DOI: 10.59518/farabimedj.2003448 ISSN: 2979-9821

Cross-Cohort Analysis Suggests Directional Elevation of the Neuronal Insulin-Signaling Transcriptional Program in Alzheimer's Disease

Duygu Aygunes Jafarı
Show PDF Cite
Objective: Reduced cerebral glucose metabolism is a defining feature of Alzheimer's disease (AD), and brain insulin resistance is widely invoked but defined at the level of receptor signaling, not transcript abundance. Whether the donor-level transcriptional signature of that axis rises, falls, or is unchanged in AD was examined using a cross-cohort, cell-type-resolved design with a pre-specified external test. Material and Methods: Four a-priori metabolic gene signatures (glucose delivery, astrocytic support, neuronal utilization, and insulin signaling; 123 genes fixed before any expression data were examined) were scored per nucleus within their own cell-type compartment and averaged to one value per donor in four single-nucleus RNA-seq cohorts (57 donor-level units). Effects were pooled by random-effects meta-analysis of Hedges' g (AD versus control) with the modified Hartung–K-napp estimator, complemented by a donor-label permutation test; pathway specificity was tested against 1000 size- and expression-matched random gene sets, and direction was examined at single-gene resolution and in one pre-specified external cohort. Results: The insulin signature showed a positive effect estimate in all four cohorts (per-cohort Hedges' g +1.11, +0.92, +0.29, +0.56). The primary few-studies estimator gave pooled g = +0.73 (95% CI −0.10 to +1.56, p = .067); the donor-label permutation p, a secondary inference, was .006, and the direction was unchanged in every leave-one-cohort-out analysis. The signature exceeded matched-random gene sets in the largest discovery cohort (empirical p = .001) and in the second-largest cohort (p = .019), but not in the two smallest (p = .55 and .28; Fisher combined p = .001). The external cohort reproduced the direction but not specificity. Neuronal utilization showed a negative estimate in all four cohorts (g = −0.49, 95% CI −1.29 to +0.32, p = .15), led by Complex I subunits. Conclusion: Across four cohorts the neuronal insulin-signaling transcriptional program shows a directionally consistent positive trend rather than the suppression a transcript-level reading of brain insulin resistance might assume. The trend is not statistically established under the primary estimator and is hypothesis-generating; because transcript abundance does not report phosphorylation, it is compatible with compensatory transcriptional feedback and marks a target for proteomic follow-up.

More from our Archive