Sex differences in the genetic regulation of the human plasma proteome

Abstract

Mechanisms underlying sex differences in the development and prognosis of many diseases remain largely elusive. Here, we systematically investigated sex differences in the genetic regulation of plasma proteome (>5800 protein targets) across two cohorts (30,307 females; 26,058 males). Plasma levels of two-thirds of protein targets differ significantly by sex. In contrast, genetic effects on protein targets are remarkably similar across sexes, with only 103 sex-differential protein quantitative loci (sd-pQTLs; for 2.9% and 0.3% of protein targets from antibody- and aptamer-based platforms, respectively). A third of those show evidence of sexual discordance, i.e., effects observed in one sex only (n = 30) or opposite effect directions (n = 1 for CDH15). Phenome-wide analyses of 365 outcomes in UK Biobank did not provide evidence that the identified sd-pQTLs accounted for sex-differential disease risk. Our results demonstrate similarities in the genetic regulation of protein levels by sex with important implications for genetically-guided drug target discovery and validation.

Fig. 1. Sex differences in the abundance of 5823 unique proteins measured by 4979 unique aptamers and 2923 unique antibody assays.

Linear regression models were used to test the association of sex with the protein abundance in each cohort. The protein targets were ordered by their effect size in males. Top panel: The top panel shows the proteins for which the plasma abundance significantly differed by sex in at least one technology (p_het < 1.01 × 10⁻⁵ for aptamer-based and p_het < 1.71 × 10⁻⁵ for antibody-based technology were used as Bonferroni-corrected thresholds respectively). The proteins were coloured blue if they had significantly higher levels in males and red if they had higher levels in females. If the protein target was significant in both of the technologies, the effect size estimate from the more significant study was displayed. The dark grey vertical lines represent the 95% confidence intervals for the effect size estimates. Bottom panel: The bars in the bottom panel represent the proteins which were targeted by both aptamer-based and antibody-based platforms. The lines were coloured lighter green if the finding was significant and directionally consistent in both technologies, yellow if the finding was significant but not directionally consistent across technologies, lilac if the finding was only significant in one of the technologies and black if the finding was not significant in any of the technologies. Results can be found in Supplementary Data 2.

Fig. 2. Forest plot of all identified sex-differential protein quantitative trait loci (sd-pQTLs) from both aptamer- (p_het < 1.01 × 10⁻¹¹) and antibody-based (p_het < 1.71 × 10⁻¹¹) technologies.

The bottom panel presents sex-discordant pQTLs (not significant (p < 5 × 10⁻⁸) in one sex or has opposing effect directions), whereas the top panel presents the remaining sex-differential pQTLs. The significant (p < 5 × 10⁻⁸) pQTLs in each sex are represented by filled circles and non-significant ones are represented by hollow circles. Linear regression models were used to identify pQTLs in each sex in each cohort. Horizontal lines represent 95% confidence interval for the effect size estimate of each finding. The sample sizes and detailed summary statistics for each of the findings can be found in Supplementary Data 3 and 4. Proteins with an asterisk (*) were measured using the aptamer-based technology, otherwise using antibody-based technology. MAF minor allele frequency.

Fig. 3. Phenotypic follow up of the identified sd-pQTLs with 365 disease outcomes with more than 2500 cases in UK Biobank.

Logistic regression models were used to calculate associations between genetic variants and disease outcomes in each sex. A Miami plot of association of 100 unique variants driving 103 sex-differential pQTLs (sd-pQTLs) with 365 disease outcomes among females on the top and among males at the bottom panel. x-axis contains each of the sd-pQTL—disease outcome pairs, ordered by their phecodes within each disease category. The associations have been coloured by disease categories. The horizontal dashed line represents a suggestive significance threshold of (p < 1 × 10⁻⁵). B Comparison of odds ratios for the sd-pQTL—disease associations which meet the suggestive significance threshold (p < 1 × 10⁻⁵) in males or females. The diagonal dashed line represents the equality line (x = y). C Comparison of −log₁₀transformed P values for the sd-pQTL—disease associations which meet the suggestive significance threshold (p < 1 × 10⁻⁵) in males or females. The diagonal dashed line represents the equality line (x = y).