Senescence-associated gene pathways are differentially expressed in equine aging-related osteoarthritis

Abstract

Objective

Osteoarthritis (OA) is a aging-associated degenerative joint disease. The objective was to determine relative senescence gene expression in joints and leukocytes of OA horses toward considering senotherapeutics to manage OA.

Methods

To define local (joint) and systemic (peripheral blood mononuclear cells [PBMCs]) senescence burden, synovial fluid cell single-cell RNA sequencing and PBMC mRNA sequencing datasets (n = 65 samples) were examined. Differential analyses were conducted using limma to compare OA versus control. A custom 3,043-gene senescence set curated from published metadata was applied to differential analyses to investigate senescence-specific pathways. Senescence genes were divided into 8 categories; scores were calculated with fast gene set enrichment analysis with P value computed via permutation and log₂ fold change ranks.

Results

Synovial fluid single-cell RNA sequencing data revealed cell type-specific heterogeneity in senescence gene expression. Fast gene set enrichment analysis pathway analysis confirmed enrichment/upregulation in inflammatory and stress-induced senescence in dendritic, cycling, CD8 T, and gamma delta T cells. Senescence-associated secretory phenotype pathways were predominantly represented in cycling cells. Senescence genes aryl hydrocarbon receptor (AHR), IL-1 receptor antagonist (IL1RN), heme oxygenase 1 (HMOX1), plasminogen activator, urokinase receptor (PLAUR), and tissue inhibitor of metalloproteinase 1 (TIMP1) were upregulated in multiple synovial fluid cell types. In contrast, genes in most senescence categories were downregulated in PBMCs.

Conclusions

Senescence pathways were differentially expressed in aged horses with OA, with upregulation of senescence genes in the joint and downregulation in PBMCs.

Clinical Relevance

Therapeutic strategies targeting senescent cells may be a disease-modifying strategy to treat equine OA.

Osteoarthritis (OA) is the most prevalent joint disease in horses, leading to chronic pain and disability.^1,2 Osteoarthritis is characterized by synovial inflammation, progressive degeneration of articular cartilage, and subchondral bone remodeling, culminating in joint stiffness and pain.³ Risk factors include aging, joint trauma, obesity, and genetic predisposition. Current management strategies including systemically administered (NSAIDS, pentosan polysulfate sodium, polysulfated glycosaminoglycans, and hyaluronate sodium) and IM-injected therapies (corticosteroids, hyaluronic acid, polyacrylamide hydrogels, and biological therapies) provide what is generally considered to be relatively short-lived symptom-modifying effects.⁴ Despite extensive investigation, the molecular underpinnings of OA remain incompletely understood. The lack of disease-modifying treatments for equine OA highlights the need to further explore mechanisms of OA progression toward improved therapeutic interventions.

Cellular senescence is a biological process associated with aging that is increasingly recognized as a key factor in human OA pathophysiology.⁵ Senescent cells enter a state of permanent growth arrest, while remaining metabolically active and secreting proinflammatory factors known as the senescence-associated secretory phenotype (SASP). This phenotype includes inflammatory cytokines, chemokines, matrix-degrading enzymes, and growth factors, all of which contribute to adjacent tissue dysfunction and chronic inflammation. While senescence can serve as a protective mechanism to prevent the proliferation of damaged cells, the accumulation of senescent cells in joint tissues disrupts homeostasis and accelerates degenerative processes.⁶ These affected cells exhibit elevated expression of markers such as p16INK4 A, p21CIP1, and senescence-associated β-galactosidase.⁷ Experimental studies⁸ provide evidence for the role of senescent cells in OA pathogenesis affecting chondrocytes, synovial fibroblasts, and osteoblasts. Notably, senescent cells have been found to be localized in sites of cartilage degradation, leading to a reduced ability to synthesize extracellular matrix and impairing the tissue’s capacity for self-repair. Secretion of SASP factors (eg, IL-6, IL-8, matrix metalloproteinase [MMP]-13) promotes chronic inflammation, contributing to synovial inflammation and fibrosis and cartilage breakdown, further exacerbating joint dysfunction. In the subchondral bone, senescent osteoblasts disrupt bone remodeling dynamics, leading to increased bone sclerosis and the formation of osteophytes. The accumulation of these dysfunctional cells perpetuates a cycle of inflammation, oxidative stress, and tissue degradation, characteristic of OA progression. The spread of senescence through paracrine signaling further amplifies these effects, as SASP factors can induce neighboring cells to enter a senescent state, thus expanding the pool of nonfunctional, inflammatory cells in the joint environment.⁹

Given the strong association between cellular senescence and OA progression in humans and laboratory species, targeting senescence may also represent an adjunctive therapeutic approach to manage OA in horses. Strategies such as senolytic drugs, which selectively eliminate senescent cells, and senomorphic agents, which suppress SASP factors, are under investigation as potential disease-modifying treatments.¹⁰ However, the association of cellular senescence in equine OA specifically has not been previously reported. In this study, we aimed to define the burden of senescence-associated genes in equine OA toward future assessment of senotherapeutics to mitigate progression. Specifically, we assessed the transcriptome of cells found in the synovial fluid (SF) and circulating peripheral blood mononuclear cells (PBMCs) in horses with OA compared to control horses for senescent gene expression using a 3,043-gene signature curated from metadata studies^11-15 that explored senescence gene signatures in high-throughput sequencing data across over 30 different cell types in human subjects. We hypothesized that horses with OA would have elevated senescent-associated patterns, which would be more prominent in SF cells compared to circulating leukocytes and provide specific gene targets for novel therapeutic or disease biomarkers.

Methods

Study design and population

To assess local gene expression in joints, single-cell transcriptomic data previously published from SF cells from osteoarthritic (n = 3) and control joints (3) were evaluated.¹⁶ To assess circulating leukocyte gene expression, mRNA sequencing data from peripheral blood mononuclear cells (PBMCs) from OA (n = 20) and 39 control horses (39) were evaluated (Unpublished dataset can be found in GEO public genomic data repository, accession number GSE313902). Horses with naturally occurring OA were enrolled as previously described (Unpublished dataset can be found in GEO public genomic data repository, accession number GSE313902). Inclusion criteria for the OA group included a clinical diagnosis based on lameness examination, radiographic evidence of joint degeneration, and/or gross pathological changes observed at the time of sampling. Control horses were free of clinical lameness and had no radiographic or gross evidence of joint disease; horses with a history of musculoskeletal injury or systemic inflammatory conditions were excluded. Horses with OA had a median age of 18 years (range, 2 to 25 years), and control horses had a median age of 3 years (range, 1 to 6 years). Protocols were approved by Colorado State University IACUC and Clinical Review Board for enrollment of sample collection from university-owned research horses and client-owned horses, respectively (protocol No. 926 and No. 3375). The study design is summarized in Figure 1.

Figure 1—

Study overview. Samples were collected from horses with and without osteoarthritis (OA), including synovial fluid (n = 3 OA joints; 3 control joints) and peripheral blood mononuclear cells (PBMCs; 20 OA; 39 control). Synovial fluid single-cell RNA sequencing (scRNAseq) identified 8 cell types that were aggregated to pseudobulk for differential expression with limma-voom, while PBMC mRNA sequencing was trimmed mean of M values (TMM)-normalized and analyzed with limma-voom for differential expression (DE). Cross-tissue analyses included differential expression, fast gene set enrichment analysis (FGSEA) pathway enrichment, and Database for Annotation, Visualization, and Integrated Discovery annotation, and senescence gene expression was evaluated using 3,043 equine senescence orthologs. Figure was created in BioRender. GSEA = Gene set enrichment analysis.

Curation of senescence gene list

The senescence gene set (3,043 genes total) was developed and combined from authoritative data-bases, including GenAge: the Aging gene database (https://genomics.senescence.info/genes/index.html) and CellAge: the Database of Senescence Genes (https://genomics.senescence.info/cells.), as well as published datasets of senescence gene signatures.^11-15 The published datasets cited all contained machine learning models that utilized historical human datasets to define senescence gene signatures in primary cells as well as immortalized cell lines. Wang et al¹⁷ specifically contained multiple curated gene lists from SenMayo, CellAge, GenAge, ASIG, SASP, AgingAtlas, SenUp, and SigRS. The complete list is summarized in Supplementary Material S1. Human senescence genes were mapped to equine orthologs using Ensembl Genes 115.¹⁸

Statistical analysis

Single-cell (pseudobulk) analysis—Cell type-specific pseudobulk count matrices were exported from Ammons et al,¹⁶ downloaded, and imported into RStudio (RStudio 2025.05.1 + 513; R version 4.5.0; Posit).¹⁹ Within each annotated cell type, cell-level unique molecular identifiers were summed per sample to create a sample-by-gene matrix. Low-count genes were filtered for a total count < 5 before modeling. Counts were transformed with voom to estimate the mean-variance relationship and precision weights, after which OA versus control differences were tested using limma linear models with empirical Bayes moderation.²⁰ Single-cell pseudobulk volcano plots by cell type were visualized at absolute log₂ fold change (log₂FC) ≥ 0.5 with raw P value ≤ .05 due to the small sample size (n = 3 per contrast). All analyses were performed in R Core Team using edgeR, limma, readr, dplyr, tibble, and ggplot2.^20-25

Bulk RNA-sequencing analysis—The bulk PBMC count matrix and associated metadata (OA vs controls; ZJ Williams, R Impastato, GK Piquini, et al, unpublished data, 2025) were compiled and downloaded and then imported into RStudio. Lowly expressed genes were filtered out for total counts < 5 across all samples. Library size normalization was performed using the trimmed mean of M values method implemented in edgeR to account for differences in sequencing depth. The “OA versus control” contrast was tested with limma’s empirical Bayes moderated t statistics. Quality control and ordination were performed on voom-transformed expression values to generate principal component analysis and sample-distance heatmaps. All analyses were performed in RStudio using edgeR, limma, readr, dplyr, tibble, and ggplot2. Volcano plots were generated with ggplot, with a significance threshold defined as log₂FC ≥ 0.5 or ≤ 0.5 and raw P value ≤ .05. The raw P value was used for significance filtering to match the single-cell analysis.

Senescence gene filtering and cross-species mapping—Following differential expression analysis with limma/EdgeR, significantly differentially expressed genes (DEGs; absolute log₂FC ≥ 0.5; P value ≤ .05) in the OA versus control comparison from both bulk and single-cell analyses were filtered against a senescence gene list assembled as described (complete list included in Supplementary Material S1). Counts for total DEGs and senescence-specific DEGs were summarized and labeled for each cell type (cycling cell, dendritic cell (DC), macrophage, neutrophil, B cell, CD4/CD8 T cell, gamma delta T cell, and mixed PBMCs.

Pathway analyses—Preranked gene set enrichment analysis (GSEApre) was conducted using only the senescence-specific significant DEGs from single-cell and bulk RNA-sequencing results.²⁶ The GSEApre was run using the weighted enrichment statistic, with filtering for a minimum of 5 genes per pathway. The MsigDB pathway collections (https://www.gsea-msigdb.org/gsea/msigdb) Hallmark, Kyoto Encyclopedia of Genes and Genomes (KEGG), Reactome, Biocarta, Pathway Interaction Database (PID), and WikiPathways were used. Results reporting included normalized enrichment score (NES), nominal P value, and false discovery rate (FDR; adjusted) q value.

Senescence-specific pathway scoring—For targeted senescence-focused interrogation in both bulk and single-cell analyses, Fast gene set enrichment analysis (FGSEA) was performed in RStudio on 8 senescence-related gene sets: stress-induced, replicative, oncogene-induced, inflammatory, metabolic, oxidative-stress response, DNA damage response, and SASP. Genes were first categorized according to Avelar et al²⁷ (CellAge). We then expanded the lists to include genes that mapped to associated pathways and related keywords from MsigDB. The following keywords were used to categorize genes: SASP combined HALLMARK INFLAMMATORY RESPONSE, HALLMARK TNFA SIGNALING VIA NFKB, HALLMARK IL6 JAK STAT3 SIGNALING with C2:CP/GO:CC sets containing SASP, secre/secret(exocytosis/secretory), exocyt, vesicle, granule, chemokine, cytokine, ECM, matrix, and CC terms for extracellular matrix/vesicle/lysosome/endosome/secretory compartments. DNA damage response merged HALLMARK DNA REPAIR, HALLMARK P53 PATHWAY, HALLMARK G2M CHECKPOINT with C2:CP/GO:BP sets containing DNA damage, DNA repair, checkpoint, ATM, ATR, BRCA, RAD51, telomer, and GO repair subclasses (nucleotide/base excision, mismatch, homologous recombination, non-homologous end joining). Oxidative stress response used HALLMARK REACTIVE OXYGEN SPECIES PATHWAY and oxidative phosphorylation together with C2:CP/GO:BP sets containing oxidative, redox, reactive oxygen, antioxid. Replicative senescence/telomere drew from C2:CP/GO:BP sets containing telomer (telomere), shelterin, telomerase, chromosome capping, and end protect. Stress-induced senescence integrated HALLMARK APOPTOSIS and HALLMARK UNFOLDED PROTEIN_RESPONSE with C2:CP/GO:BP terms for stress-activated MAPK, SAPK/JNK, p38, MAPK cascade, ER stress, UPR, and endoplasmic reticulum stress. Oncogene-induced senescence combined HALLMARK E2F TARGETS, HALLMARK MYC TARGETS (V1/V2), HALLMARK G2M CHECKPOINT with C2:CP sets containing RAS, PI3K/AKT, MAPK signaling, cell cycle, RB, E2F, and RAF/MEK/ERK. Inflammatory senescence/NF-κB merged the two Hallmark inflammatory sets with C2:CP/GO:BP terms containing NF-κB/IKB, inflammasome, TLR, and cytokine signaling. Metabolic senescence united HALLMARK MTORC1 SIGNALING, fatty acid metabolism, glycolysis, cholesterol homeostasis, and adipogenesis with C2:CP/GO:BP terms containing AMPK, oxidative phosphorylation, citrate/TCA, glycolysis, fatty acid, cholesterol, lipid metabolism, autophag/lysosome, mitochondr, one-carbon/serine-glycine, peroxisome. A list of genes including categorization is included in Supplementary Material S2. The FGSEA “fgseaMultilevel()” was then used to generate NESs for each cell type using the senescence-specific significant DEGs from each cell type.²⁸ The parameters used for filtering include a minimum of 2 hits per pathway, P value computed via permutation or multilevel Monte Carlo sampling, and significance defined as P value ≤ .05 as indicated in Figure 3.

Figure 3—

Pathway summary of senescence differentially expressed genes (DEGs) by cell type from scRNAseq data. A—Senescence pathway gene sets were defined using CellAge and MsigDB as described in Methods, and normalized enrichment scores (NES) were generated using FGSEA for each cell type. The NES is shown in a color scale, cell type is on the x-axis, and senescence category is on the y-axis. *P ≤ .05, **P ≤ .01, ***P ≤ .001, significant difference. B—Heatmap of 362 total senescence DEGs from all synovial fluid cell types, with log₂ fold change (log₂FC; OA vs control) shown in a color scale. Genes not represented in the cell type are grayed out. C to G—Pathway analysis of senescent DEGs from each cell type. Senescence DEGs from each cell type were imputed separately into preranked GSEA, the number of genes was mapped to each pathway shown in bubble size, the P value is on a color scale, and NES is shown on the x-axis. The y-axis labels pathway names. KEGG = Kyoto Encyclopedia of Genes and Genomes. SASP = Senescence-associated secretory phenotype. WP = WikiPathways.

Results

Cycling and DCs account for the majority of senescence gene changes in SF

Out of the 8 different cell types found in SF single-cell RNA sequencing, as previously described in detail by Ammons et al,¹⁶ the cycling cell population contained the most differentially expressed genes associated with senescence (130 upregulated and 162 downregulated), followed by DCs with 42 upregulated and 40 downregulated significant senescence genes (Figure 2). The remaining cell types had relatively low numbers of significant DEGs that were overlapping with the senescence gene dataset. The most represented senescence-associated gene prelamin-A_C (LMNA), which was upregulated in 7 cell types, followed by secreted phosphoprotein 1 (SSP1), and IL-1 receptor antagonist protein (IL1RN), which was upregulated in 6 cell types (Supplementary Material S3).

Figure 2—

Senescence gene expression in scRNAseq data from synovial fluid cells. Pseudobulk counts from each cell type found in synovial fluid were used for differential analysis between 3 OA-affected joints and 3 control joints. A—Graph shows total counts of significant DEGs in gray, significant senescence genes upregulated (red), and significant senescence genes downregulated (blue). Significance was defined as log₂FC ≥ 0.5 or ≤ −0.5 and raw P value ≤ .05. B and C—Volcano plots of differential analysis results comparing OA to control joints, numbers show total up- or downregulated DEGs, and red or blue dots show the genes that are included in the senescence list. The top 50 significant genes by raw P value are labeled. D to I—Volcano plots of the other 6 cell types as labeled on top, with all the significant senescence DEGs labeled in red for upregulated and blue for downregulated.

Cell type-specific senescence pathways reveal distinct upregulation in OA samples

To summarize the senescence DEGs in each cell type, senescence enrichment scores were calculated for each cell type based on the senescence classifications in “CellAge,” as well as custom lists from MsigDB. There were 1,092 genes out of the curated 3,042 senescence gene list mapped to at least 1 of 8 senescence categories (Supplementary Material S2). The “SASP genes” were upregulated most commonly in cycling cells (Figure 3); cycling cells had the most DEGs overlapping with the senescence gene set, with the most significant enrichment. Genes related to inflammatory senescence were also significantly upregulated in cycling cells. Several of the other single-cell types, despite having lower numbers of senescence DEGs, also had significant upregulation in the OA inflammatory senescence/NF-κB (DC and gamma delta T cell), stress-induced senescence (CD8 T cell), and metabolic senescence (macrophage).

To provide more information on the significant changes, gene sets that are not specific to senescence were also explored. Using the senescence DEG lists from each cell type, we mapped each list to MsigDB pathways using GSEApre. Cycling cells contained the most significant pathways; cycling cell senescence DEGs in the OA samples downregulated few immune related pathways such as interferon gamma, alpha, and RIGI (Figure 3) but largely showed strong upregulation of inflammatory pathways such as tumor necrosis factor-α, Toll-like receptor, lysosome, and hallmark inflammatory response, with a large number of genes (60 total) mapping to “innate immune system.” In contrast to the many upregulated pathways in cycling cells, DC senescence DEGs showed upregulation (OA group) of a few pathways such as IL-10, epithelial-mesenchymal transition, complement, and IL-18. Senescence DEGs were both significantly enriched in CD8 T cells and macrophages (upregulated, 6 genes) in the Reactome “cellular response to stimuli” pathway, while the other single-cell types did not have any significant pathways (P value ≤ .05; FDR P value ≤ 0.25).

Senescence genes are differentially expressed in circulating PBMCs

Applying the senescence gene list of 3,046 to the bulk RNA-sequencing PBMC data showed OA samples had significant upregulation of 167 and downregulation of 235 (P ≤ .05; absolute log₂FC ≥ 0.5) senescence genes out of a total of 2,809 significant DEGs (Figure 4). The highest upregulated genes (> 2 log₂FC) included stanniocalcin-1 (STC1) and glypican-3 (GPC3), and the most downregulated senescence genes (< −2 log₂FC) included IL-1A, galanin receptor 2 (GALR2), Wnt family member 1 (WNT1), Fos-like 1 (FOSL1), Ras related glycolysis inhibitor and calcium channel regulator (RRAD), CCL20, bone morphogenetic protein 2 (BMP2 ), and cyclin dependent kinase inhibitor 1 A (CDKN1 A), all with log₂FC < −2. The FGSEA analysis of custom senescence categories showed negative enrichment (downregulation in OA) across most senescence pathways, including several with significant P values (P ≤ .05): inflammatory senescence/NF-κB (NES, −1.22), oncogene-induced senescence (NES, −1.23), DNA damage response (NES, −1.27), oxidative stress response (NES, −1.30), and replicative senescence/telomere (NES, −1.33). Only a single category, “metabolic senescence,” indicated upregulation but was not significant. To further categorize the significant senescence DEGs, GSEA was then used, which identified 194 (KEGG, Reactome, Biocarta, PID, and Wiki) significantly enriched pathways (FDR ≤ 0.25; P value ≤ .05 minimum 5 gene). The only 3 upregulated pathways included neutrophil degranulation sensory and adhesion, whereas the remaining (eg, tumor necrosis factor, interleukin signaling, Tolllike receptor, inflammatory response, IL-18, nonobese diabetic, NF-κB, and inflammatory response) were all downregulated in OA categories.

Figure 4—

Circulating senescence genes in PBMC bulk RNA-sequencing data. Circulating senescence genes in PBMC bulk RNA-sequencing data. Bulk RNA-sequencing data from 20 OA and 39 OA horses were used for differential analysis. A—Principal component analysis (PCA) plot of the top 5,000 most variable genes, TMM normalized and voom transformed. Red dots are for OA samples, and blue dots are for control samples. B—Volcano plot of differential analysis results comparing OA to control PBMCs where numbers show all significant upregulated and all downregulated and blue and red dots show either upregulated senescence genes or downregulated senescence genes, Significance is defined as log₂FC ≥ 0.5 or ≤ −0.5 and raw P value ≤ .05. Top 50 significant senescence genes (by P value) are labeled with gene name. C—Total significant gene counts for all OA versus control (gray), upregulated senescence genes (red), and downregulated senescence genes (blue). D—FGSEA scores for senescence pathway gene sets, with NES score shown in a color scale. Significant categories with P value ≤ .05: *P ≤ .05; **P ≤ .01; ***P ≤ .001. E—Significant senescence genes were used for preranked GSEA analysis, top 30 pathways with false discovery rate adjusted P value < 0.25 and raw P value ≤ .05 are shown in the graph, and bars (x-axis) show NES scores with negative for downregulated and positive for upregulated. Color scale shows P value with gene set sizes (total genes mapped to pathway) represented as bubble size.

Discussion

This study provides the first comprehensive evidence for cellular senescence in equine OA pathogenesis, identifying upregulation of key senescence-associated pathways such as SASP, metabolic senescence, stress-induced senescence, inflammatory senescence, and replicative senescence upregulation in immune cells found in SF. Previous studies^29-33 across multiple species have established cellular senescence as a key driver of OA pathogenesis, with work in humans demonstrating that OA represents a manifestation of accelerated aging processes affecting multiple tissues simultaneously. Matrix remodeling dysregulation through coordinated changes in metalloproteinase activity results in self-perpetuating cycles of tissue damage through paracrine signaling, where SASP factors from one cell type induce senescence in neighboring cells.³¹ In the current study, immune cell populations in SF with significant upregulation of senescence-associated genes included cycling cells, macrophages, DCs, gamma delta T cells, and CD8 T cells.

Macrophages have been a commonly studied immune lineage in OA pathogenesis, with large scale single-cell sequencing studies indicating increased “proinflammatory” macrophages in the cartilage, meniscus, and synovium of OA patients.³⁴ Inflammatory or enhanced M1 macrophages recognize degraded synovial tissue that act as danger-associated molecular patterns and in turn continue to secrete cytokines, MMP, and tissue inhibitors of metalloproteinases that lead to further cartilage degradation and exasperation of OA.³⁵ Subtypes of SF-derived macrophages beyond the traditional M1/M2 classification have also been thought to be primary drivers of pain and T-cell responses in OA.^{17, 36,37} The macrophage population in the equine OA samples in this study strongly upregulated “metabolic senescence,” with genes such as malic enzyme 1, NADP dependent, cytosolic (ME1), LMNA, urokinase plasminogen activator receptor (PLAUR), and heat shock protein family a member 5 (HSPA5 ) leading the heightened pathway scoring. These genes have all been linked to macrophage activation and proinflammatory status in other disease states.^38-42 The upregulation of the metabolic senescence pathway in the equine OA macrophages mirrored findings in small animal and human data, which support the development of pathway/cell type-specific therapies that can be tested in the large animal model.

Dendritic cells have to date been a lesser studied population in OA, although experimental OA models have shown increases in DC numbers in induced OA.⁴³ Furthermore, subsets of DCs have been found to influence proinflammatory cytokine production and Th17 differentiation, thus exacerbating cartilage degradation.^44,45 In concordance with human clinical data and small animal models, the equine single-cell dataset DC population in this study indicated a significantly upregulated inflammatory senescence/NF-κB category, with leading genes such as CXCL8 (IL-8), tissue inhibitor of metalloproteinase 1 (TIMP1), IL1RN, and PLAUR driving the upregulation in OA samples. The increased inflammatory senescence in DCs could also be associated with the upregulation of inflammatory senescence in the gamma delta T-cell population, with increased heme oxygenase-1 (HM0X1) and granulocyte-macrophage colony-stimulating factor 2 (CSF1) leading to the high pathway score.

Cycling cells were defined in the context of equine OA in Ammons et al¹⁶ by expression of genes involved in cellular turnover and cell cycle progression, functional categories that directly overlap with senescence-associated pathways. This overlap in defining features between cycling cells and our senescence pathway genes could explain why cycling cells show high SASP and inflammatory senescence scores. Unfortunately, the cycling cells were composed of a “heterogenous” cell population making it difficult to draw mechanistic conclusions as to their association and role in OA in this study.

Evidence for senescence in equine cells in contexts other than OA indicates that reversing cellular senescence in horses may have benefits beyond the treatment of musculoskeletal disease.^46,47 Age-related alterations affect chondrogenic differentiation capacity of SF mesenchymal stromal cells (MSCs) and telomerase activity in articular chondrocytes.⁴⁸ Cellular proliferation capacity, viability, and regeneration potential of equine bone marrow and adipose tissue-derived MSC decline with increased donor age.⁴⁹ Advanced aging in horses has been shown to affect immune cell function, altering T-cell divisional history and inflammatory cytokine production.^50,51 Mitogenic stimulation and cell cycle arrest induce senescence in cartilage explants and coordinated senescence pathway activation across diverse synovial cell lineages.⁵² Chondrocytes isolated from joints with osteochondritis-dissecans have been shown to display enhanced senescence and oxidative stress, altered autophagy mechanisms, and enhanced expression of matrix-degrading enzymes (eg, MMP-13).⁵⁵ Aging has been shown to affect fundamental cellular signaling pathways in equine articular chondrocytes including insulin-like growth factor 1 and IL-1β signaling.⁵⁴ Metabolic disorders such as equine metabolic syndrome further impact viability, senescence, and stress factors of MSCs, providing new insights into how systemic conditions accelerate cellular aging in horses.^55-57 These findings support a role and potential broader impact of exploring interventions targeting senescence pathways in horses, particularly older horses with metabolic conditions.

Senotherapeutic approaches represent a rapidly evolving field with the potential to mitigate the impact of senescent cell burden with age-related pathologies and improve healthspan. Research to date has focused on 2 main categories: senolytic drugs to selectively eliminate senescent cells and senomorphic agents to suppress SASP factor production without killing senescent cells.⁵⁸ In the context of OA, transplantation of senescent cells into mouse knee joints induced OA-like pain, cartilage degeneration, and radiographic changes.⁵³ Selective elimination of senescent cells with senolytics reduced senescent cell accumulation, improved tissue regeneration, and restored cartilage matrix production in chondrocytes.³³ Human clinical trials are ongoing, with the combination of dasatinib and quercetin among the most extensively studied.³⁷ Detection of senescence signatures in horses in this study may provide opportunities for clinical translatability; however, establishing this as a reliable biomarker will require direct correlation of expression levels with clinical OA scores and disease progression metrics. The cell type-specific nature of senescence vulnerability and transcriptional regulators suggests that precision approaches may be more effective than broad senolytic treatments, potentially minimizing off-target effects. Further evaluation of selection, dose, and scalability of senotherapeutic approaches in horses is indicated before clinical application.

Caveats to this study warrant further discussion, including a relatively small sample size, lack of matched samples across multiomic analysis, and predominance of aged samples in the OA population, which could explain the heavily downregulated senescence pathways in the bulk PBMC sequencing dataset. It is important to note that differential gene expression of senescence-related genes does not necessarily indicate that those cells were senescent. Sample sizes evaluated here, similar to those previously reported in equine OA studies, are acknowledged to be relatively small, necessitating validation in larger cohorts. Given that the most robust senescence burden was detected locally within joint tissues, further evaluation of IA or synovium-targeted therapies is particularly warranted, as systemic approaches alone may not adequately address localized disease drivers. The findings of this study cannot be extrapolated to all clinical manifestations of OA, which may vary widely. This was a descriptive cross-sectional study describing associations of senescence burden and differential gene expression in equine OA and does not infer causality in progression.⁵⁹ The overall goal was to integrate multiomics data from multiple sources toward future studies investigating targeted interventional strategies using senotherapeutic agents. It was not the primary objective to tease out in this analysis whether age versus OA was the primary contributing factor to senescence burden due to the dichotomized nature of the older OA group compared to the younger control horses. The observation that several senescence signatures were higher locally in OA joints compared to control healthy joints supports a role for senescence in OA pathogenesis amplified locally in OA-affected joint tissues. Continued enrollment of cases to delineate the impact of other variables such as age, breed, and metabolic status from OA on senescence burden are warranted and may guide future interventional studies evaluating the route of administration and pharmacokinetics of senotherapeutic agents.

In summary, this analysis established a possible association of cellular senescence with equine OA pathogenesis. The identification of specific molecular signatures and cell populations may allow future development of senotherapeutic interventions that may offer dual-targeting opportunities for more effective treatment approaches. These findings contribute to the growing body of evidence supporting senescence as a key therapeutic target in musculoskeletal disease and provide a foundation for translating senolytic approaches from laboratory models to clinical applications in large animal species.

Supplementary Material

Supplementary materials are posted online at the journal website: avmajournals.avma.org.