Scripties UMCG - Rijksuniversiteit Groningen
 
English | Nederlands

The influence of cell type proportions on analysis of age-induced differential expression with gene expression derived from whole blood

(2018) Stutvoet, M. (Maartje)

Introduction: Changes in variable gene expression underlie ageing processes. However, the heterogeneity of whole blood and shifts in blood cell composition with age could influence gene expression-age associations found with the current models. Here, we aim to assess this influence and the efficacy of a model that more precisely corrects for blood cell composition.
Methods: We performed meta-analyses in 3165 human peripheral blood samples from four independent Dutch cohorts using two distinct linear mixed models. The Basic Model, representative for previous studies, corrected for overarching white blood cell (WBC) types. On top of that, the Correction Model corrected for 33 WBC subtypes. The significantly differentially expressed genes were further analysed with functional enrichment analyses and cell type specific gene expression.
Results: After Basic Model correction, the correlation between residual gene expression and WBC subtypes proved that gene expression-age associations found in whole blood cannot be seen independently from changes in the proportions of WBCs with age. Gene clusters found by the models were in general similar, while specific pathways varied. Opposed expression changes in different WBC subsets caused simultaneous up- and downregulation of pathways with age. Cell specific gene expression levels did not show eminent blood cell composition influences in the Basic Model. However, average gene expression in megakaryocytes (MKs) was obviously heightened in the Correction Model compared to the Basic Model.
Conclusion: In this study, we show that age, gene expression and blood cell proportions are interconnected. Our results stress the impact of tissue heterogeneity on gene expression measured, showed the difficulty in correcting for it and demonstrated its confusing effect on the interpretation of findings. Confounding by this heterogeneity should be considered for each trait studied associated with tissue composition, irrespective of the tissue gene expression was derived from.






 
To top