The 3D Genome Structure of a Blcap-Linked Silencer Loop Regulates Terminal Differentiation During Lens Fiber Cell Denucleation

Abstract

Purpose

Lens fiber cells (LFCs) remove their nuclei and organelles during terminal differentiation through a controlled process that resembles, but does not execute, apoptosis. The mechanisms that inhibit apoptotic pathways while permitting nuclear degradation remain unclear. Using Hi-C, CUT&Tag, and RNA-seq, we mapped chromatin and epigenetic changes during the LEC-to-LFC transition. We focused on a distal silencer region (DSR)–mediated chromatin loop near Blcap and its function in repressing apoptosis-related signaling to support proper nuclear clearance and lens transparency.

Methods

Hi-C sequencing mapped chromatin architecture in E16.5 mouse LECs and LFCs, focusing on apoptosis-related loci. CUT&Tag profiling for H3K27me3, CTCF, and SMC3, integrated with RNA-seq, identified a novel silencer element and an LFC-specific chromatin loop linking Blcap to the DSR. To assess function, an adeno-associated virus-delivered saCas9 dual-sgRNA system was used to delete loop anchor regions in vivo. Functional effects were validated through DNA-FISH, Western blotting, flow cytometry, and phenotypic analyses.

Results

Multiomics analysis revealed extensive chromatin remodeling during LEC-to-LFC differentiation, including A/B compartment switching, silencer reprogramming, and transcriptional changes in differentiation- and apoptosis-related genes. A specific loop tethering Blcap to the DSR was identified in LFCs. Loop disruption abolished their spatial proximity, derepressed Blcap, reduced anti-apoptotic genes Bcl2 and Bcl2l1, increased apoptosis, and caused lens opacification.

Conclusions

The Blcap-DSR chromatin loop plays a key protective role by repressing Blcap and maintaining anti-apoptotic balance in LFCs. Through silencer-associated looping and epigenetic reprogramming, LFCs achieve controlled nuclear and organelle clearance without triggering cell death, ensuring proper differentiation and lens transparency.

The differentiation of lens epithelial cells (LECs) into lens fiber cells (LFCs) is essential for the formation and maintenance of normal lens morphology, and its dysregulation may lead to lens opacity and cataract.^1,2 During terminal differentiation, LFCs maintain lens transparency by systematically eliminating organelles such as nuclei, mitochondria, endoplasmic reticulum, and Golgi apparatus.² During LFC denucleation (LFCD), fiber cells undergo nuclear condensation and a CDK1-dependent disintegration of the nuclear envelope that allows DNase IIβ entry to execute chromatin degradation.^3–5 These nuclear and structural remodeling events share certain molecular and morphological characteristics with apoptotic processes,^5–7 LFCs also express multiple regulators typically associated with apoptosis, suggesting that components of apoptotic signaling are repurposed to facilitate fiber cell differentiation.^7–11 Notably, canonical apoptotic regulators such as p53 and caspase-3, although central to programmed cell death, are also indispensable for proper lens fiber cell maturation.^5,9 Similar to erythroid differentiation,¹² although lens fiber cell differentiation involves apoptosis-like mechanisms, the molecular pathways responsible for the full execution of apoptosis appear to be negatively regulated, enabling the cells to remain viable and preserve lens transparency after the clearance of organelles and nuclei.^11,13,14 The molecular mechanisms underlying this protective regulatory program remain poorly defined.

LFCD is accompanied by extensive chromatin condensation and remodeling four and three-dimensional (3D) genome structural changes are critical to this process. Recent studies have demonstrated that dynamic alterations in 3D genome architecture regulate spatiotemporal gene expression through the formation of enhancer-promoter or silencer-promoter loops, enabling precise control of cell fate.¹⁵ For example, dynamic chromatin loops mediate silencing and activation of globin gene clusters during erythroid differentiation,¹⁶ whereas CTCF-mediated loops spatially isolate pro-differentiation genes to maintain stemness during neuronal development.¹⁷ These findings suggest that lens differentiation may similarly rely on specific 3D epigenetic architectures for regulation.^18,19 We hypothesize that chromatin structural changes during lens development may play a dual role: regulating differentiation-related gene expression while preventing aberrant activation of pro-apoptotic pathways. However, high-resolution chromatin dynamics during LEC-to-LFC transition remain uncharacterized.

In this study, we integrated Hi-C, CUT&Tag, and RNA-seq data to systematically map the 3D chromatin dynamics and epigenetic regulatory networks in E16.5 murine LECs and LFCs. Notably, we identified a “silencer-promoter” (S-P) loop associated with the pro-apoptotic factor Blcap during LFCD. Functional disruption of this loop aberrantly activated Blcap, triggered excessive apoptotic signaling in LFCs, and led to lens opacity. Our findings provide novel insights into the epigenetic mechanisms underlying lens pathologies such as cataracts and propose a new paradigm for coordinated cellular regulation during organ development.

Methods

Animals

C57BL/6J mice (JAX no. 000664; Jackson Laboratories, Bar Harbor, ME, USA) and CAST/EiJ mice (JAX no. 000928; Jackson Laboratories) were used in this study. All animal experiments were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Animal Ethics Committee of Zhongshan Ophthalmic Center. Mice were maintained under a controlled light/dark cycle with access to food and water as desired.

Isolation of Single-Cell Suspensions from LECs and LFCs

To investigate chromatin interactions in lens cells during the nuclear expulsion stage, E16.5 C57BL/6J embryonic mice were used. A standard protocol was followed to isolate the lens capsule and fiber cells.^19,20 Specifically, the lenses were dissected and the surrounding peripheral tissue was removed. The isolated lenses were washed three times with 10 mL of HBSS (no. C0219; Beyotime Institute of Biotechnology, Jiangsu, China), and placed posterior side up in a 3.5 cm culture dish (no. FCFC020-10pcs; Beyotime Institute of Biotechnology) containing M199 medium (no. M775831-500ml; Mackin Medical, Broomall, PA, USA). A small incision was made at the posterior pole of the lens capsule, and four to six small segments were peeled off from the posterior pole to the lens equator using fine tweezers. The fiber clumps were then separated by hydrolysis. Subsequently, the lens capsules and fiber clumps were digested separately in 0.05% trypsin (no. 25200072; Gibco, Thermo Fisher Scientific, Waltham, MA, USA) at 37°C for five to 10 minutes. After digestion, the cells were dissociated into a single-cell suspension by pipetting, and digestion was terminated by adding FBS (no. 25200072; Gibco, Thermo Fisher Scientific). The cells were then spun in a centrifuge at 300g for five minutes and resuspended in PBS (C10010500BT; ThermoFisher, St. Louis, MO, USA). Viable and dead cells were assessed using AO/PI staining (RE010212; Countstar, Shanghai, China) and analyzed with a fully automated fluorescent cell analyzer (Rigel S2; Countstar). The viability of the cell suspensions was consistently above 85% to 95%. To ensure sufficient statistical power, two biological replicates were prepared for each sequencing sample.

Hi-C Library Construction and Sequencing

Each biological replicate comprised LECs and LFCs isolated from 30 lenses. Cells were crosslinked at room temperature with 1% formaldehyde (no. F301839; Aladdin Scientific, Riverside, CA, USA) for 10 minutes, followed by quenching with 0.125 M glycine (no. G800880; Shanghai Macklin Biochemical Co., Ltd., Shanghai, China) for five minutes. After cell lysis, endogenous nucleases were inactivated using 0.3% SDS (no. S19770; Acmec Biochemical, Shanghai, China). Chromatin was then digested with 100 U MboI (no. R0147S; New England Biolabs, Ipswich, MA, USA) and labeled with biotin-14-dCTP (no. 19518018; Invitrogen, Carlsbad, CA, USA). Fragmented chromatin was ligated using 50 U T4 DNA ligase (no. M0202S; New England Biolabs). After reversing the crosslinks, the ligated DNA was purified using the QIAamp DNA Mini Kit (no. 51304; Qiagen, Hilden, Germany) according to the manufacturer's protocol. The DNA was subsequently sheared to 300–500 bp fragments, followed by end repair, A-tailing, and adaptor ligation. Biotin-labeled fragments were enriched using streptavidin-coated magnetic beads and then amplified via PCR. The final Hi-C libraries were quantified and sequenced on the DNBSEQ-T7 platform (BGI Group, Shenzhen, China).

Hi-C Data Analysis

Raw Hi-C sequencing reads were first subjected to quality control and adapter trimming using Trimmomatic (v0.39) (http://www.usadellab.org/cms/index.php?page=trimmomatic)²¹ yielding high-quality clean reads for downstream analysis. The processed data were then aligned to the mouse reference genome (mm39) and quality-controlled using the Juicer pipeline (v1.6) (https://github.com/aidenlab/juicer),²² resulting in whole-genome chromatin interaction matrices in .hic format. To achieve high-resolution chromatin architecture profiling, interaction matrices were normalized at 5, 10, 25, and 40 kb resolutions across all samples. A/B chromatin compartments were defined using principal component analysis (PCA) via the eigenvector module in Juicer Tools (https://github.com/aidenlab/juicertools).²² The first principal component (PC1) was typically used for compartment classification, where positive and negative PC1 values indicated compartments A and B, respectively. Chromatin loops were identified using the HICCUPS module in Juicer Tools²² using default parameters to detect locally enriched interaction signals indicative of loop structures. Finally, chromatin interaction profiles and loop architectures were visualized using Juicebox²² and Hicplotter (https://github.com/akdemirlab/HiCPlotte).²³

CUT&Tag Library Construction and Sequencing

Each biological replicate included LECs and LFCs obtained from 7 lenses. CUT&Tag libraries targeting H3K27me3, CTCF, and SMC3 were prepared using the Hyperactive Universal CUT&Tag Assay Kit (no. TD903; Vazyme Biotech Co., Ltd., Nanjing, China) and the TruePrep Index Kit V2 (no. TD202; Vazyme Biotech Co., Ltd.) following the manufacturer's instructions. Briefly, cells or nuclei were immobilized on ConA-coated magnetic beads and permeabilized using digitonin. The pA-Tn5 transposase, guided by specific antibodies, precisely binds to DNA sequences near the target protein, adding sequencing adapters to both ends of the DNA fragments. The resulting libraries were amplified by PCR, purified using AMPure magnetic beads, and quality-checked on an Agilent Bioanalyzer 2100 system (Agilent Technologies, Winooski, VT, USA). Antibodies used included anti-H3K27me3 (no. 9733S; Cell Signaling Technology, Danvers, MA, USA), anti-CTCF (no. 3418S; Cell Signaling Technology), anti-SMC3 (no. 703926; Invitrogen), and the secondary antibody Goat Anti-Rabbit IgG (H+L) (no. 35401; Cell Signaling Technology). Libraries were then sequenced on the Illumina Novaseq 6000 platform (Illumina, San Diego, CA, USA) with 150 bp paired-end reads.

CUT&Tag Data Analysis

Raw sequencing reads in FASTQ format were first quality-controlled and preprocessed using fastp (v0.20.0).²⁴ During this step, sequencing adapters, reads containing more than five ambiguous bases (“N”), and other low-quality reads were removed. Low-quality reads were defined as those in which over 40% of bases had Phred scores below 15 or reads shorter than 15 bp after trimming. The resulting high-quality reads, referred to as clean reads, were subsequently aligned to the mouse reference genome (mm39) and corresponding annotation files using Bowtie2 (v2.5.4).²⁵ Peak calling was performed using SEACR (v1.3) (https://seacr.fredhutch.org/)²⁶ with default parameters. To identify genomic regions co-occupied by CTCF and SMC3, we first processed the biological replicates for each target separately. CUT&Tag peaks from the two biological replicates of CTCF and the two replicates of SMC3 were independently called using SEACR (default parameters). For each target, reproducible peaks were obtained by intersecting the two replicate peak sets and retaining the overlapping regions. These replicate-consistent peak sets were then used for downstream co-occupancy analysis. Genomic regions co-occupied by CTCF and SMC3 were defined by computing the genomic intersection between the reproducible CTCF peak set and the reproducible SMC3 peak set using bedtools intersect (v2.31.0).²⁷ Overlapping peaks were defined as regions where a CTCF peak and an SMC3 peak shared ≥1 bp of genomic overlap. Motif enrichment analysis was performed using HOMER (v5.1) (http://homer.ucsd.edu/homer/download.html). For each CUT&Tag dataset, peaks were supplied to HOMER's findMotifsGenome.pl function.

RNA-Seq Data Analysis

Bulk RNA-seq datasets of mouse lens epithelial and fiber cells were obtained from the study by Zhao et al. (GSE113887).²⁸ The raw sequencing data were first evaluated using FastQC, and low-quality reads and adapter sequences were removed using Trimmomatic (v0.39).²¹ Cleaned reads were then aligned to the mouse reference genome (mm39) using STAR (v2.7.11b). Gene expression levels were quantified with featureCounts and normalized as FPKM (Fragments Per Kilobase of transcript per Million mapped reads). The FPKM values were log²-transformed for normalization, and gene expression distributions were visualized using violin plots and heatmaps.

SR Identification and GO/KEGG Enrichment Analysis

SRs were identified using a modified version of the ROSE algorithm (v0.1)²⁹ based on signal intensity calculation, following the approach described by Cai et al.³⁰ SR regions were annotated to their associated genes through transcription start site (TSS) overlap or chromatin loop information. Gene ontology (GO) enrichment analysis for biological processes was performed using the ClusterProfiler R package³¹ with a significance threshold of P < 0.1. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was conducted using the Metascape online platform (https://metascape.org/).³² The top 20 significantly enriched terms from both GO and KEGG analyses were selected and visualized using R.

Immunofluorescence

Mouse eyeball and lens tissues were fixed overnight at 4°C in 4% paraformaldehyde (Biosharp, BL539A), dehydrated with 30% sucrose (no. S769549; Shanghai Macklin Biochemical Co., Ltd.), and embedded in Tissue-Tek O.C.T. compound (no. 4583; Sakura Finetek USA, Inc., Torrance, CA, USA). Frozen tissues were stored at −80°C and sectioned at 15 µm thickness using a Leica CM1860 cryostat (Leica Biosciences, Wetzlar, Germany). Tissue sections were permeabilized with 0.1% Triton X-100 (no. T434388; Aladdin Scientific) and blocked with 5% BSA (no. A850222; Shanghai Macklin Biochemical Co., Ltd.). Primary antibodies were applied and incubated overnight at 4°C, followed by incubation with fluorescent secondary antibodies at room temperature for one hour. Nuclei were counterstained with DAPI (no. C1005; Beyotime Institute of Biotechnology, Jiangsu, China) for five minutes. Each step was followed by three washes in TBST (no. BL315B; BioSharp, Tallinn, Estonia), each lasting five minutes. The primary antibodies used included anti-H3K27ac (no. 8173S; Cell Signaling Technology), anti-H3K9me3 (no. 13969S; Cell Signaling Technology), and anti-H3K27me3 (no. 9733S; Cell Signaling Technology). Fluorescent secondary antibodies were obtained from Invitrogen. Immunofluorescence images were acquired using a Zeiss LSM 980 confocal microscope (Zeiss, Oberkochen, Germany).

Quantitative Real-Time PCR

High-quality total RNA was extracted from LEC and LFC using the RNeasy Mini Kit (no. 74104; Qiagen) following the manufacturer's protocol. Equal amounts of RNA were mixed with gene-specific primers and HiScript II One-Step qPCR SYBR Green Kit reagents (Q221; Vazyme Biotech Co., Ltd.) for quantitative real-time PCR (qPCR) analysis. Gene expression levels were normalized to Gapdh as an internal control. Primer sequences are listed in Supplementary Table S1.

Allele-Specific Expression Analysis of Blcap in Lens Tissue

To investigate whether Blcap exhibits imprinting in the lens, we employed intersubspecific hybrid mice generated from C57BL/6J (B6) and CAST/EiJ (Cast). An informative T/A SNP located in the last exon of Blcap, previously used to determine allelic origin in brain tissue,³³ was used in this study. Lens tissues were collected from two-week-old B6 × Cast and Cast × B6 hybrid offspring, as well as from parental strains. Lens epithelial and fiber cell masses were carefully dissected, immediately snap-frozen in liquid nitrogen, and stored at −80°C. Total RNA was extracted using the RNeasy Mini Kit (no. 74104; Qiagen) according to the manufacturer's protocol. RNA integrity was assessed with an Agilent 2100 Bioanalyzer, and cDNA libraries were prepared. Paired-end sequencing was performed on the Illumina NovaSeq platform. Sequencing reads were aligned to the mouse reference genome (mm39), and allele-specific expression was quantified using GATK ASEReadCounter to distinguish maternal and paternal allele contributions.

DNA Fluorescence In Situ Hybridization (DNA-FISH)

Frozen sections of mouse eyeballs and lens tissues were prepared as described in the ”Immunofluorescence“ part. Sections were stained using the Universal Two-Color Chromogenic DNA In Situ Hybridization Detection Kit (no. PDT0002; Pinpoease, Guangzhou, China), following the manufacturer's instructions. The probes used included a Blcap gene locus probe (chr2: 157,399,231-157,401,071; no. 536191-B1; Pinpoease) and a Blcap-DSR region probe (chr2: 15,783,475-157,184,414; no. CUS202502191-B2; Pinpoease). Final images were acquired using a ZEISS LSM 980 confocal microscope.

AAV Production

To delete the genomic region chr2:157,413,328–157,419,804 in C57BL/6J mice, we constructed an AAV2-U6-sgRNAup-U6-sgRNAdown-CMV-saCas9 plasmid. This plasmid contains sgRNAup and sgRNAdown expression cassettes driven by the U6 promoter, and a saCas9 expression cassette driven by the CMV promoter, all cloned into a pAAV2 backbone containing the inverted terminal repeats required for AAV2 packaging. AAV2 particles were packaged in HEK293T cells. Seventy-two hours after transfection, cells and supernatants were harvested, subjected to repeated freeze-thaw cycles, and treated with nucleases to remove unpackaged DNA. Viral particles were then purified using AVB affinity chromatography (no. 17372211; Cytiva, Marlborough, MA, USA), resuspended in PBS (no. C10010500BT; ThermoFisher), and sterilized through a 0.22 µm filter (no. GSWP04700; Merck, Kenilworth, NJ, USA). The final viral titer, determined by qPCR, was 0.9 × 10¹³ vg/mL. The sequences of sgRNAup and sgRNAdown are provided in Supplementary Table S1.

Anterior Chamber Injection In Mice

After anesthetizing the C57BL/6J mice, local anesthesia was applied using proparacaine hydrochloride eye drops (Alcon Laboratories, Fort Worth, TX, USA). A 33G microinjection needle (Hamilton Company, Reno, NV, USA) was then inserted into the anterior chamber at the corneal site, and carboxymethyl cellulose was injected at the angle of the anterior chamber to prevent excessive aqueous humor circulation and reduce the clearance of AAV from the anterior chamber. Subsequently, approximately 2 µL of PBS, AAV2-Ctrl, or AAV2-saCas9 was injected into the anterior chamber of 2-week-old C57BL/6J mice under a surgical microscope (OPMI Lumera I; Zeiss). AAV2-Ctrl served as a control vector carrying the same AAV2 backbone and saCas9 system without target-specific sgRNA, to exclude non-specific effects from viral delivery and Cas9 expression. After injection, ofloxacin eye ointment (Alcon Laboratories) was applied to protect the cornea. A second injection was performed two week later. Four weeks after the first injection, mouse lenses were collected for subsequent experiments.

Western Blotting

Lens-derived fiber clusters and LECs were lysed separately in RIPA buffer. The total protein concentration was determined using the Pierce 660 nm Protein Assay Reagent (no. 22660; ThermoFisher). For all Western blotting experiments, equal amounts of total protein were loaded from each sample. Proteins were separated using Tris-Glycine-SDS running buffer (no. LC2675; Invitrogen) by electrophoresis. Gels were transferred to PVDF membranes (no. FFP39; Beyotime Institute of Biotechnology) using NuPAGE transfer buffer (no. NP0006; Invitrogen). Membranes were blocked in 5% non-fat dry milk buffer (no. P0216-300g; Beyotime Institute of Biotechnology) for one hour, and then incubated overnight at 4°C with primary antibodies: anti-BLCAP (Abcepta, #AP54023), anti-EGFP (no. 50430-2-AP; Proteintech, Rosemont, IL, USA), and anti-β actin (no. 20536-1-AP; Proteintech). After washing with TBST (5 minutes × 3), membranes were incubated with secondary antibody (anti-rabbit HRP, no. ab2891; Abcam, Cambridge, MA, USA) diluted in 5% BSA at room temperature for one hour. After three washes with TBST, membranes were developed using a detection reagent kit (no. E411-04 Vazyme Biotech Co., Ltd.). The blot was imaged using a FluorChem E CCD imaging system.

Flow Cytometry

Single-cell suspensions of lens-derived fiber cells were prepared as described in the “Isolation of Single-Cell Suspensions from LECs and LFCs” section. Cells were washed twice with cold cell staining buffer (no. 20201; BioLegend, San Diego, CA, USA) and resuspended in Annexin V binding buffer (no. 422201; BioLegend) at a concentration of 0.25-1.0 × 10⁷ cells/mL. A 100 µL aliquot of the suspension was transferred to a 5 mL tube, followed by the addition of 5 µL APC-conjugated Annexin V and propidium iodide (PI; no. 421301; BioLegend). After gentle mixing, samples were incubated in the dark at room temperature for 15 minutes. Subsequently, 400 µL of Annexin V binding buffer was added. Flow cytometry was performed on a BD FACSCanto II system (BD Bioscience, Franklin Lakes, NJ, USA), with single-color compensation applied to each fluorescence channel. Data were analyzed using FlowJo software (v10), with gating strategies based on FSC/SSC to exclude debris and FSC-H/FSC-W to identify single cells while excluding aggregates.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism (v.8.0), R, and Python. As all data types were non-normally distributed, non-parametric tests were employed for all statistical comparisons. Box plots within the violin plots represent the median, 25th percentile, and 75th percentile of the data. For comparisons involving more than two groups, p-values were adjusted using the Holm correction method. P-values in the figures are denoted as *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

Data Availability

The Hi-C and CUT&Tag data analyzed in this study are available from the corresponding author upon request, under the accession Bioproject number PRJCA038181 at https://ngdc.cncb.ac.cn/gsa/. All other data are available from the corresponding author upon reasonable request.

Code Availability

The code for the analyses in this manuscript has been uploaded to GitHub and can be accessed at https://github.com/ksr157/LFCD.git.

Results

Chromatin Compaction During LFCD

In E16.5 C57BL/6J mice, the lens enters the denucleation phase, during which LFCs begin to degrade subcellular structures, including the nucleus, leading to the formation of an organelle-free zone (OFZ).^19,34 To investigate the chromatin structural changes during the denucleation process, we collected and sequenced samples from both LECs and LFCs at E16.5. Figure 1A illustrates the experimental design, where we constructed Hi-C three-dimensional chromatin maps for both LECs and LFCs and integrated transcriptomic (RNA-seq), histone modification, and CTCF/SMC3 protein interaction data.

Figure 1.

Hi-C reveals the dynamic 3D architecture of chromatin during progressive nuclear condensation in the LFCD. (A) Schematic of workflow. RNA-seq data were downloaded from public databases (GSE113887), while all other data were generated in this study. (B) Left, Immunofluorescence images of H3K27ac distribution in the lenses of E16.5 mice. Right, Representative immunofluorescence images of LEC and LFC. All scale bars represent 50 µm. (C) Left, immunofluorescence images of H3K9me3 distribution in the lenses of E16.5 mice. Right, representative immunofluorescence images of LEC and LFC. Scale bars: 50 µm. (D) Top, PC1 value of Hi-C analysis, indicating chromatin compartmentalization (red/yellow, compartment A; blue/blue, compartment B). Bottom, Hi-C contact matrices for chr1of LEC and LFC. Juicebox parameters were balanced and normalized at 25kb resolution. (E) The chromatin contact probabilities (P(s)) relative to genomic distance for LEC and LFC, respectively. (P(s) curves, P is probability and s is genomic distance). (F) Boxplot of interactions ratio of (inter-chromosomal /intra-chromosomal) per chromosome (dots). Data are presented as median interquartile range (IQR). The P values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001.

Each biological replicate yielded more than 6 billion uniquely mapped chromatin contacts, achieving a resolution of approximately 10–25kb. Consistent with previous studies, immunofluorescence (IF) staining for H3K27ac and H3K9me revealed that both euchromatin and heterochromatin were diffusely distributed in the nuclei of E16.5 mouse LECs. In contrast, chromatin in LFCs exhibited a more condensed and compacted state (Figs. 1B, 1C). A similar compressed chromatin state was observed in the nuclei of cells at the transition zone in 6 weeks mouse lenses (Supplementary Fig. S1A, S1B). Through contact probability analysis, we further observed that long-range chromosomal interactions were more frequent in LFCs, whereas LECs showed a higher enrichment of short-range interactions (Figs. 1D, 1E). Subsequently, by comparing the ratio of inter-chromosomal to intra-chromosomal interactions, we found that the ratio in LFCs was significantly higher than in LECs, suggesting that the compression of chromatin structure during the cell transition may lead to a significant increase in inter-chromosomal interaction frequency (Fig. 1F). In summary, during the denucleation of LFCs, nuclear compaction promotes the enhancement of long-range chromatin interactions.

Weakened Chromatin Compartmentalization and Compartment Switching During LFCD

To further investigate changes in higher-order chromatin organization during LFCD, we analyzed the dynamic features of A/B chromatin compartments. Previous studies have shown that compartment switching is closely associated with changes in cell state during differentiation.^35,36 In both E16.5 mouse LECs and LFCs, the genome was roughly evenly partitioned into A and B compartments (Supplementary Figs. S2A, S2B). Generally, A compartments are associated with open, transcriptionally active chromatin, whereas B compartments correlate with closed, transcriptionally inactive chromatin.³⁷ We first evaluated chromatin interactions among A-A, B-B, and A-B compartments. Although the total numbers of these interaction types showed minimal changes (Supplementary Fig. S2C), the log ratio of observed to expected interactions revealed a global increase in chromatin interactions during the transition from LECs to LFCs (Fig. 2A).

Figure 2.

LFCD A/B compartment switching is accompanied by changes in the expression of associated genes. (A) The log2[observed (obs)/expected (exp)] Hi-C contact intensity values of A–A, B–B and A–B compartment interactions in LEC and LFC. The P values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001. (B) Boxplot of the compartmentalization-strength per chromosome (dots). Data are presented as median interquartile range (IQR). The P values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001. (C) Pie chart of compartment switch types. (D) Heatmap of compartment switching, where each square represents a compartment. The color of each square indicates the vector value used to classify A/B compartments, with values greater than 0 (red) representing compartment A, and values less than 0 (blue) representing compartment B. (E) Violin and boxplots showing the gene expression levels of all genes in A to B (up) and B to A (down) compartment switching regions for each biological replicate sample. The P values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001. (F) Enriched KEGG pathways in genes within A/B compartment switching regions. (G) Heatmap displaying the expression of apoptosis-related genes (GO:0097194) in A-to-B compartment switching regions, based on the average values from three biological replicates. (H) The qPCR bar chart shows the expression levels of apoptosis-related genes (GO:0097194) in A to B compartment switching regions. Data are presented as median interquartile range (IQR). Each group consists of four biological replicates (n = 4), and all experiments were repeated three times. The P values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001. (I) Heatmap displaying the expression of apoptosis-related genes (GO:0097194) in B to A compartment switching regions, based on the average values from three biological replicates. (J) The qPCR bar chart shows the expression levels of apoptosis-related genes (GO:0097194) in B to A compartment switching regions. Data are presented as median interquartile range (IQR). Each group consists of four biological replicates (n = 4), and all experiments were repeated three times. The P values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001. (K) The point-to-point line plot shows the changes in the relative loop counts of genes in B to A compartment switching regions between LEC and LFC conditions. The P values by Wilcoxon signed-rank test, *P < 0.05, **P < 0.01, ***P < 0.001.

Further analysis demonstrated a significant decrease in compartmentalization strength in LFCs compared to LECs (Fig. 2B). Compartmentalization strength, defined as the ratio of intra-compartment to inter-compartment interactions within each chromosome, reflects the degree of spatial segregation between A and B compartments. Its reduction indicates that chromatin underwent substantial remodeling during LFC denucleation, leading to enhanced chromatin interactions and weakened functional isolation between compartments.^38–40

To explore whether this weakening of compartmentalization was accompanied by changes in compartment identity, we analyzed compartment switching during LEC-to-LFC differentiation. We found that a subset of genomic regions underwent compartment transitions from A to B or B to A (Figs. 2C, 2D). By integrating RNA-seq data from E16.5 LECs and LFCs, we observed that A-to-B switching regions were generally accompanied by downregulation of gene expression, whereas B-to-A switching regions exhibited upregulation (Fig. 2E), consistent with previous findings.^36,41

Together, these results suggest that during LFC denucleation, chromatin remodeling not only weakens compartmentalization strength but also reshapes compartment states, thereby contributing to the precise spatiotemporal regulation of gene expression.

Compartment Switching Regulates Apoptosis-Related Gene Expression During LFCD

We first systematically assessed the global gene expression changes during the LEC-to-LFC transition, revealing widespread upregulation and downregulation of genes during differentiation (Supplementary Fig. S2E). Functional enrichment analysis indicated that these differentially expressed genes were primarily involved in key pathways related to apoptosis, stress response, metabolic reprogramming, and cellular structural remodeling (Supplementary Table S2). Building on this, to further investigate the regulatory impact of A/B compartment dynamics on gene expression, we focused on genes located within compartment-switching regions and performed KEGG pathway enrichment analysis. Strikingly, genes involved in both A-to-B and B-to-A switches were significantly enriched in the apoptosis pathway (Fig. 2F). In addition to apoptosis, we also observed enrichment of multiple other pathways. Genes undergoing A-to-B transitions were enriched in stress- and survival-related pathways, including FoxO signaling, MAPK signaling, PI3K-Akt signaling, TNF signaling, autophagy, and mitophagy, as well as HIF-1 signaling and biosynthesis of cofactors, suggesting a global suppression of stress responses, cell survival, and metabolic activities in differentiating LFCs.^42–47 Conversely, genes in B-to-A transitions were enriched in pathways associated with cellular architecture and protein processing, such as phagosome, cell adhesion molecules, cytoskeleton in muscle cells, PPAR signaling, and protein processing in the endoplasmic reticulum, together with metabolic adaptations including glutathione metabolism and cytochrome P450-related xenobiotic metabolism.^48–51 These results indicate that compartment switching not only regulates apoptosis-related gene expression but also broadly coordinates stress response, metabolic reprogramming, and structural remodeling, thereby promoting the survival and maturation of LFCs during denucleation.

We then specifically analyzed the expression changes of apoptosis-related genes (based on GO:0097194) and validated key findings through qPCR experiments. In the A-to-B switching regions, several genes, including Blcap, Cecr2, Casp3, Aifm1, Igfbp3, Akt1, Bax, and Apaf1, exhibited varying degrees of downregulation in LFCs, among which Casp3 is a classical late-stage apoptosis-related gene, and previous studies have suggested that it may not be involved in LFCs differentiation or organelle degradation in mouse^6,11,52 (Figs. 2G, 2H). In contrast, within the B-to-A switching regions, genes such as Xkr4, Xkr8, Pak2, and Gcg were upregulated in LFCs; additionally, although Dffa and Dffb—two genes involved in DNA fragmentation during denucleation^5,11—did not show significant expression changes by qPCR, they were also enriched in the B-to-A compartment switching group (Figs. 2I, 2J). These findings suggest that chromatin structural remodeling during lens development coordinates the activation and repression of apoptosis-related genes, enabling controlled nuclear degradation in LFCs while preventing full apoptotic execution.

To further investigate the regulatory mechanisms linking A/B compartment switching to gene expression changes, we analyzed alterations in chromatin loop formation associated with these apoptosis-related genes. We found that, in the context of globally enhanced chromatin interactions, some apoptosis-related genes undergoing A-to-B compartment switching, particularly Blcap, exhibited a marked increase in local loop counts in LFCs (Fig. 2K). By contrast, no significant changes in looping were observed for genes undergoing B-to-A switching (Supplementary Fig. S2D). These observations imply that the repression of pro-apoptotic genes following A-to-B compartment switching may be reinforced by enhanced formation of local chromatin loops.

Remodeling of Silencing Regions During LFCD

In the 3D nuclear architecture, regulatory elements govern distal gene expression by forming chromatin loops that spatially juxtapose enhancers, promoters, or silencers with their target loci.^19,53 We hypothesized that, in LFCs, silencer-associated chromatin regions might engage in long-range interactions with apoptosis-related genes to reinforce transcriptional repression. To address this, we performed H3K27me3 CUT&Tag profiling in both LECs and LFCs to map silencing landscapes.³⁰

Following the approach used for enhancer and super-enhancer identification, we first defined H3K27me3 peaks and applied the ROSE algorithm to cluster and rank them by signal intensity.^29,30,54 The top-ranked clusters were designated as Super Silencing Regions (Super-SRs), and the remainder as Typical Silencing Regions (Typical-SRs). In LECs, we identified 16,975 silencing regions (SRs), including 274 Super-SRs, whereas in LFCs, we detected 10,009 SRs, including 180 Super-SRs. Although the total number of SRs decreased during differentiation, the relative proportion of Super-SRs remained comparable (LEC vs. LFC = 1.61% vs. 1.80%).

To explore the regulatory potential of these SRs, we integrated CUT&Tag and Hi-C data to categorize SRs based on their spatial relationship to TSSs³⁰: SRs overlapping TSSs were classified as proximal SRs (PSRs), whereas those tethered to TSSs via chromatin loops were defined as distal SRs (DSRs) (Fig. 3A). In LECs, we identified 14,631 PSRs and 4448 DSRs; in LFCs, 9253 PSRs and 995 DSRs (Fig. 3A).

Figure 3.

Silencing regions undergo remodeling during LFCD. (A) Different categories of SR associated with genes. Regions directly overlapping the TSS are defined as PSRs, whereas those anchoring the TSS through chromatin loops are defined as DSRs. (B) H3K27me3 signal strength of PSR and DSR in E16.5 mouse LFC and their associated genes. (C) Violin and box plots displaying the expression levels of genes associated with PSRs and DSRs. The values for each group represent the mean of three biological replicates (n = 3). The P-values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001. (D) The Venn diagram shows the number of shared or distinct PSRs (left) and DSRs (right) between LEC and LFC. (E) Left, immunofluorescence images of H3K27me3 distribution in the lenses of E16.5 mice. Right, representative immunofluorescence images of LEC and LFC. Scale bars: 50 µm.

Interestingly, while PSRs consistently exhibited stronger H3K27me3 enrichment than DSRs (Fig. 3B; Supplementary Fig. S3A), the overall expression levels of DSR-associated genes were still significantly lower than those of PSR-associated genes (Fig. 3C; Supplementary Fig. S3B). In LECs, 631 genes, and in LFCs, 87 genes were found to overlap between DSRs and PSRs (Supplementary Table S3). To avoid confounding effects of these overlapping genes on expression comparisons, they were excluded from the analysis of DSR- and PSR-associated gene expression. These results suggest that distal silencing may exert stronger transcriptional repression at the global level. Furthermore, PSR and DSR repertoires displayed limited overlap between LECs and LFCs (Fig. 3D), indicating a major reorganization of silencing networks during lens differentiation, including the loss of existing regions and formation of new silencer. Immunofluorescence staining of H3K27me3 in E16.5 mouse lenses supported these findings: in LECs, H3K27me3 exhibited a diffuse nuclear distribution, whereas in LFCs, it became concentrated beneath the nuclear envelope, forming a prominent perinuclear ring (Fig. 3E). Notably, this perinuclear condensation is consistent with previous observations that H3K27me3 undergoes spatial relocalization to the nuclear periphery during epigenetic reprogramming in mouse fetal germ cells,⁵⁵ as well as with studies showing that MSX1 recruits H3K27me3 to the nuclear periphery to enforce transcriptional repression during development.⁵⁶ Although the redistribution was less dramatic in adult lenses, LFCs continued to display localized H3K27me3 condensation (Supplementary Fig. S3C). Together, these results reveal that LFCD is accompanied by extensive remodeling and spatial reorganization of silencing domains, which likely contribute to the precise regulation of gene expression necessary for terminal differentiation.

DSRs Repress Apoptosis-Related Genes Via Chromatin Looping

To further investigate whether silencing regions (SRs) regulate gene expression during LFC differentiation through chromatin interactions, we analyzed SR-associated genes in both LFCs and LECs. The results showed that most SR-associated genes were closely linked to cell differentiation and tissue development processes (Fig. 3B; Supplementary Fig. S3A). GO enrichment analysis of genes associated with newly acquired DSRs and PSRs in LFCs further revealed that these genes are not only enriched in apoptotic pathways, including apoptosis, TNF signaling, and DNA damage–induced intrinsic apoptotic pathways, but also significantly associated with biological processes relevant to LFCs maturation and homeostasis (Fig. 4A; Supplementary Fig. S4A). These processes include regulation of protein metabolism and degradation (including proteasome-mediated negative regulation), protein glycosylation and glycoprotein biosynthesis, protein ubiquitination, synaptic vesicle trafficking and endocytosis, as well as negative regulation of ER stress responses.^57–59 KEGG pathway analysis indicated that these genes participate in multiple signaling pathways, such as TGF-β, Notch, Hippo, PI3K–Akt, AMPK, cAMP, Ras, and TNF pathways, as well as autophagy, glutathione metabolism, ferroptosis, cell cycle, endocytosis, and ECM–receptor interactions, all of which are closely related to LFCs differentiation, protein homeostasis, and oxidative stress responses (Fig. 4B; Supplementary Fig. S4B).^{42,44,45,47,60–66} Notably, given the repressive nature of DSR-mediated transcriptional regulation, these elements may maintain the long-term stability of lens high-abundance structural proteins by downregulating genes involved in protein degradation, ubiquitination, and glycosylation, while inhibition of autophagy and glutathione metabolism pathways may contribute to metabolic balance under oxidative stress. Moreover, repressive DSRs may finely regulate cell fate through multiple signaling pathways, highlighting the complex and systemic role of distal silencers in LFCs differentiation and functional maintenance.

Figure 4.

DSR downregulate the expression of apoptotic genes in LFCD through chromatin loops. (A) GO enrichment analysis of genes associated with LFC gained DSR. (B) Enriched KEGG pathways of genes associated with LFC gained DSR. (C, D) Top, Hi-C contact matrices for chr 12: 112670000–112680000 of LEC (C) and LFC (D) at 25 kb resolution. Middle, Akt1 gene and DSR sites along with the corresponding chromatin loops in the regions. Bottom, CTCF and SMC3 binding in the respective regions. Green shadow, LFC gained DSR-loop. (E, F) Top, Hi-C contact matrices for chr 2:156775000-157775000 of LEC (E) and LFC (F) at 25 kb resolution. Middle, Blcap gene and DSR sites along with the corresponding chromatin loops in the regions. Bottom, CTCF and SMC3 binding in the respective regions. Green shadow, LFC gained DSR-loop. (G, H) Violin and box plots displaying the expression levels of Akt1 (G) and Blcap (H). The values for each group represent the mean of three biological replicates (n = 3). The P values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001.

Given our earlier finding that some apoptosis-related genes exhibited increased loop formation after A-to-B compartment switching, we next assessed whether these newly formed loops were associated with DSRs (Fig. 2K). Considering that CTCF and SMC3 are key proteins involved in chromatin loop formation,^67,68 we integrated their CUT&Tag data. Biological replicates showed high reproducibility in both cell types. CTCF and SMC3 CUT&Tag identified slightly different numbers of peaks in LECs and LFCs (CTCF: ∼2.2M vs. ∼2.0M; SMC3: ∼1.8M vs. ∼2.0M), suggesting modest differences in CTCF binding levels and cohesin occupancy between the two cell types (Supplementary Figs. S4C, S4D). Motif analysis revealed highly consistent sequence features across samples, with NFY (CCAAT) as the most significantly enriched motif, followed by recurrent enrichment of regulatory motifs such as Twist2 and PRDM1 (Supplementary Table S4). We further identified two apoptosis-related genes, Akt1 and Blcap, that acquired new DSR-anchored chromatin loops in LFCs (Figs. 4C–F). Meanwhile, RNA-seq and qPCR analyses both demonstrated that the expression levels of Akt1 and Blcap were significantly reduced in LFCs compared to LECs (Figs. 2H, 4G). Notably, Blcap has been reported to be imprinted in mouse brain tissue.³³ To exclude potential confounding effects, we performed allele-specific cDNA sequencing and confirmed that Blcap is not imprinted in either LECs or LFCs (Supplementary Fig. S4E). Given that AKT1 is a key effector of the PI3K signaling pathway, previous studies have reported that inhibition of the PI3K/AKT1 axis induces spatiotemporal activation of autophagy, which promotes the removal of non-nuclear organelles in LFCs and contributes to the formation of the OFZ.^5,69 Therefore we speculate that DSR-anchored chromatin loops that suppress Akt1 expression may serve as a critical mechanism coordinating autophagy-dependent organelle clearance. Notably, BLCAP is a pro-apoptotic factor that regulates the cell cycle and apoptotic signaling pathways in multiple tissues.^70–73 Downregulation of BLCAP can promote cell proliferation by facilitating the G1-to-S phase transition and suppresses apoptosis, whereas overexpression of BLCAP induces S-phase arrest and apoptosis with downregulating anti-apoptotic proteins Bcl-2 and Bcl-xL.^71,72 Additionally, we further analyzed the spatial localization of Akt1, Blcap, and Bcl2 family members in mouse lens sections at E16.5 and E18.5 (Supplementary Fig. S4F). To investigate the spatial relationship between Akt1, Blcap, Bcl-2 family proteins and OFZ formation, we performed immunofluorescence on E16.5 and E18.5 lenses and generated axial gray value intensity profiles in conjunction with DAPI. The results showed that Akt1 and Blcap were expressed at significantly higher levels in the epithelial layer than in differentiating fiber cells, with their expression markedly reduced in the OFZ region; in contrast, Bcl-2, Bcl2l1, and Bcl2l2 remained detectable in fiber cells, including the OFZ region. These findings suggest that downregulation of Akt1 and Blcap together with sustained expression of Bcl-2 family members in the OFZ may jointly contribute to maintaining an anti-apoptotic-like state in lens fiber cells. Collectively, these results suggest that DSRs, by promoting chromatin loop formation, contribute to the suppression of apoptosis-related genes while coordinating the temporal order of organelle clearance, thereby promoting cell survival and maintaining homeostasis during LFCs differentiation.

Blcap-DSR Chromatin Looping Regulates Blcap Expression To Maintain Lens Homeostasis

To validate our hypothesis that the silencer–apoptosis gene axis prevents excessive apoptotic activation during LFC differentiation, we further focused on the DSR-associated chromatin loops of Blcap to dissect the underlying mechanism. We previously found that during the denucleation phase of LFCs, the Blcap locus acquired two de novo chromatin loops anchored to a DSR (Figs. 4E, F). Consistent with this, DNA FISH analysis in E16.5 mouse lenses revealed close spatial proximity between Blcap and its associated DSR in LFCs, whereas in LECs, the physical distance between these loci was markedly increased (Fig. 5A).

Figure 5.

The destruction of the Blcap-DSR loop induces lens opacity. (A) DNA FISH demonstrates the distance between the Blcap gene (green) and DSR (red) sites in E16.5 mouse LEC (left) and LFC (right). All scale bars represent 2 µm. (B) Left, 2 µL of PBS, AAV2-Ctrl, or AAV2-saCas9 was injected into the anterior chamber of two-week-old mice, and a second injection was performed two week later. Four weeks after the first injection, mouse lenses were collected for subsequent experiments. Right, The AAV2-mediated saCas9 dual sgRNA gene editing system was used to knockout the CTCF/SMC3 co-binding region at the right anchor of the Blcap-DSR loop in mouse lenses, disrupting loop formation. Conservation scores were calculated based on the 35-species vertebrate Multiz alignment using PhyloP, as provided by the UCSC Genome Browser. (C) Microscopic images of six-week-old mouse lenses under different treatments. Each group includes five biological replicates (n = 5). Bottom, All lenses with loop anchor knockout via AAV2-KO exhibited opacity. (D) DNA FISH demonstrates the distance between the Blcap gene (green) and DSR (red) sites in LEC (left) and LFC (right) of mouse lenses under vehicle treatment. Scale bars: 2 µm. (D) DNA FISH demonstrates the distance between the Blcap gene (green) and DSR (red) sites in LEC (left) and LFC (right) of mouse lenses under AAV2-KO treatment. Scale bars: 2 µm.

We suggested that the Blcap-DSR loops are critical for the spatiotemporal regulation of Blcap during lens cell differentiation. Given the essential roles of CTCF and SMC3 in chromatin loop formation,^67,68 we drew on previous studies and employed the AAV2-CRISPR-saCas9 system to delete the CTCF/SMC3-enriched region at the right anchor (Chr2: 157,413,328-157,419,804; corresponding to the site in Fig. 4F) in the mouse lens, thereby disrupting the Blcap-DSR loop (Fig. 5B). Specifically, peaks from the CTCF and SMC3 CUT&Tag datasets were identified separately, and genomic intersections between the two peak sets were computed using bedtools intersect. Regions covered by both CTCF and SMC3 peaks were defined as CTCF/SMC3 overlapping peaks, which were used to determine the precise target interval within the right loop anchor. In our experiments, we first validated the transfection efficiency of AAV2-EGFP in lens cells. Western blotting(WB) showed EGFP expression in both lens epithelial cells and fiber cells 4 weeks after two intravitreal injections of AAV2-EGFP (Supplementary Fig. S5A). Using the same methodology, we then injected AAV2-CRISPR-Cas9. Notably, compared to the blank control and AAV-Ctrl groups, disruption of the Blcap-DSR loop led to pronounced nuclear opacity in the lens (Figs. 5C, 5D). DNA FISH confirmed that this disruption increased the spatial distance between Blcap and its DSR anchor in LFCs, abolishing their previous colocalization (Figs. 5E, 5F). We speculated that disruption of the Blcap–DSR chromatin loop induces abnormal apoptosis in LFCs while also perturbing normal differentiation and denucleation processes, further affecting organelle clearance and protein homeostasis, ultimately manifesting as nuclear cataracts. Collectively, these findings highlight a critical role for Blcap-DSR mediated chromatin looping in regulating Blcap expression and maintaining lens homeostasis during fiber cell differentiation.

Disruption of the Blcap-DSR Loop Impairs the Anti-Apoptotic Program in LFCs

We hypothesize that the Blcap–DSR loop acts as a protective regulatory mechanism that restrains excessive apoptotic activation in LFCs. Specifically, the Blcap-DSR loop may inhibit Blcap expression, thereby contributing to the formation of an anti-apoptotic system during the LFCD, which in turn limits the full execution of the apoptotic program. To test this hypothesis, we conducted a detailed analysis of lenses with the loop disruption.

WB results showed that after loop disruption, Blcap expression in LFCs significantly increased (Fig. 6A). Concurrently, qPCR results showed that the expression of classical anti-apoptotic factors Bcl2 and Bcl2l1 in the Bcl2 family was significantly reduced, consistent with prior findings that elevated Blcap expression downregulates these proteins⁷³ (Fig. 6B). Previous studies have also highlighted that members of the Bcl2 family play an important role in the degradation of mitochondria, endoplasmic reticulum, and Golgi apparatus in LFCs.^7,10

Figure 6.

Blcap-DSR loop deletion leads to upregulation of Blcap expression, downregulation of the anti-apoptotic factors Bcl-2 and Bcl2l1, and ultimately induces cell apoptosis. (A) Left, Immunoblotting analysis revealed differential expression of BLCAP in LFC under different treatment; Right, The bar graph shows the relative protein levels of BLCAP. Data are presented as median interquartile range (IQR). Each group consists of five biological replicates (n = 5). The P values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001. (B) The qPCR bar chart shows the expression levels of Bcl2, Bcl2l1, Bcl2l2, and Mcl in LFC under different treatment. Data are presented as median interquartile range (IQR). Each group consists of four biological replicates (n = 5), and all experiments were repeated three times. The P values by Wilcoxon test, *P < 0.05, **P < 0.01, ***P < 0.001. (C) Representative images of Annexin/PI flow cytometry analysis of mouse lenses under different treatments (n = 4 biological replicates per group); Proportions of early (Annexin+PI−) and late (Annexin+PI+) apoptotic cells in mouse lenses under different treatments. Data are presented as median interquartile range (IQR). Adjusted P values were calculated using the Wilcoxon test with Holm correction; *P < 0.05, **P < 0.01, *P < 0.001.

To further confirm the impact of Blcap suppression relief after Blcap-DSR loop disruption on lens cell differentiation, we performed Annexin V/PI dual staining flow cytometry analysis of LFCs. The results indicated that the proportion of apoptotic LFCs was significantly higher in the loop disruption group, both in early and late stages of apoptosis (Figs. 6B, 6C). These findings suggest that loss of the Blcap-DSR loop relieves Blcap suppression, disrupts anti-apoptotic signaling, and compromises the apoptosis-resistant state required for proper differentiation during LFCD. As a result, lens cells aberrantly activate apoptotic pathways instead of maintaining the controlled, non-lethal remodeling process characteristic of normal differentiation.

Discussion

LFCs undergo selective degradation of their nuclei and other organelles at specific stages of differentiation, a process that shares certain molecular and morphological features with apoptosis but proceeds without causing cell death or loss of function.^7–11 This presents a unique and complex biological challenge. Previous studies have proposed the existence of an endogenous protective system in lens cells that restrains excessive apoptotic activation at specific stages of differentiation to ensure fiber cell survival.^11,13,14 However, the molecular mechanisms underlying this system remain unclear. In this study, by integrating Hi-C with multi-omics analyses, we uncovered pronounced chromatin reorganization events in E16.5 LFCs, including (1) global chromatin condensation, enhanced long-range interactions, and epithelial-to-fiber cell AB compartment switching; and (2) the specific formation of SRs spanning apoptosis gene clusters.

In mouse LFCDs, the degradation of nuclei and organelles is initiated almost simultaneously.⁷⁴ The nucleus first undergoes morphological alterations—changing from oval to spherical, with chromatin condensed at the nuclear periphery, small pores emerging in the nuclear envelope, and partial translocation of nuclear proteins into the cytoplasm. In line with this process, CDK1-dependent nuclear envelope breakdown has been shown to permit DNase IIβ entry for chromatin degradation, forming a core mechanism for lens fiber cell denucleation.^3,75 The nucleus then progressively shrinks and fragments into multiple apoptosis-like particles.⁷⁴ By contrast, mitochondria swell, fragment, and lose function more rapidly.³³ At this stage, we identified a novel distant regulatory element, DSR, which inhibits the expression of the apoptosis-related gene Blcap through chromatin interactions, thereby orchestrating spatially and temporally regulated anti-apoptotic responses. Furth functional perturbation experiments demonstrated that disrupting the Blcap–DSR chromatin loop was sufficient to relieve Blcap repression, leading to dysregulation of downstream anti-apoptotic programs. DNA-FISH confirmed the loss of spatial proximity, accompanied by a marked increase in Blcap transcription and protein levels, reduced Bcl2/Bcl2l1 expression, enhanced caspase activation, and the onset of lens opacification. These findings indicate that the chromatin loop itself plays a pivotal role in regulating apoptosis-related signaling during LFC differentiation.

However, additional epigenetic factors and molecular pathways in LFCs are also likely to cooperate in controlling the non-lethal program of organelle and nuclear degradation. For instance, αB-crystallin has been shown to exert anti-apoptotic functions in LECs.¹³ It has been proposed that the abundance of αB-crystallin in LFCs attenuates apoptotic cascades, guiding terminal differentiation toward organelle clearance rather than cell death.¹⁴ Consistently, proteasome-inhibition studies in LECs have demonstrated that upregulation of heat-shock proteins, including αB-crystallin and other small HSPs, suppresses caspase activation and protects lens cells from apoptosis, further supporting the existence of endogenous anti-apoptotic mechanisms in lens tissues.⁷⁶ In addition, prior studies have shown that lens fiber cells engage DNA damage–response and repair programs before denucleation.^77–79 Disruption of key DDR components, such as Nbs1, compromises chromatin processing, interferes with fiber-cell denucleation, and ultimately induces cataract, highlighting the essential role of genome-maintenance mechanisms in supporting safe, non-lethal nuclear clearance during differentiation.⁷⁷ Therefore the regulatory role of the Blcap–DSR loop should be regarded as a central, yet not exclusive, control node.

Moreover, it is noteworthy that the DSR region is enriched with H3K27me3 signals, suggesting that it may act through the classical Polycomb-mediated mechanism, serving as an initiation site for H3K27me3 spreading to establish a stable repressive environment.³⁰ However, our findings do not provide definitive evidence that DSR function is strictly dependent on Polycomb dynamics, nor can we exclude the possibility that it represents a distinct silencing mechanism. Future studies examining the binding of key PRC2 components (such as EZH2 and SUZ12) will be essential to further clarify the regulatory nature of DSR.

Our study further shows that genes associated with chromatin remodeling are not only enriched in apoptotic pathways but also in autophagy, mitophagy, protein metabolism regulation, endoplasmic reticulum stress, endocytosis/vesicle trafficking, as well as multiple signaling pathways including TGF-β, Notch, Hippo, PI3K–Akt, and AMPK, all of which are closely related to organelle clearance.^{45,60–62,69} This suggests that Blcap suppression may temporally coincide with the removal of mitochondria and other organelles, potentially occurring synchronously to provide a safe window for the cell to complete denucleation and organelle clearance without triggering full apoptotic execution. Collectively, these coordinated chromatin structural changes likely form a multilayered regulatory network that ensures terminal differentiation and organelle degradation in LFCs while maintaining cell viability.

Furthermore, we speculate that chromatin loop mechanisms similar to the Blcap–DSR interaction may not be limited to LFCs. During terminal differentiation, for instance, erythrocytes also exhibit active caspase activation, and chromatin remodeling has been shown to regulate gene expression.^80,81 However, whether chromatin loops selectively suppress key pro-apoptotic factors in these cells remains unexplored. If this mechanism proves to be conserved, it would reveal a general epigenetic strategy by which 3D genome remodeling enables non-lethal apoptotic remodeling during differentiation across tissues.

Nonetheless, this study has certain limitations. First, the specific protein complexes and dynamic assembly mechanisms underlying this chromatin loop remain unclear and warrant investigation through proteomics. Second, the lack of spatially resolved single-cell transcriptomic data limits our ability to precisely map LFC heterogeneity and the effects of loop regulation. Third, the conservation of this regulatory mechanism in the human lens remains to be verified. Future studies integrating single-cell multi-omics, spatial omics, or live imaging will facilitate the construction of high-resolution 3D genome and transcriptional dynamic maps, further validating the role of the Blcap–DSR loop in terminal differentiation and anti-apoptotic regulatory networks.

In conclusion, we systematically charted the dynamic changes during LFCD using multi-omics data and identifies an epigenetic mechanism that safeguards differentiation without cell death. Chromatin remodeling and DSR not only directly repress pro-apoptotic genes but also coordinately regulate differentiation-related genes and protein homeostasis, ensuring LFCs survival and functional maintenance during denucleation. The identification of the Blcap–DSR loop introduces a new concept of a “differentiation-coupled restriction of apoptosis” mechanism, providing new insight into how terminally differentiated cells achieve controlled nuclear and organelle remodeling while maintaining long-term survival, and offering a framework for investigating analogous regulatory strategies in other terminal differentiation systems.