Multi-Omics Mendelian Randomization Maps Lipid-Related Brain Cell-Type Associations with Kidney Disease

Abstract

Background: Cellular heterogeneity limits causal inference in complex diseases. We applied an established cell-type-stratified Mendelian randomization (csMR) framework, originally developed by Hao et al. (2024), to investigate whether lipid phenotypes affect kidney disease through distinct brain cell populations, extended by a proteome-wide mediation component. Methods: Using Bayesian colocalization (posterior probability of hypothesis 4 [PPH4], ≥0.8) and multidimensional instrumental variable selection, we evaluated cell-type-stratified MR-supported associations of five lipid traits across ten brain cell or tissue strata with kidney disease risk. Instrumental variables were derived from published genome-wide association studies and human brain single-cell expression quantitative trait locus (eQTL) maps. An exploratory proteomic mediation analysis using the Difference Method was performed with UKB-PPP as the discovery resource and deCODE as the external validation resource, with mediation signals interpreted as hypothesis-generating. For the primary csMR analysis, multiple testing was corrected separately within each lipid trait using a Bonferroni threshold of 3.33 × 10−4, corresponding to 150 cell/tissue-by-outcome tests per lipid trait. Results: In csMR analyses, cholesterol-related lipids showed cell-type-stratified associations with kidney disease, most prominently in oligodendrocyte-related analyses (max β on the log-odds scale = 1.025). LDL-related associations with broad chronic glomerular disease were most evident in excitatory neuron-related analyses. Dual-cohort proteomic analysis prioritized plasma proteins, including SNAP29 and ICAM4, as candidate protein-associated signals for exploratory mediation analysis. These findings represent genetic colocalization and Mendelian randomization-supported associations, not experimentally established mechanisms, and should be interpreted as hypothesis-generating. Conclusions: This study provides a cell-type-resolved genetic and proteomic association map linking lipid traits, brain cell strata, and kidney disease outcomes, offering hypotheses for future experimental validation.