Skip to main content
  • Research article
  • Open access
  • Published:

Genetic architecture of cardiometabolic risks in people living with HIV

A Correction to this article was published on 05 May 2021

This article has been updated



Advances in antiretroviral therapies have greatly improved the survival of people living with human immunodeficiency virus (HIV) infection (PLWH); yet, PLWH have a higher risk of cardiovascular disease than those without HIV. While numerous genetic loci have been linked to cardiometabolic risk in the general population, genetic predictors of the excessive risk in PLWH are largely unknown.


We screened for common and HIV-specific genetic variants associated with variation in lipid levels in 6284 PLWH (3095 European Americans [EA] and 3189 African Americans [AA]), from the Centers for AIDS Research Network of Integrated Clinical Systems cohort. Genetic hits found exclusively in the PLWH cohort were tested for association with other traits. We then assessed the predictive value of a series of polygenic risk scores (PRS) recapitulating the genetic burden for lipid levels, type 2 diabetes (T2D), and myocardial infarction (MI) in EA and AA PLWH.


We confirmed the impact of previously reported lipid-related susceptibility loci in PLWH. Furthermore, we identified PLWH-specific variants in genes involved in immune cell regulation and previously linked to HIV control, body composition, smoking, and alcohol consumption. Moreover, PLWH at the top of European-based PRS for T2D distribution demonstrated a > 2-fold increased risk of T2D compared to the remaining 95% in EA PLWH but to a much lesser degree in AA. Importantly, while PRS for MI was not predictive of MI risk in AA PLWH, multiethnic PRS significantly improved risk stratification for T2D and MI.


Our findings suggest that genetic loci involved in the regulation of the immune system and predisposition to risky behaviors contribute to dyslipidemia in the presence of HIV infection. Moreover, we demonstrate the utility of the European-based and multiethnic PRS for stratification of PLWH at a high risk of cardiometabolic diseases who may benefit from preventive therapies.

Peer Review reports


The number of people living with human immunodeficiency virus (HIV) infection (PLWH) worldwide has increased by 34.6% (from 27.4 million to 36.9 million) between 2000 and 2018, while acquired immune deficiency syndrome (AIDS)-related deaths have declined from 1.5 million to 940,000 annually [1]. These advances can be primarily attributed to therapeutic advances in antiretroviral therapy (ART) and improved access to ART, allowing PLWH to live longer. However, accumulating evidence suggests that PLWH are at a higher risk of cardiovascular diseases (CVD) and have increased CVD-related mortality rates than those without HIV [2,3,4,5,6]. The possible causes of increased CVD risk among PLWH include inflammation and immune activation in response to HIV infection and viremia, adverse effects of ART, and lifestyle risk factors (e.g., smoking, alcohol, and illicit drug use). However, these factors do not fully account for the increased risk of CVD in PLWH [7, 8].

Genetic variants have been identified as significant predictors of traditional CVD risk factors including cardiometabolic traits and diseases, such as dyslipidemia and lipid levels (low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), and triglycerides) [9, 10], obesity [11, 12], type 2 diabetes mellitus (T2D) [13], and myocardial infarction (MI) [14] in the general population. CVD and related disorders have been demonstrated to have polygenic modes of inheritance, meaning that common genetic variants with small effect sizes located in multiple genes contribute to variability in disease or trait risk [15, 16]. Polygenic risk scores (PRS) have been proposed to assess the cumulative burden of multiple common susceptibility loci [17, 18]. A recent study found that 8% of the population possesses a genetic predisposition that confers a more than three-fold increased risk for coronary artery disease (CAD), with the highest PRS percentiles identifying 20 times more people than found by familial hypercholesterolemia mutations at a comparable or higher risk [19,20,21]. Moreover, in randomized clinical trials, people with the highest burden of genetic risk demonstrated the most substantial clinical benefit from primary prevention (statin therapy) resulting in a roughly three-fold decrease in the number needed to treat to prevent one CAD event [22].

Despite the growing literature proposing the clinical value of PRS in the general population [23], only a few reports with limited sample sizes have demonstrated the contribution of genetic variation to cardiometabolic risk in PLWH [24,25,26]; even fewer have examined the utility of PRS in PLWH [27]. Therefore, this study aimed to identify genetic predictors of cardiometabolic traits in PLWH and systematically assess the performance of PRS derived using results from previously published well-powered genome-wide association studies (GWAS) of T2D [28], CAD [29, 30], lipids (LDL, HDL, and triglyceride levels) [31], and body mass index (BMI) [32], and genomic data from the largest ethnically diverse PLWH cohort to date with genetic information. Given the emerging interest in applying PRS to improve clinical decision making [33], this study may help shed light on the genetic predictors of cardiometabolic risk in the presence of HIV infection and improve risk stratification to identify individuals at a high risk of CVD.


Study participants

The Centers for AIDS Research Network of Integrated Clinical Systems (CNICS) cohort includes a multiethnic population of ~ 36,000 PLWH (age 18 years and older) who have received routine clinical care at one of eight sites in the USA [34]. CNICS has an ongoing genetics project in which adult PLWH across racial/ethnic backgrounds from all sites, who provided informed consent and contributed specimens to the CNICS biospecimen repository, are being genotyped. Study participants were included if their genetic data were available at the time of these analyses.

Measurement of cardiometabolic phenotypes

The CNICS data repository integrates comprehensive clinical data from sites from outpatient and inpatient encounters, including information on demographic characteristics, clinical and laboratory data, medications, and historical clinical information. Lipid levels in CNICS include HDL, LDL, and triglyceride values measured as part of routine care and, therefore, may or may not have been obtained in the fasting state. LDL was either measured directly or calculated using the Friedewald equation [35]. BMI was calculated from heights and weights as a continuous variable (kg/m2). PLWH were categorized as ART-naïve or experienced. Among participants, the initial CNICS visit dates ranged from 1995 to 2015. Between the initial and the last CNICS visits, the average follow-up period was 10.3 years (median, 9.9 years; range, 0–23 years). Most included PLWH had multiple recorded values for each lipid drawn as part of care, we used mean values. We excluded individuals who were taking lipid-lowering drugs (e.g., HMG Co-A reductase inhibitors or statins) at baseline.

T2D diagnosis in CNICS is based on the following criteria: (1) hemoglobin A1c ≥ 6.5; (2) use of a diabetes-specific medication such as insulin; or (3) use of a diabetes-related medication, which is frequently, but not exclusively, used to treat diabetes (e.g., biguanides) in the setting of also having a diabetes diagnosis [36]. We have found high sensitivity (99%) and specificity (97%) for this definition [36].

CNICS uses an established state-of-the-art approach to adjudicate [37, 38] and classify MIs based on the Universal Definition of MIs [39, 40]. Potential MIs in the centralized CNICS data repository were identified using a comprehensive set of MI diagnostic and procedure codes and elevated cardiac biomarker values to optimize the ascertainment sensitivity as previously described [37, 38]. De-identified packets were prepared that contained provider notes, electrocardiograms, laboratory reports, and results from imaging and procedures, such as cardiac catheterization. Two physicians with expertise in adjudicating cardiac events performed a centralized review of the patient data, followed by inputs from a third physician for resolving discrepancies. We included type 1 MIs, those due to atheroembolic disease, and excluded type 2 MIs due to a mismatch in the oxygen supply and demand, usually observed in the setting of sepsis or cocaine or other illicit drug-induced vasospasm [37].

Genotyping and imputation

DNA was isolated from peripheral blood mononuclear cells or buffy coats of PLWH obtained from the CNICS biorepository using the FlexiGene DNA kit (Qiagen, #51206). DNA samples were then normalized and genotyped using Illumina’s high-density custom Multiethnic Global Array (MEGA) series BeadChips. Genotyped variant calling was performed using GenomeStudio® Genotyping Module v2.0 software (Illumina®, San Diego, California, USA) and zCall [41]. PLINK v.1.9 was used to exclude single nucleotide polymorphisms (SNPs) with call rates < 95%, minor allele frequency < 1%, and deviation from Hardy-Weinberg equilibrium (p value <1E-5), as well as samples with call rates < 90%, sex discrepancies between genotype data and self-report, and pairwise identity-by-descent (pi-hat > 0.9) [42].

We inferred ethnicity on genotype data using GRAF-pop software [43], and, after excluding the human leukocyte antigen encoding region, performed principal components analysis (PCA) on the African American (AA) and European American (EA) samples separately using EIGENSOFT [44]. The estimated principal components (PCs) were included in the regression models while performing genome-wide association analysis in each ancestry group. Genotype data from each ancestry group was imputed separately using the cloud-based Michigan Imputation Server [45] and Trans-Omics for Precision Medicine, or TOPMed data, as the reference panel ( For further analysis, we only kept variants that were imputed with high quality (imputation quality score, r2 > 0.3) and passed the standard quality control procedures. The genotyped and imputed SNP counts are listed in Additional file 1: Table S1.

Genome-wide association analysis

Genome-wide association tests were conducted on each SNP using either linear or logistic regression method on imputed dosage data sets, using in-house code written in R (version 3.5.3). The tests were performed separately in European and African ancestry sub-cohorts, and then pooled using random-effects meta-analysis, implemented in the “meta” R package [46]. In addition to the first ten PCs, analyses were adjusted for site, age, sex at birth, and presence or absence of ART. A study reported that genetic associations with lipid traits differed by sex [47]; therefore, we repeated these analyses in male and female sub-cohorts separately. The results were visualized through multi-phenotype and single-phenotype mirrored Manhattan plots. HIV-specific genetic variants were defined as loci that were significant at p < 0.01 in GWASHIV and had p ≥ 0.05 in the well-powered GWASGEN, and the 99% confidence intervals (CI) for the beta coefficients in GWASHIV and GWASGEN did not overlap. Similar approach was used to detect ancestry-specific or sex-specific lipid-related variants.

Gene set enrichment analysis

Enrichr was used to perform gene set enrichment analyses using the genes containing HIV-specific variants. Enrichr database is an integrative web-based application, currently containing 335,434 annotated gene sets from 166 gene set libraries [48, 49]. UK Biobank consists of a large prospective cohort of more than 500,000 middle-aged participants with detailed information on a wide range of complex diseases, lifestyle risk factors, medical history, and physical measurements [50]. The health outcomes were adjudicated by experts for a range of disease areas. The genetic data and statistical analyses were synchronized across multiple phenotypes. We looked for enrichment in the UK Biobank GWAS version 1 ( gene set library which contains 857 terms covering 14,148 genes (122 genes per term). Adjusted p values calculated using the false discovery rate (FDR) for correction for multiple hypotheses testing [51] were reported for each term. An adjusted p < 0.05 was considered statistically significant.

Expression quantitative trait loci (eQTL) analysis

To assess the functional relevance of the newly observed associations, we tested whether HIV-specific loci are enriched among variants shown to regulate gene expression (eQTLs). We acquired eQTL data in primary CD14+ human monocytes from 432 European volunteers at baseline and after exposure to the inflammatory proxies interferon-γ (IFN-γ) or differing durations (2 h or 24 h) of lipopolysaccharide (LPS), which was profiled using the Illumina Human OmniExpress BeadChips genotyping array [52]. SNPs that were significantly associated with each trait at p < E−6 in GWASGEN of lipid profiles were excluded [31]. Furthermore, linkage disequilibrium (LD)-based pruning was performed using a threshold of r2 > 0.2. After variant-filtering, we used chi-squared tests to compare the proportion of the eQTL SNPs (eSNPs) that were associated with gene expression levels at 10% FDR, among the HIV-specific loci to the remaining non-significant SNPs.

Polygenic risk score analysis

Traditional PRS

The PRS, representing estimated genetic determinants for five traits (HDL, LDL, triglycerides, T2D, and type 1 MI) were computed following the thresholding-pruning procedure [53]. We computed PRS for EA sub-cohort of PLWH (PLWHEA) and AA sub-cohort of PLWH (PLWHAA) separately using linear combinations of the imputed genotype dosages [54], and regression coefficients from the respective summary association statistics retrieved from previously published GWAS conducted in the general population largely of European ancestry: Global Lipids Genetics Consortium (GLGC) [31]; Genetic Investigation of ANthropometric Traits (GIANT) consortium [32]; DIAbetes Genetics Replication And Meta-analysis (DIAGRAM) consortium [28]; Coronary ARtery DIsease Genome wide Replication and Meta-analysis plus the Coronary Artery Disease Genetics (CARDIoGRAMplusC4D) consortium [29]; and UKBiobank CardioMetabolic Consortium [30] (PRSGEN, Additional file 1: Table S2). For each disease/trait, we calculated eight sets of PRS using GWAS p value thresholds of 1E−1, 1E−2, 1E−3, 1E−4, 1E−5, 1E−6, 1E−7, and 1E−8 for including SNPs in the PRS derivation. Prior to the calculation for each threshold, the retrieved SNPs underwent LD-based pruning using the 1000 Genomes European and African reference populations [55] as implemented in PLINK, and highly redundant SNPs (r2 ≥ 0.5) were removed (see Additional file 1: Table S3 for the number of SNPs used to calculate each PRS). For each p value threshold, we tested associations between PRS from previously reported GWAS (Additional file 1: Table S2) and the trait of interest or disease case status and visualized it using a heatmap.

Multiethnic PRS

To derive PRS that would perform well for both PLWHEA and PLWHAA, we considered GWAS summary statistics from two training sources: (1) the GWAS conducted in the general population of European ancestry (PRSEA) and (2) the GWAS conducted in PLWHAA (PRSAA), using ten-fold cross-validation. Additionally, we derived multiethnic PRS (Additional file 1: Table S3) that combined the two training sources using a recently published method [56]. Briefly, the multiethnic PRS is defined as the linear combination of the two PRSs with mixing weights α1 and α2. That is,

$$ {\mathrm{PRS}}_{\mathrm{EA}+\mathrm{AA}}={\alpha}_1{\mathrm{PRS}}_{\mathrm{EA}}+{\alpha}_2{\mathrm{PRS}}_{\mathrm{AA}} $$

We estimated mixing weights α1 and α2 using validation data by fitting a linear regression model and computed adjusted R2 to account for the additional degree of freedom. We employed a ten-fold cross-validation, using 90% of the cohort to estimate GWAS regression coefficients and the remaining 10% of the cohort to validate predictions (using the adjusted-R2 metric with best-fit mixture weights, \( {\hat{\alpha}}_1 \) and \( {\hat{\alpha}}_2 \)) and reported an average adjusted R2 across the ten cross-validations. For each fold, we computed regression coefficients using linear regression for quantitative traits while adjusting for 10 PCs, sex, age, age2, presence or absence of ART, and site, where the PCs were estimated using only PLWHAA. For T2D and MI diagnoses that had low prevalence in our cohort, we used stratified ten-fold cross-validation, where each cross-validation had the same case-control ratio. For lipid traits, for each p value threshold, we calculated the R2 statistic derived from a fixed-effects meta-analysis of marginal associations between PRSEA + AA and the trait of interest.

Lastly, we estimated the prevalence of T2D and MI for PLWH with the highest European-based and multiethnic PRS. We applied multiple testing correction to account for the number of thresholds and PRS tested using FDR [51]. An adjusted p < 0.05 was considered statistically significant. The number of SNPs used to calculate various multiethnic PRS is reported in Additional file 1: Table S3.


The final cohort consisted of 6284 PLWH with 3095 PLWHEA and 3189 PLWHAA; both sub-cohorts were predominantly male (89% and 69%, respectively), which is consistent with the HIV epidemic in the USA (Table 1). PLWHAA had a higher prevalence of T2D (p < 0.0001, Table 1), but lower mean LDL (p < 0.0001) and triglyceride (p < 0.0001) levels and higher mean HDL levels (p < 0.0001) than PLWHEA (Table 2).

Table 1 Baseline demographic and clinical characteristics of the study cohort
Table 2 Mean (standard deviation) and mean comparison p values for lipid values stratified by European American vs. African American race in the study cohort

Figure 1 summarizes GWAS results for HDL, LDL, and triglycerides in PLWHEA alongside previously reported findings in populations of European ancestry [31]. We confirmed strong associations exceeding genome-wide statistical significance of variation in APOE (apolipoprotein E), CETP (Cholesteryl Ester Transfer Protein) with HDL levels; APOE and APOC1 (apolipoprotein C1) with LDL levels, and APOA5 (apolipoprotein A5), BUD13 (BUD13 Homolog), and TRIB1 (Tribbles Pseudokinase 1) with triglyceride levels in PLWHEA (Fig. 1, top panel; Additional file 2: Table S4). Additional associations at p < 1E−5 in both HIV and no-HIV cohorts were detected in other previously reported lipid-related genes, including LIPC (Lipase C) and AQP9 (Aquaporin 9) for HDL; NECTIN2 (Nectin Cell Adhesion Molecule 2), CELSR2 (Cadherin EGF LAG Seven-Pass G-Type Receptor 2), PSRC1 (Proline And Serine Rich Coiled-Coil 1), APOC4-APOC2 (apolipoprotein C4, C2), and TOMM40 (Translocase Of Outer Mitochondrial Membrane 40) for LDL; and LPL (Lipoprotein Lipase), ZPR1 (Zinc Finger Protein 259), and SLC18A1 for triglycerides (Fig. 1, top panel; Additional file 2: Table S4). Furthermore, we identified variants that were significant in GWASHIV but not in GWASGEN, despite having sufficient statistical power (Fig. 1, bottom panel; Additional file 2: Table S5). Specifically, we identified 12 independent loci associated with HDL levels, including intronic variants in TMTC2, CYP2B6, GRM7, BARX2, IGF2BP1, CEMIP, TNFAIP8; 11 independent loci associated with LDL levels, including intronic variants in LBR, PRKG1, RCOR1, TNIP1, PRKAG2, and seven independent loci associated with triglyceride levels, including variants in SBK1, GPR156, and CPA6 (Additional file 3: Table S5). In a subgroup analysis of PLWHAA, in addition to replicating previously reported associations of APOE, TOMM40, and NECTIN2 with LDL, HERPUD1/CETP with HDL, and APOA5 with triglycerides at the genome-wide significance level, and of APOB, CELSR2, and LDLR with LDL and LPL, LIPC, and DOCK7 with triglyceride levels at p < E−5 (Fig. 2, top panel, Additional file 4: Table S6), we found lipid-related loci that were unique to PLWHAA (Fig. 2, bottom panel, Additional file 5: Table S7). Specifically, we identified 18 independent HIV-specific loci associated with HDL, 11 with LDL, and seven with triglyceride levels in PLWHAA at p < E−5, including intergenic variants in CPA6, previously associated with total cholesterol [57] and T2D [58] in individuals of African ancestry, and PRKG1 linked to body composition [59]. Lastly, we provide further evidence suggesting sex-specific effects of lipid-related SNPs. While none of these associations achieved genome-wide statistical significance (Additional file 6: Table S8), as a group, the corresponding genes were enriched in the visceral fat deposits and the metabolic syndrome pathways using BioCarta as implemented in Enrichr [49].

Fig. 1
figure 1

Multi-phenotype, mirrored Manhattan plot of genome-wide association analysis of lipid traits in Willer et al. [31] (top) and the CNICS European American (bottom) cohorts. HDL, high-density lipoproteins, LDL, low-density lipoproteins. In the top panel, gene names are listed for loci with association p < E−5 in both cohorts. In the bottom panel, gene names are listed for loci if p < 0.01 in the CNICS cohort and p > 0.05 in the Willer et al. cohort and there is no overlap between 99% confidence intervals for the corresponding beta coefficients

Fig. 2
figure 2

Multi-phenotype, mirrored Manhattan plot of genome-wide association analysis of lipid traits in Willer et al. [31] (top) and the CNICS African American (bottom) cohorts. HDL, high-density lipoproteins, LDL, low-density lipoproteins. In the top panel, gene names are listed for loci with association p < E−5 in both cohorts. In the bottom panel, gene names are listed for loci if p < 0.01 in the CNICS cohort and p > 0.05 in the Willer et al. cohort and there is no overlap between 99% confidence intervals for the corresponding beta coefficients

Gene set enrichment analysis

Gene set enrichment analysis was performed using genes containing HIV-specific susceptibility loci identified through GWASHIV of HDL (599 genes), LDL (595 genes), and triglycerides (678 genes). We identified several significantly enriched terms in the UK Biobank GWAS (version 1) gene set library (Fig. 3). Several top enriched terms were associated with blood cell counts, body composition, fat measurements and distribution, hypertension, diabetes, mood changes, and behavioral risk factors, such as alcohol dependence and smoking. Several of these enriched terms were statistically significant in all three gene set enrichment analyses, i.e., using HIV-specific variants from GWASHIV of HDL, LDL, and triglycerides (Fig. 3).

Fig. 3
figure 3

Gene set enrichment analysis of HIV-specific susceptibility loci. Statistical overrepresentation of HIV-specific variants (GWASHIV p < 0.01, GWASGEN p > 0.05, and no overlap between 99% confidence intervals of the corresponding beta coefficients) from GWASHIV of HDL, LDL, and triglycerides was tested among numerous phenotype terms in the UK Biobank GWAS (version 1) gene set library. The y-axis is the negative log10 of the adjusted p values for each enriched gene set term. The adjusted p values were calculated using the Benjamini-Hochberg method for correction for multiple hypotheses testing

Expression quantitative trait loci

Given the association between HIV-specific lipid-related loci and immune cell counts (Fig. 3), we compared the proportion of eSNPs among the HIV-specific SNPs with the proportion of eSNPs among all remaining SNPs in various CD14+ monocyte eQTL data sets (at basal condition, IFN-γ-induced, LPS-induced for 2-h, and LPS-induced for 24-h). The eSNPs were significantly enriched among the HIV-specific SNPs for HDL and LDL (p < 0.01) for all conditions except for basal condition for LDL SNPs (Additional file 1: Table S9 and Fig. S1). For triglycerides, the enrichment was significant only in the non-induced cells.

PRS analysis

We first tested the association of various lipid levels and risk of MI or T2D in CNICS patients with PRS for corresponding traits and diseases derived from GWASGEN (Additional file 1: Table S2) at eight different GWAS p value thresholds. We detected highly significant correlations between PRS for lipid traits (HDL, LDL, and triglycerides) and corresponding phenotypes (e.g., PRSHDL and plasma HDL; Fig. 4). Furthermore, as expected, measured HDL levels were inversely correlated with PRS for LDL, triglycerides, and CAD. Measured LDL levels were positively associated with PRS for CAD and PRS for MI. T2D diagnosis was associated with higher PRS for BMI and CAD. There was a trend toward higher PRS for LDL associated with the risk of MI diagnosis.

Fig. 4
figure 4

Heat map of polygenic risk scores in the CNICS HIV cohort (European American and African American sub-cohorts combined). The scores were generated using various p value cutoffs and SNP-level effect estimates from previously published genome-wide association analyses for each trait/disease phenotype and genotyped and imputed data from the CNICS HIV cohort. The associations marked with “▲” are significant at 10% false discovery rate

For each lipid trait, we compared the variance explained (adjusted R2) by the PRSGEN [31] versus multiethnic PRSHIV separately in PLWHEA and PLWHAA (Fig. 5). PRSGEN explained up to 6% of the genetic variance in PLWHEA (Fig. 5a, x-axis), but only up to 4% in the PLWHAA sub-cohort (Fig. 5b, x-axis). Among the lipid traits, the largest variance explained by PRSGEN was for HDL in PLWHEA and for LDL in PLWHAA, whereas the smallest was for triglycerides. Moreover, in PLWHAA, using the multiethnic PRSHIV increased the R2 for LDL across all p value thresholds and for HDL, especially when variants with more stringent p values were included. In PLWHEA, PRSHIV performed as well as PRSGEN, with the highest R2 recorded for HDL across most of p value thresholds (Fig. 5).

Fig. 5
figure 5

Scatter plot comparing mean variance explained (R2) by polygenic risk scores (PRS) for lipid traits in African American and European American people living with HIV. y-axis: multiethnic PRS derived in HIV cohort. x-axis: PRS derived in the general population of European ancestry [31]. a European American PLWH. b African American PLWH. HDL, high-density lipoproteins; LDL, low-density lipoproteins; Trig, triglycerides

Lastly, to determine the predictive value of different PRS in the presence of HIV infection, we estimated the risk of T2D and MI among PLWH with the highest PRSGEN (PRSGEN for T2D and PRSGEN for MI, respectively) or the highest multiethnic PRSHIV (PRSHIV for T2D and PRSHIV for MI, respectively). For T2D, PLWHEA at the top 5% of PRSGEN_T2D had an up to 2.14-fold increased risk depending on the GWAS p value threshold used for derivation compared to the remaining 95% (Fig. 6, Additional file 7: Table S10). Stratification based on PRSGEN for T2D was unable to distinguish PLWHAA at higher risk of T2D. However, PLWHAA at the top 5% of the multiethnic PRSHIV_T2D had an up to 2.35-fold increased risk (Additional file 7: Table S10). Importantly, although PRSGEN for MI was not predictive of MI risk in PLWHAA, patients at the top 5–30% of the multiethnic PRSHIV for MI had a consistently increased risk of MI at various GWAS p value thresholds (Additional file 7: Table S10). Neither PRSGEN nor PRSHIV demonstrated any predictive ability for MI risk in PLWHEA.

Fig. 6
figure 6

Risk stratification for various polygenic risk score thresholds in European American and African American people living with HIV. OR, odds ratio. PRS-EUR, polygenic risk score derived based on the regression coefficients estimated in a European ancestry population [31]. PRS-Multi, multiethnic PRS. T2D, type 2 diabetes. MI, myocardial infarction. Asterisks denote ORs with false discovery rate-adjusted p < 0.05


In the largest genetic study in an ethnically diverse cohort of PLWH to date, we confirmed the role of numerous susceptibility loci previously associated with lipid levels in the general population of European descent [31]. In addition, we detected variants uniquely associated with lipid traits in GWASHIV and not in the large well-powered GWASGEN of 188,577 individuals [31]. These HIV-specific loci were particularly enriched in eQTLs in basal and induced monocytes and associated with blood cell counts, body metabolism, mood disorders, and predisposition to risky behaviors. Lastly, we demonstrated a predictive value of PRS derived from GWASGEN in stratifying PLWHEA to distinguish individuals at a higher risk of developing T2D, while top percentiles of multiethnic PRS derived from GWASHIV and not PRSGEN were associated with increased risk of T2D or MI in PLWHAA.

Earlier targeted genotyping studies in general population have reported the role of genome-wide significant susceptibility loci in cardiometabolic traits in PLWH. Specifically, GWAS-validated SNPs in the APOE, APOB, LDLR, and other genes have been demonstrated to contribute to dyslipidemia in the presence of HIV infection [60]. Also, several SNPs and genetic regions common across HIV-positive and HIV-negative women have been detected in association with carotid artery intima-media thickness, a subclinical marker of atherosclerosis [61]. In a series of unbiased GWAS of lipid traits, we confirmed genetic association with previously reported variants in several apolipoprotein-coding genes (APOE, APOC1, APOC2, APOC4, and APOA5), CETP, LPL, BUD13, AQP9, and CELSR2, among many others (Fig. 1, Additional file 2: Table S4).

Additionally, we detected numerous loci that were associated with lipid traits in the PLWHEA, but showed no significant signal in the large lipid GWAS conducted in a cohort of European ancestry [31] (Fig. 1, Additional file 3: Table S5). A few small GWAS studies performed in HIV-infected cohorts have identified loci associated with carotid atherosclerosis [26], subcutaneous adipose tissue volume [25], and fat loss [24]. In our study, many of the lipid-related susceptibility loci identified in GWASHIV were also linked by previous studies to HIV viral load [62], susceptibility [63], control [64], smoking behavior [65,66,67], alcohol dependence [64, 65, 68,69,70], and cannabis dependence [71,72,73], more common in PLWH than in individuals without HIV, suggesting the contribution of additional genetic variants associated with HIV infection and adverse lifestyle behaviors to dyslipidemia in this population. Importantly, HIV-specific lipid-related variants were also significantly enriched among the loci associated with blood cell counts, body composition, lifestyle risk factors (alcohol dependence and smoking), and mood disorders (Fig. 3). These findings are consistent with previous reports showing a positive correlation between lymphocyte count and LDL cholesterol levels [74]. Moreover, a shared link has been established between CAD risk and reticulocyte indices, where increased hemolysis associated with high reticulocyte counts may lead to oxidative stress and inflammation [75]. Additionally, a longitudinal relationship of depressive and anxiety symptoms with dyslipidemia and abdominal obesity has been reported [76], which can be partially explained by chronic low-grade inflammation and smoking [77]. While HIV-associated chronic inflammation has long been considered a risk factor of CVD in PLWH [78], our findings suggest that genetic variants may lead to further immune perturbations that contribute to cardiometabolic risk, especially in the presence of HIV infection. Furthermore, when we screened eQTLs in basal and induced CD14+ monocytes of healthy volunteers of European ancestry [52] for the presence of HIV-specific loci, we found significant enrichment for lipid-associated variants, further supporting a functional role of these loci in gene expression regulation of dyslipidemia in the presence of HIV infection. Validation in an independent cohort will be needed to verify the effect of HIV-specific loci on cardiometabolic diseases.

We conducted subgroup analyses to identify lipid-related genetic loci that are unique to PLWHAA (Fig. 2) or act in a sex-specific manner (Additional file 6: Table S8). While none of the associations reached genome-wide significance, we identified a number of genes that have been previously associated with total cholesterol [57] and T2D [58] in individuals of African ancestry, or linked to body composition [59]. The sex-specific genes as a group were enriched in the visceral fat deposit and the metabolic pathways. Additional analyses will be required to dissect the ancestry and sex-specific effects of these variants on metabolic traits in the presence of HIV infection.

Given the polygenic nature of CAD and its numerous risk factors, PRS-based assessment of the genetic burden across multiple susceptibility loci has demonstrated greater predictive value for disease risk and drug response than individual variants [33]. A recent study in a non-HIV cohort has shown that the CAD risk associated with a high polygenic load for lipid-increasing variants was proportional to their impact on lipid levels [79]. We showed a significant correlation of PRS for lipid traits, T2D, and MI generated based on the large European GWASGEN (Additional file 1: Table S2) with respective phenotypes in PLWH (Fig. 4). Similar to the general population, in PLWH, we observed a positive association of PRS for CAD and PRS for MI with LDL and a negative association with HDL. Our results suggest that lipid PRS could point to modifiable risk factors in the presence of HIV infection, providing additional guidance for clinical application.

However, the variance explained by PRS derived from general (predominantly European) populations in PLWHEA was > 30% lower than that explained in PLWHAA (~ 6% vs. < 4%). This finding is consistent with previous studies showing that PRS calculated using effect estimates from European GWAS were not generalizable to the African ancestry population [80]. Therefore, we calculated a multiethnic PRS, shown to significantly improve disease prediction accuracy in a non-European cohort [56], by applying weights in both EA and AA GWAS in CNICS using ten-fold cross-validation. Multiethnic PRSHIV outperformed PRSGEN in PLWHAA, especially for HDL, but not in PLWHEA (Fig. 5).

Of note, stratification based on PRSGEN for T2D was able to distinguish PLWH that were at a higher risk of T2D, with EA at the top 5% having a more than two-fold increased risk; the impact of PRSGEN for T2D on T2D risk in AA was less obvious (Fig. 6; Additional file 7: Table S10). A 2.75-fold increased risk of T2D in individuals of European ancestry at the top 5% of PRS for T2D has been previously reported [21]. However, the multiethnic PRS for T2D significantly improved T2D risk stratification in AA, but not in EA PLWH (Fig. 6).

In addition, while PRSGEN for MI was unable to significantly stratify MI risk in either ethnic subgroup, multiethnic PRSHIV demonstrated over a 3-fold increased risk in PLWHAA. Multiethnic PRSHIV for MI largely unchanged the disease risk prediction in PLWHEA. In a much larger European ancestry non-HIV cohort, a 1 standard deviation higher PRS is associated with a 33% increased risk of incident MI in participants without CAD [81]. Taken together, our findings suggest that, while the large GWAS in ethnically and racially diverse cohorts should substantially contribute to the accuracy of PRS prediction in PLWH, in the absence of such studies, multiethnic scores are feasible alternatives to identify at-risk individuals. Given that medications and intensive lifestyle interventions prevent or postpone the progression to T2D and MI [82, 83], ascertainment of PLWH with high PRS may provide an opportunity to target these interventions with increased precision.

This study has some limitations. In the general population-based cohorts used in our analyses, HIV infection-related information may not have been collected or considered during recruitment or analysis. Therefore, it is possible to have an unknown number of PLWH in these cohorts. However, the rate of HIV infection in the US population is relatively low (~ 1 in 300), and inclusion of such individuals in our analyses would bias the results toward the null. We controlled for ART presence or absence and made no distinctions across ART regimens. A thorough investigation of the effects of ART on lipids, which is a rapidly evolving field, is a big task and beyond the scope of the present analysis. Future investigations may be able to refine some of the work done in our study. We performed analyses of PRS for BMI but did not analyze the observed BMI. Many factors are associated with BMI among PLWH, including body morphology disorders and lifestyle, and fully analyzing these characteristics was beyond the scope of this study. Future work should elucidate relationships with the observed BMI. Additionally, we used the same cohort for multiethnic PRS derivation and validation; however, we do not expect over-fitting to be a concern given the small number of mixing weights optimized (up to 2) relative to the target sample size (> 3000) and given our use of adjusted R2 as the evaluation metric, similar to previously reported analyses [56]. In order to minimize the possibility of an inflated R2 prediction due to shared population stratification or familial/distant relatedness [84], we used ancestry-adjusted regression coefficients for PRS computation and ten-fold cross-validation. Despite being the largest genetic study reported in PLWH, the number of MI cases was too small to provide sufficient statistical power to assess the clinical impact of PRS. Nevertheless, we were able to demonstrate that the use of multiethnic PRS in PLWH outperformed PRS derived in largely European populations, especially for PLWHAA. Going forward, meta-analyses of PLWH cohorts should allow for validation of our findings and help assess the clinical impact of the genetic burden on disease risk.


In summary, we demonstrated that in addition to genetic loci in the lipid metabolism genes previously linked to dyslipidemia and other CAD-related risks in the general population, there are other genetic factors that can impact lipid levels by further enhancing inflammation and predisposing to mood disorders and risky behaviors, thereby contributing to dyslipidemia in the presence of HIV infection. Comprehensive polygenic risk profiling identified PLWH to be at a several-fold increased risk of T2D or MI, which may help increase the precision of ascertaining those at high risk for targeted interventions.

Availability of data and materials

The datasets generated and analyzed during the current study are available from the authors on reasonable request.

Change history



African American


Acquired immune deficiency syndrome


Antiretroviral therapy


Body mass index


Coronary artery disease


Centers for AIDS Research Network of Integrated Clinical Systems


Cardiovascular disease


European American


Expression quantitative trait locus


False discovery rate


Genome-wide association study


High-density lipoprotein cholesterol


Human immunodeficiency virus


Institutional Review Board


Kyoto Encyclopedia of Genes and Genomes


Linkage disequilibrium


Low-density lipoprotein cholesterol


Multiethnic Global Array


Myocardial infarction


Principal components analysis


People living with HIV


African American sub-cohort of people living with HIV


European American sub-cohort of people living with HIV


Polygenic risk score


Quality control


Single nucleotide polymorphism


Type 2 diabetes


  1. Global HIV and AIDS statistics - 2018 fact sheet [].

  2. Van Epps P, Kalayjian RC. Human immunodeficiency virus and aging in the era of effective antiretroviral therapy. Infect Dis Clin N Am. 2017;31(4):791–810.

    Article  Google Scholar 

  3. Smith CJ, Ryom L, Weber R, Morlat P, Pradier C, Reiss P, Kowalska JD, de Wit S, Law M, el Sadr W, et al. Trends in underlying causes of death in people with HIV from 1999 to 2011 (D:a:D): a multicohort collaboration. Lancet. 2014;384(9939):241–8.

    Article  PubMed  Google Scholar 

  4. Croxford S, Kitching A, Desai S, Kall M, Edelstein M, Skingsley A, Burns F, Copas A, Brown AE, Sullivan AK, et al. Mortality and causes of death in people diagnosed with HIV in the era of highly active antiretroviral therapy compared with the general population: an analysis of a national observational cohort. Lancet Public Health. 2017;2(1):e35–46.

    Article  PubMed  Google Scholar 

  5. Feinstein MJ, Bahiru E, Achenbach C, Longenecker CT, Hsue P, So-Armah K, Freiberg MS, Lloyd-Jones DM. Patterns of cardiovascular mortality for HIV-infected adults in the United States: 1999 to 2013. Am J Cardiol. 2016;117(2):214–20.

    Article  PubMed  Google Scholar 

  6. Freiberg MS, Chang CC, Kuller LH, Skanderson M, Lowy E, Kraemer KL, Butt AA, Bidwell Goetz M, Leaf D, Oursler KA, et al. HIV infection and the risk of acute myocardial infarction. JAMA Intern Med. 2013;173(8):614–22.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Mattevi VS, Tagliari CF. Pharmacogenetic considerations in the treatment of HIV. Pharmacogenomics. 2017;18(1):85–98.

    Article  CAS  PubMed  Google Scholar 

  8. Feinstein MJ, Hsue PY, Benjamin LA, Bloomfield GS, Currier JS, Freiberg MS, Grinspoon SK, Levin J, Longenecker CT, Post WS. Characteristics, prevention, and management of cardiovascular disease in people living with HIV: a scientific statement from the American Heart Association. Circulation. 2019;140(2):e98–e124.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. Teslovich TM, Musunuru K, Smith AV, Edmondson AC, Stylianou IM, Koseki M, Pirruccello JP, Ripatti S, Chasman DI, Willer CJ, et al. Biological, clinical and population relevance of 95 loci for blood lipids. Nature. 2010;466(7307):707–13.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. Sharma K, Baliga RR. Genetics of dyslipidemia and ischemic heart disease. Curr Cardiol Rep. 2017;19(5):46.

    Article  PubMed  Google Scholar 

  11. Goodarzi MO. Genetics of obesity: what genetic association studies have taught us about the biology of obesity and its complications. Lancet Diabetes Endocrinol. 2018;6(3):223–36.

    Article  CAS  PubMed  Google Scholar 

  12. Gonzalez-Muniesa P, Martinez-Gonzalez MA, Hu FB, Despres JP, Matsuzawa Y, Loos RJF, Moreno LA, Bray GA, Martinez JA. Obesity. Nat Rev Dis Primers. 2017;3:17034.

    Article  PubMed  Google Scholar 

  13. Ingelsson E, McCarthy MI. Human genetics of obesity and type 2 diabetes mellitus. Circ Genom Precis Med. 2018;11(6):e002090.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Kathiresan S, Srivastava D. Genetics of human cardiovascular disease. Cell. 2012;148(6):1242–57.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Timpson NJ, Greenwood CMT, Soranzo N, Lawson DJ, Richards JB. Genetic architecture: the shape of the genetic contribution to human traits and disease. Nat Rev Genet. 2018;19(2):110–24.

    Article  CAS  PubMed  Google Scholar 

  16. Torkamani A, Wineinger NE, Topol EJ. The personal and clinical utility of polygenic risk scores. Nat Rev Genet. 2018;19(9):581–90.

    Article  CAS  PubMed  Google Scholar 

  17. International Schizophrenia Consortium, Purcell SM, Wray NR, Stone JL, Visscher PM, O'Donovan MC, Sullivan PF, Sklar P. Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature. 2009;460(7256):748–52.

  18. Ripatti S, Tikkanen E, Orho-Melander M, Havulinna AS, Silander K, Sharma A, Guiducci C, Perola M, Jula A, Sinisalo J, et al. A multilocus genetic risk score for coronary heart disease: case-control and prospective cohort analyses. Lancet. 2010;376(9750):1393–400.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Nordestgaard BG, Chapman MJ, Humphries SE, Ginsberg HN, Masana L, Descamps OS, Wiklund O, Hegele RA, Raal FJ, Defesche JC, et al. Familial hypercholesterolaemia is underdiagnosed and undertreated in the general population: guidance for clinicians to prevent coronary heart disease: consensus statement of the European Atherosclerosis Society. Eur Heart J. 2013;34(45):3478–3490a.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Abul-Husn NS, Manickam K, Jones LK, Wright EA, Hartzel DN, Gonzaga-Jauregui C, O'Dushlaine C, Leader JB, Lester Kirchner H, Lindbuchler DM et al. Genetic identification of familial hypercholesterolemia within a single U.S. health care system. Science. 2016;354(6319):aaf7000.

  21. Khera AV, Chaffin M, Aragam KG, Haas ME, Roselli C, Choi SH, Natarajan P, Lander ES, Lubitz SA, Ellinor PT, et al. Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations. Nat Genet. 2018;50(9):1219–24.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Mega JL, Stitziel NO, Smith JG, Chasman DI, Caulfield M, Devlin JJ, Nordio F, Hyde C, Cannon CP, Sacks F, et al. Genetic risk, coronary heart disease events, and the clinical benefit of statin therapy: an analysis of primary and secondary prevention trials. Lancet. 2015;385(9984):2264–71.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Chen X, Yazdani S, Piehl F, Magnusson PKE, Fang F. Polygenic link between blood lipids and amyotrophic lateral sclerosis. Neurobiol Aging. 2018;67:202.e1–202.e6.

  24. Uttayamakul S, Oudot-Mellakh T, Nakayama EE, Tengtrakulcharoen P, Guergnon J, Delfraissy JF, Khusmith S, Sangsajja C, Likanonsakul S, Theodorou I, et al. Genome-wide association study of HIV-related lipoatrophy in Thai patients: association of a DLGAP1 polymorphism with fat loss. AIDS Res Hum Retrovir. 2015;31(8):792–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  25. Irvin MR, Shrestha S, Chen YD, Wiener HW, Haritunians T, Vaughan LK, Tiwari HK, Taylor KD, Scherzer R, Saag MS, et al. Genes linked to energy metabolism and immunoregulatory mechanisms are associated with subcutaneous adipose tissue distribution in HIV-infected men. Pharmacogenet Genomics. 2011;21(12):798–807.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Shrestha S, Irvin MR, Taylor KD, Wiener HW, Pajewski NM, Haritunians T, Delaney JA, Schambelan M, Polak JF, Arnett DK, et al. A genome-wide association study of carotid atherosclerosis in HIV-infected men. AIDS. 2010;24(4):583–92.

  27. Rotger M, Glass TR, Junier T, Lundgren J, Neaton JD, Poloni ES, van 't Wout AB, Lubomirov R, Colombo S, Martinez R, et al. Contribution of genetic background, traditional risk factors, and HIV-related factors to coronary artery disease events in HIV-positive persons. Clin Infect Dis. 2013;57(1):112–21.

  28. Morris AP, Voight BF, Teslovich TM, Ferreira T, Segre AV, Steinthorsdottir V, Strawbridge RJ, Khan H, Grallert H, Mahajan A, et al. Large-scale association analysis provides insights into the genetic architecture and pathophysiology of type 2 diabetes. Nat Genet. 2012;44(9):981–90.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Nikpay M, Goel A, Won HH, Hall LM, Willenborg C, Kanoni S, Saleheen D, Kyriakou T, Nelson CP, Hopewell JC, et al. A comprehensive 1,000 genomes-based genome-wide association meta-analysis of coronary artery disease. Nat Genet. 2015;47(10):1121–30.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Nelson CP, Goel A, Butterworth AS, Kanoni S, Webb TR, Marouli E, Zeng L, Ntalla I, Lai FY, Hopewell JC, et al. Association analyses based on false discovery rate implicate new loci for coronary artery disease. Nat Genet. 2017;49(9):1385–91.

    Article  CAS  PubMed  Google Scholar 

  31. Willer CJ, Schmidt EM, Sengupta S, Peloso GM, Gustafsson S, Kanoni S, Ganna A, Chen J, Buchkovich ML, Mora S, et al. Discovery and refinement of loci associated with lipid levels. Nat Genet. 2013;45(11):1274–83.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  32. Locke AE, Kahali B, Berndt SI, Justice AE, Pers TH, Day FR, Powell C, Vedantam S, Buchkovich ML, Yang J, et al. Genetic studies of body mass index yield new insights for obesity biology. Nature. 2015;518(7538):197–206.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Natarajan P, Young R, Stitziel NO, Padmanabhan S, Baber U, Mehran R, Sartori S, Fuster V, Reilly DF, Butterworth A, et al. Polygenic risk score identifies subgroup with higher burden of atherosclerosis and greater relative benefit from statin therapy in the primary prevention setting. Circulation. 2017;135(22):2091–101.

    Article  PubMed  PubMed Central  Google Scholar 

  34. Kitahata MM, Rodriguez B, Haubrich R, Boswell S, Mathews WC, Lederman MM, Lober WB, Van Rompaey SE, Crane HM, Moore RD, et al. Cohort profile: the centers for AIDS research network of integrated clinical systems. Int J Epidemiol. 2008;37(5):948–55.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem. 1972;18(6):499–502.

    Article  CAS  PubMed  Google Scholar 

  36. Crane HM, Kadane JB, Crane PK, Kitahata MM. Diabetes case identification methods applied to electronic medical record systems: their use in HIV-infected patients. Curr HIV Res. 2006;4(1):97–106.

    Article  CAS  PubMed  Google Scholar 

  37. Crane HM, Paramsothy P, Drozd DR, Nance RM, Delaney JA, Heckbert SR, Budoff MJ, Burkholder GA, Willig JH, Mugavero MJ, et al. Types of myocardial infarction among human immunodeficiency virus-infected individuals in the United States. JAMA Cardiol. 2017;2(3):260–7.

    Article  PubMed  PubMed Central  Google Scholar 

  38. Crane HM, Heckbert SR, Drozd DR, Budoff MJ, Delaney JA, Rodriguez C, Paramsothy P, Lober WB, Burkholder G, Willig JH, et al. Lessons learned from the design and implementation of myocardial infarction adjudication tailored for HIV clinical cohorts. Am J Epidemiol. 2014;179(8):996–1005.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Thygesen K, Alpert JS, White HD, Jaffe AS, Apple FS, Galvani M, Katus HA, Newby LK, Ravkilde J, Chaitman B, et al. Universal definition of myocardial infarction. Circulation. 2007;116(22):2634–53.

    Article  PubMed  Google Scholar 

  40. Thygesen K, Alpert JS, Jaffe AS, Simoons ML, Chaitman BR, White HD. Third universal definition of myocardial infarction. Eur Heart J. 2012;33(20):2551–67.

    Article  PubMed  Google Scholar 

  41. Goldstein JI, Crenshaw A, Carey J, Grant GB, Maguire J, Fromer M, O'Dushlaine C, Moran JL, Chambert K, Stevens C, et al. zCall: a rare variant caller for array-based genotyping: genetics and population analysis. Bioinformatics. 2012;28(19):2543–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience. 2015;4:7.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  43. Jin Y, Schaffer AA, Feolo M, Holmes JB, Kattman BL. GRAF-pop: a fast distance-based method to infer subject ancestry from multiple genotype datasets without principal components analysis. G3 (Bethesda). 2019;9(8):2447–61.

    Article  Google Scholar 

  44. Patterson N, Price AL, Reich D. Population structure and eigenanalysis. PLoS Genet. 2006;2(12):e190.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  45. Das S, Forer L, Schonherr S, Sidore C, Locke AE, Kwong A, Vrieze SI, Chew EY, Levy S, McGue M, et al. Next-generation genotype imputation service and methods. Nat Genet. 2016;48(10):1284–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Schwarzer G. Meta: an R package for meta-analysis. R news. 2007;7(3):40–5.

    Google Scholar 

  47. Taylor KC, Carty CL, Dumitrescu L, Buzkova P, Cole SA, Hindorff L, Schumacher FR, Wilkens LR, Shohet RV, Quibrera PM, et al. Investigation of gene-by-sex interactions for lipid traits in diverse populations from the population architecture using genomics and epidemiology study. BMC Genet. 2013;14:33.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  48. Chen EY, Tan CM, Kou Y, Duan Q, Wang Z, Meirelles GV, Clark NR, Ma'ayan A. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC bioinformatics. 2013;14:128.

    Article  PubMed  PubMed Central  Google Scholar 

  49. Kuleshov MV, Jones MR, Rouillard AD, Fernandez NF, Duan Q, Wang Z, Koplev S, Jenkins SL, Jagodnik KM, Lachmann A, et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res. 2016;44(W1):W90–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, Downey P, Elliott P, Green J, Landray M, et al. UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 2015;12(3):e1001779.

    Article  PubMed  PubMed Central  Google Scholar 

  51. Benjamini Y, Hochberg A. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J Royal Stat Soc. 1995;57(1):289–300.

    Google Scholar 

  52. Fairfax BP, Humburg P, Makino S, Naranbhai V, Wong D, Lau E, Jostins L, Plant K, Andrews R, McGee C, et al. Innate immune activity conditions the effect of regulatory variants upon monocyte gene expression. Science. 2014;343(6175):1246949.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  53. Choi SW, Shin Heng Mak T, O’Reilly PF: A guide to performing polygenic risk score analyses. BioRxiv 2020,

  54. Zhang J, Peng S, Cheng H, Nomura Y, Di Narzo AF, Hao K. Genetic Pleiotropy between nicotine dependence and respiratory outcomes. Sci Rep. 2017;7:16907.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  55. 1000 Genomes Project Consortium, Auton A, brooks LD, Durbin RM, Garrison EP, Kang HM, Korbel JO, Marchini JL, McCarthy S, McVean GA, et al. A global reference for human genetic variation. Nature. 2015;526(7571):68–74.

  56. Marquez-Luna C, Loh PR, South Asian Type 2 Diabetes Consortium, Sigma Type 2 D Consortium, Price AL. Multiethnic polygenic risk scores improve risk prediction in diverse populations. Genet Epidemiol. 2017;41(8):811–23.

  57. van Leeuwen EM, Sabo A, Bis JC, Huffman JE, Manichaikul A, Smith AV, Feitosa MF, Demissie S, Joshi PK, Duan Q, et al. Meta-analysis of 49 549 individuals imputed with the 1000 Genomes Project reveals an exonic damaging variant in ANGPTL4 determining fasting TG levels. J Med Genet. 2016;53(7):441–9.

    Article  PubMed  CAS  Google Scholar 

  58. Ng MC, Shriner D, Chen BH, Li J, Chen WM, Guo X, Liu J, Bielinski SJ, Yanek LR, Nalls MA, et al. Meta-analysis of genome-wide association studies in African Americans provides insights into the genetic architecture of type 2 diabetes. PLoS Genet. 2014;10(8):e1004517.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  59. Pulit SL, Stoneman C, Morris AP, Wood AR, Glastonbury CA, Tyrrell J, Yengo L, Ferreira T, Marouli E, Ji Y, et al. Meta-analysis of genome-wide association studies for body fat distribution in 694 649 individuals of European ancestry. Hum Mol Genet. 2019;28(1):166–74.

    Article  CAS  PubMed  Google Scholar 

  60. Rotger M, Bayard C, Taffe P, Martinez R, Cavassini M, Bernasconi E, Battegay M, Hirschel B, Furrer H, Witteck A, et al. Contribution of genome-wide significant single-nucleotide polymorphisms and antiretroviral therapy to dyslipidemia in HIV-infected individuals: a longitudinal study. Circ Cardiovasc Genet. 2009;2(6):621–8.

    Article  CAS  PubMed  Google Scholar 

  61. Shendre A, Wiener HW, Irvin MR, Aouizerat BE, Overton ET, Lazar J, Liu C, Hodis HN, Limdi NA, Weber KM, et al. Genome-wide admixture and association study of subclinical atherosclerosis in the Women's Interagency HIV Study (WIHS). PLoS One. 2017;12(12):e0188725.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  62. Ekenberg C, Tang MH, Zucco AG, Murray DD, MacPherson CR, Hu X, Sherman BT, Losso MH, Wood R, Paredes R, et al. Association between single-nucleotide polymorphisms in HLA alleles and human immunodeficiency virus type 1 viral load in demographically diverse, antiretroviral therapy-naive participants from the strategic timing of AntiRetroviral treatment trial. J Infect Dis. 2019;220(8):1325–34.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  63. Lingappa JR, Petrovski S, Kahle E, Fellay J, Shianna K, McElrath MJ, Thomas KK, Baeten JM, Celum C, Wald A, et al. Genomewide association study for determinants of HIV-1 acquisition and viral set point in HIV-1 serodiscordant couples with quantified virus exposure. PLoS One. 2011;6(12):e28632.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  64. Fellay J, Ge D, Shianna KV, Colombo S, Ledergerber B, Cirulli ET, Urban TJ, Zhang K, Gumbs CE, Smith JP, et al. Common genetic variation and the control of HIV-1 in humans. PLoS Genet. 2009;5(12):e1000791.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  65. Liu M, Jiang Y, Wedow R, Li Y, Brazel DM, Chen F, Datta G, Davila-Velderrain J, McGuire D, Tian C, et al. Association studies of up to 1.2 million individuals yield new insights into the genetic etiology of tobacco and alcohol use. Nat Genet. 2019;51(2):237–44.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  66. Brazel DM, Jiang Y, Hughey JM, Turcot V, Zhan X, Gong J, Batini C, Weissenkampen JD, Liu M, Consortium CHDE, et al. Exome chip meta-analysis fine maps causal variants and elucidates the genetic architecture of rare coding variants in smoking and alcohol use. Biol Psychiatry. 2019;85(11):946–55.

    Article  CAS  PubMed  Google Scholar 

  67. Park SL, Carmella SG, Chen M, Patel Y, Stram DO, Haiman CA, Le Marchand L, Hecht SS. Mercapturic acids derived from the toxicants acrolein and crotonaldehyde in the urine of cigarette smokers from five ethnic groups with differing risks for lung cancer. PLoS One. 2015;10(6):e0124841.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  68. Kapoor M, Wang JC, Wetherill L, Le N, Bertelsen S, Hinrichs AL, Budde J, Agrawal A, Almasy L, Bucholz K, et al. Genome-wide survival analysis of age at onset of alcohol dependence in extended high-risk COGA families. Drug Alcohol Depend. 2014;142:56–62.

    Article  PubMed  PubMed Central  Google Scholar 

  69. Evangelou E, Gao H, Chu C, Ntritsos G, Blakeley P, Butts AR, Pazoki R, Suzuki H, Koskeridis F, Yiorkas AM, et al. New alcohol-related genes suggest shared genetic mechanisms with neuropsychiatric disorders. Nat Hum Behav. 2019;3(9):950–61.

    Article  PubMed  PubMed Central  Google Scholar 

  70. de Vries PS, Brown MR, Bentley AR, Sung YJ, Winkler TW, Ntalla I, Schwander K, Kraja AT, Guo X, Franceschini N, et al. Multiancestry genome-wide association study of lipid levels incorporating gene-alcohol interactions. Am J Epidemiol. 2019;188(6):1033–54.

    Article  PubMed  PubMed Central  Google Scholar 

  71. Hubel C, Gaspar HA, Coleman JRI, Hanscombe KB, Purves K, Prokopenko I, Graff M, Ngwa JS, Workalemahu T, Consortium AWGotPG, et al. Genetic correlations of psychiatric traits with body composition and glycemic traits are sex- and age-dependent. Nat Commun. 2019;10(1):5765.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  72. Kranzler HR, Zhou H, Kember RL, Smith RV, Justice AC, Damrauer S, Tsao PS, Klarin D, Baras A, Reid J, et al. Author correction: genome-wide association study of alcohol consumption and use disorder in 274,424 individuals from multiple populations. Nat Commun. 2019;10(1):4050.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  73. Sanchez-Roige S, Fontanillas P, Elson SL, 23andMe Research Team, Gray JC, de Wit H, Davis LK, MacKillop J, Palmer AA. Genome-wide association study of alcohol use disorder identification test (AUDIT) scores in 20 328 research participants of European ancestry. Addict Biol. 2019;24(1):121–31.

  74. Oda E. Longitudinal associations between lymphocyte count and LDL cholesterol in a health screening population. J Clin Transl Endocrinol. 2014;1(2):49–53.

    CAS  PubMed  PubMed Central  Google Scholar 

  75. Astle WJ, Elding H, Jiang T, Allen D, Ruklisa D, Mann AL, Mead D, Bouman H, Riveros-Mckay F, Kostadima MA, et al. The allelic landscape of human blood cell trait variation and links to common complex disease. Cell. 2016;167(5):1415–29.e19.

  76. van Reedt Dortland AK, Giltay EJ, van Veen T, Zitman FG, Penninx BW. Longitudinal relationship of depressive and anxiety symptoms with dyslipidemia and abdominal obesity. Psychosom Med. 2013;75(1):83–9.

    Article  PubMed  CAS  Google Scholar 

  77. van Reedt Dortland AK, Vreeburg SA, Giltay EJ, Licht CM, Vogelzangs N, van Veen T, de Geus EJ, Penninx BW, Zitman FG. The impact of stress systems and lifestyle on dyslipidemia and obesity in anxiety and depression. Psychoneuroendocrinology. 2013;38(2):209–18.

    Article  PubMed  CAS  Google Scholar 

  78. Barnes RP, Lacson JC, Bahrami H. HIV infection and risk of cardiovascular diseases beyond coronary artery disease. Curr Atheroscler Rep. 2017;19(5):20.

    Article  PubMed  PubMed Central  Google Scholar 

  79. Ripatti P, Ramo JT, Mars NJ, Fu Y, Lin J, Soderlund S, Benner C, Surakka I, Kiiskinen T, Havulinna AS, et al. Polygenic hyperlipidemias and coronary artery disease risk. Circ Genom Precis Med. 2020;13(2):e002725.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  80. Martin AR, Gignoux CR, Walters RK, Wojcik GL, Neale BM, Gravel S, Daly MJ, Bustamante CD, Kenny EE. Human demographic history impacts genetic risk prediction across diverse populations. Am J Hum Genet. 2017;100(4):635–49.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  81. Howe LJ, Dudbridge F, Schmidt AF, Finan C, Denaxas S, Asselbergs FW, Hingorani AD, Patel RS. Polygenic risk scores for coronary artery disease and subsequent event risk amongst established cases. Hum Mol Genet. 2020;29(8):1388–95.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  82. Newman JD, Schwartzbard AZ, Weintraub HS, Goldberg IJ, Berger JS. Primary prevention of cardiovascular disease in diabetes mellitus. J Am Coll Cardiol. 2017;70(7):883–93.

    Article  PubMed  PubMed Central  Google Scholar 

  83. Hoskin MA, Bray GA, Hattaway K, Khare-Ranade PA, Pomeroy J, Semler LN, Weinzierl VA, Wylie-Rosett J, for the Diabetes Prevention Program Research Group. Prevention of diabetes through the lifestyle intervention: Lessons learned from the diabetes prevention program and outcomes study and its translation to practice. Curr Nutr Rep. 2014;3(4):364–78.

  84. Wray NR, Yang J, Hayes BJ, Price AL, Goddard ME, Visscher PM. Pitfalls of predicting complex traits from SNPs. Nat Rev Genet. 2013;14(7):507–15.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references


The authors thank the study participants from the Centers for AIDS Research Network of Integrated Clinical Systems (CNICS).


This work was supported by the National Heart, Lung, And Blood Institute [R01HL125027 to IP and HMC and R01HL126538 to JD], National Human Genome Research Institute [R01HG010649 to IP and HMC], National Institute on Drug Abuse [R01DA047045 to HMC and IP]. Additional support came from the National Institute of Allergy and Infectious Diseases [R24 AI067039 to MSS, P30 AI027757, and P30 AI027767 to MSS]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Author information

Authors and Affiliations



AS and IP drafted and edited the manuscript. AS, IP, HC, KH, and CML processed, analyzed, visualized, and interpreted the data. HMC, PKC, BMW, RMN, SRW, JW-N, JAD, WJL, MMK, MSS, AWillig, JJE, WCM, PWH, RDM, AWebel, and KHM participated in the cohort setup and data collection and sample processing, assisted with data acquisition and interpretation, and provided substantive comments and edits for the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Inga Peter.

Ethics declarations

Ethics approval and consent to participate

Institutional Review Boards (IRBs) at each institution have approved CNICS research activities, and a written informed consent was obtained from all participants. The ethics approval of data analysis of de-identified subjects conducted during the current study was waived by the IRB at the Icahn School of Medicine at Mount Sinai, New York, New York.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Additional file 1:

Table S1. A summary of genotyping and imputation platforms utilized to generate the final study data set. Table S2. Description of previously published genome-wide association studies (GWAS) in European populations that were used in polygenic risk score analyses. Table S3. Number of variants used to derive various polygenic risk scores for different traits and diseases. Table S9. Number of expression quantitative trait loci controlled by HIV-specific loci in CD14+ monocytes from Fairfax et al. [51]. Fig. S1. Fraction of expression quantitative trait loci among the HIV-specific loci (see the Methods section for details) compared to all other loci. HDL, high-density lipoprotein; LDL, low-density lipoprotein; TG, triglycerides; SNP, single nucleotide polymorphism. CD14, CD14+ monocytes at baseline; INF, activated CD14+ monocytes following induction with interferon-γ; LPS2, activated CD14+ monocytes following a 2-h induction with lipopolysaccharide; LPS24, activated CD14+ monocytes following a 24-h induction with lipopolysaccharide. SNP, single nucleotide polymorphism.

Additional file 2:

Table S4. List of variants significantly associated with lipid levels in European Americans in the CNICS HIV cohort (p < E-5) and in Willer et al. [31] (p < 0.05).

Additional file 3:

Table S5. List of variants significantly associated with lipid levels in European Americans in the CNICS HIV cohort (p < E-5), non-significant in Willer et al. [31] (p > 0.05) and with no overlap between 99% confidence intervals for beta coefficients between CNICS and Willer et al.

Additional file 4:

Table S6. List of variants significantly associated with lipid levels in African Americans in the CNICS HIV cohort (p < E-5) and in Willer et al. [31] (p < 0.05).

Additional file 5:

Table S7. List of variants significantly associated with lipid levels in African Americans in the CNICS HIV cohort (p < E-5), non-significant in Willer et al. [31] (p > 0.05) and with no overlap between 99% confidence intervals for the beta coefficients between CNICS and Willer et al.

Additional file 6:

Table S8. List of variants significantly associated with lipid levels in European American females (p < E-5) but not in males (p > 0.05) in the CNICS cohort with no overlap between 99% confidence intervals for beta coefficients between females and males.

Additional file 7:

Table S10. Prevalence and clinical impact of high European-based and multiethnic polygenic risk scores for type-2 diabetes (T2D) and myocardial infarction (MI) in people living with HIV.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Cheng, H., Sewda, A., Marquez-Luna, C. et al. Genetic architecture of cardiometabolic risks in people living with HIV. BMC Med 18, 288 (2020).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: