Multi-omics profiles of sex hormone-binding globulin are associated with subclinical atherosclerosis in men with HIV

Abstract

Background

Sex hormones and HIV infection both influence cardiovascular health. However, the association between sex hormones and subclinical atherosclerosis is not fully understood, especially in the context of HIV.

Methods

Among 321 men (65% with HIV) from the MACS/WIHS Combined Cohort Study, we measured 14 serum sex hormones and sex hormone-binding globulin (SHBG), assessed carotid artery plaque (IMT > 1.5 mm) using high-resolution B-mode ultrasound, and performed metagenomic sequencing on stool samples. In 312 men, we measured 986 plasma metabolites via liquid chromatography-tandem mass spectrometry and 2883 plasma proteins using the Olink Explore 3072 platform. In stratified analyses of men with (MWH) and without HIV (MWOH) and adjusting for covariates and multiple testing, we (1) examined associations of sex hormones with plaque; (2) characterized multi-omics profiles related to sex hormones; and (3) generated sex hormone-related omics scores via linear combination of related species, metabolites, and proteins, respectively, to explore whether these sex hormone-related multi-omics profiles were associated with plaque.

Results

Median age of participants was 62 years (interquartile range: 58–68), and 31.5% had carotid artery plaque. Sex hormones were differentially associated with plaque in MWH and MWOH. In MWH, an inverse association was observed between SHBG and plaque (OR = 0.60 per 1-SD increase, 95% CI: 0.41, 0.90). Furthermore, higher SHBG levels were associated with overall gut microbial composition, lower abundance of species from genera Prevotella, Fibrobacter and Coprococcus, higher levels of certain metabolites (primarily lipid and carnitine metabolites) and proteins enriched in the cell–cell adhesion pathway. Some SHBG-related species (e.g., Mediterranea massiliensis), metabolites (e.g., phosphatidylcholine-based lipids) and proteins (e.g., enriched in immune response pathway) were also associated with plaque in MWH. All three SHBG-related omics scores were inter-correlated and inversely associated with plaque in MWH. In MWOH, estrone-sulfate was positively associated with plaque (OR = 3.80, 95% CI: 1.41, 10.22) but not with any species, metabolites or proteins.

Conclusions

Higher SHBG, and related microbial species, circulating metabolites, and proteins, were inversely associated with carotid artery plaque. These findings suggested that SHBG may play a protective role in subclinical atherosclerosis in MWH.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13073-026-01709-8.

Background

With the extended lifespan of people living with HIV due to widespread effective antiretroviral therapy (ART), cardiovascular disease (CVD) has emerged as a leading cause of mortality [1]. People living with HIV face a higher risk of atherosclerosis and other cardiovascular complications, primarily driven by persistent immune activation and chronic inflammation [2]. Although the overall lifetime risk of CVD is similar for women and men in the general population [3, 4], men tend to develop CVD at an earlier age [5], whereas women have higher mortality rates and worse prognoses after acute cardiovascular events [6], suggesting significant sex differences in the development of CVD.

Sex hormones, especially estrogens and androgens, have been considered to account for some of the observed sex differences in CVD. For example, estrogens have a positive impact on lipoprotein profiles, contributing to the cardiovascular benefit experienced by premenopausal women [7]; and both androgens and estrogens regulate endothelial function and are involved in the development of CVD in a sex-specific fashion [8, 9]. Sex hormone-binding globulin (SHBG) is an important glycoprotein that binds sex hormones in the circulation, regulating the availability of free sex hormones to hormone-responsive tissues and influencing their metabolic clearance rates [10, 11]. Several studies have shown that higher SHBG levels are associated with a more favorable cardiometabolic risk profile, including lower blood pressure, triglycerides, blood glucose, and body mass index (BMI), and higher high-density lipoprotein cholesterol (HDL-C), regardless of total testosterone level [12]. These observations suggest an independent and beneficial role of SHBG in cardiovascular health [13, 14].

Subclinical atherosclerosis is associated with an increased lifetime risk of CVD [15] and all-cause mortality [16, 17], with carotid artery plaque presenting as a marker of subclinical atherosclerosis. Notably, carotid atherosclerosis is more prevalent in men than women [18]. Growing evidence demonstrates that sex hormones are associated with atherosclerosis and related subclinical markers in both men and women [19–21], but findings are inconclusive. In addition, HIV infection affects both sex hormones [22] and cardiovascular health [1]. However, associations between sex hormones and subclinical atherosclerosis remain incompletely understood, particularly among people living with HIV. We recently showed that higher serum estrogens were associated with lower prevalence of carotid artery plaque in post-menopausal women with and without HIV [23]; while such studies on men remain scarce.

Multi-omics signatures across various biological layers, including the genome, transcriptome, proteome, metabolome and microbiome, may provide insights into sex differences in disease pathogenesis, ultimately contributing to distinct health phenotypes in women and men [24]. The metabolism of sex hormones has been linked to gut bacteria, circulating metabolites and proteins, which may play a role in cardiovascular health. Specifically, gut bacteria are involved in the reactivation and degradation of estrogens and androgens, thereby affecting their excretion and circulation [25]; androgens and SHBG have been shown to exert distinct effects on metabolite profiles in women and men, with greater impact of androgens on the plasma metabolome in women than in men, while greater impact of SHBG on the urine metabolome in men than in women [26]. Moreover, sex-specific protein expression patterns have been linked to pathways involved in sex hormone metabolism [27], while sex hormones can also modulate immune responses and cytokine activity [28]. However, it remains unclear whether sex hormone-related multi-omics signatures are associated with subclinical atherosclerosis.

In the present study, we aimed to examine the association between serum sex hormones, SHBG and subclinical atherosclerosis in men with and without HIV (MWH and MWOH) within the Multicenter AIDS Cohort Study (MACS)/Women’s Interagency HIV Study (WIHS) Combined Cohort Study (MWCCS). By integrating these data with stool shotgun metagenomic data and plasma metabolomic and proteomic data, we further sought to characterize sex hormone-related multi-omics profiles and explore their relation to subclinical atherosclerosis.

Methods

Study population

MACS, now part of the MWCCS, was a prospective multicenter cohort initiated in 1984. This longitudinal cohort study enrolled men with or at risk of HIV infection from four study centers in Baltimore/Washington DC, Chicago, Pittsburgh/Columbus, and Los Angeles. Detailed study designs of MACS and MWCCS have been published previously [29–31]. Briefly, participants attended semiannual core visits that included standardized interviews, clinical measurements, physical examinations, and biological specimen collections for laboratory tests. In this analysis, participants were selected from a substudy investigating sex differences in the role of multi-omics in HIV-associated carotid artery atherosclerosis. Frozen serum samples collected from 2018 to 2021 were thawed for measurement of sex hormones, metabolites and proteins. Stool samples collected with a home-based self-collection kit [32] from 2020 to 2021, were processed for shotgun metagenomic sequencing. The most recent carotid artery ultrasound scans available, conducted between 2012 and 2013, were used for assessing subclinical atherosclerosis. As shown in Fig. 1A, we included 349 men with available data on sex hormones, gut microbiome sequencing, and carotid artery ultrasound. We excluded 15 men with a history of prostate cancer and 13 men with low sequencing depth (< 100 K reads) in their stool samples, resulting in a sample size of 321 men (208 with HIV and 113 without HIV) eligible for the main analysis.

Fig. 1.

Flow charts of study participant selection, analytical samples, and analysis strategy. A We included 321 eligible men (208 HIV + and 113 HIV −) from the Multicenter AIDS Cohort Study (MACS), now part of the MACS/WIHS Combined Cohort Study (MWCCS), a longitudinal cohort of people living with and without HIV. Shotgun metagenomics sequencing of stool samples (n = 321), plasma metabolomics (n = 312), and proteomics (n = 312) were conducted in subsets of participants, resulting in a sample size of 312 men with multi-omics data. B The analytical strategy included three major steps to examine the associations of sex hormone features and multi-omics signatures with carotid artery plaque in the context of HIV

Sex hormones measurement

Fifteen sex steroid hormones and SHBG were quantified in serum samples using validated methods [33, 34]. The measured steroids belong to four major classes: adrenal precursors, androgens, estrogens, and progesterone (Additional file 2: Table S1). Adrenal precursors include androstenedione (4-dione), dehydroepiandrosterone (DHEA), DHEA-sulfate (DHEA-S), and androst-5-enediol (Delta-5diol). Androgens include testosterone, 5α-androstane-3α,17β-diol-3-glucuronide (3αdiol-3G), 5α-androstane-3α,17β-diol-17-glucuronide (3αdiol-17G), 5α-androstane-3β,17β-diol (3β-diol), androsterone (ADT), ADT-glucuronide (ADT-G), dihydrotestosterone (DHT). Estrogens include estrone (E1), estrone-sulfate (E1-S), estradiol (E2). Quantification was performed using gas chromatography-mass spectrometry (GC–MS) for all adrenal precursors, some androgens (i.e., testosterone, ADT, DHT, and 3β-diol), some estrogens (E1 and E2), and progesterone [33]. Liquid chromatography-tandem mass spectrometry (LC–MS/MS) was used for conjugated androgens (i.e., 3αdiol-3G, 3αdiol-17G, and ADT-G) and E1-S [34]; SHBG was quantified using an enzyme-linked immunosorbent assay (ELISA) kit. Detailed experimental procedures were described previously [23]. Since free and unbound steroid hormones are considered biologically active, we calculated free E2 and free testosterone concentrations based on measured E2, total testosterone, SHBG, and an assumed constant for albumin (43 g/L) [35, 36]. To reflect aromatase-mediated conversion of androgens to estrogen, we derived the ratios of E1 to 4-dione (E1:4-dione) and E2 to total testosterone (E2:testosterone). Additionally, extents of steroid conjugation with sulfate and glucuronide groups were assessed using the ratios of E1-S to E1 (E1-S:E1), DHEA-S to DHEA (DHEA-S:DHEA), and ADT-G to ADT (ADT-G:ADT). Progesterone was excluded from the analysis due to 50% of values falling below the lower limit of quantification (LLOQ). For each sex hormone feature (referring to measured or calculated free hormones, SHBG, and hormone ratios), values below the LLOQ were replaced with half the LLOQ. All sex hormone concentrations were transformed using rank-based inverse normal transformation (INT) for statistical analyses.

Microbiome measurement and bioinformatics processing

Shotgun metagenomic sequencing was performed at the Burk lab, Albert Einstein College of Medicine, following established and highly reproducible protocols to generate species-level taxonomic and functional data [23]. The sequencing process was blinded to lab technicians, ensuring they did not know participants' individual characteristics including HIV serostatus. Briefly, DNA was extracted from stool samples using the DNeasy PowerLyzer PowerSoil Kit (Qiagen, Hilden, Germany). After quantification, DNA then underwent library construction via the KAPA HyperPlus Kit (Roche, Basel, Switzerland) using xGen Stubby Adapters and 10nt Unique Dual Index (UDI) primers (IDT, Coralville, IA). Libraries were constructed with modifications to the manufacturer’s procedures, as adapted by the Knight Lab [37]. The final indexed libraries were purified, quantified, and pooled at approximately equal concentrations. To enhance compatibility with the sequencing platform, pooled libraries are further purified and concentrated using the QIAquick PCR Purification kit (Qiagen). Libraries were then sequenced on a NovaSeq6000 (Illumina, San Diego, CA) using 2 × 150 bp paired-end chemistry.

Raw FASTQ sequence reads were processed using the standard shotgun metagenomic sequencing pipeline in Qiita [38]. Briefly, we first trimmed the low-quality bases using prinseq [39], followed by quality control using KneadData, which wraps several software packages to: (1) trim overrepresented sequences with FASTQC, (2) trim adapters and low-quality bases with Trimmomatic (i.e. left trimming bases with phred < 25 and removing any resulting orphan reads if one of the mates become < 50 bp), (3) trim repetitive sequences with Tandem Repeats Finder, and (4) remove reads mapping to the human genome with Bowtie2 [40]. We aligned the reads against the WolR1 reference database of bacterial and archaeal genomes using Woltka with the Bowtie2 aligner [40]. This process generated an operational genomic unit (OGU) table and a gene table. The sequence alignments were classified at the species taxonomic level.

Metabolomics profiling

Metabolites were measured in plasma samples using LC–MS/MS at the Broad Institute Metabolomics Platform (Cambridge, MA), an untargeted analysis as described previously [41]. We included known metabolites with a detection rate > 80% and coefficients of variation (CVs) < 20%. Missing data were imputed using half the minimum detected value across all samples per metabolite. Metabolite levels were transformed using rank-based INT prior to statistical analysis. A total of 986 plasma metabolites (from 1029 metabolites) were retained for analysis.

Proteomics profiling

Proteins were measured in plasma samples using the Olink Explore 3072 platform, which consists of 8 panels targeting biomarkers related to oncology, cardiometabolic risk factors, inflammation, and neurology [42]. We included proteins that met Olink’s batch release quality control criteria. A total of 2883 plasma proteins (from 2926 unique proteins) were retained for analysis.

Subclinical atherosclerosis

High-resolution B-mode carotid artery ultrasound with automated computerized edge detection was performed using standardized methods across all study sites [43]. Sonographers at each site received uniform training at the University of Southern California Atherosclerosis Research Unit Core Imaging and Reading Center (CIRC) [44] to ensure consistency in image acquisition and interpretation. Carotid artery plaque represents a measurable manifestation of subclinical atherosclerosis. Our primary outcome was the presence of a carotid artery plaque or lesion (binary variable), defined as at least 1 plaque characterized by focal intima-media thickness > 1.5 mm [45] at any of the imaged segments in the right carotid artery.

Covariate measurement

Covariate data, including demographic, clinical, laboratory, and anthropometric variables were measured using standardized protocols at semi-annual core study visits. For this study, all the covariate data were obtained from the visit closest to and preceding the timing of sex hormone measurement. Missing values were imputed by using available data from the nearest prior visit.

Statistical analysis

General principles

The statistical analysis consisted of three major steps (Fig. 1B). First, we examined the associations between sex hormone features and carotid artery plaque in MWH and MWOH. Second, we characterized the multi-omics profiles (i.e., omics signatures and scores) of sex hormones features, incorporating the gut microbiome, blood metabolome and proteome. Third, we examined the interplay of sex hormones and identified gut microbial species, circulating metabolites, and proteins in relation to carotid artery plaque. To account for potential misclassification of our measure of subclinical atherosclerosis, sensitivity analyses were performed using a prediction-based approach (Additional file 1: Supplementary Methods) where the outcome variable was carotid artery plaque. Given the known relationship between circulating sex hormone concentrations and HIV infection status [22], analyses were performed separately within HIV-stratified groups. To account for multiple testing, we applied a false discovery rate (FDR)-adjusted significance threshold of q < 0.1 for single-omics signature analyses. Unless specified otherwise, p-value < 0.05 was considered significant. All analyses were implemented in R (version 4.2.2).

Covariate selection

A priori covariates included age, race/ethnicity, and HIV serostatus, while additional covariates were selected by univariate analyses, with candidate covariates as predictors and inverse-normal transformed sex hormone levels as outcomes [23]. Covariates associated with at least one sex hormone feature, with a likelihood ratio test FDR-q < 0.1 in either the entire study population or any HIV-stratified group, were included as adjustment covariates in the statistical models (Additional file 1: Fig. S1). Given the high correlation between BMI and waist circumference and their similar associations with hormones, only waist circumference was retained. All candidate covariates considered in our analysis are presented in Table 1. The final covariates included in the models were age, race/ethnicity, income, educational attainment, smoking, illicit drug use, Hepatitis C virus (HCV) serostatus, hypertension medication, lipid-lowering medication, albumin, estimated glomerular filtration rate (eGFR), waist circumference, glucose, hemoglobin A_1c, total cholesterol, HDL-C, triglycerides, testosterone use, probiotics use, HIV serostatus, and antiretroviral therapy regimen (for MWH only). eGFR was estimated by serum creatinine using the race-free CKD-EPI Eq. [46]. Right-skewed variables, including blood glucose, hemoglobin A_1c, and triglycerides, were log-transformed (base e) in statistical models.

Table 1.

Characteristics of men in the MWCCS by HIV serostatus (n = 321)

	All	Men with HIV	Men without HIV	pvaluea
n	321 (100%)	208 (65%)	113 (35%)
Age, y, median [IQR]	62.0 [58.0, 68.0]	61.0 [57.0, 66.0]	66.0 [60.0, 71.0]	< 0.001
Race/ethnicity, n (%)				0.028
Black/African-American non-Hispanic	73 (22.7)	49 (23.6)	24 (21.2)
White non-Hispanic or other	223 (69.5)	137 (65.9)	86 (76.1)
Hispanic	25 (7.8)	22 (10.6)	3 (2.7)
Annual income, n (%)				0.004
$30,000 or less	124 (38.6)	90 (43.3)	34 (30.1)
$30,001—$75,000	97 (30.2)	66 (31.7)	31 (27.4)
$75,001 or more	100 (31.2)	52 (25.0)	48 (42.5)
Educational attainment, n (%)				0.011
Less than high school	16 (5.0)	13 (6.2)	3 (2.7)
Completed high school	47 (14.6)	38 (18.3)	9 (8.0)
Any college	258 (80.4)	157 (75.5)	101 (89.4)
Smoking status, n (%)				0.182
Never smoker	108 (33.6)	67 (32.2)	41 (36.3)
Former smoker	166 (51.7)	105 (50.5)	61 (54.0)
Current smoker	47 (14.6)	36 (17.3)	11 (9.7)
Alcohol use, n (%)				0.078
1 to < 4 drinks/week	228 (71.0)	155 (74.5)	73 (64.6)
4 to 13 drinks/week	61 (19.0)	32 (15.4)	29 (25.7)
More than 13 drinks/week	32 (10.0)	21 (10.1)	11 (9.7)
Illicit drug use, n (%)	150 (46.7)	105 (50.5)	45 (39.8)	0.087
Hepatitis C virus positive serostatus, n (%)	36 (11.2)	30 (14.4)	6 (5.3)	0.022
Serum albumin, g/dL, median [IQR]	4.4 [4.2, 4.6]	4.4 [4.2, 4.6]	4.3 [4.2, 4.5]	0.518
Estimated GFR, mL/min/1.73 m2, median [IQR]	78.1 [67.3, 94.3]	74.4 [62.3, 90.1]	86.0 [75.0, 96.0]	< 0.001
Serum ALT, IU/L, median [IQR]	22 [17, 28]	22 [18, 29]	21 [16, 27]	0.055
Serum AST, IU/L, median [IQR]	22 [18, 26]	23 [19, 28]	20 [18, 24]	0.008
Waist circumference, cm, median [IQR]	99.3 [90.4, 109.7]	99.2 [90.4, 109.7]	100.5 [90.5, 109.6]	0.768
BMI, kg/m2, median [IQR]	26.7 [24.0, 29.7]	26.6 [23.8, 29.8]	27.2 [24.6, 29.6]	0.327
Fasting glucose, mg/dL, median [IQR]	94 [87, 104]	95 [88, 105]	94 [87, 101]	0.579
Hemoglobin A1c, %, median [IQR]	5.5 [5.2, 5.9]	5.5 [5.1, 5.9]	5.5 [5.3, 5.9]	0.274
Total cholesterol, mg/dL, median [IQR]	164 [136, 186]	162 [135, 185]	164 [138, 193]	0.194
HDL cholesterol, mg/dL, median [IQR]	49 [40, 59]	47 [40, 56]	53 [41, 64]	0.002
Triglycerides, mg/dL, median [IQR]	107 [76, 154]	116 [83, 168]	88 [65, 135]	< 0.001
Systolic blood pressure, mmHg, median [IQR]	126 [117, 138]	125 [115, 137]	126 [119, 139]	0.329
Diastolic blood pressure, mmHg, median [IQR]	76 [69, 83]	77 [69, 84]	74 [68, 80]	0.037
Use of hypertension medication, n (%)	142 (44.2)	89 (42.8)	53 (46.9)	0.554
Use of diabetes medication, n (%)	44 (13.7)	34 (16.3)	10 (8.8)	0.09
Use of lipid-lowering medication, n (%)	178 (55.5)	116 (55.8)	62 (54.9)	0.97
Detectable HIV viral loadb, n (%)	52 (25.0)	52 (25.0)	-	-
CD4+ T-cell countb, cells/mm3, median [IQR]	638 [457, 876]	638 [457, 876]	-	-
Antiretroviral therapy regimenb, n (%)
None	5 (2.4)	5 (2.4)	-	-
PI-basedc	15 (7.2)	15 (7.2)	-	-
INSTI-basedc	106 (51.0)	106 (51.0)	-	-
NNRTI-basedc	31 (14.9)	31 (14.9)	-	-
Other	51 (24.5)	51 (24.5)	-	-
Antibiotics used, n (%)	63 (19.6)	39 (18.8)	24 (21.2)	0.697
Probiotics used, n (%)	63 (19.6)	20 (9.6)	14 (12.4)	0.561
Testosterone use, n (%)	32 (10.0)	29 (13.9)	3 (2.7)	0.002

^aP-values comparing the difference of characteristic variables between men with and without HIV are derived from Wilcoxon rank-sum test for continuous variables and χ2 test with continuity correction or Fisher’s exact test (if the expected count < 5) for categorical variables

^bHIV-related characteristics (detectable HIV viral load, CD4⁺ T-cell count, and antiretroviral therapy regimen) are only summarized in men with HIV. The lower limit of detection for HIV viral load is 20 copies/mL

^cProtease inhibitor (PI)-based: at least 1 PI and 1 nucleoside reverse transcriptase inhibitor (NRTI); integrase strand transfer inhibitors (INSTI)-based: at least 1 II and 1 NRTI; nonnucleoside reverse transcriptase inhibitor (NNRTI)-based: at least 1 NNRTI and 1 NRTI

^dSelf-reported antibiotics and probiotics use in the past 6 months

ALT alanine transaminase, AST aspartate aminotransferase, BMI body mass index, GFR glomerular filtration rate, HDL high-density lipoprotein, IQR interquartile range, MWCCS Multicenter AIDS Cohort Study/Women’s Interagency HIV Study Combined Cohort Study

Sex hormone features and carotid artery plaque

Multivariable logistic regression analyses were used to evaluate the associations between sex hormone features and carotid artery plaque separately in MWH and MWOH, adjusting for above-mentioned covariates (referred to as the “main analysis” in our result). Since carotid artery scans were performed prior to blood sample collection, sensitivity analyses were conducted to evaluate consistency of results with the main analysis. Briefly, we first established a prediction model for incident carotid artery plaque using longitudinal data from MACS participants. This model was then applied to estimate the probability of developing incident plaque at time of sex hormone measurement for men without plaque in our study, based on which the plaque status was generated. For the sensitivity analyses, logistic regression analyses were performed excluding those participants with predicted plaque. The process was repeated 100 times and the averaged effect estimates were used to estimate the association of each sex hormone feature with carotid artery plaque (see Additional file 1: Supplementary Methods for more details).

Sex hormone features and omics profiles

Gut microbiome

Overall microbiome diversity and composition. For α-diversity (within-sample variation), we used multivariable linear regression to examine the association of sex hormone features (predictors) with α-diversity (outcome), measured by observed species richness and Shannon diversity index (“phyloseq” package, R). For β-diversity (between-sample variation), we used permutational multivariate analysis of variance (PERMANOVA) to evaluate the association of sex hormone features (predictors) with overall microbial composition (outcome), measured by Jensen-Shannon Divergence (JSD) and generalized UniFrac distance (“phyloseq” and “GUniFrac” packages, R). All models were adjusted for multiple covariates as described in “Covariate selection”.

Microbial species

We included species with relative abundance > 0.01% present in at least 20% of samples, and then used Analysis of Compositions of Microbiomes with Bias Correction (ANCOM-BC, “ANCOMBC” package, R) [47] to identify species associated with each sex hormone feature, adjusting for covariates. FDR-q < 0.1 was considered statistically significant. A positive log fold change indicated an increase in species abundance with increasing hormone concentration, while a negative value indicated a decrease.

Sex hormone-related gut microbiome scores

Following strategies used in previous microbiome studies [48, 49] to represent the overall associations of each hormone with gut microbiome, and to relate with other data (i.e., carotid artery plaque, metabolomics, and proteomics data) and reduce dimensionality, hormone-specific microbiome scores were generated for those hormones associated with more than five species (FDR-q < 0.1). Centered log-ratio (clr)-transformed species abundances were standardized using Z-score, and hormone-specific microbiome scores were calculated by summing species with positive associations and subtracting species with negative associations (FDR-q < 0.1). Association direction was determined by the log fold change in ANCOM-BC. Given potential differences in hormone-associated species between MWH and MWOH, microbiome scores were created separately for each group.

Metabolomics and proteomics

We used multivariable linear regression to examine the association of sex hormone features (predictors) with plasma metabolites and proteins (outcomes). Similar to previous metabolomics and proteomics studies [50–52], hormone-specific metabolite or protein scores were created to represent the overall associations of each hormone with blood metabolome or proteome, and to relate with other data (e.g., carotid artery plaque and gut microbiome data) and reduce dimensionality. The scores were calculated, in MWH and MWOH, by summing inverse-normal transformed metabolites or proteins on a log₂ scale with positive associations and subtracting those with negative associations (FDR-q < 0.1). Association direction was determined by coefficient in multivariable linear regression.

Associations between sex hormone—related omics profiles and carotid artery plaque

For sex hormone features associated with carotid artery plaque in MWH or MWOH, multivariable logistic regression analyses were used to investigate the associations between hormone-specific omics scores and carotid artery plaque separately in MWH and MWOH, adjusting for multiple covariates (referred to as the “main analysis”). Sensitivity analyses (as described above) were conducted to evaluate consistency, with results considered consistent if the association was significant at p < 0.05 in the main analysis and p < 0.1 in sensitivity analyses with the same direction of effect. For hormone-specific omics scores associated with plaque in both main and sensitivity analysis, logistic regression was used to assess whether the individual species, metabolites, and proteins contributing to the score were associated with plaque in both main and sensitivity analysis. Spearman correlation was used to evaluate correlations between the omics signatures associated with both hormone and plaque. To gain insight into the biological functions of the identified proteins, we performed pathway enrichment analysis in Metascape (https://metascape.org).

Results

Participant characteristics

This study included 321 men (65% living with HIV), with a median age of 62 years (interquartile range: 58, 68). Compared to MWOH, MWH were younger and had lower income and educational attainment at the time of sex hormones measurement (Table 1). MWH also had a higher proportion of Hispanic participants and a lower proportion of non-Hispanic White or other participants compared with MWOH. MWH were more likely to be HCV seropositive, receive testosterone therapy, and exhibit cardiovascular risk profiles, characterized by higher levels of serum aspartate aminotransferase (AST), triglycerides, diastolic blood pressure, and lower eGFR and HDL-C levels (Table 1). Among MWH, 97.6% were receiving antiretroviral therapy and 75.0% had an undetectable viral load (< 20 copies/mL).

Sex hormone levels by HIV serostatus

Several sex hormone levels differed between MWH and MWOH. Specifically, MWH exhibited lower levels of DHEA, DHEA-S, E1, E2, and free E2, and higher levels of delta-5diol than MWOH (Table 2). These differences between HIV serostatus groups remained significant in multivariable linear regression models after adjusting for multiple potential covariates (Additional file 2: Table S2). Most sex hormones were positively correlated with each other in overall and HIV-stratified groups, whereas SHBG and certain sex hormone ratios (e.g., E2:T, E1:4-dione, ADT-G:ADT) were inversely correlated with other hormones (Additional file 1: Fig. S2).

Table 2.

Sex hormones concentrations of study participants by HIV serostatus (n = 321)^a

	Overall	Men with HIV	Men without HIV	p-valueb
n	321	208	113
Adrenal precursors
Androstenedione (4-dione) (ng/mL)	0.72 [0.56, 0.90]	0.72 [0.56, 0.87]	0.76 [0.56, 0.95]	0.412
DHEA-sulfate (DHEA-S) (ug/mL)	0.67 [0.39, 1.18]	0.64 [0.35, 1.13]	0.75 [0.46, 1.34]	0.031
Androst-5-enediol (Delta-5diol) (pg/mL)	552.78 [350.87, 805.08]	517.73 [301.78, 760.35]	634.29 [428.54, 881.94]	0.001
Dehydroepiandrosterone (DHEA) (ng/mL)	1.21 [0.82, 1.74]	1.12 [0.77, 1.61]	1.38 [0.93, 2.21]	< 0.001
DHEA-S:DHEA ratio	0.56 [0.39, 0.80]	0.57 [0.39, 0.85]	0.54 [0.41, 0.70]	0.44
Androgens
Testosterone (T) (ng/mL)	4.19 [3.13, 5.74]	4.09 [3.02, 5.64]	4.48 [3.38, 5.75]	0.109
Free T (nmol/L)c	0.24 [0.20, 0.31]	0.24 [0.19, 0.31]	0.25 [0.21, 0.32]	0.223
5α-androstane-3α,17β-diol-17-glucuronide (3αdiol-17G) (ng/ml)	3.79 [2.12, 5.84]	3.81 [1.96, 5.85]	3.74 [2.25, 5.61]	0.908
Androsterone (ADT) (pg/mL)	75.76 [25.00, 141.83]	78.26 [25.00, 142.33]	70.74 [25.00, 130.70]	0.542
5α-androstane-3α,17β-diol-3-glucuronide (3α-diol-3G) (ng/ml)	1.48 [0.94, 2.31]	1.54 [0.94, 2.45]	1.40 [0.96, 2.22]	0.622
Androsterone-glucuronide (ADT-G) (pg/mL)	30.20 [20.70, 47.00]	29.90 [18.90, 48.28]	30.30 [22.10, 42.70]	0.808
Dihydrotestosterone (DHT) (pg/mL)	317.12 [195.58, 466.09]	310.89 [183.52, 458.49]	325.43 [215.73, 481.47]	0.405
5α-androstane-3β,17β-diol (3β-diol) (pg/mL)	35.87 [10.00, 59.64]	39.14 [10.00, 65.94]	30.88 [10.00, 46.95]	0.057
ADT-G:ADT ratio	0.41 [0.18, 1.24]	0.39 [0.17, 1.24]	0.47 [0.21, 1.24]	0.449
Estrogens
Estradiol (E2) (pg/mL)	19.39 [15.69, 24.90]	18.59 [14.58, 23.96]	20.38 [17.79, 26.46]	0.011
Free E2 (nmol/L)c	1.68 [1.19, 2.28]	1.64 [1.12, 2.12]	1.75 [1.36, 2.45]	0.024
Estrone (E1) (pg/mL)	25.74 [18.33, 33.30]	24.05 [17.14, 32.20]	27.86 [21.08, 33.62]	0.009
Estrone-sulfate (E1-S) (ng/ml)	0.43 [0.27, 0.71]	0.40 [0.26, 0.71]	0.50 [0.31, 0.70]	0.102
E1-S:E1 ratio	0.02 [0.01, 0.03]	0.02 [0.01, 0.03]	0.02 [0.01, 0.03]	0.838
Other
Sex hormone-binding globulin (SHBG) (nmol/L)	43.59 [31.26, 56.55]	42.71 [28.49, 59.88]	45.18 [35.09, 52.59]	0.311
E1:4-dione ratio	35.34 [26.16, 46.53]	34.77 [24.08, 46.68]	37.39 [29.24, 44.53]	0.069
E2:T ratio	4.70 [3.71, 5.96]	4.70 [3.67, 5.88]	4.71 [3.80, 6.20]	0.579

^aThe concentrations of sex hormones are presented as median [interquartile range]

^bp-values comparing the difference of sex hormone concentrations between men with and without HIV are derived from Wilcoxon rank-sum test

^cFree estradiol (E2) and free testosterone concentrations were calculated based on measured E2, total testosterone, sex hormone-binding globulin, and an assumed constant for albumin (43 g/L)

Sex hormones and carotid artery plaque

Among MWH, we observed an inverse association between SHBG and carotid artery plaque (OR = 0.60, 95% CI: 0.41, 0.90), after adjusting for sociodemographic factors, lifestyle, clinical factors, and HIV-related variables. In MWOH, we observed a positive association between E1-S and plaque (OR = 3.80, 95% CI: 1.41, 10.22) (Fig. 2). When we examined SHBG in MWOH, and E1-S in MWH, their associations with carotid artery plaque were neither statistically significant nor directionally similar as the associations observed in the complementary HIV serostatus group. Interaction analysis revealed that HIV serostatus modified the association between E1-S and plaque (p-interaction < 0.05), with significant associations observed in MWOH. Since SHBG and E1-S were associated with carotid artery plaque in MWH and MWOH, respectively, with consistent associations across both main and sensitivity analysis (Fig. 2), subsequent multi-omics analysis focused on SHBG in MWH and E1-S in MWOH (atlas of omics signatures for other hormones are described in the Additional file 1: Supplementary Results).

Fig. 2.

Associations of sex hormone features with carotid artery plaque in men with (n = 208) and without HIV (n = 113). Among MWH, 69 men had plaque (33.2%); among MWOH, 32 men had plaque (28.3%). Main analysis was conducted using multivariable logistic regression with sex hormone features as predictors of carotid artery plaque, adjusting for age, race/ethnicity, income, educational attainment, smoking, illicit drug use, HCV serostatus, hypertension medication, lipid-lowering medication, albumin, estimated GFR, waist circumference, glucose, hemoglobin A_1c, total cholesterol, HDL cholesterol, triglycerides, testosterone use, probiotics use, and antiretroviral therapy regimen (for MWH only). Sensitivity analysis was performed by repeating the multivariable logistic regressions after excluding the predicted carotid artery plaque (100 repetitions), averaging the odds ratio, and estimating the total variance over the repeated analyses, with the same adjustment covariates listed above (details provided in the Supplementary Methods). * Sex hormone marked with an asterisk (*) indicate a significant interaction by HIV serostatus for the association between sex hormone and carotid artery plaque (p < 0.05). GFR, glomerular filtration rate; HDL, high-density lipoprotein; MWH, men with HIV; MWOH, men without HIV

Multi-omics profiles of SHBG in MWH

SHBG and gut microbiome. We characterized the association of SHBG with gut microbiome features in MWH. SHBG was not associated with gut microbiome α-diversity, as measured by observed species richness and Shannon index, after adjusting for covariates (Additional file 1: Fig. S3A). SHBG was associated with overall gut microbiome composition in PERMANOVA models, as measured by generalized Uni-Frac (R² = 0.58%, P = 0.013), but not Jensen-Shannon divergence (R² = 0.34%, P = 0.295), after adjusting for covariates (Additional file 1: Fig. S3B). At the microbial species level, of the 368 species examined, 18 species from 5 taxonomic classes were associated with SHBG in MWH (FDR-q < 0.1), using ANCOM-BC with covariates adjustment (Fig. 3A; Additional file 2: Table S3). Specifically, SHBG was associated with a lower abundance of 17 species primarily from genera Prevotella, Fibrobacter and Coprococcus, and a higher abundance of one species from genus Oscillibacter.

Fig. 3.

Multi-omics signatures associated with serum SHBG in MWH. A. Phylogenetic tree includes 18 SHBG-related microbial species from 5 taxonomic classes among MWH (n = 208). ANCOM-BC was used to identify species (present in ≥ 20% of samples) associated with SHBG, adjusting for the same covariates as in Fig. 2 for MWH. Node sizes are proportional to mean relative abundance of microbial species in MWH. Heatmap shows the log fold change from ANCOM-BC for each microbial species, indicating the natural logarithm of the fold change in the relative abundance of species for every one-unit increase in inverse-normal transformed SHBG levels. FDR-q < 0.1 is considered statistical significance in ANCOM-BC. B Volcano plot shows the association between SHBG and plasma metabolites in MWH (n = 202). Coefficients and FDR-q values were derived from multivariable linear regression models, adjusting for same covariates as in panel A. Metabolites with FDR-q < 0.1 are colored based on their Super Pathway categories, while only those with FDR-q < 0.01 are labeled by name to preserve legibility. C Similarly, the volcano plot shows the association between SHBG and plasma proteins in MWH (n = 202). Coefficients and FDR-q values were derived from multivariable linear regression models, adjusting for same covariates as in panel A. Proteins with FDR-q < 0.1 are colored based on their Olink panels, while only those with FDR-q < 0.01 are labeled by name to preserve legibility. ANCOM-BC, Analysis of Compositions of Microbiomes with Bias Correction; FDR, false discovery rate; MWH, men with HIV; SHBG, sex hormone-binding globulin

SHBG and plasma metabolites. In MWH, of the 986 metabolites examined, 102 metabolites were significantly associated with SHBG (FDR-q < 0.1), after adjusting for covariates (Fig. 3B; Additional file 2: Table S3). Specifically, SHBG was associated with higher levels of 77 metabolites (e.g., glycerophospholipids and carnitines) and lower levels of 25 metabolites (e.g., glycerolipids, sphingolipids, branched chain amino acids, and aromatic amino acids).

SHBG and plasma proteins. In MWH, of the 2882 plasma proteins examined (SHBG was excluded from the protein panel), 186 proteins were significantly associated with SHBG (FDR-q < 0.1), after adjusting for covariates (Fig. 3C; Additional file 2: Table S3). Specifically, SHBG was associated with higher levels of 177 proteins (e.g., cardiometabolic markers [IGFBP1, IGFBP2, CKB, ICAM-1, and VCAM-1]) and lower levels of 9 proteins (e.g., cardiometabolic markers [CNDP1 and F9] and inflammatory markers [ORM1, C8B, and ERBB3]). Pathway enrichment analysis of protein-coding genes (i.e., 186 SHBG-related proteins) revealed the top three significant enrichment pathways were ‘cell–cell adhesion’ (involving 39 target genes), ‘cell adhesion molecules’ (involving 36 target genes), and ‘cell–cell adhesion via plasma-membrane adhesion molecules’ (involving 18 target genes) (FDR-q < 0.001).

SHBG remained significantly associated with 9 microbial species, 37 plasma metabolites, and 43 proteins at FDR-q < 0.05 (Table S3).

SHBG, related omics scores and HIV-related factors

Using the SHBG-related microbiome species, metabolites, and proteins in MWH, we generated SHBG-related omics scores in MWH, and in MWOH for comparison. The correlations of SHBG with SHBG-related omics scores differed between MWH and MWOH (Additional file 1: Fig. S4). Specifically, SHBG showed significant positive correlations with microbiome and protein scores in MWH, whereas these correlations were null in MWOH; while the correlation between SHBG and the metabolite score was higher in MWH than in MWOH, with both being significant. Regarding other HIV-related factors, SHBG level and microbiome score differed across antiretroviral therapy regimen groups; SHBG-related metabolite score was higher in those with undetectable viral load compared to those with detectable viral load in MWH; and the SHBG-related microbiome score was higher in MWH than MWOH (Additional file 1: Fig. S5).

SHBG-related omics profiles and carotid artery plaque in MWH

We next studied among MWH whether SHBG-related omics profiles were associated with carotid artery plaque. SHBG-related microbiome score, metabolite score, and protein score had weak correlations with each other (ranging from 0.18 to 0.32, Fig. 4). All scores were inversely associated with plaque in MWH in both main and sensitivity analyses (p < 0.05 for all, Fig. 4). Among 18 species comprising the SHBG microbiome score, Mediterranea massiliensis, Coprococcus comes, and Prevotella melaninogenica, inversely related to SHBG, were positively associated with plaque (Additional file 2: Table S3). Among 102 metabolites comprising the SHBG metabolite score, 13 metabolites (mainly phosphatidylcholine-based lipids), positively related to SHBG, were inversely associated with plaque (Additional file 2: Table S3). Among 186 proteins comprising the SHBG protein score, 45 proteins (positively related to SHBG) were inversely associated with plaque (Additional file 2: Table S3). The significant enrichment pathway of these 45 proteins was ‘activation of immune response’ (involving 7 target genes, FDR-q = 0.033). Moreover, when applying p < 0.05 as a significance threshold for both the main and sensitivity analyses, 7 metabolites and 16 proteins remained statistically significant, whereas no microbial species were identified. Among the species, metabolites, and proteins associated with both SHBG and plaque, Coprococcus comes, Hex2Cer34:2;O and glutaminylglutamic acid (Glu-Gln) were correlated with a set of proteins enriched in activation of immune response pathway; Prevotella melaninogenica and certain lipid metabolites were correlated with DMP1 and IL17RB, which were enriched in the negative regulation of cell adhesion pathway (Additional file 1: Fig. S6). Exploratory bidirectional mediation analysis observed that the metabolite and protein scores mediated the inverse association between SHBG and plaque, with mediated proportions of 43.2% and 39.0%, respectively (Additional file 1: Fig. S7). However, the microbiome score did not significantly mediate the association between SHBG and plaque. We further explored the mediation effect of individual SHBG-related species/metabolites/proteins on the association between SHBG and plaque (Additional file 2: Table S4). Similarly, phosphatidylcholine-based lipids and proteins involved in activation of immune response and regulation of cell adhesion exhibited significant mediation effects, while microbial species did not.

Fig. 4.

Associations of SHBG-related omics scores with carotid artery plaque in MWH. Spearman correlations were examined among SHBG-related omics scores (n = 208 for microbiome score and n = 202 for metabolite/protein scores). Main analysis was performed to examine the associations of SHBG-related omics scores with carotid artery plaque, using multivariate logistic regression models with the adjustment for covariates described in Fig. 2 for MWH. Sensitivity analysis was performed by repeating multivariable logistic regression after excluding men with predicted plaque and averaging the estimates as described in Fig. 2. The odds ratio represents the odds of carotid artery plaque per 1-SD increase in each SHBG-related omics score. **p < 0.01; ***p < 0.001

Multi-omics profiles of E1-S in MWOH

In MWOH, E1-S was not associated with any microbial species, metabolites or proteins. Thus, omics scores were not generated for E1-S.

Discussion

We found that HIV serostatus altered the association between sex hormones and atherosclerosis in this study of MWH and MWOH. We observed an inverse association between SHBG and carotid artery plaque in MWH, and a positive association between E1-S and plaque in MWOH. By integrating multi-omics data, we further observed that certain SHBG-related microbial species, circulating metabolites and proteins were associated with plaque in MWH. These findings suggest potential biological mechanisms in which certain SHBG-related multi-omics signatures may link SHBG to subclinical atherosclerosis in MWH, which warrants further exploration in future longitudinal studies.

The cardiovascular effects of endogenous sex hormones, particularly estradiol and testosterone, have been extensively studied in an attempt to explain well-known sex differences in atherosclerosis [53, 54]. Accumulated studies have suggested that sex hormones play distinct roles in modulating vascular atherosclerosis and their levels in circulation have been correlated with carotid artery plaque composition in both women and men [18]. Estrogen is generally considered atheroprotective [55], while the impact of androgens on atherosclerosis is more controversial [20, 56–59]. We recently reported inverse associations of androgens (DHT and 3α-diol-17G) and estrogens (E1 and E2) with prevalence of carotid artery plaque in women with and without HIV. Here, we extend our research to men but find distinct associations in MWH and MWOH. SHBG showed an inverse association with carotid artery plaque in MWH, but not MWOH. However, previous studies have also reported the inverse association between SHBG and carotid artery plaque in postmenopausal women [13, 19, 55, 59–61] and men [56] living without HIV, despite these studies utilized different markers of subclinical atherosclerosis; in contrast, only one study found no association between SHBG and subclinical atherosclerosis in both women with and without HIV [62]. SHBG is a transport glycoprotein that binds testosterone and estradiol in the blood with a greater affinity for testosterone, and SHBG may have additional independent biological functions [11]. It regulates circulating hormone levels and influences cardiovascular risk both indirectly through hormone binding, and directly via SHBG receptors in certain vascular tissues (such as intimal and media layers) through cellular-level mechanisms [14, 55, 63–67]. However, we found no association between SHBG and plaque in MWOH. Inconsistent with previous studies reporting higher SHBG levels in MWH than MWOH [68, 69], our study found no significant difference in SHBG levels between the two groups. However, in multivariable regression adjusting for sociodemographic factors, lifestyle, clinical factors, and HIV serostatus, SHBG levels were non-significantly higher in MWH compared to MWOH, which was at least consistent with the direction observed in prior studies. Men in our study were somewhat older, which could have impacted the relationship of HIV and SHBG [70]. Hormonal disruptions and hypogonadism have been previously observed in MWH, related to HIV infection and/or ART [71–73]. Moreover, chronic inflammation associated with HIV infection is known to accelerate atherosclerosis [74] and may interact with hormonal disruptions [75]. Thus, the intricate interplay between HIV infection, aging, sex hormones, and subclinical atherosclerosis may contribute to differences between studies, which warrants further investigation.

In MWOH, we found a positive association between E1-S and carotid artery plaque. E1-S, a conjugated form of estrone produced via androgen peripheral aromatization and sulfation reactions, has been implicated in cardiovascular health. While estrogens generally protect against atherosclerosis in postmenopausal women [55, 61, 67], studies in men have reported an association between estradiol and increased carotid intima-media thickness [20, 56], suggesting potential sex-specific effects. Animal models indicated that male estrogen receptor-α knockout mice were less likely to develop atherosclerosis [76] and estrogen may act as a proinflammatory agent under certain conditions [54]. In addition, we did not observe significant associations between androgens and other estrogens and carotid artery plaque in MWH and MWOH.

Despite substantial epidemiological evidence linking sex hormones to atherosclerosis, the underlying mechanisms remain incompletely understood. Integrative multi-omics analysis implied a multifaceted mechanism underlying the protective association of SHBG and carotid artery plaque in MWH. Notably, Oscillibacter species, which were positively associated with SHBG, have recently been reported to associate with reduced blood cholesterol levels, suggesting potential benefits for lipid profiles and cardiovascular health [77]. A recent bidirectional two-sample Mendelian randomization analysis demonstrated a causal association between gut microbiome (exposure) and SHBG (outcome) rather than the reverse [78]. Although we identified several microbial species associated with SHBG, they did not appear to mediate the association between SHBG and plaque in the exploratory bi-directional mediation analysis. Given that numerous studies have highlighted the pivotal role of gut microbiota in the onset and progression of atherosclerosis [79–81], as well as the complexity of the interactions between gut bacteria and sex hormones in disease progression, further research is needed to understand whether gut microbiome may influence atherosclerosis through modulation of SHBG.

The metabolomics analyses elucidated that the SHBG-related metabolite score was inversely associated with carotid artery plaque in MWH. Previous studies in post-menopausal women found that SHBG was inversely associated with subclinical atherosclerosis and its progression, partially mediated by its beneficial effects on lipid profiles, including increased HDL-C and reduced LDL-C [19, 61]. In vitro evidence demonstrated that SHBG treatment significantly induced lipid degradation in differentiated 3T3-L1 adipocytes, enhancing lipolysis as indicated by increased glycerol concentrations. It also modulated gene and protein levels related to lipid metabolism, suppressing lipid accumulation independently of testosterone or estradiol [82]. SHBG has also been positively associated with HDL-C, while inversely associated with triglycerides, BMI, insulin resistance and inflammation [14, 66], which was considered a biomarker for metabolic disorders [83]. In our study, SHBG exhibited positive associations with a range of metabolites in MWH, such as glycerophospholipids and carnitines, linking to a favorable cardiovascular risk profile [84, 85]. Conversely, SHBG was negatively associated with some other metabolites, such as glycerolipids, sphingolipids, branched chain amino acids, and aromatic amino acids, which are commonly associated with an unfavorable cardiovascular risk profile [86–89]. Furthermore, a genetic study identified variants associated with SHBG concentrations, of which the genes were involved in various metabolic pathways related to lipid and carbohydrate metabolism, liver function, and type 2 diabetes [90], reinforcing a close linkage between SHBG and metabolic processes, which may be particularly important in MWH.

To our knowledge, this is the first study to examine plasma proteins associated with SHBG and their relations to subclinical atherosclerosis. In our study, SHBG was intricately associated with various cardiometabolic and inflammatory proteins, which were enriched in biological processes of cell function regulation and immune response. An in vitro study observed that SHBG reduced inflammation and lipid accumulation in adipocytes and macrophages, by markedly suppressing lipopolysaccharide-induced inflammatory cytokines production (e.g., MCP-1, TNFα, and IL-6) and phosphorylations of JNK and ERK [82], which may underlie the protective association of SHBG with cardiovascular health. Atherosclerosis is a chronic immune-inflammatory disease, with its initial stage involving the interplay of cell adhesion molecules and immune cells and facilitating the migration of leukocytes and lymphocytes into the arterial intima [91]. Our results found that SHBG-related proteins involved in immune response may play a role in the inverse association between SHBG and carotid artery plaque in MWH, which may be plausible given the crucial roles of immune factors during the development of atherosclerosis. Moreover, we found SHBG exhibited different associations with omics scores in MWH and MWOH, suggesting that HIV infection may affect the omics profile.

This study is the first multi-omics investigation to advance our knowledge of sex hormones and subclinical atherosclerosis in the context of HIV infection. The presented atlas of sex hormone and multi-omics signature associations will facilitate hypothesis generation for future study on sex hormone-related health effects. Major strengths of this study include broad and precise measurement of sex hormones, integration of multi-omics layers, including shotgun metagenomic sequencing, blood metabolomics and proteomics, as well as the use of a well-established cohort with comprehensive covariates allowing adjustment for confounders. These factors enabled us to uncover potential biological mechanisms and provide more insights compared to prior studies primarily relying on 16S rRNA sequencing and examining a narrower range of hormones. Our study also had several limitations. The observational nature of the study and cross-sectional measurement of sex hormones and multi-omics markers limit our ability to draw causal and temporal inferences. Notably, carotid artery plaque was assessed several years prior to the measurement of sex hormones and multi-omics markers, resulting in a temporal discordance between plaque ascertainment and biomarker assessment. This reversal of the expected exposure-outcome sequence may introduce potential misclassification of plaque status and precludes interpreting these biomarkers as antecedent exposures for plaque development. Specifically, individuals classified as plaque-free at the time of ultrasound may have developed plaque during the interval between ultrasound scan and biospecimen collection. Moreover, in the absence of repeated longitudinal measurements, we cannot assume the stability of circulating sex hormones or multi-omics profiles over time, further limiting their interpretability as proxies of prior exposure. To partially address this concern, we developed a prediction model for carotid artery plaque using earlier longitudinal data within this cohort and conducted sensitivity analyses by excluding participants with a higher probability of incident plaque (Additional file 1: Supplementary Methods). Nonetheless, misclassification cannot be fully excluded, and the findings should be interpreted with appropriate caution. Accordingly, our findings should be interpreted as cross-sectional associations rather than evidence of a directional or causal relationship. In addition, the exploratory mediation analyses are hypothesis-generating and warrant replication in future studies with appropriate prospective designs and temporally aligned measurements. Diet is a key determinant of gut microbiome composition and may act as an important confounder in microbiome-related associations, but was unavailable in our study. Further, the relatively smaller samples in MWOH (n = 113) compared to MWH (n = 208) may have resulted in lower statistical power to detect significant associations and limited the ability to assess effect modification by HIV. Finally, participants in the MACS are men who have sex with men with and without HIV, usually exhibiting distinct gut microbiome and metabolomic profiles compared with heterosexual men [92]. Therefore, the present findings may have limited generalizability to other populations, such as heterosexual men with and without HIV.

Conclusions

In summary, this study observed an inverse association between serum SHBG and carotid artery plaque, as well as associations between serum SHBG and altered microbial species, plasma metabolites, and proteins in MWH. Moreover, the observed inverse associations between SHBG-related microbiome, metabolite and protein scores and plaque, suggested that multi-omics signatures may play potential roles in the inverse association between SHBG and plaque. Understanding how sex hormones influence subclinical atherosclerosis in MWH and MWOH is important to enhance our understanding of sex differences in cardiovascular health in people with HIV [93–95]. While the putative anti-atherogenic role of SHBG on subclinical atherosclerosis among MWH is not fully elucidated, its relationships with microbial species, metabolites and proteins may provide further insight. Future longitudinal studies are warranted to substantiate these observations, establish temporality, and explore therapeutic targets within the multi-omics signatures to mitigate the cardiovascular risk associated with lower SHBG levels in the context of HIV.

Abbreviations

ADT

Androsterone

ADT-G

Androsterone-glucuronide

ALT

Alanine aminotransferase

ANCOM-BC

Analysis of Compositions of Microbiomes with Bias Correction

ART

Antiretroviral therapy

AST

Aspartate aminotransferase

BMI

Body mass index

Clr

Centered log-ratio

CVD

Cardiovascular disease

DHEA

Dehydroepiandrosterone

DHEA-S

Dehydroepiandrosterone-sulfate

DHT

Dihydrotestosterone

E1

Estrone

E1-S

Estrone-sulfate

E2

Estradiol

eGFR

Estimated glomerular filtration rate

ELISA

Enzyme-linked immunosorbent assay

FDR

False discovery rate

GC-MS

Gas chromatograph-mass spectrometry

HDL-C

High-density lipoprotein cholesterol

HCV

Hepatitis C virus

INT

Inverse normal transformation

JSD

Jensen-Shannon divergence

LC–MS/MS

Liquid chromatography-tandem mass spectrometry

LLOQ

Lower limit of quantification

MACS

Multicenter AIDS Cohort Study

MWCCS

MACS/Women’s Interagency HIV Study Combined Cohort Study

MWH

Men with HIV

MWOH

Men without HIV

OGU

Operational genomic unit

PERMANOVA

Permutational multivariate analysis of variance

SHBG

Sex hormone-binding globulin

Authors’ contributions

B.A.P., D.B.H., R.C.K, and Q.Q. conceived and designed the study; Y.W. performed the statistical analysis, visualization and drafted the manuscript; X.X. provided critical revision for statistical methodology; P.C, C.G. contributed to sex hormone measurements and data processing; M.U., C.B.C., R.E.G., R.D.B., and Z.W. contributed to multi-omics data processing and provided bioinformatics support; W.S.P., M.D.W., C.R.R., T.T.B., F.J.P., M.H.K., and B.E.S. contributed to cohort coordination and clinical data acquisition; H.N.H. oversaw carotid artery ultrasound reads and quality assurance; S.G. contributed to cohort data management; A.S., K.A., D.B.H., and R.C.K contributed to project funding; B.A.P. and D.B.H. supervised the study and critically revised the manuscript. All authors provided critical revision for important intellectual content of the manuscript, read and approved the final manuscript.

Funding

B.A.P., D.B.H., Q.Q., R.C.K., W.S.P., and Z.W. were supported by the National Heart, Lung, and Blood Institute (B.A.P. K01HL160146 and R03HL182350; R.C.K. R01HL148094; D.B.H. K01HL137557; R.C.K. and D.B.H. U01HL146204-04S1; Q.Q. and R.C.K. R01HL140976; Q.Q. R01HL170904; W.S.P. R01HL095129; Z.W. K01HL169019). T.T.B. is supported by K24AI120834. Data in this manuscript were collected by Multicenter AIDS Cohort Study (MACS), now (MACS)/Women’s Interagency HIV Study (WIHS) Combined Cohort Study (MWCCS). The contents of this publication are solely the responsibility of the authors and do not represent the official views of the National Institutes of Health (NIH). MWCCS (Principal Investigators): Atlanta CRS (Ighovwerha Ofotokun, Anandi Sheth, and Gina Wingood), U01-HL146241; Baltimore CRS (Todd Brown and Joseph Margolick), U01-HL146201; Bronx CRS (Kathryn Anastos, David Hanna, and Anjali Sharma), U01-HL146204; Brooklyn CRS (Deborah Gustafson and Tracey Wilson), U01-HL146202; Data Analysis and Coordination Center (Gypsyamber D’Souza, Stephen Gange and Elizabeth Topper), U01-HL146193; Chicago-Cook County CRS (Mardge Cohen, Audrey French, and Ryan Ross), U01-HL146245; Chicago-Northwestern CRS (Steven Wolinsky, Frank Palella, and Valentina Stosor), U01-HL146240; Northern California CRS (Bradley Aouizerat, Jennifer Price, and Phyllis Tien), U01-HL146242; Los Angeles CRS (Roger Detels and Matthew Mimiaga), U01-HL146333; Metropolitan Washington CRS (Seble Kassaye and Daniel Merenstein), U01-HL146205; Miami CRS (Maria Alcaide, Margaret Fischl, and Deborah Jones), U01-HL146203; Pittsburgh CRS (Jeremy Martinson and Charles Rinaldo), U01-HL146208; UAB-MS CRS (Mirjam-Colette Kempf, James B. Brock, Emily Levitan, and Deborah Konkle-Parker), U01-HL146192; UNC CRS (M. Bradley Drummond and Michelle Floris-Moore), U01-HL146194. The MWCCS is funded primarily by the National Heart, Lung, and Blood Institute (NHLBI), with additional co-funding from the Eunice Kennedy Shriver National Institute of Child Health & Human Development (NICHD), National Institute on Aging (NIA), National Institute of Dental & Craniofacial Research (NIDCR), National Institute of Allergy and Infectious Diseases (NIAID), National Institute of Neurological Disorders and Stroke (NINDS), National Institute of Mental Health (NIMH), National Institute on Drug Abuse (NIDA), National Institute of Nursing Research (NINR), National Cancer Institute (NCI), National Institute on Alcohol Abuse and Alcoholism (NIAAA), National Institute on Deafness and Other Communication Disorders (NIDCD), National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), National Institute on Minority Health and Health Disparities (NIMHD), and in coordination and alignment with the research priorities of the National Institutes of Health, Office of AIDS Research (OAR). MWCCS data collection is also supported by UL1-TR000004 (UCSF CTSA), UL1-TR003098 (JHU ICTR), UL1-TR001881 (UCLA CTSI), P30-AI-050409 (Atlanta CFAR), P30-AI-073961 (Miami CFAR), P30-AI-050410 (UNC CFAR), P30-AI-027767 (UAB CFAR), P30-AI-124414 (ERC-CFAR), P30-MH-116867 (Miami CHARM), UL1-TR001409 (DC CTSA), KL2-TR001432 (DC CTSA), and TL1-TR001431 (DC CTSA). Molecular data for the Trans-Omics in Precision Medicine (TOPMed) program was supported by the National Heart, Lung and Blood Institute (NHLBI). Metabolomics and proteomics for “NHLBI TOPMed: MWCCS: Sex differences in the role of multi-omics in HIV-associated carotid artery atherosclerosis” (phs003651) was performed at the Broad Institute and Beth Israel Metabolomics and Proteomics Platforms (3R01HL092577-06S1; 3U54HG003067-12S2; contract HHSN268201600034I—75N92020F00001, amendment P00004). Core support including phenotype harmonization, data management, sample-identity QC, and general program coordination were provided by the TOPMed Data Coordinating Center (R01HL-120393; U01HL-120393; contract HHSN268201800001I).

Data availability

MWCCS has an established process for the scientific community to apply for access to participant data and materials with such requests reviewed under coordination by the Data Analysis and Coordinating Center. These policies are described at https://statepi.jhsph.edu/mwccs/work-with-us/. The raw omics data used in this study were deposited in the NHLBI's TOPMed program and are accessible through the dbGaP repository under accession number phs003651 (https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs003651.v1.p1) [96]. The analysis code supporting the findings of this study has been deposited and is publicly available in Zenodo [97].

Declarations

Ethics approval and consent to participate

Approval was obtained from the Institutional Review Board (IRB) of the Albert Einstein College of Medicine (IRB# 2022–14425), which serves as the overall IRB approval for this study. The study was reviewed and approved by IRBs at all participating institutions (approval for the multi-omics analysis, Albert Einstein College of Medicine, 2022–14425; Bronx site, Biomedical Research Alliance of New York, 21–02-210-01E; Pittsburgh site, University of Pittsburgh, STUDY21050072; Chicago site, Northwestern University, STU00022906; Baltimore site and DACC, Johns Hopkins Bloomberg School of Public Health, 16882 and 23015; Los Angeles site, University of California, Los Angeles, 20–002292-AM-00001). All participants provided written informed consent. The research conformed to the principles of the Helsinki Declaration.

Consent for publication

Not applicable.

Competing interests

T.T.B. has served as a consultant to ViiV Healthcare, Merck, EMD-Serono, and GSK. Other authors have nothing to disclose.