Background
Despite the success of combination antiretroviral therapy (ART) in reducing HIV-associated morbidity and mortality, long-term ART is frequently associated with metabolic abnormalities including dyslipidemia, lipodystrophy, and insulin resistance [
1,
2]. These metabolic abnormalities increase the risk of cardiovascular, liver, kidney, bone, and neurological disorders and the incidence of these disorders is rising as HIV-infected populations age [
1,
3]. Mechanisms driving these abnormalities are multifactorial including effects of ART (e.g., protease inhibitor (PI)-associated dyslipidemia and nucleoside reverse transcriptase inhibitor (NRTI)-associated mitochondrial toxicity [
2‐
4]), disease related factors (e.g., CD4 T-cell depletion, inflammation, and unsuppressed viremia), and host factors (e.g. body mass index, comorbidities, and genetic predisposition) [
1,
2].
The liver plays a central role in regulating lipid, amino acid, and carbohydrate metabolism, but few studies have explored relationships between hepatic dysfunction and metabolic abnormalities in HIV-infected individuals on ART. Liver disease represents a leading cause of morbidity and mortality in HIV patients on ART, with hepatic dysfunction affecting 30-40% of patients [
5‐
7]. Ranging from mild reversible increases in hepatic enzymes to fibrosis and decompensation, hepatic dysfunction has been linked to co-infections with hepatitis B and C (HBV and HCV), ART-induced hepatotoxicity, and high prevalence of non-alcoholic fatty liver disease (NAFLD), which affects 20-70% of HIV-infected individuals [
5].
Twenty-five to 40% of HIV-infected individuals in the United States and Europe are co-infected with HCV. In these populations, HCV co-infection is associated with increased rates of lipodystrophy [
8,
9], hepatic steatosis [
10‐
12], and insulin resistance [
9,
13,
14], but lower total cholesterol (TC) and low-density lipoprotein cholesterol (LDL) [
15‐
17], particularly in patients infected with HCV genotype 3 [
11,
12,
18]. In HCV mono-infected individuals, altered cholesterol metabolism is associated with hepatic steatosis, advanced hepatic fibrosis, and poor responses to interferon-based therapy [
19‐
21]. Further studies are required to better define metabolic alterations in HIV subjects with and without HCV co-infection.
Metabolomics is the unbiased identification and quantification of small molecules in biological fluids. In the context of disease, metabolomics has been used to identify novel clinical biomarkers and therapeutic targets. Here, we performed untargeted metabolomic profiling of plasma from two independent cohorts of HIV-infected individuals with late stage disease on PI-based ART to identify a metabolite signature that distinguishes HIV-infected from healthy control subjects regardless of HCV serostatus. We also examined relationships between altered lipid metabolites and markers of inflammation, microbial translocation, and hepatic dysfunction.
Discussion
In this study, we characterized the plasma metabolome of HIV patients with advanced disease (CD4 nadir <300 cells/ul) on PI-based ART and identified a 35-metabolite signature mapping to lipid, amino acid, and nucleotide metabolism that distinguished HIV from control subjects independent of HCV serostatus. Seventy percent of these altered metabolites were lipids, including endogenous ligands for the nuclear receptors FXR, PXR, CAR, and PPAR such as bile acids, sulfated steroids, and PUFA. These receptors play important roles in regulating lipid metabolism, inflammation, and hepatic function. Lipids regulate diverse cellular processes including metabolic, inflammatory, and innate immune responses, and imbalances in lipid signaling pathways contribute to diverse pathologies including insulin resistance and cardiovascular disease [
39‐
41]. Untargeted metabolomic profiling represents a powerful tool for deconvoluting these pathways and gaining insights important for developing therapeutic strategies to reduce metabolic disease and chronic inflammation in HIV patients on ART.
In treatment-naive HIV patients studied prior to the introduction of ART, dyslipidemia was frequently detected, and characterized by hypocholesterolemia with and without hypertriglyceridemia [
42,
43]. The severity and frequency of these abnormalities were associated with decreasing CD4 counts and advanced disease [
43,
44]. In HIV patients on ART, dyslipidemia has been traditionally characterized by increased triglycerides, increased total and LDL cholesterol, and decreased HDL cholesterol [
45]. However, these lipids represent only a small subset of those in plasma [
46]. By profiling 113 plasma lipids, we identified alterations in several lipid classes including bile acids, acylcarnitines, sulfated steroids, and PUFA. Bile acids were increased in HIV patients on PI-based ART, consistent with a small previous study (n=11) [
47]. Under physiological conditions, bile acids play an important role in lipid absorption and homeostasis (reviewed in [
48]). At high concentrations, however, bile acids are proinflammatory and hepatotoxic [
48]. Acylcarnitines were also elevated in HIV subjects. Increased serum acylcarnitines, a marker of incomplete fatty acid β-oxidation, have been reported in patients with diabetes, and metabolic, cardiovascular, and mitochondrial diseases [
49‐
51]. In rodent models of obesity and insulin resistance, increased acylcarnitines have been linked to mitochondrial overload under conditions of metabolic stress [
52,
53]. Another important finding was reduced plasma sulfated steroids including DHEA-S, EpiA-S, pregnenolone sulfate, and androsterone sulfate. Reduced DHEA-S has been implicated in aging, age-related comorbidities (e.g., atherosclerosis and bone metabolism), and immune dysfunction [
54]. Furthermore, previous studies found associations between low DHEA-S and progression to AIDS, as well as dyslipidemia in HIV subjects on ART [
55,
56]. Bioactive PUFA including AA, EPA, DPA, and DHA were also decreased in the HIV plasma metabolome. PUFA depletion has been linked to hepatic triglyceride accumulation, unfolded protein responses, and increased susceptibility to endoplasmic reticulum stress [
57‐
59]. Collectively, these alterations show substantial overlap with those previously reported in NAFLD and NASH, (i.e., bile acids, acylcarnitines, LPC, PUFA, and DHEA-S) [
27‐
29], raising the possibility that mechanisms underlying development of NAFLD (i.e., lipid peroxidation and mitochondrial dysfunction) may also contribute to dyslipidemia in HIV patients on ART. However, other alterations were unique to HIV patients, including decreased saturated LCFA and eicosanoids. Thus, the plasma lipidome of HIV patients on PI-based ART has overlapping, yet distinctive features compared to NAFLD/NASH.
Mapping altered metabolites to KEGG and SMPDB pathways identified alterations in immunomodulatory pathways including tryptophan catabolism and eicosanoid biosynthesis (Figure
3). Despite long-term ART, tryptophan betaine and indoleproprionate were decreased and kynurenine and K:T ratios (an index of indolamine 2,3-dioxygenase (IDO) activity and tryptophan catabolism) were increased in HIV subjects compared to controls. Tryptophan catabolism is regulated by the balanced expression of IDO/tryptophanyl-tRNA-synthetase (TTS) [
60,
61]. Chronic IDO activation, however, drives immune dysfunction by suppressing T cell proliferation, increasing production of toxic metabolites (kynurenine and quinolinic acid), and altering the Th17/Treg balance, and has been previously linked to dysfunctional T cell responses in HIV patients [
60‐
62]. Another important finding was decreased eicosanoids (e.g., 5-HETE, PGB2, PGE2, and TXB2). Eicosanoids, derived from PUFA, play an important role in the initiation and resolution of inflammation (reviewed in [
63]). Thus, PUFA depletion and decreased eicosanoids biosynthesis may contribute to dysregulated inflammatory responses in HIV patients on PI-based ART. Correlation analysis identified negative correlations between sulfated steroids and markers of innate immune activation (IFN-α and IL-6). Mounting evidence suggests that innate immune responses regulate cholesterol/sterol metabolism. Sulfated steroids correlated inversely with plasma IFN-α, consistent with previous studies demonstrating that type I IFNs downregulate sterol metabolism during anti-viral responses [
64,
65]. Sulfated steroids also correlated inversely with IL-6. Dysregulated IL-6 production plays a role in age-related conditions, such as rheumatoid arthritis [
66,
67], osteoporosis [
68], atherosclerosis [
69], and Alzheimer's disease [
70,
71]. Consistent with our results, IL-6 has been shown to correlate inversely with serum DHEA-S in aging populations [
72]. In aging mice, dysregulated IL-6 production can be reversed by supplementing animals with DHEA-S [
73]. Thus, the relationship between IL-6 and sulfated steroids in HIV patients on PI-based ART merits further study.
The FIB-4 and APRI indices represent inexpensive, accurate (overall accuracy up to 87% and 72%, respectively), and non-invasive markers of liver fibrosis in HIV-infected individuals with and without HCV co-infection [
31‐
33,
74]. Similarly, circulating YKL-40 has been shown to correlate with the severity of fibrosis in chronic HCV infection and alcoholic liver disease [
75,
76]. Here, we found that intermediate/high FIB-4 and APRI scores and high levels of plasma YKL-40 were associated with increased plasma bile acids in HIV subjects on PI-based ART. Consistent with these findings, elevated serum bile acids have been reported in patients with cirrhosis and chronic HCV infection and were shown to correlate with the severity of liver fibrosis [
30,
77]. Bile acid levels (GCA and TCA) were positively correlated with acylcarnitines, a marker of incomplete fatty acid β-oxidation and mitochondrial dysfunction [
52,
78]. In rat models, bile acids have been shown to inhibit mitochondrial respiration in a dose dependent manner [
79‐
81] and to enhance the permeability of mitochondria to protons [
82]. Our findings, together with those of previous studies [
77,
83], suggest that plasma bile acids may represent a sensitive non-invasive marker of liver dysfunction in HIV patients.
Elevated plasma LPS has been reported in alcoholic liver disease [
84], NAFLD [
85], and inflammatory bowel disease [
86] and predicts progression to end-stage liver disease in patients with chronic HCV infection [
87]. Consistent with these findings, we found plasma LPS correlated positively with markers of hepatic function (bilirubin (E,E) and bilirubin (Z,Z)). The liver plays a key role in clearing LPS from the circulation, so increased plasma LPS may reflect decreased hepatic clearance [
85,
87,
88]. In turn, circulating LPS may promote hepatic dysfunction by increasing hepatic inflammation and fibrosis [
87]. Unexpectedly, LPS, LBP, and bilirubin ((E,E) and (Z,Z)) were negatively correlated with LCFA. LCFA deficiency has been reported in patients with advanced cirrhosis and other liver diseases (acute hepatitis, cholestasis) [
87,
89]. Thus, decreased plasma LCFA may be indicative of hepatic dysfunction in HIV patients. Together, these findings suggest complex inter-relationships between hepatic dysfunction, microbial translocation, and dyslipidemia in HIV patients on PI-based ART.
We acknowledge certain limitations of our study. We selected HIV subjects with advanced disease at increased risk for dyslipidemia (PI-based ART) and hepatic dysfunction (HCV). Further studies are required to define metabolic alterations in cohorts who are treatment-naive, at earlier stages of disease, on specific treatment regimens, with acute vs. chronic HCV infection, and with clinical evidence of metabolic abnormalities such as lipodystrophy and insulin resistance. Importantly, >60% of signature metabolites altered in HIV subjects on PI-based ART were also altered in an independent cohort of HIV subjects on NNRTI-based ART including sulfated steroids, bile acids, and LCFA (Additional file
2), indicating that many of the metabolite alterations we detected in HIV subjects on PI-based ART are not unique to this class of ART drugs. Future studies are required to better understand which metabolic abnormalities are associated with ART versus HIV infection. Individuals in both cohorts were predominantly African American. African Americans have higher rates of coronary heart disease and metabolic syndrome compared to Caucasians [
90]. Thus, further studies are needed to investigate these findings in other races/ethnicities. Finally, our study cohort was relatively small. Nonetheless, there was sufficient statistical power to identify metabolite sets with significant differences between HIV subjects with advanced disease and healthy controls. Despite these limitations, by applying stringent selection criteria, our study identified significant alterations in several important lipid classes associated with markers of inflammation and hepatic function in HIV patients on PI-based ART.
In summary, metabolomic profiling identified alterations in the plasma lipidome of HIV patients on PI-based ART, including increased bile acids and acylcarnitines and decreased sulfated steroids, PUFA, and LPC. Correlation analysis identified alterations in distinct lipid classes linked to markers of inflammation, microbial translocation, and hepatic function. Future studies are required to determine if these correlations reflect direct or indirect interactions and define the mechanisms underlying these inter-relationships. Many of the altered lipids represent important ligands of the nuclear receptor superfamily. Bile acids are endogenous ligands for FXR, which regulates bile acid biosynthesis and hepatic fatty acid and triglyceride biosynthesis [
91]. In mice, FXR activation protects against ritonavir-induced dyslipidemia and aortic plaque development, reverses insulin resistance and lipid abnormalities, and protects against hepatic steatosis [
92,
93]. Sulfated steroids and bile acids are ligands for PXR and CAR, which regulate expression of Phase I and II drug-metabolizing enzymes and transporters and hepatic lipid metabolism [
91]. Omega-3 and omega-6 PUFA modulate transcription via interactions with PPARs, which increase fatty acid oxidation and suppress lipogenesis and control inflammation by transrepression of NF-κB, NFAT, AP-1, and STATs [
39,
91]. Clinical trials have only shown modest effects of PPARα (fibrates) and PPARγ (thiazolidinediones and pioglitazone) agonists in reversing ART lipoatrophy [
94,
95], while other studies suggested these drugs may be useful in treating ART-associated diabetes and NAFLD [
96].
Authors’ contributions
EC performed the experiments and data analysis, drafted the manuscript, and prepared figures. VM performed bioinformatics and statistical analysis and prepared figures. AH and AK carried out part of the data analysis and contributed to study design and bioinformatics. SM participated in study design, data analysis, and manuscript editing. DG conceived of the study, participated in its design, coordination, and data analysis, and helped write and edit the manuscript. All authors read, participated in editing the manuscript, and approved the final manuscript.