Integrative multi-omics analyses identify PKD1 and SLC2A4 as genetically supported glycolysis-related candidate genes for rheumatoid arthritis

Rheumatoid arthritis (RA) is a chronic autoimmune disorder marked by persistent synovitis, progressive joint damage, and systemic inflammation, affecting approximately 0.5–1% of the global population (1). Despite advances in biologics and targeted therapies, many patients continue to experience disease flares and irreversible joint destruction. Beyond immune dysregulation, metabolic remodeling has emerged as a fundamental component of RA pathogenesis. Synovial fibroblasts and infiltrating immune cells reprogram their metabolism in response to inflammatory and hypoxic microenvironments, switching from oxidative phosphorylation to aerobic glycolysis (2, 3). This glycolytic shift supports energy-intensive processes like proliferation, matrix degradation, and cytokine production, which exacerbate joint inflammation. Metabolites such as lactate and succinate also accumulate locally, serving as signaling molecules that amplify inflammation and matrix remodeling (4). Consequently, metabolic adaptation not only sustains pathogenic cell function but also reshapes the RA synovial ecosystem.

Fibroblast-like synoviocytes (FLS) in RA exhibit a highly activated phenotype that mimics tumor-like behavior, including hyperproliferation, invasiveness, and resistance to apoptosis (5). This phenotype is driven by increased glucose uptake and elevated expression of glycolytic enzymes, such as HK2, LDHA, and PFKFB3 (6). Similarly, CD4+ and CD8+ T cells within the inflamed synovium undergo metabolic rewiring that enhances their effector functions, partly through HIF-1α and mTORC1-mediated pathways (7, 8). Various regulators—including metabolic enzymes and signaling proteins—can influence glycolytic switching and proinflammatory phenotypes in immune cells (7). Glycolytic inhibition in both FLS and immune cells has been shown to reduce inflammatory mediator secretion and cell migration, supporting the rationale for exploring glycolysis-targeted strategies in RA (9, 10). However, the upstream regulatory architecture linking genetic susceptibility to these glycolytic phenotypes in RA remains incompletely understood.

Recent advances in multi-omics integration have enabled systematic exploration of gene regulation in complex diseases. Summary-data-based Mendelian randomization (SMR) leverages genetic variants as instruments to infer causal links between molecular traits (e.g., DNA methylation, gene expression, protein abundance) and disease outcomes (11). SMR approaches that integrate QTL datasets—including methylation quantitative trait loci (mQTLs), expression quantitative trait loci (eQTLs), and protein quantitative trait loci (pQTLs)—with GWAS summary statistics can reveal molecular features causally associated with RA. While this strategy has been successfully applied to investigate autophagy and immune pathways in RA, its application to glycolysis remains limited. In this study, we employed a multi-layer SMR framework to investigate whether glycolysis-related molecular features are causally linked to RA risk. By integrating GWAS data from the FinnGen, UK Biobank, and GCST90129453 cohorts with multi-omics QTL datasets, we aimed to identify key methylation sites, transcripts, and proteins involved in glycolysis that may act as upstream drivers of RA pathogenesis. Here, we employ a multi-omics MR framework to systematically investigate causal roles of epigenetically regulated glycolytic genes in RA.

In this study, we conducted a comprehensive multi-omics MR analysis integrating GWAS with mQTLs, eQTLs, and pQTLs to identify glycolysis-related genes causally associated with RA. Our results revealed significant associations for 129 CpG sites, 28 transcripts, and 9 proteins with RA risk. Among these, 40 CpG sites, 11 transcripts, and 3 proteins showed robust colocalization. Notably, PKD1 and SLC2A4 demonstrated consistent multi-layer associations, supported by replication in independent RA cohorts. Additionally, integrative SMR between mQTLs and eQTLs identified potential epigenetic regulation of glycolysis-related targets, including PYGB, PFKFB2, and ALAS1. These findings support a causal chain from genetic variation through epigenetic regulation and transcriptional alterations to RA susceptibility, strengthening mechanistic links between glycolysis-related regulation and RA.

Notably, several loci contained multiple associated CpG sites within the same gene (e.g., PYGB), and in some cases CpGs within one gene showed opposing directions of association (e.g., JOSD1). Such convergent signals may reflect coordinated, biologically meaningful regional regulation (e.g., promoter/enhancer methylation changes acting in concert to influence transcription). Alternatively, multiple CpG ‘hits’ can arise from correlated methylation structure and linkage disequilibrium at a locus, where nearby probes share the same underlying genetic signal rather than representing independent functional effects. Accordingly, we interpret multi-CpG patterns as supportive of locus involvement but avoid assuming that each CpG has an independent causal role; targeted experimental dissection (e.g., site-specific epigenetic editing) will be needed to resolve CpG-level functionality.

PKD1 emerged as a prominent candidate gene potentially implicated in RA pathogenesis. PKD1 encodes polycystin-1, involved in key inflammatory signaling pathways such as VEGFR2 and NF-κB activation (21, 22). Recent evidence highlighted PKD1’s role in RA synovial inflammation through promoting fibroblast migration and vascular permeability (23). Consistent experimental findings demonstrated that PKD1 knockdown attenuated arthritis severity, further suggesting its pathogenic relevance (24). Interestingly, in cancer studies, PKD1 acts as a tumor suppressor, underscoring its complex context-dependent functions (25). Our SMR analysis revealed a significant causal association between the PKD1 gene and RA risk. Specifically, the methylation site cg07036112 within PKD1 exhibited significant negative regulatory effects in mQTL-eQTL analysis, with methylation levels inversely correlated with PKD1 expression. This negative regulation potentially leads to elevated PKD1 expression, thereby increasing RA susceptibility. Supporting these findings, whole-blood qPCR in an independent cohort demonstrated higher PKD1 mRNA expression in RA patients than in healthy controls, reinforcing its potential as a circulating biomarker.

SLC2A4, encoding the insulin-regulated glucose transporter GLUT4, is classically recognized for its role in metabolic tissues (26). Our MR analysis provided strong evidence linking genetic variants influencing hypomethylation at cg06891043 with increased SLC2A4 expression, consequently increasing RA risk. This finding supports a pathogenic role, where elevated SLC2A4 expression contributes to enhanced glycolytic flux may contribute to increased glucose uptake capacity in RA-relevant immune and stromal pathways, potentially exacerbating inflammation and invasiveness (27, 28). Furthermore, structural and docking studies in oncology contexts indicated that modulation of SLC2A4 might influence glycolytic adaptation and cell survival under inflammatory conditions (29), highlighting its relevance to RA pathology. Therapeutically, interventions targeting GLUT4 expression or its regulatory mechanisms may offer promising strategies to mitigate the metabolic dysfunction underlying RA. Consistent with our genetic and epigenetic results, our whole-blood qPCR validation confirmed increased SLC2A4 mRNA expression in RA patients compared with healthy controls, in line with the direction of effect inferred from the multi-layer MR and QTL analyses.

This whole-blood qPCR validation provides a conceptually aligned translational bridge to our blood-derived mQTL/eQTL evidence by confirming that PKD1 and SLC2A4 transcripts are increased in RA blood. Nevertheless, because this validation is cross-sectional and not genotype-stratified, it does not by itself establish that the observed expression differences are genetically mediated in the same individuals, nor does it resolve cell-type specificity. Future studies integrating paired genotyping with leukocyte subset–resolved transcriptomics and RA-relevant tissues (e.g., synovium) will be valuable to further connect the QTL layer to disease biology.

Other glycolysis-related genes, including PFKFB2, PYGB, and ALAS1, emerged with significant associations. Epigenetic regulation at the PFKFB2 locus suggests modulation of macrophage glycolysis and inflammation resolution mechanisms (30, 31). Similarly, PYGB expression, driven by genetic variation, might sustain cellular energy supply and inflammatory cell survival, paralleling roles previously described in cancer metabolism (32, 33). ALAS1, identified via mQTL-eQTL integration, potentially links mitochondrial heme biosynthesis with autophagic and AMPK signaling pathways essential for cellular survival under inflammatory stress (34). Together, these genes highlight glycolytic pathway complexity in RA, each offering mechanistic insights and therapeutic opportunities.

Functionally, the prioritized genes map onto distinct functional modules of glycolysis-related biology in RA. SLC2A4 (GLUT4) supports glucose uptake, PFKFB2 regulates a key rate-controlling step of glycolytic flux via fructose-2,6-bisphosphate, and PYGB enables glycogen mobilization to supply glycolytic substrates. Together, these findings are consistent with a disease-relevant shift toward increased glucose utilization observed in activated immune cells and pathogenic stromal populations in RA. Although not all highlighted targets are core glycolytic enzymes (e.g., PKD1 may influence metabolic programs via inflammatory signaling), the convergent genetic/QTL evidence supports dysregulated glucose metabolism as an integrated component of RA immunometabolism.

Our study systematically applied multi-omics MR integration, identifying molecular signatures across methylation, expression, and protein levels, surpassing single-layer analyses. Integration of GWAS with mQTL, eQTL, and pQTL datasets prioritized key glycolytic genes including PKD1, SLC2A4, PFKFB2, PYGB, and ALAS1, supported by robust SMR and colocalization analyses (11, 35). This comprehensive approach provided insights into methylation–expression and expression–protein regulatory cascades, exemplifying the methodological strength and robustness of the findings.

Despite the overall consistency of PKD1 and SLC2A4 signals, several other loci identified in FinnGen could not be replicated in the UKB or GCST90129453 RA cohorts. This lack of replication is likely multifactorial. Differences in sample source and collection (e.g. population-based volunteer cohort in UKB vs. health registry-based cohort in FinnGen) and cohort size/power (FinnGen’s >16k RA cases vs. UKB’s ~4.4k) may contribute to diminished signal detection in validation sets. It is also possible that varying phenotype definitions and subtle population genetic differences between cohorts affected the reproducibility of certain associations. We have thus interpreted non-validated loci with caution, recognizing that limited power or heterogeneity across cohorts could explain their absence of replication. This explicit consideration of cross-cohort differences highlights the robustness of the PKD1 and SLC2A4 findings while acknowledging the limitations and context for loci that did not replicate.

Several limitations should be considered. Firstly, our analyses primarily relied on European-derived QTL and GWAS datasets, potentially limiting generalizability to other populations. Secondly, although rigorous methods minimized pleiotropy, residual horizontal pleiotropy in summary-level data cannot be excluded. Thirdly, using blood-derived QTL data may inadequately represent synovial-specific regulatory events. Future tissue-specific and single-cell omics studies are needed. Lastly, our whole-blood qPCR validation was based on a relatively modest sample size, which may limit the precision of the effect estimates and warrants confirmation in larger, independent cohorts. Further validation through diverse populations and functional experiments remains necessary.

In conclusion, this integrative multi-omics MR analysis identifies PKD1, SLC2A4, and additional glycolysis-related genes as putative causal regulators of rheumatoid arthritis. By linking genetic variation to epigenetic and transcriptional changes associated with RA risk, these findings clarify mechanisms of metabolic reprogramming and nominate candidates for biomarker development and metabolism-targeted therapy, including potential repurposing of modulators of PKD1 or GLUT4. Looking forward, longitudinal cohorts, multimodal omics, and artificial intelligence/machine learning (AI/ML)–enabled analytics may refine causal inference, improve patient stratification, and accelerate precision medicine in RA.

We would like to thank all participants and investigators involved in the following studies: the FinnGen consortium for providing the RA GWAS summary statistics (Release R12, GWAS ID: M13_RHEUMA), the UK Biobank for the GWAS data used in the study (PheCode 714.1), and the GWAS Catalog for sharing the GCST90129453 dataset. We also thank the eQTLGen Consortium for providing the blood-derived expression QTL summary statistics, and the authors of the meta-analysis of European cohorts (Brisbane Systems Genetics Study and the Lothian Birth Cohorts) for the methylation QTL data.

The author(s) declared that financial support was received for this work and/or its publication. This work was funded by the National Natural Science Foundation of China (82474302), Changning District Science and Technology Commission (CNKW2024Y19), the Training Program for High-caliber Talents of Clinical Research at Affiliated Hospital of SHUTCM (2023LCRC24), Shanghai Municipal Health Commission (20244Y0010), Scientific Research Project of Changning District Science and Technology Commission (CNKW2022Y22) and Shanghai Science and Technology Commission Support Project (23Y11921900), and was also supported by the Changning National Master of Traditional Chinese Medicine Studio.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by Ethics Committee of Shanghai Guanghua Hospital of Integrative Medicine, Affiliated Hospital of Shanghai University of Traditional Chinese Medicine. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

XA: Methodology, Visualization, Conceptualization, Writing – original draft, Writing – review & editing, Formal Analysis. PX: Software, Data curation, Writing – original draft, Visualization, Formal Analysis. LZ: Writing – original draft, Resources, Investigation, Validation, Data curation. BX: Writing – original draft, Visualization, Software. JW: Writing – original draft, Formal Analysis, Visualization, Validation. SS: Methodology, Investigation, Writing – original draft. JX: Methodology, Investigation, Writing – original draft. CG: Methodology, Investigation, Writing – original draft. PP: Writing – original draft, Data curation, Investigation. GQ: Visualization, Validation, Formal Analysis, Writing – original draft. LJ: Writing – original draft, Data curation, Investigation. JS: Formal Analysis, Writing – original draft. XX: Writing – original draft, Formal Analysis. YC: Investigation, Writing – original draft, Data curation. SP: Writing – original draft, Formal Analysis. LR: Writing – original draft, Funding acquisition, Conceptualization, Project administration, Supervision. YB: Supervision, Writing – review & editing, Methodology, Funding acquisition. LX: Project administration, Supervision, Funding acquisition, Writing – review & editing.

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

A correction has been made to this article. Details can be found at: 10.3389/fimmu.2026.1793746↗.

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2025.1691663/full#supplementary-material↗