Extensive modulation of the circulating blood proteome by hormonal contraceptive use across two population studies

Abstract

Background: The study of circulating blood proteins in population cohorts offers new avenues to explore lifestyle-related and genetic influences describing and shaping human health.

Methods: Utilizing high-throughput mass spectrometry, we quantified 148 highly abundant proteins, functioning in the innate and adaptive immune system, coagulation and nutrient transport in 3632 blood plasma, and 500 serum samples from the CHRIS and BASE-II cross-sectional population studies, respectively. Through multiple regression analyses, we aimed to identify the main factors influencing the circulating proteome at population level.

Results: Many demographic covariates and common medications affect the concentration of high-abundant plasma proteins, but the most significant changes are linked to the use of hormonal contraceptives (HCU). HCU particularly alters amongst others the levels of Angiotensinogen and Transcortin. We robustly replicated these findings in the BASE-II cohort. Furthermore, our results indicate that combined hormonal contraceptives with ethinylestradiol have a stronger effect compared to bioidentical estrogens. Our analysis detects no lasting impact of hormonal contraceptives on the plasma proteome.

Conclusions: HCU is the dominant factor reshaping the high-abundant circulating blood proteome in two population studies. Given the high prevalence of HCU among young women, it is essential to account for this treatment in human proteome studies to avoid misinterpreting its impact as sex- or age-related effects. Although we did not investigate the influence of HCU-induced proteomic changes on human health, our data suggest that future studies on this topic are warranted.

Fig. 1. Principal Component Analysis (PCA) of the CHRIS plasma proteome data reveals that proteomes of young women cluster into two distinct groups, which are attributable to the use of hormonal contraceptives.

a Grouping of individuals on PC1 and PC2. Women (red) and men (blue) are distinguished by color. A distinct cluster of women separates from the main bulk of male and female participants, particularly along PC1, indicating the presence of a subgroup of female study participants with a strong characteristic plasma proteome. b PCA loadings on PC1 and PC2. Each arrow represents a protein, with its length and direction indicating the protein’s contribution to the respective principal component, highlighting the proteins most influential in driving variance. c Relationship between PC1 (x-axis) and participant’s age (y-axis). The cluster of women on PC1 is clearly enriched with young women below the age of 40.

Fig. 2. Associations of plasma proteins with sex, age and BMI in CHRIS participants, determined by multiple linear regression analysis.

a Sex and b age associations represented as volcano plots. Each data point represents one protein with data points shown in a red color indicating a significant association. Analysis was performed on data from n = 3,472 study participants. The coefficient (x-axis) represents the log2 difference in average abundance between women and men and in age over 10 years difference. Bonferroni-adjusted p-values (y-axis) indicate the significance of this difference. c Hierarchically clustered heatmap of coefficients for proteins found to be significantly associated with at least one BMI category. The values for the coefficients are color-coded with blue colors representing negative, orange to red positive associations. Columns 1-2, 3-2 and 4-2 contain the coefficients for the comparison of BMI category 1 (underweight; BMI < 18.5), BMI category 3 (overweight; 25 < = BMI < 30) and BMI category 4 (obese; BMI > = 30) to BMI category 2 (normal; 18.5 < = BMI < 25), respectively. Significant associations are indicated with an asterisk. d Overlap of significant protein associations with age, sex, BMI category 4 and hormonal contraceptive use (HCU), which was included in the linear models to adjust for this medication. A large overlap of significant associations is present between HCU and any other trait.

Fig. 3. Sensitivity analysis indicates that not accounting for hormonal contraceptive use may misattribute age and sex effects.

Shown are the coefficients from the linear model adjusting for hormonal contraceptive use (x-axis) against the coefficients from a linear model without that adjustment (y-axis) for Sex a, Age b and BMI (c, obese vs normal), respectively. The solid black line represents the identity line.

Fig. 4. Hormonal contraceptive use is strongly affecting the circulating blood proteome of women below the age of 40 in two independent cohorts.

Volcano plots illustrating the association between protein abundances and hormonal contraceptive use in the CHRIS a and the BASE-II d cohort. The sample sizes are n = 729 and n = 240, respectively. Each point represents one protein, red coloring indicates significant association. b and e Abundance of the protein angiotensinogen (AGT) in study participants taking hormonal contraceptives (HCU) and women and men that don’t in the CHRIS b or the BASE-II e study. Samples sizes for the 3 groups are 316, 1,623, and 1,533 in CHRIS and 91, 149, and 197 in BASE-II. c ROC (Receiver Operating Characteristics) curve demonstrating the high predictive power of AGT for HCU. f Comparison of effect sizes of all proteins for association with HCU in CHRIS and BASE-II; the solid black line represents the identity line. Correlation of data points: Spearman’s rho = 0.91.

Fig. 5. Impact of different types of combined oral contraceptives (COCs) on the plasma proteome in CHRIS.

COCs with ethinylestradiol induce more pronounced changes in protein abundances compared to those containing bioidentical estrogen, whereas those containing progesterone affect a different set of plasma proteins. a Effect sizes for COCs with ethinylestradiol (n = 248) against effect sizes for COCs with bioidentical estrogen (n = 17). b Effect sizes of COCs with ethinylestradiol against those for progesterone preparations (n = 6). Each point represents data from one protein. Solid black lines represent the linear regression fit to the data points with its slope and p-value shown in the top left corner.