Extended Data Fig. 5: Fibroblast subclustering reveals age-related cell proportion and expression changes (related to Fig. 4).

(a,d) Top enriched MSigDB Hallmark, Reactome, and Elsevier pathways gene sets of DE genes (a) and DA peaks (b) in cell clusters from 18 M vs. 3 M mice identified by hypergeometric testing. Upregulated DE genes with age are shown in red, downregulated in blue. (b) Genomic locations of significant DA peaks from 18 M vs. 3 M mice as annotated using Chipseeker. (c) Differential TF activity score with age. Significant differential motifs (P_adj < 0.05) are indicated by an asterisk. (e) Examples of DE genes with DA peaks in fibroblast clusters in 3 M vs. 18 M mice including: i) activation of Ets1, a TF that may contribute to senescent phenotypes and tumor invasiveness¹²⁴; ii) inactivation of Ace, an angiotensin I-converting enzyme gene; iii) activation of Ptges, which encodes a key enzyme in the expression of prostaglandin E2, expression. Ca molecule produced by cancer-associated fibroblasts have been shown to produce prostaglandin E2¹²⁵, and upregulation of PTEGS has been linked with hormone-dependent breast cancer growth by impacting estrogen feedback mechanisms¹²⁶; iv) activation of Enpp5, a transmembrane protein involved in nucleotide metabolism, is upregulated with age and exhibits peak opening in its promoter, and which is also overexpressed in triple negative breast cancer¹²⁷. Normalized values are shown for individual cells (Welch Two Sample t-test; ****P_Ets1 = 2.16 x10^-9, **P_Enpp5 = 0.0024, ****P_Ptges = 3.30x10^-13, ****P_Ace = 8.75x10^-12; cells with zero counts were excluded), along with a pie chart depicting the percentage of expressing cells vs. non-expressing cells. Pseudobulk snATAC-seq tracks in 3 M vs. 18 M mice are shown, along with gene structures and significant DA peaks with corresponding log₂fold changes (FC) values. (f) UMAP visualization of stromal subclusters Fib-C0-C5 and Str-C6-C10 captured by scRNA-seq, shown per age for 18 M and 3 M mice (f). (g,h) Expression of canonical marker genes in scRNA-seq stromal subclusters (g) along with top marker genes for each subcluster (h). Stromal populations could be classified into eleven subclusters (Fib-C0-C5 and Str-C6-C10): C0-C5 express fibroblast marker genes (Col1a1 + , Pdgfra + ), Fib-C5 and Str-C6 express pericyte marker genes (Rgs5 + , Des + ), Str-C7 expresses markers of skeletal muscle satellite cells (Pax7 + , Bmp4 + ), and Str-C8, Str-C9, and Str-C10 express vascular marker (Pecam1 + , Eng + ). The vascular subclusters can be further subdivided based on expression of Sox17 (Str-C8), Sele (Str-C9), and lymphatic vascular markers (Str-C10)¹²⁸. Note that while the expression of pericyte and fibroblast markers by Fib-C5 (Rgs5 + , Des + ) was suggestive of a doublet cluster, these cells specifically expressed inflammatory and contractile markers, markers of fibroblastic reticular cells (Ccl19 + ) and of mesenchymal stromal and osteolineage cells (Cxcl12 + ), potentially suggesting a specialized identity³⁸. (i) Feature plots for selected marker genes in scRNA-seq fibroblast subclusters. (j) Expression of senescence marker genes in scRNA-seq fibroblast subclusters per age. The proportions of fibroblast positive cells from 3 M and 18 M mice are shown on the right. (k) CAF gene expression signatures in 3 M vs. 18 M mice across fibroblast subclusters (Wilcoxon test; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001,****P ≤ 0.0001) with feature plots colored by expression of the signature.