Multi-omics analysis of the gut microbiome and carotid artery atherosclerosis in men with and without HIV

Summary

Background

How gut microbiota alterations may contribute to host inflammation and metabolomic profiles affecting atherosclerosis is not fully elucidated, especially in the context of HIV.

Methods

We examined associations between gut microbial features (measured by shotgun metagenomics) and subclinical carotid atherosclerosis, as assessed by high-resolution B-mode ultrasound, in 359 men from the MACS/WIHS Combined Cohort Study. We measured 822 plasma metabolites using LC–MS/MS, and up to 2866 circulating proteins by the Olink Explore 3072/384 platform (with a primary focus on 617 proteins related to inflammation and immune function).

Findings

Carotid artery plaque was detected in 115/359 men (32%). Adlercreutzia equolifaciens and Eubacterium sp3131 were associated with lower odds of plaque (OR [95% CI] = 0.57 [0.43, 0.77], 0.84 [0.76, 0.93], respectively), while Coprococcus sp13142 was associated with higher odds of plaque (OR [95% CI] = 1.14 [1.06, 1.23]). Results were consistent in men both with and without HIV. A. equolifaciens was positively correlated with HDL cholesterol and inversely correlated with systolic blood pressure. These plaque-associated microbial species were also associated with a range of circulating metabolites and inflammatory proteins. For example, A. equolifaciens positively correlated with the metabolites palmitoyl-EA and mesobilirubinogen, and inversely correlated with the pro-inflammatory chemokine CXCL9, the immune regulator CD160, and IL-24.

Interpretation

We identified gut microbial features associated with carotid artery atherosclerosis, consistent across HIV status; these associations were partially explained by specific microbiota-related metabolites and inflammatory markers. If validated, these findings suggest gut microbiota-related targets for CVD prevention.

Funding

The study was funded by the National Heart, Lung, and Blood Institute (U01HL146204-04S1, K01HL169019).

Research in context.

Evidence before this study

Emerging evidence suggested that the gut microbiota played an important role in the development of cardiovascular disease (CVD), including atherosclerosis. However, studies remain sparse in men living with HIV, who have both a higher risk of CVD and distinct alterations in gut microbiota compared with those without HIV. Circulating metabolites and host inflammatory responses may also contribute to atherosclerosis, but the interrelationships among gut microbiota, host inflammation, and immune activation, as well as the underlying mechanisms in the development of atherosclerosis, remain incompletely understood, particularly in the context of HIV.

Added value of this study

By integrating the gut microbiome shotgun metagenomics data with host circulating proteomics and metabolomics data, we investigated gut microbial features associated with carotid artery atherosclerosis in the context of HIV, and revealed potential mechanistic links through host inflammation and metabolism. We identified specific gut microbial species associated with carotid artery plaque, including Adlercreutzia equolifaciens and Eubacterium spp., which were linked to lower risk of plaque. Notably, A. equolifaciens showed favourable associations with conventional CVD risk markers, including a positive association with HDL cholesterol and an inverse association with systolic blood pressure. These plaque-associated species were also connected to circulating metabolites and inflammatory proteins: for example, A. equolifaciens was positively associated with palmitoyl-EA and mesobilirubinogen, both with anti-inflammatory or antioxidant properties; and inversely associated with the pro-inflammatory chemokine CXCL9, the immune regulator CD160, and IL-24, which can promote cell apoptosis. Importantly, the association between A. equolifaciens and plaque appeared to be partially explained by circulating metabolites and inflammatory markers, providing mechanistic insights into the gut–vascular axis in HIV.

Implications of all the available evidence

Taken together, current evidence supports the hypothesis that gut microbiota can contribute to atherosclerosis and CVD, at least in part through host inflammation and immune activation pathways. These findings advanced our understanding of the gut–vascular axis and provided important evidence to the concept that modulation of specific gut microbial species and related metabolites may have therapeutic potential for the prevention of atherosclerosis and CVD.

Introduction

Cardiovascular disease (CVD) has become a leading cause of morbidity and mortality among people living with HIV (PLWH), particularly as antiretroviral therapy (ART) has substantially extended life expectancy.^1,2 The increased risk of CVD in PLWH is attributed to a combination of chronic immune activation, persistent inflammation, and traditional cardiometabolic risk factors.^1,3,4 Prior research has shown that HIV infection is associated with the development of carotid artery plaque, an established marker of subclinical atherosclerosis that predicts increased risk for future CVD events.5, 6, 7 Accumulating evidence suggests that the gut microbiota may play a key role in host vascular physiology and the pathogenesis of atherosclerosis.8, 9, 10, 11, 12, 13 Bacterial DNA has been detected in human carotid plaques, suggesting a potential microbial translocation, from the oral and gut microbiome, in plaque development.^14,15 Furthermore, PLWH exhibit distinct alterations in gut microbial composition,16, 17, 18 highlighting the importance of understanding how gut microbiota may influence atherosclerosis and CVD among PLWH.

Gut microbial dysbiosis has been linked to changes in circulating metabolites and host inflammatory responses, both of which may contribute to atherosclerosis, although the underlying mechanisms remain incompletely understood.^8,10,19 In our prior study of women with or without HIV, we identified specific gut bacterial taxa (e.g., Fusobacterium) and their associated lipid metabolism functions (e.g., phospholipase A1 and A2) that were linked to both host plasma lipid metabolites and carotid artery atherosclerosis.²⁰ In a follow-up proteomics study, we found that associations between Fusobacterium and carotid plaque were also partially explained by serum inflammatory markers, including CXCL9, which are involved in host immune activation and microbial defence responses.²¹ These findings support the hypothesis that gut microbiota may contribute to CVD through both metabolic pathways and inflammation, particularly in PLWH.

However, findings in women may not be generalisable to men. Carotid atherosclerosis is more prevalent in men,²² and men typically develop CVD at earlier ages, while women face higher mortality and worse outcomes following acute CVD events, although lifetime CVD risk is comparable between sexes.23, 24, 25 These sex differences underscore the need for studies in men to elucidate the potential mechanisms by which gut microbiota may influence host metabolism and inflammation, thereby influencing CVD, and especially in men living with HIV, where multi-omics investigations remain scarce.

In the present study, we undertook multi-omics measurement and integrative analyses to extend the investigation of gut microbiota and its relationship with subclinical carotid atherosclerosis to men living with or without HIV. Specifically, we aimed to identify gut microbiota features (e.g., overall microbial diversity and individual taxonomic species, measured by shotgun metagenomic sequencing) associated with carotid artery plaque in male participants from the MACS/WIHS Combined Cohort Study (MWCCS). Furthermore, we examined associations between plaque-associated microbial species and host plasma metabolites (measured using liquid chromatography–tandem mass spectrometry), as well as plasma proteins (assessed via the Olink Explore 3072/384 proteomics platform). We conducted integrated analyses to explore potential mechanisms underlying the relationship between gut microbiota and subclinical atherosclerosis among PLWH.

Methods

Study population

The Multicenter AIDS Cohort Study (MACS) was a multi-site longitudinal cohort of men who have sex with men (MSM), either with or at risk for HIV infection, conducted at 4 U.S. sites,^26,27 now continuing as part of the MACS/WIHS Combined Cohort Study (MWCCS). Detailed study design of the MACS and MWCCS have been described previously.^26,27 In brief, participants currently complete annual core study visits with a comprehensive physical examination, clinical measurements, and interviewer-administered questionnaires, and provide biological specimens for laboratory tests and repository storage. Fecal samples were collected using a home-based self-collection kit^28,29 during 2020–2021, and shotgun metagenomic sequencing was performed. This analysis was led by the Bronx MWCCS site based at Albert Einstein College of Medicine and included data collected at MWCCS sites based at the University of Pittsburgh, Northwestern University, Johns Hopkins University, and University of California, Los Angeles.

In this study, we included 359 MWCCS men (217 with HIV and 142 without HIV) who met the following criteria: (1) underwent prior B-mode carotid artery ultrasound for atherosclerotic plaque assessment; (2) had gut microbiome shotgun metagenomics data; (3) had available plasma metabolite data; and (4) had proteomic marker data (Fig. 1). Both metabolomics and proteomics data were derived from thawed frozen plasma samples that were collected at closest available core visit from fecal sample collection.

Fig. 1.

Overview of the study design. The MACS/WIHS Combined Cohort Study (MWCCS) is a multicenter cohort of individuals living with or at risk for HIV infection. In this study, we included a total of 359 eligible participants (MACS men; 60.5% living with HIV) who underwent B-mode carotid artery ultrasound for atherosclerotic plaque assessment and provided fecal samples for gut microbiome profiling via shotgun metagenomics sequencing. In addition, we measured circulating metabolites using liquid chromatography–mass spectrometry (LC-MS), and proteomic inflammatory markers using the Olink Explore 3072/384 platform. We performed integrated multi-omics analyses to examine associations of gut microbial features, circulating metabolomic profiles, and proteomic inflammatory markers with carotid artery plaque, in the context of HIV infection.

Microbiome shotgun metagenomic sequencing

Shotgun metagenomic sequencing was performed at the Burk lab (Albert Einstein College of Medicine, Bronx, NY) using NovaSeq 6000 (Illumina, San Diego, CA) with established and highly reproducible protocols.^30,31 In brief, DNA was extracted from stool samples using the DNeasy PowerLyzer PowerSoil Kit (Qiagen, Hilden, Germany; cat# 12855-100). After quantification, DNA underwent library construction via the KAPA HyperPlus Kit (Roche, Basel, Switzerland; cat# 07962428001) using xGen Stubby Adaptors (IDT, Coralville, IA; cat# 10005924 and 10 nt Unique Dual Index (UDI) primers (IDT, Coralville, IA; cat# 10008054). Libraries were constructed with optimised protocols by the Knight lab (University of California, San Diego).³¹ To enhance compatibility with the sequencing platform, pooled libraries were further purified and concentrated using the QIAquick PCR Purification kit (Qiagen, Hilden, Germany; cat# 28104).

The 2 × 150 bp paired-end raw FASTQ sequence reads were processed using the standard shotgun metagenomic sequencing pipeline in Qiita.³² We trimmed low-quality bases using prinseq 0.20.4,³³ followed by quality control using KneadData 0.12.0, which wraps several software packages to: (1) identify overrepresented sequences with FASTQC 0.11.9, (2) trim adaptors and low-quality bases with Trimmomatic 0.39 (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 4.09, and (4) remove reads mapping to the human genome with Bowtie 2.³⁴ We aligned the reads against the WolR1 reference database of bacterial and archaeal genomes using Woltka, and Bowtie 2³⁴ was selected as the alignment tool. After microbial taxonomic assignment, downstream taxonomic analyses were performed at the species level. α-diversity indices (Shannon index, Chao-1 index and Faith's PD index) and β-diversity (weighted UniFrac distances) were calculated using Qiita.³²

Metabolomics profiling

We measured circulating metabolites in plasma samples using liquid chromatography-tandem mass spectrometry (LC-MS/MS) at the Broad Institute Metabolomics Platform (Cambridge, MA). Two separate LC-MS/MS methods were performed to measure lipids and polar metabolites in each sample, using Thermo Exactive Focus and Q-Exactive HF mass spectrometers coupled to a Shimadzu Nexera ×2 U-HPLC system, as previously described.35, 36, 37 Raw data from Orbitrap mass spectrometers were processed using Progenesis QI version 3.0 (NonLinear Dynamics) for feature alignment, untargeted signal detection, and signal integration. Metabolite abundances were estimated using the area under the curve. Targeted processing of known metabolites was conducted using TraceFinder software (version 3.1, Thermo Fisher Scientific; Waltham, MA), based on matching accurate mass, chromatographic retention time, and MS/MS fragmentation patterns to the reference library at the Broad Institute Metabolomics Platform. We included a total of 822 annotated metabolites in the current analysis, and all metabolites had coefficient variation <20% and missing rate <20%. Metabolites with missing data (under detectable levels) were imputed with ½ the minimum value for a given metabolite.

Proteomics profiling

Proteomics profiling was performed in plasma samples at the Beth Israel Proteomics Platform (Boston, MA) using Olink® Explore 3072/384 (Olink, Boston, MA), which consists of 8 panels targeting biomarkers related to cardiometabolic risk factors, inflammation, oncology and neurology. Protein concentrations were reported as normalised protein expression (NPX) units, which are Ct values from the PCR read-out and normalised by the subtraction of values for extension control, as well as an inter-plate control. The scale was shifted using a correction factor (normal background noise) and log² scaled.^38,39 After quality control based on internal and external Olink controls and standard QC thresholds, we included 2866 plasma proteins in the downstream statistical analyses.

Carotid artery plaque ascertainment

High-resolution B-mode carotid artery ultrasound with automated computerised edge detection⁷ had been performed in 2012–2013 to detect carotid artery plaque. A standardised protocol was used by all study sites, and focal plaque measures were obtained at a centralised reading center (University of Southern California, Los Angeles, CA),^7,40 to ensure consistency in image acquisition and interpretation. We defined a focal plaque as an area with localised intima-media thickness >1.5 mm in any of eight segments in the right carotid artery.⁴¹ Because carotid artery plaque assessment occurred several years before omics measurements, we employed prediction modelling to account for the difference in timing, as described below and in the Supplementary Methods and Fig. S8.

Assessments of HIV infection and other variables

Demographic, behavioural, clinical, and laboratory variables were collected using standardised protocols at core study visits. HIV serostatus was ascertained by enzyme-linked immunosorbent assay and confirmed by Western blot. HIV-specific parameters included CD4+ T-lymphocyte cell counts, HIV-1 plasma viral loads, and ART use.⁴² Conventional CVD risk factors assessed included body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), serum triglycerides, total cholesterol, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, fasting glucose, and haemoglobin A1c.⁴³ In this study, sex was self-reported by participants at their baseline visit. We defined MSM status based on self-reported receptive anal intercourse since the last study visit. In this study, all covariates were obtained from the core visit closest in time to fecal sample collection.

Statistics

Gut microbiota and prevalent carotid artery plaque

We examined associations of overall microbiome composition with carotid artery plaque. The Kruskal–Wallis test was applied to compare differences in microbial α-diversity indices by carotid artery plaque status. Permutational multivariate ANOVA (PERMANOVA) and principal-coordinate analysis (PCoA) were carried out for microbial β-diversity analyses.

We then examined associations of gut microbiota taxonomic features with plaque. Logistic regression was performed to examine multivariable-adjusted associations between gut microbial species (with central log ratio (CLR) transformation conducted for species-level taxonomic abundance) and carotid artery plaque, adjusting for age, race/ethnicity, study site, antibiotic use, income, education, BMI, smoking, alcohol, HIV serostatus and ART use. These covariates were selected as potential confounders based on their known or plausible associations with the gut microbiome and cardiometabolic health. Race/ethnicity was included to account for potential confounding arising from social, environmental, and structural factors that may influence microbial composition and cardiometabolic outcomes. We excluded species present in <20% of the population or with average relative abundance <0.001%. We controlled the false discovery rate (FDR) at 10%. Since carotid artery scans were performed prior to fecal sample collection, sensitivity analyses were conducted to obtain robust results. Briefly, we employed a prediction model for incident carotid artery plaque that was established using longitudinal data from MACS participants. This model was applied to estimate the probability of developing incident plaque at time of omics measurement for men without plaque in our study, based on predicted plaque status. In sensitivity analyses, logistic regression analyses were performed excluding participants with predicted plaque. The process was repeated 100 times, and effect estimates were averaged across the 100 datasets to estimate associations of gut microbiota taxonomic features with carotid artery plaque (see Supplementary Methods for additional details).

Spearman correlation was employed to estimate correlation coefficients among the identified carotid artery plaque-associated species, and traditional CVD risk factors including BMI, SBP, DBP, triglycerides, total cholesterol, LDL cholesterol, HDL cholesterol, fasting glucose and haemoglobin A1c.

To further explore potential HIV-specific results, we conducted stratified logistic regression analyses by HIV serostatus, focussing on associations between microbial species and carotid artery plaque. In addition, we examined the associations of plaque-associated species with HIV serostatus using multivariable regression models, adjusting for the aforementioned covariates.

Metabolomics and proteomics profiles and carotid artery plaque

We applied partial least squares discriminant analysis (PLSDA)⁴⁴ to separately examine associations between overall circulating metabolomics profiles (based on 822 metabolites) and carotid artery plaque, as well as between overall proteomic profiles (based on 2866 proteins) and plaque status.

Identification of gut microbiota-associated circulating metabolites and proteomic inflammatory markers

In addition to global analyses of metabolomic and proteomic profiles, we performed partial Spearman correlation analyses to examine associations between microbial species and specific individual circulating metabolites and proteomic inflammatory markers, adjusting for age, race/ethnicity, study site, antibiotic use, income, education, BMI, alcohol consumption, smoking status, and HIV serostatus. To ensure comparability with prior work conducted in women,²¹ the current analyses focused specifically on 822 circulating metabolites, and 617 proteomic inflammatory markers (Olink Explore “Inflammation” panels). Network analysis was conducted using a weighted correlation network analysis implemented in the WGCNA 1.72,⁴⁵ to examine interrelationships among microbiota-associated biomarkers and to assign metabolomic modules.

Integrated analyses of gut microbiota, microbial-associated omics profiles and carotid artery plaque

We performed multivariable logistic regression analyses to investigate associations of carotid artery plaque with microbial-associated circulating metabolites and proteomic inflammatory markers (inverse normal-transformed), adjusting for the aforementioned covariates. We then further adjusted for microbial metabolites and proteomic markers in logistic regression models to examine whether these factors may partially explain the associations between microbial species and carotid artery plaque, adjusting for the aforementioned covariates. In addition, we examined associations of these microbial-associated circulating metabolites and proteomic markers with HIV serostatus using multivariable regression models, adjusting for the aforementioned covariates.

In addition, we conducted a comprehensive multi-omics factor analysis (MOFA)⁴⁶ integrating the gut microbiome, circulating metabolomics, and proteomics datasets.

The Benjamini-Hochberg false discovery rate (FDR) method was used for multiple testing correction. Statistical analyses were performed using R 4.0.2, unless otherwise stated.

Ethics

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 Institutional Review Boards 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. All experimental methods comply with the Helsinki Declaration.

Role of funders

The funders played no role in the design of the study, the collection of data, the analysis and interpretation of data, or the writing of the manuscript.

Results

Gut microbiota and carotid artery plaque

Table 1 shows characteristics of the 359 participants (217 with HIV and 142 without HIV). Carotid artery plaque was detected in 115 of 359 men (32%). As expected, compared with men without plaque, participants with plaque were older and more likely to have higher levels of CVD risk factors (e.g., higher blood pressure). The majority of men with HIV reported ART use (96% and 88%, in participants without or with plaque, respectively) and had an undetectable HIV-1 viral load (<20 copies/mL, 75% and 83%, respectively).

Table 1.

Characteristics of study participants by carotid artery plaque status.

	Carotid artery plaque−	Carotid artery plaque+	P value
Number of participants	n = 244	N = 115
Age, y, median (IQR)	62 (57–67)	67 (61–71)	<0.001
Body mass index, kg/m2, median (IQR)	26.6 (24.2–30.0)	26.5 (24.3–30.4)	0.470
Race/ethnicity, n (%)			0.116
African-American	80 (33)	28 (24)
White	156 (64)	79 (69)
Hispanic/Other	8 (3)	8 (7)
Annual income <$12,000, n (%)	21 (9)	14 (12)	0.384
Education less than high school, n (%)	105 (43)	41 (35)	0.216
Smoking status, (%)			0.064
Never smoker	80 (33)	28 (24)
Current smoker	24 (10)	20 (17)
Former smoker	131 (54)	65 (57)
Antibiotic use, n (%)	48 (20)	23 (20)	0.909
Systolic blood pressure, mmHg	125 (115–135)	128 (119–142)	0.045
Diastolic blood pressure, mmHg	75 (68–83)	74 (69–83)	0.845
Triglycerides, mg/dL	101 (72–141)	113 (74–170)	0.736
Total cholesterol, mg/dL	162 (136–187)	157 (133–183)	0.714
HDL cholesterol, mg/dL	48 (40–59)	48 (41–58)	0.379
LDL cholesterol, mg/dL	86 (67–109)	81 (67–103)	0.480
Glucose, mg/dL	94 (87–105)	96 (90–107)	0.938
HbA1C, percent	5.4 (5.1–5.8)	5.5 (5.2–5.9)	0.638
HIV-specific characteristics
Living with HIV, n (%)	142 (58)	75 (65)	0.209
Detectable HIV-1 viral load (>20 copies/mL), among PLWH, n (%)	36 (25)	13 (17)	0.190
Low CD4+ T-cell count (<500 cells/μL), among PLWH, n (%)	43 (30)	29 (39)	0.217
ART use, among PLWH, n (%)	136 (96)	66 (88)	0.197

n = 359 men from the MACS/WIHS Combined Cohort Study (MWCCS).

Data are presented as count (%) for categorical variables or median (IQR) for continuous variables unless otherwise noted.

Abbreviations: ART, antiretroviral therapy; HDL, high-density lipoprotein; HbA1C, haemoglobin A1c; HIV, human immunodeficiency virus; IQR, interquartile range; LDL, low-density lipoprotein; PLWH, people living with HIV.

Antibiotic use was based on self-reported medication records at the time of closest core visit.

We did not find significant associations of α-diversity indices (Shannon, Chao-1 and Simpson's indices) (all P > 0.05, Fig. S1A) or β-diversity (measured using weighted UniFrac distances) (Fig. S1B; R²< 0.1, P > 0.05, PERMANOVA analysis) with carotid artery plaque.

In individual taxonomy analyses, after multivariable adjustment, we detected significant differences in the CLR-transformed abundance of Eubacterium sp3131, A. equolifaciens, and Coprococcus sp13142 between men with and without plaque (Fig. 2A and Table S1). Higher levels of Coprococcus sp13142 were associated with elevated odds of carotid artery plaque (OR [95% CI] = 1.14 [1.06, 1.23], per SD increment in CLR-transformed abundance) while higher levels of Eubacterium sp3131 and A. equolifaciens were associated with lower odds of plaque (OR [95% CI] = 0.57 [0.43, 0.77], 0.84 [0.76, 0.93], respectively) (all FDR <0.1). We obtained highly consistent results in the sensitivity analysis (Fig. S2 and Table S2). The integrated phylogenetic tree and correlation analyses indicated that these microbial species were not significantly correlated with one another, nor did they share close phylogenetic relationships (Fig. 2B).

Fig. 2.

Associations between microbial taxa and carotid artery plaque status (n=359). (A) Associations between species and plaque status (multivariable logistic regression). Results are odds ratios (ORs) and 95% confidence intervals (CIs) for carotid artery plaque per standard deviation increment of CLR transformed abundance of gut microbial species, adjusted for age, race/ethnicity, study site, antibiotic use, income, education, BMI, current smoking, alcohol, HIV status and ART use. (B) Integrated phylogenetic tree. Taxa from inner to outer circle represent bacterial kingdom to species level. Blue and red fonts represent positive and negative associations between species and carotid artery plaque, respectively. (C) Partial Spearman correlations between plaque-associated bacterial species and traditional CVD risk factors, adjusted for the aforementioned covariates (∗P < 0.05, partial Spearman correlation). Abbreviations: BMI, body mass index; SBP, Systolic blood pressure; DBP, Diastolic blood pressure; HDL, high-density lipoprotein; LDL, low-density lipoprotein; HbA1C, haemoglobin A1c.

We examined correlations between these plaque-associated microbial species and conventional CVD risk factors, and found that A. equolifaciens correlated positively with HDL cholesterol (r = 0.26, P < 0.001, partial Spearman correlation) and inversely with SBP (r = −0.17, P = 0.009, partial Spearman correlation) (Fig. 2C). No substantial correlations were identified with CVD risk factors for Coprococcus sp13142 or Eubacterium sp3131.

In analyses stratified by HIV serostatus, we observed consistent results across these strata—men with and without HIV (Table S3), and we did not observe significant effect modification by HIV serostatus (Table S3). None of the three plaque-associated microbial species showed significant differences by HIV serostatus after adjustment for covariates (Table S4).

As shown in Table S5, the distribution of receptive anal intercourse did not differ significantly between individuals with and without carotid artery plaque (P = 0.168, χ² test). We also conducted sensitivity analyses in which we additionally adjusted for MSM status. As shown in Fig. S3, the associations between microbial species and carotid artery plaque status remained materially unchanged.

Metabolomic and proteomic profiles and carotid artery plaque

We performed global analyses to investigate the associations between overall circulating metabolomic profiles and carotid artery plaque, as well as overall proteomic profiles and plaque, separately.

For metabolomic profiles, partial least squares discriminant analysis (PLSDA) of all 822 circulating metabolites revealed a potential distinction between participants with and without plaque, although the two groups were not fully separated (Fig. S4A). Detailed properties and annotation information for the 822 metabolites are summarised in Table S6.

For proteomic profiles, PLS-DA of all 2866 proteomic markers demonstrated a clear separation between participants with and without plaque (Fig. S5A).

Gut microbiota-associated circulating metabolites and proteomic markers

We next examined correlations of identified plaque-associated bacterial species with individual circulating metabolites (Fig. 3A) and individual proteomic markers (Fig. 3B). The plaque-associated microbial species were linked to a range of circulating metabolites spanning several major metabolite classes, including sphingolipids, glycerophospholipids, fatty acyls, tetrapyrroles and derivatives, and purine nucleosides (Fig. S4B). The relationships among these microbe-related metabolites are shown in Fig. S6. Since these microbiota-related sphingolipids (i.e., sphingomyelins [SMs]) were highly correlated with each other and structurally similar, we created an SM module score to represent them in downstream analyses.²⁰ This was also the case for phosphatidylcholines (PCs). The potentially protective A. equolifaciens and Eubacterium sp3131 were both significantly correlated with mesobilirubinogen (a metabolic product of bilirubin with antioxidant properties⁴⁷), as well as with the SM and PC modules. In addition, A. equolifaciens was positively associated with the metabolites palmitoyl-EA (PEA, known for its antihypertensive and anti-inflammatory properties⁴⁸) and adenosine. In contrast, the unfavourable Coprococcus sp13142 was negatively associated with mesobilirubinogen. We then examined associations between microbiota-related metabolites and carotid artery plaque (Fig. 3C). Consistent with the protective associations of A. equolifaciens and Eubacterium sp3131 with prevalent carotid plaque, mesobilirubinogen—which was positively correlated with both species—also showed a significant protective association with plaque (OR [95% CI] = 0.41 [0.19, 0.89], P = 0.024, logistic regression), after adjusting for demographic and behavioural variables. Associations of palmitoyl-EA and adenosine with plaque were also in the protective direction, although not statistically significant. We did not observe significant associations between microbiota-related lipids (i.e., SM or PC modules) and plaque. None of these microbial-associated metabolites showed significant associations with HIV serostatus (Table S7).

Fig. 3.

Gut microbiota, circulating metabolites, proteomic inflammatory markers, and carotid artery plaque status (n=359). Associations of identified plaque-associated bacterial species with (A) host circulating metabolites and (B) proteomic inflammatory markers. Partial Spearman correlations were performed with adjustment for age, race/ethnicity, study site, antibiotic use, income, education, body mass index, alcohol use, smoking, and HIV status. (C) Associations of microbial-associated circulating metabolites and proteomic inflammatory markers with plaque status, adjusted for the aforementioned covariates. Results are odds ratios and 95% confidence intervals (CIs) for carotid artery plaque per standard deviation change in the CLR transformed abundance of each metabolite, metabolite module, or inflammatory marker. The complete lists of metabolites comprising the PC and SM modules, together with their inter-metabolite correlation structure, are provided in Fig. S6. PCs, phosphatidylcholines; SMs, sphingomyelins.

In addition to global proteomic analyses involving all 2866 circulating proteins, we also examined associations between plaque-associated microbial species and specific individual proteomic markers. Prior evidence has shown that gut microbiota may contribute to atherosclerosis through modulation of host immune activation and inflammation.^13,21 Notably, in our previous study of women living with HIV, we identified microbial-associated inflammatory proteins such as CXCL9,²¹ which may partially explain the associations between microbial taxa and carotid artery plaque. To keep comparability with prior work in women, the current analysis focused specifically on 617 proteomic inflammatory markers (Fig. S5B). We found that the unfavourable Coprococcus sp13142 was positively associated with the plasma inflammatory marker AOC1. Both of the potentially protective species, A. equolifaciens and Eubacterium sp3131, were inversely correlated with plasma inflammatory markers CXCL9, CD6, CD160, and LY9. In addition, A. equolifaciens was also inversely associated with IL-24, IGLC2, and MZB1. We further examined the associations between microbiota-related proteomic inflammatory markers and carotid artery plaque (Fig. 3C). After adjusting for demographic and behavioural variables, CXCL9 was significantly positively associated with plaque (OR [95% CI] = 1.51 [1.03, 2.21]; P = 0.037, logistic regression), while LY9 showed a marginally positive association (OR [95% CI] = 1.51 [0.97, 2.34]; P = 0.068, logistic regression). Among the microbial-associated proteomic inflammatory markers, CD160 and LY9 were significantly elevated in men with HIV compared to those without (linear regression, P = 0.001 and P = 0.003, respectively, Table S8). CXCL9 also showed a marginal positive association with HIV serostatus (linear regression, P = 0.070).

Integrated analyses of gut microbiota, microbial-associated omics profiles and carotid artery plaque

To examine whether the identified microbial-related metabolites and proteomic inflammatory markers could potentially explain the associations between gut microbial species and carotid artery plaque, we further included these omics features—separately and jointly—as covariates in multivariable logistic regression models (Fig. 4).

Fig. 4.

Associations between microbial species and plaque status, with and without adjustment for microbial-related metabolites and/or microbial-related proteomic inflammatory markers (n=359). Data are odds ratios (ORs) and 95% confidence intervals (CIs) for carotid artery plaque per standard deviation increment of CLR-transformed abundance of each gut microbial species. Model 1: adjusted for age, race/ethnicity, study site, antibiotic use, income, education, body mass index, current smoking, alcohol, HIV status and antiretroviral therapy use. Model 2: adjusted for covariates in Model 1 + circulating metabolites relevant to each specific species. Model 3: adjusted for covariates in Model 1 + proteomic inflammatory markers relevant to each specific species. Model 4: adjusted for covariates in Model 1 + circulating metabolites and proteomic inflammatory markers relevant to each specific species.

The associations between these three identified microbial species (Eubacterium sp3131, A. equolifaciens, and Coprococcus sp13142) and carotid plaque were attenuated after adjusting for their corresponding taxa-related metabolites (Model 2). When adjusting for their corresponding taxa-related proteomic inflammatory markers (Model 3), the associations were attenuated to a lesser extent and remained statistically significant. When both metabolites and proteomic markers associated with each microbial species were simultaneously included as covariates (Model 4), a more pronounced attenuation was observed, and for A. equolifaciens and Eubacterium sp3131, the associations with plaque were further attenuated and no longer statistically significant. These results suggest that both microbiota-associated metabolites and proteomic inflammatory markers (especially metabolites), may partially explain the observed associations between gut microbiota and carotid artery plaque.

In addition, our MOFA analysis indicated that all three omics layers contributed meaningful variance to the latent factors, with proteomics explaining the largest proportion of variance (37.1%), followed by the gut microbiome (29.1%) and circulating metabolomics (26.8%) (Fig. S7A). Factor 1 was predominantly driven by proteomic features and showed the strongest association with atherosclerotic disease status (P < 0.001, linear regression). Factor 2 was mainly shaped by gut microbiome variation and demonstrated a marginal association with plaque (P = 0.08, linear regression). Circulating metabolomics contributed most to Factors 3, 4, and 5; among these, Factor 4 showed a marginal association with plaque (P = 0.09, linear regression) (Fig. S7B and C). Collectively, the MOFA results demonstrated that gut microbiome, metabolomics, and proteomics each contribute complementary information associated with carotid plaque.

Discussion

In this study, our integrative analyses incorporated shotgun metagenomics, untargeted metabolomics, and proteomics data, extending our prior investigations of gut microbiota and its relationship with atherosclerosis in women living with or at risk for HIV to men. We identified several gut microbial species and their related metabolites and inflammatory markers that were significantly associated with carotid artery plaque. These findings shed light on the complex interactions among microbiota, microbial-related omics features, and host biology involved in subclinical cardiovascular disease, and provide new evidence for potential microbiota-related mechanisms contributing to atherosclerosis in the context of HIV.

A. equolifaciens is a Gram-positive, obligately anaerobic bacterium recognised for its anti-inflammatory properties both in vitro and in vivo.^49,50 The Adlercreutzia genus, including A. equolifaciens, is known for its ability to produce equol,⁵⁰ a metabolite of the soy isoflavone daidzein, which functions as a phytoestrogen and has been shown to exert beneficial effects on vasodilation and nitric oxide metabolism,^51,52 potentially improving vascular health.²⁰ In our previous study, we reported protective associations between the Adlercreutzia and atherosclerosis in women with and without HIV.²⁰ In the present study, we extend our research to men and observe consistent associations, suggesting a broader protective role for A. equolifaciens in atherosclerosis across sexes.

We identified several circulating metabolites associated with A. equolifaciens. Notably, this species was positively associated with mesobilirubinogen, adenosine, and palmitoyl-ethanolamide (PEA), which may partially explain its protective effects against plaque. Mesobilirubinogen is a tetrapyrrolic compound formed during the microbial reduction of bilirubin in the intestine and serves as an intermediate in the catabolic pathway of haem degradation.⁴⁷ Its precursor, bilirubin, has well-documented antioxidant and anti-inflammatory properties and has been associated with a reduced risk of CVD events.⁵³ Our findings suggest that A. equolifaciens may have the capacity to convert bilirubin into mesobilirubinogen. Consonant with previous reports, we found that mesobilirubinogen itself demonstrated a significant protective association with carotid artery plaque. Adenosine, another metabolite correlated with A. equolifaciens, is a purine nucleoside that acts as a potent vasodilator and exerts anti-inflammatory and athero-protective effects primarily through activation of the A2A receptor.^54,55 Additionally, PEA, a fatty acid amide also associated with A. equolifaciens, has been reported to have anti-inflammatory, analgesic, and neuroprotective effects.⁴⁸ While its direct cardiovascular effects remain underexplored, the known anti-inflammatory actions of PEA may indirectly support vascular health.

Our study also identified inverse associations between A. equolifaciens and specific inflammatory markers, offering insights into its potential anti-inflammatory mechanisms. Notably, A. equolifaciens was inversely associated with several inflammatory markers, including CXCL9, LY9, and IGLC2. Of note, we observed that some inflammatory markers, including CXCL9, were significantly positively associated with carotid artery plaque. This finding aligns with our previous results in women living with HIV²¹ and is now extended to men, underscoring the consistency across populations with HIV. CXCL9 is a pro-inflammatory chemokine produced by endothelial cells in response to interferon-gamma (IFN-γ) stimulation.⁵⁶ It promotes atherosclerosis by recruiting T-cells through the CXCR3 receptor, contributing to vascular inflammation and CVD pathogenesis.^57,58 The observed inverse association between A. equolifaciens and CXCL9 suggests its potential favourable role in modulating immune-mediated vascular injury. IGLC2 (immunoglobulin lambda constant 2) is believed to influence antigen binding activity and it is a marker of active immunoglobulin synthesis.⁵⁹ Elevated levels of IGLC2 have been identified as a potential biomarker for the progression of cardiovascular pathology associated with type 2 diabetes in a recent targeted proteomic study.⁶⁰ This evidence suggests that IGLC2 may be involved in inflammatory processes contributing to CVD pathogenesis. LY9 (lymphocyte antigen 9), also known as CD229, is part of the signalling lymphocytic activation molecule (SLAM) family. It modulates T-cell and natural killer T (NKT)-cell activation.⁶¹ Although direct evidence linking LY9 to CVD is limited, increased expression of LY9 has been noted in certain inflammatory states,⁶² and its resultant enhancement of chronic inflammation may contribute to development of atherosclerosis and other cardiovascular conditions. In summary, the inverse associations of A. equolifaciens with these specific inflammatory markers further reveals its potential anti-inflammatory effects, possibly resulting in promotion of vascular health. Our findings suggest that A. equolifaciens may play a role in mitigating inflammation-related aspects of host metabolic disorders.

We found protective associations between A. equolifaciens and plaque in both men with and without HIV, underscoring a potential cardiovascular protective effect of A. equolifaciens regardless of HIV serostatus. We also evaluated associations between A. equolifaciens and conventional CVD risk factors. Notably, A. equolifaciens was positively correlated with HDL cholesterol and inversely correlated with SBP, two well-established markers of cardiovascular health, further supporting its proposed protective function. Consistent with our findings, a recent population-based study also demonstrated that genus Adlercreutzia is associated with a reduced risk of hypertension.⁶³ Taken together, these findings highlight the potential of A. equolifaciens and its associated metabolites in modulating systemic inflammation and reducing CVD risk. Further investigations are warranted to explore the causal pathways linking this bacterium to vascular health, and to evaluate whether modulation of A. equolifaciens or its functional metabolic outputs could serve as a therapeutic strategy for preventing or managing atherosclerosis and related cardiovascular diseases.

Our study also identified another gut bacterial species, Eubacterium sp3131, which was inversely associated with carotid artery plaque in both men living with or at risk for HIV. The genus Eubacterium represents a diverse group of anaerobic bacteria commonly found in the human gut, with several species previously linked to beneficial impacts on cardiovascular health.⁶⁴ For example, Eubacterium eligens and Eubacterium hallii are known producers of short-chain fatty acids (SCFAs),^64,65 which exert anti-inflammatory effects, while Eubacterium coprostanoligenes has been shown to reduce cholesterol levels by converting cholesterol to coprostanol, a less absorbable sterol.⁶⁶ In our study, Eubacterium sp3131 appeared to possess the capacity to generate the protective metabolite mesobilirubinogen from bilirubin. As mentioned earlier, mesobilirubinogen was significantly inversely associated with carotid artery plaque, suggesting a common metabolic pathway that may contribute to vascular protection. We observe significant negative associations between Eubacterium sp3131 and specific inflammatory markers, including plaque-associated CXCL9 and LY9. These findings provide important clues for elucidating the anti-inflammatory properties of Eubacterium sp3131 and its potential protective mechanisms against atherosclerosis.

Another interesting finding of this study is the positive association between Coprococcus sp13142 and carotid artery plaque in both men living with or without HIV. Coprococcus sp13142 is a Gram-positive, obligately anaerobic bacterial species. Although prior studies of this species are limited and direct evidence linking it to CVD is scarce, the broader Coprococcus genus has been implicated in cardiovascular health. In support of our findings, the Bogalusa Heart Study in Louisiana, based on 16S rRNA gene sequencing, reported associations between the relative abundance of Coprococcus and lifetime CVD risk profiles, suggesting that variation in Coprococcus abundance may influence long-term cardiovascular outcomes.⁶⁷ However, there is also evidence that other species of the Coprococcus genus, such as Coprococcus eutactus, can produce SCFAs like butyrate.⁶⁸ These controversial reports highlight the functional complexity within the Coprococcus genus, where different species may exert divergent effects on host CVD development. Further research is warranted to elucidate the potential complex roles of different Coprococcus species in the pathogenesis of CVD and to reveal underlying mechanisms that may explain these effects.

In our previous study of women with HIV, we observed positive associations between Fusobacterium (particularly F. nucleatum) and carotid plaque.^20,21 Similarly, in the present study of men, we observed a marginal positive association between F. nucleatum and plaque. We previously showed that this association could be partially explained by the inflammatory marker CXCL9. Consistent with that, in the current study, CXCL9 was also positively associated with plaque in men. These findings, together with the findings on A. equolifaciens, suggested some consistency for women and men in plaque-associated microbial taxa and microbial-related inflammatory markers in the context of HIV. Given the differences in socioeconomic and demographic between these populations, the consistency of findings across these studies further supports the robustness of these associations.

Our study has several limitations. Given its observational nature, causal inference cannot be established without additional evidence. Although the three -omics measurements (microbiome, metabolome, and proteome) were obtained at a closely aligned time, the assessment of carotid artery plaque occurred prior to the measurement of multi-omics data. However, we performed a sensitivity analysis using a previously established prediction model for carotid artery plaque within this cohort to address this potential source of bias. The limited sample size, particularly among men without HIV, may have restricted our ability to adequately assess effect modification by HIV serostatus. Since our findings were derived from a cohort of men with or at risk of HIV, they should be interpreted with caution, as exploratory and hypothesis-generating, and validation in independent cohorts and other populations is warranted.

In summary, this study identified altered gut microbial species, circulating metabolites, and plasma inflammatory markers in relation to carotid artery atherosclerosis among men with or at risk for HIV. A. equolifaciens and Eubacterium sp3131, which were associated with several beneficial plasma metabolites such as mesobilirubinogen, exhibited anti-inflammatory properties and were protectively associated with carotid plaque. Notably, we found that certain microbiota-related metabolites and inflammatory markers could partially explain the observed microbial associations with atherosclerosis. These findings offer new evidence and insights into the interrelationships among the gut microbiome, host inflammation, metabolism, and subclinical cardiovascular disease among people with HIV. If validated, this work may help inform future research and interventions targeting gut microbiota and their functional outputs as potential therapeutic avenues for CVD prevention in individuals with or at risk for HIV.

Contributors

Study concept and design: D.B.H., Z.W., Q.Q. and R.C.K. Cohort coordination and acquisition of data: D.B.H., A.S., K.A., Z.W., Q.Q., B.A.P., R.C.K., T.S., C.S., W.S.P., T.T.B., H.N.H., C.B.C., and R.D.B. Data access and verification: D.B.H., Z.W and Y.W. Analysis and interpretation of data: Z.W., Y.W., D.B.H., Q.Q., and R.C.K. Drafting of the manuscript: Z.W. and D.B.H. Critical revision of the manuscript for important intellectual content: All authors. Obtained funding: D.B.H., R.C.K., R.D.B., Z.W. and Q.Q. Study supervision: R.C.K., R.D.B., Q.Q. and D.B.H. All authors read and approved the final version of the manuscript.

Data sharing statement

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).

Declaration of interests

T.T.B. has served as a consultant to ViiV Healthcare, Merck, EMD-Serono, GSK, and Jannsen.

F.J.P. has served as a consultant and/or on the Speakers’ Bureau for ViiV Healthcare, Gilead Sciences, Merck, and EMD-Serono.

B.E.S. has served as a consultant to Gilead Sciences.

Other authors have nothing to disclose.