Single-cell and spatial transcriptomics map senescent vascular cells in arterial remodeling during atherosclerosis in mice

Abstract

Growing evidence suggests that the induction of cellular senescence in vascular cells is causally linked to the etiology of cardiovascular diseases. To investigate systematically the heterogeneity of senescent vascular cells in atherosclerosis, we used a high-fat diet and PCSK9 overexpression to induce atherosclerosis in a senescence reporter mouse model (p16-tdTomato^+/-) and performed single-cell RNA sequencing on whole aortas. Using the SenMayo and CellAge gene sets, we identified four clusters of vascular smooth muscle cells (VSMCs), fibroblasts and T cells enriched in features of senescence, which were reduced upon treatment with the senolytic agent ABT-737. We then derived a global senescence signature of atherosclerosis including Spp1, Ctsb and Tnfrsf11b mRNAs. We validated the enrichment of these mRNAs in senescence by using spatial transcriptomics in a second mouse model of atherosclerosis and senolysis (Ldlr^-/-; p16-3MR), as well as by analyzing in vitro models of human VSMC senescence. Our results uncover a vascular-specific transcriptomic signature of senescence that may be exploited for tracking and treating age-related vascular diseases.

Extended Data Fig. 1 |. Atherosclerotic plaque pathology analysis and quality control analysis.

a, Quantification of plaque size presented by sex and condition (ND, HFD, and HFD + ABT-737). For female mice, ND (n = 4), HFD (n = 3), and HFD + ABT-737 (n = 2) and HFD vs. ND, p = 0.0006 and HFD + ABT-737 vs. HFD, p = 0.576, whereas for male mice, ND (n = 4), HFD (n = 2), and HFD + ABT-737 (n = 3) and HFD vs. ND, p = 0.0004 and HFD + ABT-737 vs. HFD, p = 0.8876. b, Quantification of necrotic core area presented by sex and condition (ND, HFD, and HFD + ABT-737). For female mice, HFD (n = 3) and HFD + ABT-737 (n = 2) HFD + ABT-737 vs. HFD, p = 0.1929, whereas for male mice, HFD (n = 2) and HFD + ABT-737 (n = 2) and HFD + ABT-737 vs. HFD, p = 0.4985. c, Quantification by plaque fibrous cap thickness presented by sex and condition (ND, HFD, and HFD + ABT-737). For female mice, HFD (n = 3) and HFD + ABT-737 (n = 2) and HFD + ABT-737 vs. HFD, p = 0.2033, whereas for male mice, HFD (n = 2) and HFD + ABT-737 (n = 2) and HFD + ABT-737 vs. HFD, p = 0.4552. d, Representative H&E staining of aortic roots from ND, HFD, and HFD + ABT-737. The scale bar is 200 μm. e, Quantification of collagen fibers (Aniline blue) area / area of aortic root presented by sex and condition (ND, HFD, and HFD + ABT-737). For female mice, HFD (n = 2) and HFD + ABT-737 (n = 2) HFD + ABT-737 vs. HFD, p = 0.2467, whereas for male mice, HFD (n = 2) and HFD + ABT-737 (n = 2) and HFD + ABT-737 vs. HFD, p = 0.7036. f, Representative senescence-associated beta galactosidase (SA-β-gal) activity staining of aortic roots from ND, HFD, and HFD + ABT-737. The scale bar is 200 μm. g, Quantification of SA-β-gal (ND, HFD, and HFD + ABT-737). ND (n = 5), HFD (n = 5) and HFD + ABT-737 (n = 5) and for significance, HFD vs. ND, p ≤ 0.0001 and HFD + ABT-737 vs. HFD, p = 0.069. h, Pulse wave velocity (PWV) measurements presented by sex and condition (ND, HFD, and HFD + ABT-737). For female mice, ND (n = 3), HFD (n = 2), and HFD + ABT-737 (n = 4) and HFD vs. ND, p = 0.748 and HFD + ABT-737 vs. HFD, p = 0.9597, whereas for male mice, ND (n = 3), HFD (n = 2), and HFD + ABT-737 (n = 2) and HFD vs. ND, p = 0.0294 and HFD + ABT-737 vs. HFD, p = 0.1001. i, Quality control table of the scRNA-seq samples including estimated number of cells, fraction reads in cells, mean reads per cell, median UMI counts per cell, median genes per cell, and total genes detected. j, Heatmap of normalized expression levels of classical senescence associated mRNAs either enriched in HFD and reduced by ABT-737 treatment, or conversely, reduced by HFD and increased by ABT-737 treatment in mice (Lmnb1 and Lbr mRNAs). Significance was established using Two-Way ANOVA (for data in graphs a-c, e, and h) and One-Way ANOVA (for data in graph g) with multiple comparisons. *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001.

Extended Data Fig. 2 |. scRNA-seq analysis of senescence gene expression.

a, UMAPs of p16-tdTomato mRNA expression in ND, HFD, and HFD + ABT-737 treated mice. b, UMAP plot of Cdkn2a mRNA expression across all clusters and conditions (left). Violin plot of Cdkn2a mRNA expression across all clusters (right). c, UMAP plot of Cdkn1a mRNA expression across all clusters and conditions (left). Violin plot of Cdkn1a mRNA expression across all clusters (right). d, GSEA SenMayo plots comparing HFD vs ND and HFD + ABT-737 vs HFD for Clusters 15, 20, 21, 3, 16, and 27. e, GSEA CellAge plots comparing HFD vs ND and HFD + ABT-737 vs HFD for Clusters 6, 17, 7, and 11.

Extended Data Fig. 3 |. VSMC phenotype analysis.

a, Dot plot representation of the gene expression profile for VSMC phenotype mRNA markers for cluster 0 subclustered VSMCs. b, Dot plot representation of the gene expression profile for VSMC phenotype mRNA markers for cluster 12 subclustered VSMCs.

Extended Data Fig. 4 |. Analysis of UMI counts, mitochondrial and ribosomal gene expression in VSMC, fibroblast, and T cell clusters with senescence features.

a-c, UMAPs of scRNA-seq UMI counts (a), mitochondrial gene expression (b) and ribosomal gene expression (c) across cluster 0 subclustered VSMCs in ND, HFD, and HFD + ABT-737 treated mice. d-f, UMAPs of scRNA-seq UMI counts (d), mitochondrial gene expression (e) and ribosomal gene expression (f) across cluster 12 subclustered VSMCs in ND, HFD, and HFD + ABT-737 treated mice. g-i, UMAPs of scRNA-seq UMI counts (g), mitochondrial gene expression (h) and ribosomal gene expression (i) across cluster 2 subclustered fibroblasts in ND, HFD, and HFD + ABT-737 treated mice. j-l, UMAPs of scRNA-seq UMI counts (j), mitochondrial gene expression (k) and ribosomal gene expression (l) across Cluster 9 subclustered T cells in ND, HFD, and HFD + ABT-737 treated mice.

Extended Data Fig. 5 |. Vascular senescence score on individual cell clusters.

a, UMAPs of vascular-senescence scoring across cluster 0 VSMCs in ND, HFD, and HFD + ABT-737. b, UMAPs of vascular-senescence scoring across cluster 12 VSMCs in ND, HFD, and HFD + ABT-737. c, UMAPs of vascular-senescence scoring across cluster 2 fibroblasts in ND, HFD, and HFD + ABT-737. d, UMAPs of vascular-senescence scoring across cluster 9 T cells in ND, HFD, and HFD + ABT-737.

Extended Data Fig. 6 |. Gene score of the most prominent transcripts expressed in senescent vascular cells.

UMAPs visualization of the expression of the 17 vascular senescent transcripts individually across all clusters. The arrows indicate the senescent cell clusters for VSMCs (V), fibroblasts (F), and T cells (T) as well as non-senescent macrophages (M) in which the transcripts were expressed.

Extended Data Fig. 7 |. Doxorubicin-induced cellular senescence in wild type mice.

a, RT-qPCR analysis of the levels of Cdkn2a and Lmnb1 mRNAs in aortic roots of mice treated with vehicle, doxorubicin, and doxorubicin followed by ABT-737. b, RT-qPCR analysis of the levels of Ltbp2, Sparc, Mgp, Lum, and Cd9 mRNAs in aortic root tissue as described (a). c, RT-qPCR analysis of the levels of Apoe, Lgals3, Cst3, Tnfrsf11b, and Prg4 mRNAs in aortic root tissue as described (a). d, RT-qPCR analysis of the levels of Mmp3, Ctss, Lcp1, and Comp mRNAs in aortic root tissue as described (a). In a-d, data represent the means ± SD for DMSO, n = 4; Doxo, n = 4; Doxo + ABT-737, n = 3 (females), and DMSO, n = 4; Doxo, n = 4; Doxo + ABT-737, n = 4 (males). n represents biological replicates. In panel (a), Cdkn2a, Doxo vs. DMSO, p = 0.0162 (females) and p = 0.002 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.0226 (females) and p = 0.0034 (males); Lmnb1, Doxo vs. DMSO, p = 0.5863 (females) and p = 0.1807 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.1013 (females) and p = 0. 3449 (males). In panel (b), Ltbp2, Doxo vs. DMSO, p = 0.0004 (females) and p ≤ 0.0001 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.0007 (females) and p ≤ 0.0001 (males); Sparc, Doxo vs. DMSO, p ≤ 0.0001 (females) and p ≤ 0.0001 (males), whereas Doxo + ABT-737 vs. Doxo, p ≤ 0.0001 (females) and p = 0.0002 (males); Mgp, Doxo vs. DMSO, p = 0.9928 (females) and p = 0.9928 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.7181 (females) and p > 0.9999 (males); Lum, Doxo vs. DMSO, p = 0.0018 (females) and p = 0. 0002 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.0018 (females) and p = 0.001 (in males); Cd9, Doxo vs. DMSO, p = 0.0014 (females) and p = 0. 0011 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.0031 (females) and p = 0.0019 (males). In panel (c), Apoe, Doxo vs. DMSO, p = 0.0194 (females) and p = 0. 3607 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.5746 (females) and p = 0.4692 (males); Lgals3, Doxo vs. DMSO, p ≤ 0.0001 (females) and p ≤ 0.0001 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.0002 (females) and p ≤ 0.0001 (males); Cst3, Doxo vs. DMSO, p = 0.0008 (females) and p ≤ 0.0001 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.0021 (females) and p ≤ 0.0001 (males); Tnfrsf11b, Doxo vs. DMSO, p = 0.3051 (females) and p = 0.9339 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.1425 (females) and p = 0.9833 (males); Prg4, Doxo vs. DMSO, p ≤ 0.0001 (females) and p = 0.0177 (males), whereas Doxo + ABT-737 vs. Doxo, p ≤ 0.0001 (females) and p = 0.0283 (males). In panel (d), Mmp3, Doxo vs. DMSO, p ≤ 0.0001 (females) and p ≤ 0.0001 (males), whereas Doxo + ABT-737 vs. Doxo, p ≤ 0.0001 (females) and p ≤ 0.0001 (males); Ctss, Doxo vs. DMSO, p = 0.0007 (females) and p = 0.0025 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.0017 (females) and p = 0.005 (males); Lcp1, Doxo vs. DMSO, p ≤ 0.0001 (females) and p ≤ 0.0001 (males), whereas Doxo + ABT-737 vs. Doxo, p ≤ 0.0001 (females) and p ≤ 0.0001 (males); Comp, Doxo vs. DMSO, p = 0.0005 (females) and p = 0.0029 (males), whereas Doxo + ABT-737 vs. Doxo, p = 0.0027 (females) and p = 0.004 (males). Significance was established using One-Way Anova for each sex, *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001.

Extended Data Fig. 8 |. Human VSMC senescence validation.

a, Phase contrast micrographs of SA-βgal activity (blue) staining in proliferating (P), replicative senescence (RS), doxo-treated, IR-treated, oxLDL-treated, and CoCl2-treated human VSMCs for 7–10 days. b, RT-qPCR analysis of the levels of CDKN2A and IL8 mRNAs in human VSMCs treated as described in (a). c, RT-qPCR analysis of the levels of LTBP2, SPARC, and MGP mRNAs in human VSMCs treated as described in (a). d, RT-qPCR analysis of the levels of LUM, CD9, APOE, LGALS3, CST3 mRNAs in human VSMCs treated as described in (a). e, RT-qPCR analysis of the levels of TNFRSF11b, PRG4, MMP3, CTSS, and LCP1 mRNAs in human VSMCs treated as described in (a). f, Table of human scRNA-seq metadata analysis from atherosclerotic tissue. Each column represents a VSMC phenotype identified in the metadata analysis, and each column includes the vascular senescent scoring genes that were expressed in the indicated VSMC phenotype. In b-e, data represent the means ± SD for P, n = 3; IR, n = 3; oxLDL, n = 3; CoCl2, n = 3; Doxo, n = 3; RS, n = 3; n represents biological replicates. In panel (b), CDKN2A, IR (p = 0.1334), oxLDL (p = 0.0191), CoCl2 (p = 0.001), Doxo (p = 0.4446), RS (p ≤ 0.0001) vs. P; IL8, IR (p = 0.0007), oxLDL (p = 0.0014), CoCl2 (p = 0.0003), Doxo (p = 0.0008), RS (p = 0.0027) vs. P. In panel (c), LTBP2, IR (p = 0.002), oxLDL (p = 0.0021), CoCl2 (p ≤ 0.0001), Doxo (p ≤ 0.0001), RS (p = 0.4192) vs. P; SPARC, IR (p = 0.2171), oxLDL (p ≤ 0.0001), CoCl2 (p = 0.0002), Doxo (p = 0.0106), RS (p = 0.0019) vs. P; MGP, IR (p = 0.0201), oxLDL (p = 0.001), CoCl2 (p = 0.0134), Doxo (p = 0.0015), RS (p = 0.0043) vs. P. In panel (d), LUM, IR (p ≤ 0.0001), oxLDL (p = 0.0002), CoCl2 (p ≤ 0.0001), Doxo (p = 0.0003), RS (p ≤ 0.0001) vs. P; CD9, IR (p = 0.0002), oxLDL (p = 0.0296), CoCl2 (p = 0.0151), Doxo (p = 0.0291), RS (p = 0.0002) vs. P; APOE, IR (p = 0.0171), oxLDL (p = 0.0004), CoCl2 (p = 0.016), Doxo (p = 0.0486), RS (p = 0.1703) vs. P; LGALS3, IR (p = 0.2702), oxLDL (p = 0.0014), CoCl2 (p = 0.007), Doxo (p = 0.0284), RS (p = 0.0014) vs. P; CST3, IR (p = 0.9809), oxLDL (p ≤ 0.0001), CoCl2 (p = 0.1202), Doxo (p = 0.087), RS (p = 0.0209) vs. P. In panel (e), TNFRSF11b, IR (p = 0.0019), oxLDL (p ≤ 0.0001), CoCl2 (p = 0.001), Doxo (p = 0.0003), RS (p ≤ 0.0001) vs. P; PRG4, IR (p = 0.0595), oxLDL (p = 0.0005), CoCl2 (p = 0.0008), Doxo (p = 0.0011), RS (p = 0.0021) vs. P; MMP3, IR (p = 0.0054), oxLDL (p = 0.0007), CoCl2 (p = 0.0003), Doxo (p = 0.0024), RS (p = 0.0098) vs. P; CTSS, IR (p = 0.0085), oxLDL (p ≤ 0.0001), CoCl2 (p = 0.0922), Doxo (p = 0.0019), RS (p = 0.0003) vs. P; LCP1, IR (p = 0.0029), oxLDL (p = 0.0033), CoCl2 (p = 0.0003), Doxo (p = 0. 0008), RS (p = 0.0006) vs. P. Significance was established using Shapiro-Wilk test first, followed by Welch’s t-test. *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001.

Extended Data Fig. 9 |. Human WI-38 fibroblast senescence validation.

a, Phase contrast micrographs of SA-βgal activity (blue) staining in human WI-38 fibroblasts that were either proliferating (P), rendered senescent by replicative senescence (RS), or by treatment with Doxo, IR and additional culture for 7–10 days. b, RT-qPCR analysis of the levels of CDKN2A, CDKN1A, GDF15, LMNB1 and IL8 mRNAs in WI-38 fibroblasts treated as described in (a). c-f, RT-qPCR analysis of the levels of SPP1, SERPINE2, CTSB, TNFRSF11B, PRG4, LTPBP2, FTH1, MGP, LUM, CD9, APOE, LGALS3, CST3, SPARC, THBS1, MMP3, CTSS, and LCP1 mRNAs in human WI-38 fibroblasts treated as described in (a). In b-f, data represent the means ± SD for P, n = 3; IR, n = 3; n = 3; Doxo, n = 3; RS, n = 3; n represents biological replicates. In panel (b), CDKN2A, IR (p = 0.0099), Doxo (p ≤ 0.0001), RS (p = 0.0005) vs. P; CDKN1A, IR (p ≤ 0.0001), Doxo (p ≤ 0.0001), RS (p ≤ 0.0001) vs. P; GDF15, IR (p = 0.001), Doxo (p ≤ 0.0001), RS (p ≤ 0.0021) vs. P; LMNB1, IR (p = 0.001), Doxo (p = 0.0002), RS (p = 0.0009) vs. P; IL8, IR (p ≤ 0.0001), Doxo (p = 0.0005), RS (p = 0.0003) vs. P. In panel (c), SPP1, IR (p = 0.0085), Doxo (p = 0.0035), RS (p = 0.0071) vs. P; SERPINE2, IR (p = 0.0014), Doxo (p = 0.0025), RS (p = 0.0026) vs. P; CTSB, IR (p = 0.0003), Doxo (p = 0.0002), RS (p = 0.0007) vs. P; TNFRSF11B, IR (p = 0.0013), Doxo (p = 0.0003), RS (p ≤ 0.0001) vs. P; PRG4, IR (p = 0.0135), Doxo (p = 0.0043), RS (p = 0.0039) vs. P. In panel (d), LTPBP2, IR (p = 0.0003), Doxo (p = 0.0012), RS (p = 0.0006) vs. P; FTH1, IR (p = 0.0025), Doxo (p = 0.0038), RS (p = 0.0016) vs. P; MGP, IR (p = 0.0007), Doxo (p ≤ 0.0001), RS (p ≤ 0.0001) vs. P; LUM, IR (p = 0.0005), Doxo (p = 0.0059), RS (p = 0.0014) vs. P; CD9, IR (p = 0.0051), Doxo (p = 0.7688), RS (p = 0.0008) vs. P. In panel (e), APOE, IR (p = 0.0008), Doxo (p = 0.007), RS (p ≤ 0.0001) vs. P; LGALS3, IR (p = 0.0108), Doxo (p = 0.0229), RS (p ≤ 0.0001) vs. P; CST3, IR (p = 0.0026), Doxo (p ≤ 0.0001), RS (p ≤ 0.0001) vs. P; SPARC, IR (p = 0.0021), Doxo (p = 0.002), RS (p = 0.0004) vs. P; THBS1, IR (p = 0.0003), Doxo (p = 0.0114), RS (p ≤ 0.0001) vs. P. In panel (f), MMP3, IR (p = 0.0015), Doxo (p = 0.0007), RS (p = 0.0003) vs. P; CTSS, IR (p = 0.0002), Doxo (p = 0.0023), RS (p = 0.0019) vs. P; LCP1, IR (p = 0.0003), Doxo (p = 0.9169), RS (p ≤ 0.0001) vs. P.Significance was established using Shapiro-Wilk test first, followed by Welch’s t-test. *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001.

Extended Data Fig. 10 |. Spatial transcriptomic profiling of arterial segments.

a, H&E staining (top) and Masson’s trichrome staining (bottom) of brachiocephalic arterial sections from Ldlr^−/− + GCV, Ldlr^−/−;3MR + Vehicle, and Ldlr^−/−;3MR + GCV treated mice. The scale bar is 100 μm. b, Plaque cross sectional area, c, plaque fibrous cap thickness measurements and d, Masson’s trichrome analysis of Aniline blue-positive area over total plaque area in brachiocephalic arterial sections from Ldlr^−/− + GCV (n = 4), Ldlr^−/−;3MR + Vehicle (n = 4), and Ldlr^−/−;3MR + GCV (n = 4) treated mice. In panel (b), Ldlr^−/−;3MR + Vehicle vs. Ldlr^−/− + GCV (p = 0.8299) and Ldlr^−/−;3MR + GCV vs. Ldlr^−/−;3MR + Vehicle (p = 0.7933). In panel (c), Ldlr^−/−;3MR + Vehicle vs. Ldlr^−/− + GCV (p = 0.5444) and Ldlr^−/−;3MR + GCV vs. Ldlr^−/−;3MR + Vehicle (p = 0.3969). In panel (b), Ldlr^−/−;3MR + Vehicle vs. Ldlr^−/− + GCV (p = 0.7628) and Ldlr^−/−;3MR + GCV vs. Ldlr^−/−;3MR + Vehicle (p = 0.9994). e, H&E staining (top) and Masson’s trichrome staining (bottom) of the descending aortic sections from Ldlr^−/− + GCV, Ldlr^−/−;3MR + Vehicle, and Ldlr^−/−;3MR + GCV treated mice. The scale bar is 50 μm. f-h, Plaque cross sectional area (f), plaque fibrous cap thickness measurements (cap is underlined with yellow dashed line) (g), and Masson’s trichrome analysis of Aniline blue-positive area over total plaque area in sections of the descending aorta (h) from Ldlr^−/− + GCV, Ldlr^−/−;3MR + Vehicle, and Ldlr^−/−;3MR + GCV treated mice. In panel (f) and (g), Ldlr^−/− + GCV (n = 5), Ldlr^−/−;3MR + Vehicle (n = 4), and Ldlr^−/−;3MR + GCV (n = 6); in panel (f), Ldlr^−/−;3MR + Vehicle vs. Ldlr^−/− + GCV (p = 0.9706) and Ldlr^−/−;3MR + GCV vs. Ldlr^−/−;3MR + Vehicle (p = 0.08), whereas in (g), Ldlr^−/−;3MR + Vehicle vs. Ldlr^−/− + GCV (p = 0.379) and Ldlr^−/−;3MR + GCV vs. Ldlr^−/−;3MR + Vehicle (p = 0.0021). In panel (h), Ldlr^−/− + GCV (n = 4), Ldlr^−/−;3MR + Vehicle (n = 4), and Ldlr^−/−;3MR + GCV (n = 5), and significance Ldlr^−/−;3MR + Vehicle vs. Ldlr^−/− + GCV (p = 0.9867) and Ldlr^−/−;3MR + GCV vs. Ldlr^−/−;3MR + Vehicle (p = 0.0263) i, Reactome pathway analysis of genes upregulated in cap compared to normal wall across all conditions (left) and of genes upregulated in the normal wall compared to the cap (right). j, Reactome pathway analysis of genes upregulated in the core compared to the normal wall across all conditions. In b-d and f-h, data represent the means ± SD and n represents biological replicates. Significance was established using One-Way Anova with multiple comparisons (b-d and f-h) or Fisher-Irwin test (i and j). *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001.

Fig. 1 |. Senolytics improve ECM remodeling during atherosclerosis.

a, Schematic depicting the experimental design of our study. Briefly, p16-tdTomato^+/− mice were injected with PCSK9 AAV; 2 weeks later, mice were given either an ND or HFD for 16 weeks. At that time, a subset of HFD mice were treated with either a vehicle or ABT-737 for an additional 8 weeks while remaining on the diet before they were sacrificed for aortic single-cell isolation and sequencing. b, Representative images of aortic roots from ND, HFD and HFD + ABT-737 mice, stained with Oil Red O for lipid deposition examination and Masson’s trichrome to visualize collagen. The scale bar is 200 μm. c, Quantification of the relative lesion area using Oil Red O staining of the aortic roots of mice from the ND, HFD and HFD + ABT-737 groups (ND, n = 6; HFD, n = 4; HFD + ABT-737, n = 5; HFD versus ND, P ≤ 0.0001 and HFD + ABT-737 versus HFD, P = 0.9146). d, Quantification of the necrotic core area. e, Quantification of the cap thickness. f, Quantification of the ratio of the collagen fiber (aniline blue-positive) area to the area of the aortic root (sample size for the necrotic core area, cap thickness and collagen fibers (HFD, n = 4; HFD + ABT-737, n = 4)); significance for the necrotic core area (HFD + ABT-737 versus HFD, P = 0.0312), cap thickness (HFD + ABT-737 versus HFD, P = 0.0091) and collagen fibers (HFD + ABT-737 versus HFD, P = 0.0244). g, PWV measurements in mice in the ND, HFD and HFD + ABT-737 groups (ND, n = 6; HFD, n = 4; HFD + ABT-737, n = 5; HFD versus ND, P = 0.0261 and HFD + ABT-737 versus HFD, P = 0.0452). h, Bio-Plex analysis of circulating SERPINE1, CD93, ICAM-1, BAFF, MMP3 and CXCL1 in serum from mice in the ND, HFD and HFD + ABT-737 groups (SERPINE1 (ND, n = 7; HFD, n = 5; HFD + ABT-737, n = 4), HFD versus ND, P ≤ 0.0001 and HFD + ABT-737 versus HFD, P = 0.9205; CD93 (ND, n = 7; HFD, n = 5; HFD + ABT-737, n = 6), HFD versus ND, P = 0.0413 and HFD + ABT-737 versus HFD, P = 0.4241; ICAM-1 (ND, n = 6; HFD, n = 5; HFD + ABT-737, n = 4), HFD versus ND, P = 0.7365 and HFD + ABT-737 versus HFD, P = 0.9998; BAFF (ND, n = 6; HFD, n = 5; HFD + ABT-737, n = 4), HFD versus ND, P = 0.2129 and HFD + ABT-737 versus HFD, P = 0.0722; MMP3 (ND, n = 7; HFD, n = 5; HFD + ABT-737, n = 5), HFD versus ND, P ≤ 0.001 and HFD + ABT-737 versus HFD, P = 0.2665; CXCL1 (ND, n = 7; HFD, n = 5; HFD + ABT-737, n = 6), HFD versus ND, P = 0.0261 and HFD + ABT-737 versus HFD, P = 0.9693). i, Volcano plot of pseudobulk RNA-seq analysis comparing differential gene expression between HFD and ND aortas. j, Volcano plot of pseudobulk RNA-seq analysis comparing differential gene expression between HFD + ABT-737 and HFD aortas. k, Dot plot representation of pathway analysis depicting differentially expressed mRNAs associated with the complement and coagulation pathways (the green gradient represents shared genes between pathways; darker and lighter green are areas of unique genes). l, Dot plot representation of pathway analysis depicting differentially expressed mRNAs associated with the TGFβ, ECM and EMT signaling pathways (the green gradient represents shared genes between pathways; darker and lighter green are areas of unique genes). Data represent the means ± s.d. from indicated biological replicates. Significance was established using Student’s t test (d–f) or one-way ANOVA (c, g and h). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. Panel a created with BioRender.com.

Fig. 2 |. A subset of aortic cell clusters is enriched for senescence-associated genes.

a, Dot plot analysis of the top 4 genes expressed in each cluster for cell type determination. b, UMAP plot of aortic cells from all samples, color-labeled by major cell types. c, Bar graph of the distribution (%) of all clusters across conditions (ND, HFD and HFD + ABT-737). d, Cell distribution (%) of top changing clusters (0, 12, 23, 4, 8, 19 and 24) in the ND, HFD and HFD + ABT-737 conditions. e, Bar graph depicting the GSEA of the SenMayo panel, graphed by enrichment score for clusters with increased enrichment with HFD (red) and reduced enrichment with ABT-737 treatment (cyan). f, Representative GSEA plots demonstrating the increase in enrichment with HFD and decrease with ABT-737 for clusters 0, 2, 9 and 12. g, Bar graph depicting the GSEA of the CellAge panel, graphed by enrichment score for clusters with increased enrichment with HFD (red) and reduced enrichment with ABT-737 treatment (cyan). h, Representative GSEA plots demonstrating the increase in enrichment with HFD and decrease with ABT-737 for clusters 0, 2, 9 and 12. Fibro, fibroblasts; Mono, monocytes; ECs, endothelial cells; Macro, macrophages; Neutro, neutrophils; NES, normalized enrichment score.

Fig. 3 |. Identifying senescent subpopulations in VSMC clusters.

a, UMAP subclustering of cluster 0 (VSMCs) from the ND, HFD and HFD + ABT-737 mouse groups. Subclusters were then analyzed for populations that increased in the HFD condition and were diminished by ABT-737 treatment; subcluster 5 (red) was considered potentially senescent, and the percentage (subcluster 5/total cluster 0) is indicated for each condition. b, Dot plot analysis of the top 50 RNAs most highly expressed in subcluster 5 compared to all subclusters. c, Dot plot analysis of the top 50 RNAs most highly expressed in subcluster 5 compared across the ND, HFD and HFD + ABT-737 groups for cluster 0. d, Reactome pathway analysis of subcluster 5. e, UMAP subclustering of cluster 12 (VSMCs) from the ND, HFD and HFD + ABT-737 mouse groups. Subclusters were then analyzed for populations that increased in the HFD condition and were diminished by ABT-737 treatment; subcluster 4 (red) was considered potentially senescent, and the percentage is indicated for each condition. f, Dot plot analysis of the top 50 RNAs most highly expressed in subcluster 4 compared to all subclusters. g, Dot plot analysis of the top 50 RNAs most highly expressed in subcluster 4 compared across the ND, HFD and HFD + ABT-737 groups for cluster 12. h, Reactome pathway analysis of subcluster 4. In d and h, statistical significance was established using a Benjamini–Hochberg multiple test correction.

Fig. 4 |. Identifying senescent subpopulations in fibroblast and T cell clusters.

a, UMAP subclustering of cluster 2 (fibroblasts) from the ND, HFD and HFD + ABT-737 mouse groups. Subclusters were then analyzed for populations that increased in the HFD condition and were diminished by ABT-737 treatment; subcluster 3 (red) was considered potentially senescent, and the percentage is indicated for each condition. b, Dot plot analysis of the top 50 RNAs most highly expressed in subcluster 3 compared to all subclusters. c, Dot plot analysis of the top 50 RNAs most highly expressed in subcluster 3 compared across the ND, HFD and HFD + ABT-737 groups for cluster 2. d, Reactome pathway analysis of subcluster 3. e, UMAP subclustering of cluster 9 (T cells) from ND, HFD and HFD + ABT-737 mice. Subclusters were then analyzed for populations that increased in the HFD condition and were diminished by ABT-737 treatment; subcluster 5 (red) was considered potentially senescent, and the percentage is indicated for each condition. f, Dot plot analysis of the top 50 RNAs most highly expressed in subcluster 5 compared to all subclusters. g, Dot plot analysis of the top 50 RNAs most highly expressed in subcluster 5 compared across the ND, HFD and HFD + ABT-737 groups for cluster 9. h, Reactome pathway analysis of subcluster 5. In d and h, significance was established using a Benjamini–Hochberg multiple test correction.

Fig. 5 |. Validation of mouse vascular senescence markers in diverse models of vascular senescence.

a, Table of shared senescence-associated mRNAs in all vascular senescent clusters and the compartments in which the respective encoded proteins reside. b, UMAPs of vascular senescence scoring across all clusters using the genes from a, with black arrows pointing out senescent VSMCs (V), fibroblasts (F) and T cells (T). c, Schematic of doxorubicin (Doxo)-induced senescence in wild-type mice. d,e, Traditional senescence markers Cdkn1a, Il6 and Gdf15 mRNAs (d) and new senescence markers Spp1, Serpine2, Ctsb, Fth1 and Thbs1 mRNAs (e) were then measured by RT–qPCR analysis (for d and e, DMSO, n = 4; Doxo, n = 4; Doxo + ABT-737, n = 3 (females), whereas DMSO, n = 4; Doxo, n = 4; Doxo + ABT-737, n = 4 (males); Cdkn1a, Doxo versus DMSO, P = 0.0029 (females) and P = 0.0039 (males), whereas Doxo + ABT-737 versus Doxo, P = 0.0054 (females) and P = 0.0369 (males); Il6, Doxo versus DMSO, P = 0.0007 (females) and P = 0.0008 (males), whereas Doxo + ABT-737 versus Doxo, P = 0.0009 (females) and P = 0.066 (males); Gdf15, Doxo versus DMSO, P = 0.0111 (females) and P = 0.0011 (males), whereas Doxo + ABT-737 versus Doxo, P = 0.0164 (females) and P = 0.0015 (males); Spp1, Doxo versus DMSO, P ≤ 0.0001 (females) and P = 0.0012 (males), whereas Doxo + ABT-737 versus Doxo, P = 0.0002 (females) and P = 0.0037 (males); Serpine2, Doxo versus DMSO, P = 0.0015 (females) and P = 0.0025 (males), whereas Doxo + ABT-737 versus Doxo, P = 0.0026 (females) and P = 0.0025 (males); Ctsb, Doxo versus DMSO, P = 0.0019 (females) and P ≤ 0.0001 (males), whereas Doxo + ABT-737 versus Doxo, P = 0.0045 (females) and P ≤ 0.0001 (males); Fth1, Doxo versus DMSO, P ≤ 0.0001 (females) and P = 0.002 (males), whereas Doxo + ABT-737 versus Doxo, P ≤ 0.0001 (females) and P = 0.0064 (males); Thbs1, Doxo versus DMSO, P = 0.0004 (females) and P ≤ 0.0001 (males), whereas Doxo + ABT-737 versus Doxo, P = 0.0009 (females) and P ≤ 0.0001 (males)). f, Schematic of the hVSMC validation experiment. g,h, RT–qPCR analysis of proliferating (P) cells compared to those rendered senescent by exposure to IR, oxLDL, CoCl₂ (a hypoxia mimetic) or doxorubicin and cultured for an additional 7–10 days, or those rendered replicatively senescent (RS) by culture through exhaustion of population doublings. Traditional senescence markers CDKN1A, GDF15 and LMNB1 mRNAs (g) and new senescence markers SPP1, SERPINE2, CTSB, FTH1 and THBS1 mRNAs (h) were measured by RT–qPCR analysis (for g and h, P, n = 3; IR, n = 3; oxLDL, n = 3; CoCl₂, n = 3; Doxo, n = 3; RS, n = 3; CDKN1A, IR (P = 0.0136), oxLDL (P ≤ 0.0001), CoCl₂ (P = 0.0002), Doxo (P = 0.0022), RS (P ≤ 0.0001) versus P; GDF15, IR (P = 0.0004), oxLDL (P = 0.0049), CoCl₂ (P ≤ 0.0001), Doxo (P ≤ 0.0001), RS (P = 0.0165) versus P; LMNB1, IR (P ≤ 0.0001), oxLDL (P = 0.0049), CoCl₂ (P = 0.0002), Doxo (P ≤ 0.0001), RS (P ≤ 0.0001) versus P; SPP1, IR (P ≤ 0.0001), oxLDL (P = 0.0007), CoCl₂ (P = 0.0002), Doxo (P ≤ 0.0001), RS (P = 0.0022) versus P; SERPINE2, IR (P = 0.0011), oxLDL (P ≤ 0.0001), CoCl₂ (P = 0.0021), Doxo (P = 0.0002), RS (P = 0.0051) versus P; CTSB, IR (P = 0.0023), oxLDL (P ≤ 0.0001), CoCl₂ (P = 0.0006), Doxo (P = 0.0012), RS (P = 0.0004) versus P; FTH1, IR (P = 0.0002), oxLDL (P = 0.0012), CoCl₂ (P ≤ 0.0001), Doxo (P = 0.0002), RS (P = 0.0008) versus P; THBS1, IR (P = 0.0028), oxLDL (P = 0.002), CoCl₂ (P = 0.0139), Doxo (P = 0.0002), RS (P = 0.0006) versus P). In d and e, data represent the means ± s.d. from indicated biological replicates. Significance was established using one-way ANOVA for each sex. In g and h, data represent the means ± s.d. from n = 3 biological replicates. Significance was established using the Shapiro–Wilk test first, followed by Welch’s t test. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. Panels c and f created with BioRender.com.

Fig. 6 |. Spatial transcriptomic profiling confirms the vascular senescence signature in the cap and core of atherosclerotic plaques.

a, Schematic of experiments with p16–3MR mice on an Ldlr^−/− background. b, Brachiocephalic artery with ROIs labeled. ROIs include the cap, core, media and normal wall. c–h, Volcano plot representations of the RNAs identified by spatial transcriptomic sequencing in the cap versus normal wall (c) and the core versus normal wall (d) across all conditions; ROIs comparing the senolytic (Ldlr^−/−; 3MR + GCV) and atherosclerotic (Ldlr^−/− + GCV) conditions in the cap (e), core (f), media (g) and normal wall (h). i–l, Volcano plot representations of the RNAs identified by spatial sequencing in the core versus normal wall across all conditions in each ROI comparing the senolytic (Ldlr^−/−; 3MR + GCV) and atherosclerotic (Ldlr^−/−; 3MR + vehicle) conditions in the cap (i), core (j), media (k) and normal wall (l). m,n, Heatmap analysis of normalized RNAs from transcriptomes measured in the normal wall and cap (m) and the media and core (n) from the three conditions. Statistical significance was established using a Benjamini–Hochberg multiple test correction.

Fig. 7 |. smFISH visualization of vascular senescence-associated mRNAs in the atherosclerotic core.

a, Representative images of the atherosclerotic core from HFD and HFD + ABT-737 mice, stained for Spp1, Ctsb and Fth1 mRNAs and CD45 protein. b–d, Quantification of Spp1 (b), Ctsb (c) and Fth1 (d) mRNAs per field of atherosclerotic core HFD and HFD + ABT-737 sections. Columns from left to right in a: DIC (differential interference contrast) image (BF, brightfield); raw merged z stacks of the atherosclerotic cap for RNA probes labeled with Texas Red; raw merged z stacks for CD45 protein labeled with Cy5; merge of RNA (pseudocolored as red) and protein (pseudocolored as green) overlaid on DAPI staining (pseudocolored as blue). The scale bar is 5 μm. White arrows indicate representative cells positive for the described protein and RNA. Two mice per condition were used for a total of eight (HFD) and nine (HFD + ABT-737) independent field measurements for Spp1 mRNA, four (HFD) and eight (HFD + ABT-737) independent field measurements for Ctsb mRNA, and seven (HFD) and ten (HFD + ABT-737) independent field measurements for Fth1 mRNA. Significance was established using Student’s t test. **P ≤ 0.01.

Fig. 8 |. smFISH visualization of vascular senescence-associated mRNAs in the atherosclerotic cap.

a, Representative images of the atherosclerotic cap from HFD and HFD + ABT-737 mice, stained for Spp1, Ctsb and Fth1 mRNAs and αSMA protein. b–d, Quantification of Spp1 (b), Ctsb (c) and Fth1 (d) mRNAs per field of atherosclerotic cap HFD and HFD + ABT-737 sections. Columns from left to right in a: DIC image; raw merged z stacks of the atherosclerotic cap for RNA probes labeled with Texas Red; raw merged z stacks for αSMA protein labeled with Cy5; merge of RNA (pseudocolored as red) and protein (pseudocolored as green) overlaid on DAPI staining (pseudocolored as blue). The scale bar is 5 μm. White arrows indicate representative cells positive for the described protein and RNA. Two mice per condition were used for a total of eight (HFD) and six (HFD + ABT-737) independent field measurements for Spp1 mRNA, four (HFD) and five (HFD + ABT-737) independent field measurements for Ctsb mRNA, and four (HFD) and seven (HFD + ABT-737) independent field measurements for Fth1 mRNA. Significance was established using Student’s t test. **P ≤ 0.01.