Cell clusters in human hand articular cartilage

We obtained 10 articular cartilage specimens (i.e., five osteoarthritic and five nonosteoarthritic specimens) from the interphalangeal joints of five donors (Fig. 1a and Table S1) whose pathological conditions were confirmed by macroscopic observation and Safranin-O/Fast Green staining (Fig. 1b). Among them, 105 142 individual cells passed the strict quality filtering process for subsequent analysis (Fig. S1), with 52 197 cells originating from osteoarthritic cartilage and 52 945 from nonosteoarthritic cartilage (Fig. 1c). We identified 20 putative clusters, including 18 chondrocyte clusters and two rare clusters, using the unsupervised method (Fig. 1d). The cluster-specific differentially expressed genes (DEGs) are shown in Table S2. Most chondrocyte clusters align with the following 10 chondrocyte subpopulations according to established markers (Fig. 1e)20,21,22,23,24: effector chondrocytes (EC, Cluster 1 and 2); prehypertrophic chondrocytes (preHTC, Cluster 3 and 4); regulatory chondrocytes (RegC, Cluster 5, 6, and 7); prefibrocartilage chondrocytes (preFC, Cluster 8); fibrocartilage chondrocytes (FC, Cluster 9, 10, and 11); proliferating chondrocytes (ProC, Cluster 12); hypertrophic chondrocytes (HTC, Cluster 13); mitochondrial chondrocytes (MTC, Cluster 16); homeostatic chondrocytes (HomC, Cluster 17 and 18); and cartilage progenitor cells (CPC, Cluster 19). Remaining Clusters 13 and 14 were grouped as a novel subpopulation (inflammatory chondrocytes, InflamC) according to their distinct expression of genes related to the inflammatory response, immune system process and immune response, i.e., CCL20, CCL2, NOS2, and MMP3 (Fig. 2a). Immunohistochemistry (IHC) analysis revealed a predisposed distribution of InflamC in the superficial zone of articular cartilage (Fig. 2b). In addition, two small subpopulations were detected, including a subpopulation of macrophages (Mac, Cluster 20) that specifically expressed IL1B, CD74, CD68, and HLA-DRA and a subpopulation of H-type endothelial cells (EndC, Cluster 21) that highly expressed PECAM1 and EMCN (Fig. 1e and Fig. 2c, d). To better understand the specific characteristics of hand chondrocytes, we compared transcriptomic differences between these chondrocyte subpopulations and knee chondrocytes.20 The results showed a close relationship of EC, RegC, ProC, preHTC, HTC and HomC in knee cartilage and hand cartilage (Fig. S2), which is consistent with the correlation results for cell subpopulations in hand cartilage only (Fig. S1h). In addition, FC in Ji et al.’s dataset showed notable differences from other subpopulations but exhibited the highest similarity to the FC observed in this study (Fig. S2).

Fig. 1

ScRNA-seq atlas of cell clusters in human hand articular cartilage. a Schematic workflow of single-cell RNA sequencing. b Representative macroscopic and microscopy images of hand osteoarthritic and nonosteoarthritic cartilage of hand interphalangeal joints. Arrows indicate locations of cartilage degeneration. c t-SNE embedding plot of cells colored according to disease status. All clusters contained cells from both hand osteoarthritic and nonosteoarthritic samples. d t-SNE embedding plot for 105 142 cells derived from paired hand articular cartilage samples of five donors. e Heatmap showing the top 10 discriminative genes of each cell cluster in hand articular cartilage. Genes indicated in the right column were adopted for annotation for each subpopulation. ScRNA-seq single-cell RNA sequencing, EC effector chondrocytes, preHTC prehypertrophic chondrocytes, RegC regulatory chondrocytes, preFC prefibrocartilage chondrocytes, FC fibrocartilage chondrocytes, HTC hypertrophic chondrocytes, InflamC inflammatory chondrocytes, MTC mitochondrial chondrocytes, HomC homeostatic chondrocytes, CPC cartilage progenitor cells, Mac macrophages, EndC endothelial cells

Fig. 2

Signatures and putative spatial distribution of InflamC, CPC, Mac and EndC. a Expression of selected marker genes (combined for annotation) for InflamC visualized by a feature plot. b Representative immunohistochemistry (IHC) staining for MIP-3α and iNOS in hand articular cartilage tissues (n = 4). Scale bar, left, 100 μm; right, 20 μm. ***P < 0.001, ****P < 0.000 1. c Violin plot of selected marker genes for CPC, Mac and EndC. d Representative IHC staining of hPTTG, IL-1β and PECAM-1. Arrows indicate positive cells in the cartilage tissue. Scale bar, left, 100 μm; right, 10 μm. InflamC inflammatory chondrocytes, CPC cartilage progenitor cells, Mac macrophages, EndC endothelial cells, MIP-3α macrophage inflammatory protein 3 alpha, iNOS inducible nitric oxide synthase, hPTTG human pituitary tumor-transforming gene 1 protein, IL-1β interleukin-1 beta, PECAM-1 platelet endothelial cell adhesion molecule

InflamC and FC are potentially key chondrocyte subpopulations in hand OA

Although no statistically significant differences were found between hand OA and non-OA cartilage, preFC, InflamC, and FC showed a trend toward increased numbers in the cartilage of hand OA joints compared with non-OA joint cartilage. In contrast, the proportion of EC, MTC and HomC tended to be lower in the cartilage of hand OA joints (Fig. 3a). As shown in Fig. 3b and Table S3, InflamC was found to play an exclusive role in the response to cytokine stimuli and the innate immune response. Marker genes of preFC and FC were preferentially enriched for the extracellular matrix (ECM) or structural organization. HomC was enriched in nucleic acid and RNA metabolic processes, consistent with a previous report of human knee osteoarthritic cartilage.20 Notably, compared to the other subpopulations, InflamC and FC were enriched in many inflammatory signaling pathways, e.g., reactive oxygen species pathway, TNFα signaling via NFκB, interleukin and interferon-mediated pathways (Fig. S3). Gene set variation analysis (GSVA) of matrisome genes revealed heterogeneous performance in ECM composition and regulation by distinct chondrocyte subpopulations (Fig. 3c, d). ECs, HTCs and ProCs preferentially expressed genes encoding secreted factors, whereas preFCs and RegCs tended to express ECM regulators and proteoglycans. Furthermore, CPC and FC seemed to undergo evident changes in ECM gene expression between nonosteoarthritic and osteoarthritic status, suggesting a potential role in hand OA (Fig. 3e, f).

Fig. 3

Hand OA-related alteration of proportion in subpopulations and their potential function. a Differential analysis of subpopulation abundance between hand osteoarthritic and nonosteoarthritic cartilage. The Wilcoxon matched-pairs signed rank test was used for analysis, and a P value < 0.05 was considered significant. b GO enrichment analysis of DEGs from each chondrocyte subpopulation, showing distinct potential functions in hand articular cartilage tissue. c and d Heatmap and radar map showing the performance of 7 matrisome gene sets among each chondrocyte subpopulation. e and f Heatmap and radar map showing evident changes in matrisome genes in FC and CPC between hand osteoarthritic and nonosteoarthritic status. OA osteoarthritis, EC effector chondrocytes, preHTC prehypertrophic chondrocytes, RegC regulatory chondrocytes, preFC prefibrocartilage chondrocytes, FC fibrocartilage chondrocytes, HTC hypertrophic chondrocytes, InflamC inflammatory chondrocytes, MTC mitochondrial chondrocytes, HomC homeostatic chondrocytes, CPC cartilage progenitor cells, DEGs differentially expressed genes, ECM extracellular matrix

After identifying the functional role of each subpopulation alone, we next sought to decode the pattern of intercellular communication using CellChat analysis. As expected, ECM-receptor communication was shown to occur among chondrocytes (Fig. S4a–d). We observed a dominant role of FC in Notch and platelet-derived growth factor pathways based on ligand‒receptor interactions with EndC, suggesting its potential role in angiogenesis associated with hand OA (Fig. S4e–h). Considering that InflamC specifically expressed genes related to the recruitment and proliferation of macrophages, we next focused on interaction between InflamC and Mac. ICAM1 and VCAM1, marker genes of InflamC, were inferred to act on cellular contact and adhesion between InflamC and Mac (Fig. 4a, b). Furthermore, secretion of proinflammatory molecules such as IL1B, SPP1, CD99, and C3 by Mac might act on their corresponding receptors on InflamC, probably eliciting downstream inflammatory processes (Fig. 4c). Collectively, our results support unique and cooperative roles for the identified cell subpopulations in maintaining hand articular cartilage homeostasis and suggest a potential role for InflamC and FC in hand OA.

Fig. 4

CellChat analysis of the interaction between InflamC and Mac. a Bubble plot showing the molecular pattern of interaction between InflamC and Mac. b Chord plot showing communication between InflamC and Mac. Arrows indicate ICAM1 and VCAM1 in intercellular communication. c Heatmap emphasizing the ICAM, VCAM, IL1, SPP1, CD99, and COMPLEMENT pathways activated by the InflamC-Mac interaction. InflamC inflammatory chondrocytes, Mac macrophages

Ferroptosis is a key enriched pathway in hand osteoarthritic cartilage

Subpopulation-specific differences in gene expression between osteoarthritic and nonosteoarthritic cartilage are presented in Table S4. Given the potentially important role of InflamC and FC in hand OA, we focused on the difference between these two subpopulations. FC exhibited extensive alterations, with upregulation of a series of chondrocyte catabolic genes (e.g., MMP2, MMP3, and MMP13) and inflammatory response-related genes (e.g., CCL20, CD99, and TNFAIP6) in osteoarthritic cartilage (Fig. 5a). Seven genes, including inflammatory response-related genes (e.g., CCL20 and TNFAIP6), were upregulated in InflamC from osteoarthritic cartilage (Fig. 5b). Gene Ontology (GO) analysis revealed that upregulated genes of FC in osteoarthritic cartilage were enriched in platelet-derived growth factor binding and collagen catabolic and metabolic processes and that upregulated genes of InflamC were enriched in regulation of cell migration and immune response (Fig. 5c). Furthermore, Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of osteoarthritic FC and InflamC indicated significant enrichment of many ECM-degrading pathways, such as protein digestion and absorption and mineral absorption (Fig. 5d). Notably, genes upregulated in InflamC and FC of hand osteoarthritic cartilage were both significantly enriched in the ferroptosis pathway (Fig. 5d).

Fig. 5

Alteration of expression patterns and enrichment of ferroptosis in hand OA. a Volcano plot showing DEGs of FC between hand osteoarthritic and nonosteoarthritic cartilage ( | log fold change | >0.5; P < 0.05). b Volcano plot showing DEGs of InflamC between hand osteoarthritic and nonosteoarthritic cartilage ( | log fold change | >0.5; P < 0.05). c GO enrichment analysis of upregulated genes of FC and InflamC in hand OA. d KEGG enrichment analysis of upregulated genes of FC and InflamC in hand OA. Arrows indicate ferroptosis enriched among upregulated genes in osteoarthritic FC and InflamC. e GSEA showing extensive enrichment of ferroptosis in hand osteoarthritic FC and InflamC. f GSEA showing extensive enrichment of PI3K-AKT-mTOR signaling and inflammatory response in hand osteoarthritic FC and InflamC, respectively. g Violin plot showing differential expression of FTH1 and HMOX1 in all subpopulations between hand osteoarthritic and nonosteoarthritic cartilage. OA osteoarthritis, DEGs differentially expressed genes, FC fibrocartilage chondrocytes, InflamC inflammatory chondrocytes, GO Gene Ontology, KEGG Kyoto Encyclopedia of Genes and Genomes, GSEA gene set enrichment analysis

We next conducted gene set enrichment analysis (GSEA) and identified significant enrichment of the ferroptosis gene set in both InflamC and FC in hand osteoarthritic cartilage, which highly suggests that ferroptosis plays a role in hand OA (Fig. 5e). In addition, genes related to the inflammatory response and the PI3K-AKT-mTOR pathway were highly enriched in InflamC and FC in osteoarthritic cartilage, respectively (Fig. 5f). DEG analysis of core ferroptotic genes revealed significant elevation in FTH1 and HMOX1 in FC, InflamC and Mac (Fig. 5g and Fig. S5). Notably, FTH1 was consistently upregulated in both InflamC and FC, indicating iron overload in these cells during OA states (Fig. 5g). Furthermore, IHC analysis revealed that the protein level of FTH1 in human hand osteoarthritic cartilage was significantly elevated compared with that in nonosteoarthritic cartilage in the superficial and middle layers (Fig. S6). Together, these results demonstrate essential alterations in gene expression patterns in hand OA. Elevation of FTH1 expression as well as enrichment of the ferroptosis pathway might represent a novel and vital molecular mechanism in hand OA.

Comparative analysis of hand OA cartilage and knee OA cartilage reveal that FTH1 and ferroptosis play key and unique roles in hand OA

Due to the observed differences in structure and stress environment, the cellular and molecular alterations occurring during hand OA and knee OA are expected to be different. Thus, we performed comparative analysis of our data and previous scRNA-seq data on knee OA.20 We first analyzed alteration of the cellular proportion in knee OA cartilage. The proportions of EC and HomC showed a significant decrease, whereas those of preHTC were significantly increased (Fig. S7). These results are consistent with observations for hand OA cartilage, in which EC and HomC showed a similar trend (Fig. 3a). However, FC, which exhibited an increasing trend in hand OA cartilage, was not significantly changed in knee OA. These results, together with the fact that InflamC was identified specifically in hand OA cartilage, represent a difference in cellular components between hand OA and knee OA (Fig. S7).

Subsequently, we compared DEGs in all cells during knee OA and hand OA. There were eight upregulated genes in hand OA cartilage, including TNFAIP6, COL3A1, CCL20, FTH1, SERPINE2 and VCAM1 (Fig. S8a), and 59 upregulated genes in knee OA cartilage, including S100A4, COL3A1, HTRA1, TGFBI, OGN and ASPN (Fig. S8b). GO enrichment analysis showed that genes involved in calcium-mediated signaling, regulation of cell migration, motility and locomotion were upregulated in hand OA cartilage (Fig. S8c). Genes involved in collagen and extracellular matrix organization were upregulated in knee OA cartilage (Fig. S8e). Notably, KEGG analysis revealed that the upregulated genes in both hand OA and knee OA were enriched in protein and mineral absorption and the AGE-RAGE signaling pathway (Fig. S8d); the ferroptosis pathway, which was identified as the key pathway in hand OA, was specifically enriched in hand OA cartilage (Fig. S8f). Intersection of the DEGs of knee OA and hand OA revealed seven shared DEGs, nine hand OA-specific DEGs, and 86 knee OA-specific DEGs (Fig. S8g). Functional analysis indicated the shared DEGs to be enriched in cell migration, motility and protein digestion and absorption (Fig. S8h, i). Notably, GO analysis showed that hand OA-specific DEGs were enriched in age-dependent responses to reactive oxygen species and oxidative stress (Fig. S8j). Moreover, KEGG analysis revealed that hand OA-specific DEGs were enriched in ferroptosis, in which FTH1 was upregulated specifically in hand OA cartilage (Fig. S8k). Knee OA-specific DEGs were enriched in the extracellular matrix and protein digestion, consistent with previous studies (Fig. S8l, m).

As FCs were identified as key subpopulations in hand OA cartilage, we also performed DEG analysis for FCs between early-stage (stage 0 and 1) and late-stage (stage 3 and 4) knee OA cartilage. Upregulated genes of FC in late knee OA cartilage include COL3A1, COL6A3, IFI27, IGFBP7, THY1 and S100A10 (Fig. S9a). Functional enrichment analysis revealed that similar to hand OA cartilage, the upregulated genes of FC in late knee OA cartilage were enriched in extracellular matrix, extracellular vesicles, protein digestion and absorption (Fig. S9b, c). Intersection of the DEGs of FC in late knee OA and hand OA revealed nine shared DEGs, 47 hand OA-specific DEGs, and 502 knee OA-specific DEGs (Fig. S9d). Functional analysis showed the shared DEGs to be enriched in collagen catabolic processes, extracellular matrix organization and protein digestion and absorption (Fig. S9e, f). GO analysis showed that hand OA-specific DEGs and knee OA-specific DEGs were both enriched in the extracellular matrix and structure organization (Fig. S9g, i). Notably, KEGG analysis revealed that hand OA-specific DEGs were enriched in ferroptosis but that knee OA-specific DEGs were not. FTH1, which was identified as a key molecule of ferroptosis in hand OA, was uniquely and significantly upregulated in hand OA cartilage (Fig. S9h, j).

Taking all these results together, comparative analysis between hand OA cartilage and knee OA cartilage revealed that both cellular and molecular alterations during hand OA were quite different from those during knee OA. FC and InflamC, as well as the ferroptosis pathway, especially FTH1, play key and unique roles in cartilage degeneration in hand OA and might be specific targets for future intervention.

MR analysis of the causal association between upregulated FTH1 mRNA expression and the risk of hand OA in UK Biobank

Given that FTH1 was consistently and significantly upregulated in InflamC and FC from osteoarthritic cartilage, we subsequently sought to investigate the causal association between FTH1 mRNA expression and the risk of hand OA. We included 332 668 individuals of European descent from UK Biobank for MR analysis. The selection process of participants is presented in Fig. S10a, and the characteristics of the included individuals are presented in Table S5. Of the individuals, 2 418 (0.73%) individuals had a diagnosis of hand OA.

The selected genetic instrumental variables (Table S6) explained 65.6% of the variance in FTH1 mRNA expression. Univariate two-sample MR analysis showed that a genetic predisposition toward higher expression of FTH1 mRNA significantly increased the risk of hand OA (odds ratio [OR] = 1.07 per standard deviation [SD] increase in FTH1 expression, 95% confidence interval [CI]: 1.02–1.11, P = 0.005) (Table 1). Moreover, the direction of the effect estimate was consistent across the four MR methods assessed (i.e., inverse-variance weighted [IVW], MR‒Egger, weighted median, and MR-PRESSO) (Fig. 6a and Table 1). We did not observe heterogeneity in the present analysis (PCochran’s Q = 0.999), and MR‒Egger intercepts indicated limited evidence of directional pleiotropy (P = 0.491).

Table 1 The effect of genetically predicted FTH1 mRNA expression on hand OAFig. 6

Mendelian randomization analysis of the causal association between FTH1 mRNA expression and the risk of hand OA and cross-sectional analysis of the positive association between serum ferritin (encoded by FTH1) levels and the prevalence of hand OA. a Two-sample Mendelian randomization analysis of genetically predicted FTH1 mRNA expression in hand OA. b Schematic illustration of this study. MR Mendelian randomization, OA osteoarthritis, SNP single-nucleotide polymorphism, XO study Xiangya Osteoarthritis Study

Serum ferritin level and prevalence of hand OA in Xiangya Osteoarthritis Study

Ferroptosis is a form of regulated cell death initiated by perturbations of the intracellular microenvironment, which relies on iron availability.25 Our current findings, e.g., enrichment of the ferroptosis pathway and upregulated expression of FTH1 mRNA in osteoarthritic cartilage by scRNA-seq analysis and a causal association between upregulated FTH1 mRNA expression and the risk of hand OA by MR analysis, prompted us to investigate whether iron load is elevated in patients with hand OA. For this purpose, we measured serum ferritin, a biomarker for body iron stores,26 in Xiangya Osteoarthritis Study participants.18,19 A total of 1 241 participants were included (50.9% were female, mean age 62.8 years). The selection process is outlined in Fig. S10b. Of these participants, 392 (31.6%) had hand OA. The baseline characteristics according to hand OA status are shown in Table S7.

As indicated in Table 2, compared with the lowest quintile of serum ferritin, the crude ORs for hand OA were 1.07 (95% CI: 0.74–1.56), 1.07 (95% CI: 0.74–1.56), 1.27 (95% CI: 0.88–1.84), and 1.55 (95% CI: 1.08–2.22) in the second, third, fourth and highest quintiles of serum ferritin, respectively (P-for-trend = 0.037). Adjustment for age, sex and BMI did not change the results materially (P-for-trend = 0.005).

Table 2 Association between serum ferritin levels and the prevalence of hand OA