The association between GLP-1R expression and cardiovascular-kidney-metabolic-related diseases in non-diabetic and non-obese population: evidence triangulation using Mendelian randomization, observational and polygenic score association analysis Authors: Zhihe Chen, Qian Yang, Jiajia Liu, Huiling Zhao, Hui Ying, Haoyu Liu, Lanhui Huang, Hong Lin, Shuangyuan Wang, Mian Li, Yangbo Sun, Tiange Wang, Zhiyun Zhao, Min Xu, Yuhong Chen, Yu Xu, Jieli Lu, Guang Ning, Yanyun Gu, Weiqing Wang, Jing Li, Yufang Bi, Jie Zheng Published: 2026 ABSTRACT BackgroundGlucagon-like peptide-1 receptor (GLP-1R) agonists are emerging as promising therapies for cardiovascular-kidney-metabolic (CKM) related diseases in individuals with type 2 diabetes mellitus (T2DM) or obesity. But their effects in non-obese and non-diabetic individuals are unclear. This study triangulates evidence using Mendelian randomization (MR), polygenic scores (PGS) and observational analyses to estimate the associations of GLP-1R expression with chronic kidney disease (CKD), heart failure (HF) and metabolic dysfunction-associated steatotic liver disease (MASLD).MethodsFor the MR analysis, instruments mimicking GLP-1R expression were identified using pancreas-specific cis-expression quantitative trait loci from GTEx (N ≤ 305). MR-Robust method was used as the primary MR approach. PGS and observational analyses were performed both in non-diabetic and non-obese individuals separately. A genome-wide association study (GWAS) for MASLD (14,231 cases and 348,091 controls) was performed in the general population using data from UK Biobank.ResultsGLP-1R expression showed robust effects on CKD (odds ratio [OR] 0.96, 95%CI 0.95 to 0.97, q = 1.7 × 10− 10 ), HF (OR = 0.96, 95%CI 0.94 to 0.97, q = 2.5 × 10− 8) and MASLD (OR = 0.96, 95%CI 0.93 to 0.98, q = 1.3 × 10− 3) in the general population. Consistent results were observed in validation analyses. Furthermore, PGS and observational analyses among non-T2DM and non-obese individuals found little evidence to support its association with CKD, HF or MASLD. GWAS analysis identified eight conditionally independent variants associated with MASLD, in which rs563199662 was a new signal located at TFPI region.ConclusionsThis study provides multilayered evidence for GLP-1R expression in mitigating CKD, HF and MASLD risks in the general population, while de-prioritized its effect on CKM-related diseases in non-obese and non-diabetic individuals. Further clinical trials are needed to validate the effects of GLP-1R agonists in relative health population.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12916-026-04888-8. Background Cardiovascular-kidney-metabolic (CKM) syndrome, or cardiometabolic renal syndrome, is a newly proposed concept that describes the systemic pathological condition involving the complex interaction of metabolic risk factors, chronic kidney disease (CKD), and cardiovascular diseases (CVD) [1]. Of note, metabolic dysfunction-associated steatotic liver disease (MASLD), a state of metabolic dysfunction, is closely related to CKM and its progression can lead to CVD and CKD complications [2, 3], but not well studied using genetic tools yet. High prevalence of CKM in adults is associated with multi-organ dysfunction, increased adverse cardiovascular outcomes, and higher morbidity and mortality [4]. Of this, drug intervention supporting the simultaneous management of CKM syndrome is an attractive therapeutic approach. Obesity and type 2 diabetes mellitus (T2DM) are two key metabolic risk factors for CKM and MASLD [1, 5]. Thus, therapeutic approaches targeting metabolic risk factors, e.g. glucose levels, may show protective effect on CKM-related diseases [6]. Glucagon-like peptide-1 receptor (GLP-1R) agonists, e.g. dulaglutide, albiglutide, and semaglutide, are a class of anti-diabetic drugs exerting their glucose-lowering effects primarily through the pancreas by enhancing the glucose-dependent insulin excretion and suppressing glucagon release, while also contributing to delay of gastric emptying and promoting satiety [7]. The U.S. Food and Drug Admistration and the European Medicines Agency have approved the use of GLP-1R agonists in treating diabetes, overweight and obesity [8], indicating its great promise in managing CKM. Some studies have shown that GLP-1R agonists may protect against cardiovascular diseases, mortality and metabolic dysfunction–associated steatohepatitis in diabetic and obese populations [9, 10]. Recently, SELECT and ESSENCE randomized controlled trial (RCT) indicated that semaglutide improved cardiovascular endpoints in patients with heart failure (HF) and impreoved liver histologic results in patients with metabolic dysfunction–associated steatohepatitis and moderate or advanced liver fibrosis, respectively [11, 12]. However, these clinical trials primarily involved disease-specific subgroups. It is unclear whether GLP-1R agonism protects CKM-related diseases in the general population, especially in those without T2DM and/or obesity. Mendelian randomization (MR) uses germline genetic variants as instruments to estimate causal effects between an exposure and an outcome, which is not generally susceptible to the reverse causation or residual confounding inherent in conventional observational studies [13, 14]. Recent MR studies have investigated GLP-1R expression on cancers and HF, supporting the repurposing of GLP-1R agonists to prevent these diseases [15, 16]. We refined instrument selection to accurately proxy the effects of GLP-1R expression in its functional tissue, pancreas, by estimating its effect on T2DM and obesity simultaneously. Multiple state-of-the-art genetic and observational epidemiology approaches, e.g. polygenic score (PGS) association and observational association analyses, were applied to triangulate evidence and get reliable associations of GLP-1R activities with CKM-related diseases [17]. In the current study, we conducted a genome-wide association study (GWAS) of MASLD using the newly proposed clinical definition [5], as well as collected GWAS summary statistics of other CKM-related diseases, including CKD- and CVD-related outcomes. We then selected instruments that proxy GLP-1R expression using tissue-specific expression quantitative trait loci (eQTLs) and systematically investigated the causal effect of genetically proxied GLP-1R expression on CKM-related diseases in the general population using MR. To triangulate evidence, we performed PGS association analysis and observational analysis to explore the potential associations in non-diabetic and non-obese individuals using data from UK Biobank, with the aim of extending the usage of GLP-1R agonists to those healthy individuals without diabetes and obesity. Methods Study design The design of this study was as follows. (Fig. 1) First, for the MR analyses, we selected instruments proxying GLP-1R expression of GLP-1R using pancreas specific eQTLs from Genotype-Tissue Expression (GTEx) v8 [18, 19]. As a positive control analysis, we validated our genetic variants by estimating its associations with T2DM and body mass index (BMI). Then the casual estimates of GLP-1R expression on CKD, HF, CAD and MASLD outcomes were calculated using MR-Robust, while other five MR methods had been applied for validation. Second, to assess whether the causal effects we observed are independent of metabolic benefits, a multivariable MR (MVMR) analysis adjusting glycated haemoglobin (HbA1c) and BMI was performed using PC-GMM method. Third, PGS associations analysis of GLP1R gene expression in pancreas and observational analysis of protein levels of GLP-1R against CKM-related diseases were performed in general, non-diabetic and non-obese individuals using data from UK Biobank. These were used to explore whether the association we observed in the general population can also be observed in the relatively healthy individuals, which may inform the extension of scope of GLP-1R agonism on CKM prevention. Fourth, to expand the genetic landscape of MASLD, a new GWAS was conducted for MASLD using updated definition. Study design and overview. This study included three major sections. First, instruments proxying GLP-1R expression were selected using pancreas-specific eQTLs from Genotype-Tissue Expression v8. As a positive control analysis, we validated our genetic variants by estimating its associations with T2DM and BMI. Then the casual estimates of GLP-1R expressionon CKD, CVD and MASLD outcomes were calculated. MR-Robust was applied as the primary method in the MR analysis with other five methods (MR-RAPS, dIVW, PC-GMM, weighted median and IVW) as validation methods. Second, to determine whether the causal effects are independent of metabolic benefit, MVMR adjusting for HbA1c and BMI was performed. PGS analyses were performed to test the association of polygenic score of GLP-1R on CKM outcomes in non-diabetic and non-obese populations of UK Biobank. Observational analysis of plasma protein levels of GLP1R on CKM outcomes in non-diabetic and non-obese individuals were conducted in UK Biobank to validate the PGS analysis findings. Lastly, a new GWAS was conducted for MASLD using upadated definition to expand the genetic landscape of MASLD. Abbreviations: GLP-1R, Glucagon-like peptide-1 receptor; CKD, chronic kidney disease; CVD, cardiovascular disease; HF, heart failure; CAD, coronary artery disease; MASLD, metabolic dysfunction-associated steatotic liver disease; T2DM, type 2 diabetes mellitus; HbA1c, glycated haemoglobin A1c; BMI, body mass index; MR, Mendelian randomization; MR-RAPS, MR-Robust Adjusted Profile Score; dIVW, debiased inverse-variance weighted; PC-GMM, Principal Component Generalized Method of Moments; IVW, inverse-variance weighted; MVMR, multivariable Mendelian randomization; PGS, polygenic score UK Biobank and Proteomics data in the UK Biobank Access to the UK Biobank was granted under application number 104,086. Detailed information on UK Biobank and UK Biobank Pharma Proteomics Project is summarized in the Additional file 1: Supplementary Method [20, 21]. Mendelian randomization analysis Selection for genetic predictors of GLP-1R expression We used pancreas-specific cis-eQTLs for the targeting gene to proxy GLP-1R expression, since pancreas is the functional organ for GLP-1R agonists [22]. Summary-level data for cis-eQTLs were obtained from GTEx v8. Only one variant being retrieved under strict genome-wide significance (P < 5 × 10− 8). To ensure sufficient instruments for MR sensitivity methods, we relaxed the threshold, selecting variants showing marginal associations (P < 1 × 10− 3) with GLP-1R expression. Then we clumped variants at a linkage disequilibrium (LD) r2 < 0.01 to remove variants with pair-wise correlation between each other. After above process, three genetic variants were selected as instruments for GLP-1R expression, and their F-statistics were calculated to assess instrumental strength. To comprehensively evaluate potential tissue-specific effects and address the widespread expression of GLP-1R, we systematically expanded our selection of genetic instruments beyond the pancreas. We extracted cis-eQTLs from eight additional biologically relevant tissues, including heart left ventricle, heart atrial appendage, artery coronary, kidney cortex, brain hypothalamus, pituitary, small intestine terminal ileum, and stomach. For the univariable and multivariable Principal Component Generalized Method of Moments (PC-GMM) analysis, we adopted a different instrument selection strategy to fully leverage its methodological advantage in handling correlated variants and capturing maximal genetic variance. Specifically, we selected variants with a threshold of P < 0.05 and a LD threshold of r2 < 0.95, allowing PC-GMM to construct principal components from a dense set of genetic markers, thereby improving estimation efficiency. In the debiased Inverse Variance Weighted (dIVW) analysis, the Condition ( (average F-statistic − 1) * sqrt(# SNPs)) statistics exceeded 50, which met the dIVW method requirements for estimator reliability [23]. Study outcomes To provide a comprehensive evaluation, we selected four categories of CKM-related outcomes for this study, including metabolic diseases, CKD outcomes, CVD outcomes and MASLD. For metabolic diseases, T2DM, HbA1c, BMI and body weight were included as positive control outcomes. CKD-related outcomes comprised CKD, estimated glomerular filtration rate (eGFR), blood urea nitrogen (BUN), microalbuminuria (MA) and urinary albumin-to-creatinine ratio (UACR). HF, HF subtypes (non-ischemic HFpEF and non-ischemic HFrEF) and coronary artery disease (CAD) was chosen as CVD outcomes. Except MASLD, GWAS summary statistics were obtained from public studies. Collectively, seven diseases and seven relative quantitative biomarkers were selected. Descriptive information of the study phenotypes in this study was provided in Additional file 1: Supplementary Method [24–42]. Univariable mendelian randomization analysis of GLP-1R expression on CKM-related outcomes in the general population We conducted a two-sample MR to assess the causal relationships between GLP-1R expression and CKM-related outcomes. Since we used a P value cut-off of 1 × 10− 3 for instrument selection, weak instrument bias may be introduced [43]. To address this issue, we designated MR-Robust [44] as the primary MR method due to its resilience against outliers and invalid instruments. To further validate the robustness of our findings, particularly against weak instrument bias, we employed three advanced approaches including dIVW [23], MR-Robust Adjusted Profile Score (MR-RAPS) [45], and univariable PC-GMM [46]. These methods are specifically validated to provide reliable estimates even in the presence of weak instruments, pleiotropy or extreme outliers. Additionally, standard IVW and Weighted Median analyses were conducted to provide a comprehensive assessment of the causal estimates. Multivariable mendelian randomization analysis We performed multivariable PC-GMM to estimate the direct effects GLP-1R expression on CKM-related diseases adjusting for HbA1c and BMI to determine whether the causal effects we observed are independent of metabolic benefit. We reported the effects on binary outcomes as odds ratio (OR) and 95% confidence interval (CI) of disease per 1-standard deviation (SD) increase in gene expression levels of GLP-1R in pancreas. The effects on continuous outcomes were reported as SD unit change of the continuous outcome per 1-SD increase in gene expression levels of GLP-1R in pancreas. To account for multiple comparisons, the Benjamini-Hochberg False Discovery Rate (FDR) was calculated based on MR-Robust results across all outcomes. FDR-adjusted significance levels are reported as q-values throughout the manuscript, whereas unadjusted p-values are presented as p-values. Evaluation of MR assumptions MR have three core assumptions: (1) the genetic variants are robustly associated with the exposure; (2) there is no common causes (confounders) between the instrument and outcome; (3) the genetic variants are only associated with the outcome through the exposure under study [47]. To assess the potential violation of the first assumption, F-statistics were used to confirm the strength of genetic variants. Meanwhile, four weak instrument robust methods were applied to minimize possible bias. Cochran’s Q test utilizing IVW methods was applied to detect heterogeneity. Heterogeneity was considered using the p-value threshold < 0.05. To evaluate the potential horizontal pleiotropy, the MR-Egger intercept test was employed as a sensitivity analysis. A p for intercept test < 0.05 was considered indicative of pleiotropy. Steiger filtering was conducted to test the directionality of effects and minimize the possibility of reverse causality [48]. To rigorously assess the exclusion restriction assumption of MR, we performed a PheWAS for the selected instruments using the IEU OpenGWAS database [49] and the AstraZeneca PheWAS Portal [50]. For any targets identified, we conducted a MR analysis using cis-eQTLs extracted from eQTLGen to evaluate their causal effects on CKM-related diseases. Our study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization (STROBE-MR) guidelines [51, 52].(Additional file 2: STROBE-MR checklist) The MR statistical analyses were conducted using TwoSampleMR (v.0.5.8), MendelianRandomization (v0.9.0) and ieugwasr (v.1.0.0) packages in R software (v.4.3.1). Polygenic score association of GLP-1R expression with CKM-related diseases Polygenic Scores for GLP-1R To develop the PGS of GLP-1R in UK Biobank participants, we utilized three SNPs identified as instrumental variables from MR analysis as weight. The PGS was constructed by summing the number of risk alleles for these three SNPs, each weighted according to its effect size derived from GTEX v8. This score represents the aggregate genetic influence on the expression of GLP1R in pancreas, and was subsequently used for association analysis with the outcome of interest. Definition of CKM-related diseases included in the PGS association model CKD and HF was identified by UK Biobank Category 1712, which contains information of the first occurrence of certain diseases. Definition of MASLD was detailed in the GWAS analysis section. Participants with improbable event date (e.g. before or matching date of birth, in the future or no event date available) were excluded. Polygenic score association analysis We have restricted our analyses to participants of European ancestry, excluding individuals of any other ancestries or with sex mismatch. Analyses were conducted in the general population and subsequently in specific subgroups. To further explore the associations in non-diabetic and non-obese individuals, we performed subgroup analyses by excluding participants with T2DM (identified using UK Biobank Category 1712 first occurrence data) and obesity (BMI ≥ 30 kg/m2) before defining any specific disease (CKD, HF, CAD, or MASLD) or selecting controls, respectively. For the association analyses, participants were stratified into three tertiles based on their PGS distribution, ranging from the lowest to the highest scores. For each specific outcome analysis, we excluded prevalent cases at baseline and applied a six-month lag period, censoring events that occurred within the first six months of follow-up. PGS association analyses were performed using Cox Proportional Hazards (PH) models, with time to event (from the time of patient enrollment to the occurrence of the event) as timescale, the PGS tertiles as a predictor and the outcome of interest (CKD, HF, CAD and MASLD) as the dependent variable. We utilized two models: one corrected for array and the first 10 principal components, and the other further adjusted for age and sex. Hazard ratios (HRs) with 95% CIs were reported for each model. Observational analysis of plasma protein levels of GLP-1R on CKM outcomes in non-diabetic and non-obese individuals in UK Biobank Plasma protein levels of GLP-1R In UKB-PPP, plasma protein measurements were reported as normalized protein expression values after being transformed using a log-base-2 scale. Missing values in the proteomics data were imputed using the K-Nearest Neighbors algorithm. Subsequently, the protein levels underwent rank-based inverse normal transformation to mitigate skewness and ensure a normal distribution. Observational analysis In the observational analysis, we restricted the sample to participants of European ancestry and excluded those involved in the COVID-19 sub-study to ensure data consistency. The observational analyses were also conducted in the general population and subsequently in the non-diabetic and non-obese subgroups. Participants with T2DM and obesity (BMI ≥ 30 kg/m2) were excluded before defining any specific disease (CKD, HF, CAD, or MASLD) or selecting controls, respectively. For the association analysis, participants were stratified into three tertiles based on the ascending order of their plasma protein levels. For each outcome, we excluded participants with a history of the respective disease at/before baseline and those who developed the outcome within the first six months of follow-up. Time to event was calculated from enrollment to the date of incident diagnosis, death, or the end of follow-up (censoring date). Cox proportional hazards regression models were employed to estimate HRs and CIs for the middle and highest tertiles, using the lowest tertile as the reference. We constructed a minimally-adjusted model (controlling for age and sex) and a lifestyle-adjusted model (additionally controlling for three lifestyle covariates). Lifestyle covariates included alcohol intake frequency (0 = never, 1 = special occasions only, 2 = one to three times a month, 3 = once or twice a week, 4 = three to four times a week, 5 = daily or almost daily), the Townsend index of deprivation (higher score representing greater levels of deprivation) and smoking status (0 = never, 1 = previous, 2 = current). Outcomes were defined using the same approach as the PGS association analysis. Participants with pre-baseline onset of the event of interest were excluded. Cox PH models were run using the “survival” (v3.5-8) R package. Genome-wide association analysis of Metabolic Dysfunction–Associated Steatotic Liver Disease Phenotype definition Cases of MASLD were ascertained based on a composite definition aligning with the latest international consensus [5], which necessitates the coexistence of hepatic steatosis and metabolic dysfunction. Briefly, participants with either electronic health records (ICD-10: K76.0, K75.8, K74.6 and K76.9; ICD-9: 571.8, 571.9 and 571.5) or a fatty liver index [53, 54] (FLI) ≥ 60 and with any of the five cardiometabolic risk factors, were defined as MASLD cases. Participants were excluded if they had records of excessive alcohol consumption, viral hepatitis, or other specific liver diseases (e.g., autoimmune hepatitis, Wilson’s disease) [55]. Genome-wide association analysis A total of 362,322 general European individuals (14,231 cases and 348,091 controls) were included in this analysis. GWAS was performed using REGENIE v4.1, which could handle unbalanced case-control ratio, population structure and relatedness [56]. Sex, age, the first 10 principal components and genotyping batch were included as covariates. The conditional analysis of the top association signals was conducted using GCTA-COJO [57, 58]. SNP-based heritability and and the genetic correlation between MASLD and non-alcoholic fatty liver disease (NAFLD) [55] was estimated using LD score regression (LDSC) 1.0.1 [59]. To differentiate polygenicity from confounding by population stratification, we calculated genomic inflation coefficient (λGC) and estimated the LDSC intercept [60]. SNPs identified by GWAS were mapped and annotated to their gene using the FUMA [61] with the default settings. GWAS Catalogue associations for the top loci were performed in FUMA. More detailed description of GWAS methodology are provided in Additional file 1: Supplementary Method [62, 63]. Results Reliability of instruments to proxy GLP-1R expression Three genetic variants proxying GLP-1R expression were listed in Additional file 3: Table S1. Total F-statistics of the three instruments was 20, which indicates an acceptable instrument strength. In the positive control analysis, we found that higher expression of GLP1R gene in pancreas decreased the risk of T2DM (OR 0.96, 95%CI 0.96 to 0.97, q = 6.3 × 10− 22). (Additional file 3: Table S2) Meanwhile, results indicated the protective effects of GLP-1R expression on reduced HbA1c levels (β -0.007, 95%CI -0.010 to -0.004, q = 4.2 × 10− 7), BMI (β -0.014, 95%CI -0.017 to -0.011, q = 4.5 × 10− 23) and body weight (β -0.015, 95%CI -0.019 to -0.011, q = 1.7 × 10− 14). (Additional file 3: Table S2). These results further validated the reliability of our instruments in proxy the effect of GLP-1R expression. The PheWAS analysis identified that two of the GLP-1R instruments were also associated with the expression levels of KCNK17 and SAYSD1 (p < 5 × 10− 8). MR analysis found no evidence to support the effects of KNCK17/SAYSD1 expressions on CKD, HF, CAD or MASLD.( Additional file 3: Table S3) In addition, queries in the AstraZeneca PheWAS Portal to revealed no significant associations between these two genes and the study outcomes or established risk factors, mitigating concerns that these secondary pathways introduce bias to our primary findings. Effects of pancreatic GLP-1R expression on CKM-related outcomes Genetically proxied GLP-1R expression showed a protective effect on CKD (OR 0.96, 95%CI 0.95 to 0.97, q = 1.7 × 10− 10). (Fig. 2a) Other five MR models also validated this effect. Little evidence was found to support the existence of heterogeneity. (Additional file 3: Table S4) The MR results also supported the protective effect of GLP-1R expression on eGFR (β 0.001, 95%CI 0.000 to 0.002, q = 7.1 × 10− 3), BUN (β -0.007, 95%CI -0.012 to -0.003, q = 9.0 × 10− 4) and UACR (β -0.002, 95%CI -0.004 to -0.001, q = 1.1 × 10− 4). (Fig. 2b and Additional file 3: Table S4) The direction of the causal estimates was entirely consistent across all three outcomes. Fig. 2Casual effect of pancreas-specific GLP-1R expression on CKM-related outcomes. (a) Forest plot (left) and heatmap (right) showing the causal estimates for binary disease outcomes, expressed as Odds Ratios (OR) with 95% confidence intervals (CI). (b) Forest plot (left) and heatmap (right) showing the causal estimates for continuous outcomes, expressed as Beta coefficients with 95% CI. In the forest plots, the MR-Robust method results are presented; dots represent point estimates and error bars represent 95% CIs. Red dots indicate a positive association, while blue dots indicate a negative association. The heatmaps display the results across other five analysis methods (MR-RAPS, dIVW, PC-GMM, Weighted median, and IVW). The color reflects the direction of the OR or Beta (red for positive, blue for negative). An asterisk indicates p < 0.05. Abbreviations: CKD, chronic kidney disease; HF, heart failure; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; CAD, coronary artery disease; MASLD, metabolic dysfunction-associated steatotic liver disease; eGFR, estimated glomerular filtration rate; BUN, blood urea nitrogen; UACR, urine albumin-to-creatinine ratio; MA, microalbuminuria; MR, Mendelian randomization; MR-RAPS, MR-Robust Adjusted Profile Score; dIVW debiased inverse-variance weighted; PC-GMM, Principal Component Generalized Method of Moments; IVW, inverse-variance weighted Casual effect of pancreas-specific GLP-1R expression on CKM-related outcomes. (a) Forest plot (left) and heatmap (right) showing the causal estimates for binary disease outcomes, expressed as Odds Ratios (OR) with 95% confidence intervals (CI). (b) Forest plot (left) and heatmap (right) showing the causal estimates for continuous outcomes, expressed as Beta coefficients with 95% CI. In the forest plots, the MR-Robust method results are presented; dots represent point estimates and error bars represent 95% CIs. Red dots indicate a positive association, while blue dots indicate a negative association. The heatmaps display the results across other five analysis methods (MR-RAPS, dIVW, PC-GMM, Weighted median, and IVW). The color reflects the direction of the OR or Beta (red for positive, blue for negative). An asterisk indicates p < 0.05. Abbreviations: CKD, chronic kidney disease; HF, heart failure; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; CAD, coronary artery disease; MASLD, metabolic dysfunction-associated steatotic liver disease; eGFR, estimated glomerular filtration rate; BUN, blood urea nitrogen; UACR, urine albumin-to-creatinine ratio; MA, microalbuminuria; MR, Mendelian randomization; MR-RAPS, MR-Robust Adjusted Profile Score; dIVW debiased inverse-variance weighted; PC-GMM, Principal Component Generalized Method of Moments; IVW, inverse-variance weighted For, CVD outcomes, genetically proxied GLP-1R expression was associated with lower risk of HF (OR 0.96, 95%CI 0.94 to 0.97, q = 2.5 × 10− 8). (Fig. 2a) As for the subtypes of HF, we found no evidence to support the effect of GLP-1R expression on non-ischemic HFpEF or non-ischemic HFrEF. (Fig. 2a and Additional file 3: Table S5) In addition, evidence was found to support the association of GLP-1R expression with increasd risk of CAD (OR 1.01, 95%CI 1.00 to 1.02, q = 4.7 × 10− 2). (Additional file 3: Table S5) Using results of our GWAS analysis for MASLD, we observed an effect of GLP-1R expression on reducing MASLD risk (OR 0.96, 95%CI 0.93 to 0.98, q = 1.3 × 10− 3). (Fig. 2a and Additional file 3: Table S6) No significant evidence of heterogeneity was detected in sensitivity analyses. Consistently, results were replicated in validation datasets, showing similar effect directions and magnitudes.(Additional file 2: Table S7) No pleiotropic effects was detected in the intercept test. (Additional file 3: Table S8) In addition, Steiger filtering analysis showed that genetic instruments explained more variance in the exposure than in the outcome, indicating the appropriate direction of causality. (Additional file 3: Table S9) Tissue-specific Mendelian randomization analysis Genetic variants proxying GLP-1R expression in other tissues were listed in Additional file 3: Table S10. In contrast to the significant associations observed using pancreatic cis-eQTLs, MR analyses using instruments derived from the eight extra-pancreatic tissues yielded no statistically significant associations with the studied outcomes (Additional file 3: Table S11). Multivariable Mendelian randomization analysis To evaluate whether the protective effect of genetically proxied GLP-1R expression on CKM-related diseases is independent of metabolic risk factors, we employed the PC-GMM multivariable Mendelian randomization approach. Notably, after adjusting for HbA1c and BMI, the association was substantially attenuated. The protective effect observed in the univariable analysis was no longer statistically significant for CKD (OR 0.96, 95% CI 0.92 to 1.00), HF (OR 0.99, 95% CI 0.95 to 1.03) and MASLD (OR 0.98, 95% CI 0.92 to 1.05).( Additional file 3: Table S12) This results suggesting that the influence of pancreas-specific GLP-1R expression on CKM-related diseases may be largely driven by or mediated through glycemic and weight control. Association of polygenic score of GLP-1R with CKM-related diseases In this analysis, the study included 442,240 participants in the general population analysis, with 402,639 non-diabetic and 334,024 non-obese individuals retained for the subgroup assessments. Detailed baseline demographics and clinical characteristics are shown in Additional file 3: Table S13. In the general population, individuals in the highest PGS tertile showed a significantly decreased risk of CKD (HR 0.97, 95%CI 0.93 to 0.99, p = 3.8 × 10− 2) and HF (HR 0.96, 95%CI 0.93 to 1.00, p = 5.5 × 10− 2) compared to those in the reference tertile. Notably, these genetic associations with CKD and HF were attenuated in two subgroup analyses. No evidence of association with CKD, HF, CAD, or MASLD among non-diabetic and non-obese participants. (Fig. 3a and Additional file 3: Table S14). Fig. 3Associations of polygenic score of GLP-1R and plasma protein levels of GLP-1R with CKM-related diseases.(a) Associations of polygenic risk scores of GLP-1R with CKD, HF, CAD and MASLD in UK Biobank Europeans. (b) Associations of GLP-1R protein levels with CKD, HF, CAD and MASLD in UKB-PPP Europeans. Hazard ratios (HRs) were calculated across tertiles (T1–T3) of the respective PGS and protein levels within the general population, as well as in non-diabetic and non-obese subgroups. The color of the circles represents the HR, with blue indicating a reduced risk (HR < 1.0) and red indicating an increased risk (HR > 1.0) relative to the reference group (T1). Fully adjusted models were presented for both analyses. Abbreviations: UKB-PPP, UK Biobank Pharma Proteomics Project; CKD, chronic kidney disease; HF, heart failure; CAD, coronary artery disease; MASLD, metabolic dysfunction-associated steatotic liver disease; PGS, polygenic score Associations of polygenic score of GLP-1R and plasma protein levels of GLP-1R with CKM-related diseases.(a) Associations of polygenic risk scores of GLP-1R with CKD, HF, CAD and MASLD in UK Biobank Europeans. (b) Associations of GLP-1R protein levels with CKD, HF, CAD and MASLD in UKB-PPP Europeans. Hazard ratios (HRs) were calculated across tertiles (T1–T3) of the respective PGS and protein levels within the general population, as well as in non-diabetic and non-obese subgroups. The color of the circles represents the HR, with blue indicating a reduced risk (HR < 1.0) and red indicating an increased risk (HR > 1.0) relative to the reference group (T1). Fully adjusted models were presented for both analyses. Abbreviations: UKB-PPP, UK Biobank Pharma Proteomics Project; CKD, chronic kidney disease; HF, heart failure; CAD, coronary artery disease; MASLD, metabolic dysfunction-associated steatotic liver disease; PGS, polygenic score Observational association of plasma protein levels of GLP-1R with CKM-related diseases In observational analyses, up to 41,110 general, 37,230 non-diabetic and 31,058 non-obese individuals were included.The demographic and phenotypic information for the individuals involved in the observational analysis was presented in Additional file 3: Table S15. The analysis of protein levels revealed no significant associations with studied diseases among non-diabetic and non-obese participants. (Fig. 3b and Additional file 3: Table S16) To bridge the findings from our genetic analyses with the observational analyses, we further investigated the relationship between the PGS and plasma protein levels. We found no significant correlation between the pancreatic tissue-specific PGS and measured plasma protein levels. (correlation = 0.006, p = 0.25) This suggests that these two metrics likely capture distinct biological activity of GLP-1R. Consequently, observational analysis should be viewed as independent validations for MR and PGS analyses. Given multiple layers of evidence were provided by our study, we summarised that GLP-1R expression reduced the risk of CKD, HF and MASLD in the general population, although this protective effect was not observed in non-T2DM or non-obese subpopulations. The highlight evidence of the top findings were summarized in Table 1. Table 1Evidence synthesis of effects of GLP-1R on CKM-related outcomesStudyMRPGS analysisObservational analysisPopulationGeneralGeneralNon-daibeticNon-obeseGeneralNon-diabeticNon-obeseOR (95%CI)HR (95%CI) CKD 0.96(0.95,0.97)0.97(0.93,0.99)0.97(0.93,1.01)0.99(0.95,1.03)1.00(0.91,1.11)0.97(0.86,1.10)1.03(0.90,1.17) HF 0.96(0.94,0.97)0.96(0.93,1.00)0.98(0.94,1.03)0.99(0.94,1.04)0.97(0.86,1.08)0.97(0.84,1.11)0.95(0.82,1.10) CAD 1.01(1.00,1.02)1.01(0.97,1.05)0.99(0.95,1.04)1.03(0.99,1.08)0.96(0.85,1.08)0.93(0.81,1.06)0.96(0.83,1.12) MASLD 0.96(0.93,0.98)0.96(0.92,1.01)0.98(0.92,1.04)0.99(0.93,1.07)0.89(0.75,1.06)0.89(0.73,1.10)0.87(0.69,1.11)*Results of PGS association analyses (Tertile 3 comapred to Tertile 1) were adjusted for age, sex, array and the first 10 principal components. Results of observational analyses (Tertile 3 comapred to Tertile 1) presented in the table are from the lifestyle-adjusted model, which was adjusted for age, sex, and three lifestyle factors (the Townsend index of deprivation, smoking status and alcohol intake frequency)Abbreviations: MR, Mendelian randomization; PGS, polygenic score; OR, odds ratio; HR, hazard ratio; CI, confidence interval; HR, hazard ratio; CKD, chronic kidney disease; HF, heart failure; CAD, coronary artery disease; MASLD, metabolic dysfunction-associated steatotic liver disease Table 1Evidence synthesis of effects of GLP-1R on CKM-related outcomesStudyMRPGS analysisObservational analysisPopulationGeneralGeneralNon-daibeticNon-obeseGeneralNon-diabeticNon-obeseOR (95%CI)HR (95%CI) CKD 0.96(0.95,0.97)0.97(0.93,0.99)0.97(0.93,1.01)0.99(0.95,1.03)1.00(0.91,1.11)0.97(0.86,1.10)1.03(0.90,1.17) HF 0.96(0.94,0.97)0.96(0.93,1.00)0.98(0.94,1.03)0.99(0.94,1.04)0.97(0.86,1.08)0.97(0.84,1.11)0.95(0.82,1.10) CAD 1.01(1.00,1.02)1.01(0.97,1.05)0.99(0.95,1.04)1.03(0.99,1.08)0.96(0.85,1.08)0.93(0.81,1.06)0.96(0.83,1.12) MASLD 0.96(0.93,0.98)0.96(0.92,1.01)0.98(0.92,1.04)0.99(0.93,1.07)0.89(0.75,1.06)0.89(0.73,1.10)0.87(0.69,1.11)*Results of PGS association analyses (Tertile 3 comapred to Tertile 1) were adjusted for age, sex, array and the first 10 principal components. Results of observational analyses (Tertile 3 comapred to Tertile 1) presented in the table are from the lifestyle-adjusted model, which was adjusted for age, sex, and three lifestyle factors (the Townsend index of deprivation, smoking status and alcohol intake frequency)Abbreviations: MR, Mendelian randomization; PGS, polygenic score; OR, odds ratio; HR, hazard ratio; CI, confidence interval; HR, hazard ratio; CKD, chronic kidney disease; HF, heart failure; CAD, coronary artery disease; MASLD, metabolic dysfunction-associated steatotic liver disease Evidence synthesis of effects of GLP-1R on CKM-related outcomes *Results of PGS association analyses (Tertile 3 comapred to Tertile 1) were adjusted for age, sex, array and the first 10 principal components. Results of observational analyses (Tertile 3 comapred to Tertile 1) presented in the table are from the lifestyle-adjusted model, which was adjusted for age, sex, and three lifestyle factors (the Townsend index of deprivation, smoking status and alcohol intake frequency) Abbreviations: MR, Mendelian randomization; PGS, polygenic score; OR, odds ratio; HR, hazard ratio; CI, confidence interval; HR, hazard ratio; CKD, chronic kidney disease; HF, heart failure; CAD, coronary artery disease; MASLD, metabolic dysfunction-associated steatotic liver disease Identification of genetic variants associated with MASLD We have performed a GWAS analysis of European ancestry in UK Biobank (14,231 cases and 348,091 controls). The SNP-based heritability of MASLD was 1.8% ± 0.13. With an elevated genomic inflation factor (λGC = 1.104) and a LD score regression intercept of 1.015 ± 0.003 indicating minimal population stratification, the genetic architecture of MASLD was found to be polygenic. Using bivariate LD score regression, we compare the genetic correlation between MASLD and NAFLD. Thoug a high genetic correlation (rg) of 0.86 ± 0.17 was obtained, such results still indicated a minor genetic difference of the two diseases. GWAS analysis identified eight conditionally independent genetic signals associated with MASLD using GCTA-COJO, which met conventional genome-wide significance (P < 5 × 10− 8) (Additional file 3: Table S17). The Manhattan plot and Quantile-quantile plot were presented in Additional file 4: Fig. S1. The top association signal, rs738408 (p = 2.4 × 10− 64), is a synonymous exonic SNV located at PNPLA3 gene region. Beyond the top hit, rs1260326 and rs429358 were identified as nonsynonymous exonic SNVs, implying a direct impact on protein structure. The remaining five variants were located in non-coding regions. Specifically, rs2642442 and rs56094641 were annotated as intronic variants, while rs563199662, rs2980888, and rs756350040 were classified as intronic variants within non-coding RNA. Based on the GWAS Catalogue associations, the variant rs563199662 near TFPI emerged as a novel finding among these significant loci. (Additional file 3: Table S18) The genotype-specific proteomic scan identified a distinct set of plasma proteins significantly differentiated between C/G carries and C/C carries. (Additional file 3: Table S19 and Additional file 4: Fig. S2) Subsequent KEGG pathway enrichment analysis revealed that these downregulated proteins were predominantly clustered to pathways including pancreatic secretion, cytokine-cytokine receptor interaction and JAK-STAT signaling pathway. (Additional file 3: Table S20 and Additional file 4: Fig. S2) Discussion In this study, using multiple layers of evidence with different source of data and assumptions, we observed an protective effect of the expression of GLP-1R on CKD, HF and MASLD, but not on CAD and non-ischemic HF subtypes in the general population. However, multivariable MR analyses revealed that this protective effect was attenuated after adjusting for HbA1c and BMI. Consistent with this, subgroup analyses based on both PGS and observational data showed no significant protective effect in non-diabetic or non-obese individuals. This raises concerns about efficacy and clinical value to extend the usage of GLP-1R agonists to treat CKM outcomes in healthy population. A new GWAS of MASLD, which resulted in a total of eight conditionally independent signals, of which one (rs563199662 near TFPI region) have not been reported to be associated with NAFLD before. Benefits in kidney function have been focal points in the usage of GLP-1R agonists. Previous retrospective cohort study found that compared to dipeptidyl peptidase-4 inhibitor, GLP1-R agonist could lower the mortality rate among patients with T2DM and advanced CKD [64]. The FLOW RCT reported the outstanding effect of semaglutide in mitigating risk of clinically important kidney outcomes and death in diabetic individuals [65]. Importantly, semaglutide was newly approved for reducing the risk of kidney disease worsening in diabetes patients. All these studies support the causal effects of GLP-1R agonism on protecting kidney function and reducing the risk of CKD in T2DM. Our study showed that GLP-1R expression reduced the risk of CKD and protected kidney functions in the general population. However, our PGS association and observational analyses showed that this protective effect was not found among non-diabetic or non-obese individuals. This implies that the protective effect of GLP-1R agonism on kidney outcomes is likely through its glucose or weight loss benefits rather than a direct protective effect on kidney function. Although GLP-1R agonists were reported to prevent atherosclerotic CVD events, their efficacy on HF were not clear. One meta-analysis of trials of cardiovascular outcomes indicated that GLP-1R agonists could reduce the risk of hospitalization for HF by 9% [66]. One study leveraging data from 22 RCTs found no evidence to support its protective effect on HF readmissions [67]. Our finding showed that the pancreas-specific GLP-1R expression reduced risk of HF. And we found on evidence supporting the effect of GLP-1R expression on non-ischemic HF subtypes. While a 52-weeks RCT showing that treatment with semaglutide relieved symptoms and physical limitations in patients with HFpEF [68], some other studies have explored the association between GLP-1R agonists and the core phenotype of HFpEF - left ventricular ejection fraction, but reported distinct effects [69]. Existing study estimating association of GLP-1R agonists usage with HFrEF reported a neutral effect [70]. It is noteworthy that use of GLP-1R agonists was reported to be associated with higher heart rate [71], while higher heart rate was reported to be associated with worse HErEF outcomes [72], which implies a safety concern of usage of GLP-1R agonists in patients with HFrEF. However, results of the SELECT trial revealed a protective effect of semaglutide on composite HF endpoints in overweight/obese patients with HF, regardless of HF subtypes [11]. It should be noticed non-ischemic HFpEF represents a distinct pathophysiological endotype, whereas broad diagnosis of HFpEF likely encompasses ischemia-driven pathways. As such, the association between GLP-1R expression on HF subtypes cannot be entirely ruled out. The effect of GLP-1R agonists on cardiovascular diseases worth more attentions. Most of the recent clinical trials suggested broad cardiovascular benefits of GLP-1R agonists [73]. Nevertheless, emerging evidence has highlighted certain cardiovascular concerns associated with GLP-1R agonists usage. For instance, recent studies hypothesize that semaglutide may induce cardiac atrophy in experimental models [74], and clinical usage has been associated with elevated heart rates [71]. Our study indicated that genetically predicted increased GLP-1R expression was associated with a slightly higher risk of CAD in the general population. This finding must be interpreted with caution. First, the genetic evidence cannot be directly compared with clinical trial evidence since we were estimating the effect of GLP1R gene expression rather than the direct drug effect of GLP-1R agonists. Second, the MR were performed in the general population, while the trials were performed in obese and/or diabetic individuals. These two populations have different physiological and pathological states. We found evidence to support the associations between GLP-1R expression and MASLD. Previous clinical trials have explored the potential effects of GLP-1R agonists in treating MASLD, results from recent study showing emergingly promising benefits of GLP-1R agonists in treating MASLD. Recently, a RCT incorporating 94 MASLD patients reported a protective effect of GLP-1/glucagon dual receptor agonist on liver fat content and hepatic inflammation [75]. Meanwhile, a retrospective cohort study suggested that use of GLP-1R agonists could reduce risk of progression to cirrhosis and mortality among patients with MASLD and diabetes [76]. What’s more, there are series of ongoing clinical trials exploring the effect of GLP-1R agonists on treating MASLD [77], which will provide additional insights into the therapeutic benefit of GLP-1R agonists. However, according to our findings, caution is required when considering the usage of GLP-1R agonists to prevent MASLD in the non-T2DM and non-obese individuals. Our GWAS analysis report eight variants robustly associated with MASLD risk. While most of these genetic variants were found to be directly associated with NAFLD and liver-related phenotypes [78–81], this GWAS analysis further supported the robustness of these loci and revealed important information on the genetic architecture of MASLD. Rs563199662, identified in our study, is a potentially novel susceptible variant for MASLD. Our analysis identified a set of downstream protesins enriched in JAK-STAT signaling pathway and cytokine-cytokine receptor interaction, which are critical mediators of the systemic inflammatory response and are well-established drivers of hepatic inflammation and fibrosis progression [82, 83]. Notably, TFPI regulates coagulation and has emerging roles in lipid metabolism [84], mechanisms which are plausibly relevant to MASLD pathogenesis. Given the well-established association between obesity and MASLD progression, we hypothesize that genetic variation at this locus may contribute to MASLD risk by modulating adiposity or metabolic dysfunction and could serve as a potential indirect therapeutic target for MASLD. Our study has several strengths. First, we selected three instruments with acceptable instrument strengthen (F statistics = 20). The positive control analysis further showed that our instruments were associated with reduced risk of T2DM, and reduced levels of HbA1c, BMI and body weight. All of these outcomes are known to be regulated by GLP-1R agonists. Therefore, we generated and offered a set of genetic variants with reliability and versatility. Second, we employed a variety of methods including MR, PGS association and observational analyses to investigate the association between GLP-1R expression and CKM-related diseases across different sub-populations. By utilizing these study designs, we increased the reliability of our findings, extended the generalizability of GLP-1R agonism on CKM-related diseases in the general population, and provided new insights into the appropriate populations for design of future trials of GLP-1R agonists. There are several limitations in our study. First, genetic instruments we selected did not reach genome-wide significance, which may introduce the weak instruments bias. However, the instrument strengthen measurement, F-statistic, of the instruments was over 20, which indicating good instrument strength. Moreover, we have applying multiple MR methods allowing the presence of weak instruments, which further minimised the influence of instrument strength. Moreover, in our validated dIVW MR analyses, the built-in F-statistics met the recommended criteria and further confirmed the reliability of MR estimators. (see Additional file 3: Tables S2, S4, S5 and S6) Second, it should be noticed that non-ischemic HF subtypes defined in HERMES represents a stricter subset of the broader HF subtypes. Given this phenotypic heterogeneity, we have interpreted these results with cautions. Third, as the current ICD-10 diagnostic criteria were unable to accurately differentiate HF subtypes, we did not include corresponding analyses in our PGS and observational studies. Fourth, we acknowledge the potential limitations regarding MASLD phenotyping precision. Our case definition relied on electronic health records, ICD codes, and the FLI index, which could overlook asymptomatic cases or lead to misclassification. Fifth, we acknowledge potential power limitations for specific outcomes with lower event rates in PGS and observational subgroup analyses. Lastly, analyses in our study were predominantly constructed among European individuals. As such, the generalizability of our findings to other ancestry groups is restricted. Conclusions In conclusion, we identified causal roles of GLP-1R expression on CKM related phenotypes including CKD, HF and MASLD in the general population. Our study provides evidence to support the usage of GLP-1R agonists as potential intervention target for the prevention of CKD, HF and MASLD. In contrast, our study did not provide conclusive evidence to support the protective effect of GLP-1R expression on CAD, especially in the non-T2DM and non-obese individuals. A more conservative approach is warranted when prescribing this medication to pre-diabetic, overweight, and even healthy populations. Future RCTs targeting non-disease population is warranted to establish robust evidence and guide future clinical guidelines on the usage of GLP-1R agonists. Supplementary Information Below is the link to the electronic supplementary material. Additional file 1: Supplementary Methods Additional file 1: Supplementary Methods Additional file 2: Figures S1-S2. Additional file 2: Figures S1-S2. Additional file 3: Tables S1-S21. Additional file 3: Tables S1-S21. Additional file 4: STROBE-MR checklist Additional file 4: STROBE-MR checklist