PerspectiveHuman Genomics

Searching for sex differences

See allHide authors and affiliations

Science  11 Sep 2020:
Vol. 369, Issue 6509, pp. 1298-1299
DOI: 10.1126/science.abd8340

The behemoth effort, started a decade ago, known as the Genotype-Tissue Expression (GTEx) Consortium aims to discover how DNA variation affects gene expression across human tissues (1, 2). As part of this consortium, on page 1331 of this issue, Oliva et al. (3) find that more than one-third of genes show sex-biased expression in at least one tissue. Four other GTEx studies, on pages 1318, 1334, 1333, and 1332 of this issue, respectively, discuss the effects of gene regulation in human tissues (4), identify functional rare genetic variation (5), study predictors of telomere length (6), and report cell type–specific gene regulation (7). What is especially notable about Oliva et al. is the careful analysis, which revealed that in addition to reported genetic and hormonal effects (8), there are cell type–specific sex differences in tissue composition. Furthermore, their work highlights that rather than being strictly dimorphic, interindividual variation results in overlapping distributions of gene expression between the sexes.

It has been hypothesized that selection shaped sex differences in immune function in response to the evolution of pregnancy and the placenta in mammals, beginning more than 90 million years ago and contributing to the observed sex differences in diseases today, including a female bias in autoimmune disease and male bias in most cancers (9). Sex differences in gene expression are broadly shared across mammals, but their role in shaping sex differences in disease etiology has not been thoroughly explored. Oliva et al. report that genes that show differences between sexes are enriched for multiple pathways, including in immune responses and cancer. Furthermore, they identify sex differences in a cluster of genes that target histone H3 lysine 27 trimethylation (H3K27me3) sites; these histone marks have also been reported to show sex-differentiated expression in the placenta (10). Oliva et al. provide a comprehensive baseline for sex differences in gene expression in unaffected tissues that can be used for future comparisons with diseased tissues. These observations may also inform about which pathways are most important in sex differences in disease etiology and aid in the development of targeted therapies.

Sex differences in gene expression vary across the genome and between individuals, as represented by these heatmaps.

(GRAPHIC) V. ALTOUNIAN/SCIENCE AND M. ATAROD/SCIENCE; (DATA) GTEX CONSORTIUM

Oliva et al. identified hundreds to thousands of genes (1.3 to 12.9% of the genes expressed per tissue) that show sex differences in gene expression in any given tissue but found that the effect for each individual gene is subtle (the median fold change in expression was just 1.04). This is after accounting for the cellular composition of tissues that came from males versus females. The authors hypothesize that this sex difference in cell type composition—particularly of immune-related cells, such as monocytes and neutrophils [previously reported by (11)]—may contribute to the underlying sex-specific dysregulation for some diseases. These sex differences in cell type composition affect estimates of gene expression and, if unaccounted for, can skew results that compare groups with unequal sex ratios. Future studies that include samples from males and females will now need to account for cell type composition in addition to sex chromosome complement and hormonal environment. This is because different sex ratios in cell type in cases versus controls may drive the gene expression signal more than the phenotype of interest.

Although the genes with the highest fold change in expression were found on the X chromosome, the X chromosome contains only 4% of genes with sex-differential expression; the remaining 96% are spread across the genome (3). This is important because the X chromosome is often excluded from genome-wide analyses (12), but in doing so, studies may be missing genes with the largest effects. Additionally, Oliva et al. call attention to the importance of autosomal (non–sex chromosome) gene regulation in contributing to sex differences in humans. Given this, it is also noteworthy that they show that sex-biased autosomal gene expression is not very specific for predicting the sex of the donor from which the sample was taken (84% accurate, with 56% specificity), emphasizing how labile sex-biased gene expression is across people.

The GTEx Consortium has generated an invaluable resource through the generous involvement of patients and their families. However, like many consortia, sampling biases hinder investigation of interindividual variation. Details about the sampling are, with much appreciation, made transparent by the consortium on the GTEx Portal (13). Just over two-thirds (67.1%) of the samples are from males. This means that studies are unevenly statistically powered to detect sex differences. It also means that studies that use the GTEx data as a reference set for comparison with disease state, for example, should take into account the relative proportion of samples from males and females (in both sets) because the relative sex effects on gene expression may not be the same in both.

Also, more than half of the samples come from people 50 years and older. This means that the samples are skewed toward understanding gene expression in tissues that have had many different exposures, potentially contributing to the observed interindividual variation, and does not reflect expression of tissues across the life span. Considering variation across the life span is especially critical for understanding how puberty and menopause, for example, affect gene regulation between the sexes.

Last, representation of global human genetic variation is low, with nearly 85% of samples collected from white people of European descent. There is a dearth of information about genetic variation and gene expression outside of a narrow range of recent genetic ancestries (14). This is critical for human health because inferences about genetic risk from one group of people with recent shared ancestry often do not generalize to others (15).

Given these limitations of the samples, it is even more surprising—and should be motivating to human geneticists—how much interindividual variation is observed in gene expression among the people included in the GTEx Consortium. This should be a call to projects to expand the representation of human variation in future studies.

References and Notes

Acknowledgments: The author is supported by the National Institute of General Medical Sciences of NIH (R35GM124827).
View Abstract

Stay Connected to Science

Navigate This Article