Impact of weight loss on cancer-related proteins in serum: results from a cluster randomised controlled trial of individuals with type 2 diabetes

Abstract

Background: Type 2 diabetes is associated with higher risk of several cancer types. However, the biological intermediates driving this relationship are not fully understood. As novel interventions for treating and managing type 2 diabetes become increasingly available, whether they also disrupt the pathways leading to increased cancer risk is currently unknown. We investigated the effect of a type 2 diabetes intervention, in the form of intentional weight loss, on circulating proteins associated with cancer risk to gain insight into potential mechanisms linking type 2 diabetes and adiposity with cancer development.

Methods: Fasting serum samples from participants with diabetes enrolled in the Diabetes Remission Clinical Trial (DiRECT) receiving the Counterweight-Plus weight-loss programme (intervention, N = 117, mean weight-loss 10 kg, 46% diabetes remission) or best-practice care by guidelines (control, N = 143, mean weight-loss 1 kg, 4% diabetes remission) were subject to proteomic analysis using the Olink Oncology-II platform (48% of participants were female; 52% male). To identify proteins which may be altered by the weight-loss intervention, the difference in protein levels between groups at baseline and 1 year was examined using linear regression. Mendelian randomization (MR) was performed to extend these results to evaluate cancer risk and elucidate possible biological mechanisms linking type 2 diabetes and cancer development. MR analyses were conducted using independent datasets, including large cancer meta-analyses, UK Biobank, and FinnGen, to estimate potential causal relationships between proteins modified during intentional weight loss and the risk of colorectal, breast, endometrial, gallbladder, liver, and pancreatic cancers.

Findings: Nine proteins were modified by the intervention: glycoprotein Nmb; furin; Wnt inhibitory factor 1; toll-like receptor 3; pancreatic prohormone; erb-b2 receptor tyrosine kinase 2; hepatocyte growth factor; endothelial cell specific molecule 1 and Ret proto-oncogene (Holm corrected P-value <0.05). Mendelian randomization analyses indicated a causal relationship between predicted circulating furin and glycoprotein Nmb on breast cancer risk (odds ratio (OR) = 0.81, 95% confidence interval (CI) = 0.67-0.99, P-value = 0.03; and OR = 0.88, 95% CI = 0.78-0.99, P-value = 0.04 respectively), though these results were not supported in sensitivity analyses examining violations of MR assumptions.

Interpretation: Intentional weight loss among individuals with recently diagnosed diabetes may modify levels of cancer-related proteins in serum. Further evaluation of the proteins identified in this analysis could reveal molecular pathways that mediate the effect of adiposity and type 2 diabetes on cancer risk.

Funding: The main sources of funding for this work were Diabetes UK, Cancer Research UK, World Cancer Research Fund, and Wellcome.

Keywords: Cancer; DiRECT; Diabetes; Mendelian randomization; Obesity; Weight loss.

Fig. 1

Comparison of serum protein levels at 12 months follow-up between intervention groups (n = 260). Betas reflect the mean difference between allocation groups (protein measures were standardised and normalised prior to analyses using rank-based inverse normal transformation) while adjusting for protein levels at baseline. Labelled proteins are those that pass a Holm corrected P = 0.05. Values greater than 0 indicate an increase in serum protein seen in the intervention arm and values less than 0 reflect a decrease in serum protein levels. GPNMB: Glycoprotein Nmb; FURIN: Furin; WIF1: WNT Inhibitory Factor 1; TLR3: Toll-like receptor 3; PPY: Pancreatic prohormone; ERBB2: Receptor tyrosine-protein kinase erbB-2; HGF: Hepatocyte growth factor; ESM-1: Endothelial cell-specific molecule 1; RET: Proto-oncogene tyrosine-protein kinase receptor Ret.

Fig. 2

Estimated effect of circulating proteins on cancer risk in Mendelian randomization analyses (odds ratio (OR) and 95% confidence interval; n = 98,715–260,832 depending on cancer). All analyses used the Wald ratio model. Where estimates are missing the genetic instruments for the exposure were not present in the outcome dataset and thus MR could not be performed. GPNMB: Glycoprotein Nmb; FURIN: Furin; WIF1: WNT Inhibitory Factor 1; TLR3: Toll-like receptor 3; PPY: Pancreatic prohormone; ERBB2: Receptor tyrosine-protein kinase erbB-2; HGF: Hepatocyte growth factor; ESM-1: Endothelial cell-specific molecule 1; RET: Proto-oncogene tyrosine-protein kinase receptor Ret; CI = confidence interval.