referencing the human reference sequence hg19. Hi-C sequencing data in human neural progenitor cells (hNPCs) reveal that the MCOR locus is located within a 250 kb TAD (T2) flanked by larger TADs of 1.2 Mb (T1) on the left and 500 kb (T3) on the right. Insulation scores and CTCF ChIP-seq tracks below the heatmap delineate TAD boundaries. The MoDLE Hi-C map predicted using CTCF ChIP-seq data (WT Predict) demonstrates a similar chromatin organization. Removal of CTCF peaks within the MCOR deletion (MCOR Predict) results in the expansion of the T2 TAD into T3, encompassing SOX21.

(B) Magnified view of the predicted WT and MCOR heatmaps, gene track, and insulation scores calculated using hicFindTADs tool on MoDLE highlighting chromatin interaction changes due to MCOR-causal deletions. The projection of the most distant deletion boundaries (MAX) and the critical MCOR deletion (MIN) shows all deletions impact the boundary between T2 and T3 in the wild-type condition. A single asterisk (^∗) denotes the critical MCOR region from the smallest MCOR-causing deletion in Family FR1 (Fares-Taie et al.⁴). A double asterisk (^∗∗) indicates the deletion in the patient with elevated TGFB2 levels in their AH. The dashed blue line shows the SOX21 position across all panels.

(C) Virtual circular chromosome conformation capture (4C) sequencing plot anchored on the SOX21 promoter shows SOX21 interacting within the T3 domain in hNPCs (red arrows). The critical MCOR deletion shifts and enhances SOX21 promoter interactions to the T2 domain, particularly with DCT (blue arrows).

(D) Chromatin interaction heatmap and CTCF ChIP-seq track from mouse neural progenitor cells (mNPC) at chromosome 14qE4 show a 3D organization similar to chromosome 13q32.1 in humans, with the MCOR region located within a 200 kb TAD (t2), flanked by larger TADs (t1 and t3), each spanning 800 kb; positions refer to the mouse reference sequence mm10. Below the mNPC heatmap, MoDLE predictions with normal CTCF ChIP-seq data (mWT Predict) suggest t2 is partitioned into two subTADs (t2.1 and t2.2). This slight difference does not affect the pattern of chromatin reorganization in the MCOR condition compared to the human counterpart. When the three CTCF peaks in the MCOR deletion are removed (mMCOR Predict), the t2.2 subTAD expands into t3 to include Sox21, whose promoter gains interaction with the region comprising Dct, as depicted in (F).

(E) Magnified view of the predicted heatmaps form the WT (mWT) and MCOR (mMCOR) mice, gene track insulation scores calculated using hicFindTADs tool on MoDLE highlighting chromatin interaction changes caused by deleting the mouse region homologous to the critical MCOR-causal deletion in humans. This region crosses the boundary between t2.2 and t3 in the wild-type condition. The dashed blue line marks Sox21 position across the panels.

(F) Virtual 4C plot anchored on the mouse Sox21 promoter shows interaction within t3 in the wild-type condition (red arrow) and a gain of interaction within t2, particularly with Dct (blue arrows).

The human MCOR locus at 13q32.1 aligns with a broad syntenic region on mouse 14qE4

By comparing genomic sequences and NPC-derived Hi-C maps from human and mouse species, we observed that human chromosome 13q32.1 forms a synteny block with mouse chromosome 14qE4. This block exhibits conserved gene composition, organization, and chromatin contacts at and around the MCOR locus, which is situated within a TAD domain (t2) aligning with T2 in the human sequence (Figures 1A and 1D). Utilizing MoDLE, we generated a predictive mouse NPC (mNPC) Hi-C map showing a partition of t2 into two subTADs, t2.1 and t2.2 (mWT Predict; Figures 1D and 1E). Interestingly, MoDLE predicted that deleting the mouse sequence aligning with the critical MCOR region (UCSC Genome Browser: GRCm38/mm10_chr14:118,125,761–118,160,875) results in a TAD reorganization that mirrors the reorganization observed in hNPCs to include Sox21 in t2.2. (mMCOR Predict; Figures 1D and 1E). Further similar to humans, assessing chromatin interactions under wild-type and MCOR conditions by anchoring the analysis at the Sox21 promoter revealed novel interactions with the region comprising Dct at the left boundary of t2.1 (Figure 1F).

Homozygosity for the critical MCOR causing deletion is lethal in C57BL/6 mice

Based on the synteny between human chromosome 13q32.1 and mouse chromosome 14qE4, and the similar TAD reorganization observed under the MCOR condition in both species, we generated a C57BL/6 mouse model of the MCOR disease by ablating the UCSC Genome Browser: GRCm38/mm10_chr14:118,125,761–118,160,875 sequence aligning with the critical MCOR region (B6.cΔMCOR mouse; cΔ indicates the critical deletion; Figure S2). This model was used to study the impact of the deletion on chromatin interactions, iris development, and its potential relevance to GLC and high myopia.

Mating of heterozygous B6.cΔMCOR/+ failed to produce homozygous B6.cΔMCOR/cΔMCOR animals. Collection of embryos from 8.5 to 10.5 embryonic days (E) showed that homozygous animals ceased developing around E8.5 and regressed by E9.5 (data not shown). Further analysis of E9 embryos revealed that homozygous mutants displayed noticeable abnormalities, including smaller size and varying degrees of cardiac hypertrophy, and one individual exhibited pericardial effusion consistent with compromised cardiac function. Additional anomalies included hypoplastic optic and otic vesicles, enlargements in specific facial structures, and irregularities along the body axis (Figure S3A).

For in-depth analysis of these abnormalities in homozygous embryos, serial histological sections were conducted (Figure S3B). Mutant embryos exhibited pyknotic nuclei in key regions, including the neural tube, hindbrain, and neural-crest-derived tissues such as branchial arches and spinal primordia. Massive pyknosis was observed in these neural-crest-derived tissues. Additionally, scattered pyknotic cells were found in mesenchyme, neural tube, and endodermal regions throughout the embryos (Figure S3B). Although the mutant embryos displayed a well-regionalized and differentiated heart, the presence of red blood cells was noticed in the pericardium, potentially indicating myocardium issues. While the myocardium in mutants was thinner compared to controls, their developmental delay complicated conclusive assessments of abnormal differentiation (Figure S3B). Of note, placenta sections showed no apparent differences between mutants and controls (data not shown), suggesting that observed cardiovascular abnormalities are not linked to placental defects.

Together these observations, in particular the pattern of pyknosis in mutant embryos, support defective neural crest migration and differentiation.⁴² In contrast to homozygous animals, heterozygous littermates (B6.cΔMCOR/+; referred to thereafter as the MCOR mouse model) developed normally (Figure S3A).

Heterozygosity for the critical MCOR causing deletion causes reduced pupil size

The irises of two-month-old B6.cΔMCOR/+ mice were non-transilluminable (data not shown), but pupillometry showed a statistically significant moderate reduction (p < 0.01) in baseline pupil size compared to wild-type littermates (B6.WT; Figure 2A), while their response to light remained unaffected (Figure S4).

Figure 2.

Pupillary and molecular phenotype of the B6.cΔMCOR/+ mouse

(A) Assessment of pupil diameter in B6.cΔMCOR/+ mice using a pupillometer reveals a moderate but statistically significant reduction in baseline pupil size compared to B6.WT animals (^∗∗p < 0.01, n = 8 and 7 two-month-old animals per group, respectively). The graph plots individual data points with mean values and SEM error bars. The p value was calculated using one-way/two-way ANOVA.

(B) Analysis of the impact of the critical MCOR deletion on the expression of Dct, Tgds, Gpr180, Sox21, and Abcc4 mRNAs using RNA-seq datasets from IrCB of B6.cΔMCOR/+ and B6.WT mice shows that the deletion induces ectopic expression of Sox21; Note that the abundance of Tgds and Gpr180 is halved in B6.cΔMCOR/+ mice carrying the heterozygous deletion of the MCOR region. Erros bars depict SEM. The p values were calculated using an unpaired t test (∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant).

(C) Real-time qPCR analysis of Dct and Sox21 abundance in IrCB RNA extracts from newborn B6.cΔMCOR/+ and B6.WT mice (n = 5 and 3, respectively) are consistent with RNA-seq data analysis. The graph plots individual data points with mean values and SEM error bars. Statistical significance was determined using Fisher's PLSD test following ANOVA (∗∗∗p < 0.01; ns, not significant).

(D) Western blot analysis of IrCB protein extracts shows detection of SOX21 in B6.cΔMCOR/+ but not in B6.WT counterparts. β-actin (BACT) serves as the loading control, and C+ denotes glial cells which express endogenously SOX21.

Sox21 and its protein product are ectopically expressed in IrCB tissue of the MCOR mouse model

To assess the influence of the critical MCOR-causing deletion on gene expression in IrCB tissue, we conducted transcriptome sequencing on newborn B6.cΔMCOR/+ and B6.WT specimens. Focusing on the four genes within the reorganized t2 (Dct, Tgds, Gpr180, and Sox21; ENSEMBL: ENSMUSG00000022129, 22130, 22131, and 61517, respectively) and the closest gene in reorganized t3 (Abcc4; ENSEMBL: ENSMUSG00000032849), we observed that mRNA levels of Tgds and Gpr180 in B6.cΔMCOR/+ animals were approximately half of those in B6.WT specimens (0.49 and 0.60, respectively; p < 0.001; Figure 2B), consistent with heterozygosity for the deletion, which removes one copy of each of them. Interestingly, Sox21 showed expression in the IrCB of B6.cΔMCOR/+ but not B6.WT (p < 0.01), while the expression of Dct and Abcc4 remained consistent between heterozygous and wild-type counterparts (Figure 2B). These observations strongly suggest that the critical MCOR-causing deletion induces ectopic expression of Sox21.

Real-time qPCR and western blot analyses were conducted on IrCB RNA and protein extracts from adult mice using specific primers for Sox21 and a highly specific SOX21 antibody,⁴³ respectively. These analyses demonstrated the expression of Sox21 (Figure 2C) and the presence of its protein product in B6.cΔMCOR/+ mice, conspicuously absent in B6.WT littermates (Figure 2D), indicating continuous ectopic expression.

SOX21 is specifically detected in cells expressing DCT in the IrCB of the MCOR-mouse model

Immunohistochemistry (IHC) analysis of the IrCB in C57BL/6 animals is challenging due to the high pigmentation of the tissue. To circumvent potential alterations to IrCB integrity caused by depigmentation protocols, we opted to generate albino cΔMCOR/+ and WT lines by mating B6.cΔMCOR/+ animals with tyrosinase (Tyr)-negative Swiss albino (CFW) mice. Real-time qPCR analysis of Sox21 expression in irises of two-month-old animals from the resulting lines, referred to as SW.cΔMCOR/+ and SW.WT, revealed continued ectopic expression, despite the inactivation of tyrosinase (Figure S5). We performed IHC analysis on IrCB sections from adult animals using antibodies specific to SOX21 and DCT, respectively. The dopachrome tautomerase DCT (UniProt: EC 5.3.2.3), also recognized as TYRP2, operates as a melanogenic enzyme located just downstream of tyrosinase and serves as a distinctive marker for melanocyte lineages.^44,45 It is present in the pigmented epithelium layers of the IrCB, respectively, the iris posterior and CB anterior layers. IHC demonstrated robust staining of DCT in the posterior iris and anterior CB epithelia in both SW.cΔMCOR/+ and SW.WT animals (Figure 3).

Figure 3.

SOX21 and DCT expression in the IrCB of newborn albino SW.cΔMCOR/+ and SW.WT mice

Immunohistochemistry analysis of iris and CB in newborn SW.cΔMCOR/+ and SW.WT mice illustrate the detection of SOX21 and DCT in the PEL of the iris and the AEL of the CB.

AEL, anterior epithelium layer; CB, ciliary body; CP, ciliary process; DM, dilator muscle; PEL, posterior epithelium layer; S, iris stroma. Scale bars: 20 μm.

Interestingly, in SW.cΔMCOR/+ animals, SOX21 detection was limited to DCT-expressing epithelium layers of the IrCB (Figure 3). At this age, we observed no SOX21 staining in the AEL of the iris, which is the region from which the dilator muscle forms during the embryonic life (Figure 3). Consistent with the ectopic expression of Sox21 in mutant mice, SOX21 was undetectable in the iris and CB of SW.WT littermates (Figure 3).

Combining RNAscope in situ hybridization (ISH) of Sox21 mRNA and immune staining of DCT, we investigated their spatiotemporal pattern of expression from 10 embryonic days to birth (postnatal day 0, P0) in SW.WT and SW.cΔMCOR/+ eyes (Figure 4). In WT eyes, Sox21 mRNA was detected in the proximal part of the optic vesicle (OV) at E10 and in the dorsal neuroectoderm at E10.5 and E11. However, from E11.5 onwards, although DCT staining became evident, Sox21 mRNA was no longer detectable, except in the surface ectoderm forming the eyelid and hair follicles, beginning around E15 (Figure S6). In mutant eyes, Sox21 mRNA was initially detectable in the proximal part of the OV at E10 and in the dorsal neuroectoderm at E10.5 and E11, similar to its expression in WT littermates. However, from E10.5 onwards, Sox21 mRNA was observed in the outer layer of the optic cup and later in the pigmented structures derived from this layer (Figure 4A). These pigmented structures in the developing eye include the retinal pigmented epithelium (RPE), as well as the AELs of the iris (reversed in adults) and CB (Figure 4B). They consistently expressed DCT, which became readily detectable from E11.5 onward (Figures 4A and 4B).

Figure 4.

Spatiotemporal pattern of expression of Sox21 and DCT in the developing eye of albino SW.cΔMCOR/+ and SW.WT mice

(A) In situ hybridization (ISH) combined with immunohistochemistry analysis of Sox21 mRNA and DCT during eye development, spanning from the invagination of the optic vesicle (OV) into the optic cup (OC) to the formation of the IrCB. Sox21 mRNA is stained in red, and DCT is stained in green. Autofluorescence mainly arising from blood vessels appears in cyan. In SW.cΔMCOR/+ eyes, Sox21 expression begins at E10.5, covering the entire outer layer of the OC. This layer develops into the retinal pigment epithelium (RPE) and the anterior epithelium of both the ciliary body (CB) and iris (AEL) at its edge (outlined by white dotted lines) by E15.5. Sox21 aberrant expression precedes the appearance of DCT in the developing eye. As the RPE differentiates, DCT expression decreases, mirroring the decline in Sox21 expression during the late stages of development. In the anterior epithelium of the ciliary body (aCB), both genes maintain stable expression from development to maturation, unlike the iris anterior epithelium (AEL) and iris posterior epithelium (PEL). DCT and Sox21 are expressed during the prenatal period in the developing AEL but not postnatally. In contrast, in the PEL, expression of both genes begins prenatally and continues into the postnatal period. In SW.WT eyes, Sox21 expression initiates in the proximal region of the developing OV at embryonic day 10 (E10) but is absent in the forming OC) (depicted by the orange bar in the schematic summary); note that Sox21 is detectable by E15.5 in the eyelid and hair follicles (white arrows). AEL, anterior epithelium layer; CB, ciliary body; CP, ciliary process; DM, dilator muscle; L, lens; PEL, posterior epithelium layer; R, retina; S, iris stroma; SM, sphincter muscle. Scale bars: 100 μm.

(B) Schematic summary of Sox21 and DCT expression dynamics during iris development showing that under normal conditions, Sox21 expression is confined to the OV stage (orange bar), whereas in mutant eyes, it persists throughout eye development (red bars) and mirrors DCT expression (green bars); note that at the earliest stage (E10), Sox21 mRNA precedes detection of DCT, possibly due to enhancer priming before gene transcription initiation. Interestingly, the expression patterns of Sox21 and DCT in mutant iris epithelia correspond to reported changes in iris pigmentation in pigmented mice: initially heavily pigmented AEL loses melanin postnatally, while the PEL gains melanin centrifugally, reaching full pigmentation in adulthood. These observations suggest that DCT regulates Sox21 expression in mutant animals. The dilator muscle (depicted in dark blue) begins forming shortly after the sphincter (light blue), around E17.5 and E18.5, respectively. Thus, during the initial stages of dilator muscle development, Sox21 expression is present in the AEL, which contributes to the dilator muscle (DM). The cessation of Sox21 expression in the AEL by P0 in mice indicates a likely early developmental defect underlying the disease. The shaded box denotes the period from birth to 2 months, where information on iris AEL pigmentation and the stages of DM development are currently limited. aCB, ciliary body anterior epithelium; AEL, iris anterior epithelium; C, cornea; DM, dilator muscle; EL, eyelid; IL, optic cup inner layer; L, lens; LP, lens placode; LV, lens vesicle; NE, neuroectoderm; OC, optic cup; OL, optic cup outer layer; OV, optic vesicle; pCB, ciliary body posterior epithelium; PEL, iris posterior epithelium layer; R, retina; RPE, retinal pigment epithelium; S, iris stroma; SE, surface ectoderm; SM, sphincter muscle. Scale bars: 100 μm.

The absence of Sox21 expression in the non-pigmented posterior epithelium layer (PEL) of both the developing and adult CB, along with the loss of both Sox21 and DCT expression in the AEL of adult irises compared to developing irises (Figures 4A and 4B), further strengthens the evidence for the co-expression of Sox21 and DCT in SW.cΔMCOR/+ eyes.

Integrated analysis of SOX21 targets and gene deregulation in MCOR mouse reveals disrupted genes linked to iris development and MCOR symptoms

We investigated the impact of the loss of the critical MCOR-causing region on the expression of genes located outside of TADs 1 and 2 by analyzing RNA-seq data from the IrCB of new-born B6.cΔMCOR/+ and B6.WT mice. This study detected 2,500 differentially expressed genes (DEGs ≥1.5-fold change [FC], p < 0.05; Figure S7A). Gene ontology enrichment analysis revealed a significant enrichment of genes associated with crucial processes such as cellular commitment, development, differentiation, and migration, particularly in the context of neural and sensory systems. Notably, the top pathways identified include cell fate commitment, forebrain development, regulation of cell differentiation, retina morphogenesis, regulation of neuron differentiation, and neuron projection development (Figure S7B). Consistent with these findings, previous research has highlighted the role of SOX21 in regulating adult neurogenesis in vivo, contributing to the generation of new neurons by repressing Hes5 expression.⁴³ Notably, our study reveals a downregulation of Hes5 in the B6.cΔMCOR/+ IrCB (FC = 0.56, p < 0.01; Figure S7), aligning with the established regulatory function of SOX21. Our study identified additional deregulated neurogenesis-associated genes whose protein products interact with SOX21. Notably, Sox2, crucial for anterior segment formation in eye development,^{46,47,48,49,50,51} showed significant downregulation (FC = −1.85, p < 0.01; Figure S8).

Furthermore, we noted the upregulation of key genes linked to major signaling pathways in iris development, including WNT/β-catenin, TGFB, and bone morphogenetic protein (BMP) signaling pathways (Figure S9).¹ Our investigation into the WNT/β-catenin pathway highlights the substantial elevation of Wnt2b expression (FC = 3.57, p < 0.01 Figure S9), a gene crucial for specifying iris progenitor cells toward a myoepithelial fate.¹ In the context of BMP and TGFB signaling pathways, we observed upregulation in Bmp7 (FC = 1.92, p < 0.05 Figure S9), implicated in iris smooth muscle generation,¹ and Tgfb2 (FC = 1.6, p < 0.01), known for its role in IrCB specification, as well as in GLC^9,10,11,12 (Figure S9), and potentially associated with high myopia.^13,14

Furthermore, we observed an enrichment of genes associated with the regulation of smooth muscle cell proliferation, though with a lower significance ranking (top 45; Figure S7B). Notably, our analysis unveiled a decrease in expression of Des (FC = −2.0, p = 0.012), encoding desmin intermediate filaments known to be deficient in the iris anterior epithelium layer of MCOR-affected individuals.^3,52 To explore if the smaller pupil size in B6.cΔMCOR/+ is linked to reduced DES levels, we used immunohistochemistry with DES and smooth muscle actin (SMA) antibodies as control on mouse iris sections. SMA stained both sphincter and dilator muscle fibers (Figure S6), but DES antibodies failed to detect the protein in both B6.cΔMCOR/+ and B6.WT irises.

Tgfb2 is a direct target of SOX21 in the IrCB of the MCOR mouse model

To search for SOX21 target genes among the deregulated ones, we conducted ChIP-seq analysis using the SOX21 antibody on IrCB from newborn B6.cΔMCOR/+ mice.

This analysis uncovered 25 DNA regions throughout the genome (p ≤ 0.1), comprising 10 intragenic (all intronic), 14 intergenic, and 1 in promoter-TSS regions, respectively (Table 1). We examined deregulated expression (≥1.5-fold; p ≤ 0.05 B6.cΔMCOR/+ versus B6.WT animals) among these genes and the nearest genes in cases where the binding occurred outside a gene. This search identified a unique gene: Tgfb2; the remaining genes either exhibited no expression in the IrCB or showed no deregulation (Table 1). SOX21 binding to Tgfb2, as determined by ChIPseq, was unequivocal (p < 0.0001) and strongly supported by JASPAR⁵³ analysis, which identified a consensus SOX21-binding sequence in the 252 bp intronic region specified by ChIP-seq (UCSC Genome Browser: GRCm38/mm10_chr1:186,698,304–186,698,555; 5.9 kb downstream of the consensus donor splice site of the 16-kb-long intron 1; Figure 5A). Consistent with an association between SOX21 and Tgfb2, analysis of IrCB RNA-seq datasets revealed dysregulation of the TGFB2 signaling pathway in B6.cΔMCOR/+ mice compared to B6.WT littermates (Figure 5B).

Table 1.

DNA regions identified by ChIP-seq analysis for SOX21 binding in the IrCB of B6.cΔMCOR/+ mice

Gene	Gene description	Chr	Peak score	Annotation	Distance to TSS	Nearest promoter ID	RNA-seq
Hipk2	homeodomain interacting protein kinase 2	6	104.51/26.18	intron (GenBank: NM_010433, intron 2 of 14)	76,864	GenBank: NM_001136065	ND
Gm13580	predicted gene 13580	2	46.62	intergenic	−8,724	GenBank: NR_046065	ND
Gm17644	predicted gene 17644	1	43.95	intergenic	−22,756	GenBank: NR_045297	ND
Tgfb2	transforming growth factor, beta 2	1	40.86	intron (GenBank: NM_009367, intron 1 of 6)	7,563	GenBank: NM_009367	×1.6
Mydgf	myeloid derived growth factor	17	40.64/5.14	intron (GenBank: NM_172624, intron 10 of 20)	−16,198	GenBank: NM_080837	ND
Olfr854	olfactory receptor 854	9	37.19	intergenic	−8,090	GenBank: NM_146522	ND
Tex14	testis expressed gene 14	11	24.63	intron (GenBank: NM_031386)	−4,435	GenBank: NM_001199293	ND
Gm4736	predicted gene 4736	6	21.86/8.66	intergenic	−99,620	GenBank: NM_053251	ND
Slf1	SMC5-SMC6 complex localization factor 1	13	20.03	intron (GenBank: NM_134071, intron 5 of 20)	25,853	GenBank: NM_134071	ND
4930521E06Rik	RIKEN cDNA 4930521E06 gene	1	18.78	intergenic	−206,375	GenBank: NR_040602	ND
Ctdsp1	C-terminal domain small phosphatase 1	1	17.45	intergenic	−4,548	GenBank: NM_153088	ND
Myo7a	myosin VIIA	7	16.1	intron (GenBank: NM_001256081, intron 39 of 48)	49,182	GenBank: NM_001256083	ND
AA545190	EST AA545190	6	15.75	intergenic	788,031	GenBank: NR_033776	ND
Popdc3	popeye domain containing 3	10	15.56	intergenic	−109,987	GenBank: NM_024286	ND
Gdnf	glial cell-line-derived neurotrophic factor	15	14.96	intergenic	−113,169	GenBank: NM_001301332	ND
1700128A07Rik	RIKEN cDNA 1700128A07 gene	14	14.8	intergenic	−588,884	GenBank: NR_045938	ND
Pcdh9	protocadherin 9	14	14.16	intergenic	1,798,871	GenBank: NM_001081377	ND
Adgrl4	adhesion G protein-coupled receptor L4	3	14.16	intergenic	−303,096	GenBank: NM_133222	ND
Nsmce2	non-SMC element 2 homolog (MMS21)	15	14.11	intron (GenBank: NM_026746)	51,552	GenBank: NM_026746	ND
Wls	Wntless homolog (Drosophila)	3	13.73	intron (GenBank: NR_037590)	12,420	GenBank: NR_037590	ND
Nrarp	notch-regulated ankyrin repeat protein	2	12.17	promoter-TSS (GenBank: NM_025980)	−24	GenBank: NM_025980	ND
Cpa6	carboxypeptidase A6	1	11.31	intron (GenBank: NM_177834)	208,422	GenBank: NM_001289497	ND
Llph	LLP homolog, long-term syNDptic facilitation factor	10	11.22	intergenic	51,806	GenBank: NM_025431	ND
Cartpt	CART prepropeptide	13	10.95	intergenic	−2,410	GenBank: NM_001081493	ND
2810404M03Rik	RIKEN cDNA 2810404M03 gene	8	10.1	intron (GenBank: NR_045497)	143,286	GenBank: NR_045497	ND

The genes listed in this table are restricted to those associated with a peak score greater than 10, corresponding to a p value of 0.1 or higher. The peak location is presented relative to the genomic distance from the transcription start site (TSS). Specifically, the 25 sites from the SOX21 ChIP sequencing data were identified in one or more samples, each of which comprised 16 irises. ND, not deregulated (threshold set at 1.5-fold change).

Figure 5.

SOX21 binding sequences in mouse and human TGFB2 and in vitro and in vivo analysis of TGFB2 expression relative to SOX21 expression

(A) The DNA sequence targeted by SOX21 in the first intron of Tgfb2, identified by ChIP-seq analysis in the mouse B6.cΔMCOR/+, is shown on the left. The corresponding sequence from the human genome assembly GRCh38, located in the first intron of TGFB2 is presented on the right panel. The mouse and human SOX21-consensus motifs, according to the JASPAR⁵³ database, are displayed at the bottom of each panel. The mouse binding sequence (JASPAR: PB0069.1) of SOX21 is highlighted in blue, found in both the human and mouse sequences in the intron. The human motif (JASPAR: MA0866.1), shown in green, is also present in the TGFB2 intron 1, suggesting a potential binding of SOX21 in this location.

(B) RNA-seq heatmap focusing on genes in the TGFB pathway reveals consistent overexpression of Tgfb2 in all four B6.cΔMCOR/+ irises, despite some variations observed among individuals within the same group.

(C) RT-PCR analysis in human glioma cells shows a significant decrease in TGFB2 abundance when SOX21 expression is knocked down using siRNA, compared to treatment with non-targeting siRNA. This supports the role of SOX21 in regulating TGFB2 expression. The graph plots individual data points with mean values and SEM error bars. The statistical significance was assessed using an unpaired t test (∗∗p < 0.01; ∗∗∗∗p < 0.0001).

(D) Analysis of TGFB2 concentration in the AH of an individual affected with MCOR from Family FR1 initially reported by Fares-Taie et al.⁴ whose deletion is shown in Figure 1, and 11 controls using ELISA show an accumulation of the protein in the MCOR subject compared to controls. The graph plots individual data points (each representing TGFB2 levels from a single eye) with mean values and SEM error bars. Statistical significance was not assessed due to the availability of only one MCOR individual.

The iris and ciliary body are both heterogeneous tissues composed of several cell populations, which makes in vitro culture and genetic manipulation highly challenging. Thus, to substantiate the link between SOX21 and TGFB2 (MIM: 190220) expression, we used the human glioma cell line U-251, which endogenously expresses both genes. Notably, the ChIP-seq-specified region is conserved in the human TGFB2 intron 1 orthologous region, containing numerous potential transcription-factor binding sites (UCSC Genome Browser: GRCh38/hg38_chr1:218,352,441–218,352,576; Figure 5A). In U-251 glioma cells, we showed that silencing the expression of SOX21 using siRNAs resulted in a statistically significant reduction of TGFB2 levels, as determined by real-time qPCR (Figure 5C). In contrast, transfection with a non-targeting siRNA did not alter mRNA levels of SOX21 or TGFB2 (Figure 5C), indicating that the reduction in TGFΒ2 expression specifically resulted from SOX21 silencing, further corroborating observations in the mouse IrCB.

TGFB2 concentration is elevated in the AH of an MCOR individual

The ectopic expression of Sox21 in the CB, along with SOX21 binding to a regulatory region of Tgfb2 and its upregulation in B6.cΔMCOR/+ IrCB, prompted us to measure TGFΒ2 concentration in the AH of 12-month-old B6.cΔMCOR/+ and B6.WT mice using an ELISA assay. Concentrations, measured from pooled AH from both eyes, exhibited high variability in both WT and mutant eyes. Analysis of a large number of samples revealed no statistical difference between the two groups (Figure S10).

However, upon analyzing the AH of a 45-year-old individual with non-glaucomatous MCOR, carrying a 50.5 kb deletion that includes the critical MCOR region at 13q32.1 (Family FR1; Fares Taie et al.⁴; Figure 1B), we observed a significant increase in TGFB2 levels compared to 11 individuals without MCOR or GLC (Figure 5D). All AH samples were collected on the same day during cataract surgery. It is interesting that the TGFB2 level was also high in the AH sample from the MCOR individual’s second eye surgery, two weeks after the first one.

This observation provides additional support for the connection between SOX21 and TGFB2, emphasizing the role of TGFB2 in MCOR.

Discussion

Anatomical analyses of postmortem eyes and iris specimens from individuals affected by MCOR have shown iris thinning, stromal atrophy, and a lack of stromal contractile processes in the anterior iris epithelium, suggesting potential muscular development issues.³ However, the underlying mechanisms behind these findings remain unclear. The same holds true for understanding why the disease is often associated with high axial myopia and high IOP GLC. It has been suggested that closing the eyelids to reduce glare from iris transillumination could lead to form-deprivation myopia; however, the absence of a clear link between myopia and oculocutaneous albinism, which also involves iris transillumination and light sensitivity, challenges this hypothesis. Similarly, while high axial myopia and anatomical features of microcoric chamber angles, which may impact AH flow, are proposed contributing factors for GLC, these alone cannot fully account for the condition, as evidenced by the 30% prevalence of GLC in microcoric individuals who consistently display these anomalies.

Our research utilized recent genetic insights into the disease and CRISPR-Cas9 technology to create a mouse model, supporting the role of the SOX21 transcription factor in both abnormal iris development and other eye issues. The role of SOX21 in eye development was documented in the chick⁵⁴ and zebrafish⁵⁵ but not in the mouse. In the chick, Sox21 is transiently activated during the early stages of OV morphogenesis and specification in the lens and retina but is not expressed thereafter⁵⁴; Sox21 expression in the eye ceases before the onset of iris development. Loss of Sox21 function in the chick, similar to findings in zebrafish, disrupts normal lens development,⁵⁶ whereas it has no effect on the mouse eye,⁵⁷ which aligns with our immune histological analysis detecting no expression of SOX21 in the anterior segment of the developing mouse eye.

In our study, we show that deleting the sequence in the mouse genome, equivalent to the critical MCOR region in humans, results in constitutive Sox21 expression. Remarkably, we observe colocalization of Sox21 and the melanogenic marker DCT in both developing and adult eyes. According to the predicted 3D structural organization, Dct and Sox21 are located in two adjacent neighboring TADs, the boundaries of which are disrupted by MCOR-causing deletions, while the genes themselves remain unaltered (Figure 1). In line with prior research associating TAD disruptions with developmental defects,⁵⁸ we propose that the loss of TAD boundaries in MCOR may initiate a reorganization, potentially activating Sox21 through Dct enhancers (Figure 1). The delayed detection of DCT (E11.5) compared to the detection of Sox21 mRNA (E10.5) is consistent with studies demonstrating that enhancer priming precedes the initiation of gene transcription.^59,60,61

In line with the melanogenic DCT expression pattern, we detected ectopic expression of Sox21 in the pigmented PEL of the iris in adult MCOR mouse model, while it was absent in the AEL responsible for dilator muscle formation.^62,63 Contrastingly, during embryonic development, the iris AEL is pigmented and expresses DCT. While literature lacks reports on iris muscle development in mice, in rats, the dilator muscle initiates development around E18.5 and continues postnatally.⁶² In our study, we found DCT in the iris AEL of E18.5 and P0 WT mice and both DCT and Sox21 in MCOR mice. This suggests that abnormal Sox21 expression in the AEL of the MCOR-mouse iris coincides with the onset of dilator muscle development and decreases during postnatal development, along with the loss of DCT expression (Figures 4A and 4B). These findings might elucidate histological observations of microcoric iris specimens, displaying poor differentiation, disordered fibers lacking myofilaments and intermediate filaments, or even complete growth inhibition, which suggest issues in the terminal differentiation stages of the anterior iris epithelium (reviewed by Angee et al.).³ In the mouse, our histological analysis failed to detect such anomalies. This could be attributed to the minimal iris phenotype of the mutant mouse, potentially linked to reduced Sox21 expression, divergent timelines in iris dilator muscle development, or dissimilarities in the 3D architecture of the region encompassing the MCOR locus between humans and mice. While the iris phenotype in the mutant mouse is minimal, possibly due to the dilator muscle playing a less significant role in mice compared to humans, it is noteworthy that we observed a reduced expression of Des in the IrCB of newborn mutant mice. While we couldn’t directly confirm the reduced abundance of DES in the iris AEL due to the lack of suitable antibodies for immunocytochemistry analysis, this observation aligns with previous studies on iris anatomy. These studies have reported the absence or severe reduction of DES positive fibers in the anterior epithelium of individuals with microcoria.^3,52

Furthermore, RNA-seq analysis revealed the deregulation of key genes known to be crucial in signaling pathways important for iris development, such as Wnt2b. In the early stages of eye development in chicken and mouse embryos, Wnt2b expression is concentrated at the marginal tip of the developing retina, with a notable increase in the dorsal part of this region, responsible for forming the pigmented epithelium of the iris.¹ Interestingly, this specific region coincides with the location of ectopic expression of Sox21 observed in B6.cΔMCOR/+ mice (Figures 4 and S6). The lack of SOX21 binding to Wnt2b regulatory sequences, as revealed by ChIPseq analysis, suggests that the deregulation is likely indirect to Sox21 induction. Nonetheless, the observed spatial correlation between Sox21 expression and the regional modulation of Wnt2b provides valuable insights into the regulatory mechanisms governing early eye development and the formation of the iris pigmented epithelium.

Among the deregulated genes identified by RNA-seq, Tgfb2 is the sole gene found to bind SOX21 in mutant IrCB. This binding seems to occur through a consensus binding sequence located in intron 1. This observation aligns with a growing body of evidence across various organisms, suggesting that for a subset of genes, the critical sequences regulating expression are not primarily situated in the promoter but within introns within the first kilobase of transcribed sequences.⁶⁴

Presently, the specific mechanisms behind defective muscle development in both the MCOR mouse model and individuals with MCOR remain elusive. One plausible hypothesis suggests the involvement of TGFB2. This factor is notably secreted within the periocular mesenchyme and plays a pivotal role in the development of various ocular components, including the anterior chamber—critical for AH outflow—and the stroma of the iris and CB.⁶⁵ Although the iris dilator muscle originates from the optic cup margins, which derives from the neuroectoderm, its myofibers manifest within the stroma. Therefore, an aberrant expression of TGFB2 within the iris could potentially disrupt stromal development, subsequently affecting the proper formation of the dilator muscle within this context. This proposed link underscores the importance of investigating the interplay between TGFB2 signaling and muscle development in the iris, offering a potential avenue for understanding the muscle-related anomalies observed in the context of MCOR. Remarkably, several studies reported significantly elevated levels of TGFB2 in the AH of individuals with POAG, as well as in cultured glaucomatous cell strains and isolated human glaucomatous trabecular meshwork (TM) tissues.^11,12 The cause and cellular source of TGFB2 in glaucomatous eyes remain elusive. However, studies have demonstrated that cells from the TM, through which AH is drained out at the iridocorneal angle, express an active TGF receptor complex and respond to exogenous TGFB2.^66,67 This has been reported to result in increased synthesis of extracellular matrix (ECM) proteins, accumulation of which increases resistance to aqueous outflow, leading to elevated IOP and GLC in humans, primates, cats, and mice.^{9,10,11,12,68}

This knowledge, combined with (1) the histopathologic examination of TM biopsies from two microcoric brothers with elevated IOP, which revealed ECM accumulation,⁶⁹ and (2) an elevated concentration of TGFB2 in the AH of an individual with MCOR, implies that GLC associated with MCOR may be initiated by SOX21. The strength of this inference is heightened by our findings, which provide evidence of ectopic Sox21 expression in the CB—where the AH is produced—specifically in the non-pigmented PEL.⁷⁰ Furthermore, a recent study has uncovered a significant correlation between elevated TGFB2 levels in the AH and axial elongation.^13,14 This discovery not only suggests a connection between TGFB2 and, by extension, SOX21 but also extends the potential link to high myopia.

Finally, we demonstrate that homozygosity for the critical MCOR-causing deletion is lethal in the mouse. While the possibility of Sox21 playing a role cannot be ruled out, it is more likely that lethality results from the homozygous deletion of Tgds. Notably, biallelic mutations in TGDS (MIM: 616146) have been documented to lead to a severe developmental disorder characterized by the Pierre Robin anomaly known as Catel-Manzke syndrome (MIM: 616145). Intriguingly, to our knowledge, no instances of homozygosity or compound heterozygosity for loss-of-function alleles were reported, suggesting that such a situation could be lethal. An alternative explanation could consider the aberrant accumulation of pyknotic nuclei in developing ectodermal and neuroectodermal derivatives such as rhombomeres, neural-crest-derived cranial ganglia, and placodes observed as a cause of lethality in mutant embryos. From a morphogenetic standpoint, the pathophysiological accumulation of pyknotic cells is linked to embryotoxicity,^42,71 attesting of teratogenic or pathogenic conditions and accounts for cell degeneration initiated by apoptosis. In neuronal and glial-derived tissues in the central and peripheral nervous system, pyknosis is interpreted as signs of cell degeneration that could be enhanced through the activation of Ca2+-activated proteases⁷² or metalloproteases.⁷³ Notably, tissues exhibiting intense accumulation of pyknotic cells coincide with active production of TGFB2 during normal development.^74,75,76 In mice lacking Tgfb2, inhibiting this pathway prevents neuroblastic lineages from apoptosis.⁷⁷

In conclusion, our study demonstrates that the critical submicroscopic deletion associated with MCOR potentially disrupts the 3D architecture of the region, leading to modified gene interactions. This alteration results in the ectopic expression of the Sox21 transcription factor in Dct-expressing pigmented cells within the iris and CB of MCOR mice. Importantly, we suggest that in MCOR, both GLC and high myopia stem directly from SOX21-mediated TGFB2 overexpression in the CB rather than arising due to abnormal iris development. Consequently, we propose that SOX21 serves as a crucial link, connecting iris malformation, high myopia, and GLC in MCOR. This positions MCOR as an invaluable model for scrutinizing eye development and unraveling the underlying mechanisms of common myopia and POAG.

Data and code availability

Supporting RNA-seq and ChIP-seq data are available in the BioStudies database (http://www.ebi.ac.uk/biostudies) under accession numbers S-BSST1320 and S-BSST1472 via the following links: https://www.ebi.ac.uk/biostudies/studies/S-BSST1320?key=fe81ffbe-98e2-467b-97ac-ffb55d563d5e and https://www.ebi.ac.uk/biostudies/studies/S-BSST1472?key=9283105e-2710-4f3c-8fe6-663a1f57939a, respectively.

Acknowledgments

This research has been generously supported by grants from the Agence Nationale de la Recherche (ANR#-20-CE 12-0019-01), the Institut National de la Santé et de la Recherche Médicale”(Inserm), MSDAVENIR (DEVO-DECODE program), the Fondation Visio, and the Association Retina France.

Author contributions

J.-M.R. and L.F.T. designed the project. B.P., J.-L.V., and P.D. generated the MCOR mice. C.A. and E.E. performed and interpreted the molecular, histological, and imaging experiments. D.H. analyzed the Hi-C-seq data. C.K. and S.V.C. performed the pupillometry experiments. E.D., J.K., P.C., J.P., and N. Chassaing provided clinical data. M.Z. and C.B.-F. performed sequencing of the RNA samples. N. Cagnard performed bioinformatic analyses. B.N. and J.C. provided technical assistance. C.A. wrote the manuscript and prepared the figures with contributions from all co-authors. J.-M.R. and L.F.T. reviewed all of the data and edited the manuscript.

Declaration of interests

The SOX21-TGFB2 pathway in iris development, axial myopia, and GLC has been officially patented under the title “Methods and pharmaceutical compositions for treating ocular diseases” (WO/2021/245224). The inventors of this patent are J.-M.R., L.F.T., B.N., C.A., and J.K.

Declaration of generative AI and AI-Assisted technologies in the writing process

During the preparation of this work the author(s) used ChatGPT in order to edit English. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Published: September 17, 2024