This study aimed to systematically elucidate the shared genetic architecture and causal relationship between AP and T2D. The analytical workflow comprised four sequential steps [Figure 1]: (1) estimating the genome-wide genetic correlation between AP and T2D using linkage disequilibrium score regression (LDSC); (2) identifying pleiotropic SNPs jointly associated with both traits using the pleiotropy analysis under the composite null hypothesis (PLACO); (3) annotating the pleiotropic SNPs to independent genomic loci and candidate genes using the functional mapping and annotation (FUMA) platform, followed by PPI network construction, hub gene identification, and systematic functional enrichment analyses (Gene Ontology, GO and Kyoto Encyclopedia of Genes and Genomes, KEGG); and (4) evaluating the direction of causal relationships using bidirectional two-sample MR analysis.

All analyses used publicly available summary-level data from GWAS conducted in individuals of European ancestry. Steps were taken to ensure no sample overlap between the AP and T2D datasets.

To assess the shared genetic basis between AP and T2D, we first performed LDSC. The analysis revealed a significant positive genetic correlation between AP and T2D [r_g = 0.415, standard error (SE) = 0.082, P = 4.73 × 10^-7]. This finding suggests a moderate yet statistically robust shared genetic architecture between AP and T2D.

We next applied PLACO to identify SNPs jointly associated with both AP and T2D. This analysis identified 73 SNPs that reached genome-wide significance for pleiotropy (P_PLACO < 5 × 10^-8) [Supplementary Table 1 and Supplementary Figure 1].

To delineate the shared genomic architecture, the 73 pleiotropic SNPs were annotated and clustered into independent loci using the FUMA platform. This process defined 12 independent genomic risk loci [Supplementary Table 2]. Within these loci, a total of 23 candidate genes were initially mapped. Of these, 22 candidate genes were prioritized based on the strength of the genetic association signal within their respective loci [Table 1]. TCF7L2 exhibited the most significant pleiotropic association. Other top-ranked genes included IGF2BP2, GCKR, and a gene cluster on chromosome 19 encompassing APOE, APOC1, and TOMM40.

To elucidate the functional relationships among the 22 candidate pleiotropic genes, a PPI network was constructed using the STRING database. The resulting network contained 22 nodes (genes) connected by 46 edges, representing significant functional associations [Figure 2]. The network exhibited a high average node degree (4.18) and local clustering coefficient (0.629), indicating a densely interconnected structure. Notably, the PPI enrichment (P < 1.0 × 10^-16) confirmed that the observed interactions are significantly greater than expected by chance, strongly supporting the biological coherence of this gene set.

Topological analysis of the network identified key hub genes central to the network’s overall connectivity. Based on the MCC algorithm, HHEX, IGF2BP2, CDKAL1, and TCF7L2 were ranked as the top hub genes [Supplementary Table 3]. Furthermore, a distinct, tightly interconnected module on chromosome 19, comprising APOE, APOC1, and TOMM40, was evident within the network, underscoring the shared role of lipid metabolism and mitochondrial function in AP and T2D. Additional genes such as GCKR and FTO also demonstrated high connectivity, serving as key connectors within the overall network architecture.

To link the genetic findings to underlying biological mechanisms, we performed GO and KEGG enrichment analyses on the 22 candidate pleiotropic genes [Supplementary Figure 2]. The findings revealed significant enrichment in lipid and metabolic pathways. KEGG analysis identified “Cholesterol metabolism” [hsa04979, adjusted P-value (P.adj) = 0.036] as the most significantly enriched pathway. Similarly, GO BP analysis highlighted crucial pathways including “phospholipid efflux” (GO:0033700), “neutral lipid metabolic process” (GO:0006638), “acylglycerol metabolic process” (GO:0006639), and the negative regulation of phosphate and phosphorus metabolic processes (all P.adj < 0.05). These enrichments were predominantly driven by APOE, APOC1, GCKR, and HHEX genes, corroborating the hypothesis that genetically driven lipid dysregulation is a core shared mechanism between T2D and AP.

To dissect the causal direction underlying the observed genetic correlation, we performed bidirectional two-sample MR analyses between AP and T2D.

Genetically predicted T2D was causally associated with increased risk of AP. Using 104 independent genetic variants as IVs (Supplementary Table 4, all F-statistics > 10), the primary IVW method revealed that genetic predisposition to T2D was associated with an 8.0% increase in the odds of AP [odds ratio (OR) = 1.08, 95% confidence interval (CI): 1.00-1.16, P = 0.040]. The direction of effect was consistent across complementary MR methods, with the weighted median approach yielding a significant estimate (OR = 1.14, 95%CI: 1.02-1.26, P = 0.018) [Figure 3 and Supplementary Figure 3A]. Sensitivity analyses supported the robustness of this causal inference. Although Cochran’s Q test indicated moderate heterogeneity (P = 0.017), there was no significant evidence of horizontal pleiotropy (MR-Egger intercept P = 0.477). The MR-PRESSO analysis found no significant outliers (P = 0.356), and leave-one-out analysis confirmed that the overall effect was not driven by any single influential SNP. Importantly, the Steiger directionality test empirically confirmed our causal assumption: the IVs for T2D explained significantly more variance in T2D (R² = 1.34%) than in AP (R² = 0.03%), yielding a highly significant correct causal direction (P_Steiger < 0.001).

In contrast, we found no statistically significant evidence for a reverse causal effect of AP on T2D. Using 4 independent genetic instruments for AP (Supplementary Table 5, all F-statistics > 10), the IVW estimate showed no statistically significant causal effect (OR = 0.97, 95%CI: 0.91-1.04, P = 0.418). All sensitivity analyses yielded consistently non-significant results, with no evidence of heterogeneity (Cochran’s Q P = 0.460) or directional pleiotropy (MR-Egger intercept P = 0.724) [Figure 3 and Supplementary Figure 3B].

Collectively, these MR results revealed an asymmetric causal relationship: genetic predisposition to T2D confers a causal risk for developing AP, whereas genetic liability to AP does not appear to causally influence T2D risk.

The robust genome-wide genetic correlation between AP and T2D (r_g = 0.415, SE = 0.082, P = 4.73 × 10^-7), corroborated by the identification of 12 independent pleiotropic risk loci and 22 candidate genes [Table 1], confirms that a substantial portion of the observed clinical co-occurrence is rooted in shared genetic factors, beyond common environmental exposures. The pleiotropic loci we identified converge on biological pathways central to both diseases. Notably, TCF7L2, the top-ranked gene, reaffirms the pivotal role of Wnt signaling and pancreatic beta-cell dysfunction^[21]. The PPI network analysis further transformed this list of candidate genes from a static catalog into a dynamic functional map. The identification of HHEX, IGF2BP2, CDKAL1, and TCF7L2 as network hubs suggests these genes act as critical regulators within the shared genetic program, potentially orchestrating downstream effects on insulin secretion, cell proliferation, and inflammatory responses.

A particularly notable finding is that the APOE/APOC1/TOMM40 gene cluster on chromosome 19 forms a tightly interconnected module within the pleiotropic PPI network [Figure 2]. This module is not merely a bystander but represents a functional module with direct relevance to the pathophysiology of both conditions. APOE is a master regulator of lipoprotein metabolism, with alleles influencing circulating triglyceride levels^[22], a well-established, direct trigger for AP. Simultaneously, APOE expression in macrophages regulates inflammation and tissue repair via extracellular vesicles^[23]. Mitochondrial dysfunction is implicated in both insulin resistance in T2D^[24] and in acinar cell injury and necrosis during AP^[25]. TOMM40, encoding an essential mitochondrial import subunit, sits at the crossroads of these pathways, as it is crucial not only for organellar function but also for modulating hepatic lipid and plasma lipoprotein levels^[26]. The physical linkage and strong co-expression of these genes suggest their coordinated action may influence pancreatic health by modulating two intersecting vulnerabilities: lipid homeostasis and cellular energetics or stress resilience. This provides a plausible genetic mechanism whereby variants in this cluster simultaneously elevate risk for hypertriglyceridemia-driven AP and for T2D associated with metabolic stress.

The bidirectional MR analysis yielded a clear and asymmetric causal inference [Figure 3]: genetic predisposition to T2D significantly increased AP risk, whereas genetic liability to AP did not significantly alter T2D risk. This asymmetry provides genetic evidence to reframe the clinical AP-DM relationship. We propose a two-hit model in which T2D-associated genetic variants establish a baseline state of metabolic vulnerability. This vulnerability compromises the exocrine ability of the pancreas to withstand secondary insults, thereby lowering the threshold for developing pancreatitis. Specific variants in genes such as GCKR (influencing hepatic lipid and glucose metabolism)^[27] and FTO (linked to adiposity)^[28] exemplify genetic factors that may predispose individuals to this vulnerable state. The lack of a reverse causal effect suggests that the development of DM following AP is not driven by the common genetic variants that predispose to AP itself. This supports the understanding that diabetes of the exocrine pancreas (DEP) is largely a sequela of anatomical and inflammatory pancreatic damage, distinguishing it etiologically from classic T2D. This distinction is important for accurate disease classification and management.

Our findings have direct translational implications. First, they reinforce that intensive management of shared modifiable risk factors is paramount in individuals with or at high genetic risk for T2D, as it may prevent both T2D progression and AP episodes. Second, the identified pleiotropic genes and pathways reveal potential targets for novel therapeutic strategies aimed at enhancing metabolic resilience and reducing pancreatic inflammation. Importantly, our findings lay the groundwork for implementing polygenic risk scores (PRS) in clinical risk stratification. A PRS derived from the T2D risk-increasing alleles identified here could be used to identify a subset of diabetic patients, or even high-risk prediabetic individuals, who carry an exceptionally high genetic burden associated with AP risk. This genetically high-risk cohort could benefit from targeted surveillance (e.g., periodic pancreatic enzyme or triglyceride checks), personalized lifestyle interventions, and earlier pharmacologic management of lipids, moving toward a paradigm of genetically informed prevention of pancreatitis.

Several limitations of our study should be acknowledged to ensure a comprehensive interpretation of the findings. First, our analyses were restricted to GWAS summary statistics from European ancestry populations. While this restriction minimized population stratification bias, it inherently limits the generalizability of our findings to non-European ancestry groups. Given the known differences in lipid metabolism and diabetes prevalence across global populations, future multi-ancestry GWAS and cross-ethnic meta-analyses are warranted. Second, although we conducted extensive MR sensitivity analyses, there was a notable disparity in the number of IVs used (104 for T2D vs. only 4 for AP). This limited number of instruments for AP significantly constrained the statistical power of the reverse MR analysis. Therefore, while our data suggest a lack of reverse causality (AP leading to T2D), this null finding should be interpreted with caution until larger, better-powered AP GWAS datasets become available. Third, our study identified associations and provided causal estimates at the population level, but it did not include experimental validation. The precise biological mechanisms and cooperative regulatory networks through which these specific variants (e.g., the APOE/APOC1/TOMM40 cluster) exert their pleiotropic effects cannot be fully resolved through in silico analyses alone. Functional follow-up studies, particularly those utilizing Clustered Regularly Interspaced Short Palindromic Repeats-edited cellular models (CRISPR-edited cellular models), single-cell multi-omics of the human pancreas, and cell-type-specific investigations, are highly warranted to link these genetic findings to underlying biological mechanisms. Finally, while we employed robust methods to minimize confounding variables, residual horizontal pleiotropy can never be entirely ruled out in MR frameworks. Future prospective cohort studies incorporating PRS and detailed lipidomics are needed to fully translate our genetic findings into improved patient outcomes.

Our study provides genetic evidence for a shared etiological basis between AP and T2D, pinpointing pleiotropic loci, notably the APOE/APOC1/TOMM40 cluster, and highlighting core lipid and cholesterol metabolism pathways identified via functional enrichment that link metabolic dysregulation to both conditions. A unidirectional causal pathway from T2D to AP was identified, providing genetic evidence to distinguish classic T2D from diabetes secondary to pancreatic disease. These findings highlight the potential of genetic risk profiling to identify high-risk individuals who may benefit from targeted monitoring and early triglyceride management, supporting a proactive approach to pancreatic health in diabetes care.