Transfer learning of an in vivo-derived senescence signature identifies conserved and tissue-specific senescence across species and diverse pathologies

Abstract

Senescent cells (SnCs) contribute to normal tissue development and repair but accumulate with aging where they are implicated in a number of pathologies and diseases. Despite their pathological role and therapeutic interest, SnC phenotype and function in vivo remains unclear due to the challenges in identifying and isolating these rare cells. Here, we developed an in vivo-derived senescence gene expression signature using a model of the foreign body response (FBR) fibrosis in a p16Ink4a-reporter mouse, a cell cycle inhibitor commonly used to identify SnCs. We identified stromal cells (CD45-CD31-CD29+) as the primary p16Ink4a expressing cell type in the FBR and collected the cells to produce a SnC transcriptomic signature with bulk RNA sequencing. To computationally identify SnCs in bulk and single-cell data sets across species and tissues, we used this signature with transfer learning to generate a SnC signature score (SenSig). We found senescent pericyte and cartilage-like fibroblasts in newly collected single cell RNAseq (scRNASeq) data sets of murine and human FBR suggesting populations associated with angiogenesis and secretion of fibrotic extracellular matrix, respectively. Application of the senescence signature to human scRNAseq data sets from idiopathic pulmonary fibrosis (IPF) and the basal cell carcinoma microenvironment identified both conserved and tissue-specific SnC phenotypes, including epithelial-derived basaloid and endothelial cells. In a wound healing model, ligand-receptor signaling prediction identified putative interactions between SnC SASP and myeloid cells that were validated by immunofluorescent staining and in vitro coculture of SnCs and macrophages. Collectively, we have found that our SenSig transfer learning strategy from an in vivo signature outperforms in vitro-derived signatures and identifies conserved and tissue-specific SnCs and their SASP, independent of p16Ink4a expression, and may be broadly applied to elucidate SnC identity and function in vivo.

Cellular senescence promotes progenitor cell expansion during axolotl limb regeneration

“…Although mechanistic links between ribosome biogenesis and senescence induction have been previously established, this represents a relatively under-explored aspect of senescence and its prevalence in other in vivo contexts merits further investigation. Further, we uncover the heterogeneous subpopulations that comprise the total bSC population (Figure 3J) and find that bSCs consist predominantly of cells of connective tissue and macrophage-origins, with minor contributions from other cell types, consistent with previous reports of retention of cell-type identity following senescence induction (Cherry et al, 2022). As our characterizations relied on nanoparticle-based isolation, there is a possibility that macrophages could be non-specifically labelled due to their phagocytic activity (Hall et al, 2017).…”

Age-associated Senescent - T Cell Signaling Promotes Type 3 Immunity that Inhibits Regenerative Response

“…To identify the signaling components that maybe responsible for impaired wound healing in old animals, and evaluate the mechanistic of type 3 immune skewing with aging, we investigated age-associated TF activation in T cells and their communication in the signaling networks. We developed a transfer learning algorithm to identify senescent cells in single cell datasets using a senescence signature derived from a p16-Cre;Ai14 reporter model 52 . Transfer learning identified 3 stromal clusters expressing a high senescent signature, with the Mgp hi fibroblast cluster showing the highest senescence signature enrichment ( Fig.…”