Ezh2 Shapes T Cell Plasticity to Drive Atherosclerosis

Human carotid artery plaques were collected from patients who underwent carotid endarterectomy. Written informed consent was given by each patient. The study protocol was approved by the regional ethical committee. After surgery, the biopsies were frozen or transferred to RNAlater and processed in the Munich Vascular Biobank.^29,30 Paraffin-embedded sections were subjected to immunohistochemical staining, as well as RNA extraction and gene expression analysis (Supplemental Material).

We retrieved single-cell gene expression data from the data sets generated as part of Alsaigh et al³¹ and Bashore et al,³² available through the Gene Expression Omnibus (accession codes GSE159677 and GSE253904, respectively). Pipeline for data analysis and visualization is described in the Supplemental Material.

Ezh2^fl/fl mice containing LoxP (locus of crossover P1)–flanked exons 14 and 15 of the Ezh2 gene were provided by Stuart H. Orkin, Boston Children’s Hospital.³³ To establish CD4⁺ and CD8⁺ cell-specific KO mice, Ezh2^fl/fl mice were bred with Cd4^cre (No. 017336; Jackson Laboratory)³⁴ or Cd8^cre mice (provided by Andreas Thiel, Charité Berlin).³⁵ These mice were backcrossed ≥10 times with Apoe^-/- mice (No. 002052; Jackson Laboratory) and genetic background was confirmed by genome-wide single nucleotide polymorphism analysis (Charles River). To induce atherosclerosis, mice were fed a Western-type diet (21% fat, 0.2% cholesterol; Ssniff) for 6 to 8 weeks. Mice were bred and housed in environmentally enriched cages with a 12-hour light/12-hour dark cycle at the animal facility at Ludwig Maximilians Universität München, according to institutional guidelines. To minimize confounders, animals were kept in standard conditions at all times. The experiments were performed and analyzed in a blinded fashion at all stages. The mice included in the experiments were monitored once per week according to a scoring system for animal disease and pain, which was approved by the local authorities. All animal experiments were approved by the local ethical committee for animal experimentation (TVA 55.2Vet-2532.Vet_02-17-180 and 02-22-150). The total number of mice used is indicated in Table S7.

Data are presented as mean±SD, unless noted differently. For data sets with n>6, Gaussian distribution was examined using the Shapiro-Wilk test. Comparisons between 2 groups were analyzed by Student t test or Mann-Whitney U test, when the assumption of normal distribution was violated. To reduce the risk of inflation of the α error, multiple (>3) comparisons were corrected with the false discovery rate approach described by Benjamini, Krieger, and Yekutieli, with a cutoff of 5%. Categoric variables were compared using χ² test or Fisher exact test, using GraphPad Prism v.10 software (GraphPad Software Inc, USA) and IBM SPSS Statistics v.29.0.2 (IBM Corporation, USA). Individual data points were excluded if identified as outliers by the ROUT (robust regression and outlier removal) test (Q=1%) or excluded for genotyping and technical issues (eg, damaged sections). A 2-sided P value <0.05 was considered statistically significant.

Analysis of RNA-sequencing data from human carotid atherosclerotic plaques in the Munich Vascular Biobank²⁹ revealed a 1.7-fold increase (P=2.0×10^-7) in EZH2 expression in advanced atherosclerotic plaques compared with early lesions (proximal adjacent tissue; Figure 1A). Additionally, quantitative polymerase chain reaction (qPCR) analysis revealed significantly higher EZH2 expression in unstable versus stable atherosclerotic plaques, classified according to established clinical and histomorphological criteria³⁰ (Figure S1A and S1B). To determine the cellular source of EZH2 within the human atherosclerotic plaque, we analyzed publicly available single-cell gene expression data (GSE253904) of 73 833 cells from 18 patients who underwent carotid endarterectomy.³² Using established markers, we could identify all major immune cell types, as well as endothelial and smooth muscle cell clusters (Figure S1C). As visualized in the uniform manifold approximation and projection (UMAP), EZH2 was primarily expressed in a distinct T cell subcluster, characterized by co-expression of proliferative genes (TOP2A [topoisomerase IIα], TYMS [thymidylate synthetase], MKI67 [marker of proliferation Ki-67], and PCLAF [PCNA clamp associated factor]), which we termed EZH2⁺ T cells, whereas lower expression was observed in other clusters, including Tregs, natural killer (NK) cells, dendritic cells, and macrophages (Figure 1B through 1E; Figure S1C). The prominent expression of EZH2 in T cells compared with other cell types was confirmed in an independent scRNA-seq data set of human atherosclerotic plaques (GSE159677),³¹ validating our findings (Figure S1D and S1E). These findings suggest that T cell–derived EZH2 contributes to atherosclerosis development and progression, warranting further investigations into its mechanistic relevance.

Because T cells are the major source of EZH2 in human plaques, we studied T cell–specific Ezh2 deficiency in atherosclerosis-prone Apoe^-/- mice (Ezh2^cd4-KO) and respective littermate controls (Ezh2^cd4-WT [wild type]). Splenic T cells of atherosclerotic Ezh2^cd4-KO mice showed a reduction of Ezh2 at both the transcript (−72%) and protein (−90%) levels, as assessed by qPCR and immunofluorescence, confirming the specificity of the model (Figure S2A through S2C). Western blot and mass spectrometry–based analysis of histones isolated from CD4⁺ T cells revealed a significant decrease in H3K27 di- and trimethylation in Ezh2^cd4-KO mice (Figure S2C and S2D) compared with Ezh2^cd4-WT littermates. Other components of the PRC2, such as Ezh1, Suz12, and Jarid2, as well as the demethylases Utx (Kdm6a) and Jmjd3 (Kdm6b) remained unaffected in CD4⁺ T cells (Figure S2A), suggesting no counter-regulation related to Ezh2 deficiency. Body weight, basic hematological parameters, and plasma lipid profiles were comparable between Ezh2^cd4-KO mice and littermates (Table S1).

In the absence of T cell Ezh2, atherosclerotic plaque area was significantly decreased in aortic roots and arches of both female and male mice (Figure 2A; Figure S3A and S3B), compared with their Ezh2^cd4-WT littermates. Ezh2^cd4-KO mice had less advanced plaques (defined as fibrous cap atheroma) and a larger percentage of initial plaques (defined as intimal xanthoma and pathological intimal thickening),³⁶ suggesting that T cell Ezh2 deficiency slowed plaque progression (Figure 2B). These findings were also reflected by changes in plaque composition: plaques of Ezh2^cd4-KO mice had decreased collagen and macrophage content (Figure 2C and 2D) and no difference in α-SMA⁺ (α-smooth muscle actin) cell content but a trend toward fewer CD4⁺ T cells was observed (Figure 2E and 2F). The amount of lesional Foxp3⁺ (forkhead box protein 3) Tregs was strongly reduced in Ezh2^cd4-KO mice (Figure 2G). Collectively, these data indicate that Ezh2 abrogation in T cells reduces atherosclerotic plaque size in both sexes, accompanied by altered plaque composition despite reduced Tregs.

To further elucidate the effects of T cell Ezh2 deficiency on the immune system, we analyzed the amount and composition of major immune cell (sub)populations in blood and lymphoid organs. Ezh2 deficiency solely affected the number of CD3⁺ T cells, whereas B cell, monocyte, and neutrophil counts remained unchanged in blood (Figure 3A). In-depth analysis of the T cell compartments in blood, lymph nodes, and spleen demonstrated that CD4⁺ T cells were predominantly affected: Ezh2^cd4-KO mice exhibited a decrease in CD4⁺ T cells, whereas CD8⁺ T cells were mainly unaltered (Figure 3B; Figure S3C). To further detail subpopulation differences in T cells, we employed scRNA-seq of splenic CD3⁺ cells and identified 11 clusters of T cells (Figure 3C), characterized by using common expression markers (Figure S4A). Ezh2^cd4-KO mice showed reduced percentages of naive CD4⁺ and CD8⁺ T cell, Ccl5⁺ cells and Tregs, but increased percentages of CD4⁺ and CD8⁺ memory T cells, and especially iNKT cells (Figure 3D). These results were validated by flow cytometry: analysis of splenic CD4⁺ T cells from Ezh2^cd4-KO mice showed a shift from naive to effector memory T cells and a reduction in Tregs (Figure 3E; Figure S3D). Likewise, splenic CD8⁺ T cells displayed a reduced naive fraction and a prominent central memory population (Figure 3D and 3F; Figure S3E). To better detail the profound increase in CD4⁺ effector T cells, we analyzed subpopulations of effector memory cells using inflammatory chemokine markers, Cxcr3 (C-X-C motif chemokine receptor 3) and Ccr6 (C-C motif chemokine receptor 6), which are preferentially expressed on Th1 (type 1 T helper) and Th17 (type 17 T helper) Acells, respectively.^37–40 Accordingly, we observed a shift from Th1 (Cxcr3⁺Ccr6^-) to Th2 (Cxcr3^-Ccr6^-) cells in Ezh2^cd4-KO mice (Figure 4A; Figures S3F and S7A). In line, we observed a trend toward elevated concentrations of the Th2-associated cytokines Il-4 and Il-13 in the plasma of Ezh2^cd4-KO mice (Figure 4B). To confirm CD4⁺ T cells as a cellular source of these cytokines, we stimulated CD4⁺ T cells in vitro and subsequently measured cytokine release. The supernatant analysis demonstrated CD4⁺ T cells from Ezh2^cd4-KO mice secreted 39× more Il-4 and 8× more Il-13 compared with their Ezh2^cd4-WT counterparts, with no change observed in the Th1-associated cytokine Ifn-γ ([interferon γ] Figure 4C). Remarkably, we further observed a 17-fold increase of Il-4 transcripts in the descending aorta of Ezh2^cd4-KO mice, whereas Ifn-γ and Il-13 remained unchanged (Figure 4D). These data suggest that deficiency of Ezh2 in T cells favors the activation of systemic and local type 2 immune responses, possibly underlying the reduction in atherosclerosis.

To unveil the relevance of the changes in CD4⁺ T and CD8⁺ T cells for atherosclerosis upon T cell Ezh2 deficiency, a second animal model with a CD8-specific Ezh2 deficiency (Ezh2^cd8-WT and Ezh2^cd8-KO) was employed. CD8⁺ T cells displayed a 60% reduction in Ezh2 but not Ezh1 transcripts (Figure S5A). Using Western blot analysis, we confirmed Ezh2 deficiency in CD8⁺ T cells (Figure S5B). No significant differences were observed in basic hematological parameters or cholesterol levels of Ezh2^cd8-KO mice and their WT littermates (Table S2). Plaque burden and development were unaltered in Ezh2^cd8-KO mice fed a Western-type diet for 8 weeks compared to littermate controls (Figure S5C and S5D). Major immune cell populations in blood and lymphoid organs were unaffected by Ezh2 deficiency in CD8⁺ T cells (Figure 5E). Moreover, proportions of CD4⁺ and CD8⁺ T cells in blood, lymph nodes, and spleen were similar in Ezh2^cd8-KO mice and littermate controls (Figure S5F). Considering the profound effects on T cell populations observed in Ezh2^cd4-KO mice, we recapitulated the in-depth analysis of the T cell compartments. Splenic CD4⁺ T cells from Ezh2^cd8-KO mice were mainly unaffected (Figure S5G), whereas splenic CD8⁺ T cells from Ezh2^cd8-KO mice showed slight reductions in naive and effector memory subsets along with an increase in central memory T cells (Figure S5H). No differences were observed in Tregs (Figure S5G). Notably, plasma levels of type 2 cytokines (ie, Il-4 and Il-13) were unchanged in Ezh2^cd8-KO mice (Figure S5I). Together, these data suggest that the CD8⁺ T cell phenotype observed in our Ezh2^cd4-KO model is not intrinsically related to Ezh2 deficiency in CD8⁺ T cells, but likely results from the effects of Ezh2 deficiency in CD4⁺ T cells. The type 2 immune response, and specifically the Il-4 environment induced by Ezh2 deficiency in CD4⁺ T cells, likely affects the phenotype and functionality of CD8⁺ cells by skewing them toward (innate) memory traits.

Because T cells secrete cytokines with a profound effect on neighboring cells, we investigated whether T cells from Ezh2^cd4-KO mice would regulate the polarization of surrounding macrophages. Indeed, when cultured with the supernatant from stimulated CD4⁺ T cells from Ezh2^cd4-KO mice, bone marrow–derived macrophages were polarized toward an anti-inflammatory phenotype, made evident via increased macrophage Arg1 (arginase 1) mRNA and protein expression (Figure 4E and 4F). Additionally, inducible nitric oxide synthase (iNos), a key marker of inflammatory macrophages, was reduced after stimulation with Ezh2^cd4-KO T cell supernatant (Figure 4G and 4H). In vivo, we found fewer iNos⁺ macrophages in plaques of Ezh2^cd4-KO mice (Figure 4I). These data together with the elevated aortic Il-4 cytokine transcripts suggest that Ezh2 deficient CD4⁺ T cells may induce an advantageous, anti-inflammatory plaque macrophage phenotype, likely contributing to a reduction of atherosclerosis.

To unravel mechanisms driving the identified effects on immune responses, we performed RNA sequencing on splenic CD4⁺ T cells isolated from Ezh2^cd4-KO mice and Ezh2^cd4-WT controls using an efficient sequencing method (ie, Prime-seq).⁴¹ Of 23 160 detected genes, we observed differential expression for 435 transcripts for a false discovery rate <5% (Figure 5A). Notably, the 187 upregulated genes in our transcriptomic analysis overlapped with those found to be regulated by H3K27me3 marks detected by chromatin-immunoprecipitation (ChIP) and sequencing in multiple cell types, including lymphoid subtypes (ENCODE [Encyclopedia of DNA Elements] Histone Modification 2015; Figure 5B), thus corroborating that transcriptional changes are likely to reflect lower H3K27me3 repressive marks upon Ezh2 deletion. Beyond confirming the increase in Il-4 and other genes involved in type 2 immune response (Gene Ontology [GO]: 0042092; P=3.3×10^-6), we identified an enrichment (factor >1.5) in pathways and processes relevant to lymphocyte differentiation and lymphocyte-mediated immunity. Surprisingly, pathways related to NK cell–mediated toxicity were also enriched in our analyses (Figure 5C). These findings show that T cells, in which Ezh2 is deleted, may possess unique features, and share traits resembling NK cells. We refined our analysis by delineating a network of regulated genes encoding for proteins biophysically interacting with ≥2 other members. Two densely interconnected clusters were identified by MCODE (Molecular Complex Detection) algorithm and GO analysis on the network confirmed the significantly higher representation of terms related to NK cell–mediated toxicity and lymphocyte-mediated immunity (Figure 5D).

The prevalent enrichment of lymphocyte differentiation and NK cell pathways prompted an unbiased analysis of our results in the context of cell-specific transcriptional signatures. The most significant overlap of upregulated genes with cell identity in most murine atlases was NK cells rather than conventional T cell subtypes (Figure 5E). Notably, the differentially expressed genes overlapped with the variable genes in a scRNA-seq data set ([GSE152786] odds ratio, 14.3; P=2.4 × −10^-13) of thymic iNKT cells.⁴² This distinct T cell subset shows phenotypic and functional properties similar to NK cells and differentiates in the thymus under specific transcriptional programs driven by sequential transcription factors.^43–45 Hence, we analyzed the overlap between the transcription factors dictating iNKT development and the 29 transcription factors differentially regulated in our transcriptomic analysis, and found Zbtb16 (zinc finger and BTB domain containing 16), the gene coding for Plzf (promyelocytic leukemia zinc finger), to be strongly upregulated in Ezh2-deficient CD4⁺ T cells (Figure 5F). To assess the direct role of Ezh2 and the H3K27me3 mark in gene repression in T cells, we performed H3K27me3 ChIP on splenic CD3⁺ T cells isolated from both Ezh2^cd4-KO and Ezh2^cd4-WT mice, followed by qPCR. This analysis revealed a significant enrichment of H3K27me3 at the transcription starting site of the Zbtb16 gene in Ezh2^cd4-WT mice, which was significantly diminished in Ezh2^cd4-KO mice (Figure 5G). A similar pattern was observed for the Il-4 gene (Figure 5G), reinforcing evidence for a lack of transcriptional repression of Il-4 and Zbtb16 genes in the absence of H3K27me3 in Ezh2^cd4-KO mice. Finally, we detected significant H3K27me3 enrichment at the regulatory promoter region of the ZBTB16 gene in publicly available human ChIP-seq data (GSE18927 and GSE187334; Figure 5H), thus translating our findings on this EZH2-dependent regulatory pathway to humans. Together, our data suggest that EZH2 represses the transcriptional program, leading to iNKT differentiation by epigenetically repressing PLZF.

Our scRNA-seq data showed a strong increase in the fraction of iNKT cells in Ezh2^cd4-KO mice (Figure 3D), which was validated by flow cytometry demonstrating a 3.8-fold increase in the iNKT cell proportion in the spleen of Ezh2^cd4-KO mice (Figure 6A). iNKT cells are CD1d-restricted innate-like adaptive lymphocytes capable of responding quickly (ie, minutes to hours) on antigenic lipid stimulation.⁴⁶ Analogous to conventional T cells, iNKT cells differentiate into distinct subsets, namely iNKT1, iNKT2, and iNKT17 cells.⁴⁷ More detailed analysis revealed that splenic iNKT cells of Ezh2^cd4-KO strongly express Plzf, the hallmark of iNKT2 cells (Figure 6B). Further phenotyping confirmed a striking shift from iNKT1 cells in Ezh2^cd4-WT mice toward Plzf-expressing iNKT2 cells in Ezh2^cd4-KO mice (Figure 6C; Figure S6A). This suggests that in the absence of T cell–specific Ezh2, iNKT cells are not only vastly expanded, but also preferentially differentiate toward an iNKT2 phenotype. Likewise, the expression of Zbtb16 was significantly upregulated in atherosclerotic aortas of Ezh2^cd4-KO mice (Figure 6D), indicating that Plzf-expressing iNKT2 cells accumulate in atherosclerotic plaques on Ezh2 deletion in T cells. This finding aligns with the 2.0-fold decrease of ZBTB16 expression in human advanced versus early lesions (P=2.1×10^-6) in the Munich Vascular Biobank with a high-grade negative bivariate correlation with EZH2 expression (Spearman ρ=−0.750; P=7.2×10^-38; Figure 6E and 6F), suggesting that the atheroprotective role of ZBTB16 and iNKT2 cells is translatable to humans. Notably, iNKT populations and Plzf expression in the spleen of Ezh2^cd8-KO mice were similar to their WT littermates (Figure S5J), supporting our finding that the effects observed are attributable to Ezh2 deficiency in CD4⁺—and not CD8⁺—T cells.

Similar to conventional T cell populations, iNKT cells originate from CD4⁺/CD8⁺ double-positive precursors in the thymus. In contrast to conventional T cells that differentiate in the periphery, iNKT cells differentiate in the thymus into functionally distinct iNKT1, iNKT2, or iNKT17 subsets.⁴⁸ Contrary to the significant decreases in T cell numbers in the circulation and secondary lymphoid organs of Ezh2^cd4-KO mice (Figure 3A and 3B), we observed a massive accumulation of CD3⁺ T cells in the thymus of Ezh2^cd4-KO mice (Figure 7A), which could be ascribed to differences in the thymic development of T cells. We observed a reduction in the proportion of double-positive cells (Figure 7B), along with an accumulation of double-negative 1 and 4 populations (Figure 7C), leading to a profound decrease in thymic CD4⁺ T cells in Ezh2^cd4-KO mice, whereas thymic CD8⁺ T cells were only slightly increased (Figure 7D). Foxp3 expression was barely detectable in the thymus of Ezh2^cd4-KO mice (Figure S6D), indicating that the profound reduction of systemic and lesional Tregs can be attributed to the disturbed T cell development in the thymus. To unravel the source of the type 2 immune response in Ezh2^cd4-KO mice, we examined the cytokine milieu in the thymus. In parallel to the periphery, we observed a profound increase in Il-4 transcripts and iNKT cells in the thymus of Ezh2^cd4-KO mice (Figure 7E and 7F). Corroborating our findings in the spleen, Ezh2 deficiency in CD4⁺ T cells skewed the iNKT cells toward an iNKT2 phenotype (Figure 7G; Figure S6B), which likely explains the increase in Il-4 in the thymus.

Overall, our data suggest that the increased abundance of iNKT2 cells in the thymus of Ezh2^cd4-KO mice is the source of Il-4 and a driving force for the type 2 immune response and subsequent CD4⁺ and CD8⁺ T cell polarization in the periphery and atherosclerotic aorta.

To discriminate between the canonical and potential noncanonical, chromatin-independent role of EZH2 in the context of atherosclerosis, we investigated the cellular localization of EZH2 in T cells in atherosclerotic lesions using immunofluorescence staining. In both murine and human plaque specimens, we mainly observed the presence of EZH2 in the nucleus. However, we could confirm weak expression of EZH2 in the cytoplasm of lesional T cells (Figure 8A and 8B). These findings suggest mainly canonical (chromatin-dependent), but also noncanonical activity of EZH2 in T cells in human and mouse atherosclerotic plaques.

Previous studies found that Ezh2 controls actin polymerization,⁴⁹ cell adhesion, and migration through direct methylation of extranuclear substrates.²⁸ In support of these reports, we could demonstrate a profound reduction in actin filaments in CD4⁺ T cells isolated from Ezh2^cd4-KO mice (Figure 8C). Moreover, we observed a strong impairment in the migratory capacity of these Ezh2-deficient CD4⁺ T cells toward the chemokines Ccl19 and Ccl22, compared with CD4⁺ T cells from Ezh2^cd4-WT mice (Figure 8D). Altogether, our data suggest an additional noncanonical, chromatin-independent function of Ezh2 in T cells, likely contributing to attenuated atherosclerosis in Ezh2^cd4-KO mice.