Purine Metabolites and Carnitine Biosynthesis Intermediates Are Biomarkers for Incident Type 2 Diabetes

Abstract

Context: Metabolomics has the potential to generate biomarkers that can facilitate understanding relevant pathways in the pathophysiology of type 2 diabetes (T2DM).

Methods: Nontargeted metabolomics was performed, via liquid chromatography-mass spectrometry, in a discovery case-cohort study from the Malmö Preventive Project (MPP), which consisted of 698 metabolically healthy participants, of whom 202 developed T2DM within a follow-up time of 6.3 years. Metabolites that were significantly associated with T2DM were replicated in the population-based Malmö Diet and Cancer-Cardiovascular Cohort (MDC-CC) (N = 3423), of whom 402 participants developed T2DM within a follow-up time of 18.2 years.

Results: Using nontargeted metabolomics, we observed alterations in nine metabolite classes to be related to incident T2DM, including 11 identified metabolites. N2,N2-dimethylguanosine (DMGU) (OR = 1.94; P = 4.9e-10; 95% CI, 1.57 to 2.39) was the metabolite most strongly associated with an increased risk, and beta-carotene (OR = 0.60; P = 1.8e-4; 95% CI, 0.45 to 0.78) was the metabolite most strongly associated with a decreased risk. Identified T2DM-associated metabolites were replicated in MDC-CC. Four metabolites were significantly associated with incident T2DM in both the MPP and the replication cohort MDC-CC, after adjustments for traditional diabetes risk factors. These included associations between three metabolites, DMGU, 7-methylguanine (7MG), and 3-hydroxytrimethyllysine (HTML), and incident T2DM.

Conclusions: We used nontargeted metabolomics in two Swedish prospective cohorts comprising >4000 study participants and identified independent, replicable associations between three metabolites, DMGU, 7MG, and HTML, and future risk of T2DM. These findings warrant additional studies to investigate a potential functional connection between these metabolites and the onset of T2DM.

Figure 1.

Metabolite features from different metabolite classes associated with incident T2DM in the MPP (N = 698). logOR is the 10 log of the OR calculated from logistic regression models, and –logp is the negative 10 log of the P value calculated from the logistic regression models. Metabolite features with an annotation confidence at level 4 are marked as unknown. Metabolite features with annotation confidence of 2 or 3 are colored according to their metabolite class, and metabolites with annotation confidence 1 are additionally named in the figure.

Figure 2.

T2DM-associated metabolites belong to several metabolite classes. Metabolites that are defined as being at a level 4 metabolite annotation confidence are characterized as unknown (N = 42). Metabolites with an annotation confidence of ≥3 are characterized according to their metabolite class (N = 36).

Figure 3.

Correlation matrix of intermetabolite correlations in the MPP (N = 698). Correlations are Spearman ρ coefficients.

Figure 4.

Partial Spearman correlations between annotated T2DM-associated metabolites and BMI in the MPP (N = 698). Positive correlations are marked in blue and negative correlations in red.

Figure 5.

Partial Spearman correlations between annotated T2DM-associated metabolites and fasting glucose levels in the MPP (N = 698). Positive correlations are marked in blue and negative correlations in red.