Inflammatory protein signatures as predictors of insulin resistance and cognitive function: a one-year longitudinal study in the mPRIME cohort

Large-scale studies have demonstrated a temporal relationship between insulin resistance (IR) and peripheral inflammation, with emerging evidence also suggesting a link to cognitive decline. However, most of the existing research has focused on individual inflammatory proteins, while the role of inflammatory protein signatures associated with IR and cognition remains unexplored. In this study, we investigated a) whether specific inflammatory protein signatures are associated with IR; b) whether baseline inflammatory protein signatures predict changes in Homeostatic Model Assessment for IR (HOMA-IR) after one-year follow-up, and c) whether baseline IR and/or specific inflammatory signatures predicted cognitive performance at follow-up. In 110 healthy adults from the mPRIME study, 45 inflammatory markers (measured by OLINK Proximity Extension Assay), HOMA-IR (as an index of IR), and cognitive function (as measured with an extensive neuropsychological test battery) were assessed at baseline and one-year follow-up. Inflammatory protein signatures were defined using hierarchical cluster analysis. Regression and cross-lagged analyses examined relationships between inflammation, HOMA-IR and cognitive changes over time. Next to the cumulative score, two inflammatory protein signatures were identified, related to 1) cytokine signalling, 2) immune cell migration. At baseline, the cytokine signalling signature correlated with HOMA-IR; together with BMI, CCL7 and CCL4 were consistently selected in >99% of models, with the penalised regression signature explaining 38,2% of HOMA-IR variance (mean out of sample R2). Cross-lagged analyses showed that baseline HOMA-IR predicted immune cell migration and IL-6 levels at follow-up, while CRP predicted future HOMA-IR. Cognitive performance (global cognitive performance, working memory z scores but not executive functioning) were nominally associated with the cytokine signalling signature and baseline HOMA-IR but did not survive FDR correction. No consistent association was found for inflammation, HOMA-IR and the one-year change in cognitive performance. Our findings contribute to a better understanding of the specific inflammatory proteins involved in the temporal association between IR and inflammation: while CRP predicts HOMA-IR, the more specific signature linked to cytokine-signalling did not; instead, a high HOMA-IR predicted higher inflammation related to immune cell migration. We discovered proteins including CCL4, CCL7 and IL-10 associated with HOMA-IR, which should be explored further.