Atherosclerosis Profiling Reveals BHLHE40 as a Candidate Modulator of VSMC.
May 13, 2026DOI: 10.1161/circresaha.125.326821Journal ArticleAHA
Written by: Chinyere O IbikunleEnrique J GarciaChenyi XueEunyoung KimHanying YanJohana CoronelLucie Y ZhuJian CuiAllen ChungLauren E FriesNadja SachsRobert C BauerLars MaegdefesselMingyao LiAlan R TallAlexander C BashoreMuredach P Reilly
Introduction
Atherosclerosis is a chronic inflammatory disease characterized by a heterogeneous cellular composition, including vascular smooth muscle cells (VSMCs), fibrotic cells, immune cells, and extracellular matrix components1. VSMCs can comprise 50% or more of the cellular content within atherosclerotic plaques2, highlighting their role in the initiation and progression of the disease. A hallmark of atherosclerotic development is the phenotypic modulation of VSMCs. In healthy arteries, VSMCs reside in the medial layer and exhibit a contractile phenotype, defined by the expression of smooth muscle α-actin (ACTA2), myosin heavy chain (MYH11), and other proteins that facilitate vascular tone and contractility3. During atherogenesis, these cells migrate from the media to the intima, undergoing a phenotypic switch to a synthetic state, characterized by downregulation of contractile proteins, increased proliferative and migratory capacity, and enhanced production of extracellular matrix components4. While this phenotypic plasticity is essential for vascular repair, it also significantly contributes to the pathogenesis of atherosclerosis.The factors driving VSMC phenotypic switching are multifaceted, involving a complex interplay of genetic, epigenetic, and environmental stimuli. Growth factors, cytokines, and mechanical forces are among the myriad factors implicated in VSMC phenotypic modulation5. For instance, platelet-derived growth factor (PDGF), secreted by endothelial cells, macrophages, and VSMCs themselves, is a potent mitogen for VSMCs, driving their proliferation and migration6. The implications of VSMC phenotypic modulation in atherosclerosis are profound. Synthetic VSMCs contribute to plaque stability by synthesizing and depositing collagen in the fibrous cap2. Conversely, their proliferation can contribute to intimal thickening and luminal narrowing, while their production of extracellular matrix-degrading enzymes can precipitate plaque rupture, leading to acute vascular events7.Recent advances in lineage-tracing techniques coupled with single-cell RNA sequencing (scRNA-seq) technologies have revealed substantial heterogeneity of VSMC phenotypes within atherosclerotic plaques, challenging the binary classification of ‘contractile’ and ‘synthetic’ phenotypes8–10. Several studies suggest that VSMCs can transdifferentiate into cells expressing macrophage-associated markers, contributing to the inflammatory environment within plaques2,10,11. These insights underscore the nuanced roles VSMCs play in atherosclerosis, from fostering plaque stability to promoting lesion instability. The concept of VSMC-derived macrophage-like cells for example challenges the conventional dogma of distinct cellular contributions to atherosclerosis, suggesting a more dynamic interplay of VSMCs in disease progression. However, reported frequencies of VSMC-derived macrophage-like cells vary widely, ranging from rare to abundant, likely reflecting differences in experimental design, including composition and diet selection, mouse models, and cell isolation and sequencing protocols. Moreover, while lineage-tracing studies have demonstrated co-expression of VSMC and macrophage markers, such analyses may overestimate macrophage-like differentiation due to partial recombination, antibody cross-reactivity, or ambiguous thresholds for lineage marker detection. While debate remains regarding the extent to which VSMCs transdifferentiate into bona fide macrophage-like cells, it is well-established that many lesional foam cells are derived from VSMCs. Greater precision in distinguishing VSMC-derived foam cells from macrophage-like cells is therefore essential for clarifying their distinct roles and regulation in atherosclerosis.
In the present study, we performed a comprehensive and novel analysis of the phenotype heterogeneity of VSMCs in atherosclerosis by leveraging the power of cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) to simultaneously profile the transcriptome and cell surface phenotype of cells within lesions. To ensure rigorous identification of cells of VSMC origin, we applied quantitative thresholds to the transcript expression of the fluorescent lineage tag, defining an optimal cutoff that minimized false-positive assignment. Our findings reveal that the conversion of VSMCs into mature macrophage-like cells is uncommon. Instead, most VSMCs transition into two modulated VSMCs that resemble fibroblast-like cells, and a smaller proliferative subset that emerges as atherosclerosis advances. Notably, we confirm that a majority of foam cells within lesions originate from VSMCs, and identify a novel transcription factor regulator, BHLHE40, of VSMC phenotype modulation and foam cell formation.
Methods
A comprehensive description of all methods and materials used in this research can be found in the Supplemental Materials.
All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Columbia University under protocol number AABQ5576. All Mouse strains including Ldlr−/− (The Jackson Laboratory, 002207), ROSA26LSL-ZsGreen1+/− (The Jackson Laboratory, 007906), and Myh11-CreERT2 (The Jackson Laboratory, 019079) were purchased and bred to generate a hyperlipidemic smooth muscle cell lineage tracing strain: Ldlr−/− Myh11-CreERT2 ROSA26 LSL-ZsGreen1+/− 10. Only male mice were used since the Myh11-CreERT2 was integrated into the Y chromosome. To confer atherosclerosis, the mice were fed a Western Diet (WD) for the specified durations (Research Diet: D12079Bi). C57BL/6J (The Jackson Laboratory, 000664) were fed a regular chow diet. Mice were housed under a 12-hour light/12-hour dark cycle with ad libitum access to their diet and water. Prior to tissue collection, mice were sedated with 5% isoflurane and euthanized by cervical dislocation as specified in our IACUC protocol.
Results
The prevalence of VSMC-derived macrophage-like cells in atherosclerosis remains controversial8–10,15,18. To comprehensively investigate the cellular heterogeneity underlying atherosclerosis development and progression, we analyzed a dataset we previously generated that integrates CITE-seq lesion profiling with a specialized mouse model for VSMC lineage tracing10,14,54. In this model, tamoxifen induction leads to the expression of the fluorescent reporter ZsGreen1 in all VSMCs that express Myh11, and any subsequent phenotypically modulated progeny, and mice were fed a Western Diet (WD) for 0, 8, 16, and 26 weeks. At each time point, aortas were enzymatically digested, and CITE-seq was performed using an antibody cocktail targeting 119 cell surface proteins. Cells expressing ZsGreen1 (ZsGreen1+) and those lacking ZsGreen1 (ZsGreen1−) were separated via Fluorescence-Activated Cell Sorting (FACS), validated by ZsGreen1 transcript expression, and analyzed at the single-cell level (Fig. 1A).
Clustering of the combined data revealed 23 distinct cell populations (Fig. 1B). This included macrophage subtypes, T cells, neutrophils, proliferative cells, endothelial cells, contractile VSMCs, modulated VSMCs, chondromyocytes, fibroblasts, and myofibroblasts as confirmed by a comprehensive evaluation of DE genes (Extended Data Set 1) and proteins (Extended Data Set 2). Visualization of cell clusters based on ZsGreen1 expression confirmed that multiple cell populations, including VSMC1, VSMC2, VSMC3, modulated VSMC, proliferative VSMC, and chondromyocytes were almost completely of VSMC origin (Fig. 1C). Myofibroblasts and chondromyocytes were initially clustered together (Suppl Fig. 1A) but were distinguished by VSMC lineage tracing and origin, with only chondromyocytes being derived from VSMCs. Importantly, very few cells clustered with macrophages that were of VSMC origin.
Given potential for technical errors and biases in single-cell flow cytometry studies18, we used a transcript encoding the fluorescent reporter protein to perform orthogonal validation of our FACS-based separation of VSMC-derived and non-VSMC cells. We created a tailored reference genome that included the ZsGreen1-WPRE sequence from the Ai6 reporter (Suppl Fig. 1B). To minimize autofluorescence artifact and background noise, we established a threshold for ZsGreen1 transcript expression utilizing VSMCs at the 0-week time point (Suppl Fig. 1C). We categorized all cell types as either VSMC or non-VSMC derived, employing both FACS-based separation and ZsGreen1 transcript threshold demonstrating high consistency between both approaches (Fig. 1D and 1E). Using this new rigorous and validated FACS protocol, we quantified the proportion of VSMC-derived macrophage-like cells at 26 weeks of WD and found that only 3.62% of macrophages were ZsGreen1+ (Fig. 1F). Among all macrophages, ZsGreen1+ VSMC-derived macrophage-like cells were identified to be 1-3% in the FACS-defined analysis, compared to approximately 2% using the ZsGreen1 transcript threshold. Furthermore, the FACS analysis may underestimate the proportion of ZsGreen1+ fibroblasts, identifying them at around 1%, whereas the ZsGreen1 transcript analysis indicated a higher proportion of approximately 5% (Fig. 1G).
To investigate published conflicting findings, we re-examined our prior data (Pan et al.10) using the new ZsGreen1 transcript threshold approach. Clustering of all cell types revealed three macrophage subsets, two fibroblast subsets, two contractile VSMCs, two modulated VSMCs, and myofibroblasts (Suppl Fig. 1D), consistent with reported literature. Reanalyzing the dataset from Pan et al.10 revealed a significant mismatch between the original FACS (Suppl Fig. 1E) and new ZsGreen1 transcript (Suppl Fig. 1F) approaches. There were substantially fewer transcript-defined ZsGreen1+ macrophages and, to a lesser extent, fibroblasts compared to those identified by the original FACS protocol. The proportion of ZsGreen1+ macrophages decreased markedly from 30-60% to 1-2%. A similar but less marked trend was noted for fibroblasts, from 20-30% to approximately 5% ZsGreen1+ (Suppl Fig. 1G). Quantification of other cell populations revealed no substantial differences in ZsGreen1+ proportions among contractile VSMCs, modulated VSMCs, and myofibroblasts between methods. To assess whether our newly deployed FACS gating strategy was overly stringent, we sorted ZsGreen1+, ZsGreen1−, and ZsGreen1int (intermediate) populations and performed scRNA-seq (Suppl Fig. 1H). This analysis revealed that the ZsGreen1int population consisted of a mixture of ZsGreen1+ and ZsGreen1− cells, but was predominately ZsGreen1−, indicating that our gating of ZsGreen1+ was stringent to minimize misclassification of ZsGreen1− cells as VSMC-derived (Suppl Fig. 1I and 1J). Thus, the discrepancy between our present study and the previous report is likely due to differences in FACS gating strategies, which in the prior report may have inadvertently classified autofluorescent ZsGreen1− or ZsGreen1int cells, especially macrophages, as VSMC lineage-positive. Overall, analyses of both datasets suggest that macrophage-like cells are estimated to be approximately 2% of VSMC origin, while fibroblast-like cells appear to be around 5% of VSMC origin.
To further validate our finding of rare VSMC-derived macrophage-like cells and to localize them in lesions, we performed dual RNAscope and immunofluorescence staining on lesions from 16-weeks (Fig. 1H) and 26-weeks WD fed mice (Fig. 1I). At both time points, we observed little overlap between CD68 (protein), a canonical macrophage marker, and ZsGreen1 (RNA), marked VSMC-derived cells. Quantification of data across multiple lesions showed that approximately 2.7% of cells at 16 weeks WD and 4% at 26 weeks WD were CD68+ZsGreen1+ (Fig. 1J). These findings confirm that VSMC-derived macrophage-like cells are present but represent a very rare population in advanced mouse atherosclerosis.
Transcriptomic data (Fig. 2A) distinguished cell types based on established canonical gene markers. In parallel, we leveraged CITE-seq dataset to identify the most highly expressed surface proteins across cell types (Fig. 2B; Extended Data Set 2 and 3). While immune and endothelial cells were defined using established immunophenotypic markers, VSMC immunophenotyping has been limited. Here, we identified several promising novel surface proteins uniquely expressed by VSMCs and modulated VSMCs. Consistent with our previous publication14, CD200 was highly expressed across all VSMC-derived populations, while CD49b emerged as a specific marker of modulated and proliferative VSMCs. Immunofluorescence in advanced atherosclerotic lesions confirmed colocalization of ZsGreen1 and CD49b predominantly in the intima (Fig. 2C). Quantification of lesions at 16 and 26 weeks WD showed that at least one-fifth of ZsGreen1+ cells were ZsGreen1+CD49b+ modulated VSMC at both time points (Fig. 2D). These findings highlight the utility of high-dimensional surface phenotyping to identify and isolate distinct lesion cell populations.
As expected, the cellular composition of the aorta shifted markedly with disease progression with emergence of multiple VSMC-derived cell states (Fig. 2E). There was marked expansion of ZsGreen1+ modulated VSMCs, proliferative VSMC-derived cells, and chondromyocytes, accompanied by a decline in contractile VSMC 1 and 2 (Fig. 2F). Resident macrophages decreased, whereas foamy macrophages expanded as lesions progressed (Fig. 2G). Most fibroblast subpopulations exhibited a modest decline over time (Fig. 2H). In addition, ZsGreen1+ chondromyocytes increased, while ZsGreen1− myofibroblasts declined (Fig. 2I). These findings highlight the cellular heterogeneity, expansion of VSMC-derived cell types and VSMC phenotypic modulation in the evolving lesion environment.
To resolve finer cell sub-phenotypes13,18,55,56 during atherosclerosis progression, we performed a deeper re-clustering of all macrophage subtypes (Suppl Fig. 2) and all ZsGreen1+ VSMC-lineage cells (Fig. 3) identified in the initial clustering in Fig. 1.
Re-clustering of macrophages revealed ten populations, including five macrophage subsets and another five subsets consisting of monocytes, dendritic cells, and a small number of fibroblasts clustering with macrophages in our initial analysis (Suppl Fig. 2A). We identified all macrophage subtypes previously reported in a recent scRNA-seq meta-analysis of leukocytes in atherosclerosis57 (Suppl Fig. 2B) except for the interferon-inducible (IFNIC) macrophage population. Consistent with our initial analysis (Fig. 1), each macrophage subpopulation contained very few cells that were of VSMC origin, even as disease progressed (Suppl Fig. 2C and 2D). Both canonical macrophage genes (Suppl Fig. 2E) and proteins (Suppl Fig. 2F) were detected in ZsGreen1+ VSMC-derived macrophage-like cells, albeit at slightly lower levels, especially for cavity macrophages, compared to ZsGreen1− myeloid-derived macrophages. Because the number of ZsGreen1+ macrophage-like cells within each cluster was minimal and not restricted to a specific subtype, these findings require cautious interpretation.
Deep re-clustering of ZsGreen1+ VSMC-lineage cells revealed ten subpopulations (Fig. 3A) including five contractile VSMC subpopulations, two modulated VSMC subpopulations, chondromyocytes, a small cluster of macrophage-like VSMCs, and proliferative VSMCs. Although relatively small populations, the macrophage-like and proliferative VSMCs were readily apparent. We compared our findings to a recent integrative analysis of VSMC-lineage-traced datasets in atherosclerosis18 (Fig. 3B). DE genes from the integrative analysis highlighted both shared and unique features relative to our data. Modulated VSMC 1 and the proliferative subpopulation were enriched in genes associated with fibrochondrocytes, immune responses, and synthetic extracellular matrix-secreting VSMCs. However, proliferative VSMCs were uniquely characterized by high expression of cell cycle-related genes. Interestingly, our chondromyocyte population lacked canonical marker genes associated with chondromyocytes in the integrated analysis. Macrophage-like VSMCs expressed Igf1, a marker previously linked to TNNT2+ (Cardiac muscle troponin T) VSMCs, but no other additional marker was uniquely associated with a specific VSMC subset in our dataset. We did not identify any VSMC subpopulation with a distinct interferon-related gene signature.
DE gene analysis revealed distinct transcriptional signatures among the VSMC-derived subtypes (Fig. 3C). Although modulated VSMC 1 and modulated VSMC 2 shared several overlapping features, Modulated VSMC 1 showed elevated expression of Vcam1, Lum, and Dcn genes commonly associated with phenotypic modulation. In contrast, Modulated VSMC 2 was enriched for Col14a1, suggesting a specialized role in extracellular matrix (ECM) remodeling. The Proliferative VSMC subtype was defined by high expression of cell cycle–regulating genes, including Ube2c, Birc5, and Top2a, indicating active proliferation. We also identified a rare VSMC-derived macrophage-like population, characterized by expression of leukocyte and macrophage-associated genes such as Cd48 and Adgre1 (Fig. 3C), as well as surface proteins CD11b, CD45, and CD48 (Fig. 3D). Compared to authentic blood-derived or resident macrophages (ZsGreen1−), these ZsGreen1+ macrophage-like cells retained low-level expression of contractile VSMC genes and exhibited slightly reduced Adgre1 expression, although protein levels of CD11b, CD45, and CD48 were comparable (Fig. 3C and 3D).
CITE-seq protein analysis (Fig. 3D) confirmed that all VSMC subpopulations expressed CD200, confirming its role as a lineage marker for phenotypically modulated VSMCs14. The two modulated VSMC subpopulations were further distinguished by their expression of CD54, a transmembrane glycoprotein involved in cellular adhesion58. Meanwhile, the proliferative VSMC population was marked by Ly6A/E, a protein traditionally associated with stem cells, which we previously identified in a VSMC-derived subpopulation and labeled as intermediate SEM (stem cell, endothelial cell, monocyte) cells10.
To explore functional differences among the modulated VSMC 1, modulated VSMC 2, and proliferative VSMC, we performed Ingenuity Pathway Analysis (IPA) with their top DE genes (Fig. 3E). This revealed a substantial overlap among these subpopulations, with enrichment in inferred functions related to collagen metabolism, fibrosis, and extracellular matrix organization. The modulated VSMC 2 subpopulation showed enrichment in the binding and uptake of ligands by scavenger receptors, implying a role in lipid uptake and metabolism. Consistent with our initial clustering (Fig. 2), we observed expansion of the modulated VSMC 1 subpopulation, as well as the emergence of proliferative VSMC during disease progression at 16 weeks and 26 weeks of WD feeding (Fig. 3F and 3G).
Given the evidence that VSMCs can uptake lipids during atherosclerosis and adopt a foam cell-like state24,59,60, we investigated the genomic features of VSMC-derived foam cells through bulk RNA-seq on LipidTOX-stained aortic tissues from Ldlr−/−ZsGreen1+/−Myh11-CreERT2 mice after 16 and 26 weeks of WD. We developed a flow cytometry sorting strategy to distinguish between foam (LipidTOX+) and non-foam (LipidTOX−) VSMCs (ZsGreen1+) and non-VSMCs (ZsGreen1−) (Fig. 4A). Consistent with existing literature24, our analysis showed that at 0 weeks WD, VSMC-derived foam cells were virtually absent. However, by 26 weeks, over one-third of the VSMCs exhibited significant lipid uptake. At 16 and 26 weeks of WD feeding, more than 75% of foam cells originated from VSMCs (Fig. 4B).
Despite similar cellular levels of lipid accumulation (Fig. 4C), DE gene analysis revealed that VSMC-derived (ZsGreen1+) foam cells exhibit a markedly distinct transcriptional profile to non-VSMC (ZsGreen1−) at both 16- (Fig. 4D) and 26-weeks of WD (Fig. 4E), particularly in VSMC and macrophage marker expression. The non-VSMC foam cells had high expression of classical macrophage (Cd68 and Csfr1) and foam cell markers (Cd36, Fabp5, Trem2). In contrast, VSMC-derived foam cells exhibited comparatively lower levels of expression of these genes and retained expression of Myh11, Acta2, and Tagln (Fig. 4F and 4G). Unbiased analyses of gene expression in foam and non-foam ZsGreen1+ and ZsGreen1− populations at 16 and 26 weeks (Fig. 4F and 4G) showed that VSMC-derived foam cells maintained a transcriptional signature distinct from macrophages, characterized by persistent expression of VSMC-related genes and limited activation of macrophage-associated programs even at advanced stages of atherosclerosis. In combination with our CITE-seq analyses, these findings indicate that while many VSMC-derived cells can uptake and accumulate lipids to adopt a foamy phenotype, only a small fraction of VSMC-derived cells fully adopt a macrophage-like genomic and proteomic profile.
Relative to VSMC-derived non-foam cells, VSMC-derived foam cells exhibited an upregulation of genes involved in lipid metabolism (Abca1, Abcg1, Trem2, Cd36, Fabp5, Lpl), VSMC phenotypic modulation (Vcam1, Ly6a, Lgals3, Lum, Ly6c1, Dcn), a few classical macrophage markers (Lyz2, Il6, Folr2, Cd68), and proliferation (Top2a, Birc5, Ube2c, Slurp1, Mki67) at both 16 weeks (Fig. 5A, Fig. 5C) and 26 weeks (Fig. 5B, Fig. 5D) of WD. In parallel, VSMC contractile genes Cnn1 and Myh11 were also downregulated (Fig. 5D). Pathway analysis using all DE genes in this population showed activation of numerous cell cycle-related pathways (Fig. 5E) akin to the proliferative VSMC subpopulation identified in our CITE-seq analysis (Fig. 3E). Together, these findings indicate that VSMC-derived foam cells exhibit coordinated upregulation of lipid metabolism genes, increased proliferation and inflammation genes while undergoing phenotypic modulation, overlapping with multiple subsets of phenotypically modulated VSMCs identified in our CITE-seq data.
To identify transcriptional regulators governing the development of VSMC-derived foam cells, we performed upstream regulator analyses at 16 weeks (Fig. 5F) and 26 weeks (Fig. 5G), revealing substantial overlap (Fig. 5H). Eleven transcriptional regulators, including BHLHE40, CEBPB, FOXO1, JUN, KLF6, and YAP1, were shared across time points, indicating sustained activation from early disease to advanced atherosclerosis. Among these regulators, Bhlhe40 also had a two-fold increase in mRNA levels in VSMC-derived foam cells at both time points of WD (Fig. 5I and 5J).
To determine which subpopulations in our CITE-seq analysis align with VSMC-derived foam cells, we examined the expression of lipid metabolism genes across the CITE-seq dataset (Fig. 6A). This revealed several VSMC-derived subpopulations were enriched in genes associated with lipid metabolism. VSMC-derived macrophage-like cells exhibited high expression of lipid metabolism genes including Abca1, Lipa, Lpl, Trem2, Lrp1, and Apoe, consistent with macrophage-like lipid metabolism functions. Abca1 was also highly expressed in the modulated VSMC 1 and proliferative VSMC subpopulations. Among the lipid metabolism genes, Lrp1 and Apoe were consistently expressed across all modulated VSMC subpopulations. These findings suggest that foam cell features are found in multiple modulated VSMC subtypes and are consistent with the upregulation of genes associated with multiple cellular phenotypes in VSMC foam cells in bulk RNA-seq (Fig. 5).
To explore the inferred developmental trajectory and transcriptional regulation of VSMCs in atherosclerosis, we conducted a pseudotime analysis on the CITE-seq data from VSMC-lineage cells described in Fig. 3, combining all contractile VSMC subtypes into a single population, as their overlapping gene expression profiles made it challenging to determine the precise cluster from which the trajectory originates. Our analysis revealed three distinct trajectories, each progressing from contractile VSMCs through chondromyocytes to modulated VSMC2, and subsequently diverging to terminate at either modulated VSMC1 (lineage 1), proliferative VSMCs (lineage 2), or macrophage-like VSMCs (lineage 3) (Fig. 6B). To identify putative transcriptional regulators that control the developmental trajectory of lineage 1 and 2, we performed an upstream regulator analysis (Fig. 6C). This revealed several potential regulators that overlapped with the upstream regulator analysis of bulk RNA-seq VSMC foam cells, including Klf6, Cebpb, and Bhlhe40. In our CITE-seq data, Klf6 and Cebpb were highly expressed across all VSMC subpopulations, while Bhlhe40 was specific to modulated VSMC subpopulations (Fig. 6D). Complementary master regulatory analysis on the VSMC subtypes using ARACNe identified BHLHE40 as a putative regulator selectively active in modulated and proliferative VSMCs, but not in contractile VSMCs (Fig. 6E). Using immunofluorescence, we detected BHLHE40 in the intima of atherosclerotic lesions at both 16 weeks and 26 weeks of WD (Fig. 6F). Furthermore, Bhlhe40 expression increased progressively in modulated states along inferred pseudotime trajectory (Fig. 7A). Taken all together, these findings implicate BHLHE40 as a previously uncharacterized putative candidate transcriptional regulator of VSMC foam cells and phenotypic modulation in atherosclerosis.
To understand the role of BHLHE40 in VSMC phenotypic modulation, we used Cell Oracle to perform in silico perturbation analysis of CITE-seq data modelling the effect of Bhlhe40 knockout and overexpression. Simulated Bhlhe40 knockout blocked phenotypic modulation of VSMCs, while overexpression promoted VSMCs transition to a modulated state (Fig. 7B).
To further elucidate the mechanism by which Bhlhe40 promotes VSMC phenotypic switching, we performed in vitro bulk RNA-seq during siRNA targeting of Bhlhe40 (siBhlhe40) in primary mouse aortic VSMCs. Phenotypic modulation of VSMCs was induced by treating cultured cells with TNFα and methyl-β-cyclodextrin (MBD) cholesterol. This in vitro “atherogenic stress” treatment is associated with decreased expression of VSMC contractile genes (Myh11, Acta2, Cnn1, and Tagln), increased expression of Ly6a, Lum, Vcam1, and Dcn, and an increase in lipid metabolism gene expression (Cd68, Abca1, and Cd36) (Fig. 7C – 7F), consistent with prior studies10,53,61 and data for in vivo SMC phenotype modulation. Transcriptomic profiling revealed reproducible shifts in gene expression between untreated cells and TNFα/MBD cholesterol treated cells. Heatmaps of ranked gene expression (Suppl Fig. 3A–3E) demonstrate highly concordant expression patterns across biological replicates, supporting the reproducibility of the TNFα/MBD cholesterol effects. Bhlhe40 siRNA knockdown modulated these transcriptional responses under both basal and TNFα/MBD-cholesterol treated conditions (Suppl Fig. 3D, Extended Data Set 4, 5, and 6). Independent RT-PCR validation confirmed consistent reductions in Cd68, Abca1 and Cd36 following Bhlhe40 knockdown (Suppl Fig. 4A); Vcam1 showed a trend toward decreased expression that did not reach significance under conservative permutation-based analysis (Suppl Fig. 4A). Furthermore, Bhlhe40 knockdown significantly reduced CD68, CD36, and ABCA1 protein levels in the treated states (Suppl Fig. 4B and 4C), suggesting that BHLHE40 blunts VSMC phenotypic transitions and foam cell–like features.
IPA and Gene Set Enrichment Analysis (GSEA) revealed a strong activation of viral and bacterial response pathways with TNFα/MBD treatment along with significant down regulation of extracellular matrix processes and muscle identity/differentiation pathways (Suppl Fig. 3F and 3H, Extended Data Set 7, 11 and 12). In contrast, knockdown of Bhlhe40 attenuated the expression of genes associated with phenotypic modulation and macrophage identity and function (Cd68, Abca1, Vcam1, Dcn, IL6 and Ly6c) (Fig. 7D). Furthermore, upstream regulator analysis with IPA identified several transcription factors predicted to promote VSMC phenotypic modulation both in vitro and in vivo, including KLF6, BHLHE40, STAT1, IRF3, IRF7, ZBTB10 and JUN (Extended Data Set 8 and 10).
Bhlhe40 silencing during TNFα/MBD-cholesterol treatment produced a distinct transcriptional profile characterized by enrichment of pathways related to cell division, DNA replication, and cell cycle checkpoints (Suppl Fig. 3G and 3I, Extended Data Set 9, 13 and 14). This was accompanied by suppression of immune-mediated pathways, with GSEA further revealing coordinated decreases in lipid metabolism and cholesterol biosynthesis pathways, consistent with transcriptional repression of cholesterol biosynthesis genes including Acat2, Acat3, Msmo1, and Mvd (Fig. 7G, Suppl Fig. 3G and 3I). Transcription factor activity analysis of Bhlhe40 silencing during phenotypic modulation in vitro confirmed the predicted decrease in BHLHE40 activity but also revealed a decrease in activity of other transcription factor previously predicted to be increased in foamy VSMC (KLF6, STAT1, IRF3, IRF7, ZBTB10) (Extended Data Set 10). A strong predicted inhibition of p53 (TP53) activity was also observed, consistent with prior reports showing downregulation in contractile SMC and upregulation in modulated SMC populations10. Additional transcription factors potentially associated with VSMC phenotypic switching (E2FA, MYC, TOX, and STAT3) were also predicted to be involved, suggesting a complex role for BHLHE40 during phenotypic modulation. Overall, these findings highlight the role of BHLHE40 in regulating VSMC phenotypic modulation under atherogenic conditions and are consistent with simulated Bhlhe40 perturbation and overexpression in CITE-seq data from our in vivo mouse model.
To extend findings to human atherosclerosis, we re-analyzed our recently published CITE-seq dataset of human carotid atherosclerosis13, with a specific focus on phenotypically modulated VSMCs and macrophage populations. VSMCs, fibrotic-type cells (Fig. 8A) and macrophages (Fig. 8B) were isolated. Building upon our initial annotations, we refined our classification to identify multiple contractile VSMC subtypes, two fibroblast populations (CD26+ and THSD4+), myofibroblasts, and foamy VSMCs, along with endothelial cells undergoing endothelial-to-mesenchymal transition (endoMT EC) (Fig. 8A). Within the macrophage compartment, as previously reported, we identified IL1Bhi, C1Qhi, proliferative, and ACTA2+ subsets (Fig. 8B and C). Since lineage tracing is not practical in human atherosclerosis, canonical gene markers were used to identify phenotypically modulated VSMCs (Fig. 8D). The foamy VSMC population exhibited low expression of the VSMC contractile gene TAGLN but showed moderate expression of lipid metabolism-associated genes, including LPL, NR1H3, CD36, and APOE. In contrast, ACTA2+ macrophages expressed all major VSMC contractile genes (ACTA2, TAGLN, MYH11, CNN1), but at a much lower level than other VSMC subsets, and exhibited lower expression of lipid metabolism genes relative to other macrophage subtypes. The ACTA2+ macrophages expressed hallmark macrophage proteins (Fig. 8E), VSMC and fibroblast markers as well as markers commonly associated with phenotypically modulated VSMCs (VCAM1, DCN, and LUM). Consistent with the mouse data, the proportion of these putative VSMC-derived macrophage-like cells in human plaques was modest (~6%; Fig. 8C).
Last, we found that the expression pattern of BHLHE40 was comparable to our observations in murine models (Fig. 8F) and was specifically observed in cells displaying fibrotic and foamy phenotypes, while contractile VSMCs showed no detectable BHLHE40 expression. Immunofluorescence staining of atherosclerotic human coronary arteries showed colocalization of MYH11 and BHLHE40 (Fig. 8H, Suppl Fig. 5). Immunohistochemical analysis of stable and unstable carotid atherosclerotic plaques revealed BHLHE40 expression (Fig. 8G). Although the limited sample size precluded formal statistical analysis, BHLHE40 staining appeared more prominent in stable plaques.
Discussion
We present a comprehensive cellular analysis of mouse atherosclerosis across distinct stages of disease progression. By combining the lineage tracing of VSMCs and multi-omic profiling, we provide novel insights into VSMC phenotypic modulation in atherosclerosis. We validate the expansion of a modulated VSMC subpopulation resembling a fibromyocyte population and identify a distinct proliferative VSMC subset that expands in advancing atherosclerosis. Using rigorous orthogonal validations, we document a rare VSMC population exhibiting macrophage-like characteristics. Supported by convergent computational, expression, and knockdown evidence, we identify Bhlhe40 as a putative transcriptional regulator of VSMC phenotypic transitions and foam cell formation. We further validate that the majority of foam cells within the lesions originate from VSMCs. Analysis of human atherosclerotic datasets suggests the presence of VSMC-derived foam cells in human lesions, a similar subcellular expression pattern of BHLHE40 in human plaques, and a low-frequency population of apparent VSMC-derived ACTA2+ macrophages, mirroring mouse models observations.
The developmental trajectory and phenotypic diversity of VSMCs have been extensively studied over the past several decades. Early work proposed that VSMCs can transition from a contractile to a synthetic phenotype, existing along a dynamic continuum between these states, particularly in the context of atherosclerosis62. The advent of high-dimensional bulk and single-cell technologies has substantially refined and extended our understanding of VSMC phenotypic modulation8–10,14,18,55,63–66. However, these approaches have also introduced new uncertainties and prompted debate regarding the full spectrum of cellular phenotypes and functions that VSMCs can adopt in disease.
Here, we address these discrepancies using rigorous multi-omics and orthogonal approaches to define VSMC-derived phenotypes and ensure reproducibility. First, we implemented a stringent flow cytometry gating strategy to eliminate false-positive ZsGreen1+ cells and validated their identity through transcriptomic profiling. Second, we used ZsGreen1 transcript detection as an independent and robust VSMC lineage tracing strategy, similar to Sharma et al.18. Third, we performed RNA-scope in-situ hybridization on aortic lesions from mice at late and advanced stages of disease. Across 16- and 26-weeks WD timepoints, less than 5% of cells within the lesions were ZsGreen1+Cd68+, indicating VSMC-derived macrophage-like cells are a minor population in both developing and mature lesions. Collectively, these approaches provide robust support for the rare occurrence of VSMC-derived macrophage-like cells in mouse and human lesions, helping clarify this area of controversy.
We focused next on lipid uptake by VSMCs during atherosclerosis progression. VSMCs can internalize lipids and transition into foam cells, a hallmark of advanced atherosclerosis. Consistent with previous reports66,67, the majority of foam cells in late-stage mouse lesions were VSMC-derived, with similar findings in human lesions. Critically, hyperlipidemia-driven modulation produced distinct transcriptomic profiles between contractile VSMCs and their foam cell counterparts. Integrating single-cell and bulk RNA-seq data revealed substantial cellular heterogeneity within the VSMC foam cell continuum, with modulated, proliferative, and macrophage-like VSMCs subsets sharing upregulated lipid metabolism genes. Although VSMC-derived foam cells exhibit some macrophage-associated expression features, they lack most macrophage gene expression and protein signatures, making functional equivalence unlikely. These findings support the dynamic and multifaceted nature of VSMC-to-foam cell modulation during atherosclerosis progression.
A central goal of single-cell profiling in atherosclerosis is to identify regulators that drive pathogenic cellular phenotypes and represent potential novel therapeutic targets. Several regulators have been proposed and experimentally validated, including TCF218, KLF49, and retinoic acid signaling10. Building on this, we integrated trajectory inference with upstream and master regulator analyses to prioritize candidate transcriptional regulators controlling terminally differentiated VSMC phenotypes. Based on these inferential analyses, a key candidate that emerged was BHLHE40, a basic helix-loop-helix transcription factor that regulates cell proliferation, differentiation, and death68. BHLHE40 is relevant to several atherosclerosis cell types, including macrophages and T cells, where it plays a proinflammatory role69, endothelial cells, where it is regulated by SREBP2, a key modulator of cholesterol biosynthesis70, and hepatocytes where it works with LXRα and C/EBPβ to induce SREBP-1c and enhance fatty acid synthesis71. Bhlhe40 expression is also induced by retinoic acid, regulating LXR/RXR target genes72, notable given our prior work linking retinoic acid signaling to VSMC phenotype switching10. More recently, loss of BHLHE40 and its family member BHLHE41 in human iPSC-derived microglia and human THP-1 macrophages increased expression of LXR/RXR target genes involved in lipid accumulation and cholesterol efflux73.
Motivated by these observations, we investigated a novel role for Bhlhe40 in regulating VSMC phenotypic switching and foam cell formation. Upstream regulator analysis predicted increased BHLHE40 activity, supported by elevated Bhlhe40 expression in modulated VSMCs identified by CITE-seq and bulk RNA-seq of VSMC-derived foam cells, with comparatively lower expression in contractile VSMCs. In silico perturbation using CellOracle predicted that simulated Bhlhe40 knockdown attenuated, while overexpression promoted, VSMC phenotypic modulation. Consistent with these findings, our in vitro Bhlhe40 knockdown effects were most pronounced under pro-atherogenic stress conditions of TNFα and cholesterol exposure, reducing expression of lipid metabolism genes (Abca1 and Cd36) and phenotypic modulation markers (Cd68) and showing a trend toward reduced Vcam1 expression that did not reach significance under conservative permutation-based analysis. Bulk RNA-seq of cultured VSMC revealed that knockdown of Bhlhe40 during TNFα and cholesterol treatment led to reversal of immune activation, activation of cell cycle and DNA replication and a decrease in activity of transcription factors previously predicted to be increased in VSMC foam cells. This expression pattern differs from that observed during knockdown of Bhlhe40 in microglia and THP-1 macrophages, potentially reflecting cell-type-specific responses in an atherogenic milieu. These findings support a role for BHLHE40 in contributing to the phenotypic modulation of VSMCs toward a foam cell-like state during atherosclerosis.
To translate mouse findings to humans, we performed deeper analyses of our published human multimodal atlas of human atherosclerosis13. We identified distinct subsets of ACTA2+ macrophages and VSMC-derived foam cells, and observed a BHLHE40 expression pattern closely mirroring our mouse findings. Although lineage tracing is not feasible in humans, BHLHE40 was enriched in foamy and fibrotic cell populations likely to be partly VSMC-derived based on their transcriptional profiles. We validated these findings by immunostaining both coronary and carotid atherosclerotic lesions, where BHLHE40 colocalized with VSMCs at both anatomical sites.
In conclusion, we present a comprehensive multimodal characterization of VSMC phenotypic diversity across atherosclerosis progression in mice and complex human lesions, prioritizing BHLHE40 as a candidate regulator of VSMC phenotypic modulation and foam cell formation, with conserved expression in human disease. Whether BHLHE40 acts independently or in concert with other transcription factors to drive VSMC phenotypic modulation remains to be determined.
AC
Dr. Arlene Chung
Emergency Medicine · General Emergency MedicineUniversity of VermontarleneBurlington, VT, USA