Ankylosing spondylitis (AS) is a disabling chronic arthritis resulting from a combination of factors. Genetic susceptibility does not fully explain the etiology of AS, and the effect of the interplay of genes, sex, microorganisms, mechanical stress, and additional lifestyle and environmental factors on increased susceptibility to AS is unclear, which adds to the complexity of treatment (1). Targeted biologic therapies are the mainstay treatment for patients with AS who are difficult to treat with nonsteroidal anti-inflammatory drugs (2). In recent years, the successful introduction of monoclonal antibodies against AS (interleukin [IL]-17A and tumor necrosis factor-α [TNF-α]) has exemplified numerous vital pathological pathways. Nonetheless, <50% of patients respond favorably to IL-17A and TNF-α blockade. Currently, curative treatment for AS is lacking, and many patients have to manage symptoms through lifelong medication, which can induce adverse reactions (3). Consequently, exploring novel diagnostic biomarkers and therapeutic agents is crucial to provide more effective diagnostic methods and therapeutic options for patients with AS, thereby improving patient prognosis.

Human proteins are crucial in diverse biological processes and represent the main drug targets. Nelson et al. (4)revealed that protein drug targets linked to genetic association-supported diseases are likely to be approved for marketing, with a possible twofold increase. In recent years, Mendelian randomization (MR) analyses have been extensively applied to develop and repurpose drug targets (5). Genetic instrumental variable analysis involves the use of single nucleotide polymorphisms (SNPs) from genome-wide association studies (GWAS) as the genetic approach to estimate the causal relationship between the exposure and the outcome. In contrast to observational studies, MR avoids the effects of confounding factors. Progress in plasma high-throughput genomic and proteomic technologies has allowed MR-based strategies to facilitate the identification of candidate therapeutic targets for diverse disorders (e.g., type 1 diabetes and breast cancer) (6, 7). However, MR analyses that integrate GWAS with protein quantitative trait locus (pQTL) data, an approach that could provide important insights for early disease diagnosis and drug target discovery, are lacking in AS research. Consequently, the present study aimed to integrate large-scale GWAS with pQTL data using MR analysis to investigate plasma proteins that could be candidate biomarkers or therapeutic targets of AS.

We established the precise control of AS by discovering potential drug targets from plasma proteins. Figure 1 shows the study design. First, MR analysis of two samples was performed, and eight potential drug proteins were screened after FDR correction. The plasma protein data were extracted from plasma proteomics-related publications, whereas the AS data were obtained from the GWAS data of the International Genetics of Ankylosing Spondylitis Consortium (IGAS) (8). Secondly, we performed colocalization analysis to verify the robustness of the genetic associations between plasma proteins and AS. Third, the relationship between the identified proteins was analyzed using the protein-protein interaction (PPI) network to identify potential therapeutic targets. Fourth, a sensitivity analysis, including Replication and meta-analysis, was conducted to ensure the accuracy and directionality of the identified associations. Finally, we assessed the potential adverse effects of identified drug proteins on other phenotypes using phenome-wide Mendelian randomization analysis.

The development of new therapeutic agents for ankylosing spondylitis is challenging. One of the main reasons for this difficulty is the incomplete understanding of the pathophysiology of AS. The human proteome is the primary therapeutic target, and protein drug targets possess significant clinical value. Therefore, an integrative analysis was conducted to identify novel anti-AS therapeutic targets to evaluate the causative proteins of AS based on prior GWAS (21).

The “causality” detected through MR could be genetic confounding, or horizontal pleiotropy because of LD (9). Thus, only cis-pQTL were used as instruments to limit horizontal pleiotropy bias because they directly affect the translation and transcription of the gene of interest. In this study, we analyzed the potential pathogenic relationship between plasma proteins and the risk of AS and identified potential therapeutic targets among plasma proteins for AS. After adjusting for FDR, the plasma proteins ERAP1, IL12B, IL18R1, IL23R, IL7R, CCL8, TNFAIP6, and TYMP showed a causal relationship with the risk of AS. The Steiger filtering method were used to validate the directionality of the causal relationships (22). These methods showed no evidence of reverse causation for the proteins identified in the primary analysis. This reinforces the finding that IL23R, IL7R, and TYMP are likely contributors to the pathogenesis of AS, suggesting that these proteins play a causative role in the disease’s development rather than being a consequence of it. The PPI networks revealed that IL23R and IL7R interact with current AS drug targets TNF-α and IL17, indicating their potential as new therapeutic targets. Additionally, database searches identified three approved drugs—tipiracil hydrochloride, ruxolitinib, and celecoxib—targeting TYMP, IL7R, and IL23R, respectively, suggesting their potential for repurposing in AS treatment (23). In the replication analysis section, we attempted to replicate the study results using FinnGen AS GWAS as the outcome. We found that the MR result for IL23R remained significant in the replication analysis. As for IL23R and TYMP, although the direction of the MR results was consistent with the discovery results, they did not reach significance in the replication cohort. This may be due to the smaller sample size of the replication cohort, which reduces statistical power. In addition, our MR PheWAS results indicate that developing these potential drug proteins appears to be safe, as there is little statistical evidence to suggest that they will produce harmful side effects (24).

IL23R is implicated in the IL-23/IL-17 axis, which plays a crucial role in the pathogenesis of AS. Studies have shown that IL-23 stimulates the expansion and maintenance of Th17 cells, leading to increased levels of IL-17, a proinflammatory cytokine involved in AS (25, 26). Genetic variants in IL23R have been associated with AS susceptibility, highlighting it as a potential therapeutic target. IL7R is essential for T-cell development and homeostasis. Variants in IL7R have been linked to multiple autoimmune diseases, including AS. The IL-7 receptor is involved in the survival and proliferation of T cells, which are critical in the inflammatory processes underlying AS (27). Targeting IL7R could modulate immune responses and potentially ameliorate AS symptoms. Of the eight AS-related proteins identified in the present study, ERAP1, IL12B, and IL23R were shown to be related to AS via numerous animal experiments and clinical studies, with ERAP1 suggested to be the second most potent gene related to AS (28, 29). This further establishes the reliability of our study of ERAP 1 as a drug target for treatment of AS. ERAP1 and HLA-B27 contribute to 70% of familial genetic factors for AS (30). ERAP1 polymorphisms have been shown to affect AS susceptibility in HLA-B27+ people (31). Thus, ERAP1 can exert synergistic effects on factors such as HLA-B27 and is related to abnormal peptide processing and incorrect antigen presentation, leading to AS susceptibility. According to one case−control association study, protective genetic variants are related to a decrease in ERAP1 and ERAP2 function and inhibition of cell surface major histocompatibility complex I expression (8). ERAP1 and ERAP2 variants can affect associated peptide numbers to decrease the accelerated HLA-B27 folding rate, thus aggravating ER stress while accelerating AS progression (32). Mei et al. (33) reported that AS patients had higher serum IL-17 and IL-23 levels than did normal participants. IL-23, an IL-12-associated cytokine, may promote T helper (Th)17 cell growth and differentiation (34). Th17 cells are involved in the pathogenesis of AS (35) and are also implicated in psoriasis, inflammatory arthritis, and Crohn’s disease (36–38). IL-12B is likely related to the pathogenesis of AS because of its effects on IL-23R+CD4+ T cells and preferential stimulation of the above cells to release IL-17 predominantly (39). The present systematic study based on MR studies precisely addresses the inability of previous animal experiments to infer causality. Moreover, we identified novel AS-associated proteins (TYMP, CCL8, and TNFAIP6). TNFAIP6, also known as TSG 6, is a class of proteins induced by TNF-α that is secreted during acute inflammatory responses and is associated with inflammation and tissue remodeling (40). Previous studies have reported that TNFAIP6 has anti-inflammatory effects in an experimental mouse model of arthritis (41), but further evidence to support this observation is lacking. Moreover, the present study suggested that TNFAIP6 may have similar anti-inflammatory protective effects on AS via MR, providing a strong reference for the TNFAIP family as the causative genes for AS. CCL8 is a monocyte chemokine that can interact with CCR1, CCR2B, and CCR3; it also regulates tumor occurrence, antiviral infections, and inflammatory immunity in the host (42). Consistent with the present findings, one study suggested that CCL8 is upregulated in the serum of patients with AS and may be a useful biomarker for predicting active inflammation in the sacroiliac joints of patients with AS (43). TYMP catalyzes reversible thymidine phosphorylation and is suggested to have a critical effect on angiogenesis, tumor growth, migration, and invasion (44). A recent study showed that TNF-α strongly stimulates TYMP expression in fibroblast-like synoviocytes. Thus, we hypothesize that TNF-α may affect AS by inducing TYMP expression.

Previously, Zhao et al. (20)studied drug targets for ankylosing spondylitis using the UK Biobank-PPP database. Given that the pQTL data used by Zhao et al. differs from the data source in our study, we conducted a comparative analysis of their findings. The results indicate that the target proteins we identified are entirely different from theirs. For instance, the absence of IL23R protein measurement in the UK Biobank led to its omission in Zhao et al.’s study, thereby extending the data coverage scope in our research. Additionally, the MR analysis results for TYMP in Zhao et al.’s study indicated potential significance (P=0.0037), suggesting a potential causal role of this protein in the risk of AS. However, after p-value correction, TYMP was not selected as a candidate drug protein in their study. Furthermore, our discovery of drug-target candidate proteins (IL23R, TYMP, and IL7R) passed rigorous screening, with co-localization analysis strongly supporting their shared causal variants with AS in the region (PPH4 > 0.90). Moreover, our MR-PheWAS results showed almost no statistical evidence of adverse side effects for these candidate drug proteins, indicating that their development seems to be safe. Overall, these findings demonstrate the feasibility and practicality of our candidate proteins in drug development. We believe that the screening of these targeted proteins and the analysis of adverse effects can provide valuable references for research teams developing new targeted drugs for AS.

However, this study has certain limitations. First, since the instrumental variables included in this study mainly consist of a single cis-acting SNP and lack trans-pQTLs, it is not possible to determine the sensitivity test results at this level, which affects pleiotropy and heterogeneity analyses. Nevertheless, the F-statistics for our selected SNPs were all greater than 10, indicating minimal weak instrument bias. Second, the coloc method assumes single causal variants, which may not be accurate, and we are unable to assess AS progression genes (as GWAS identifies AS susceptibility genes). Therefore, our focus is on finding targets for disease prevention rather than treatment. Third, this study’s exposure and outcome data came from groups with European heritage. To translate these findings into practical applications, more study on non-European ancestries is needed in order to apply the conclusions to other locations, including Asia, Africa, and the Americas. Fourth, the results of the PPI study are suggestive rather than definitive, despite the fact that we were able to identify certain connections between pathogenic proteins and therapeutic targets of existing MS medications. It is need to do further research involving people who are healthy and patients with AS to confirm these correlations.

Our study identified eight plasma proteins significantly associated with ankylosing spondylitis risk using Mendelian randomization and colocalization analyses. We provided robust evidence for the causal roles of IL23R, IL7R, and TYMP, highlighting them as promising therapeutic targets. Phenome-wide MR analysis also evaluated potential side effects of these targets. Despite limitations, our findings enhance understanding of the genetic architecture of AS and support future research to validate these targets for clinical application.