The inner nuclear membrane protein LEMD3 organizes the 3D chromatin architecture to maintain vascular smooth muscle cell identity

Abstract

Maintaining the contractile phenotype of vascular smooth muscle cells (VSMCs) is critical for vascular homeostasis. However, the role of the 3D chromatin architecture in regulating VSMC identity remains elusive. A genome-scale CRISPR screen identifies LEMD3 as a potential regulator to maintain VSMC identity. Lemd3 deficiency in VSMCs results in the loss of the contractile phenotype and exacerbates intimal hyperplasia in mice. Protein interactome analysis reveals that LEMD3 interacts with CBX3, a principal reader of H3K9me2/3, subsequently anchoring heterochromatin at the nuclear periphery. Employing the DNA polymer model based on Hi-C data, whole-chromosome simulations demonstrate that Lemd3 depletion disturbs the chromatin structure. Multi-omics analysis further reveals that Lemd3 depletion alters the genome conformation as the increase of inter-TAD (topologically associated domain) interactions at the boundaries of A and B compartments, which correlates with decreased chromatin accessibility and repressed expression of VSMC contractile genes. This study reveals that LEMD3 organizes the 3D chromatin architecture by anchoring heterochromatin at the nuclear periphery to maintain the VSMC contractile identity.

Introduction

Vascular smooth muscle cells (VSMCs) exhibit a quiescent, contractile phenotype, as they highly express contractile markers, such as MYH11 (also known as smooth muscle myosin heavy chain 11), calponin, transgelin (also known as SM22α), and α-smooth muscle actin, to maintain vascular homeostasis under physiological conditions¹, while contractile VSMCs transiently dedifferentiate to acquire the migratory and proliferative capabilities for vasculogenesis during the physiological processes of vascular development and regeneration². Moreover, in response to abnormal cues in the vascular microenvironment, such as alterations in mechanical stretching, excessive production of growth factors, extracellular matrix remodeling, and reactive oxygen species (ROS) accumulation^{3,4,5,6,7,8,9}, VSMCs suffer from the loss of a contractile phenotype and subsequently transform into a series of pathological phenotypes related to various vascular diseases, including atherosclerosis, vascular calcification, intimal hyperplasia, and aortic aneurysm^10,11,12,13. Thus, maintaining the contractile identity of VSMCs is beneficial for preventing VSMC phenotypic switching and related vascular diseases. Although some critical transcription factors or repressor-driven gene profiling transitions, as well as epigenetic modulations, including noncoding RNA, DNA methylation and histone acetylation, have been reported to orchestrate VSMC phenotypic switching^{14,15,16,17,18}, the mechanism by which VSMCs maintain their contractile phenotype remains poorly understood. Current knowledge is mainly limited to the fact that serum response factor (SRF) and its coactivator myocardin exert their transcriptional activities as well as epigenetic regulation on histone variant H2A.Z and H3K4me2 modification to facilitate VSMC contractile identity^19,20,21,22.

Compelling evidence has suggested that cell identity programs result from the interaction between transcription factors and the chromatin landscape that they encounter. Chromatin adopts a complex three-dimensional (3D) conformation within the nucleus. Recent high-throughput chromosome conformation capture (Hi-C) analyses of comprehensive interaction maps over large regions or whole genomes have indicated that the genome is hierarchically organized into chromosome territories, A/B compartments, topologically associated domains (TADs), and chromatin loops²³. The intermingling of chromosome territories, A-B compartment switching and dynamic changes in TADs are involved in the alteration of the 3D chromatin architecture, which further affects chromatin accessibility, thereby modulating transcription factor-dependent gene expression²⁴. Notably, TADs are fundamental 3D genome units that engage in dynamic higher-order inter-TAD interactions²⁵. However, investigations of inter-TAD interactions are relatively limited, and the modulation of inter-TAD interactions remains poorly understood.

The 3D chromatin structure transitions are commonly recognized during development and cell differentiation^23,26 and are often dysregulated in disease processes^27,28,29,30. In the cardiovascular system, dynamic changes in 3D chromatin structure have been reported to be involved in cardiomyocyte maturation, cardiac remodeling, and dilated cardiomyopathy^27,31,32. However, to date, the role of the 3D chromatin architecture in the vascular system is still elusive. Here, we identified LEMD3, an inner nuclear membrane protein, as a critical regulator that maintains the VSMC contractile phenotype using a genome-scale CRISPR knockout screen. Further protein interactome screening revealed that LEMD3 interacted with CBX3, a principal reader of the heterochromatin-associated histone modification H3K9me2/3, subsequently anchoring heterochromatin at the nuclear periphery, whereas Lemd3 depletion caused the repositioning of heterochromatin from the nuclear periphery toward the interior validated via immunofluorescence staining. Through Hi-C technology and whole-chromosome modeling, we further revealed that Lemd3 depletion disturbed the 3D chromatin architecture, namely, enhanced inter-TAD interactions at the boundaries of A and B compartments, which correlated with the downregulation of contractile marker genes, as well as the subsequent loss of the contractile phenotype. Thus, LEMD3 organized the 3D chromatin architecture to maintain the contractile identity of VSMCs.

Results

A genome-wide CRISPR knockout screen identifies LEMD3 essential for maintaining the vascular smooth muscle cell contractile phenotype

To identify genes critical for the maintenance of the VSMC contractile phenotype, we designed a genome-scale CRISPR knockout screen in the mouse vascular smooth muscle cell line MOVAS. First, we engineered a MOVAS cell line expressing an endogenous fluorescent reporter using CRISPR/Cas9 knock-in technology to insert a P2A-EGFP (enhanced green fluorescent protein) cassette before the stop codon of the endogenous contractile phenotype gene ACTA2 to yield the coexpression of the ACTA2 protein and EGFP (Fig. 1a). Next, we validated whether EGFP expression was regulated consistently with the previously well-recognized altered pattern of ACTA2 expression during the VSMC phenotypic transition in the knock-in cell line. Flow cytometry analysis revealed that the fluorescence intensity of EGFP was increased in VSMCs with the transforming growth factor-β (TGF-β)-or serum starvation-primed contractile phenotype but was significantly decreased following platelet-derived growth factor BB (PDGF-BB)-induced VSMC synthetic phenotypic switching (Fig. 1b), suggesting that EGFP efficiently served as a sensitive reporter for the VSMC phenotypic transition in the knock-in MOVAS cell line.

Fig. 1: A genome-wide CRISPR knockout screen identifies genes essential for maintaining the contractile phenotype of VSMCs.

a Schematic diagram showing the generation of an EGFP reporter-expressing MOVAS cell line. P2A, a self-cleaving peptide. b Representative flow cytometry and quantification of EGFP expression in the reporter MOVAS cell line in response to 48 h of TGF-β treatment (blue line), serum starvation (red line), or PDGF-BB treatment (green line). WT, wild-type; TGF-β, transforming growth factor-β; PDGF-BB, platelet-derived growth factor BB. n = 3 independent experiments. c Schematic diagram of the genome-scale CRISPR knockout screen. d Flow cytometry analysis of EGFP reporter knock-in MOVAS cells that were transduced with or without the pooled lentivirus encoding a genome-wide sgRNA library with two rounds of cell sorting. Wild-type (WT) MOVAS cells were utilized as a negative control for EGFP fluorescence. e Identification of the top candidate genes using the MAGeCK algorithm. Pink circles represent the top candidate genes that have been previously reported to regulate the VSMC phenotype. MAGeCK, Model-based Analysis of Genome-wide CRISPR/Cas9 Knockout. RRA score, robust rank aggregation score. f RT‒qPCR analysis of Acta2, Cnn1, Tagln, and Myh11 gene expression in the A7r5 rat VSMC cell line transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 24 h. The data are presented as the relative fold changes to siRNA_scramble (n = 5 independent experiments). Error bars indicate s.e.m. Statistical analysis was performed using ordinary one-way ANOVA with Tukey’s multiple comparisons test for panel b and two-sided unpaired Student’s t test for (f).

We further generated stable Cas9 expression in the EGFP reporter cell line via lentivirus transduction to apply CRISPR technology (Supplementary Fig. 1a). Then, the EGFP reporter MOVAS cells with stable Cas9 expression were targeted with a mouse genome-scale CRISPR knockout (GeCKO) v2 library containing 6 sgRNAs per gene (total sgRNA size of 130,209)³³, followed by two successive rounds of flow cytometry sorting. Since we aimed to target the genes whose knockout caused loss of the contractile phenotype, we preferentially sorted the cells with reduced EGFP fluorescence intensity (EGFP^low population) (Fig. 1c). Consequently, the EGFP^low population with 18.2% proportion was enriched following two rounds of sorting (Fig. 1d and Supplementary Fig. 1b). Next-generation sequencing of the population revealed sgRNA enrichment for 109 genes (MAGeCK³⁴ sgRNA enrichment score <10⁻³) (Fig. 1e, and Supplementary Data 1). Among the top 8 genes, Acta2 served as the positive control, while Nf1 and Nf2 have been previously reported to maintain VSMC contractile identity^35,36, indicating the reliability of the genome-wide CRISPR knockout screen (Fig. 1e). We subsequently focused on the remaining 5 genes with unknown functions in VSMCs. Using specific small interfering RNAs (siRNAs), we validated that knockdown of these 5 genes, including Cand1, Ajuba, Gdi1, Xpo6 and Lemd3, resulted in significant downregulation of VSMC contractile markers, including Acta2, Cnn1, and Tagln, suggesting the loss of the contractile phenotype (Fig. 1f and Supplementary Fig. 2). Interestingly, previous evidence has shown that Lemd3 deficiency in mice results in abnormal vascular development and embryonic lethality³⁷, suggesting a potential role for LEMD3 in the vascular system, but without further investigation thus far. Therefore, we preferentially explored the role of Lemd3 in VSMC identity and vascular homeostasis.

LEMD3 maintains the contractile phenotype of vascular smooth muscle cells in vitro

We firstly clarified the nuclear localization of LEMD3 in VSMCs. As shown by the immunocytochemical staining on rat VSMCs, LEMD3 was substantially expressed and mainly localized in the nuclei of VSMCs (Fig. 2a). Meanwhile, we validated it in human arteries by immunohistochemical staining, as evidenced by the obvious staining of LEMD3 in the cellular nuclei of arterial media mainly containing VSMCs (Fig. 2b). Following the transfection of the EGFP-LEMD3 plasmid into the A7r5 VSMCs line, we observed that LEMD3 fused with EGFP was localized mainly to the nuclear membrane in VSMCs (Fig. 2c), consistent with the role of LEMD3 as an inner nuclear membrane protein³⁸. Next, we explored the expression alterations of LEMD3 during VSMC phenotypic switching. As results, serum starvation and rapamycin priming VSMCs into contractile phenotype significantly upregulated LEMD3 expression (Fig. 2d, e and Supplementary Fig. 3a, b), whereas PDGF-BB causing VSMC dedifferentiation markedly downregulated LEMD3 expression (Fig. 2f and Supplementary Fig. 3c). Moreover, we validated the regulation of LEMD3 expression in vivo. As a consequence, LEMD3 expression was reduced at 3 days after wire injury in mouse carotid arteries compared to sham arteries (Fig. 2g). These results suggested that LEMD3 expression positively correlated with the contractile phenotype of VSMCs.

Fig. 2: LEMD3 maintains the contractile phenotype of VSMCs.

a Representative images of immunocytochemical staining for LEMD3 in primary rat VSMCs. Scale bar = 100 µm. n = 3 independent experiments. b Representative images of immunohistochemical staining for LEMD3 in human internal mammary arteries. The region between the dashed lines corresponds to the medial area. Scale bar = 20 µm. n = 3 independent samples. c Representative images of immunofluorescence staining of rat VSMCs transfected with the pEGFP-N1-LEMD3 plasmid or control plasmid. The nuclei were stained with DAPI (blue). Scale bar = 5 µm. This experiment was repeated three times with similar results. d Representative Western blotting and quantification of LEMD3, ACTA2, CNN1, TAGLN in the A7r5 rat VSMC cell line under serum starvation (48 h). GAPDH was used as an internal control. The data are presented as the relative fold changes to control (Ctrl) group (n = 6 independent experiments). e Representative Western blotting and quantification of LEMD3, ACTA2, CNN1, TAGLN in the A7r5 rat VSMC cell line under rapamycin treatment (100 nM, 48 h). GAPDH was used as an internal control. The data are presented as the relative fold changes to vehicle group (n = 6 independent experiments). f Representative Western blotting and quantification of LEMD3, ACTA2, CNN1, TAGLN in the A7r5 rat VSMC cell line under PDGF-BB treatment (10 ng/ml, 48 h). GAPDH was used as an internal control. The data are presented as the relative fold changes to vehicle group (n = 6 independent experiments). g Representative western blotting and quantification of LEMD3 in the lysates of sham-operated and wire-injured mouse carotid arteries at 3 days after surgery. GAPDH was used as an internal control. Data were presented as relative fold change to Sham (n = 4 independent mice). h Representative Western blotting and quantification of LEMD3, ACTA2, CNN1, and TAGLN in the A7r5 rat VSMC cell line transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 12 h, followed by serum starvation treatment (48 h). GAPDH was used as an internal control. The data are presented as the relative fold changes to siRNA_scramble (n = 4 independent experiments). i Representative Western blotting and quantification of LEMD3, ACTA2, CNN1, and TAGLN in the A7r5 rat VSMC cell line transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 12 h, followed by rapamycin treatment (100 nM, 48 h). GAPDH was used as an internal control. The data are presented as the relative fold changes to siRNA_scramble (n = 4 independent experiments). j Representative images of phalloidin staining and quantification of F-actin (red) in primary rat VSMCs transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 48 h. DAPI staining (blue) indicates the nucleus. Scale bar = 25 µm. n = 3 independent experiments. k Collagen gel contraction assays using primary rat VSMCs transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 48 h. n = 3 independent experiments. l EdU incorporation assays of primary rat VSMCs transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 48 h. n = 6 independent experiments. Error bars indicate s.e.m. Statistical analysis was performed using two-sided unpaired Student’s t test for (d-g), two-way ANOVA followed by Bonferroni’s multiple comparisons test for (h) and (i), the χ² test for (j) and the Mann‒Whitney U test for (k) and (l).

Next, we investigated whether LEMD3 is required for maintenance of the contractile phenotype of VSMCs. Lemd3 knockdown suppressed serum starvation- or rapamycin-induced upregulation of contractile gene expression at both mRNA and protein levels in rat and human VSMCs (Fig. 2h-i and Supplementary Fig. 4). Moreover, F-actin staining revealed that Lemd3 silencing resulted in VSMC morphological changes from an elongated contractile pattern to a polygonal synthetic pattern (Fig. 2j), while collagen gel contraction assays also revealed that Lemd3 knockdown reduced the contractility of VSMCs (Fig. 2k). These results indicated that Lemd3 silencing led to the loss of the contractile phenotype in VSMCs. VSMC proliferation and migration have been reported to be enhanced following phenotypic switching³⁹. Accordingly, we evaluated the proliferation and migration of Lemd3-silenced VSMCs. EdU (5-ethynyl-2’-deoxyuridine) incorporation assays suggested that Lemd3 silencing promoted VSMC proliferation (Fig. 2l). However, Lemd3 knockdown did not affect VSMC migration, as evidenced by the results of the scratch wound assay and transwell migration assay (Supplementary Fig. 5a-b). Conversely, Lemd3 overexpression reversed PDGF-BB-induced loss of VSMC contractile phenotype (Supplementary Fig. 5c-d). Together, these results collectively supported that LEMD3 maintains the contractile identity of VSMCs.

Lemd3 deficiency causes the loss of the vascular smooth muscle cell contractile phenotype in vivo

To further validate the role of LEMD3 in VSMCs in vivo, we constructed tamoxifen-induced smooth muscle cell-specific Lemd3 deletion mice (Lemd3^SMKO mice) by intercrossing Myh11-CreER^T2 mice with Lemd3^flox/flox mice in which exon 2 was flanked with loxP sites (Fig. 3a and Supplementary Fig. 6b). Tamoxifen was administered to 8-week-old Myh11-CreER^T2 Lemd3^flox/flox mice to achieve smooth muscle cell-confined depletion of Lemd3 compared with Lemd3^WT control mice (littermate Myh11-CreER^T2 mice), which received a comparable tamoxifen treatment (Supplementary Fig. 6a). Lemd3^SMKO mice presented no significant differences in body weight or systolic blood pressure (101.8 ± 4.1 mmHg vs. 103.2 ± 3.9 mmHg) compared with Lemd3^WT control mice (Supplementary Table 1), whereas ultrasonography of mouse aortas revealed a slight expansion in segments of the ascending aorta and aortic arch in Lemd3^SMKO mice (Supplementary Fig. 6c-d). As measured by western blotting and immunohistochemistry, Lemd3 deficiency significantly downregulated the expression of ACTA2, CNN1 and TAGLN in mouse aortas, suggesting a loss of the contractile phenotype (Fig. 3b-c). Accordingly, aortic rings and mesenteric resistance arteries isolated from Lemd3^SMKO mice exhibited reduced contraction in response to phenylephrine compared with those isolated from Lemd3^WT mice (Fig. 3d and Supplementary Fig. 6e). Since the in vivo studies only included male mice due to the insertion of Myh11-CreER^T2 transgene into Y chromosome, primary aortic smooth muscle cells were alternatively isolated from Lemd3^flox/flox female and male mice to clarify the potential sex-specific difference in the role of LEMD3 in VSMCs. Adenovirus encoding Cre recombinase was applied in the isolated mouse primary VSMCs to induce Lemd3 knockout. As results, Lemd3 deficiency comparably downregulated the contractile gene expression in both female and male background VSMCs (Supplementary Fig. 7a-d), excluding the potential sex-specific difference.

Fig. 3: Lemd3 depletion results in the loss of the contractile phenotype in VSMCs in vivo.

a Schematic view of the generation of tamoxifen-inducible smooth muscle cell-specific Lemd3 knockout mice. LoxP sites are indicated by red triangles. b Representative western blot and quantification of protein expression in the aortas of 12-week-old Lemd3^WT mice and Lemd3^SMKO mice. GAPDH was used as an internal control. The data are presented as the relative fold changes compared with Lemd3^WT mice. n = 6 mice per group. c Representative images and quantification of immunohistochemical staining for ACTA2 and CNN1 in the thoracic aortas from 12-week-old Lemd3^WT mice and Lemd3^SMKO mice. Scale bar = 100 µm. n = 4 mice per group. d Contraction of aortic rings isolated from 12-week-old Lemd3^WT mice and Lemd3^SMKO mice in response to phenylephrine (Phe). n = 5 mice per group. e Left, Hematoxylin and eosin (H&E) staining of the carotid arteries of sham-operated and wire-injured Lemd3^WT mice and Lemd3^SMKO mice on Day 28 after surgery. The inner circles represent the neointimal areas, and the outer circles represent the medial and neointimal areas. Scale bar = 100 µm. Right, quantitative analysis of the intima areas, the media areas, intima to media ratio, and external elastic lamina (EEL) circumference in H&E-stained sections of the carotid arteries of wire-injured Lemd3^WT mice and Lemd3^SMKO mice at 28 days after surgery. n = 6 mice per group. f EdU staining of the carotid arteries of Lemd3^WT mice and Lemd3^SMKO mice at 28 days after wire injury. The white arrows mark the EdU-positive cells in the neointima and medial area. Mouse vascular elastic lamina (EL) exhibits green autofluorescence. Scale bar = 100 µm. n = 6 mice per group. HE hematoxylin and eosin, EEL external elastic lamina. Error bars indicate s.e.m. Statistical analysis was performed using two-sided unpaired Student’s t test with Welch’s correction for (b-d) and the Mann‒Whitney U test for (e) and (f).

Next, we explored the protective role of LEMD3 in VSMCs under pathological conditions. We generated a carotid wire injury model in 12-week-old male tamoxifen-induced Lemd3^SMKO mice and Lemd3^WT mice to induce VSMC phenotypic switching and neointima formation. At 28 days postinjury, the mouse carotid arteries were harvested (Supplementary Fig. 7e). H&E staining of wire-injured carotid arteries showed that Lemd3 deficiency markedly exacerbated intimal hyperplasia, as evidenced by the elevations in the neointima areas and the neointima/media ratios in Lemd3^SMKO carotid arteries compared with those in Lemd3^WT controls (Fig. 3e). In addition, EdU staining revealed the greater cell proliferation within vascular wall in carotid arteries from Lemd3^SMKO mice than those from control mice at 28 days after wire injury (Fig. 3f). Since the major cell type in arterial media and neointima is VSMCs as well as the increased cell proliferation is also a hallmark of VSMC dedifferentiation⁴⁰, these in vivo data collectively suggested that Lemd3 depletion exacerbates cell proliferation and intimal hyperplasia after vascular injury, possibly due to the loss of the VSMC contractile phenotype.

LEMD3 interacts with CBX3

Previous studies have reported that LEMD3 interacts with SMAD2/3 and represses TGF-β/SMAD2/3 signaling^41,42, while compelling evidence indicates that TGF-β/SMAD2/3 signaling maintains the contractile phenotype of VSMCs⁴³. To further verify whether TGF-β/SMAD2/3 signaling was involved in LEMD3-regulated VSMC phenotypic switching, we explored the alteration of TGF-β/SMAD2/3 signaling upon Lemd3 knockdown in rat VSMCs. As results, TGF-β treatment significantly induced SMAD3 phosphorylation (Supplementary Fig. 8a, b), whereas Lemd3 knockdown further enhanced TGF-β-induced SMAD3 phosphorylation (Supplementary Fig. 8a, b). Of interest, although the silencing of Lemd3 amplified TGF-β/SMAD3 signaling, Lemd3 knockdown conversely abolished TGF-β-upregulated contractile gene expression in VSMCs (Supplementary Fig. 8c, d). Furthermore, we applied SIS3, a selective inhibitor of SMAD3, and found that SIS3 did not influence on the suppression of Lemd3 knockdown on TGF-β-induced VSMC contractile phenotype (Supplementary Fig. 8c, d), suggesting that the effect of LEMD3 on VSMC phenotypic switching was independent on TGF-β/SMAD signaling.

To explore the underlying mechanism by which LEMD3 maintains the VSMC contractile phenotype, we performed an interactome analysis to screen potential LEMD3-binding proteins. HEK293T cells were transfected with a FLAG-tagged LEMD3 plasmid, and the LEMD3-binding proteins were enriched via immunoprecipitation with an anti-FLAG antibody followed by liquid chromatography‒tandem mass spectrometry (LC‒MS) identification (Fig. 4a). As a result, 67 proteins were identified as potential LEMD3-binding proteins (Supplementary Data 2). Among them, SMAD2/3 and CBX3 (chromobox protein homolog 3) have been reported to maintain the VSMC contractile phenotype⁴⁴. Since we excluded the involvement of TGFβ/SMAD signaling in the role of LEMDs in VSMCs, we next investigated the role of the potential interaction between LEMD3 and CBX3 in maintaining VSMC contractile identity.

Fig. 4: LEMD3 interacts with CBX3.

a Schematic diagram of the LEMD3 interactome. b Coimmunoprecipitation (Co-IP) assays of HEK293T cells cotransfected with LEMD3-FLAG and HA-CBX3 plasmids for 48 h. The lysates were immunoprecipitated with a control IgG antibody, anti-FLAG antibody, or anti-CBX3 antibody, followed by immunoblotting with the indicated antibodies. c Co-IP assays of endogenous LEMD3 and CBX3 in rat VSMCs. The lysates were immunoprecipitated with a control IgG antibody, anti-LEMD3 antibody, or anti-CBX3 antibody, followed by immunoblotting with the indicated antibodies. d Schematic illustration of LEMD3 domain depletion mutations used to evaluate the interaction with CBX3. The presence or absence of the deletion mutant binding to CBX3 was defined as + or −, respectively. e HEK293T cells were cotransfected with the CBX3 plasmid and the full-length LEMD3-FLAG, ΔC-terminal-FLAG, or ΔRRM-FLAG plasmid for 48 h. The cell lysates were immunoprecipitated with an anti-CBX3 antibody, and the precipitates were analyzed by immunoblotting with an anti-FLAG antibody. f Coimmunoprecipitation (Co-IP) assays of HEK293T cells cotransfected with RRM-FLAG and HA-CBX3 plasmids for 48 h. The lysates were immunoprecipitated with a control IgG antibody, or anti-FLAG antibody, followed by immunoblotting with the indicated antibodies. g, h Representative western blotting and quantification of VSMC contractile marker expression in the A7r5 rat VSMC cell line transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 12 h, followed by transfection with the pcDNA3.1, LEMD3-FLAG, or ΔRRM-FLAG plasmid for 48 h. GAPDH was used as an internal control. The data are presented as the relative fold changes compared with that of the pcDNA3.1+vehicle group (n = 5 independent experiments). i, j Representative western blotting and quantification of VSMC contractile marker expression in the A7r5 rat VSMC cell line transfected with the pcDNA3.1 or LEMD3-FLAG plasmid for 12 h, followed by transfection with the scrambled siRNA (20 nM) or siRNA targeting Cbx3 (20 nM) for 48 h. GAPDH was used as an internal control. The data are presented as the relative fold changes compared with that of the pcDNA3.1+siRNA_scramble group (n = 3 independent experiments). Data shown in (b, c, e and f) are from one representative of three independent experiments with similar results. Error bars indicate s.e.m. Statistical analysis was performed using two-way ANOVA followed by Bonferroni’s multiple comparisons test for (h) and (j).

We first validated that LEMD3 interacted with CBX3 in HEK293T cells overexpressing CBX3 and FLAG-tagged LEMD3 via coimmunoprecipitation assays (Fig. 4b). Furthermore, the endogenous interaction between LEMD3 and CBX3 was also verified in rat VSMCs and mouse aortas (Fig. 4c and Supplementary Fig. 9a). Moreover, PDGF-BB treatment efficiently downregulated LEMD3 expression, thereby reducing the interaction between LEMD3 and CBX3 in VSMCs (Supplementary Fig. 9b), associated with the loss of contractile phenotype caused by PDGF-BB stimulation. Then, we further explored the specific binding motif of LEMD3 that interacts with CBX3 by subcloning serial deletion mutants of LEMD3, including a C-terminal deletion mutation (ΔC-terminal, aa 1–876) and an RNA recognition motif (RRM) deletion mutation (ΔRRM, aa 1–784), into the pcDNA3.1-FLAG plasmid. HEK293T cells were transfected with plasmids encoding CBX3 and full-length LEMD3-FLAG, ΔC-terminal-FLAG, or ΔRRM-FLAG, followed by coimmunoprecipitation analysis with an anti-CBX3 antibody. Full-length LEMD3 and the ΔC-terminal mutant, but not the ΔRRM mutant, efficiently immunoprecipitated with CBX3, suggesting that LEMD3 interacts with CBX3 via its RRM domain (Fig. 4d-e). Further NanoBit-based binding assays consistently confirmed that the RRM domain was the binding motif of LEMD3 that interacted with CBX3 (Supplementary Fig. 9c). We subcloned RRM domain into the pcDNA3.1-FLAG plasmid and transfected HEK293T cells with plasmids encoding CBX3 and RRM-FLAG. Coimmunoprecipitation analysis revealed that RRM domain interacted with CBX3 (Fig. 4f).

To further confirm whether the interaction with CBX3 mediated the effect of LEMD3 to maintain the VSMC contractile phenotype, we transfected rat VSMCs with the pcDNA3.1-FLAG vector, FLAG-tagged LEMD3, or FLAG-tagged ΔRRM mutant plasmid followed by vehicle or PDGF-BB treatment. Unlike full-length LEMD3, overexpression of the ΔRRM mutant that is unable to bind to CBX3 did not reverse the PDGF-BB-induced loss of the VSMC contractile phenotype at either the mRNA or protein level (Supplementary Fig. 9d and Fig. 4g, h). To specifically target endogenous LEMD3-CBX3 interaction, we transfected rat VSMCs with plasmid encoding RRM-FLAG. RRM domain overexpression blocked the endogenous LEMD3-CBX3 interaction and decreased the contractile protein levels in VSMCs (Supplementary Fig. 9e, f). Furthermore, we silenced Cbx3 or Lemd3 in VSMCs, with concomitant overexpression of Lemd3 or Cbx3, respectively. As results, Cbx3 silencing efficiently reversed Lemd3 overexpression-enhanced SMC contractile protein expression (Fig. 4i, j), while knockdown of Lemd3 conversely inhibited Cbx3 overexpression-upregulated contractile protein expression in VSMCs as well (Supplementary Fig. 9g), indicating the interdependent relationship between LEMD3 and CBX3 in the maintenance of VSMC contractile phenotype. These results collectively suggested that the interaction of LEMD3 with CBX3 plays a critical role in maintaining the VSMC contractile phenotype.

LEMD3 anchors heterochromatin at the nuclear periphery and maintains the 3D chromatin architecture

CBX3 belongs to the heterochromatin protein 1 (HP1) family, which consists of three members: HP1α (encoded by Cbx5), HP1β (encoded by Cbx1) and HP1γ (encoded by Cbx3)⁴⁵. As the principal reader of the heterochromatin-associated histone modification H3K9me2/3, the HP1 family is involved in heterochromatin organization and subsequently influences the 3D chromatin structure^46,47,48. Considering the interaction between LEMD3 and CBX3, we explored whether LEMD3 affects heterochromatin organization. By performing immunofluorescence staining for the repressive histone modification H3K9me3, as a potential indicator of heterochromatin, we observed that the fluorescence intensity along the nuclear diameter exhibited a significant shift in the radial distribution of H3K9me3 from the nuclear periphery toward the interior upon Lemd3 knockdown in rat VSMCs (Fig. 5a, b). However, Lemd3 knockdown in rat VSMCs did not affect the total amount of H3K9me3, as evidenced by western blot analysis (Fig. 5c). To further explore the impact of Lemd3 knockdown on the global landscape of H3K9me3, we performed ChIP sequencing of rat VSMCs transfected with scrambled siRNA or siRNA targeting Lemd3. H3K9me3 signaling exhibited no statistically changes through either genome-wide profiling or locus-specific analysis upon Lemd3 knockdown (Fig. 5d and Supplementary Fig. 10). Meanwhile, we further verified the similar alteration in nuclear distribution of H3K9me2 in VSMCs upon Lemd3 knockdown through immunofluorescence staining (Supplementary Fig. 11a, b). Overall, altered nuclear distribution of H3K9me2/3 suggested that Lemd3 silencing reduced heterochromatin perinuclear anchoring.

Fig. 5: LEMD3 mediates heterochromatin perinuclear anchoring and maintains the global chromatin organization.

a Left, representative images of immunofluorescence staining for H3K9me3 (green) and Lamin B1 (red) in rat VSMCs transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 48 h. DAPI staining (blue) indicates the nucleus. Scale bar = 10 µm. n = 3 independent experiments. Right, quantification of immunofluorescence staining intensities for H3K9me3. Upper, the intensity ratio of nuclear interior to periphery; Lower, total amount of nuclear intensities. Six cells were randomly selected in each experiment. n = 18 cells. b Fluorescence intensity along the white line in a single cell was measured using ImageJ software for the H3K9me3 (green) channel. Scale bar = 5 µm. This experiment was repeated three times with similar results. c Representative western blotting and quantification of H3K9me3 levels in rat VSMCs transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 48 h. Histone H3 was used as an internal control. The data are presented as the relative fold changes to siRNA_scramble (n = 3 independent experiments). d ChIP-seq data (n = 2) showing the genomic occupancy of H3K9me3 in the A7r5 rat VSMC cell line transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 48 h. e, f Hi-C analysis (n = 2) of the A7r5 rat VSMC line transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) for 48 h. e The 3D whole-nucleus maximum entropy model (line rendering) and plane model of the nucleus (bond rendering) of WT and Lemd3 KD VSMCs. Purple represents the A compartment, and green represents the B compartment. f Curve diagram showing the ratio of A/B compartments from the center to the periphery of the nuclei. Blue represents WT, and red represents Lemd3 KD. g Left, representative electron microscopy images of aortic smooth muscle cells from 12-week-old Lemd3^WT mice and Lemd3^SMKO mice. The yellow arrows mark the perinuclear heterochromatin. E elastic lamina. Scale bar = 2 μm. Right, Quantitative analysis of the width of perinuclear heterochromatin of aortic smooth muscle cells from 12-week-old Lemd3^WT mice and Lemd3^SMKO mice. Four vascular smooth muscle cells were randomly selected from each mouse aortic section. n = 6 mice per group. Error bars indicate s.e.m. Statistical analysis was performed using the Mann‒Whitney U test for (a) and (g), and two-sided unpaired Student’s t test for (c).

To further explore whether the reduction in heterochromatin perinuclear anchoring caused by Lemd3 knockdown altered the global chromatin architecture, we performed a high-throughput chromosome conformation capture (Hi-C) assay. Based on principal component analysis (PCA) of Hi-C data, the genome can be divided into A and B compartments⁴⁹, which primarily overlap euchromatin and heterochromatin, respectively⁵⁰. We used parameters based on the maximum entropy principle to construct 3D polymer models at 100-kb resolution using Hi-C data⁵¹. The parameters of the polymer model are encoded in the energy function, and their values are determined iteratively. The parameters are adjusted to match the experimentally determined contact frequencies, ensuring compatibility with experimental data while minimizing biases. Our polymer models positioned compartments in a B-A-B order from the center to the periphery of the nuclei (Fig. 5e). This result was supported by imaging experiments and 3D genome structure calculations at single-cell resolution, which revealed that the chromosomes packed together to form an outer B compartment ring, an inner A compartment ring, and an internal B compartment region around the hollow nucleoli in all cells, based on the previous definition standards^24,52. From certain views of the model, we observed that chromatin appeared to be compressed toward the core of the nucleus upon Lemd3 knockdown (Fig. 5e). As expected, H3K9me3 was predominantly found in B compartments (Supplementary Fig. 11c), which are reported to harbor inactive chromatin^53,54. In the polymer models of Lemd3 KD VSMCs, the perinuclear B compartment was skewed toward the nuclear interior compared with that of WT VSMCs (Fig. 5f), which was consistent with the alterations of nuclear distribution of the heterochromatin mark H3K9me2/3 upon Lemd3 knockdown. Furthermore, we analyzed the chromatin organization of the aortic smooth muscle cells in Lemd3^WT mice and Lemd3^SMKO mice via electron microscopy. The electron-dense heterochromatin was distributed along the nuclear envelope (NE) in the aortic smooth muscle cells of Lemd3^WT control mice (Fig. 5g). However, perinuclear heterochromatin was decreased in the aortic smooth muscle cells of the Lemd3^SMKO mice (Fig. 5g). Taken together, these findings indicate that Lemd3 depletion results in the repositioning of heterochromatin and disturbances in the 3D chromatin architecture.

To investigate whether LEMD3 anchored H3K9me3-modified heterochromatin at the nuclear periphery dependently on the interaction with CBX3, we firstly confirmed that Cbx3 silencing phenocopied the effect of Lemd3 knockdown to cause the redistribution of H3K9me3-modified heterochromatin from the nuclear periphery toward the interior in VSMCs (Supplementary Fig. 11d, e). Next, we transfected rat VSMCs with plasmid encoding RRM-FLAG to block the endogenous LEMD3-CBX3 interaction (Supplementary Fig. 9e), and also found that H3K9me3-modified heterochromatin redistributed from the nuclear periphery toward the interior upon RRM domain overexpression (Supplementary Fig. 11f, g), suggesting that the interaction of LEMD3 with CBX3 might mediate the anchoring of H3K9me3-modified heterochromatin at the nuclear periphery.

Lemd3 depletion increases inter-TAD interactions at the boundaries of A and B compartments and decreases contractile-related gene expression

We further explored the alterations in the 3D chromatin architecture at the scales of the compartments and topologically associated domains (TADs) upon Lemd3 depletion. At the compartment level, the vast majority of the genome remained unchanged in A/B compartment organization (Supplementary Fig. 12a). Recent studies revealed that long-range 3D aggregation of B compartment interactions facilitates the exclusion of gene-rich regions from the B compartment during cell development⁵⁵, whereas interactions between A and B compartments are increased in tumors⁵¹. Interestingly, we discovered that long-range interactions of B-B compartments were decreased and that short-range interactions between nearby A-B compartments were more frequent in certain chromatin regions upon Lemd3 knockdown (Fig. 6a and Supplementary Fig. 12b). Furthermore, in regions with a strong compartment pattern, knockdown of Lemd3 resulted in a relative increase in inter-compartment interactions compared to intra-compartment interactions (Supplementary Fig. 12c), suggesting that the change in inter-compartment interactions may be concentrated within a smaller specific range of the genome. Therefore, we deepened the resolution of the analysis and identified TADs of the Hi-C normalized matrix at 40-kb resolution. We identified 259 WT-specific TAD boundaries, 184 KD-specific TAD boundaries and 4071 common TAD boundaries. The TAD boundaries exhibited no significant difference between the Lemd3 KD and control groups or between the biological replicates (Supplementary Fig. 12d). These findings suggested that the knockdown of Lemd3 does not affect the overall TAD structure in VSMCs. We then classified genome-wide interactions into intra-TAD interactions, interactions between adjacent TADs and interactions between non-adjacent TADs. Interestingly, we found that Lemd3 knockdown resulted in obvious changes in inter-TAD interactions (Fig. 6b). Overall, the contacts of non-adjacent TAD pairs were much weaker than those of adjacent TAD pairs (Fig. 6c). Consequently, our attention was directed toward interactions between adjacent TADs, particularly those located at the boundaries of A and B compartments. We found that 1845 of the 4323 adjacent TAD pairs exhibited altered interactions upon Lemd3 knockdown (Fig. 6b). Both the upregulation and downregulation of the interactions between neighboring TADs within A or B compartments were observed in Lemd3-silenced VSMCs, whereas adjacent TAD interactions at the boundaries of A and B compartments were exclusively increased by Lemd3 knockdown (Fig. 6d-e). Examples of the contact map of a 3-Mb region on chromosome 6 and 11 showed that neighboring TAD-TAD interactions at the A-B boundaries were markedly increased (Fig. 6f, g).

Fig. 6: The Hi-C analysis indicates that Lemd3 depletion changes inter-TAD interactions.

a Heatmap showing the alterations in long-range compartment interactions in the region of 50–100 Mb on chromosome 8 upon Lemd3 depletion. The red box indicates increased interactions between A-B compartments, and the black box indicates decreased interactions between B-B compartments. The bin size is 100 kb. b Bar plots showing the percent change in TAD-TAD interactions. c Box plot showing the average contacts of adjacent and non-adjacent TAD pairs. Box plots are defined in terms of the minimum values and maximum values (whiskers), median (center line) and 25th and 75th percentiles (bounds of box). Adjacent: n = 4323 TAD pairs; Non-adjacent: n = 259868 TAD pairs. d Contour plots showing the distribution of adjacent TAD pairs with decreased or increased interactions upon Lemd3 depletion. e Histogram showing the proportion of adjacent TAD pairs with increased or decreased interactions upon Lemd3 depletion. Heatmap showing the alterations in adjacent TAD-TAD interactions at the A-B boundary in the region of 21‒24 Mb on chromosome 6 (f) and the region of 74‒77 Mb on chromosome 11 (g) upon Lemd3 depletion. The dotted boxes denote inter-TAD interactions at the A-B boundary. The bin size is 40 kb. Statistical analysis was performed using the Mann‒Whitney U test for the data in (c).

Recent studies revealed close connections between TAD boundaries, chromatin accessibility and gene transcription^56,57,58,59. Thus, we additionally performed RNA-seq and ATAC-seq to assess alterations in gene expression and chromatin accessibility upon Lemd3 knockdown (Fig. 7a). RNA-seq revealed that Lemd3 silencing upregulated 1169 genes and downregulated 1431 genes (Supplementary Fig. 13a, Supplementary Data 3), whereas ATAC-seq revealed that chromatin accessibility was predominantly increased at intergenic regions and decreased at promoter regions, with 29,202 increased peaks and 1467 decreased peaks upon Lemd3 knockdown (Supplementary Fig. 13b, c). Interestingly, ATAC-seq signals were negatively correlated with inter-TAD interactions at the boundaries of the A and B compartments (Fig. 7b). Gene set enrichment analysis (GSEA) and heatmap of the downregulated genes identified via RNA-seq revealed that Lemd3 depletion inhibited smooth muscle contraction and differentiation pathways (Supplementary Fig. 13d-e, Supplementary Data 4). Pathway enrichment analysis of the genes assigned to the downregulated peaks via ATAC-seq also revealed that Lemd3 depletion inhibited smooth muscle contraction and differentiation pathways (Supplementary Fig. 13f and Supplementary Data 5). Conversely, the genes assigned to the upregulated ATAC-seq peaks were functionally associated with positive regulation of smooth muscle cell proliferation (Supplementary Fig. 13g and Supplementary Data 6). These findings are consistent with the data shown in Fig. 2, suggesting that Lemd3 silencing downregulates VSMC contractile gene expression and promotes VSMC proliferation. The integrated analysis of ATAC-seq and RNA-seq data identified 143 genes, including the contractile-related genes Acta2, Cnn1, Tagln, Myh11, Col4a1, etc., with decreased chromatin accessibility and downregulated mRNA expression (Fig. 7c, Supplementary Fig. 14 and Supplementary Data 7), suggesting that LEMD3 maintained the chromatin accessibility of these genes and facilitated their transcription. Further bioinformatics analysis of the potential transcription factor-binding motifs in the DNA sequences of ATAC-seq peaks mapped to these 143 genes suggested that the transcription factors MEF2A/C/D, SRF, and TEAD3/4 which have been reported to orchestrate VSMC contractile identity^60,61,62,63, might be involved in LEMD3-facilitated contractile marker gene expression (Fig. 7d and Supplementary Data 8). In addition, we found that among these 143 genes, 62 genes were located in the TADs at A-B boundaries which were most significantly enriched in the muscle cell differentiation pathway and smooth muscle contraction pathway (Fig. 7e and Supplementary Data 9). Together, these data suggested that Lemd3 depletion substantially increased inter-TAD interactions at the boundaries of A and B compartments and decreased the chromatin accessibility of possible MEF2/SRF/TEAD-binding DNA motifs, thereby downregulating VSMC contractile-related gene expression.

Fig. 7: Lemd3 depletion increases inter-TAD interactions at the boundaries of A and B compartments and downregulates smooth muscle contractile gene expression.

a Schematic representation of topologically associating domains (TADs). RNA-seq (n = 3) was used to measure the transcriptional output. ATAC-seq (n = 3) was used to identify chromatin accessibility. b Scatter plot showing the correlation between the fold change of ATAC-seq signals and the fold change of Hi-C contact frequencies at the boundaries of the A and B compartments upon Lemd3 depletion. PCC Pearson’s correlation coefficients. c Venn diagram showing the overlap between the genes whose expression was downregulated and the genes around the closed ATAC peaks. d Transcription factor binding motif analysis of the DNA sequences of closed ATAC peaks corresponding to the genes downregulated upon Lemd3 depletion. e Bar plots showing the enriched GOBP pathways of the genes located in the TADs at the A-B boundaries with both decreased chromatin accessibility and downregulated mRNA expression. Hi-C contact maps (n = 2), histone ChIP-seq tracts (n = 2), ATAC-seq tracts (n = 3) and RNA-seq tracts (n = 3) surrounding the gene loci of Tagln (f) and Acta2 (g). The black lines delineate TAD boundaries. The bin size is 40 kb. h Scatter plot showing the statistical results of LEMD3-modulated human homologous genes. Three coordinate axes represent the distance to the nearest A-B boundary of human gene locus, the gene sequence identity compared to homologs in the rat genome, and the P value of SNPs associated with CAD from the statistical results of GWAS catalog, respectively. Dark red dots represented genes which were located at the A-B boundaries and possessed SNPs associated with CAD. Light red dots represented genes which were located at the A-B boundaries but did not possess SNPs associated with CAD. Black dots represent genes which possessed SNPs associated with CAD but were not located at the A-B boundaries. Gray dots represent genes which were not located at the A-B boundaries and did not possess SNPs associated with CAD. SNPs, single nucleotide polymorphisms. CAD, coronary artery disease. i Hi-C contact maps, ChIP-seq tracts for histone modification and regional association plot surrounding the gene locus of COL4A1 in human genome. Hi-C data were from human coronary artery smooth muscle cells, and histone modification data were from human smooth muscle cells derived from the H9 cell line. The SNPs were calculated by van der Harst P, et al. The black lines delineate TAD boundaries and the bin size is 40 kb. The black arrows indicated inter-TAD interactions at the A-B compartment boundary near the COL4A1 gene locus. CAD, coronary artery disease.

To further confirm the relationship of inter-TAD interactions at the A-B boundaries with the modulation of VSMC contractile gene expression, we generated Hi-C contact maps, H3K9me3 ChIP-seq tracts, H3K27ac ChIP-seq tracts (mapping active promoter and enhancer regions), ATAC-seq tracts and RNA-seq tracts surrounding several representative gene loci. The TADs in which the contractile-related gene loci, including Tagln, Acta2, Col4a1, Myh11 and Myocd were located, had more contacts with the neighboring TADs at the A-B boundaries upon Lemd3 knockdown (Fig. 7f, g and Supplementary Fig. 15a-c). Increased interactions between the active and inactive chromatin domains were correlated with decreased chromatin accessibility and/or gene expression (Fig. 7f, g and Supplementary Fig. 15a-c). Moreover, we performed nascent RNA capture and detected the transcription-dependent RNA synthesis of Tagln and Acta2 upon Lemd3 depletion. As results, knockdown of Lemd3 caused the reduced RNA synthesis of Tagln and Acta2 in VSMCs, supporting that these contractile gene expression levels were directly regulated by chromatin accessibility-dependent transcription (Supplementary Fig. 15d). Since MYOCD functions as a master transcriptional regulator in maintaining VSMC contractile identity, we accordingly verified whether LEMD3 maintained VSMC contractile phenotype in a MYOCD-dependent manner. Of interest, MYOCD overexpression slightly but not completely reversed the downregulation of contractile gene expression upon Lemd3 knockdown in VSMCs (Supplementary Fig. 16), suggesting MYOCD-independent mechanism, such as the direct regulation of chromatin architecture and accessibility, substantially contributed to LEMD3-modulated VSMC identity.

Next, we explored whether the modulation pattern of inter-TAD interactions at the A-B boundaries on gene expression also potentially existed in human VSMCs or related to vascular pathologies. In accordance, we re-analyzed Hi-C data of human coronary artery smooth muscle cells from Zhao Q et al.⁶⁴. We mainly focused on 143 genes whose expressions were maintained by LEMD3 in rat genome (Fig. 7c), and found 139 homologous genes in human genome (Supplementary Fig. 17a and Supplementary Data 7). Based on human Hi-C data, 40 of these homologous genes were located in the TADs at A and B compartment boundaries (Fig. 7h), even exhibiting obvious inter-TAD interactions (black arrows, Fig. 7i), in human VSMCs as well, as illustrated by the contractile-related genes COL4A1 and MYH11 (Fig. 7i and Supplementary Fig. 17b), suggesting that these human genes might be regulated by inter-TAD interactions at A-B boundaries similarly as their homologous genes in rat genome. Furthermore, according to previously reported genome-wide association studies (GWAS) of coronary artery disease (CAD)^65,66,67, we identified that 5 genes (Fig. 7h), including COL4A1 and MYH11, located in the TADs at A and B compartment boundaries in human genome and possessed single nucleotide polymorphisms (SNPs) associated with CAD (Fig. 7i and Supplementary Fig. 17b). Taken together, Lemd3 depletion increases inter-TAD interactions at the boundaries of A and B compartments and downregulates VSMC contractile-related gene expression, which might correlate with human vascular diseases, such as CAD.

Discussion

The current understanding of the regulatory mechanisms of the VSMC phenotypic transition limitedly focuses on the transcriptional regulation of different VSMC identity-specific transcription factors and epigenetic regulatory mechanisms, including noncoding RNAs, DNA methylation, and histone modifications^16,68,69,70. In addition to these mechanistic processes, whether other critical factors and biological processes are also involved in VSMC identity modulation has long fascinated vascular biologists. In the present study, we utilized a genome-wide CRISPR knockout screen to identify candidates potentially maintaining the VSMCs contractile phenotype. Furthermore, we confirmed that the inner nuclear membrane protein LEMD3 is a significant regulator that maintains the VSMC contractile phenotype. LEMD3 interacts with CBX3, the reader of the H3K9me3 histone modification in heterochromatin, and thus anchors heterochromatin at the nuclear periphery to organize the 3D chromatin architecture and facilitate contractile-related gene expression in VSMCs. Lemd3 deficiency in VSMCs results in subtle alterations in chromatin organization at the TAD level and decreases chromatin accessibility and the expression of VSMC contractile-related genes, some of which have substantial GWAS trait-associated loci for coronary artery disease, thereby resulting in the loss of the contractile phenotype and exacerbating postinjury neointima formation in mice (Fig. 8). Our study revealed a novel mechanism for the modulation of VSMC identity in which organizing the 3D chromatin architecture maintains the VSMC contractile phenotype and vascular homeostasis.

Fig. 8: Schematic illustration of the role of LEMD3 in organizing the 3D chromatin architecture to maintain the contractile phenotype of vascular smooth muscle cells (VSMCs).

Inner nuclear membrane protein LEMD3 binds to CBX3, a principal reader of H3K9me3, subsequently anchoring heterochromatin at the nuclear periphery. Lemd3 depletion causes the repositioning of heterochromatin from the nuclear periphery toward the interior in VSMCs. Furthermore, Lemd3 depletion alters the genome conformation as the increase of inter-TAD interactions at the boundaries of A and B compartments, which correlates with the repression of  VSMC contractile gene expression. Overall, LEMD3 organizes the 3D chromatin architecture by anchoring heterochromatin at the nuclear periphery to maintain the VSMC contractile identity.

Although the 3D chromatin architecture plays a critical role in determining cell fate in development and cancer research, the role of the 3D chromatin architecture in the cardiovascular system has not been fully explored. In general, A/B compartment switching and TAD reorganization substantially contribute to the modulation of the 3D chromatin architecture. A/B compartment switching is often observed in the differentiation and reprogramming of embryonic/pluripotent stem cells, whereas intra-TAD interactions and loop formation mediated by CTCF and cohesion generally indicate enhancer‒promoter communication, which facilitates gene expression^59,60. Recently, several studies have suggested that 9.8% of the genome undergoes A/B compartment switching during cardiomyocyte differentiation³¹, and chromatin loop-mediated enhancer‒promoter interactions are altered in dilated cardiomyopathy²⁷. To the best of our knowledge, the current study is the first to reveal the role of the 3D chromatin architecture in the regulation of VSMC identity and vascular homeostasis. Intriguingly, we globally observed that the inter-TAD interactions were changed more obviously, compared with A/B compartment switching or alterations in intra-TAD interactions, upon Lemd3 knockdown using Hi-C analysis in the present study. Although inter-TAD interactions are infrequently observed compared with intra-TAD interactions^71,72, a few studies have revealed the nonnegligible effects of inter-TAD interactions on regulating chromatin accessibility and gene expression⁵⁵. For example, the nuclear matrix protein HNRNPU maintains the 3D genome architecture globally in mouse hepatocytes, whereas the depletion of HNRNPU leads to decreased TAD boundary strengths and increased inter-TAD interactions⁷³. In addition, loss of lamins causes altered interactions among TADs but not the overall TAD structure in mouse embryonic stem cells, and gene transcription changes are correlated with alterations in interactions between active and inactive chromatin domains⁷⁴. Our study revealed that Lemd3 depletion increased inter-TAD interactions at the boundaries of A and B compartments and reinforced the importance of inter-TAD interactions in the regulation of contractile gene expression and VSMC identity. SRF functions as the master transcription factor to control VSMC identity. Of interest, we found that the chromatin region in which Srf gene locus located exhibited reduced inter-TAD interaction within A compartment upon Lemd3 knockdown in VSMCs (Supplementary Fig. 18a). Since inter-TAD interactions within A compartment are generally considered to facilitate chromatin accessibility and gene expression^75,76, we accordingly observed the downregulated accessibility and expression of Srf gene upon Lemd3 silencing in ATAC-Seq and RNA-Seq (Supplementary Fig. 18a, b). Thus, the inter-TAD interactions within A compartment might also be involved in LEMD3-modulated chromatin architecture and VSMC contractile phenotype.

Our study revealed a pivotal role for LEMD3 in regulating VSMC identity. To date, approximately 35 integral membrane proteins have been shown to localize at the inner nuclear membrane of mammalian cells, but none of these proteins have been explored in VSMCs. Through an unbiased genome-wide CRISPR knockout screening strategy, we identified LEMD3 as a potential regulator of the maintenance of the VSMC contractile phenotype. LEMD3 is the longest member of the LEM domain protein family characterized by the presence of a bihelical motif called the LEM domain. Regarding its known modulatory effects, LEMD3 interacts with the nuclear lamina (lamin A/C) at the nuclear envelope–lamina–chromatin interface and functionally recruits chromatin-modifying proteins such as HDAC3 and transcriptional regulators such as BAF to suppress gene transcription⁷⁷. In addition, LEMD3 has been reported to interact with R-SMADs through its carboxyl terminal RRM domain and subsequently antagonize both TGF-β-SAMD2/3 and bone morphogenic protein (BMP)-SMAD1/5/8 signaling⁷⁸. Consistently, our interactome screen also detected the interaction of LEMD3 with R-SMADs, including TGF-β-responsive (SMAD2/3) and BMP-responsive (SMAD1/5) SMADs. Since TGF-β signaling facilitates smooth muscle contractile gene expression in VSMCs⁴³, the LEMD3-SMAD2/3 interaction, which suppresses TGF-β signaling, cannot explain the function of LEMD3 in maintaining the VSMC contractile phenotype well. Meanwhile, we also found the inhibition of TGF-β/SMAD signaling did not affect VSMC alterations upon Lemd3 silencing, excluding the potential involvement of LEMD3-SMAD2/3 interaction in regulating VSMC phenotype switching. Interestingly, our ATAC-seq data revealed that Lemd3 depletion upregulated BMP signaling and the bone development pathway (Supplementary Fig. 13g), which could be explained well by the loss of LEMD3 and the SMAD1/5 interaction. This finding was consistent with clinical case reports that Lemd3 loss-of-function mutations lead to osteopoikilosis, Buschke–Ollendorff syndrome and melorheostosis in humans⁷⁹. However, the role of LEMD3 in maintaining the contractile phenotype of VSMCs is not well understood in the context of the upregulation of BMP signaling upon Lemd3 depletion. Accordingly, we explored the critical role of the LEMD3-CBX3 interaction in the maintenance of VSMC contractile identity. Previous studies have reported that CBX3 maintains the VSMC contractile phenotype and inhibits postinjury neointima formation by increasing SRF transcriptional activity and repressing Notch3 expression, respectively^44,80. However, the exact mechanism and direct role of CBX3 in regulating VSMCs transcription factor activity and gene expression are poorly understood. CBX3 is the principal reader of H3K9me2/3, one of the major repressive histone modifications, which frequently exist in the inactive domain of chromatin, such as heterochromatin⁸¹. The binders of these repressive histone modifications are generally located at the nuclear periphery through interactions with nuclear envelope proteins, thereby determining the perinuclear anchoring of heterochromatin. For example, HP1α, which belongs to the same protein family as CBX3, recognizes H3K9me3 in heterochromatin, whereas the inner nuclear membrane protein lamin B receptor (LBR) interacts with H3K9me3-bound HP1α and functions as an anchoring site of heterochromatin for perinuclear distribution^82,83. In the present study, we found that the inner nuclear membrane protein LEMD3 also functions as an anchoring site and interacts with H3K9me3-bound CBX3 to anchor heterochromatin at the nuclear periphery. The interaction between CBX3 and LEMD3 organized the 3D genome in VSMCs, which might be not only involved in the maintenance of the VSMC contractile phenotype but also essential for embryonic vascular development, since both CBX3 knockout and LEMD3 knockout cause abnormal vascular development^37,44, possibly due to the dysfunction of VSMC differentiation, although further investigation is needed. Of interest, RRM domain, the binding domain of LEMD3 with CBX3, has also been reported to mediate the interaction of other proteins (e.g., R-SMAD, transcription regulators GCL and BTF) with LEMD3^77,84. Considering overexpression of RRM domain may also influence on the binding of other proteins to LEMD3 in VSMCs, we could not completely exclude that LEMD3 anchors heterochromatin at the nuclear periphery and maintains VSMC contractile phenotype also dependently on other interacting proteins, although we demonstrated that the effect of LEMD3 on VSMC phenotype modulation was independent on TGF-β/SMAD signaling (Supplementary Fig. 8). To further clarify the specific role of LEMD3-CBX3 interaction in heterochromatin anchoring as well as the maintenance of VSMC contractile phenotype, the specific amino acid residues mediating LEMD3-CBX3 interaction need to be identified, through unbiased RRM-interacting protein profiling and subsequent site-directed mutagenesis assays of RRM domain to compare the differences among the interactions with all RRM-interacting proteins, which requires further investigation in future studies.

In addition, we did not explore the role of VSMC-specific LEMD3 in female mice, because Myh11-CreER^T2 transgene is inserted into the Y chromosome. Alternatively, we excluded the potential sex-specific difference using primary aortic smooth muscle cells isolated from male and female Lemd3^flox/flox mice, as evidenced by the loss of contractile phenotype in both male and female background VSMCs upon Cre-overexpressed adenovirus-mediated Lemd3 knockout, whereas additional in vivo evidence might require further investigation on other mouse line including female VSMC-specific Lemd3 knockout mice. Of interest, SMC-targeting Cre lines driven by the Myh11 promoter also display high activity in visceral SMCs⁸⁵. Thus, loss of Lemd3 in intestinal SMCs might impair intestinal contraction and consequently cause intestinal obstructions. However, we did not observe the significant phenotype of intestinal obstructions in Lemd3^SMKO mice till 16-week age under basal or physiological conditions. Further investigation might be required to explore the role of LEMD3 in visceral SMCs, especially under some intestinal pathological conditions.

Taken together, our study revealed that the inner nuclear membrane protein LEMD3 organizes the 3D chromatin architecture by anchoring heterochromatin at the nuclear periphery through its interaction with CBX3, a principal reader of the heterochromatin-associated histone modification H3K9me3. Lemd3 depletion caused the repositioning of heterochromatin from the nuclear periphery toward the interior, as well as the enhancement of inter-TAD interactions at the boundaries of A and B compartments in chromatin, which correlated with the downregulation of contractile-related genes and the subsequent loss of the contractile phenotype. Since LEMD3-modulated genes involve some GWAS trait-associated loci for coronary artery disease, LEMD3 dysfunction might correlate with the pathogenesis of coronary artery disease. Thus, LEMD3 functions as a vital regulator to maintain the VSMC contractile identity and prevent vascular disease by organizing the 3D chromatin architecture.

Methods

Experimental materials

An antibody against LEMD3 (ab121854) for western blot analysis (dilution 1:1000) and immunohistochemical staining (dilution 1:100) was purchased from Abcam (Cambridge, UK). An antibody against LEMD3 (orb107113) for co-immunoprecipitation (Co-IP) assays was purchased from Biorbyt (Cambridge, UK). An antibody against Lamin B1 (66095-1-Ig, dilution 1:200) for immunofluorescence staining was purchased from Proteintech Group, Inc. (Wuhan, China). Antibodies against ACTA2 (ab5694), CNN1 (ab46794), and TAGLN (ab14106) for western blotting (dilution 1:1000) and immunohistochemical staining (dilution 1:200) were purchased from Abcam (Cambridge, UK). Antibodies against SRF (16821-1-AP, dilution 1:1000) and GAPDH (60004-1-Ig, dilution 1:5000) for western blotting were purchased from Proteintech Group, Inc. (Wuhan, China). An antibody against CBX3 (ab217999) for western blotting (dilution 1:1000) and Co-IP was purchased from Abcam (Cambridge, UK). The anti-FLAG-M2 (F3165) antibody used for the Co-IP assay was purchased from Sigma‒Aldrich (St. Louis, MO, USA). An antibody against H3K9me3 (ab176916) for immunofluorescence staining (dilution 1:2000), western blotting (dilution 1:1000), and ChIP-Seq was purchased from Abcam (Cambridge, UK). An antibody against H3K9me2 (ab176882, dilution 1:500) for immunofluorescence staining was purchased from Abcam (Cambridge, UK). An antibody against histone H3 (ab1791, dilution 1:1000) for western blotting was purchased from Abcam (Cambridge, UK). An antibody against H3K27ac (ab4729) for ChIP-Seq was purchased from Abcam (Cambridge, UK). IRDye-conjugated secondary antibodies (dilution 1:10000) for western blot analysis were purchased from Rockland, Inc. (Gilbertsville, USA). Normal mouse IgG (sc-2025) used as a negative control for the co-IP assay and immunostaining was obtained from Santa Cruz Biotechnology, Inc. (Dallas, USA). Normal rabbit IgG (2729S) used for the co-IP and immunostaining assays was obtained from Cell Signaling Technology (Boston, USA). YF633-phalloidin (BN10053) was purchased from Biorigin, Inc. (Beijing, China). 4,6-Diamidino-2-phenylindole (DAPI) (62248), Alexa Fluor 488-conjugated goat anti-rabbit IgG (A-11008, dilution 1:1000), and Alexa Fluor 555-conjugated goat anti-mouse IgG (A-21422, dilution 1:1000) for immunofluorescence staining were purchased from Thermo Fisher Scientific (Rochester, USA). PDGF-BB (100-14B-50) and TGF-β (100-621C-2) were purchased from PeproTech, Inc. (Rocky Hill, USA). Rapamycin (HY-10219) and selective SMAD3 inhibitor SIS3 (HY-13013) were purchased from MCE (Shanghai, China). The Amaxa^® Cell Line Nucleofector^® Kit V (VCA-1003) used for electrotransfection was purchased from Lonza (USA). The Lipofectamine^® RNAiMAX reagent (13778075) used for siRNA transfection was purchased from Invitrogen (Carlsbad, USA). The JetPEI Transfection reagent used for plasmid transfection was purchased from Polyplus-transfection SA (Strasbourg, France). BrdU (5-bromo-2-deoxyuridine) (19–160) was purchased from Sigma–Aldrich (St. Louis, MO, USA). A SanPrep column DNA gel extraction kit used for the purification of DNA fragments from PCR or agarose gels was obtained from Sangon Biotech, Inc. (Shanghai, CN). The ClonExpress II One-Step Cloning Kit was purchased from Vazyme Biotech Co. (Nanjing, CN). T1/Mach1-T1 phage-resistant chemically competent E. coli (ZC102-2) was purchased from ZOMANBIO, Inc. (Beijing, China). The NucleoBond Xtra Midi Plus Kit used for plasmid isolation from E. coli was purchased from MACHEREY-NAGEL, Inc. (Bethlehem, PA, USA). The NanoBiT^® protein–protein interaction assay kit (N2014) was purchased from Promega Corporation (Madison, WI, USA).

Cell lines

Primary rat vascular smooth muscle cells (VSMCs) were isolated from the thoracic aortas of 150–180 g male Sprague‒Dawley rats via collagenase digestion⁸⁶. Primary rat VSMCs were cultured in low-glucose Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, USA) supplemented with 10–20% fetal bovine serum (FBS), and cells at passages 3–6 were used in subsequent experiments. Primary human artery smooth muscle cells (SMCs) isolated from arteries of the human umbilical cord were maintained in Nutrient Mixture F12 Ham Kaighn’s Modification (F12K, N6760-10X1L, Sigma Aldrich) supplemented with 10% FBS, 10% SMC Growth Medium (Cell Applications), 100 units/ml penicillin and 100 μg/ml streptomycin. Primary human artery SMCs at passages 4 to 6 were used in subsequent experiments. The mouse aortic smooth muscle cell line MOVAS (CRL-2797), rat aortic smooth muscle cell line A7r5 (CRL-1444) and HEK293T cells (CRL-3216) were purchased from the American Type Culture Collection (ATCC). MOVAS, A7r5 and HEK293T cells were cultured in high-glucose DMEM supplemented with 10% FBS, 100 units/ml penicillin and 100 μg/ml streptomycin. All the cells were maintained in a water-saturated 5% CO₂ incubator at 37 °C.

Generation of the EGFP reporter-expressing MOVAS cell line

A custom-designed CRISPR guide oligo was cloned and inserted into the PX458 plasmid, which contains a cassette for the transient expression of SpCas9. The upstream and downstream homologous arms of the Acta2 gene and P2A-EGFP expression cassette were amplified via PCR and subsequently cloned and inserted into the pUC19 plasmid to obtain the homology template. MOVAS cells were transfected with homology templates and PX458 CRISPR‒Cas9 plasmids using the Amaxa^® Cell Line Nucleofector^® Kit V (Lonza, USA). Ninety-six hours after transfection, EGFP-positive cells were sorted via flow cytometry to obtain single-cell clones. Single-cell clones containing the endogenous Acta2-P2A-EGFP gene sequence were further verified by sequencing analysis.

The following sgRNAs were used in this study:Acta2-sgRNA-1: 5’-GGACTTAGAAGCATTTGCGG-3’; Acta2-sgRNA-2: 5’-AACAGGAATACGACGAAGCT-3’.

Lentivirus production

We used the lentiCas9-Blast plasmid (Addgene 52962) and the mouse CRISPR Gecko v2 pooled library (Addgene 1000000052) to generate lentiviruses. For each 10 cm dish, HEK293T cells at 90% confluence were transfected with 6.8 μg of the plasmids of interest mentioned above, 5.2 μg of psPAX2 (Addgene 12260), and 3.4 μg of pMD2.G (Addgene 12259) using 35 μl of Lipofectamine 3000 (Thermo Fisher Scientific L3000150), 30 μl of P3000 Enhancer (Thermo Fisher Scientific L3000150), and 2.5 ml of Opti-MEM (Thermo Fisher Scientific 31985070). The media were changed 6 h after transfection. The supernatant containing the lentivirus was harvested at 48 h posttransfection, filtered with a 0.45-μm PVDF filter, and stored at −80 °C.

Genome-wide CRISPR knockout screening

The EGFP reporter-expressing MOVAS cell line was transduced with lentiCas9-Blast. Blasticidin (10 µg/ml) was added to the cells at 72 h after infection, and the mixture was maintained in culture for 7 days to screen the cells stably overexpressing Cas9, which was further verified by western blot analysis. A total of 240 million MOVAS cells that stably expressed EGFP and Cas9 were subsequently seeded into twenty 10 cm plates and infected with the pooled GeCKO v2 mouse lentiviral library (over 300 × coverage) at an MOI of 0.3 to ensure that most cells took up only one sgRNA. Puromycin (10 µg/ml) was added to the cells at 72 h after infection, and the cells were maintained in culture for 7 days. The cells were subsequently sorted via two rounds of flow cytometry to enrich the EGFP^low population.

Genomic DNA was purified from the EGFP^low population, as well as from the initial infected population, using the Gentra Purogene Kit (Qiagen). Integrated sgRNAs were enriched by PCR amplification, with eight replicate PCRs each with 1 µg of template per reaction, to maintain complex sampling of the cellular population. Adapter sequences and per-sample barcodes were added to the libraries via a second round of PCR. The libraries were pooled and sequenced on an Illumina MiSeq instrument using 150-bp single-end reads. The enriched genes were identified from the sgRNA sequencing results using Model-based Analysis of Genome-wide CRISPR-Cas9 Knockout (MAGeCK). The MAGeCK algorithm can prioritize enriched genes by comparing the sgRNAs in EGFP^low cells to those in unsorted cells. Briefly, the read counts of each sgRNA from different samples were normalized to adjust for the effects of the library size and read count distribution. The enriched genes were subsequently identified by searching for genes whose sgRNAs ranked consistently higher using robust rank aggregation (RRA). Genes with smaller RRA values ranked higher in the knockout screen.

Flow cytometry

For the analysis of the EGFP fluorescence intensity during screening, the cells were suspended in PBS supplemented with 2% FBS for direct detection. The cells were then analyzed on an ARIA-SORP sorter (Becton Dickinson, USA). The EGFP fluorescence intensity was detected using the FITC channel. The data were analyzed using FlowJo software.

Human and animal samples

Human internal mammary artery samples were obtained from Peking University People’s Hospital, Beijing, China. The study involving human tissue was approved by the Medical Ethics Committee of Peking University People’s Hospital. Internal mammary arteries were obtained from patients that underwent coronary artery bypass/valve replacement. The informed consent form was signed by the donors before collection. All sample collection and experimental procedures followed national laws and international ethical and technical guidelines. The entire procedure was examined and approved by the Ethics Board at the Institute of Peking University People’s Hospital (approval no. 2023PHB170-001). Human internal mammary artery samples were immediately placed in 4% formalin, embedded in paraffin, and made into serial sections for subsequent immunohistochemical staining.

Lemd3^flox/flox mice in which exon 2 was flanked with LoxP sites were generated from C57BL/6J mice using the CRISPR/Cas9 system. Genotyping was performed by PCR using the following two primers: forward-5’-CGGACAGTGAGCGAGGCATT-3’ and reverse-5’-AATGTAAACAAACGAGGTAAGAAGC-3’. The PCR program was as follows: 94 °C, 2 min; 98 °C, 10 s; 60 °C, 30 s; 68 °C, 30 s (35 cycles); 68 °C, 10 min; and 4 °C, hold. The length of the PCR product was 217 bp (+) or 183 bp (-). The Myh11-CreER^T2 mice were kindly provided by Prof. Wei Li from Peking University People’s Hospital⁸⁷. Since the bacterial artificial chromosome containing Myh11-CreER^T2 was inserted into the Y chromosome, only male mice were used. We intercrossed Myh11-CreER^T2 Lemd3^flox/wt male mice with Lemd3^flox/wt female mice to obtain offspring Myh11-CreER^T2 Lemd3^flox/flox male mice as well as the littermate Myh11-CreER^T2 male mice. Eight-week-old male Myh11-CreER^T2 Lemd3^flox/flox mice and Myh11-CreER^T2 mice were treated with 75 mg/kg tamoxifen for 5 consecutive days to generate smooth muscle cell-specific Lemd3 knockout (Lemd3^SMKO) mice and Lemd3^WT control mice, respectively. The primers used for Myh11-CreER^T2 genotyping were as follows: SMWT1-5’-TGACCCCATCTCTTCACTCC-3’, SMWT2-5’-AACTCCACGACCACCTCATC-3’ and phCREAS1-5’-AGTCCCTCACATCCTCAGGTT-3’. The PCR program was as follows: 95 °C for 1 min; 95 °C for 10 s, 52 °C for 30 s, and 72 °C for 45 s (35 cycles); 72 °C for 5 min; and hold at 4 °C. The length of the PCR product was 287 bp (+). All experimental animals were maintained in standard cages in an SPF environment with a 12-h light/dark cycle in the Department of Laboratory Animal Science, Peking University Health Science Center. All studies were conducted in compliance with the guidelines of the Animal Care and Use Committee of Peking University.

Primary aortic smooth muscle cell isolation

VSMCs were isolated from the thoracic aortas from 6- to 10-week-old Lemd3^flox/flox mice as described before⁸⁸, with minor modifications. Briefly, we separated the thoracic aortas from the mice. The isolated aortas were washed twice in ice-cold PBS and cultured in 1 mL 0.2% collagenase I solution in Ham’s F12 medium at 37 °C for 30 min. The adventitia was stripped away from the aorta using forceps under microscopic guidance. The aortas were opened longitudinally and the endothelial cells were gently scraped off. The aortas were then divided into small pieces, placed at the bottom of the culture dish, and cultured for several days in DMEM/F-12 medium containing 20% FBS at 37 °C in a humidified atmosphere with 5% CO₂. Cells that had migrated from the explants were collected and maintained in the growing medium. VSMCs at passages 3–6 were used for further experiments. Purity of VSMCs was confirmed via positive staining for ACTA2.

Real-time quantitative polymerase chain reaction (RT‒qPCR)

Total RNA was extracted using TRIzol reagent (Vazyme Biotech Co., Nanjing, China), and equal amounts (1 μg) were reverse-transcribed into cDNA using a reverse transcription kit (Vazyme Biotech Co., Nanjing, China). Quantitative real-time PCR (RT‒qPCR) was performed using 2 × SYBR Green PCR mix (Vazyme Biotech Co., Nanjing, China) according to the manufacturer’s instructions. All amplification reactions were conducted in 40 cycles and were performed in triplicate. The data were analyzed via the ΔΔCT method. All samples were normalized to β-actin. The primers used for RT‒qPCR are presented in the Supplementary Table 2.

Immunohistochemical and immunofluorescence staining

For immunohistochemical staining, cultured cells were fixed with 4% paraformaldehyde for 15 min and then permeabilized with 0.5% Triton X-100 for 20 min, while tissue sections were directly permeabilized with 0.5% Triton X-100 for 20 min. Cells or sections were incubated with primary antibodies at 4 °C overnight and then with secondary antibodies before being stained with a DAB kit (ZSGB-BIO, Beijing, China). The nuclei were counterstained with hematoxylin. Cells or sections incubated with species-matched IgG were used as negative controls. Images were obtained by using a Zeiss microscope. Immunohistochemical images were quantified by using ImageJ software. We first selected the positive area according to the same criteria, and then use the average brightness multiplied by the area as its expression intensity, and results were expressed as the ratio of relative intensity compared with the controls.

For immunofluorescence staining, cultured cells were fixed with 4% paraformaldehyde for 15 min and then permeabilized with 0.5% Triton X-100 for 20 min. The sections were directly permeabilized with 0.5% Triton X-100 for 20 min. The cells or sections were incubated with primary antibodies at 4 °C overnight, followed by an incubation with a secondary Alexa Fluor 488-conjugated goat anti-rabbit IgG at a 1:1000 dilution and secondary Alexa Fluor 555-conjugated goat anti-mouse IgG (Thermo Fisher Scientific, Rochester, USA) at a 1:1000 dilution for 1 hour at room temperature. The nuclei were stained with DAPI at a 1:1000 dilution for 10 min at room temperature. For F-actin staining, the cells were fixed with 4% paraformaldehyde for 15 min and then permeabilized with 0.5% Triton X-100 for 20 min. Next, the cells were incubated with YF633-phalloidin at a 1:100 dilution for 1 h at room temperature. The nuclei were stained with DAPI at a 1:1000 dilution for 10 min at room temperature. The fluorescence signals were monitored using a confocal laser scanning microscope (Leica Microsystems, Wetzlar, Germany).

Western blotting

Cells or mouse tissues were lysed in RIPA buffer (Beyotime, Shanghai, China), and whole-cell protein samples were extracted. Protein concentrations were evaluated using a BCA protein assay kit (Beyotime, Shanghai, China). Equal amounts of total protein were resolved on 8–12% SDS‒PAGE gels and subsequently transferred onto nitrocellulose membranes (PALL Inc., Wilmington, DE, USA). The membranes were blocked with 5% milk or BSA in TBST. After being incubated with primary antibodies at 4 °C overnight, the membranes were incubated with IRDye-conjugated secondary antibodies (Rockland, Inc., Gilbertsville, PA, USA) for 1 hour at room temperature. The immunofluorescence signals were subsequently detected using an Odyssey infrared imaging system (LI-COR Biosciences, Lincoln, NE, USA).

siRNA transfection

Small interfering RNAs (siRNAs) targeting rat Lemd3, Cand1, Ajuba, Gdi1 or Xpo6, and scrambled siRNAs were designed and synthesized by GenePharma (Suzhou, China). Rat VSMCs were transfected with siRNAs (20 nM) using RNAiMax reagent (Invitrogen, CA, USA) according to the manufacturer’s protocol. The list of siRNA sequences is presented in the Supplementary Table 3.

Collagen gel contraction assay

Primary rat VSMCs transfected with scrambled siRNAs (20 nM) or siRNAs targeting Lemd3 (20 nM) were resuspended in DMEM containing 10% FBS at a density of 5 × 10⁵ cells/ml. A collagen lattice was prepared by mixing the cell suspension and an ice-cold collagen gel solution at a volume ratio of 1:4. Then, 0.5 ml of the mixture was added to a well of a 24-well plate, followed by an incubation at 37 °C for 1 h. After collagen polymerization, 1 ml of DMEM was added on top of each collagen gel lattice. Twenty-four hours later, the plates were scanned, and the area of the gel in each well was analyzed using ImageJ software.

Wire myography

Wire myography was performed to assess the contractility of the aortic rings and mesenteric resistance arteries from 12-week-old Lemd3^WT mice and Lemd3^SMKO mice. Briefly, the aortic rings and mesenteric resistance arteries were isolated and removed from the mice following euthanasia. The dissected vessels were immediately placed in Krebs‒Henseleit buffer. After the removal of additional connective tissues, the aortic rings and mesenteric resistance arteries were sectioned to approximately 1.5 mm in length and subsequently subjected to vascular tension measurements. The cumulative dose response to phenylephrine (Phe) was obtained to characterize vasocontraction. The data are presented as percentages of the maximal contraction induced by 60 mM KCl.

In vitro EdU incorporation assay

A BeyoClick™ EdU Cell Proliferation Detection Kit (Beyotime, Shanghai, China) was used to detect cell proliferation in accordance with the manufacturer’s instructions. Primary rat VSMCs were transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM). 48 h after transfection, 20 μΜ EdU was added, and the cells were subsequently incubated for another 3 h. Following fixation with 4% paraformaldehyde for 15 min and permeabilization with 0.5% Triton X-100 for 20 min, the cells were incubated with Click reaction cocktail for 30 min and then stained with DAPI at a 1:1000 dilution for 10 min. The cells were observed and photographed with an inverted fluorescence microscope (Olympus, Japan).

Migration assay

Cell migration was determined by performing wound healing and transwell migration assays. For the wound healing assay, primary rat VSMCs transfected with the scrambled siRNA or siRNA targeting Lemd3 were seeded in six-well plates and wounded by manually scraping the cells with a 200 µl pipette tip. The medium was replaced with FBS-free DMEM supplemented with or without PDGF-BB (20 ng/ml). The migration areas were monitored at 0 h and 12 h postwounding under a microscope. For the transwell migration assay, primary rat VSMCs transfected with the scrambled siRNA or siRNA targeting Lemd3 were seeded in the top chambers of 8-µm pore size transwells in plates (Corning Costar, Lowell, MA) in FBS-free DMEM. In addition, 0.6 ml of DMEM supplemented with 20% FBS was added to the well of the plate (lower compartment). After a 12-h incubation at 37 °C, the cells on the upper surface of the membrane were removed via gentle abrasion with a cotton bud, and the cells that migrated to the lower surface were fixed with 4% paraformaldehyde for 15 min and stained with 0.1% crystal violet. The cells were subsequently washed with PBS several times to remove the excess dye. The mean number of cells on the lower side of the surface was counted in 4 randomly chosen high-power fields (HPFs) under a light microscope. ImageJ software was used to measure the migration area and cell number.

Mouse carotid artery wire injury

A carotid artery wire injury model was generated in 12-week-old male Lemd3^WT mice and Lemd3^SMKO mice. Briefly, after isolation of the left common carotid artery, carotid clamps were used to block the blood flow of the left common carotid artery and the left internal carotid artery. Next, a curved flexible wire (0.38 mm) was inserted into the left external carotid artery to induce carotid artery injury. The wire was subsequently removed, and the left external carotid artery was tied proximally to the aortic arch. The mouse carotid arteries were harvested at 28 days after wire injury. The mouse carotid arteries were collected for hematoxylin‒eosin staining and immunofluorescence staining.

Quantification of neointima formation

The mouse carotid arteries were harvested and embedded in Tissue-Tek OCT (Sakura Finetek, Staufen, Germany). Every other section (7 μm each) was collected within a standardized distance from the bifurcation point (1000 to 2000 μm, the segments marked with curly brackets in Supplementary Fig. 7e) and these sections were divided into 20 slides, so that every slide contained eight sections at 140 μm intervals^89,90. Following H&E staining on the cross-sections, the neointima was defined as the region between the lumen and internal elastic lamina. The medial wall was defined as the region between the internal and external elastic laminas. The areas of the neointima and media on cross-section of H&E-stained artery segments were measured with a computerized image analysis system (Image-Pro Plus 6.0 software) by two independent investigators in a blinded manner. For quantitative analysis of neointima formation, the mean of eight sections on one slide was taken for each mouse.

Measurement of blood pressure

Blood pressure was measured in Lemd3^WT mice and Lemd3^SMKO mice before and after surgery using a CODA Mouse & Rat Tail-Cuff Blood Pressure System (Kent Scientific Co., Connecticut, USA) according to the scientific statement from the American Heart Association⁹¹. The equipment was kept clean and free from foreign scents and blood odors. The investigator was blinded with respect to the experimental groups to perform the measurements, while the mice were tested in a randomized manner. The mice underwent 7 consecutive days of training sessions from 1 to 5 P.M. each day to become accustomed to the tail-cuff procedure. The measurements were then recorded by a single investigator at the same time on 3 consecutive days. Five measurements were performed daily on each mouse, and the average of 15 measurements represented the systolic blood pressure of each mouse.

In vivo EdU incorporation assay

A carotid artery wire injury model was generated in 12-week-old male Lemd3^WT mice and Lemd3^SMKO mice as described above. The mice were sacrificed at 28 days after surgery. The mice were intraperitoneally injected with 50 mg/kg 5-ethynyl-2’-deoxyuridine (EdU) 18 h prior to sacrifice. Arterial sections were washed with PBS and incubated with 0.5% Triton X-100 for 20 min to permeabilize the cell membrane. The sections were incubated with Click reaction cocktail for 30 min and then stained with DAPI at a 1:1000 dilution for 10 min. The sections were observed and photographed with an inverted fluorescence microscope (Olympus, Japan).

LEMD3 interactome

LEMD3-binding proteins were identified via liquid chromatography tandem mass spectrometry (LC‒MS/MS)⁹². Briefly, HEK293T cells were transfected with the pcDNA3.1 or LEMD3-FLAG plasmid. Forty-eight hours after transfection, the cells were lysed in lysis buffer (50 mM Tris-HCl [pH=7.4], 150 mM NaCl, 2 mM EDTA, and 0.3% NP-40). The cell lysate samples were incubated with anti-FLAG M2 affinity gel (A2220, Sigma‒Aldrich) according to the manufacturer’s instructions. The anti-FLAG M2 affinity gel-bound proteins were eluted with FLAG peptide (F3290, Sigma). The final eluent was collected and resolved on an SDS-PAGE gel. The gel was fixed with a 30% ethanol:10% acetic acid solution, washed with 10% ethanol, and subjected to standard reduction, alkylation, and in-gel tryptic digestion. The resulting peptides were lyophilized and reconstituted in 0.1% (v/v) aqueous formic acid. Peptide separation was performed using an Easy-nanoLC system coupled online to a LTQ Orbitrap Velos Pro mass spectrometer. Chromatography utilized a reversed-phase column with mobile phase A (0.1% formic acid in water) and mobile phase B (0.1% formic acid in acetonitrile) at a flow rate of 300 nL/min. The elution gradient was programmed as follows: 2% to 40% B over 70 min, 40% to 95% B over 5 min (70–75 min), followed by a 5-min hold at 95% B. Eluting peptides were ionized via a nano-electrospray ionization source (spray voltage: 2.2 kV; capillary temperature: 300 °C) and analyzed by the LTQ Orbitrap Velos Pro. Full-scan MS spectra (m/z 350–2000) were acquired in the Orbitrap analyzer. In data-dependent acquisition mode, the 15 most intense precursor ions from each full scan were sequentially selected for CID-MS/MS fragmentation in the linear ion trap, using a normalized collision energy of 35%. Dynamic exclusion was enabled with a duration of 30 s to prevent repeated sequencing of abundant ions. Protein identification was performed using Proteome Discoverer software (version 1.4) with database searches against the UniProt Human Proteome database. Search parameters included a precursor mass tolerance of 10 ppm, a fragment mass tolerance of 0.6 Da, and trypsin specificity allowing up to two missed cleavages. Fixed modification was set as carbamidomethylation of cysteine (+57.021 Da), while oxidation of methionine (+15.995 Da) was specified as a variable modification. All peptide-spectrum matches (PSMs) were filtered at a strict false discovery rate (FDR) threshold of ≤ 1%. Two independent biological replicates of LC‒MS were performed. Each replicate included an experimental sample and a negative control sample. Samples from HEK293T cells transfected with the pcDNA3.1 plasmid were used as the negative controls.

Plasmid construction and transfection

The cDNA fragments encoding FLAG-tagged full-length human LEMD3 (NCBI reference sequence: NM_014319.5, LEMD3-FLAG, aa 1–911), LEMD3 ΔC-terminal mutant aa 1–876 and LEMD3 ΔRRM mutant aa 1–784, as well as human CBX3 (NCBI reference sequence: NM_016587.4), were subsequently cloned and inserted into the pcDNA3.1 plasmid. Cells cultured in 6-well plates were transfected with 2 μg of plasmid per well using jetPEI (Polyplus-transfection SA, Strasbourg, France).

Coimmunoprecipitation (Co-IP)

HEK293T cells transfected with LEMD3-FLAG and CBX3 plasmids were lysed with immunoprecipitation (IP) buffer (50 mM Tris-HCl [pH=7.4], 150 mM NaCl, 2 mM EDTA and 0.3% NP-40) supplemented with a protease inhibitor cocktail to analyze the binding of LEMD3 and CBX3. The lysates were incubated with either anti-FLAG M2 antibody (2 μg), anti-CBX3 antibody (2 μg) or negative control IgG antibody (2 μg) at 4 °C overnight. Then, Protein A/G agarose beads (Santa Cruz, CA, USA) were added and incubated for another 4 h at 4 °C. The precipitated proteins were eluted from the beads by resuspending the beads in 2× SDS‒PAGE sample loading buffer. After boiling for 10 min at 95 °C, the protein samples were analyzed via western blotting.

Cultured rat VSMCs and mouse aortic tissues were lysed with IP buffer supplemented with a protease inhibitor cocktail to analyze the endogenous binding of LEMD3 and CBX3. The lysates were incubated with either an anti-LEMD3 antibody (2 μg), anti-CBX3 antibody (2 μg) or negative control IgG antibody (2 μg) at 4 °C overnight. Then, Protein A/G agarose beads (Santa Cruz, CA, USA) were added and incubated for another 4 h at 4 °C. The precipitated proteins were eluted from the beads by resuspending the beads in 2 × SDS‒PAGE sample loading buffer. After boiling for 10 min at 95 °C, the protein samples were analyzed via western blotting.

NanoBiT-based binding assay

The NanoBiT^® protein–protein interaction assay kit (N2014) was purchased from Promega Corporation (Madison, WI, USA). SmBit-CBX3 expression constructs were generated by cloning CBX3 into the pBiT2.1-N vector containing the N-terminal SmBit. LgBit-LEMD3, LgBit-ΔC-terminal, and LgBit-ΔRRM expression constructs were generated by cloning LEMD3, the ΔC-terminal mutant, and the ΔRRM mutant into the pBiT1.1-N vector containing the N-terminal LgBit, respectively. HEK293T cells were transfected with empty vector or SmBit-CBX3 and LgBit-LEMD3, LgBit-ΔC-terminal, or LgBit-ΔRRM plasmids in white 96-well cell culture plates. Thirty-six hours after transfection, the culture media were removed and replaced with HBSS containing 1 mg/ml bovine serum albumin (BSA). The substrate furimazine was then added to each well. The plate was incubated for 30 min, and luminescence was measured using a Thermo Scientific Varioskan Flash reader.

Transmission electron microscopy (TEM)

The descending aortas of the Lemd3^WT mice and Lemd3^SMKO mice were isolated, cut into 3 mm rings, and fixed with 3% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4) for 3 h. After being washed in cacodylate buffer, the tissues were postfixed with 1% OsO₄, dehydrated through an ethanol series, infiltrated with propylene oxide and embedded in Epon 812. Ultrathin sections were cut on an ultramicrotome (LKB Huxley, Stockholm, Sweden) and counterstained with uranyl acetate and lead citrate. Images were acquired using a transmission electron microscope (HITACHI H-7000, Tokyo, Japan). We randomly selected four vascular smooth muscle cells from each mouse aortic section to measure the width of perinuclear heterochromatin. The width of perinuclear heterochromatin was quantified by measuring the width of the binarized chromatin from 1D intensity profiles along the nuclear perimeter, sampled at every 10 perimeter pixels⁹³, by two independent investigators in a blinded manner.

Bulk RNA-seq library preparation

The A7r5 rat VSMC line was transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM). RNA was extracted from rat VSMCs using the Qiagen RNeasy Mini Kit according to the manufacturer’s instructions. The RNA-seq libraries were prepared using the NEBNext Ultra RNA Library Pre Kit for Illumina and sequenced on the Illumina NextSeq platform.

ChIP-seq library preparation

A7r5 rat VSMCs transfected with the scrambled siRNA or siRNA targeting Lemd3 were crosslinked with 1% formaldehyde for 10 min at room temperature and quenched by an incubation with glycine added at a final concentration of 125 mM for 5 min. The fixed cells were resuspended in lysis buffer (1% SDS, 5 mM EDTA, and 50 mM Tris-HCl, pH 8.1) containing protease inhibitors and then subjected to 30 cycles (30 s on and 30 s off) of sonication (Bioruptor, Diagenode) to generate chromatin fragments of ~300 bp in length. Lysates were diluted in buffer (1% Triton X-100, 2 mM EDTA, 150 mM NaCl, and 20 mM Tris-HCl, pH 8.1) containing protease inhibitors. For immunoprecipitation, the diluted chromatin was incubated with normal IgG (control), H3K9me3, or H3K27ac antibodies (3 μg) overnight at 4 °C with constant rotation, followed by an incubation with 50 μl of 50% (v/v) Protein A/G Sepharose beads for an additional 2 h. The beads were washed with the following buffers: TSE I (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 150 mM NaCl, and 20 mM Tris-HCl, pH 8.0); TSE II (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 500 mM NaCl, and 20 mM Tris-HCl, pH 8.0); and TSE III (0.25 M LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mM EDTA, and 10 mM Tris-HCl, pH 8.0). The pulled-down chromatin complexes were eluted with TE (1 mM EDTA and 10 mM Tris-HCl, pH 8.0), and the inputs were decrosslinked at 55 °C for 12 h in elution buffer (1% SDS and 0.1 M NaHCO₃). The DNA was purified with a QIAquick PCR Purification Kit (QIAGEN). ChIP-seq libraries were prepared using the VAHTS^® Universal Pro DNA Library Prep Kit for Illumina (Vazyme Biotech Co., Nanjing, China) and sequenced on the Illumina NextSeq platform.

ATAC-seq library preparation

ATAC-seq libraries were prepared according to the standard ATAC-seq protocol with minor modifications⁹⁴. Briefly, 50,000 cells transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) were lysed on ice in a tube containing 50 μl of cold lysis buffer. The nuclei were collected and subjected to a transposition reaction in a preheated metal bath at 37 °C. Then, the DNA fragments were extracted and purified with a PCR purification kit. The DNA was then resuspended in 29 μl of ddH₂O. ATAC-seq libraries were then prepared using the High-Sensitivity Open Chromatin Profile Kit for Illumina (N248, Novoprotein) and sequenced on the Illumina NextSeq platform.

Hi-C library preparation

The experiment was performed according to the in situ Hi-C protocol⁹⁵. Briefly, A7r5 rat VSMCs transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM) were washed with PBS and crosslinked with 1% formaldehyde, and the reaction was quenched with 200 mM glycine. The cells were lysed, the DNA was then cut with DpnII, and the overhangs were filled with a biotinylated base. The free ends were then ligated together in situ. The crosslinks were reversed, the DNA was sheared to 300–500 bp, and then the biotinylated ligation junctions were recovered with streptavidin beads. The sequencing libraries were generated using a standard Illumina library construction protocol. Briefly, ends of sheared DNA were repaired, and the blunt ends were added to an “A” base to ligate with Illumina’s adapters that have a single “T” base overhang. The DNA was subsequently amplified via PCR for 8‒12 cycles. Products were purified using the AMPure XP system and sequenced on the Illumina HiSeq X Ten platform.

Bulk RNA-seq analysis

The raw RNA-seq reads were mapped to the rat genome version rn7 using TopHat (v2.1.1)⁹⁶, and the expression value (number of raw reads) for each gene was determined via the software HTSeq2 (v2.2.1). Normalized (using the upper quartile method) expression values and differentially expressed genes were determined using edgeR. Genes with |logFC | >1 and adjusted P value < 0.05 were considered differentially expressed genes (DEGs). We used the function bamCoverage in deepTools (3.5.1)⁹⁷ and normalized the bam files by a scaling factor. We used Reads Per Kilobase per Million mapped reads (RPKM) to generate a BigWig file that contained the RNA-seq signal (read density) at each base pair across the genome.

ChIP-seq analysis

Paired ChIP-seq raw reads were trimmed with TrimGalore (v0.32), then aligned to the rn7 genome using bwa-mem2 (v2.0pre2), and converted to SAM files. The SAM files were then sorted and converted to bam format using SAMtools (v1.15.1). PCR duplicates were removed using gatk (v4.2.6.1) MarkDuplicates. Finally, bam files were filtered using samtools view with “-f 2 -q 24” parameters. We used danpos (v2.4)⁹⁸ to call peaks and generate normalized wig files with the parameters “danpos.py dregion --smooth_width 0 -c 25000000 -u 1 --frsz 200 --extend 200 --region_dis 3000 --pheight 1e-10” for the ChIP-seq data. We used the wigToBigWig (v4) tool to convert wig files to BigWig files. We used the top 10,000 widest H3K9me3 peaks in the siRNA_scramble group to plot a heatmap of the average signal in both groups using deepTools (3.5.1)⁹⁷ with the functions “computeMatrix reference-point” and “plotHeatmap”.

ATAC-seq analysis

Paired raw ATAC-seq reads were trimmed with TrimGalore (v0.32), aligned to the rn7 genome using bwa-mem2 (v2.0pre2), and then converted to SAM files. The SAM files were then sorted and converted to bam format using SAMtools (v1.15.1). PCR duplicates were removed using gatk (v4.2.6.1) MarkDuplicates. Finally, bam files were filtered using samtools view with “-f 2 -q 24” parameters. We merged three biological replicates using “samtools merge”. We used danpos (v2.4)⁹⁸ to call peaks and generate normalized wig files with the parameters “danpos.py dpeak -c 25000000 -u 1 --frsz 200 -m 1 --smooth_width 0 --pheight 1e-20” for ATAC-seq. We used the wigToBigWig (v4) tool to convert wig files to BigWig files. We used the csaw R package⁹⁹ to obtain different ATAC-seq peaks using sliding windows at 300 bp with a cutoff (adjusted P value < 0.05). Enrichment for known TF-binding sites of ATAC peaks was performed using hypergeometric optimization of motif enrichment (HOMER)¹⁰⁰. We first lifted the ATAC peak position from the rn7 version to the rn6 version of the genome. Furthermore, findMotifsGenome.pl was run on peak regions with the parameters “rn6 -size 200 -mask -p 24 -S 100 -len 6,8”.

Hi-C data analysis

For all Hi-C datasets of rat VSMCs, we used the HiCExplorer (v3.7.2)¹⁰¹ analysis pipeline. The read pairs were aligned to the rn7 genomes using the bwa-mem2 algorithm. The aligned read pairs were assigned to DpnII restriction fragments. Invalid pairs with dangling ends were excluded. The genome was divided into bins of specific sizes, and valid pairs were counted per bin. We used 100, 40, 20, and 10 kb bin sizes to generate raw and iterative correction and eigenvector decomposition (ICE) normalized matrices. For the Hi-C analysis of human samples, we downloaded the data in hic format from the GEO database, which contains Hi-C contact maps with resolutions of 10, 25, 100 kb, and converted them into 40 kb-resolution contact maps and saved them in h5 format by using the “hicConvertFormat” and”hicMergeMatrixBins”. A 100 kb matrix was used to identify compartments via the juicer_tools (v1.6)¹⁰² eigenvector, and a 40 kb matrix was used to define TAD boundaries by calculating the insulation score using “hicexplorer hicFindTADs”.

For TAD-TAD contacts, we classified genome-wide interactions into intra-TAD interactions, interactions between neighboring TADs and interactions between non-neighboring TADs (long-range interactions), and counted the contact values within each region. Differences in TAD-TAD contacts between the WT group and the Lemd3 KD group were compared with edgeR. We utilized “bedtools intersect” from bedtools (v2.18) with the parameters “F = 1e-9, f = 0.5” to retrieve the genes located within each of these three types of TAD pairs based on the chromatin locus. The surrounding TADs were divided into three types, A-A, B-B, and A-B (also the boundaries of A/B compartments), based on the eigenvector of the TAD position. The ATAC signal peaks in the TADs at the boundaries of A and B compartments were sorted according to the fold change in Hi-C contacts of their TADs after knockdown of Lemd3. We calculated the averages of the fold change of ATAC signal and the fold change of Hi-C contacts corresponding to every 200 ATAC peaks, and then calculated the Pearson’s correlation coefficients between the two factors.

We pooled two Hi-C samples from the same group. The log2Ratio matrix was obtained with “hicexplorer hicCompareMatrices”. Visualization of the pooled sample was implemented using “hicexplorer hicPlotMatrix” and “hicexplorer hicPlotTADs”.

DNA polymer modeling

We applied a computational approach to derive polymer models and 3D structures for pooled samples⁵¹. We studied the genome organization of 100-kb genomic segments that represent each bead. This model provides a more detailed representation of the genome and enables us to characterize the spatial localization of the A and B compartments. In our simulations, we modeled a total of 42 chromosomes, representing the full complement of chromosomes in a diploid mammalian nucleus. These include 21 homologous pairs, with each pair consisting of two identical chromosomes to represent sister chromatids. Due to the limitations of Hi-C data in distinguishing between sister chromatids, we assumed that both copies within each pair share the same A/B compartment profile and spatial organization. Each chromosome is represented as a coarse-grained polymer chain, with beads corresponding to 100 kb genomic segments, and interactions are derived from experimentally inferred contact probabilities. This polymer model uses an ensemble of structures, instead of a single, unique conformation, to reproduce Hi-C data. The ensemble of structures is assumed to follow a Boltzmann distribution with a potential energy function ${U}_{{ME}(r)}$, the expression of which can be derived according to the maximum entropy principle and is provided below. The parameters of the polymer model are solely encoded in the energy function. Molecular dynamics simulations were conducted to collect structures consistent with the energy function and the Boltzmann distribution.

The energy function of the chromosome model adopts the following form:

$${U}_{{ME}(r)}=U\left(r\right)+{U}_{{ideal}}\left(r\right)+{U}_{{compt}}(r)$$

where $r$ represents the 3D conformation of the entire genome. By definition, $U\left({{{\rm{r}}}}\right)$ and ${U}_{{ideal}}\left(r\right)$ are generic potentials shared by all chromosomes, and ${U}_{{compt}}(r)$ describes compartment-type-specific interactions within the same chromosome and between different chromosomes.

Specifically, $U\left({{{\rm{r}}}}\right)$ is the energy function for a confined homopolymer and consists of four terms: ${U}_{{bond}}$, ${U}_{{angle}}$, ${U}_{{sc}}$ and ${U}_{c}$. ${U}_{{bond}}$ is the bonding potential between neighboring beads. ${U}_{{angle}}$ is the angular potential applied among every three neighboring beads to define the persistence length of the polymer. ${U}_{{sc}}$ is a soft-core potential applied to all the nonbonded pairs to enforce the excluded volume effect among genomic loci. ${U}_{c}$ models a spherical boundary and is introduced to mimic the confinement effect applied by the nuclear envelop onto chromosomes. $I$ is an index over different chromosomes, and $i$ and $j$ are indices over all pairs of nonbonded chromatin beads from chromosome $I$.

$$U\left(r\right)={\sum}_{i\in I}[{U}_{{bond}}\left({r}_{i,i+1}\right)+{U}_{{angle}}\left({r}_{i,i+1},{r}_{i+1,r+2}\right)+{U}_{c}\left({r}_{i}\right)]+{\sum}_{i,j\in I,j > i}{U}_{{sc}} ({r}_{{ij}})\left)\right.$$

${U}_{{ideal}}\left({r}_{I}\right)$ measures the average contact probability for genomic loci separated at a given sequence length of $|j-i|$. It is independent of the compartment type.

$${U}_{{ideal}}\left({r}_{I}\right)={\sum}_{i,j\in I}{\alpha }_{{ideal}}|j-i|f({r}_{{ij}})$$

${U}_{{compt}}(r)$ describes compartment-type-specific interactions within the same chromosome ${U}_{{intra}}(r)$ and between different chromosomes ${U}_{{inter}}(r)$ as follows:

$${U}_{{compt}}\left(r\right)={U}_{{intra}}\left(r\right)+{U}_{{inter}}(r)$$

$${U}_{{intra}}\left(r\right)={\sum}_{i,j\in I}{a}_{{intra}}({T}_{i}^{I},{T}_{j}^{I})f({r}_{{ij}})$$

$${U}_{{inter}}\left(r\right)={\sum}_{i,j\in I}{a}_{{inter}}({T}_{i}^{I},{T}_{j}^{I})f({r}_{{ij}})$$

where $f({r}_{{ij}})$ determines the contact probability of a genomic pair with a spatial distance of ${r}_{{ij}}$. $f\left({{{\rm{r}}}}\right)$ was defined as follows:

$$f\left({{{\rm{r}}}}\right)=\left\{\begin{array}{c}\frac{1}{2}\left[1+\tanh \left(\sigma \left({r}_{c}-r\right)\right)\right],{ifr}\le {r}_{c}\\ \frac{1}{2}{(\frac{{r}_{c}}{r})}^{4},{ifr} > {r}_{c}\hfill\end{array}\right.$$

where ${r}_{c}$ = 2 and $\sigma$ = 2. The parameters ${\alpha }_{{ideal}}$, ${a}_{{intra}}$ and ${a}_{{inter}}$ are entirely encoded in the energy function, their values were determined iteratively so that the simulated structures reproduce Hi-C contact maps to match the experimentally determined contact frequencies. ${{{{\rm{U}}}}}_{{{{\rm{ME}}}}({{{\rm{r}}}})}$ can be shown as the least biased potential to reproduce these experimental constraints following the maximum entropy principle. These ensure that the following ensemble averages determined with the simulated genome conformation match the corresponding experimental constraints calculated using Hi-C data; we used the parameters from Qi and Zhang¹⁰³.

Details of the molecular dynamic simulation and analysis

Molecular dynamics simulations were performed using the software package LAMMPS (v2018). The simulations were maintained at a constant temperature of $T$ = 1 via Langevin dynamics with a damping coefficient of $\gamma$ = 10.

We first placed all the chromosomes consecutively on a cubic lattice with an edge length of $0.9{{{\rm{R}}}}/\sqrt{3}$, where ${{{\rm{R}}}}$ is the radius of the spherical confinement introduced to ensure a volume fraction of 0.1, to generate an initial configuration for these simulations. This configuration was subsequently equilibrated during a 100,000-step-long simulation under a time step of ${dt}$ = 0.0001 and a 10,000-step-long simulation under a time step of ${dt}$ = 0.001 to relax both the topology and energy of the polymer structures. A 4,000,000-step-long simulation with a time step of ${dt}$ = 0.01 was subsequently performed. The genome conformations were saved every 2000 timesteps to calculate the ensemble averages. The last state from this equilibration trajectory was then used to visualize the genome structure with VMD (194a57).

We divided the radius of the simulated nucleoplasmic space from the core to the perinuclear space into 50 equal parts of equal length, and the whole space was divided into 50 parts of unequal volume from inside to outside to compare the ratio of the A/B compartments from the core to the perinuclear region. We calculated the ensemble averages from the last saved 500 genome conformations, defined the number of compartments per unit volume as the density of the compartments, and counted the A/B compartments in each layer of the nucleoplasmic space in the ensemble averages.

Functional enrichment analysis

For the pathway enrichment analysis, P values were obtained using an empirical phenotype-based permutation test, and pathways with P value < 0.05 were selected. The P values were corrected for multiple hypotheses using the Benjamini–Hochberg method. Pathways associated with the Gene Ontology biological process (GOBP) with adjusted P value < 0.05 were included. The GO gene sets were obtained using the msigdbr R package. We analyzed the genes around different peak signals identified via ATAC-seq and the genes on the boundary of A/B compartments through pathway enrichment analysis on GOBP.

We performed gene set enrichment analysis (GSEA) on GOBP gene sets using the GSEA function with the parameters “minGSSize = 5, maxGSSize = 1200, p value Cutoff = 0.05, pAdjustMethod = “BH”, and by = “fgsea””. GSEA of the DEGs identified via RNA-seq was performed with the clusterProfiler R package (v4.4.4)¹⁰⁴.

Nascent RNA capture

A STAR-Click Nascent RNA Capture Prep kit (Foreverstar Biotech, Beijing, China) was used to detect RNA synthesis in accordance with the manufacturer’s instructions. The rat VSMCs were transfected with the scrambled siRNA (20 nM) or siRNA targeting Lemd3 (20 nM). 24 hours after transfection, 1 mΜ of 5-ethynyl uridine (EU, an alkyne-modified nucleoside) was added, and the cells were subsequently incubated for another 2 h. After the incubation, total RNA or mRNA labeled with EU were isolated and used in a copper catalyzed click reaction with an azide-modified biotin, which created a biotin-based handle for capturing nascent RNA transcripts on streptavidin magnetic beads. The captured transcripts were used as a template for reverse transcriptase-mediated cDNA synthesis for subsequent analysis using qPCR.

Analysis of homologous genes in the rat and human genome

We found homologous genes in the human genome corresponding to rat genes from the Ensembl database and homologene R package. The R package is a wrapper for the homologene database by the National Center for Biotechnology Information (NCBI)¹⁰⁵. It allows searching for gene homologs across species. We use the python module biopython¹⁰⁶ to obtain the homologous gene sequences in the NCBI Entrez database, store them in fasta format respectively, and, via a step forward, use the blastn command line tool¹⁰⁷ to perform pairwise sequence alignment of the homologous sequences, with the parameter set as follows “-max_target_seqs 100 -word_size 11 -gapopen 5 -gapextend 2”. For each pair of homologous genes, we weighted the gene sequence identity of the sequence fragments aligned according to the sequence length.

Identity of single nucleotide polymorphisms (SNPs) associated with disease

We first downloaded the statistics of reported SNPs associated with coronary artery disease from the GWAS catalog¹⁰⁸. We mapped these SNPs to the human genome to find their homologs of LEMD3-modulated rat genes. In addition, we downloaded the computational results of the corresponding GWAS studies from the IEU OpenGWAS project¹⁰⁹ based on the PMID (PubMed ID: 29212778)⁶⁷ of the papers reporting these SNPs. These results were stored in vcf format, and we read the vcf format files by vcfR R package¹¹⁰ and used the CMplots R package¹¹¹ to visualize the SNPs near the COL4A1 and MYH11 gene loci. The SNPs are presented in the Supplementary Data 10.

Statistical analysis

The data are presented as the means ± standard errors of the means (s.e.m.) and were analyzed using GraphPad Prism 8.0 software (GraphPad Software, San Diego, CA, USA). The Shapiro‒Wilk normality test was used to determine whether the data were normally distributed. For two-group comparisons of normally distributed data, we applied two-sided unpaired Student’s t test for data with similar variances or with Welch’s correction if equal standard deviations were not assumed through an F test. In addition, for comparisons of more than two groups of normally distributed data, we applied ordinary one-way ANOVA with Tukey’s multiple comparisons test for data with similar variances or Brown–Forsythe and Welch’s ANOVA with Games–Howell’s multiple comparisons test if equal standard deviations were not assumed. The Mann‒Whitney U test was used for two groups or the Kruskal‒Wallis test with Dunn’s multiple comparisons test were applied for more than two groups of data that were not normally distributed. The detailed statistical analysis applied to each experiment is presented in the corresponding figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.