Renal function dominates the landscape of the metabolome in type 1 diabetes

Background and aims: Kidney disease (KD) is a major burden in type 1 diabetes (T1D). Although problems in kidney function directly affect blood levels, it is still
unclear, which metabolite pathways are perturbed, and whether these perturbations bear prognostic potential. We did a data-driven analysis of associations in metabolites with the goal of, both, predicting the development of diabetic complications as well as to uncovering the underlying molecular biology.
Materials and methods: Plasma levels of 75 metabolites were analysed in a cross-sectional study of 586 persons with T1D. A machine learning method, the
graphical LASSO, was used in an unbiased way to mine for associations in the mixed set of metabolite measurements and complete clinical variables shown in
the Figure.
Results: A dense co-regulation network between metabolites and clinical variables was discovered (Figure). Metabolites grouped into four main clusters: amino acids (top-right), free fatty acids (top-left), glucose metabolism (bottom-right) and sugars derivatives (bottom-left). The highest number of associations between metabolites and clinical variables were in eGFR, which was associated with 11 metabolites. Particularly, sugar derivatives myo-inositol, ribitol and ribonic acid were inversely associated with eGFR but were independent of HbA1c and other clinical variables. In addition, the balance of ketone bodies was disturbed, and amino acids valine and isoleucine were associated with eGFR.
The medians (quartiles) of eGFR, AER and duration of T1D were 83.7 (63.1; 101.4) ml/min/1.73 m2, 17.0 (8.0; 62.3) mg/24 h and 35.5 (25.2; 44.6) years,
respectively. 25 % and 30 % had a history of micro- and macro-albuminuria, respectively. The entire network of associations is shown in the Figure as follows: Nodes are measured variables and edges (lines) are inferred associations (width: strength; red: positive, blue: inverse). Circular and rectangular nodes, respectively, are metabolites and clinical variables. Node size is the degree of the node (number of associations with the node). The highest-degree clinical node is highlighted in yellow and other nodes are colored by the respective Spearman r to the yellow node.
Conclusion: Pathway analysis of a large T1D cohort revealed a dense fabric of metabolites associating with clinical variables. Remarkably, the highest number of metabolite associations were to eGFR and not to other clinical variables that are known to influence major metabolite pathways. This finding indicates that KD leads to drastic changes in blood metabolite levels. Particularly, sugar derivatives and ketone bodies were perturbed in kidney disease. Blood levels of certain sugar derivatives are regulated by the kidney, suggesting potential for their use as an early biomarker of progression of KD.