Systematic analysis of gene expression in human brains before and after death

Abstract

Background: Numerous studies have employed microarray techniques to study changes in gene expression in connection with human disease, aging and evolution. The vast majority of human samples available for research are obtained from deceased individuals. This raises questions about how well gene expression patterns in such samples reflect those of living individuals.

Results: Here, we compare gene expression patterns in two human brain regions in postmortem samples and in material collected during surgical intervention. We find that death induces significant expression changes in more than 10% of all expressed genes. These changes are non-randomly distributed with respect to their function. Moreover, we observe similar expression changes due to death in two distinct brain regions. Consequently, the pattern of gene expression differences between the two brain regions is largely unaffected by death, although the magnitude of differences is reduced by 50% in postmortem samples. Furthermore, death-induced changes do not contribute significantly to gene expression variation among postmortem human brain samples.

Conclusion: We conclude that postmortem human brain samples are suitable for investigating gene expression patterns in humans, but that caution is warranted in interpreting results for individual genes.

Figure 1

ANOVA test results. Numbers indicate number of probe sets with expression significantly influenced by brain region, source of sample material, and their interaction. The interaction term is significant when the expression changes due to death differ significantly in the two brain regions examined (see Material and methods). Numbers in brackets indicate the percentage of significant probe sets compared to the total number included in the analysis. Overlapping regions include probe sets with more than one significant term.

Figure 2

Scatter plot of expression differences between cortex and hippocampus in resection (x-axis) and autopsy (y-axis) samples. Expression differences were calculated as base two logarithm transformed ratios of gene expression values. All probe sets showing significant differences in expression levels between the two brain regions, either in the autopsy or in resection samples, are plotted: (a) according to Student's t test; (b) according to SAM. Red dashed lines represent linear regression results and black dotted lines represent expected regression lines with the slope = 1.