Unveiling CTRB2, RSPO3, KLOTB, and ROR1 as obesity-pancreatic disease association proteins: a comprehensive Mendelian randomization study

Abstract

Background

Obesity is recognized as a prominent contributing factor for pancreatic diseases; however, the mechanisms remain elusive. We aimed to identify the mediating role of circulating proteins in these associations.

Methods

A two-step Mendelian randomization (MR) was conducted to investigate associations between nine obesity indicators, thousands of circulating proteins, with three pancreatic diseases (acute pancreatitis, chronic pancreatitis, and pancreatic carcinoma). Colocalization analyses were performed to validate these associations. Protein mediating networks among obesity indicators and pancreatic diseases were investigated by mediation analysis.

Results

Genetically predicted circulating levels of 4, 2, and 2 proteins were associated with acute pancreatitis, chronic pancreatitis, and pancreatic carcinoma, respectively. In mediation analysis, decreased chymotrypsin B2 (CTRB2) levels mediated 1.03% (95% CI [confidence interval] 0.02%–2.03%) of the effects of body mass index on acute pancreatitis. Increased R-spondin 3 (RSPO3) levels mediated the effects of body mass index (2.95%, 95% CI 0.18%–5.73%), body fat percentage (4.53%, 95% CI 1.11%–7.96%), waist–hip ratio (8.48%, 95% CI 3.11%–13.86%), and visceral adipose tissue (3.93%, 95% CI 0.64%–7.22%) on acute pancreatitis. We also found increased klotho beta (KLOTB) levels mediated the effects of waist–hip ratio (7.01%, 95% CI 3.30%–10.71%) and visceral adipose tissue (8.98%, 95% CI 4.55%–13.41%) on chronic pancreatitis, and decreased receptor tyrosine kinase-like orphan receptor 1 (ROR1) levels mediated the effects of body mass index (10.39%, 95% CI 3.36%–17.42%) and visceral adipose tissue (6.29%, 95% CI 1.00%–11.58%) on pancreatic carcinoma.

Conclusions

The MR suggests that circulating CTRB2, RSPO3, KLOTB, and ROR1 proteins may mediate associations between obesity and pancreatic diseases.

Introduction

Acute pancreatitis, chronic pancreatitis, and pancreatic carcinoma represent some of the most severe and life-threatening pancreatic diseases, posing significant burdens on human health [1]. Moreover, these conditions are intricately interlinked, wherein recurrent acute pancreatitis can progress to chronic pancreatitis [2], which in turn heightens the risk for pancreatic carcinoma [3]. However, current treatment options for pancreatic diseases are limited, often requiring invasive surgical procedures that carry risks and complications [4]. With limited therapeutic options available, elucidating the mechanisms underlying pancreatic disease development and identifying potential therapeutic targets are crucial for disease prevention and treatment.

Extensive evidence shows associations of obesity and obesity-related conditions with the increased risk of pancreatic diseases [5–8]. A meta-analysis study including 10 articles demonstrated that per 5 kg/m² increase in body mass index (BMI) was related to a 1.18-fold increase in the risk of acute pancreatitis [9]. Another meta-analysis of 56 studies found that each 5 kg/m² increase in BMI was associated with a 1.17-fold increased relative risk for pancreatic carcinoma [10]. Moreover, a cohort study with 62 chronic pancreatitis patients found that over half of people were overweight or obese [8]. While significant progress has been made in understanding the molecular mechanisms underlying the association between obesity and pancreatic diseases, there are still many aspects that require further clarification, particularly in identifying mediating pathways with potential drug targets [11, 12]. The identification of protein mediators plays a pivotal role in furthering the progress of novel drug development for pancreatic diseases induced by obesity, given the direct involvement of proteins in gene regulation and their frequent utilization as targets for pharmacological interventions [13]. Evidence from observational studies indicated that patients with obesity exhibited elevated levels of various proteins, including C-reactive proteins, in comparison to patients without obesity [14, 15]. Proteins including leptin and adiponectin have been shown to play important roles in the obesity-induced pancreatic carcinoma in numerous animal experiments [11]. Nevertheless, the identities of specific proteins that may mediate effects of obesity on pancreatic diseases remain to be fully understood. Moreover, the causal inference from observational evidence is constrained by confounding factors and the potential for reverse causation.

Mendelian randomization (MR) is an epidemiological approach applied to investigate causal relationships between exposures and outcomes [16]. Genetic variations are extensively used as instrumental variables for exposure in MR studies. Confounding and reverse causation can be minimized because genetic variants are randomly allocated at conception (before disease onset) and remain unaffected by environmental factors [16]. Valid instruments for MR need to meet three key requirements: (i) instruments are strongly associated with exposure; (ii) instruments are not associated with confounding factors and (iii) instruments influence the risk of the outcome directly through the exposure, rather than through other pathways [16]. Two additional assumptions required are that the exposure-mediator effects and mediator-outcome effects should be linear without interactions; and the effects between the exposure, mediator, and outcome should be homogenous [17, 18]. In this study, we conducted MR analyses to identify the effects of circulating proteins in mediating the associations of nine obesity-related indicators with acute pancreatitis, chronic pancreatitis, and pancreatic carcinoma.

Methods

Study design

As shown in Figure 1, we first identified plasma proteins associated with pancreatic diseases (Step 1) and assessed whether these proteins act as mediators in the relationship between obesity and pancreatic diseases (Step 2). We further evaluated the druggability of the significant mediational proteins and phenome-wide MR was employed to predict potential on-target side effects of drug targets. The ethics of included studies were approved by the relevant committees. This work has been reported in line with the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) criteria [19].

Figure 1.

Study design. AP = acute pancreatitis, CP = chronic pancreatitis, PC = pancreatic carcinoma, FDR = false discovery rate.

Circulating protein data source

Genetic associations with 4,907 circulating protein levels were extracted from a genome-wide association study (GWAS) from the deCODE study, which comprised up to 35,559 Icelanders [20]. GWAS for each protein was conducted with adjustments for age, sex, and sample age [20]. Cis-region was defined as ±1 Mb from each plasma protein-encoding gene. Significant (P < 5 × 10⁻⁸) and independent (linkage disequilibrium threshold of r² < 0.1 in a 10,000 kb window, according to the 1000G linkage disequilibrium reference data set for Europeans) cis-acting single-nucleotide polymorphisms (SNPs) were then selected as instrumental variables [21]. Instruments within the human major histocompatibility complex region (MHC, chromosome 6: 28,477,797–33,448,354) were excluded from our study as its complex linkage disequilibrium (LD) structure may lead to horizontal pleiotropy. This set of instrumental variables was used in the discovery analysis. For replication, we utilized data from the Fenland study in the UK [22], which evaluated 4,979 plasma proteins in 10,708 British individuals. The same criteria for the selection of instrumental variables were applied.

Outcome data source

The outcomes were three pancreatic diseases including acute pancreatitis, chronic pancreatitis, and pancreatic carcinoma. Summary-level data for acute pancreatitis were extracted from a GWAS meta-analysis of the UK Biobank [23], the Estonian Biobank, and the FinnGen study [24] which included 10,630 cases and 844,679 controls [21]. We conducted a meta-analysis of GWASs for chronic pancreatitis and pancreatic carcinoma with data from the UK Biobank and the FinnGen study, yielding a total of 3,834 cases and 737,174 controls for chronic pancreatitis, as well as 2,005 cases and 680,509 controls for pancreatic carcinoma. The UK Biobank is a large cohort study consisting of more than 500,000 participants in the UK from 2006 to 2010 [23]. Summary statistics for the UK Biobank study were gathered from the Lee lab [25]. The FinnGen study is a growing project consisting of genetic data from Finnish biobanks and health data from the Finnish health registries [24]. We utilized the R9 data releases in this study. The detailed information can be found in Supplementary Table S1. Meta-analyses of GWASs were performed under a fixed-effects model in METAL software [26]. Genomic control correction was applied to account for small amounts of population stratification or unaccounted-for relatedness [26].

Obesity-related indicators

Nine obesity-related indicators were selected as exposures, including BMI, body fat percentage (BFP), visceral adipose tissue (VAT), waist–hip ratio (WHR), whole body fat mass (WBFM), whole body fat-free mass (WBFFM), fat mass index (FMI), fat-free mass index (FFMI), and pancreatic fat (PF). Summary-level data for BMI and WHR were derived from a meta GWAS of the Genetic Investigation of Anthropometric Traits consortium and the UK biobank study, which included up to 806,834 individuals of European ancestries [27]. Data on BFP [28], VAT [29], FMI [30], FFMI [30], WBFM [31], and WBFFM [31] were obtained from the UK Biobank study, involving 331,291 to 492,983 European individuals. Data on BFP were obtained from Medical Research Council Integrative Epidemiology Unit at the University of Bristol (MRC-IEU) [32], while data of WBFM and WBFFM were gathered from the Neale lab. Genetic associations with PF were extracted from GWAS of 25,617 European-ancestry individuals from the UK Biobank [33]. BMI is calculated by dividing weight by the square of height [27]. WHR is calculated as the ratio of waist circumference to hip circumference [27]. BFP is measured by a body composition analyzer [32]. WBFM and WBFFM are measured by bioelectrical impedance [31]. FMI and FFMI were computed by dividing fat mass or fat-free mass by the square of height [30]. VAT is estimated using a machine-learning method with a training dataset measured by dual-energy X-ray absorptiometry [29]. PF is estimated using a machine-learning method with MRI scans [33]. Comprehensive details regarding the genetic instruments are presented in Supplementary Table S1. We extracted SNPs associated with these obesity indicators at the genome-wide significance level (P < 5 × 10⁻⁸) from the aforementioned GWASs. The LD of the chosen SNPs for each trait was assessed using the 1,000 Genomes European reference panel. After removing SNPs in LD (r² ≥  0.001) and in the MHC region, the residual SNPs were used as instrumental variables for obesity indicators in MR analysis.

Statistical analysis

MR analysis

The inverse variance weighted method (IVW) with a multiplicative random-effect model and the Wald ratio method were used as the primary statistical methods in the current study. The IVW method was used when more than one SNPs were included in the analysis, whereas the Wald ratio method was utilized when only one SNP was included in the analysis [34]. Notably, cis-acting SNPs were used as instrumental variables exclusively when circulating proteins served as the exposure [21]. It is because cis-acting SNPs reside near the transcription start site of the protein-coding gene and are more likely to directly influence the protein levels. The IVW method provides the highest precision if all SNPs are valid or when the pleiotropy is balanced [35]. Cochran’s Q test was used to assess heterogeneity. The Wald ratio is a straightforward method that directly uses the effect size estimates of a single SNP to infer the relationship between the exposure and the outcome [36]. The MR-Egger regression [37] and weighted median methods [38] were performed as sensitivity analyses to evaluate the robustness of results and detect directional pleiotropy. The MR-Egger regression can detect unbalanced horizontal pleiotropy through the intercept test and provide less precise estimates. Consistent causal estimates are obtained with the weighted median method when valid instruments contribute to over 50% of the weight. In the discovery analysis, multiple comparisons were corrected for the false discovery rate (FDR), and associations with an adjusted P value < 0.05 were considered significant. For replication, we performed MR analyses utilizing data from the Fenland study for proteins that were deemed significantly associated with pancreatic diseases in the discovery stage. We regarded P value < 0.05 in the replication analysis as a successful replication. Proteins identified (with adjusted P < 0.05 and median to high support evidence of colocalization) in the discovery analysis were selected in the mediation analyses.

Two-step MR analyses were conducted using the coefficient method in order to estimate the mediation effects of circulating plasma proteins between obesity-related indicators and pancreatic diseases (Figure 1) [39]. The effects of the exposures on the mediators (β_{Obesity-Protein}) and the total effects of the exposures on the outcomes (β_{Obesity-Pancreatic-disease}) were calculated by the IVW method. We adopted the estimates of the proteins on pancreatic diseases (β_{Protein-Pancreatic-disease}) from the discovery stage in Step 1. Mediation effect (β_mediation) was calculated by multiplying β_{Obesity-Protein} and β_{Protein-Pancreatic-disease} [39]. Subsequently, we calculated the proportion of the mediation effects (β_mediation/β_{Obesity-Pancreatic-disease}). Proteins that had significant associations (adjusted P value < 0.05) in both steps of the two-step MR analysis were included. To ensure the meaningfulness of the mediation effect, we excluded the following combinations: (i) those with an insignificant β_{Obesity-Pancreatic-disease} (P value ≥ 0.05); (ii) those with β_mediated × β_{Obesity-Pancreatic-disease} < 0, indicating inconsistent directions; and (iii) those where the absolute value of β_mediation exceeded the absolute value of β_{Obesity-Pancreatic-disease}, suggesting a potential suppression effect. We employed the coefficient product method to assess the mediation effect and calculated the confidence interval for the proportion of the mediating effect using the error propagation method [40]. Additionally, we calculated the P value for the mediation effect, with the null hypothesis stating that the causal effect of obesity-related indicators on pancreatic diseases is not mediated by the protein. The protein mediators were considered as potential drug targets and further analyzed in Step 3. Analyses were performed using the TwoSampleMR package v0.5.7 in R software (version 4.2.0).

Colocalization analyses

To further identify the shared causal loci between plasma proteins and the three pancreatic diseases, and to avoid false positives owing to LD, we conducted a colocalization analysis of selected proteins. Colocalization analysis was performed using the coloc R package [41]. We evaluated whether SNPs located within ± 500 kb of the lead pQTL (defined as the protein quantitative trait loci with the smallest P value) in the deCODE dataset shared a similar casual variant with each pancreatic disease [42]. The parameter settings (P1 = P2 = 1 × 10⁻⁴, P12 = 1 × 10⁻⁵) were used in the colocalization analyses. P1 and P2 were the prior probabilities that an SNP was associated with either of these two traits, and P12 was the prior probability that an SNP was related to both two traits [41]. Different posterior probabilities correspond to five different hypotheses: PP H0, SNP had no association with either trait; PP H1, and PP H2, SNP was only associated with trait 1 or trait 2; PP H3, two independent SNPs were associated with trait 1 and trait 2 separately; PP H4, the shared SNP was associated with trait 1 and trait 2 [41]. We considered proteins with a posterior probability of hypothesis 4 (PP H4) > 0.8 as indicating high colocalization support, while proteins with 0.5 < PP H4 < 0.8 were categorized as having medium colocalization support. Additionally, we validated proteins with high and medium colocalization support in the Fenland dataset, and all of them were included in Step 2.

Identification of druggable targets

We queried the DrugBank [43], Dependency Map [44], and Open Targets [45] to assess the druggability of the protein mediators. For druggable proteins, we recorded the drug names, development stages, and their therapeutic indications. Proteins were classified into four categories based on the stage of drug development: (i) approved (protein-targeting drugs have been approved); (ii) in clinical trials (protein-targeting drugs are still undergoing clinical trials); (iii) preclinical (protein-targeting drugs have been confirmed but have not yet undergone clinical trials); (iv) druggable (no drug designed to target this protein specifically, but the protein is a druggable target).

Phenome-wide MR analyses

We conducted phenome-wide MR analyses to investigate the impact of potential drug-targeting proteins on other diseases and the potential side effects of the drugs. The potential drug-targeting proteins from the deCODE study selected in Step 2 were used as exposures. GWAS summary data of 2,272 phenotypes in the FinnGen R9 datasets were used as the outcomes in the phenome-wide MR analyses. Briefly, MR analyses were conducted for each protein in the exposure and each phenotype in the outcome separately. The FDR was calculated on the phenome-wide MR results for each protein. Phenotypes with adjusted P value < 0.05 were considered significant.

Results

Step 1. Identify pancreatic disease-associated proteins

We evaluated the associations between circulating plasma protein levels and the risk of acute pancreatitis, chronic pancreatitis, and pancreatic carcinoma in proteome-wide MR analyses. The summary of the results is shown in Figure 2. 1,728, 1,738, and 1,743 proteins were analyzed for acute pancreatitis, chronic pancreatitis, and pancreatic carcinoma, respectively. We identified 48 proteins significantly associated with acute pancreatitis, 26 with chronic pancreatitis, and 26 with pancreatic carcinoma after multiple testing corrections (Table 1 and Supplementary Table S2). For each standard deviation increase in protein levels, the odds ratio ranged from 0.39 to 1.21 for acute pancreatitis, 0.72 to 2.75 for chronic pancreatitis, and 0.16 to 2.14 for pancreatic carcinoma. Supplementary Figure S1 depicts overlaps among proteins associated with each pancreatic disease. The effects for significant proteins were directionally consistent in the weighted median and MR-Egger analyses (Supplementary Table S2). We did not identify heterogeneity among SNPs of these proteins. Pleiotropy was detected in 2, 3, and 2 circulating proteins in the analyses of acute pancreatitis, chronic pancreatitis, and pancreatic carcinoma (Supplementary Table S2). The direction of associations between the identified proteins and pancreatic diseases was consistent in the replication analyses (Supplementary Table S3).

Figure 2.

Scatter plot illustrating the effect of genetically predicted circulating protein levels on pancreatic diseases. Proteomics data from the deCODE study are regarded as the exposure. (A) The effect of proteins on acute pancreatitis. (B) The effect of proteins on chronic pancreatitis. (C) The effect of proteins on pancreatic carcinoma. The y-axis corresponds to the logarithm transformation of the P values of the inverse variance weighted analyses.

Table 1.

Association between identified circulating proteins and pancreatic diseases

Gene	Protein name	OR (95% CI) a	P value	Colocalization analysis PP H4
Discovery: Proteins in the deCODE dataset
Acute pancreatitis
CTRB2	CTRB2	0.92 (0.91, 0.94)	1.29 × 10−13	0.978
UGT1A6	UGT1A6	0.81 (0.76, 0.87)	2.63 × 10−8	0.889
UGT1A1	UGT1A1	0.79 (0.72, 0.86)	1.15 × 10−7	0.897
RSPO3	RSPO3	1.21 (1.11, 1.31)	8.57 × 10−6	0.855
Chronic pancreatitis
KLB	KLOTB	1.11 (1.06, 1.16)	6.69 × 10−7	0.954
CTRB2	CTRB2	0.93 (0.90, 0.97)	7.56 × 10−5	0.585
ABO	BGAT	1.05 (1.02, 1.07)	1.40 × 10−4	0.640
Pancreatic carcinoma
ABO	BGAT	1.17 (1.13, 1.22)	8.88 × 10−19	0.962
ROR1	ROR1	0.73 (0.64, 0.85)	3.10 × 10−5	0.774
Replication: Proteins in the Fenland dataset
Acute pancreatitis
CTRB2	CTRB2	0.93 (0.91, 0.95)	2.00 × 10−11	0.974
UGT1A6	UGT1A6	0.86 (0.77, 0.96)	8.61 × 10−3	0.897
UGT1A1	UGT1A1	0.82 (0.77, 0.89)	2.24 × 10−7	0.885
RSPO3	RSPO3	1.11 (1.04, 1.19)	3.25 × 10−3	0.679
Chronic pancreatitis
KLB	KLOTB	1.10 (1.06, 1.15)	1.96 × 10−7	0.307
CTRB2	CTRB2	0.94 (0.91, 0.96)	6.73 × 10−6	0.715
ABO	BGAT	–	–	–
Pancreatic carcinoma
ABO	BGAT	–	–	–
ROR1	ROR1	0.63 (0.50, 0.80)	1.06 × 10−4	0.835

^aThe OR of outcome disease was scaled to one standard deviation increase in genetically predicted gene expression levels. OR = odds ratio, CI = confidence interval.

Colocalization analyses were conducted between pancreatic diseases and identified proteins (Table 1 and Supplementary Table S4). For acute pancreatitis, four proteins including chymotrypsin B2 (CTRB2), UDP glucuronosyltransferase family 1 member A6 (UGT1A6), UDP glucuronosyltransferase family 1 member A1 (UGT1A1), R-spondin 3 (RSPO3) showed high colocalization evidence (PP H4 > 0.8). For chronic pancreatitis, klotho beta (KLOTB) showed high support evidence of colocalization, whereas CTRB2 and histo-blood group ABO system transferase (BGAT) had medium support (0.5 < PP H4 < 0.8). For pancreatic carcinoma, BGAT had high support evidence for colocalization, while receptor tyrosine kinase-like orphan receptor 1 (ROR1) had medium support evidence. All proteins were validated in the Fenland dataset except RSPO3 (medium colocalization), KLOTB (lacking colocalization), and BGAT (not exist). Proteins with adjusted P value < 0.05 and medium to high support for colocalization were included in Step 2.

Step 2. Identify protein mediators between obesity and pancreatic disease

We conducted two-step MR analyses to quantify the proportion of the total impact of obesity on pancreatic diseases mediated by the circulating plasma proteins identified in Step 1. After adjusting for multiple comparisons, genetically predicted BMI, BFP, VAT, WHR, WBFM, and WBFFM were associated with an increased risk of one or more pancreatic diseases (Figure 3 and Supplementary Table S5). The MR results of the associations of obesity-related indicators with circulating protein levels are presented in Supplementary Table S6. The MR results of obesity-related indicators on circulating protein levels were also replicated (Supplementary Table S7).

Figure 3.

Heat map for associations between obesity indicators and genetically predicted circulating protein levels and risk of pancreatic diseases. This figure shows the MR results of the association between obesity and circulating protein levels and risk of pancreatic diseases. GWAS data of proteins are sourced from both the deCODE and Fenland studies. Significant associations (adjusted P value < 0.05) were marked by the star. BMI = body mass index, BFP = body fat percentage, WHR = waist–hip ratio, VAT = visceral adipose tissue, FMI = body fat mass index, FFMI = body fat-free mass index, WBFM = whole body fat mass, WBFFM = whole body fat-free mass, PF = pancreatic fat.

After integration, four proteins were observed to be the mediators between the obesity-related indicators and pancreatic diseases (Table 2 and Figure 4). Specifically, decreased CTRB2 levels mediated 1.03% (95% CI 0.02%–2.03%) effects of BMI on acute pancreatitis. The increased RSPO3 levels mediated the effect of BMI (2.95%, 95% CI 0.18%–5.73%), BFP (4.53%, 95% CI 1.11%–7.96%), WHR (8.48%, 95% CI 3.11%–13.86%), and VAT (3.93% 95% CI 0.64%–7.22%) on acute pancreatitis. We also found that the increased KLOTB levels mediated the effect of WHR (7.01%, 95% CI 3.30%–10.71%) and VAT (8.98%, 95% CI 4.55%–13.41%) on chronic pancreatitis, and the decreased ROR1 levels mediated the effect of BMI (10.39%, 95% CI 3.36%–17.42%) and VAT (6.29%, 95% CI 1.00%–11.58%) on pancreatic carcinoma. The four circulating plasma proteins were defined as potential drug targets and further analyzed in Step 3.

Table 2.

Mediation proportion of proteins selected in Step 1

Exposure	Mediation protein name	Outcome	Mediation proportion (%)	95% CI lower bounds (%)	95% CI upper bounds (%)	P value	Protein expression a
BMI	CTRB2	Acute pancreatitis	1.03	0.02	2.03	0.046	Decrease
BMI	RSPO3	Acute pancreatitis	2.95	0.18	5.73	0.037	Increase
BFP	RSPO3	Acute pancreatitis	4.53	1.11	7.96	0.009	Increase
WHR	RSPO3	Acute pancreatitis	8.48	3.11	13.86	0.002	Increase
VAT	RSPO3	Acute pancreatitis	3.93	0.64	7.22	0.019	Increase
WHR	KLOTB	Chronic pancreatitis	7.01	3.30	10.71	2.12 × 10−4	Increase
VAT	KLOTB	Chronic pancreatitis	8.98	4.55	13.41	7.16 × 10−5	Increase
BMI	ROR1	Pancreatic carcinoma	10.39	3.36	17.42	0.004	Decrease
VAT	ROR1	Pancreatic carcinoma	6.29	1.00	11.58	0.020	Decrease

^aThe levels of circulatory proteins, either increase or decrease, in the process of obesity leading to pancreatic diseases. BMI = body mass index, BFP = body fat percentage, WHR = waist–hip ratio, VAT = visceral adipose tissue, CI = confidence interval.

Figure 4.

Mediation proportion of proteins between obesity and pancreatic diseases. (A) CTRB2 and RSPO3 mediated the effects of obesity indicators on acute pancreatitis. (B) KLOTB mediated the effects of obesity indicators on chronic pancreatitis. (C) ROR1 mediated the effects of obesity indicators on pancreatic carcinoma.

Step 3. Follow-up analyses for CTRB2, RSPO3, KLOTB, and ROR1

CTRB2, KLOTB, and ROR1 were classified as druggable by a ligand-based assessment from the dependency map (Supplementary Table S8). Rosmantuzumab, an RSPO3 inhibitor, has been investigated in clinical trials (NCT02482441, Phase I) for the treatment of advanced solid tumors and acute myeloid leukemia. Drugs activating KLOTB were designed for diabetes, fatty liver diseases, sclerosing cholangitis, and so on (Supplementary Table S8). Furthermore, a ROR1 inhibitor was investigated in clinical trials for treating pancreatic carcinoma and other tumor diseases (Supplementary Table S8).

To evaluate the potential adverse effects of the drugs, we performed phenome-wide MR analyses (Supplementary Figure S2 and Supplementary Table S9). There was no trait significantly associated with CTRB2 or ROR1 (adjusted P value < 0.05). Genetically predicted higher RSPO3 levels were associated with increased risks of several genitourinary diseases, whereas higher RSPO3 levels were associated with decreased risks of certain injury-related traits. Genetically predicted higher KLOTB levels were associated with increased risks of some injury traits, mental disorders, musculoskeletal traits, and neoplasms.

Discussion

We performed proteome-wide MR analyses and utilized the colocalization analyses to ascertain the effect of over 4,000 circulating plasma proteins on pancreatic diseases. Four, two, and two protein-acute pancreatitis, protein-chronic pancreatitis, and protein-pancreatic carcinoma associations were uncovered, respectively. Subsequently, two-step MR analyses were performed to investigate their mediation role in the process of linking obesity-related indicators to pancreatic diseases. We identified four circulating proteins including CTRB2, RSPO3, KLOTB, and ROR1 as potential drug targets for obesity-induced pancreatic diseases, which are illustrated in Figure 4.

In the current study, we uncovered that genetically predicted circulating CTRB2 levels were associated with decreased risks of acute pancreatitis and chronic pancreatitis. CTRB2 is a subtype of chymotrypsin, which belongs to a group of digestive proteases secreted by the pancreas. Consistent with our findings, previous studies have highlighted the protective effect of CTRB2 in pancreatitis. In a GWAS involving 3,609 individuals with chronic pancreatic diseases, researchers found that a 16.6 kb inversion at the CTRB1-CTRB2 locus was associated with a decreased risk of chronic pancreatitis [46]. Evidence from observational studies suggested that increased expression of CTRB2 levels exerted a protective effect against pancreatitis [47]. Our research further unveiled that the decreased circulating CTRB2 levels mediated the effect of BMI on acute pancreatitis. Specifically, an elevated BMI was associated with decreased CTRB2 levels, consequently leading to an increased risk of acute pancreatitis. Since the activation of trypsinogen to trypsin in the pancreas is a key initiator of pancreatitis, the degradation of trypsinogen by chymotrypsin (including CTRB2) is an important antitrypsin defense mechanism in the pancreas [48]. Increased CTRB2 expression, thereby promoting the degradation of human anionic trypsinogen, may play a protective role in pancreatitis [46]. PheWAS-MR results in this study showed that CTRB2 was not associated with any other traits, suggesting no side effects for drugs targeting CTRB2. In general, CTRB2 might be a potential drug target for obesity-related pancreatitis despite the absence of any current drug developments targeting CTRB2. These findings underscore the potential therapeutic significance of CTRB2 in addressing obesity-related pancreatitis, warranting further exploration and drug development efforts in this field.

As a member of the R-spondin family, RSPO3 functions as a positive regulator of Wnt/beta-catenin and Wnt/planar cell polarity signaling pathways [49]. While previous studies concentrated on investigating the impact of RSPO3 on tumor development and aggressiveness [50], its role in pancreatitis has remained elusive. A cytological experiment revealed that RSPO3, together with pro-inflammatory IL-1 beta, can enhance the permeability of endothelial cells [51]. This heightened permeability of vascular endothelial cells, leading to the extravasation of blood components and immune cell transmigration, constitutes a primary pathogenic mechanism in numerous inflammatory diseases [51]. This could potentially explain the association between RSPO3 and an elevated risk of acute pancreatitis. In this study, we identified that increased circulating RSPO3 levels mediated the effect of BMI, BFP, WHR, and VAT on acute pancreatitis. While the link between obesity and circulating RSPO3 has not been extensively examined, it has been observed that a high-fat diet can lead to the up-regulation of RSPO3 in animal experimentation [52]. Notably, rosmantuaumab has been developed as an RSPO3 inhibitor for the treatment of advanced solid tumors and metastatic colorectal cancer [53]. Considering the mediating effect of RSPO3 found in this study, exploring the potential use of this drug in the context of obesity-related acute pancreatitis holds significant promise and warrants further investigation.

Results from the current study also indicated that circulating KLOTB levels acted as a mediator in the association of obesity measured by WHR and VAT with chronic pancreatitis. KLOTB (also called β-Klotho) is a beta type of Klotho and plays a necessary role in the binding of fibroblast growth factor (FGF). As a key mediator of the regulatory effects of FGF, KLOTB is required for energy expenditure and glucose metabolism, especially in liver and pancreas [54, 55]. A previous experimental study has suggested that the elimination of the KLOTB gene leads to a decrease in the secretion of digestive enzymes stimulated by FGF [56], which may provide an explanation for the observed association between KLOTB and chronic pancreatitis as identified in our MR study. While relations between obesity and KLOTB have rarely been studied, evidence from animal experiments revealed that a high-fat diet could upregulate the level of KLOTB in female mice [57]. Further research is necessary to provide a comprehensive understanding of the underlying mechanism by which KLOTB mediates the relationship between obesity and chronic pancreatitis.

ROR1 was a tyrosine kinase-like orphan receptor that was expressed in cancer cell lines but almost not expressed in all normal tissues. A study employed immunohistochemistry to investigate the expression of cell-surface ROR1 in diverse human tumor tissues, revealing its presence in pancreatic cancer cells [58]. Experiments involving cells and mice found intratumoral cells exhibiting high levels of ROR1 [59]. Inconsistent with previous research, our findings indicate that circulating ROR1 levels were associated with decreased risks of pancreatic carcinoma. This discrepancy could be explained that the increased levels of ROR1 may exert a protective role in pancreatic carcinoma. Besides, incomplete control for confounding or reverse causation bias could not be ruled out in observed evidence. Our MR investigation uncovered the relationship between ROR1 and pancreatic carcinoma and further revealed the mediator role of ROR1 in the association of obesity (measured by BMI and VAT) with pancreatic carcinoma.

Previous studies have explored pancreatic cancer-related proteins. A previous study identified 38 candidate protein biomarkers for pancreatic cancer in MR method, using proteomic data from the INTERVAL study and data from the European descent from the Pancreatic Cancer Cohort Consortium (PanScan) and the Pancreatic Cancer Case-Control Consortium (PanC4) [60]. Additionally, another proteome-wide association study used comprehensive protein genetic prediction models as instruments to assess the associations between genetically predicted blood concentrations of proteins and the risk of pancreatic ductal adenocarcinoma [61]. Going further, we partially confirmed these previous studies (e.g. BGAT protein) and we supplement new evidence for pancreatitis and pancreatic carcinoma associated proteins in proteome-wide MR and colocalization analysis methods. In addition, our comprehensive study reaffirmed the strong associations of obesity (measured by BMI, WHR, VAT, and BFP) with the risk of pancreatitis and pancreatic carcinoma, identified four protein mediators and revealed the potential avenues for therapeutic interventions. The implementation of non-pharmacological interventions (specifically those that encourage physical activity and dietary modifications) among individuals with obesity has the potential to alleviate the pancreatic disease burden owed to obesity, which can be achieved through the up-regulation of proteins, such as CTRB2 and ROR1, as well as the dysregulation of proteins like RSPO3 and KLOTB. The modulation of these proteins presents a promising pathway toward ameliorating pancreatic diseases stemming from obesity; however, additional in-depth investigations are imperative to validate and harness the therapeutic potential of these proteins in real-world medical practices.

The strength of this study lies in the following aspects. First, we conducted the MR analyses which were less prone to the unmeasured confounding and reverse causation compared to observational studies. The colocalization analysis was performed which was an effective method to reveal whether two traits share the same casual genetic variant and strengthen the correlation between them. Second, the effect of four circulating plasma proteins on pancreatic diseases was replicated in a variety of sensitivity analyses and replication analyses, ensuring the robustness of the results. Third, nine obesity-related indicators were included to comprehensively assess the impact of obesity in different aspects. Furthermore, a variety of analyses were conducted to evaluate the druggability and side effects of the protein mediators.

Several limitations should be noted. First, we conducted all analyses based on European individuals so that we could minimize the biases caused by ancestry, necessitating further exploration of protein mediation effects in other ancestries. Secondly, due to the utilization of summary-level GWAS data, the stratification of analyses based on covariates of interest, such as sex, was not feasible. Additionally, our MR analysis was constrained to assessing the impact of exposure (when measured) on the lifelong risk of the outcome. Besides, the assessment of the diverse impacts of exposure on outcomes across various stages of life was inconclusive. Statistical power also merits consideration. For example, some proteins were excluded from the MR analysis as they lacked available instrumental variables owing to the statistical power of corresponding GWASs. The limited sample size of the outcome datasets may have constrained the ability to identify mild associations between proteins and pancreatic disease. As a consequence, it is possible that certain associated proteins with potential overlap may not have been identified. Expanding the sample sizes for exposure and outcome could facilitate the identification of additional protein-disease associations. For example, future studies could further utilize more comprehensive pancreatic cancer genetic databases including PanScan and PanC4 to validate and expand our findings. In addition, we primarily used cis pQTLs as instrumental variables to infer relationships between proteins and pancreatic diseases in this study. However, future research could consider employing trans pQTLs as alternative instrumental variables to capture broader genetic regulatory effects. Trans pQTLs are typically located in distal regions of the target gene and may reveal more complex gene–protein interaction networks. Ultimately, we have identified four circulating proteins that mediate the effects of obesity indicators and pancreatic diseases. However, further investigation is necessary to elucidate the molecular mechanism underlying these circulating proteins.

Conclusion

We performed proteome-wide MR analyses, colocalization analyses, and mediation analyses in this study and identified circulating plasma CTRB2, RSPO3, KLOTB, and ROR1 as significant mediators linking obesity to pancreatic diseases. Notably, some existing drugs targeting these proteins had fewer side effects. The results obtained from our study provide illumination on the potential mechanisms that underlie the correlation between obesity and pancreatic diseases, thereby presenting valuable insights for the identification of novel potential drug targets.

Supplementary data

Supplementary data is available at Gastroenterology Report online.

Authors’ contributions

All authors were responsible for conception and were involved in manuscript writing. C.Z., X.R., T.C., Y.Z., S.Y., X.L., J.Z., X.C., contributed to the methodology. C.Z., X.R., T.C., conducted formal analysis. C.Z., X.R., contributed to the data curation. J.C., X.W., and D.Z., contributed to funding acquisition. All authors read and approved the final version of the manuscript.

Funding

C.Z. is supported by National Natural Science Foundation of China [No. 82270667]. X.W. is supported by National Natural Science Foundation of China [Nos U23A20492 and 8217033803]. D.Z. is supported by National Natural Science Foundation of China [No. 82170559]. J.C. is supported by the Natural Science Fund for Excellent Young Scholars of Hunan Province [No. 2025JJ40083], and the Natural Science Foundation of Changsha [No. kq2502174].

Conflicts of interest

The authors declared no conflict of interest.

Data availability

All data analysed are publicly available.