Synovial Dysregulation in Ankle Osteoarthritis: Molecular Insights and Pathogenetic Pathways

Abstract

Osteoarthritis of the ankle is a debilitating condition affecting nearly 20% of adults in the United States. The tissue, cellular, and molecular basis of this disease remains poorly understood, particularly the role of the synovial membrane. This prospective case-control study examined the gene expression profile of synovial tissue in ankle osteoarthritis to uncover biological pathways altered in the disease. Synovial biopsies were obtained from twenty-six patients with ankle osteoarthritis undergoing joint replacement and eight patients with minimal or no evidence of osteoarthritis. Bulk RNA-sequencing and differential gene expression analysis were performed, followed by biological pathway enrichment, unsupervised cluster, and network-based analysis. Synovium from ankle osteoarthritis patients showed distinct gene expression patterns compared to controls, with enrichment in pathways related to extracellular matrix remodeling and tissue morphogenesis. Age and a prior history of ankle fracture were identified as significant factors influencing the transcriptomic profile of osteoarthritic synovium. After accounting for these variables, transcriptomic clustering perfectly distinguished osteoarthritic from non-osteoarthritic synovium and revealed upregulation of pathways involved in muscle development, neural signaling, immune response, and carbohydrate metabolism. Conversely, pathways related to ribosome production, protein synthesis, and cellular signaling were downregulated. Multiple WNT signaling pathways remained significantly enriched in osteoarthritic ankle synovium. This study identifies a distinct gene expression profile in the synovium of ankle osteoarthritis patients, highlighting key molecular pathways possibly involved in disease progression.

Introduction

Osteoarthritis (OA) is a debilitating condition affecting over 300 million people worldwide.¹ While the knee and hip are most commonly affected, ankle OA accounts for 1-4% of OA cases,² often following trauma² and progressing more rapidly than OA in other lower limb joints.³ Despite its significant impact, ankle OA remains understudied, with only 2 of 43 studies in a systematic review of lower limb OA specifically addressing the ankle.⁴

Current OA treatments are palliative, focusing mostly on symptom management, as no clinically viable disease-modifying approaches are available.⁵ Research often extrapolates findings from knee and hip OA to other joints, including the ankle, overlooking critical differences in biomechanics, cartilage properties, and etiologies.⁶ Specifically, ankle cartilage is thinner and is characterized by a denser extracellular matrix (ECM) with higher proteoglycan and water content than knee cartilage.⁶ These distinctions, coupled with the earlier onset and trauma-driven etiology of ankle OA, underscore the need for joint-specific research and treatment approaches.

OA has long been viewed as mechanically induced wear and tear of the cartilage⁷. Newer evidence suggests that a myriad of molecular pathways and processes constitute the etiology of OA disease development, and disease initiation involves contributions of tissues such as subchondral bone⁸ and synovium⁹. Regarding the ankle specifically, little is known about the role of the synovium in joint homeostasis and disease despite its critical function for joint nutrition, cartilage maintenance, and immune/inflammation modulation^{9; 10}. Given its important contribution to OA disease in other contexts, understanding synovial biological function in the ankle is a new and important area of focus.

RNA sequencing (RNA-seq) offers a powerful method to uncover global gene expression patterns, enabling the identification of novel biomarkers, pathways, and therapeutic targets.¹¹ Transcriptomic studies have revealed unique molecular signatures in rheumatoid arthritis, hip OA, and knee OA,^{12; 13} challenging the assumption that findings from one joint can be generalized to another. To date, no study has comprehensively examined the synovial transcriptome in ankle OA. To fill this critical knowledge gap, we performed the first transcriptomic analysis of ankle OA synovium versus non-OA controls to identify disease-specific molecular pathways that could inform targeted, joint-specific approaches to ankle OA.

1. Methods

2.1. Design

Prospective case-control study; level of evidence: 3.

2.2. Study Participants

Following institutional review board approval (COMIRB #19-2855), 510 consecutive adult patients scheduled for primary total ankle joint replacement (diagnosis: end-stage ankle OA), ankle arthrodesis (diagnosis: end-stage ankle OA), or ankle arthroscopy (diagnosis: ankle instability, impingement, or painful hardware) from 2020-2022 at University of Colorado-associated hospitals were screened for participation (Table 1). Patients’ suitability and OA severity were classified by the surgeon’s clinical and surgical assessment using the Outerbridge grading system¹⁴ (Fig1). Synovial specimens were collected intraoperatively by the surgeon under direct visualization, either during arthrotomy (OA patients undergoing joint replacement) or arthroscopy (non-OA patients). Synovium was carefully dissected and separated from adjacent structures (bone, cartilage, capsule) prior to flash freezing in liquid nitrogen. Post-surgery, specimens graded Outerbridge 4 from ankle replacement and arthrodesis surgeries were classified as “OA”, while those graded as Outerbridge 0 or 1 from ankle arthroscopy were classified as “non-OA”. The final analysis included 26 OA and eight non-OA patients (Fig1). Clinical metadata, including smoking status (never, current, former), age, sex, BMI, history of an ankle fracture (yes, no), and use of cholesterol-lowering drugs (yes, no), were prospectively collected.

Table 1:

Inclusion and exclusion criteria of subject screening and enrollment.

Inclusion of Ankle Osteoarthritis Patients
Clinical Criteria	Inclusion Criteria   ■ Patient ≥ 18 years of age   ■ Diagnosis of end-stage ankle OA and scheduled for joint replacement surgery or arthrodesis    OR   ■ Diagnosis of ankle instability, impingement, or painful hardware and scheduled for ankle arthroscopy	Exclusion Criteria   ■ Presence of autoimmune comorbidities, inflammatory disease of the joint (rheumatoid arthritis, systemic lupus, Crohn’s disease, psoriatic arthritis, gout, Reiter’s disease, ankylosing spondylitis, active infection)   ■ Scheduled for revision surgery
Specimen Criteria	■ Surgical discard of a minimum on 20mg synovial joint tissue

Figure 1:

(A) STROBE Diagram of subject screening and enrollment.

1.3. Comprehensive mRNA Analysis

RNA was extracted using TRIzol^™ following the manufacturer’s protocol (ThermoFisher) and purified (GeneJET RNA Purification Kit). RNA quality and quantity were assessed using a Bioanalyzer 2100 and Nanodrop. Samples with an RNA Integrity Number (RIN) ≥4 and RNA yield ≥10μg were progressed to library construction. messenger RNA (mRNA) was purified using poly-T oligo-attached magnetic beads. After fragmentation, first-strand cDNA was synthesized using random hexamer primers, followed by second-strand cDNA synthesis with dUTP for a directional library or dTTP for a non-directional library. Libraries were quantified using Qubit and real-time PCR, while size distribution was assessed with a Bioanalyzer. The quantified libraries were pooled, and bulk 150bp, paired-end RNA-seq was performed using the Illumina NovaSeq6000 platform. Reads were aligned to the GRCh38/hg38 human reference genome using Hisat2 v2.0.5.¹⁵ Data containing adapter, ploy-N (unresolvable bases >10%), and low-quality reads (reads containing >50% of low-quality (Qphred<=5) bases) were removed. Quality metrics, including sequencing error rates and nucleotide distributions, were evaluated prior to downstream analysis. The bulk RNA raw data and annotations are available in the Gene Expression Omnibus (GEO) with the accession number GSE293276.

1.4. Bioinformatic Analysis

1.4.1. Differential Gene Expression and Pathway Analyses

Differential gene expression analysis was performed between end-stage osteoarthritic and non-OA ankle synovium using DESeq2¹⁶ in RStudio (v2024.09.0+375). The Benjamini-Hochberg method adjusted for multiple testing with a significance cutoff of p-adjusted <0.05. Genes with an absolute fold-change ≥2 were analyzed further (Supplemental Table 1). Gene Ontology (GO)¹⁷, KEGG¹⁸, and Reactome¹⁹ enrichment analyses were performed using the clusterProfiler R package²⁰, with gene length bias correction applied (Supplemental Table 2). Gene set enrichment analysis was performed using the local version of the GSEA tool (http://www.broadinstitute.org/gsea/index.jsp). Pathway terms with corrected p<0.05 were considered significantly enriched. The biological process (BP), cellular component (CC), and molecular function (MF) terms were annotated, respectively.

2.4.2. Principal Component Analysis and Covariate Assessment

A PCA was performed on log-transformed, normalized gene counts to identify potential confounding factors. Data were scaled prior to PCA to ensure comparability across genes. The first principal component (PC1) was extracted and integrated with clinical metadata (age, sex, BMI, smoking status, cholesterol-lowering drug use, and fracture history). To evaluate the relationship between PC1 and clinical metadata, a multiple linear regression (MLR) model was fitted using PC1 as the dependent variable and clinical metadata as the independent variable (Supplemental Fig1).

2.4.3. Cluster Analysis with Covariate Adjustment

To identify unbiased clusters within the gene expression dataset, we employed dimensionality reduction using Uniform Manifold Approximation and Projection (UMAP)²¹, incorporating age and fracture history as additional input features alongside gene expression values. The optimal number of clusters was determined using a combination of elbow plots and silhouette scores (Supplemental Fig2).

2.4.4. Gene Interaction Network Construction and Centrality Analysis

Centrality analysis was performed on gene expression data after adjusting for clinical covariates, age, prior ankle fractures, and cluster assignment. To do so, variation attributable to these covariates was removed using the removeBatchEffect function from the limma package in R.²² Differential expression contrasts between clusters were then applied using the contrasts.fit function, and empirical Bayes smoothing (eBayes) was performed to stabilize variance estimates. Significantly enriched pathways were identified for up- and downregulated genes using a custom enrichResult object, facilitating pathway enrichment analysis across the GO, KEGG, and REACTOME databases. To assess the influence of covariates, we compared differential gene expression between unadjusted (univariate) and covariate-adjusted (multivariate) datasets in a scatter plot (Fig3B). Genes deviating from the diagonal tangent x=y indicate expression patterns influenced by age, fracture history, or cluster assignment (Fig2B). A Gene Interaction Network was then constructed from filtered edge lists of significantly regulated genes mapped to enriched pathways across GO, KEGG, and REACTOME databases. Gene relationships within enriched pathways were analyzed using the igraph package,²³ and pathway-gene networks were visualized using the tkplot() function from the igraph package in R.²³ Centrality metrics, including degree (DC), betweenness (BC), and eigenvector centrality (EC), were calculated to quantify node importance and identify hub genes with high connectivity or influence within the network (Supplemental Table 3).

Figure 3:

Analysis of gene expression data after adjusting for clinical covariates “age” and “prior ankle fractures”, and cluster assignment: (A) Unbiased cluster analysis of OA and non-OA gene expression data after adjusting for clinical covariates. OA and non-OA patient samples robustly align with the unbiased clusters. (B) Scatter Plot between log normalized gene count data from univariate analysis on the x-axis and multivariate analysis on the y-axis. Genes clustering closest to the x=y tangent show minimal expression changes after multivariate adjustment. (C) Pathway-Gene-Network including significantly up- or downregulated GO, KEGG, and REACTOME pathways. Degree, betweenness, and eigenvector centrality were calculated to identify hub genes with high connectivity or influence within the network.

Significant pathways were grouped into Pathway Groups based on shared biological themes. Node size represents degree centrality, where larger nodes indicate genes with higher connectivity. Node color reflects betweenness centrality, with darker colors highlighting genes that serve as key intermediaries within the network. Edges (connections) indicate relationships between genes and pathways.

Figure 2:

Analysis of un-adjusted gene expression data: (A) Heatmap analysis of significant differentially expressed genes in ankle OA synovium compared to non-OA controls. (B) Volcano Plot displaying the distribution of differentially expressed genes across all samples, comparing OA synovium vs. non-OA synovium. The X-axis indicates the log fold-change of gene expression, and the Y-axis displays the adjusted p-value (padj) representing the statistical significance of those differences. Thresholds included a padj ≤ 0.05 and an absolute log fold-change ≥ 2.

(C-D) Gene set enrichment analysis of OA samples compared to healthy control samples. Significant Reactome pathway enrichment (C), KEGG pathways enrichment (D), and GO term enrichment (E) show a high representation of pathways and functions related to ECM remodeling and tissue morphogenesis, as well as enrichment of processes related to innate and adaptive immune activation across all three databases.

2.5. Statistical Analysis

Demographic data were analyzed using GraphPad Prism. Fisher’s exact test was used to compare categorical variables between groups. Numerical variables were analyzed using paired t-tests when variances and residuals were normally distributed, as confirmed by the Shapiro-Wilk test. If normality assumptions were violated, the Mann-Whitney U test was applied. Differences between groups were considered significant at a p-value<0.05.

2. Results

3.1. Osteoarthritic Ankle Synovium Shows Distinct Gene Expression

To define the molecular profile of osteoarthritic ankle synovium, we analyzed synovial specimens from patients with and without end-stage ankle OA. Demographic features were similar between groups, though the control group was significantly younger and had significantly shorter surgery times due to the nature of their procedures and patient spectrum for ankle instability (Table 2).

Table 2:

Participant demographic summary. The ankle OA cohort demonstrated significant differences in age (p<0.001), and average tourniquet time (p<0.001). Fisher’s exact test was used to compare categorical variables between groups, when both variances and residuals were normally distributed, as confirmed by the Shapiro-Wilk test. If normality assumptions were violated, the Mann-Whitney U test was applied

Demographics
	Non-OA	OA	p-value
Average Age (±SD), years *	33.63(±14.93)	65.81(±8.96)	<0.0001
Average BMI (±SD), kg/m2	27.88(±5.49)	27.82(±5.52)	0.897
Sex (m/f)	3/5	16/10	0.42
	Ethnicity
	Non-OA	OA
Hispanic/Latino	0	1	>0.99
Not Hispanic/Latino	8	25
	Race
	Non-OA	OA
African American/Black	0	1	0.42
American Indian or Alaska Native	0	0
White	7	25
Asian	0	0
Native Hawaiian or Other Pacific Islander	0	0
Other	1	0
Past ankle Fractures
	Non-OA	OA
History of Past Ankle Fracture	0	6	0.30
No Past Ankle Fractures	8	20
Smoking Status
	Non-OA	OA
Never Smoker	5	17	0.82
Former Smoker	3	7
Average Pack Year (±SD)	9(±9.9)	9.03(±6.91)
Current Smoker	0	2
Pack Per Day (±SD)	N/A	0.63(±0.53)
Diabetes Mellitus
	Non-OA	OA
No Diabetes	7	25	0.42
Pre-Diabetes	1	1
Type 1 Diabetes	0	0
Type 2 Diabetes	0	0
Surgical Information
	Non-OA	OA	p-value
Total Ankle Arthroplasty (TAA)	0	25
Ankle Arthrodesis	0	1
Ankle Arthroscopy	8	0
Average Tourniquet Time (±SD), minutes *	51.25(±11.85)	115.1(±30.81)	<0.0001
Average ASA Score (±SD)	1.5(±0.53)	1.96(0.53)	0.054

^*Notes statistically significant difference, corrected p-value ≤ 0.05. SD = Standard Deviation.

Heatmap analysis of the RNA-seq data revealed distinct gene expression profiles in osteoarthritic synovium compared to non-OA controls (Fig2A). Differential gene expression (DEG) analysis identified 1,100 out of 38,339 detected genes with significant changes in OA (adjusted p<0.05, log2 fold-change >2), of which 915 were upregulated, and 185 were downregulated (Fig2B, Supplemental Table 1). Among the most differentially expressed genes by fold-change, we observed marked upregulation of collagens (COL2A1: 580-fold, COL6A5: 138-fold) and immunoglobulin components (IGHA1: 433-fold, IGHG1: 221-fold, IGKC: 144-fold. Table 3A). There was also a significant increase in multiple components of the IL-6 cytokine family, with the most substantial increase seen in IL11 (61-fold, Table 3B), alongside activation of several different WNT proteins (Table 3C). Notably, we observed the most substantial increases in WNT7B (40-fold) and WNT5A (11-fold) within the WNT-pathway family (Table 3C). Pathway enrichment analysis revealed a significant overrepresentation of processes related to extracellular matrix (ECM) remodeling (GO: extracellular matrix organization padj= 7.28×10⁻¹³, extracellular structure organization padj=7.28×10⁻¹³; Reactome: Extracellular matrix organization padj=1.22×10⁻¹⁵), tissue morphogenesis (GO: collagen metabolic process padj=2.99×10⁻⁷, connective tissue development padj=8.28×10⁻⁷; Reactome: Collagen formation padj=1.48×10⁻⁹; Collagen biosynthesis and modifying enzymes padj=1.17×10⁻⁸ Fig2C–E), and immune activation in osteoarthritic ankle synovium as compared to non-OA controls across GO, KEGG, and Reactome databases (GO: leukocyte migration padj=9.05×10⁻⁸, positive regulation of leukocyte chemotaxis padj=1.3×10⁻⁶; Reactome: Interleukin-6 family signaling padj=0.024; Collagen biosynthesis and modifying enzymes padj=1.4×10⁻⁸; KEGG: Cytokine-cytokine receptor interaction padj=0.0002; Fig2C–E, Supplemental Table 2)

Table 3:

Differentially expressed genes in OA synovium vs healthy controls, sorted by fold-change. (A) 15 most differentially expressed genes (DEGs). (B) DEGs of the IL-6 family cytokines. (C) DEGs related to the Wnt pathway.

A	most differentially expressed genes (by fold change)
	Gene Name	FoldChange	padj
	COL2A1	580.7	1.91E-08*
	IGHA1	432.6	5.63E-10*
	IGHG1	221.1	9.01E-11*
	CA9	211.5	6.42E-07*
	ODAM	179.6	7.96E-06*
	IGKC	143.7	5.05E-10*
	COL6A5	138.2	1.62E-06*
	IGKV3-15	134.2	2.82E-03*
	IGKV1-5	129.8	1.54E-04*
	IGHG3	128.0	3.21E-06*
	IBSP	124.8	6.06E-06*
	IGHA2	113.8	2.13E-06*
	IGKV3-20	111.6	8.99E-05*
	IGHV3-23	105.5	6.42E-03*
	RUFY4	98.9	2.69E-07*

B	Wnt pathway related differentially expressed genes
	Gene Name	FoldChange	padj
	WNT7B	40.3	3.05E-03*
	WNT5A	10.7	1.98E-10*
	KREMEN2	7.6	2.93E-04*
	WIF1	6.3	0.02*
	DKK3	5.7	4.53E-08*
	WNT9A	4.8	1.29E-06*
	WNT3	4.0	2.05E-06*
	WNT10A	3.5	0.11
	WNT2	3.2	0.24
	WISP1	2.5	3.11E-03*
	WNT9B	2.5	0.04*
	DKK4	2.4	0.38
	WISP3	2.4	0.16
	WNT4	2.3	8.57E-03*
	DKK2	2.0	0.05
	KREMEN1	2.0	1.15E-04*
	WNT16	2.0	0.39
	WNT1	1.9	0.55
	WNT10B	1.8	0.03*
	WNT5B	1.8	9.64E-03*
	WNT7A	1.7	1.00
	WISP2	1.7	0.08
	WNT11	1.2	0.61
	DKK1	1.0	0.99
	WNT2B	0.9	0.76
	WNT8A	0.9	1.00
	WNT8B	0.9	0.91
	WNT6	0.5	0.48
	WNT3A	0.4	1.00

C	IL-6 family Cytokines, differentially expressed genes
	Gene Name	FoldChange	padj
	IL11	61	4.68E-06*
	IL6	11.3	2.10E-07*
	OSM	9	2.70E-05*
	LIF	3.1	1.85E-04*
	CTF1	3	1.28E-04*
	CLCF1	2.8	5.50E-09*
	IL6R	0.9	0.44
	IL6ST	0.7	0.01

^*Notes statistically significant difference, corrected p-value ≤ 0.05.

3.2. Unbiased Clustering Identifies Osteoarthritis-Specific Transcriptomic Patterns

To address potential confounding due to clinical heterogeneity, we evaluated the impact of age, sex, BMI, smoking status, cholesterol-lowering drug use, and prior ankle fractures on gene expression. Multiple linear regression identified age and fracture history as significant confounders (Supplemental Fig1). Further, we examined transcriptomic differences between OA and non-OA ankle synovium while adjusting for age and fracture history via an unbiased cluster analysis after dimensionality reduction of the gene expression data using UMAP.

K-mean cluster analysis, based on the first two UMAP dimensions, identified two distinct clusters corresponding precisely to our biological groups: end-stage OA synovium and non-OA synovium (Fig3A, Adjusted Rand Index = 1, Purity Index = 1, Silhouette Score = 0.52). A silhouette score of 0.52 for two clusters indicated a moderately well-defined clustering structure. This was supported by the UMAP visualization, which displayed two visually distinct groups. While the elbow plot suggested the potential for 3 to 4 clusters, increasing K>2 resulted in lower silhouette scores, smaller cluster sizes, and less interpretable separations. Based on these findings, we performed k-means clustering with K=2, using optimal seed selection guided by the Adjusted Rand Index and Purity Score to ensure robust cluster identification. This clustering approach confirmed the segregation of OA and non-OA samples, independent of covariates and OA outcome.

3.3. WNT Signaling is a Significant Signature in Osteoarthritic Ankle Synovium

To assess the impact of covariate adjustment on differential gene expression, we compared univariate and multivariate gene expression analyses using a scatter plot (Fig3B). WNT pathway transcripts showed robust upregulation in osteoarthritic ankle synovium, with WNT pathway transcripts closely aligned with the y=x tangent. In contrast, collagen transcripts were clustered below this line, suggesting some sensitivity to covariate adjustments. IL6-family cytokines and immunoglobulins exhibited the most pronounced variability, reflecting their dependency on age and fracture history (Fig3B).

3.4. Multivariate Expression Analysis Highlights Key Pathways in Osteoarthritic Ankle Synovium

To identify influential pathways in ankle OA, we performed network enrichment analysis across KEGG, GO, and REACTOME databases, integrating co-expression network analyses from the adjusted multivariate gene expression dataset. Pathway enrichment analysis revealed a pronounced upregulation of neurosensory pathways in osteoarthritic synovium, with significant enrichment in the detection of chemical stimulus involved in sensory perception (DC=27, BC=4,246, EC=0.98) and sensory perception of smell (DC=27, BC=389, EC=1). Notably, multiple olfactory receptors (e.g., OR10A2, OR10H5, OR1C1) emerged as hub genes (Fig3C, Supplemental Table 3). While alignment dispersion artifacts may contribute due to their high sequence homology, the strong network-level enrichment suggests biological relevance. Myogenic pathways exhibited high betweenness scores in OA synovium, including muscle contraction (DC=23, BC=3,470, EC=0.001), muscle organ development (DC=22, BC=3,327, EC=0.001), and cytoskeleton in muscle cells (DC=18, BC=2,343, EC=0.0001), showing several myosin chains as central genes (Fig3C, Supplemental Table 3). Other upregulated pathways included carbohydrate metabolism, humoral immune responses, inhibition of posttranscriptional processes, and structural tissue formation, with keratin genes emerging as central regulators (Fig3C, Supplemental Table 3). Conversely, post-transcriptional regulation, protein synthesis, and cellular growth and signaling pathways were significantly downregulated, forming highly connected hub networks (RNA splicing: DC=449, BC=1,277,823, EC= 0.11; ribonucleoprotein complex biogenesis: DC=475, BC=1,034,793, EC=0.11; Translation: DC=287, BC=984,328, EC=0.05; cell growth: DC=460, BC=704,763, EC=0.6, Fig3C, Supplemental Table 3). Hub genes included ribosomal proteins, translation initiation factors, and elongation factors. Pathways involved in innate and cellular adaptive immunity, lipid metabolism, and protein metabolism were also downregulated (Fig3C). Notably, pathways governing cell adhesion and activation emerged as key regulatory influencers based on their high eigenvector scores (regulation of cell-cell adhesion: DC=450, BC=345,108, EC=1; positive regulation of cell adhesion: DC=453, BC=437,208, EC=0.98; positive regulation of cell activation: DC=357, BC=183,905, EC=0.89; Fig3C, Supplemental Table 3).

3. Discussion

OA has traditionally been viewed as a uniform disease process across joints despite apparent differences in clinical presentation and progression.^{3; 24} While hip and knee OA have been extensively studied, ankle OA remains relatively understudied, particularly regarding the role of the synovium in disease development. Our analysis begins to address this gap, uncovering distinct gene expression patterns in ankle OA synovium, characterized by altered WNT and IL6-Family Cytokine expression, alongside metabolic, immunologic, cell signaling, and ECM expression changes.

Our analysis identified a clear transcriptional separation between OA and non-OA ankle synovium, confirming that ankle OA is characterized by distinct gene expression patterns. Unbiased cluster analysis revealed that samples grouped strongly according to disease status, reinforcing the presence of a unique synovial molecular signature in OA. This separation suggests that ankle OA may not merely be a downstream consequence of cartilage degeneration but might involve an active and distinct synovial pathology.

Further supporting this distinction, our Pathway and DEG analysis of ankle OA synovium revealed significant enrichment in ECM-related and morphogenetic pathways. These findings align with post-traumatic OA studies in equine weightbearing joints, which reported enrichments in ECM organization, angiogenesis, and cell signaling.²⁵ Interestingly, our results contrast with gene expression profiles of hip and knee OA, which noted enrichment in proteolysis, collagen catabolism, and ECM disassembly.²⁵ The distinct transcriptional landscape of ankle OA synovium suggests that joint-specific factors, such as biomechanics, injury mechanisms, and synovial-cartilage crosstalk, may contribute to unique disease processes in the ankle.

Together, these findings provide robust evidence that osteoarthritic ankle synovium exhibits distinct gene expression patterns compared to non-OA controls. This supports the need for joint-specific OA research and serves as a starting point for deeper network-based analyses, as explored in the next section, where we examine the key molecular pathways that characterize osteoarthritic synovium after adjusting for covariates.

Age and prior ankle fracture emerged as significant influences on synovial gene expression, underscoring the importance of patient history in shaping the local molecular environment. While age is a known risk factor for osteoarthritis and has been linked to morphological and transcriptomic changes in synovial tissue²⁶, our finding that prior fracture contributes to distinct expression patterns suggests a unique molecular trajectory for post-traumatic osteoarthritis in the ankle. Supporting this, Haslauer et al. observed increased proteolytic enzyme expression in knee synovium post-trauma,²⁷ indicating shared mechanisms of trauma-induced joint degeneration.

Our data further revealed downregulation of pathways related to post-transcriptional regulation, protein production, and cellular growth, signaling, and activation – hallmarks of cellular senescence.²⁸ Senescent cells have been identified in knee OA synovium²⁹ and may similarly contribute to ankle OA pathology. Interestingly, metabolic shifts such as increased glycolysis have been described previously,³⁰ which aligns with our observations. Additionally, our network analysis revealed significant downregulation of ribosomal biogenesis and rRNA metabolic processes, paralleling findings in OA chondrocytes showing impaired ribosomal function.³¹ This suggests that ribosomal dysfunction may not be limited to cartilage but also involves the synovium. Chabronova et al. demonstrated that osteoarthritic synovial fluid alters chondrocyte protein synthesis by depleting small nucleolar RNAs (snoRNAs) in osteoarthritic cartilage, impairing ribosome translation³². This suggests potential synovial-cartilaginous crosstalk in the ribosomal regulation during ankle OA pathogenesis. However, the synovial role in epi-transcriptomic regulation of cartilage remains unclear and requires further research.

Beyond metabolic alterations, we observed immunological changes with increased humoral immune responses, reduced innate and cellular adaptive immunity, and decreased cell-cell adhesion. Synovial macrophages have been discussed as an important contributor to OA pathogenesis and OA-related pain.³³ Interestingly, our network analysis revealed enrichment of neurosensory pathways with several olfactory receptors (OR) emerging as hub genes. While ORs are best known for their role in olfactory tissues, they are also expressed in macrophages, modulating tissue immunity³⁴ and contributing to inflammatory conditions such as atherosclerosis³⁵. Their role in ankle osteoarthritis remains unexplored and warrants further investigation.

Notably, immunoglobulin transcripts were among the most upregulated genes, highlighting a possible role for plasma cells in ankle osteoarthritis. While increasing evidence suggests that the adaptive immune system is more than just a bystander in the inflammatory OA process, the contribution of immunoglobulin-producing plasma cells is still poorly understood. Clonal analysis of synovial B-cell antibodies suggests an antigen-driven process³⁶, similar to rheumatoid arthritis, and autoantibodies targeting cartilaginous proteins have been detected in OA³⁷; however, their role in the synovium and the drivers of their clonal expansion remain to be elucidated.

We also observed consistent activation of various WNT signaling pathways in osteoarthritic ankle synovium, with WNT7B emerging as the most upregulated family member. The WNT-family encodes secreted signaling proteins critical for skeletal development, chondrogenesis, and chondrocyte organization,³⁸ and its involvement in OA progression is well established^{39; 40}. Notably, WNT7B has been implicated in inflammatory and degenerative joint diseases in murine arthritis models,^{41; 42} although its specific role within the synovium remains poorly understood. Inflammatory cytokines such as IL-6, TNF-α, and IL-1β have been shown to induce WNT7B expression,⁴³ suggesting potential upstream regulation by synovial inflammation. In our dataset, we observed elevated IL-6 expression in OA ankle synovium, while TNF-α and IL-1β levels remained unchanged. Interestingly, we detected a 61-fold increase in IL-11, a cytokine known to be elevated in osteoarthritic synovial fluid⁴⁴ and implicated in subchondral bone remodeling in hip OA.⁴⁵ Given its close association with WNT signaling, particularly WNT5A,⁴⁶ their potential interplay in the ankle OA pathogenesis warrants further investigation.

Finally, we observed myogenic gene expression in OA ankle synovium, similar to findings by Yeremenko et al., who identified a myogenic transcriptomic signature in the synovial sub-lining of spondylarthritis patients.⁴⁷ Notably, this signature was unrelated to hypervascularization, raising questions about the origin of myogenic cells in the synovium.⁴⁷ While synovial tissue lacks skeletal muscle cells, it harbors pluripotent stem cells capable of differentiating into myogenic lineages.⁴⁸ Timmons et al. similarly noted a myogenic signature in brown preadipocytes, linking it to the metabolic demands of brown adipose tissue.⁴⁹ Impaired metabolic flexibility in synovial fibroblasts⁵⁰ may drive this myogenic-like signature in OA; however, its role in OA pathogenesis or synovial metabolism remains unclear, warranting further study.

The primary limitation of this study is the relatively small sample size of our RNA-seq dataset, which may reduce statistical power and limit the ability to detect subtle differential gene expression patterns. Small sample sizes can also affect the estimation of biological variability and dispersion parameters, potentially impacting the robustness of our findings. To mitigate these challenges, we incorporated a network and centrality analysis approach, which enhances individual gene-level signals by examining broader interaction patterns. This method prioritizes gene modules with consistent interaction trends, allowing us to identify biologically relevant pathways despite potential limitations in individual gene-level power. While this approach enabled the identification of key pathways and hub genes, larger studies are needed to capture more nuanced transcriptional changes that may have been missed in our analysis.

A second limitation of this study is the inherent demographic differences between the OA and control groups, particularly with respect to age. These differences contributed significantly to the variation observed in the gene expression dataset. Since patients undergoing ankle arthroscopy for conditions such as ankle instability or impingement tend to be younger than those undergoing joint replacement surgery or arthrodesis for end-stage OA, complete age matching was not feasible. To account for these baseline differences, we incorporated age and prior fracture history as covariates in the second part of our analysis focused on our centrality analysis. However, the possibility of residual or unmeasured confounding remains and should be considered when interpreting the results.

Finally, we need to acknowledge the observational nature of RNA sequencing data. While the observed synovial changes are likely shaped by a multidirectional synovial-cartilaginous crosstalk, the methodological limitations of RNA sequencing prevent us from establishing causality. Further research incorporating functional validation experiments is needed to clarify the directionality of these interactions and their role in disease progression.

4. Conclusion

We have demonstrated for the first time that osteoarthritic ankle synovium exhibits a distinct pathological gene expression profile compared to healthy synovium, identified through unbiased cluster analysis. This includes a myogenic signature, specific WNT signaling proteins, IL6-family cytokines, metabolic alterations, immune modulation, and impaired post-transcriptional processing. Despite a limited sample size, our findings provide valuable insights into the complex molecular landscape of osteoarthritic ankle synovium. With the novel information provided here, further mechanistic studies are needed to clarify the directionality of these interactions and their role in disease progression, therapeutics, and prevention. Our transcriptomic findings underscore the need for joint-specific OA research and highlight pathways that may drive or respond to disease progression, providing a foundation for future investigations into ankle OA pathogenesis and treatment.

Supplementary Material

SFig2

Supplemental Figure 2:

(A) Elbow plot of k-means cluster analysis of OA and non-OA gene expression data after adjusting for clinical covariates. Number of clusters (1-10) is shown on the x-axis, the total within-cluster sum of squares for each cluster is shown on the y-axis. The rate of decrease in WCSS significantly diminishes at 3 to 4 clusters. (B) The silhouette scores are shown for clusters 1-4.

SFig1

Supplemental Figure 1:

Multiple linear regression model between Principle Component 1 and clinical metadata age, fracture history, smoking status (current vs non-smoker; former vs non-smoker), Cholesterol-lowering drugs, sex (male vs female), and BMI.

Supp_Table_1

Supplemental Table 1:

Gene Count Matrix with Differential Gene Expression Analysis: DEG analysis was performed between end-stage osteoarthritic and non-OA ankle synovium. The Benjamini-Hochberg method adjusted for multiple testing.

Supp_Table_2

Supplemental Table 2:

Significantly enriched GO, KEGG, and Reactome pathways: Gene Ontology, Reactome, and KEGG enrichment analysis of differentially expressed genes was implemented by the clusterProfiler R package, in which gene length bias was corrected. Pathways with corrected P value less than 0.05 were considered significantly enriched by differential expressed genes.

Supp_Table_3

Supplemental Table 3:

Gene Interaction Network and Centrality Metrics: The Gene Interaction Network was constructed from filtered edge lists of significantly regulated genes across GO, KEGG, and REACTOME databases. Centrality metrics, including degree, betweenness, and eigenvector centrality, were calculated to quantify node importance.

Clinical Significance:

These findings provide valuable insights into the molecular mechanisms driving ankle osteoarthritis and have the potential to inform the development of novel targeted therapeutic strategies.