The mediation of circulating inflammatory proteins in the causal pathway from immune cells to osteoarthritis

Abstract

Background

Observational studies have found that immune cells and circulating inflammatory proteins play a dual role in the progression of osteoarthritis, but the exact mechanism remains unclear. Therefore, this study aimed to investigate whether the causal relationship between immune cells and Knee OA is mediated by circulating inflammatory proteins.

Method

A two-sample Mendelian Randomization analysis was conducted involving 731 immune cells, 91 inflammatory proteins and OA, utilising summary-level data from genome-wide association studies. The causal relationships between immune cells, inflammatory proteins and OA were sequentially analysed by multivariate Mendelian Randomization and validated using Bayesian weighted Mendelian Randomization. Subsequently, sensitivity analyses were conducted, employing Cochran’s Q test to assess heterogeneity, MR-Egger tests to assess pleiotropy, and Steiger directionality tests to rule out reverse causality. Lastly, a two-step approach was employed to ascertain the proportion of inflammatory proteins that mediate immune cell-mediated effects in OA.

Result

By integrating the inverse variance weighting method with the Bayesian weighting algorithm and conducting sensitivity analyses to exclude unstable factors, we identified 33 immune cell types with significant causal associations with OA (16 with a positive causal effect and 17 with a negative causal effect), as well as six inflammatory proteins (three promoting OA and three protective against OA). Further two-step analysis revealed that these six inflammatory proteins mediated the effects of nine immune cell phenotypes on OA. Among them, TRAIL exhibited the highest mediation proportion (10.99%) in the pro-OA effect of CD25 + IgD + CD24 + B cells and mediated 7.13% of the protective effect of CD16 − CD56 + NKT cells against OA.

Conclude

This study provides a comprehensive investigation of the causal relationships between immune cells and Knee OA, and estimates the mediating role of circulating inflammatory proteins. These findings contribute to identifying high-risk populations for OA and offer new insights for early prevention and clinical intervention.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13018-025-06374-y.

Highlights

• Mendelian Randomization and Bayesian weighted analysis explore the causal relationships between inflammatory proteins, immune cells, and OA.

• Mediation analysis explores the role of circulating inflammatory proteins in mediating the effects of immune cells on OA.

• Provides evidence for early prevention and intervention in OA.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13018-025-06374-y.

Introduction

Osteoarthritis (OA) is a common chronic degenerative disease of the joints characterized by progressive damage to articular cartilage, bone redundancy formation, structural changes in the subchondral bone, as well as alterations in the integrity of the synovium and joint capsule [1]. The incidence and prevalence of OA are increasing with the aging of the population and increasing life expectancy, placing a heavy burden on the healthcare system and significantly affecting the quality of life of patients [2, 3]. Although the exact pathogenesis of OA is not fully understood, factors such as aging, obesity, genetics, and low-grade inflammation have been suggested as key factors [4]. Therefore, it is important to diagnose OA in a timely manner and explore its pathogenesis in depth.

Circulating inflammatory proteins play a key role in the pathogenesis and development of OA. These proteins are responsible for modulating immune responses and inflammatory processes, thereby promoting or inhibiting joint tissue destruction and repair. Some typical pro-inflammatory cytokines, such as tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β), and IL-6, play an important role in the pathophysiology of OA, which has been demonstrated [5–7]. These inflammatory factors not only directly affect the stability and function of articular cartilage, but also promote an inflammatory response in the synovium and joint capsule, leading to pain, swelling and impaired joint function.

In recent years, the role of immune mechanisms in the pathogenesis and development of OA has received increasing attention. Recently, the classical definition of OA has shifted to an inflammatory disease driven by abnormal innate and adaptive immune responses [8]. Innate disease-free cells, such as natural killer (NK) cells, macrophages, and mast cells, play the most important pathogenic role in the early inflammatory response, whereas adaptive disease-free cells, such as CD4 + Th1 lymphocytes and antibody-producing B cells, contribute significantly to the onset of the inflammatory response [1]. It was also found that significant macrophage infiltration and alterations in T and B cell populations were present in OA synovium [9]. Regulatory T cells (Treg) were elevated in peripheral blood and synovial fluid in patients with OA, whereas synovial biopsies showed reduced levels of B cells and elevated levels of mast cells [10].

There are complex interactions and interactions between inflammatory proteins such as cytokines and chemokines and immune cells. Various immune cells, such as macrophages, T cells and synovial fibroblasts, coordinate the inflammatory response in the joint microenvironment by releasing pro-inflammatory cytokines and matrix-degrading enzymes [11]. Meanwhile, inflammatory proteins are also capable of influencing the migration, activation, and function of immune cells, thereby regulating the immune response and inflammatory process. This reveals a mediating role for immune cells in the regulation of OA by inflammatory factors. However, although the importance of these interactions has been recognized, the specific mechanisms of action and effects are still not fully understood and further studies are needed to elucidate them.

Mendelian randomization (MR) analysis utilizes single nucleotide polymorphisms (SNPs) from GWAS as instrumental variables (IVs) and is considered to be an effective method for assessing causality [12, 13]. The advantage of MR analysis is that it utilizes genotypes determined at the time of fertilization, minimizing bias due to confounding factors and reverse causality [14]. Therefore, in this study, publicly available summary statistics of 91 circulating inflammatory proteins, 731 immune cells, and OA were obtained from a recent genome-wide association study (GWAS) and analyzed by two-sample MR (TSMR), inverse MR, Bayesian weighted Mendelian randomization, BWMR) and two-step MR to assess the causal associations between circulating inflammatory proteins and the two main clinical types of OA (Knee OA) and to explore the level of immune cell mediation in the inflammatory protein-to-osteoarthritis pathway.

By combining genetic data from a large population-based cohort with detailed clinical phenotypes of OA, we hope to reveal the potential mechanisms of immune cell-mediated circulating inflammatory proteins in the pathogenesis and progression of OA. The results of this study may provide new targets for the treatment and prevention of OA and new strategies for improving clinical outcomes and quality of life of patients.

Materials and methods

Study design

The use of SNPs as IVs in MR analyses to assess causal associations between exposures (e.g., Immune cells and Circulating inflammatory proteins) and outcomes (Knee OA) requires that three basic assumptions be satisfied (Figure 1a). First, IVs must be strongly associated with exposure; second, IVs are not associated with any confounding factors; and finally, IVs influence outcomes only through exposure [15]. In causal analysis, the effect of an exposure on an outcome can be divided into total and indirect effects. The total effect refers to the direct influence of the exposure on the outcome, while the indirect effect refers to the influence mediated through intermediate factors. In this study, we used immune cells as the exposure, OA as the outcome, and selected circulating inflammatory proteins as potential mediators. We employed two-sample MR(TSMR) and multivariable MR methods to elucidate the potential mediating role of circulating inflammatory proteins in the causal relationship between immune cells and Knee OA (Fig. 1b). Specifically, we sequentially investigated the causal effects of immune cells on OA, inflammatory proteins on OA, and immune cells on inflammatory proteins. Using the product-of-coefficients method, we calculated the indirect effect of the exposure on the outcome, thereby elucidating the specific role of circulating inflammatory proteins in this causal pathway. The overall study design is shown in Fig. 1c. This causal analysis adhered to the Strengthening the Reporting of Observational Studies in Epidemiology, (STROBE) guidelines to ensure the scientific rigor and standardization of the study [16]. This study is not a clinical trial; therefore, a clinical trial number is not applicable

Fig. 1.

Schematic diagrams depicting the experimental design. a Three major assumptions to be followed in MR analysis; b Schematic of the methodology for a two-step MR analysis; c Overall technical line of this study

Data source

GWAS summary statistics for circulating inflammatory proteins were derived from a meta-analysis involving 11 cohorts totaling 14,824 participants of European ancestry, which finalized a total of 91 circulating inflammatory proteins as measured by the Olink Target platform on genomic genetic data and plasma proteomic data [17], and the complete data for each inflammatory protein were uploaded to the public open database GWAS Catalog (https://www.ebi.ac.uk/gwas/) and are freely available (login IDs GCST90274758 through GCST90274848).

Genetic data on immune cell traits were derived from a GWAS analysis of 272 blood immune cell-related factors in 3757 European individuals [18]. This analysis covered 731 immune traits including 118 absolute cell counts, 389 mean fluorescence intensities of surface antigens, 32 morphological parameters and 192 relative cell counts. These traits were further categorized into 7 cell groups including 43 monocytes, 190 B cells, 64 CD cells, 64 myeloid cells, 124 TBNK cells, 79 T cell maturation stages and 167 Treg cells. The relevant complete data are available from the OpenGWAS database (https://gwas.mrcieu.ac.uk/) (login IDs GCST90001391 through GCST90002121).

The summary statistics for OA were obtained from the largest GWAS meta-analysis published to date, which aggregated data from nine populations [19], involving 826,690 individuals (including 177,517 OA patients), and covering 11 OA-related traits. In this study, we selected a common types of osteoarthritis: Knee OA.

IVs selection

In this study, we initially screened SNPs associated with immune cells, inflammatory proteins, and OA by genome-wide association analyses as IVs for MR analyses. To ensure that the IVs met the first assumption of MR (i.e., that the IVs were strongly associated with the exposure variables), we employed a P < 5 × 10^{− 8} threshold for screening SNPs. However, the number of instrumental variables obtained under this rigorous screening criterion was more limited. To ensure the inclusion of a sufficient number of significant SNPs and improve the statistical power of the analysis, we relaxed the P-value threshold to P < 1 × 10^{− 5}. However, for immune cells acting as mediators, the threshold of P < 5 × 10^{− 8} was still applied [20]. However, genetic variants with similar genomic locations are more likely to be co-inherited, resulting in a higher than random probability that they occur on the same chromosome. Therefore, we used the clump_data() function in the R package TwoSampleMR to constrain the linkage disequilibrium condition (set kilobase pairs (kb) = 10,000, r² = 0.001) to ensure that the selected SNPs are independent [21].

To further ensure the accuracy of the data, potential duplicates or palindromes were removed using the harmonise_data() function with the parameter set to ‘action = 2’. The F-test value was then calculated for each SNP using the formula F = (N-K-1)R²/K(1-R²) to assess the strength of the correlation between loci and exposure factors. According to the criteria, if the F value was less than 10, the SNP was considered a weak instrumental variable and was excluded. In the formula, R² is the proportion of variance explained by the IVs, N is the sample size, and K is the number of SNPs [22]. Through this rigorous screening and validation process, we ensured that the selected IVs had sufficient statistical validity and independence for subsequent Mendelian Randomization analyses.

MR analysis

This study utilized the “TwoSampleMR” package to conduct MR analyses on immune cells, inflammatory proteins and OA. When the exposure included only a single SNP, the Wald ratio method was used to estimate causal effects [22, 23]. The Wald ratio method is appropriate for single instrumental variables, as it directly calculates the ratio of the SNP-outcome association to the SNP-exposure association, providing a straightforward and reliable estimate. For exposures with two or more instrumental variables, the inverse-variance weighted (IVW) method was applied [24]. The IVW method assumes all SNPs are valid instrumental variables and combines estimates from multiple SNPs using a meta-analytic approach. By pooling information from multiple independent variants, this method improves statistical power and enhances precision [24]. To further validate causal inference under the IVW framework, Bayesian weighted Mendelian randomization (BWMR) analysis was also employed. BWMR effectively eliminates violations of instrumental variable assumptions through posterior covariance adjustment, thereby improving the reliability of causal effect estimates [25]. Results were expressed per standard deviation (SD) increase in exposure levels and reported as odds ratios (ORs) with 95% confidence intervals (CIs). A p-value of less than 0.05 was considered indicative of a suggestive causal relationship.

Sensitivity analysis

In a concerted effort to ascertain the fidelity of causal attributions and to obviate any deleterious factors encumbering causal hypotheses, a battery of stringent quality control measures was instituted, appraising the sensitivity, heterogeneity, and multiplicity of study findings. These encompassed Cochran’s Q test, deploying the IVW method to derive p-values aimed at rectifying heterogeneity-induced biases, alongside MR-Egger intercepts to rectify horizontal gene pleiotropy following outlier exclusion [26]. Furthermore, the impact of individual single nucleotide SNP drivers on randomized estimates was rigorously scrutinized via leave-one-out analyses.

Additionally, we conducted the Steiger directionality test to confirm the forward causal trajectory between the exposure and outcome [27, 28]. For the Steiger directionality test, a P-value less than 0.05 suggests a forward relationship between the exposure and the outcome.

Mediation analysis

To assess the mediating effect of circulating inflammatory proteins, we first calculated the total effect (β_all) of immune cells on OA, followed by the indirect effect of inflammatory proteins on OA (β2) and the indirect effect of immune cells on inflammatory proteins (β1). Next, we computed the product of β1 and β2 to obtain the mediation effect (β1β2). Finally, to quantify the mediating role of circulating inflammatory proteins in the causal relationship between immune cells and OA, we used the mediation ratio (β1β2/β_all). This ratio reflects the proportion of the total effect that is mediated by inflammatory proteins in the causal pathway between immune cells and OA [29].

Results

Selection of IVs

Based on the preset screening criteria (P < 1 × 10^− 5, kb = 10,000, r² = 0.001), 731 immune cells had significant IVs ranging from 3 to 759, 91 inflammatory proteins had significant IVs ranging from 6 to 54. Detailed information on SNPs, effector alleles, allele frequencies and beta values for each IV is stored in S1 Table. The smallest F value for these SNPs was 19.508, indicating the lowest probability of weak instrumental bias.

Causal effects of immune cells on KneeOA

In the multivariable MR analysis, we employed the IVW method as the primary analytical approach and identified 45 immune cells with significant causal associations with Knee OA (Fig. 2a; S2 Table). Subsequently, we conducted BWMR analysis on these 45 immune cells, excluding nine with p-values greater than 0.05, resulting in a final selection of 36 immune cells (Fig. 2b). To further eliminate potential confounding factors, we performed sensitivity analyses on the 36 causal associations identified by both IVW and BWMR (Fig. 2c). Heterogeneity of the instrumental variables was assessed using Cochran’s Q statistic, revealing significant heterogeneity in three inflammatory proteins: TRAIL, IL-6, and TNF-α (P < 0.05), suggesting that their causal effect estimates may vary due to different genetic instruments. Additionally, to evaluate horizontal pleiotropy, we conducted the MR-Egger intercept test, which detected no significant horizontal pleiotropy in any immune cells, indicating that directional pleiotropy had minimal impact on the causal effect estimates.

Fig. 2.

Forest plot showing causal relationship between immune cell and Knee OA a Circle plots of MVMR results of immune cell with Knee OA ; b Circle plots of BWMR results of immune cell with Knee OA; c Forest plots of the results under the two MR methods analyses of the proteins with a Pval of less than 0.05 (OR greater than 1 indicate a positive correlation and less than 1 indicates a negative correlation)

Through the above analysis, we identified 33 immune cell types significantly associated with Knee OA, including 21 blood protein measurements, 6 lymphocyte counts, 5 leukocyte counts, and 1 myeloid white cell count. Among them, 16 immune cell types (such as IgD- CD27- %B cell, IgD- CD24- B cell %lymphocyte, TD CD4+ %CD4+, and CD19 on IgD- CD27- B cell) showed a positive association with Knee OA (OR > 1), suggesting their potential role in promoting the onset of Knee OA. In contrast, 17 immune cell types, including CD8br AC, CD4 + CD8dim %lymphocyte, and CD28 + CD45RA- CD8br %T cell, exerted a negative regulatory effect on Knee OA (OR < 1). Furthermore, the direction of β estimates for these causal associations remained consistent between the MVMR and BWMR methods (Fig. 2c), further strengthening the robustness of the findings. Additionally, Steiger’s test (Fig. 2c) indicated that the genetic susceptibility to Knee OA did not affect any immune cell characteristics, suggesting no reverse causality (P < 0.05).

Causal effects of circulating inflammatory proteins on KneeOA

In line with the previous methods, we performed BWMR to validate the results from the MVMR analysis, identifying 6 out of 91 circulating inflammatory proteins that were significantly causally associated with Knee OA (p < 0.05) (Fig. 3a, b). The results showed that the genetically predicted C-X-C motif chemokine ligand 1 (CXCL1) (OR: 1.050; 95% CI 1.010–1.091; p = 0.014), Interleukin-17 C (IL-17 C) (OR: 1.046; 95% CI 1.004–1.090; p = 0.031), and Neurotrophin-3 (NT-3) (OR: 1.058; 95% CI 1.009–1.110; p = 0.019) showed a positive association with Knee OA, indicating that higher expression levels of these proteins correspond to an increased risk of Knee OA. In contrast, Knee OA was negatively regulated by Interleukin-10 receptor subunit alpha (IL-10Rα) (OR: 0.939; 95% CI 0.889–0.992; p = 0.024), transforming growth factor beta-1 (TGF-β1) (OR: 0.931; 95% CI 0.890–0.975; p = 0.002), and Tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) (OR: 0.959; 95% CI 0.931–0.988; p = 0.005). The β estimates reflecting these causal relationships remained consistent across both the MVMR and BWMR methods (Fig. 3c). Furthermore. we conducted sensitivity analyses to account for potential confounding factors. The results indicated no significant heterogeneity, (Cochran’s Q test P > 0.05) or horizontal pleiotropy, (Egger intercept P > 0.05), and no evidence of reverse causality was detected, (Fig. 3c). These findings suggest that our results are highly robust and not influenced by potential biases or heterogeneity.

Fig. 3.

Circulating inflammatory proteins with a causal relationship to Knee OA. a Inflammatory proteins identified by the MVMR method as having a causal relationship with Knee OA; b BWMR validation of significant causal relationships between inflammatory proteins and Knee OA; c Forest plot of the results for six proteins analyzed using two MR methods

Mediated Mendelian randomization analysis

Based on immune cells and circulating inflammatory proteins that showed significant causal associations with Knee OA, we further employed the MVMR approach to explore the potential causal relationships between these two factors. Specifically, immune cells were treated as exposure variables and inflammatory proteins as outcome variables to assess their direct associations while accounting for potential confounders. As a result, we identified 13 causal relationships involving 9 immune cell types and 6 inflammatory proteins (Fig. 4). Notably, our findings suggest that multiple immune cell types may exhibit causal relationships with the same inflammatory protein. For instance, both CD28 + CD45RA − CD8br %T cells and CD16 − CD56 on NKT cells positively regulated CXCL1 expression. Additionally, 6 immune cell types, including HLA-DR on CD14 + CD16 + monocytes, IgD on IgD + CD38br cells, and CD39 + CD4 + AC, were causally associated with TRAIL, with 4 showing negative associations and 2 showing positive associations.

Fig. 4.

The relationship between immune cells and inflammatory proteins that have significant associations with Knee OA.

To further validate the robustness of these findings, we conducted a series of sensitivity analyses, including heterogeneity testing, horizontal pleiotropy assessment, and reverse causality testing. The results indicated that our causal estimates were not significantly affected by instrumental variable heterogeneity, directional pleiotropy, or potential reverse causality (Fig. 4). These rigorous validation steps enhance the reliability of our conclusions and provide a solid foundation for further investigating the mechanistic pathways linking immune responses, inflammatory mediators, and OA progression.

Potential Mediation effect calculation

In the mediation analysis, a necessary condition for inflammatory proteins to function as mediators is that the indirect effect of immune cells on Knee OA through inflammatory proteins must be directionally consistent with the overall effect of immune cells on Knee OA. Therefore, after establishing the causal relationships among nine immune cell types, six inflammatory proteins, and Knee OA, we proceeded with effect estimation. Specifically, we calculated the total effect (β_all) of immune cells on Knee OA, the effect of immune cells on inflammatory proteins (β1), and the effect of inflammatory proteins on Knee OA (β2). The indirect effect (β1β2) was then determined, and the mediation proportion was quantified by dividing the indirect effect by the total effect.

The results of this analysis are presented in Table 1. We identified two inflammatory proteins mediating the relationship between five immune cell phenotypes and Knee OA. Notably, TRAIL mediated the effect of four immune cell types on Knee OA. Specifically, it accounted for 10.99% of the pro-OA effect of CD25 on IgD + CD24 + B cells and 6.72% of the effect of Activated & secreting CD4 regulatory T cells %CD4 + T cells in promoting Knee OA. Conversely, TRAIL also mediated 7.13% of the inhibitory effect of CD16 − CD56 on NKT cells and 4.11% of the effect of IgD on IgD + CD38br cells in suppressing Knee OA progression. The second key mediator identified was IL-17 C, which accounted for 4.28% of the protective effect of HLA-DR on CD14 + CD16 + monocytes against disease progression.

Table 1.

Mediation proportion of inflammatory proteins in the causal effect of immune cells on knee OA

Immune cell	Protein	β_all	β1	β2	Mediation proportion	P
CD25 on IgD + CD24 + B cell	TRAIL	− 0.0314	− 0.0420	0.0013	0.0107	10.99%
CD16-CD56 on NKT	TRAIL	0.0340	− 0.0420	− 0.0014	− 0.0186	7.13%
Activated & secreting CD4 regulatory T cell %CD4 + T cell	TRAIL	− 0.0096	− 0.0420	0.0004	0.0056	6.72%
HLA DR on CD14 + CD16 + monocyte	IL17C	− 0.0166	0.0310	− 0.0005	− 0.0115	4.28%
IgD on IgD + CD38br	TRAIL	0.0293	− 0.0420	− 0.0012	− 0.0288	4.11%

Discussion

In recent years, research on the pathological mechanisms of OA has advanced significantly, with increasing evidence indicating that immune cells play a critical role in OA onset and progression [30–32]. Although numerous observational studies have demonstrated a strong association between immune cells and OA, the presence of confounding factors often makes it difficult to establish causality. Moreover, inflammatory responses in OA exhibit a complex bidirectional regulatory pattern, suggesting that circulating inflammatory proteins may serve as mediators in the immune cell-regulated OA process [33]. However, the precise mechanisms by which immune cells influence OA through inflammatory proteins remain largely unexplored [34]. To address this gap, our study employs Mendelian randomization to construct a causal inference model using genetic instrumental variables. This approach aims to elucidate the direct effects of immune cells on OA as well as the potential mediating role of circulating inflammatory proteins, thereby providing theoretical insights for a deeper understanding of OA pathogenesis and the development of targeted intervention strategies.

Our study systematically evaluated the causal relationship between 731 immune cell phenotypes and Knee OA. The results identified 16 immune cells significantly associated with OA onset, including 9 B cells, 11 T cells, and 13 other immune cells (such as dendritic cells, monocytes, and NKT cells). Previous studies have also demonstrated the involvement of B cells and T cells in OA pathogenesis, with T cell infiltration in the joints recognized as a hallmark of OA. A 2020 study involving 30 OA patients and 45 healthy controls, utilizing peripheral blood flow cytometry, reported no significant differences in the overall proportions of T cells, CD8⁺ T cells, and B cells. However, notable alterations in subset distributions were observed, including increased levels of CD4⁺ T cells, the CD4⁺/CD8⁺ T cell ratio, CD8⁺CD45RA⁻ T cells, CD4⁺ T cells expressing different CCR7 levels, T helper 17 (Th17) cells, and follicular helper T cells (Tfh2) [35]. Additionally, the CD4⁺/CD8⁺ T cell ratio in OA patients reached 5:1, compared to approximately 2:1 in normal synovial tissue, suggesting a significant shift in the T and B cell repertoire in OA [36]. Monocytes, based on specific surface marker expression, can be classified into three major subtypes: CD14 + + CD16⁻ (classical monocytes, CM), CD14 + + CD16⁺ (intermediate monocytes, ITM), and CD14⁺dimCD16⁺⁺ (non-classical monocytes, NCM) [37]. Classical monocytes predominate within the monocyte population and play a crucial role in pathological clearance and inflammatory responses. A study highlighted the importance of CCR2 in dorsal root ganglion (DRG) neurons in regulating macrophage infiltration and pain sensitivity in OA [38]. Furthermore, another study demonstrated that CCR2-deficient mice exhibited reduced monocyte/macrophage infiltration into joint tissues, conferring partial protection against OA [39]. Similarly, inhibition of the CCL2/CCR2 signaling pathway significantly reduced macrophage accumulation, synovial inflammation, and cartilage damage [40]. In addition, previous research showed that CCL4L2 promotes M1 polarization, thereby aggravating inflammatory responses and suppressing TSPC differentiation [41]. Beyond these immune cell types, other immune components also play pivotal roles in OA pathogenesis. NKT cells, which bridge innate and adaptive immunity, have garnered increasing attention for their dual regulatory role in OA. By secreting various cytokines, NKT cells can either promote or suppress inflammation and apoptosis [42]. Moreover, regulatory B cells (Bregs) may mitigate local inflammation and slow cartilage degeneration by secreting anti-inflammatory cytokines such as IL-10 [43]. Although previous studies have elucidated the role of immune cells in OA pathogenesis, our findings reveal some differences in immune cell phenotypes. These discrepancies likely arise from our more refined classification of immune cell subpopulations, which allowed us to detect subtle differences previously overlooked. Nonetheless, prior research confirms the critical role of immune cells in OA. Building on this, our study refines and expands the immune cell landscape in OA, offering novel insights into the immunoregulatory mechanisms driving OA development and progression.

Given the critical role of immune cells in OA, inflammatory proteins may play a crucial regulatory role in mediating OA development. Increasing evidence suggests that circulating inflammatory proteins are not only associated with OA severity but may also serve as mediators of immune cell effects on OA. In this study, we identified six inflammatory proteins with causal relationships with OA, including IL-10Rα, TGF-β1, and TRAIL. IL-10 and its receptor, IL-10Rα, are well known for their anti-inflammatory and immunoregulatory properties, particularly in cartilage protection. Studies have shown that IL-10 is highly expressed in both healthy and OA chondrocytes and can inhibit the pro-inflammatory effects of IL-1 and TNF-α, thereby mitigating cartilage degeneration. Furthermore, IL-10 and its receptor IL-10R exhibit the highest expression levels in fetal cartilage [44], suggesting their potential regulatory role in cartilage development and OA pathogenesis.

TRAIL (APO-2 L) is a cytokine family protein that primarily induces apoptosis signaling pathways by binding to death receptors (TRAIL-R1/DR4 and TRAIL-R2/DR5) [45]. Studies have demonstrated that TRAIL overexpression in OA chondrocytes activates apoptosis-related factors such as caspase-3 and PARP, indicating its potential involvement in OA pathogenesis [46]. Additionally, moderate mechanical stimulation has been shown to regulate TRAIL expression through the NF-κB/NLRP3 pathway, exerting a protective effect on OA to some extent [47]. TNF is a multifaceted cytokine produced by immune cells, intricately involved in the pathogenesis of a wide range of physiological processes. Its influence extends across the spectrum of inflammation regulation, anti-tumour defences and immune system equilibrium [48, 49]. TNF was originally identified for its capacity to induce tumour necrosis, which is the basis of its nomenclature. However, it has evolved beyond its original role within the immune system, becoming a central mediator of a vast array of physiological and pathological processes. It is notable that TNF exists in two principal forms: TNF-α and TNF-β (also known as LT-α) each exert distinct regulatory functions within the intricate tapestry of cellular signalling and immune modulation.TNF-α is a product of various immune cells, including macrophages, monocytes, and T lymphocytes, among others. In contrast, TNF-β is primarily generated by activated T lymphocytes [50]. Both isoforms of TNF are intricately involved in a multitude of physiological and pathological processes, including inflammatory reactions, immune modulation, and apoptosis. Research has demonstrated that TNF plays a multifaceted role in the perpetuation of chronic joint inflammation, through the activation of NF-kB [51]. Additionally, TNF has been shown to instigate macrophage activation and to foster osteoclast proliferation and differentiation. These intricacies are intimately entwined with the genesis of OA-related pain. It is noteworthy that in a sodium iodoacetate (MIA)-induced murine model of OA pain, both synovial tissues and the joint capsule exhibited heightened expression levels of TNF-α and IL-6 from day 1 to 28, with the zenith observed on day 4 [52]; In addition, levels of pro-inflammatory cytokines TNF-α and IL-6 were also significantly elevated in both the femoral head tissue and serum of patients with steroid-induced osteonecrosis of the femoral head [53].

To further investigate the mediating role of inflammatory proteins in the effect of immune cells on OA, we conducted a mediation analysis and identified TRAIL and IL-17 C as significant mediators. Interestingly, TRAIL was found to mediate the causal relationship between four immune cell types and OA. Specifically, CD25 on IgD + CD24 + B cells and Activated & Secreting CD4 Regulatory T cells (%CD4 + T cells) exhibited a positive association with Knee OA while being negatively associated with TRAIL. Additionally, TRAIL itself showed a negative correlation with Knee OA. These findings suggest that elevated expression levels of these two immune cell types may reduce TRAIL expression, which in turn increases the risk of developing Knee OA. The downregulation of TRAIL might weaken its protective effects in maintaining cartilage homeostasis and inhibiting pathological chondrocyte apoptosis, thereby contributing to OA progression. In contrast, the expression levels of CD16 − CD56 on NKT cells and IgD on IgD + CD38br cells were positively correlated with both TRAIL and OA. This indicates that the upregulation of these immune cell types may enhance TRAIL expression, which could exert a protective effect against Knee OA. Given that TRAIL is known to modulate immune responses and regulate apoptosis-related pathways, its increased expression might mitigate OA pathogenesis by promoting chondrocyte survival and limiting inflammatory damage. These results highlight the complex interplay between immune cells, inflammatory mediators, and OA progression, emphasizing the critical role of TRAIL as a key intermediary in immune-mediated OA pathophysiology.

This study employs Mendelian randomization to assess the causal relationship between immune cells and Knee OA, effectively minimizing reverse causality and controlling for potential confounding factors. By leveraging large-scale GWAS data, this study eliminates the need for long-term follow-up, thereby conserving time and resources while identifying immune cells and inflammatory proteins causally associated with Knee OA, providing valuable insights for future research. Although the mediating effect of inflammatory proteins between immune cells and Knee OA is limited, their significance in immune responses and disease progression should not be overlooked. Further exploration of the roles of these inflammatory proteins may facilitate the identification of potential therapeutic targets and offer new perspectives for the early diagnosis and personalized treatment of Knee OA. In the future, integrating the causal relationship between immune cells and Knee OA with the mediating role of inflammatory proteins may aid in the development of novel immunomodulatory therapies, ultimately improving patient prognosis and quality of life.

The present study stands as a pioneering endeavor in unraveling the intricate interplay of immune cell mediation within the causal nexus between specific circulating inflammatory factors and OA. Despite uncovering noteworthy associations, several caveats warrant consideration. Firstly, the utilization of pooled genetic statistics sourced primarily from individuals of European ancestry may constrain the extrapolation of findings to other demographic cohorts, thereby potentially attenuating the robustness of the results against population-specific confounders. Secondly, the relaxation of the screening threshold for SNPs to 1e−5, while aimed at maximizing the identification of IVs, necessitates a delicate balance between efficacy and stringency. Finally, while MR tools offer invaluable insights into causal inference, the imperative of corroborating findings through biologically controlled experiments remains paramount. Integrating these considerations will fortify the validity and applicability of our findings, facilitating a deeper understanding of the pathophysiological mechanisms underpinning OA etiology.

Conclusions

In summary, our study meticulously explored the intricate causal interplay between immune cell traits, circulating inflammatory proteins and Knee OA. Leveraging the robust analytical framework of Two-step MR, we unveiled a compelling array of five-group mediating relationships, shedding light on the nuanced role of circulating inflammatory proteins in this dynamic process. This groundbreaking finding holds immense promise in guiding the identification of potential biomarkers and therapeutic targets for OA, paving the way for more targeted interventions and personalized treatment strategies in the management of this debilitating condition.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Ye Yuan and Yanrong Liu conceived and designed the project; Xingyu Liang is responsible for acquiring, cleaning, and managing the genetic data and phenotypic data required for the research. Ye Yuan conducts the primary statistical analysis for Mendelian randomization, selects appropriate tools and methods, and interprets the analysis results. Sining Kang assists with the selection and implementation of analytical methods, reviews the analysis code, and checks the results. Ye Yuan was responsible for drafting the manuscript, with revisions made by both Ye Yuan and Yanrong Liu on important content; All authors had full access to the data in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

No funding.

Data availability

The circulating inflammatory protein data used for the MR analysis in this study are freely available from the GWAS Catalog (https://www.ebi.ac.uk/gwas/) under the IDs GCST90274758-GCST90274848. Data for 731 immune cell phenotypes can be accessed from the OpenGWAS database (https://gwas.mrcieu.ac.uk/) with IDs ranging from GCST90001391 to GCST90002121. The GWAS data for osteoarthritis can be freely accessed through The Genetics of Osteoarthritis consortium (https://msk.hugeamp.org/downloads.html).

Declarations

Ethics approval and consent to participate

Ethical approval was not required for the study involving humans in accordance with the local legislation and institutional requirements. Written informed consent to participate in this study was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and the institutional requirements.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.