Background

Metabolic and vascular diseases commonly coexist. However, the pathophysiological mechanisms linking both diseases are not well understood. A possible link is the dysregulation of vasoactive peptides that could be implicated in both vascular and metabolic homeostasis [1], such as endothelin-1 (ET-1) and adrenomedullin (ADM).

ET-1, a 21-amino acid peptide primarily secreted by vascular endothelial cells, is a potent vasoconstrictor and pro-inflammatory peptide [2]. ET-1 has been implicated in the pathogenesis of several chronic diseases, including hypertension and chronic kidney disease [3]. In addition to its known effect on the vascular function, ET-1 also limits insulin actions in skeletal muscles and adipocytes leading to insulin resistance and impaired glucose tolerance [4,5,6]. ADM, a 52-amino acid peptide that belongs to the calcitonin gene–related peptide family, is a vasodilator. ADM is secreted by a variety of different cells, including vascular endothelial cells, smooth muscle cells, adventitial fibroblasts as well as adipocytes [7, 8]. In a later investigation, adipose tissue was suggested to be the major source of ADM [8]. ADM has several metabolic actions, including counteracting oxidative stress–induced insulin resistance [9, 10] and inhibition of insulin secretion from the pancreatic islets [11].

Measurements of circulating concentrations of ET-1 and ADM are very difficult because of the short half-life, the existence of binding proteins, and other technical difficulties. Therefore, assays have been developed to measure the inactive fragments of ET-1 and ADM as the surrogates, C-terminal-proendothelin-1 (CT-proET-1) and mid-regional-proadrenomedullin (MR-proADM), respectively, which are biologically stable and are correlated with the active peptides in equimolar concentrations [12, 13].

In epidemiological studies using a cross-sectional design, both CT-proET-1 and MR-proADM were positively associated with the metabolic syndrome, insulin resistance and prevalent type 2 diabetes [14,15,16]. Previously, using a prospective study design, we have also shown that higher circulating concentrations of CT-proET-1 and MR-proADM were associated with increased insulin resistance, suggesting that both vasoactive peptides could play a role in the pathogenesis of type 2 diabetes [17]. However, most of the existing prospective studies failed to provide evidence for an association of both CT-proET-1 and MR-proADM with incident type 2 diabetes [17,18,19,20]. So far, only two prospective studies reported a positive association between CT-proET-1 and incident type 2 diabetes [21, 22]. Thus, we aimed to examine the putative association of CT-proET-1 and MR-proADM with incident type 2 diabetes by performing a prospective cohort study with a larger sample size using data from the multinational Biomarkers for Cardiovascular Risk Assessment in Europe (BiomarCaRE) Consortium [23] in tandem with a two-sample Mendelian randomisation study using published data on genetic variants that are specific for CT-proET-1 or MR-proADM, to allow a more robust analysis.

Methods

Study population

BiomarCaRE is an EU-funded consortium that aims to determine the value of established and emerging biomarkers in improving risk estimation of cardiovascular disease. BiomarCaRE relies on the Monitoring of Trends and Determinants in Cardiovascular Diseases (MONICA) Risk Genetics Archiving and Monograph (MORGAM) Project [24], which includes harmonized data from a large number of population-based cohorts. All participating cohorts were approved by local ethical review boards and written informed consent was obtained from all study participants. The study was conducted according to the Declaration of Helsinki.

In the prospective cohort study, we included three BiomarCaRE population-based cohorts involving 12,006 participants initially without diabetes and cardiovascular diseases and with follow-up data on type 2 diabetes. The exclusion criteria are described in Additional file 1: Fig. S1. The participating cohorts were the FINRISK Study (Finland), the Prospective Epidemiological Study of Myocardial Infarction (PRIME) Belfast (UK), and the Cooperative Health Research in the Region of Augsburg Study (KORA) F4 (Germany). An overview of each participating cohort is provided in Additional file 1: Table S1. The following harmonized variables were available for each cohort: age, sex, body mass index (BMI), waist circumference, systolic and diastolic blood pressure, antihypertensive medication, smoking status, total and high-density lipoprotein (HDL) cholesterol and diabetes status.

Ascertainment of type 2 diabetes cases

We defined prevalent diabetes as a documented diagnosis of diabetes at baseline, either identified by record linkage or through self-report of the participants that were verified by medical chart review or through information obtained from the treating physician. Incident type 2 diabetes was defined as a new diagnosis of type 2 diabetes during follow-up, either identified by record linkage or through self-report of the participants initially without diabetes at baseline that were verified by medical record review or through information obtained from the treating physician. Details of the assessment of type 2 diabetes in each participating cohort are provided in Additional file 1: Table S1.

Laboratory measurements

Baseline concentrations of CT-proET-1 and MR-proADM were measured from plasma with immunoluminometric assay (BRAHMS/Thermo Fisher Scientific, Hennigsdorf, Berlin, Germany) on the BRAHMS KRYPTOR automated system. The data were measured centrally in the MORGAM/BiomarCaRE core laboratory for FINRISK and PRIME Belfast (in 2008) and locally for KORA F4 (in 2010). The cohort-specific intra- and interassay coefficients of variation for CT-proET-1 and MR-proADM are described in Additional file 1: Table S2. Laboratory procedures for other diabetes-related biomarkers used in the analyses are provided in Additional file 1: Text S1.

Statistical analysis

Measurement values below the limit of detection (LOD) (N = 121 for CT-proET-1 and N = 166 for MR-proADM, all from the FINRISK study) were set to the lower LOD (i.e. 9.44 pmol/l for CT-proET-1 and 0.24 nmol/l for MR-proADM). Other missing values of the vasoactive peptides or missing values of diabetes risk factors (Additional file 1: Table S3) were handled with multiple imputation by chained equations (MICE), performed using R package mice [25], version 3.13. The imputation was done separately for each cohort. A total of 200 imputed data sets were created. Additional variation due to imputation was taken into account according to the Rubin’s rules for multiple imputation [26].

Descriptive statistics are reported for the participants stratified by incident type 2 diabetes during follow-up and shown as frequency (percentage) for categorical variables and as mean (standard deviation (SD)) for continuous variables. Continuous variables with skewed distributions are presented as geometric mean (antilog SD).

The associations of both CT-proET-1 and MR-proADM with incident type 2 diabetes were estimated by calculating hazard ratios (HRs) with 95% confidence intervals (95% CIs) in Cox proportional hazard (PH) models. The models were stratified by study cohort and were adjusted for age (continuous, in years) and sex (men/women) in model 1 and were further adjusted for current smoking (yes/no), total and HDL cholesterol (continuous, in mmol/l), actual hypertension (yes/no) and BMI (continuous, in kg/m2) in model 2. Actual hypertension was defined as having systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg or using antihypertensive medication. The distribution of MR-proADM was right-skewed (Additional File 1: Fig. S2) and thus was log-transformed to approximate normality. Both peptides were (0,1)-standardized to estimate the HRs per 1-SD increase. To further evaluate whether other diabetes-related biomarkers might account for the observed associations, we additionally included the baseline measurement of estimated glomerular filtration rate (eGFR), insulin, high-sensitivity C-reactive protein (hsCRP), leptin, and fasting glucose individually and simultaneously in model 2. The PH assumption was tested by plotting scaled Schoenfeld residuals against follow-up time for each covariate. No indication of non-proportionality was observed.

We tested for interactions of both peptides with BMI, sex and actual hypertension by creating cross-product terms and evaluating the significance level. Additionally, we also tested for the interaction with waist circumference as an alternative to BMI. False discovery rate (FDR) with the Benjamini–Hochberg method was used to correct for multiple testing. An interaction was considered relevant at FDR < 0.05. Subgroup analyses were conducted by examining the associations across BMI (≥ 30 kg/m2 vs < 30 kg/m2), waist circumference (men: ≥ 102 cm, women: ≥ 88 cm vs men: < 102, women: < 88 cm), sex (men vs women) and actual hypertension (yes vs no) categories. We also calculated the associations of CT-proET-1 and MR-proADM with incident type 2 diabetes for each participating cohort. Heterogeneity in the association across cohorts were examined by testing the interaction by study cohort and by examining Cochran’s Q and I2.

To examine the associations of genetically predicted CT-proET-1 and MR-proADM with type 2 diabetes risk, we performed two-sample univariate Mendelian randomisation analyses using results from published genome-wide association (GWA) studies. We identified single nucleotide polymorphisms (SNPs) with effects specific to either CT-proET-1 or MR-proADM at a P-value < 5E-8 as the genetic instrumental variables (IVs) from a published GWA study of European ancestry from Verweij, et al. [27]. Estimates of the genetic association with type 2 diabetes were extracted from meta-analyses of GWA studies for populations of European ancestry by Mahajan et al. [28] and Bonàs-Guarch et al. [29], depending on the data availability. The procedure for the Mendelian randomisation analysis is provided in detail in Additional file 1: Text S2.

All statistical analyses were performed using R version 4.0.3 [30]. P-values less than 0.05 were considered statistically significant.

Results

Baseline characteristics of study participants

Baseline characteristics of participants according to incident type 2 diabetes status during follow-up are summarized in Table 1. During a median follow-up time of 13.8 years (interquartile range of 4.8), 862 out of 12,006 participants developed type 2 diabetes. Participants who developed type 2 diabetes were more frequently men. At baseline, in comparison to non-cases, the cases of incident type 2 diabetes were on average older, had higher concentrations of CT-proET-1 and MR-proADM, had a higher BMI and waist circumference, were more frequently hypertensive, had lower eGFR, had lower concentrations of HDL cholesterol and higher concentrations of total cholesterol, hsCRP, insulin and leptin. Participant characteristics for each participating cohort are presented in Additional file 1: Table S4.

Table 1 Participant characteristics in the total study population and stratified by incident type 2 diabetes status
Full size table

Associations of CT-proET-1 and MR-proADM with incident type 2 diabetes

Both CT-proET-1 and MR-proADM were positively associated with incident type 2 diabetes in the overall study population. The HRs [95% CIs] in model 1 were 1.30 [1.21; 1.39] per 1-SD increase of CT-proET-1 and 1.57 [1.45; 1.69] per 1-SD increase of log MR-proADM. The associations were attenuated, but remained statistically significant after additional adjustment for diabetes risk factors according to model 2 (HR [95% CI]: 1.10 [1.03; 1.18] per 1-SD increase of CT-proET-1 and 1.11 [1.02; 1.21] per 1-SD increase of log MR-proADM). The association for CT-proET-1 remained stable when we further adjusted for eGFR, insulin, hsCRP, leptin, and fasting glucose (Table 2 and Additional File 1: Table S5). However, it was no longer significant for MR-proADM when insulin, hsCRP or leptin were added to the model (Table 2). The associations of CT-proET-1 and MR-proADM with incident type 2 diabetes were also examined in each participating cohort with a negligible level of heterogeneity (Additional File 1: Figs. S3 and S4, respectively).

Table 2 Association of CT-proET-1 and MR-proADM with incident type 2 diabetes
Full size table

We observed significant interactions of MR-proADM with BMI and waist circumference with respect to the association with incident type 2 diabetes (FDR < 0.05) (Table 3). When stratified by BMI, the positive association between MR-proADM and incident type 2 diabetes was only significant in obese participants. The HRs [95%CIs] per 1-SD increase of log MR-proADM were 1.19 [1.05; 1.34] in obese and 1.02 [0.90; 1.15] in non-obese participants. The results were similar when we stratified by waist circumference (Table 3). In an analysis where we further adjusted for eGFR, insulin, hsCRP and leptin, the association between MR-proADM and incident type 2 diabetes was attenuated, but remained significant in obese participants (HRs [95% CIs] per 1-SD increase of log MR-proADM: 1.14 [1.01; 1.29] in obese and 1.00 [0.88; 1.13] in non-obese participants). No significant differences could be detected in the association between MR-proADM and incident type 2 diabetes across sex and hypertension categories (Table 3).

Table 3 Subgroup analysis of the association of CT-proET-1 and MR-proADM with incident type 2 diabetes
Full size table

For CT-proET-1, no relevant interactions with BMI, waist circumference, sex and hypertension were observed with respect to incident type 2 diabetes under FDR < 0.05 (Table 3). The distribution of CT-proET-1 and MR-proADM by subgroup are presented in Additional file 1: Fig. S5.

Mendelian randomisation analysis

We identified one SNP that is specific for CT-proET-1 in the EDN-1 gene (rs5370) and one SNP that is specific for MR-proADM in the ADM gene (rs2957692) and included them as the genetic IVs. The genetic associations with each vasoactive peptide and with type 2 diabetes were extracted from the previously mentioned GWA studies [27,28,29].

In line with the findings from the time-to-event analysis, our Mendelian randomisation analysis showed a significant positive association between genetically predicted CT-proET-1 and type 2 diabetes risk. The OR [95% CI] was 1.12 [1.03; 1.22]. Conversely, we did not observe a significant association between genetically predicted MR-proADM and type 2 diabetes risk. The OR [95%CI] for MR-proADM was 0.97 [0.74; 1.27]. Sensitivity analyses using the likelihood-based method yielded similar results (OR [95%CI]: 1.12 [1.02; 1.22] for CT-proET-1 and 0.96 [0.73; 1.27] for MR-proADM) (Table 4).

Table 4 Results for the two-sample Mendelian randomisation analysis
Full size table

Discussion

In the current study, we observed that higher concentrations of both CT-proET-1 and MR-proADM were significantly associated with a higher incidence of type 2 diabetes. This is the first study to demonstrate a positive association between MR-proADM and incident type 2 diabetes independently of classical diabetes risk factors and that the association was more apparent in obese than in non-obese individuals. Using Mendelian randomisation approaches, we added further evidence that genetically predicted CT-proET-1 was significantly associated with a higher risk of type 2 diabetes. No significant association between genetically predicted MR-proADM and type 2 diabetes risk was documented.

Our results corroborate the few existing prospective analyses reporting a positive association of CT-proET-1 with incident type 2 diabetes [21, 22] and added evidence for a similar association for MR-proADM, particularly in obese individuals. In a previous study using data from 7953 participants of the Prevention of Vascular and Renal End-stage Disease Cohort [18], the authors also reported a significant positive association between MR-proADM and incident type 2 diabetes in a model adjusted for age and sex. However, the association was no longer significant in a multivariable model adjusted for classical diabetes risk factors. Compared with the previous prospective studies, our study represents the largest population-based cohort study examining the association of CT-proET-1 and MR-proADM with incident type 2 diabetes.

Factors underlying the association between CT-proET-1 and incident type 2 diabetes are not well understood. Studies conducted thus far have demonstrated that overexpression of ET-1 directly limits insulin actions. In skeletal muscles, the activation of endothelin receptor type-A by ET-1 suppresses insulin-mediated Akt phosphorylation and reduces glucose uptake [5, 31]. ET-1 also disrupts insulin-regulated glucose transporter 4 translocation to the plasma membrane [32]. In adipocytes, ET-1 blocks free fatty acid uptake and induces lipolysis, resulting in increased free fatty acid concentrations [33, 34]. Moreover, the interplay between increased free fatty acids and impaired glucose uptake may further exacerbate the dysregulation of lipid metabolism and energy homeostasis in insulin-resistant states [33]. Conversely, the inhibition of ET-1 signalling improves insulin sensitivity [35, 36]. Altogether, these biological effects suggest that ET-1 promotes insulin resistance and impaired glucose tolerance and thereby increases the risk of type 2 diabetes.

Furthermore, ET-1 is a potent vasoconstrictor, which plays an important role in the pathogenesis of hypertension and chronic kidney disease [3, 37], both are known to be associated with type 2 diabetes. ET-1 signalling also has been linked to increased leptin production [38] and stimulates the secretion of pro-inflammatory cytokines known to be involved in the development of metabolic disorders [6, 39]. However, in the present study, the positive association between CT-proET-1 and incident type 2 diabetes remained stable after additional adjustment for eGFR, insulin, hsCRP, leptin, and fasting glucose suggesting other possible explanations.

With regard to ADM, the underlying mechanisms linking higher concentrations with an increased risk of type 2 diabetes seem to be less straightforward. Evidence from in vivo and in vitro studies suggest that ADM could counteract insulin resistance through its antioxidant effects and the inhibition of insulin secretion [9, 11]. The latter notion also implicates ADM in maintaining insulin homeostasis [11]. ADM also has anti-inflammatory actions [40]. In obesity, ADM expression is upregulated in adipocytes and circulating ADM concentrations are increased [10]. Evidence from previous epidemiological studies also suggest positive associations of ADM with BMI and waist circumference [41, 42]. Factors that upregulate ADM production in obesity are incompletely understood. In an experimental study using a euglycaemic-hyperinsulinemic clamp technique, acute hyperinsulinemia was demonstrated to induce circulating ADM concentrations in obese, but not in lean individuals [43]. This evidence could explain the more apparent association of MR-proADM with incident type 2 diabetes in obese than in non-obese individuals seen in the current study. Furthermore, oxidative stress, insulin resistance, low-grade inflammation and dyslipidaemia, conditions that are commonly found in obesity, were also associated with increased MR-proADM concentrations [16, 44]. The increased ADM release in adipocytes seems to be a compensatory action attempting to restrain insulin homeostasis rather than a causal factor of insulin resistance thus, type 2 diabetes. Of note, in our data, the association between MR-proADM and incident type 2 diabetes was attenuated when we further controlled for insulin, hsCRP and leptin. We also did not observe a significant association between genetically predicted MR-proADM and type 2 diabetes risk in our Mendelian randomisation analysis. However, a non-significant association is not evidence for no association. Further studies are needed to confirm our findings.

Our study has some limitations that should be considered. As only single measurements of CT-proET-1 and MR-proADM were available at baseline we could not take into account the intra-individual variation. This could have led to exposure misclassification and regression dilution bias. In the current study, the harmonized data on several known diabetes risk factors, such as physical activity, diet and family history of diabetes, were lacking, which could have led to some degree of residual confounding. Our study participants were predominantly of European descent, which means that further studies need to confirm our findings in other ethnic groups. Finally, due to a very limited number of genetic IVs, we were unable to perform more robust analyses for our Mendelian randomisation.

Our study also has several strengths including the prospective, population-based design and the long-term follow-up with a median of 13.8 years. The use of harmonized data from the population-based cohorts participating in the BiomarCaRE project allows us to include a large sample size. To our knowledge, our study represents the so far largest population-based cohort study examining the association of CT-proET-1 and MR-proADM with incident type 2 diabetes. Furthermore, standardized epidemiological and laboratory procedures based on individual level data also allow for the best possible data analyses, including thorough adjustments for different diabetes risk factors and subgroup analyses.

Conclusions

In conclusion, higher concentrations of CT-proET-1 and MR-proADM were associated with incident type 2 diabetes. However, the positive association between MR-proADM and incident type 2 diabetes seemed to be modified by body composition, with a more apparent association in obese than in non-obese individuals. Our Mendelian randomisation analysis further suggests a probable causal link between CT-proET-1 and type 2 diabetes. These findings raise the possibility that ET-1 might be implicated in the pathogenesis of type 2 diabetes. Future studies are warranted to examine the utility of both peptides in risk stratification of type 2 diabetes for a better health care decision and their potential as targets for antidiabetic therapy.