Metabolic syndrome (MetS), a set of metabolic risk factors including obesity, dysglycemia, and dyslipidemia, is associated with increased colorectal cancer (CRC) risk. A putative biological mechanism is chronic, low-grade inflammation, both a feature of MetS and a CRC risk factor. However, excess body fat also induces a proinflammatory state and increases CRC risk. In order to explore the relationship between MetS, body size, inflammation, and CRC, we studied large panels of inflammatory and cancer biomarkers. We included 138 participants from the Västerbotten Intervention Programme with repeated sampling occasions, 10 years apart. Plasma samples were analyzed for 178 protein markers by proximity extension assay. To identify associations between plasma protein levels and MetS components, linear mixed models were fitted for each protein. Twelve proteins were associated with at least one MetS component, six of which were associated with MetS score. MetS alone was not related to any protein. Instead, BMI displayed by far the strongest associations with the biomarkers. One of the 12 MetS score-related proteins (FGF-21), also associated with BMI, was associated with an increased CRC risk (OR 1.71, 95% CI 1.19–2.47). We conclude that overweight and obesity, acting through both inflammation and other mechanisms, likely explain the MetS-CRC connection.
Metabolic syndrome (MetS) is associated with numerous adverse health outcomes, such as cardiovascular disease (CVD), type 2 diabetes mellitus (T2DM), and several types of cancer [
One hallmark of MetS is a state of chronic inflammation. Several previous studies have reported protein biomarkers associated with MetS [
We hypothesized that the connection between MetS and CRC is driven by inflammation, with body composition as an important component, and that inflammatory proteins associated with MetS would therefore also associate with CRC risk. Using a unique collection of repeated samples from the Västerbotten Intervention Programme (VIP) in northern Sweden, we analyzed large panels of inflammatory and cancer biomarkers in relation to MetS and its components. The MetS-related biomarkers identified were examined in relation to the risk of developing CRC.
All study participants were selected from the Västerbotten Intervention Programme (VIP), initiated in 1985 and still ongoing [
Samples for this study were originally selected as part of a prospective study of CRC biomarkers. All CRC cases had to have a verified CRC diagnosis within five years after the latest sampling (excluding samples collected within three months of diagnosis) and have at least two available blood samples in the biobank. All but one case set had samples collected ten years apart. We selected an equal number of control subjects, matched on age (±12 months), sex, and sampling dates (±12 months). Controls had to be cancer-free at the latest follow-up (Dec. 31, 2014). For both cases and controls, only samples collected after at least eight hours of fasting were included, and none of the samples had previously been thawed. After all inclusions and exclusions, the study included repeated samples from 69 prospective CRC cases and 69 matched controls, resulting in 276 samples analyzed.
The project was approved by the Regional Ethical Review Board of Umeå University, Sweden. All VIP participants provide a written informed consent before donating their samples for research purposes, and they retain the right to withdraw that consent at any time in the future.
MetS components were measured as body mass index (BMI), triglyceride levels, total cholesterol levels, mid-blood pressure (mean of systolic and diastolic blood pressure), and fasting glucose levels. The variables were scaled to mean 0 and standard deviation (SD) 1 (z-transformed) separately for sex and sampling occasion. Due to a skewed distribution, triglycerides were log transformed first. We calculated a composite MetS score by summing all scaled variables except total cholesterol, which could distort the score depending on the proportions of HDL and LDL/VLDL within the total cholesterol measurement. The MetS score was also scaled separately by sex and sampling occasion. As a sensitivity analysis, we defined a dichotomous MetS variable according to the International Diabetes Federation criteria of obesity (BMI ≥ 30 kg/m2) and at least two of elevated triglyceride levels (≥1.7 mmol/l or lipid-lowering medication), hypertension (SBT ≥ 130 mm Hg or DBT ≥ 85 mm Hg or antihypertension medication), and elevated fasting glucose levels (≥5.6 mmol/l or self-reported diabetes).
All 276 samples were analyzed simultaneously for 178 unique protein biomarkers on two predesigned Proseek Multiplex® immunoassay panels (Olink Proteomics, Uppsala, Sweden) related to inflammation and cancer (all proteins are listed in Supplementary Table
All computations were conducted in R v.3.3.1 (R Foundation for Statistical Computing, Vienna, Austria).
Associations between protein markers and MetS were determined by fitting linear mixed models for each protein using the lme function in the lmer R-package. The mixed models included participant as a random factor (random intercept) and MetS and covariates as fixed factors. Two models were fitted for each protein, one including MetS score and one including each individual MetS component. Other covariates adjusted for the models were CRC status (case, control), age (continuous), sex (male, female), physical activity (5-level scale from never to >3 times/week), smoking (non-, current, and ex-smoker), and level of education (elementary school, upper secondary school, and university). The contribution of MetS and its components to protein variance was tested by an analysis of variance approach using the anova.lme function. For evaluation of variation in protein levels within and between individuals, we calculated intraclass correlations (ICC), defined as the proportion of total variance due to variation between individuals, using the variance estimates from the mixed models. We also calculated variance explained by fixed factors (
We assessed MetS and lifestyle-adjusted associations between the MetS-associated proteins by calculating partial Spearman’s correlations on the estimated residuals from the mixed models using the core function on pairwise complete observations.
All
We also examined MetS and the MetS-associated proteins in relation to CRC risk using conditional logistic regression models stratified on the matched case sets. Odds ratios (ORs) were estimated per 1SD change in the MetS variables. To evaluate whether associations differed depending on the follow-up time from sampling to CRC diagnosis, we tested for an interaction between sampling time point, MetS, and the protein variables.
Characteristics of the participants at the first and second sampling occasion are presented in Table
(a) Characteristics of study participants, stratified by sampling occasion. (b) Characteristics of study participants, stratified by colorectal cancer (CRC) status at the second sampling occasion.
Characteristics | Sample 1 ( |
Sample 2 ( |
|
---|---|---|---|
Median (range) | Median (range) | ||
Age | 49.9 (30.0–52.4) | 59.9 (40.0–60.5) | — |
Height | 173 (157–195) | 172 (155–196) | <0.001 |
Weight | 76.5 (51–128) | 80.0 (51–139) | <0.001 |
Triglycerides, mmol/l | 1.0 (0.4-5.2) | 1.1 (0.5–3.8) | 0.98 |
Total cholesterol, mmol/l | 5.5 (3.5–9.5) | 5.5 (3.4–10.1) | 0.18 |
Diastolic bp | 79.0 (60.0–120.0) | 80.0 (60.0–110.0) | 0.12 |
Systolic bp | 120.0 (94.0–180.0) | 126.0 (90.0–179.0) | 0.04 |
Fasting plasma glucose, mmol/l | 5.4 (4.0–6.9) | 5.5 (1.0–13.1) | 0.003 |
BMI | 25.3 (18.8–41.3) | 26.0 (18.4–44.9) | <0.001 |
Categorized BMI ( |
— | — | 0.08 |
<18.5 (underweight) | 0 | 1 (1) | — |
18.5–24.9 (normal) | 62 (45) | 52 (38) | — |
25–29.9 (overweight) | 60 (43) | 56 (41) | — |
30+ (obese) | 16 (16) | 29 (21) | — |
MetS (yes, %) | 12 (9) | 24 (17) | 0.05 |
aNumber of samples.
bPaired
Characteristics | Case ( |
Control ( |
|
---|---|---|---|
Median (range) | Median (range) | ||
Age (at sampling occasion) | 59.9 (40.3–60.4) | 59.9 (40.0–60.5) | — |
Height | 172.5 (156–196) | 172 (155–191) | 0.99 |
Weight | 81.0 (56–114) | 78.0 (51–139) | 0.95 |
Triglycerides, mmol/l | 1.1 (0.6–3.5) | 1.1 (0.5–3.8) | 0.99 |
Total cholesterol, mmol/l | 5.4 (3.8–7.6) | 5.6 (3.4–10.1) | 0.03 |
Diastolic bp | 82.0 (62.0–105.0) | 80.0 (60.0–110.0) | 0.14 |
Systolic bp | 126.0 (90.0–163.0) | 126.0 (90.0–179.0) | 0.80 |
Fasting plasma glucose, mmol/l | 5.7 (1.0–11.1) | 5.5 (4.1–13.1) | 0.73 |
BMI | 26.1 (21.0–36.0) | 26.0 (18.4–44.9) | 0.82 |
Categorized BMI ( |
— | — | 0.12 |
<18.5 (underweight) | 0 | 1 (1) | — |
18.5–24.9 (normal) | 22 (32) | 30 (43) | — |
25–29.9 (overweight) | 34 (50) | 22 (32) | — |
30+ (obese) | 12 (18) | 16 (23) | — |
MetS score | −0.02 (−0.60–0.54) | −0.06 (−0.80–0.51) | 0.26 |
MetS (yes, %) | 9 (13) | 15 (22) | — |
All data refer to characteristics at the second sampling as depicted in Table
bPaired
Out of 160 proteins that passed quality control, six were associated with the MetS score: three (TNFSF14, HGF, and FGF-21) with MetS score and BMI, two (SCF and ERBB2) with MetS score and triglyceride levels, and one (Furin) with MetS score, BMI, and triglyceride levels. An additional five proteins were associated with BMI alone (TNFSF10, SEZ6L, IL-6, FGF-BP1, and ESM-1), and one (OPG) was associated with total cholesterol (Figure
Significant associations between metabolic syndrome (MetS) and its components and each protein. Connections illustrate significant contributions to protein variance (Bonferroni corrected
List of proteins significantly associated with MetS score or one of its components.
Protein name | UniProt number | Associated with | Direction of association | Original Olink panel |
---|---|---|---|---|
IL-6 | P05231 | BMI | Positive | Inflammation panel and oncology panel |
TNFSF10 | P50591 | BMI | Positive | Inflammation panel and oncology panel |
SEZ6L | Q9BYH1 | BMI | Negative | Oncology panel |
FGF-BP1 | Q14512 | BMI | Negative | Oncology panel |
ESM-1 | Q9NQ30 | BMI | Negative | Oncology panel |
FGF-21 | Q9NSA1 | MetS and BMI | Positive | Inflammation panel |
TNFSF14 | O43557 | MetS and BMI | Positive | Inflammation panel |
HGF | P14210 | MetS and BMI | Positive | Inflammation panel and oncology panel |
Furin | P09958 | MetS, BMI, and triglycerides | Positive | Oncology panel |
ERBB2 | P04626 | MetS and triglycerides | Positive | Oncology panel |
SCF | P21583 | MetS and triglycerides | Negative | Inflammation panel and oncology panel |
OPG | O00300 | Total cholesterol | Positive | Inflammation panel |
Volcano plots for metabolic syndrome (MetS) (a) and BMI (b). The dashed line indicates the Bonferroni-adjusted significance threshold. Coefficients are interpreted as SD change in protein levels by 1SD change in MetS score and BMI, respectively.
The proportion of variance in protein levels explained by the fixed factors in our models,
Contribution to variance explained by metabolic syndrome (MetS) and other included covariates. (a) Variance explained by including MetS (
Furin and HGF showed the strongest associations with MetS. The relations were driven largely by BMI, with 20–30 percentage points of the variation explained by the fixed factors attributed to BMI (Figure
Partial correlations between the MetS-associated proteins are presented in Supplementary Figure S4. Almost all correlations were positive. Two clusters of more correlated proteins were present: cluster 1: SCF, OPG, ERBB2, SEZ6L, ESM-1, and FGF-BP1; cluster 2: HGF, TNFSF14, Furin, IL-6, FGF-21, and TNFSF10.
None of the MetS components was significantly associated with CRC risk in conditional logistic regression models adjusting for age, sex, and sampling date by case-set stratification, and additionally by smoking status and educational level by regression (Figure
Associations between metabolic syndrome (MetS) score, MetS components, protein levels, and CRC risk. All odds ratios (ORs) were calculated using conditional logistic regression, stratified for the case-sets. For MetS, odds ratios (ORs) were adjusted for smoking status and education level. For protein models, ORs are adjusted for smoking status and education level. ORs adjusted are additionally adjusted for the MetS score. MetS score OR and MetS score P are ORs and corresponding
Metabolic syndrome is becoming increasingly common, and many studies indicate a direct association between MetS and the risk of developing CRC and other forms of cancer, likely driven, at least in part, by body composition and inflammation [
Interestingly, five of the 12 proteins identified were associated with BMI only, all of which were included in the predefined oncology protein marker panel due to a potential relation to cancer (with or without an inflammatory connection). However, of the six proteins associated with MetS score, there was an equal distribution between inflammatory and cancer proteins. Thus, body composition likely contributes to cancer development not only through chronic inflammation but also through other pathways.
Of the MetS-associated proteins, only one, which was also associated with BMI, was positively associated with CRC risk, FGF-21 (fibroblast growth factor 21). Inclusion of MetS strengthened the association between FGF-21 and CRC risk and attenuated the association between MetS and CRC risk, suggesting a mediating effect. Increases in BMI and MetS score contributed to a significant amount of the FGF-21 protein level variation, and it was the protein with the largest change in level per MetS score increase. Consistent with our observations, the associations between FGF-21, BMI, and MetS have been previously described [
Two proteins were strongly associated with both MetS and BMI, namely, HGF (hepatocyte growth factor) and Furin. Both of these proteins have been previously implicated in MetS [
The protein most strongly associated with CRC risk was ESM-1 (endothelial cell-specific molecule-1), which was inversely associated with both BMI and CRC in our dataset but not associated with MetS. ESM-1 has previously been shown to be overexpressed in CRC patients and associated with poor prognosis [
One other protein, SEZ6L (seizure 6-like protein), was also inversely associated with both BMI and CRC. It controls synaptic connectivity and motor coordination and is also a substrate for the
Weaknesses of our study include lack of a central obesity measurement (substituted with BMI) and HDL cholesterol measurements. Total cholesterol was evaluated in relation to protein measurements but not included in the MetS score definition because of the conflicting roles of HDL and LDL, both of which contribute to total cholesterol. In addition, CRP (c-reactive protein), an established and important marker of inflammation previously connected to MetS [
Major strengths of the study included the use of high-quality blood samples collected prospectively with respect to CRC diagnosis, with all participants fasting for at least eight hours prior to sampling, and with no previous thaw-freeze cycles. The VIP cohort also provided a unique opportunity to use repeated samples from both cases and time-matched controls, allowing us to account for intra-individual variation. Finally, an important strength of the study was the large number of protein biomarkers evaluated simultaneously using a highly sensitive platform.
In our study MetS does not, in itself, appear to contribute to the inflammation cancer-connection. MetS score was associated with six different proteins in our investigation. However, all were also associated with BMI and/or triglyceride levels. Of the individual MetS components assessed, BMI displayed by far the strongest associations with inflammatory and cancer biomarkers. Although external replication is needed, our data indicate that the relationship between MetS and CRC risk is likely driven primarily by excess body fat, acting through both pro-inflammation and other pro-carcinogenic mechanisms.
The authors declare that they have no conflicts of interest.
Sophia Harlid and Robin Myte contributed equally to this work.
This study was funded by the Lion’s Cancer Research Foundation, Umeå University; the Cancer Research Fund in Northern Sweden; the Swedish Society of Medicine; the Swedish Cancer Society; the Young Scientist and other research grants from the County Council of Västerbotten, Sweden, through the regional agreement between Umeå University and Västerbotten County Council in cooperation in the field of medicine, odontology, and health; and the Faculty of Medicine at Umeå University, Umeå, Sweden.