Abstract

The immunological, biochemical and molecular mechanisms associated with poor immune recovery are far from known, and metabolomic profiling offers additional value to traditional soluble markers. Here, we present novel and relevant data that could contribute to better understanding of the molecular mechanisms preceding a discordant response and HIV progression under suppressive combined antiretroviral therapy (cART). Integrated data from nuclear magnetic resonance (NMR)-based lipoprotein profiles, mass spectrometry (MS)-based metabolomics and soluble plasma biomarkers help to build prognostic and immunological progression tools that enable the differentiation of HIV-infected subjects based on their immune recovery status after 96 weeks of suppressive cART. The metabolomic signature of ART-naïve HIV subjects with a subsequent late immune recovery is the expression of pro-inflammatory molecules and glutaminolysis, which is likely related to elevate T-cell turnover in these patients. The knowledge about how these metabolic pathways are interconnected and regulated provides new targets for future therapeutic interventions not only in HIV infection but also in other metabolic disorders such as human cancers where glutaminolysis is the alternative pathway for energy production in tumor cells to meet their requirement of rapid proliferation.

Introduction

Combined antiretroviral therapy (cART) successfully controls HIV viremia and improves the prognosis of HIV-infected subjects due to the decrease in HIV replication below detectable levels and due to CD4+ T-cell recovery [1]. However, up to one-fourth of HIV-infected subjects who start cART when they are severely immunosuppressed do not experience sufficient CD4+ T-cell increases after 2 years of successful viral suppression. They are so-called ‘patients with incomplete recovery or discordant response’, ‘immunological non-responders (INR)’ or ‘poor responders’ [2–4]. In addition, although the concept of a poor immune response remains vague and the prevalence of INR varies between 10 and 40% depending on the cohort studied [4–8], it is well established that patients receiving cART with CD4+ T-cell counts persistently below 250 cells/µl are associated with worse long-term prognosis, including a higher risk of progression toward AIDS and non-AIDS-defining clinical events and death [9–11].

Previously, our group has retrospectively investigated the immune characteristics preceding poor CD4 recovery during cART in a cohort of INR patients. Increased expression of CD4+ T-cell turnover-related markers (ki67/CD95) denoted an increased cycling and proliferative status in INRs before cART onset [12]. This phenomenon, together with an early profound immune dysregulation affecting Treg and Th17 cells [12,13], could contribute to the observed pro-inflammatory status that likely compromises the immune reconstitution of INR subjects. However, the integrated immunological, biochemical and molecular mechanisms associated with poor immune recovery are far from known. Investigations regarding the biologic mechanisms associated with this condition are needed to develop a useful tool for early detection of potential poor responders in daily clinical care and to investigate potential therapeutic targets.

The use of nuclear magnetic resonance (NMR) and mass spectrometry (MS) allows for the fast and reproducible quantitation of several molecules simultaneously, offering additional value to the standard methodologies for diagnosis or clinical prognosis in the study of HIV infection [14]. We and others have recently demonstrated that plasma metabolomics is a powerful tool to monitor natural HIV evolution or the effect of treatment as well as the metabolic diseases and chronic inflammation associated with HIV infection [15–19]. Our previous studies revealed that NMR-based metabolomics is a powerful instrument to identify a baseline lipoprotein profile associated with poor immune recovery [15] and dyslipidemia development in HIV-infected patients [16]. Untargeted plasma metabolomics profiling linked lipid abnormalities to circulating markers of inflammation, microbial translocation and hepatic function in HIV-infected subjects with advanced disease on protease inhibitor (PI)-based cART [17], whereas targeted metabolomics identified an increased rate of glycolysis in HIV-infected CD4+ T cells [18] and alterations in the catabolism of branched amino acids associated with disease progression [19].

Because our previous NMR-based metabolomics analyses helped to define a baseline metabolomic signature of HIV-infected subjects with low nadir [15], we hypothesized that the combination of NMR-based lipoprotein profile and MS-based metabolomics could help to elucidate the molecular mechanisms associated with the increased CD4+ T-cell turnover and immune dysregulation preceding poor immune recovery in our previously characterized cohort of HIV subjects [12,13]. Thus, we searched for potential predictive metabolomics markers of a late immune response before cART onset and HIV progression associated with discordant response after cART. For this aim, we performed circulating NMR- and MS-based HIV plasma metabolomics in 41 cART-naïve HIV-infected subjects who were subsequently followed-up for 96 weeks under cART.

Methods

Study design and participants

The study included 41 HIV-infected subjects aged ≥18 years and naïve to ART drugs recruited from the cohort of the Spanish AIDS Research Network (CoRIS) [20], which is an open cohort, multicenter cohort of patients newly diagnosed with HIV infection in the hospital or treatment centre, over 13 years of age, and naïve to antiretroviral treatment. For the present study, we analyzed two groups of pre-cART samples from HIV-positive subjects along with the main risk factors associated with immune discordance [12] kindly provided by the HIV Biobank integrated in the Spanish AIDS Research Network (RIS) [21]. The study subjects included patients who started cART with counts of <200 CD4+ cells/µl but did not achieve more than 250 CD4+ cells/µl after 96 weeks of suppressive treatment (INR subjects, n=18) and a control group of patients who had also started cART with <200 CD4+ cells/µl but achieved more than 250 CD4+ cells/µl under the same conditions (immunological responder (IR) subjects, n=23). Available follow-up samples (INR subjects, n=9 and IR subjects, n=8) after 96 weeks under suppressive cART were also analyzed. During this period, patients were receiving a combination of two nucleoside reverse transcriptase inhibitors (NRTI) plus a non-NRTI (NNRTIs) and/or a PI or a combination of both. Only two patients of each group were exposed to zidovudine (AZT), a known highly toxic NRTI-containing regimen. Data collection included basic clinical background and demographic information, as well as CD4+ T-cell count, HIV-1 RNA viral load measurements and AIDS-defining events. Regarding viral load determinations, at least four viral load determinations throughout that period were measured, and all of which had to be below 40 HIV RNA copies/ml, except for the first 6 months following treatment initiation, where higher values were allowed [12]. Patients taking lipid-lowering agents (such as statins, fibrates or ezetimibe), antidiabetic drugs (such as metformin, sulfonylureas, DDP-4 inhibitors or insulin and analogs), psychotropic drugs (including antipsychotics or antidepressants) and other drugs with several known metabolic effects were excluded. Additionally, a group of healthy HIV-uninfected (n=9) volunteers, matched by gender (all male) and age [43 (40-48) years], was also analyzed for reference values. The present study was carried out in accordance with the recommendations of the Ethical and Scientific Committees from each participating institution. All subjects gave written informed consent in accordance with the World Medical Association Declaration of Helsinki.

Plasma soluble markers

Markers of systemic inflammation [interleukin (IL)-6, high-sensitivity C-reactive protein (hsCRP)], immune activation or suppression [interferon γ-induced protein (IP)-10, soluble CD14 (sCD14), transforming growth factor-β (TGF-β)], microbial translocation [lipopolysaccharide (LPS)], endothelial dysfunction [soluble intercellular adhesion molecules (sICAM-1) and soluble vascular cell adhesion molecule (sVCAM-1)], platelet activation [soluble CD40 ligand (sCD40L)] and coagulation (D-dimer) were analyzed as described elsewhere [12]. Previous pre-cART data from several individual soluble parameters were partially published in a previous study including some patients from the same cohort [12] whereas values achieved at 96 weeks of cART were exclusively of the present work.

NMR lipoprotein measurements

All 1H NMR spectra were recorded at 310 K on a Bruker Avance III 600 spectrometer operating at a proton frequency of 600·20 MHz. Lipid concentrations (i.e., triglycerides and cholesterol), lipoprotein sizes and particle numbers for very low-density lipoprotein (VLDL) (38.6–81.9 nm), low-density lipoprotein (LDL) (18.9–26.5 nm) and high-density lipoprotein (HDL) (7.8–11.5 nm) classes, as well as the particle numbers of nine subclasses, namely, large, medium, and small VLDL, LDL and HDL (a total set of 26 variables) were measured in 2D spectra from diffusion-ordered NMR spectroscopy (DOSY) experiments using the Liposcale test [22]. Briefly, the cholesterol and triglyceride concentrations of the main lipoprotein fractions were predicted using partial least squares (PLS) regression models. Then, the methyl proton resonances of the lipids in lipoprotein particles were decomposed into nine Lorentzian functions representing nine lipoprotein subclasses and the mean particle size of every main fraction was derived by averaging the NMR area of each fraction by its associated size. Finally, we calculated the particle numbers of each lipoprotein main fraction by dividing the lipid volume by the particle volume of a given class and we used the relative areas of the lipoprotein components used to decompose the NMR spectra in order to derive the particle numbers of the nine lipoprotein subclasses.

Metabolomic analysis

Metabolomic profiling was performed by the Centre for Omic Sciences (COS) (http://omicscentre.com/) using ultra-high-performance liquid chromatography coupled to quadrupole-time of flight high-resolution MS (UHPLC-(ESI)qTOF) using positive and negative ionization and hydrophilic interaction liquid chromatography coupled to quadrupole-time of flight high-resolution MS (UHPLC-(ESI)qTOF) using positive ionization. The instrument was an Agilent 1290 Infinity UHPLC coupled to 6550 i-funnel qTOF, and chromatographic columns were an Acquity UPLC® HSS T3 C18 (100 × 2.1 mm, 1.8 µm) for UHPLC positive ionization, an Acquity UPLC® BEH C18 (100 × 2.1 mm, 1.8 µm) for UHPLC negative ionization and an Acquity UPLC® BEH HILIC (100 × 2.1 mm, 1.7 µm) for HILIC positive ionization, all from Waters, U.S.A. The chromatographic analyses were performed under gradient elution using ultrapure water (0.1% formic acid) and acetonitrile for the UHPLC positive ionization method, ultrapure water (1 mM ammonium fluoride) and acetonitrile for the UHPLC negative ionization method and ultrapure water (50 mM ammonium acetate) and acetonitrile for the HILIC positive ionization method. The flow rate was 0.4 ml/min, the column temperature was set at 25°C and the injection volume was 2 µl (+4°C) for all methods. The acquisition range was between 100 and 1000 m/z at 1.5 spectra/s.

The extraction of plasma metabolites was conducted by protein precipitation using methanol:water (8:1). A total of 450 μl of extracting solution was added to 50 μl of serum and were mixed for 10 s and sonicated for 2 min. Next, the solution was kept for 10 min on ice and then centrifuged at 15000 rpm at +4°C. The remaining supernatants were transferred to LC vials for analysis.

For tentative identification of metabolites, Mass Profinder software (Agilent) was used for data deconvolution and Mass Profiler Professional (Agilent) was used for data alignment for all samples. The aligned features were submitted to the ID Browser module for identification by their exact mass and retention time using the Metlin Personal Compound Database (Agilent). A total of 99 metabolomic entities were initially identified, and pre-processing data analysis was performed to exclude variables with constant or single values. Missing values for a given metabolite were detected and replaced by the median value of the metabolite. PLS discriminant analysis (PLS-DA) was performed in the Metaboanalyst web portal (www.metaboanalyst.ca). Random Forest (RF) and hierarchical clustering of signature metabolites associated with immune recovery were performed using the R. Both, human Metabolome Database (HMDB) (http://www.hmdb.ca/) [23] and the MetabolomeXchange web portal (http://www.metabolomexchange.org) were used to confirm the presence of metabolites identified in the human body and to obtain detailed information about the metabolites found in our study.

Statistical analysis

The normality of the distribution of the variables was assessed using a Kolmogorov–Smirnov test. Medians and interquartile ranges or means and standard deviations were used to summarize the continuous variables, and comparisons between the groups were performed with non-parametric Kruskal–Wallis (KW) and/or Mann–Whitney (MW) tests for unpaired samples and a Wilcoxon ttest for paired samples (W). The categorical variables are summarized as frequencies and percentages, and their associations with immune recovery were determined using the χ2 test. Spearman correlation coefficients and the corresponding P-values were calculated to assess the associations between CD4 counts and the estimated plasma lipoprotein, metabolome and lipidome entities and soluble parameters. RF analyses were performed as multivariate tests to identify the variables that best partitioned the overall study population according to immune recovery predisposition. The RF interpretations are represented using the mean decrease in accuracy (MDA) variable, which estimates how much excluding (or permuting) each variable reduces the accuracy of the model during the out-of-bag error calculation phase. The variables with large MDAs were selected, and logistic regression models that combined the statistically significant variables were generated in both univariate and multivariate tests. Statistical analyses were performed using SPSS (version 21.0, SPSS Inc., Chicago, IL) and the R software computing environment (https://www.r-project.org/). The graphical representations are based on both the graphical environment of R and GraphPad Prism software (version 5.0, GraphPad Inc., San Diego, CA). The metabolite–protein interaction network and functional enrichment analyses were generated using the Search Tool for Interactions of Chemicals (STITCH) database, version 5.0 [24]. The results were considered significant at P<0.05.

Results

Patient characteristics

Forty-one cART-naïve patients were included in the study. The baseline clinical characteristics of the HIV-infected subjects (n=41) are presented in Table 1. All patients were males, with a median age of 42 (34–51) years, with low CD4+ counts [86 (63–161) cells/µl] and plasma viral loads [4.86 (4.41–5.37) log HIV RNA copies/ml]. Both the INR and the IR group had similar frequencies of hepatitis C virus (HCV) co-infection (7.5%) and AIDS-related illness (24.4%). Based on the classification criteria, after 96 weeks under suppressive cART, INR subjects achieved 210 (176–235) CD4+ T cells/µl and a CD4/CD8 T-cell ratio of 0.32 (0.23–0.41), whereas IR subjects achieved 436 (360–570) CD4+ T cells/µl and a CD4/CD8 T-cell ratio of 0.38 (0.31–0.48).

Table 1
Pre-cART characteristics of HIV-infected subjects
All patients (n=41)INR subjects (n=18)IR subjects (n=23)P-value
Clinical characteristics 
Age at cART initiation (years) 42 (34–51) 42 (33–55) 41 (33–49) 0.52 
Risk factor    0.03 
  Heterosexual 11 (26.8) 8 (19.5) 3 (7.3)  
  Homo/Bisexual 23 (56.1) 6 (14.6) 17 (41.5)  
  Intravenous drug abuse 4 (9.8) 1 (2.4) 3 (7.3)  
  Other 1 (2.4) 1 (2.4)  
  Unknown 2 (4.9) 2 (4.9)  
CD4+ T-cell count (cells/µl) 86 (63–161) 81 (44–143) 144 (74–166) 0.17 
CD8+ T-cell count (cells/µl) 645 (574–1083) 598 (578–923) 889 (538–1642) 0.35 
CD4/CD8 T-cell ratio 0.13 (0.08–0.26) 0.11 (0.07–0.27) 0.15 (0.10–0.25) 0.46 
Plasma HIV RNA load (log copies/ml) 4.86 (4.42–5.32) 4.96 (4.13–5.29) 4.77 (4.54–5.32) 0.95 
AIDS-related illness    0.34 
  Yes 10 (24.4) 5 (12.2) 5 (12.2)  
  No 25 (61.0) 12 (29.3) 13 (31.7)  
  Unknown 6 (14.6) 1 (2.4) 5 (12.2)  
HCV co-infection    0.41 
  Positive 3 (7.5) 2 (5.0) 1 (2.5)  
  Negative 29 (72.5) 13 (32.5) 16 (40.0)  
  Unknown 8 (20.0) 2 (5.0) 6 (15.0)  
All patients (n=41)INR subjects (n=18)IR subjects (n=23)P-value
Clinical characteristics 
Age at cART initiation (years) 42 (34–51) 42 (33–55) 41 (33–49) 0.52 
Risk factor    0.03 
  Heterosexual 11 (26.8) 8 (19.5) 3 (7.3)  
  Homo/Bisexual 23 (56.1) 6 (14.6) 17 (41.5)  
  Intravenous drug abuse 4 (9.8) 1 (2.4) 3 (7.3)  
  Other 1 (2.4) 1 (2.4)  
  Unknown 2 (4.9) 2 (4.9)  
CD4+ T-cell count (cells/µl) 86 (63–161) 81 (44–143) 144 (74–166) 0.17 
CD8+ T-cell count (cells/µl) 645 (574–1083) 598 (578–923) 889 (538–1642) 0.35 
CD4/CD8 T-cell ratio 0.13 (0.08–0.26) 0.11 (0.07–0.27) 0.15 (0.10–0.25) 0.46 
Plasma HIV RNA load (log copies/ml) 4.86 (4.42–5.32) 4.96 (4.13–5.29) 4.77 (4.54–5.32) 0.95 
AIDS-related illness    0.34 
  Yes 10 (24.4) 5 (12.2) 5 (12.2)  
  No 25 (61.0) 12 (29.3) 13 (31.7)  
  Unknown 6 (14.6) 1 (2.4) 5 (12.2)  
HCV co-infection    0.41 
  Positive 3 (7.5) 2 (5.0) 1 (2.5)  
  Negative 29 (72.5) 13 (32.5) 16 (40.0)  
  Unknown 8 (20.0) 2 (5.0) 6 (15.0)  

Data are presented as n (%) or median (interquartile range). Categorical data were compared by means of a χ2 test, whereas continuous data were compared using the MW test.

AIDS was diagnosed according the CDC1993 criteria. P-value <0.05 was considered significant and is highlighted in bold.

Pre-cART profile of soluble biomarkers predicting poor immune recovery after ART

Lipoprotein profile, lipidome and metabolome entities were analyzed in order to characterize groups of study. Previous pre-cART data from several individual soluble parameters were also included in order to improve the prognostic value of the metabolomics signature [12]. As expected, among a selection of soluble biomarkers of inflammation, immune activation/suppression, microbial translocation, endothelial dysfunction, platelet activation and coagulation, only higher IL-6 levels and a greater number of subjects having hsCRP ≥ 5 mg/l where found in INR subjects compared with IR subjects (Supplementary Table S1). Additionally, whereas no differences were found regarding the baseline pre-cART Liposcale lipoprotein characterization (Supplementary Table S2), the metabolite PLS-DA score plot, including lipidomic markers, was different between INR subjects and IR subjects before cART onset (Figure 1A). In fact, univariate analysis revealed an elevated concentration of l-tyrosine and a decreased concentration of l-glutamate and phosphatidylcholine PC (16:1) at baseline associated with later poor immune recovery (Figure 1B).

Baseline integrated data associated with poor immune recovery

Figure 1
Baseline integrated data associated with poor immune recovery

(A) Two-dimensional PLS-DA score plot of metabolomics and lipidomics analyses distinguishes INR subjects (red dots) from IR subjects (green dots). (B) Entities (arbitrary units) that clearly distinguish INR subjects from IR subjects in the univariate test (MW U test). Data are presented as mean ± S.E.M. (C) Plot shows the MDA from RF analysis, ranking variables (MDA > 5.5) according to their prognostic importance for immune recovery. Abbreviations: a.u., arbitrary unit; dGMP, deoxyguanosine 5′-monophosphate.

Figure 1
Baseline integrated data associated with poor immune recovery

(A) Two-dimensional PLS-DA score plot of metabolomics and lipidomics analyses distinguishes INR subjects (red dots) from IR subjects (green dots). (B) Entities (arbitrary units) that clearly distinguish INR subjects from IR subjects in the univariate test (MW U test). Data are presented as mean ± S.E.M. (C) Plot shows the MDA from RF analysis, ranking variables (MDA > 5.5) according to their prognostic importance for immune recovery. Abbreviations: a.u., arbitrary unit; dGMP, deoxyguanosine 5′-monophosphate.

Next, RF analysis was used as a multivariate method to rank variables (integrated data) that constitute the best predictors of immune recovery (Figure 1C). In particular, RF provided additional evidence indicating that IL-6 and PC (16:1) were among the most discriminatory parameters (MDA > 30) between INR subjects and IR subjects before cART onset. Palmitoylcarnitine (PalC) concentrations also had a strong classification power in the multivariate model despite not appearing to be significantly different in the univariate test (Figure 1B), and increased plasma hsCRP ≥ 5 mg/l was the best classifier (MDA > 58). We next evaluated which metabolic pathways were affected by the interaction of these selected metabolites and proteins (Figure 2A). Surprisingly, in the biological process category, microbial translocation was one of the most-enriched terms along with the expected inflammatory (IL-6) and acute-phase response (hsCRP), suggesting a connection between the expression of pro-inflammatory cytokines and microbial translocation in the discordant response (Figure 2B). Gene ontology of cellular component categories further confirmed the enrichment of IL-6 complex and T-cell complex, as well as, their connection to neuronal components (Figure 2B). Notably, ciliary neurotrophic factor receptor (CNTFR), which supports the survival of neurons, is closely related to IL-6 receptor. Thus, the enriched KEGG pathway revealed a strong relationship between pre-cART immune recovery not only with immune processes (hematopoietic cell lineage, Jak-STAT signaling pathway, T-cell receptor) but also with key regulatory factors of T-cell glycolysis (hypoxia-inducible factor 1 (HIF-1) signaling pathways) inflammatory disorders (inflammatory bowel disease (IBD), NALFD), and alanine, aspartate and glutamate metabolism (Figure 2B).

Network interactions associated with baseline poor immune recovery

Figure 2
Network interactions associated with baseline poor immune recovery

(A) Network of metabolite–protein interaction generated with the set of 35 parameters that constitute the best predictors of immune recovery according the RF model. The names of the integrated data (soluble protein markers and metabolites) were used as input in the STITCH database. (B) The STITCH database identified different functional enrichments associated with our network using the false discovery rate (FDR). The strong biological process and KEGG pathways included the expected IL-6-mediated signaling pathways but also microbial translocation, inflammatory bowel disease (IBD) and glutamate metabolism, among others. Glutamate metabolism refers to alanine, aspartate and glutamate metabolism. Abbreviation: NAFLD, non-alcoholic fatty liver disease.

Figure 2
Network interactions associated with baseline poor immune recovery

(A) Network of metabolite–protein interaction generated with the set of 35 parameters that constitute the best predictors of immune recovery according the RF model. The names of the integrated data (soluble protein markers and metabolites) were used as input in the STITCH database. (B) The STITCH database identified different functional enrichments associated with our network using the false discovery rate (FDR). The strong biological process and KEGG pathways included the expected IL-6-mediated signaling pathways but also microbial translocation, inflammatory bowel disease (IBD) and glutamate metabolism, among others. Glutamate metabolism refers to alanine, aspartate and glutamate metabolism. Abbreviation: NAFLD, non-alcoholic fatty liver disease.

HDLs and glutamate metabolism are strongly associated with immune recovery after cART

Next, we examined lipoprotein alterations and analyzed soluble parameter changes after suppressive cART therapy in a group of 17 HIV-infected subjects with available follow-up samples after 96 weeks on stable cART (Supplementary Table S3). INR subjects (n=9) had increased HDL cholesterol and increased HDL particle sizes, mainly due to increased medium HDL-P, compared with IR subjects (n=8) (Figure 3A). Regarding the soluble plasma markers, INR subjects had reduced concentrations of TGF-β after suppressive cART (Figure 3A and Supplementary Table S4).

Plasma lipoprotein metabolism and metabolome associated with immune recovery after cART

Figure 3
Plasma lipoprotein metabolism and metabolome associated with immune recovery after cART

(A) Plasma lipoprotein characteristics and TGF-β concentration associated with immune recovery after 96 weeks of cART. Data are presented as box and whiskers plots (min to max). (B) Two-dimensional PLS-DA scatter plot constructed with plasma metabolomic and lipidomic entities clearly distinguishes IR subjects (green dots) from INR subjects (red dots). (C) Data presented as mean ± S.E.M. from metabolites differentially expressed between INR subjects from IR subjects in the univariate U-test. (D) Plot shows the MDA from RF analysis ranking parameters (MDA > 10) according to their prognostic importance for immune recovery. (E) The STITCH database illustrated the association between CD4, ‘glutamate’ and ‘TGF-β family’ using the false discovery rate (FDR).

Figure 3
Plasma lipoprotein metabolism and metabolome associated with immune recovery after cART

(A) Plasma lipoprotein characteristics and TGF-β concentration associated with immune recovery after 96 weeks of cART. Data are presented as box and whiskers plots (min to max). (B) Two-dimensional PLS-DA scatter plot constructed with plasma metabolomic and lipidomic entities clearly distinguishes IR subjects (green dots) from INR subjects (red dots). (C) Data presented as mean ± S.E.M. from metabolites differentially expressed between INR subjects from IR subjects in the univariate U-test. (D) Plot shows the MDA from RF analysis ranking parameters (MDA > 10) according to their prognostic importance for immune recovery. (E) The STITCH database illustrated the association between CD4, ‘glutamate’ and ‘TGF-β family’ using the false discovery rate (FDR).

Next, we also studied the effect of successful immune restoration on the HIV plasma metabolome and lipidome. A supervised two-dimensional PLS-DA score plot clearly distinguished INR subjects from IR subjects (Figure 3B). Univariate analysis revealed a significantly decreased concentration of 5-aminolevulinic acid, eicosapentaenoic acid, γ-glγu-leu, glycolic acid, isopimaric acid and l-glutamate in INR subjects compared with IR subjects (Figure 3C). In this case, RF ranked HDL-C, LDL-C, HDL-P, and l-glutamate as the top parameters accounting for immune recovery after cART (Figure 3D).

Correlation analyses between CD4+ T-cell counts and metabolomics signature after cART

Discordant response is clearly defined by CD4+ T-cell counts after cART, so we performed Spearman correlation analysis to confirm and detect the relationship between metabolomics and CD4+ T-recovery. Correlation analyses of integrated data showed significant direct association of glycocholic acid (ρ = 0.51, P=0.04), isopimaric acid (ρ = 0.74, P=0.04), and citrulline (ρ = 0.60, P=0.03) to CD4+ T-cell count achieved after 96 weeks of suppressive cART. By contrast, CD4+ T-cell count was inversely correlated to dl-pipecolic acid (ρ = −0.94, P<0.01) and microbial translocation (LPS, ρ = −0.61, P=0.03). In fact, networking modeling confirmed the relationship between glutamate metabolism not only to 5-aminolevulinic acid, citrulline [0.02 (0.02–0.03) in INR-subjects and 0.03 (0.01–0.03) in IR-subjects] and dl-pipecolic acid [0.07 (0.07–0.07) in INR-subjects and 0.03 (0.02–0.03) in IR-subjects], but also to ‘TGF-β family’ via interaction with CCR2-CCL2 and retinoic acid axis (Figure 3E).

Longitudinal evaluation of lipoprotein metabolism and plasma soluble parameters

Plasma NMR-based lipoprotein values achieved at 96 weeks of cART in HIV-infected subjects were compared with a control group of healthy volunteers (H-subjects) [1012 (801–1164) CD4+ T cells/µl]. INR subjects had increased LDL-C (ΔLDL-C = 31%) and LDL-P (ΔLDL-P = 28%), although their values remained significantly lower compared with the control group (Figure 4A and Supplementary Figure S1). In contrast, IR subjects reached similar LDL-C (ΔLDL-C = 19%) values and LDL-P (ΔLDL-P = 18%) values as healthy subjects (Figure 4A). Regarding HDL metabolism, INR subjects suffered a surprising increase in HDL-C (ΔHDL-C = 42%) and HDL-P (ΔHDL-C = 12%), achieving comparable values to healthy subjects, whereas in IR subjects these parameters remained lower compared with the reference group (P=0.03 and P<0.01, respectively), despite the same longitudinal increasing trend (ΔHDL-C = 16% and ΔHDL-C = 7%) (Figure 4B and Supplementary Figure S1).

Longitudinal evolution of the lipoprotein profile in a follow-up subgroup of 8 INR and 9 IR subjects in comparison with a group of healthy volunteers

Figure 4
Longitudinal evolution of the lipoprotein profile in a follow-up subgroup of 8 INR and 9 IR subjects in comparison with a group of healthy volunteers

(A) Both, INR subjects and IR subjects exhibited increased LDL-C and LDL-P after 96 weeks of cART. Whereas INR subjects values remained significantly lower compared with the control group, IR subjects reached similar values to healthy volunteers (H-subjects). (B) HDL-C and HDL-P exhibited a surprising slight increase in INR subjects achieving comparable values to healthy subjects, but remained lower in IR subjects compared with healthy subjects. Comparisons between groups were performed with MW tests for unpaired samples and Wilcoxon ttest for paired samples (W). Abbreviations: B, pre-cART value; P, value after 96 weeks of cART.

Figure 4
Longitudinal evolution of the lipoprotein profile in a follow-up subgroup of 8 INR and 9 IR subjects in comparison with a group of healthy volunteers

(A) Both, INR subjects and IR subjects exhibited increased LDL-C and LDL-P after 96 weeks of cART. Whereas INR subjects values remained significantly lower compared with the control group, IR subjects reached similar values to healthy volunteers (H-subjects). (B) HDL-C and HDL-P exhibited a surprising slight increase in INR subjects achieving comparable values to healthy subjects, but remained lower in IR subjects compared with healthy subjects. Comparisons between groups were performed with MW tests for unpaired samples and Wilcoxon ttest for paired samples (W). Abbreviations: B, pre-cART value; P, value after 96 weeks of cART.

Next, we explored longitudinal changes in plasma soluble parameters associated with immune recovery. Both groups experienced a decrease in soluble markers of immune activation (interferon γ-induced protein 10 (IP-10)) and endothelial dysfunction (ICAM-1, VCAM-1), although the differences were only significant in INR subjects (Supplementary Figure S2). In contrast, markers of platelet activation (CD40L) were slightly increased after ART therapy, and again, differences were only significant in INR subjects.

Metabolite restoration in relation to immune recovery

To simplify data analysis, pre-processing was performed to include metabolites that were significantly different throughout the follow-up in HIV-infected subjects and metabolites that were significantly different between healthy volunteers and HIV-infected subjects at 96 weeks of cART. By this approach, four metabolites were significantly altered throughout the follow-up in INR subjects, nine metabolites were significantly altered throughout the follow-up in IR subjects and eight metabolites were exclusively associated with HIV infection (Supplementary Table S5).

Hierarchical clustering of this set of 35 metabolites clearly distinguished HIV-infected subjects from H-subjects after 96 weeks of cART. Furthermore, hierarchical clustering clearly distinguished INR subjects from IR and healthy subjects, indicating an anomalous immune response after suppressive cART. Notably, after cART, IR subjects had similar metabolomic signatures as non-HIV-infected healthy volunteers (Figure 5).

Hierarchal clustering of metabolites altered during cART

Figure 5
Hierarchal clustering of metabolites altered during cART

Hierarchical clustering of a set of 35 metabolites that were different throughout the follow-up in both groups of HIV patients and different between healthy volunteers and HIV-infected patients at 96 weeks of cART.

Figure 5
Hierarchal clustering of metabolites altered during cART

Hierarchical clustering of a set of 35 metabolites that were different throughout the follow-up in both groups of HIV patients and different between healthy volunteers and HIV-infected patients at 96 weeks of cART.

Discussion

Recently, we demonstrated NMR-based metabolomics as a powerful tool to identify dyslipidemia development in HIV-infected subjects [16] and a baseline lipoprotein profile associated with poor immune recovery [15]. Here, we show that integrated data from NMR-based lipoprotein profiles, MS-based metabolomics and soluble plasma biomarkers builds prognostic and immunological tools that allow the differentiation of HIV-infected subjects based on their immune recovery status after 96 weeks of stable cART. Inflammation, glutaminolysis and structural and compositional changes in lipoproteins were associated with late immune recovery in INR subjects with increased CD4+ T-cell turnover and immunological dysregulation. To our knowledge, this is the first study to evaluate plasma predictive markers and disease progression using NMR and MS-based metabolomics along with traditional soluble markers in a longitudinal study of HIV-infected subjects initiating cART with a subsequent low CD4+ T-cell counts (less than 200 cells/µl), and followed up for 96 weeks of cART. The current study is consistent with our previous observation [12] that associated higher levels of IL-6/hsCRP and increased CD4+ T-cell turnover (ki67/CD95 expression) in INRs before cART onset, and offers new insight into the metabolic pathways involved.

Plasma IL-6, hsCRP ≥ 5 mg/l, l-tyrosine, l-glutamate and PC (16:1) by univariate model, and hsCRP, IL-6 and PalC by multivariate model (RF), were identified as predictive markers of late immune recovery. PalC is a long-chain acyl fatty acid ester of carnitine that has associated with decreased Treg cell number in Salmonella infection [25] due to apoptosis regulation and hence associated with pro-inflammatory effects on the immune response [26]. Consistent with this, increased plasma PalC and IL-6 concentrations, as well as a greater number of subjects having hsCRP ≥ 5 mg/l, were the best classifiers of late poor immune recovery in the RF model. Accordingly, we have recently described increased levels of IL-6 and hsCRP preceding poor CD4 T-cell recovery [12]. For instance, IL-6 activates the STAT3 signaling pathway, boosting the up-regulation of the Th17-promoting transcription factor RORγt which is also directly induced by the transcription HIF-1 [27,28]. HIF-1 is a key metabolic sensor expressed in CD4+ T cells that regulate the Th17/Treg balance by up-regulating glycolytic metabolism in an mTOR-dependent manner [29,30]. Consistent with this, our previous work also demonstrated that increased frequencies of Th17 cells precede the discordant response to cART, suggesting that immune dysregulation affecting these subsets is critical in INRs to cART and could be linked to their inflammatory state [13]. T-cell activation drastically increases the metabolic demands, down-regulating the characteristic pathways of quiescent cells via the induction of the transcription factor Myc as well as HIF-1 via the PI3/Akt/mTOR and NF-κB pathways [27,31]. In effect, increased glycolytic metabolism was previously associated with low CD4+ T-cell counts and abnormally high levels of immune activation due to metabolic depletion of CD4+ T-cell counts in HIV subjects [32]. Additionally, expression of NF-κB is activated by PC turnover, and thus, PC breakdown assumes a major role in HIV replication [33]. Interestingly, HIV replication in INR subjects under suppressive cART may be related to the existence of a population of metabolically active CD4+ OX40+ T cells that are highly susceptible to HIV infection [34], and we recently observed that such a population was up-regulated both before and after cART initiation in INR subjects [35]. Thus, according to our results, we suggest that cell cycle regulation and inflammatory molecules are tightly interrelated and could induce the activation of key regulatory factors of T-cell glycolysis in HIV patients with poor immune response before starting cART.

On the other hand, the increased concentrations of plasma l-tyrosine could also be related to increased CD4+ T-cell activation and the production of pro-inflammatory cytokines in INR subjects [36,37]. l-tyrosine is a non-essential amino acid and a precursor of brain catecholamines (dopamine, norepinephrine and epinephrine), that could be produced by lymphocytes in an autocrine or paracrine fashion [38]. It is known that dopamine reduces the suppressive and trafficking activities of Treg cells via the ERK signaling pathway [36]. Additionally, overexpression of tyrosine hydroxylase (TH), the enzyme responsible for catalyzing the conversion of l-tyrosine into l-DOPA, facilitates a shift in T-helper cell differentiation and function toward Th2 cells direction, promoting the production of anti-inflammatory cytokines [37]. Therefore, we could hypothesize that INR subjects have diminished TH activity, accumulating higher circulating plasma l-tyrosine and, as a consequence, increasing CD4+ T-cell exhaustion and promoting a Th1/Th2 shift toward Th1 cells, which is also consistent with the excess of pro-inflammatory cytokines in patients with poor immune recovery.

We have recently reported increased expression of CD4+ T-cell turnover-related markers (Ki67/CD95) in INR subjects before cART initiation [12]. Here, we observed that these high rates of cellular turnover likely maintain the l-glutamate concentration far from what can be offset even after 96 weeks of cART, which may be associated with damage of the gastrointestinal (GI) tract and microbial translocation. Our results revealed that CD4+ T-cell count was positively correlated with glycocholic acid, a primary bile acid, and the concentration of citrulline, a biomarker of intestinal diseases and enterocyte function [39,40]. CD4+ T-cell count was also inversely correlated with both LPS, a bacterial by-product already associated with higher glucose uptake in monocytes [41], and dl-pipecolic acid, a product of lysine degradation [42]. Thus, damage of the GI tract and microbial translocation could be associated with worse prognosis of immune recovery in patients receiving cART. In this context, l-glutamate is an ATP-producing substrate for enterocytes as well as a precursor for citrulline synthesis [39], which is also negatively influenced by hsCRP, likely due to HIV replication, with depletion of CD4+ T cells leading to enterocyte loss [40]. Bile acids directly impact the adaptive immune response via inhibition of Th1 activation via Vitamin D receptor (VDR) signaling [43]. Furthermore, d-Pipecolic acid is metabolized from lysine by intestinal bacteria, and their uptake into the mitochondria was related to the development of hepatic encephalopathy via the induction of apoptosis in neuronal cells [44]. Additionally, high levels of circulating LPS were previously related to elevated CD4+ T-cell turnover, suggesting an increased translocation of microbial products though the GI mucosa in HIV incomplete responders [45,46].

Both protein and lipid compositional changes in HDL particles were previously associated with structural and functional changes in these lipoproteins in HIV infection [47]. Here, we observed increased HDL metabolism (HDL-C and HDL-P) in INR subjects after 96 weeks of cART that was surprisingly even higher than in healthy volunteers. HDL particles are heterogeneous in their shape, size, and surface charge with diverse functionality. HDLs are known to be the principal mean to transport lipid from extrahepatic cells back to the liver for degradation or recycling, but growing evidence challenges the role of HDL and its components in glucose metabolism via a variety of mechanisms [48]. HDL activates Akt phosphorylation in ApoA-I transgenic mice enhancing both glycolysis and glucose oxidation in mouse muscle cells and facilitating glucose uptake through an ABCA1/AMPK-dependent mechanism. In addition, although it is well recognized that HDL exerts anti-inflammatory effects reducing the secretion of several cytokines and chemokines, recent evidence also suggests a clear pro-inflammatory effect of HDL on macrophages. Disruption of membrane lipid rafts by HDL action increases the production of pro-inflammatory cytokines via the NF-kB/STAT1-IRF1 axis [49]. Furthermore, increased HDL particles could be related to the down-regulation of TGF-β, which is critical for the suppression of Th1 cell differentiation. In vitro evidence suggests that HDL can decrease the TGF-β1-mediated induction of α-smooth muscle actin expression and TGF-β1-induced collagen deposition in human aortic endothelial cells [50]. Consistent with these data, our results revealed decreased TGF-β concentrations in INR subjects compared with IR subjects after 96 weeks of cART. Thus, taken together, our data suggest that substantial remodeling of up-regulated HDL particles contribute to combat lipid accumulation from the cell membrane components of apoptotic cells, skewing the Th1/Th2 balance toward Th1 cells and thus promoting an inflammatory status and enhancing the expression of key glycolysis genes in INR subjects.

Our study has some notable limitations. The number of patients per groups was relatively small, a phenomenon that is offset by the study design that makes the groups of study more comparable and also because the sample size in untargeted metabolomics is considered less relevant than in other types of studies. Also, no standard definition for the immunological response is available and therefore the threshold of 250 of CD4+ T cells/μl could seem somewhat arbitrarily set for the present study. In this regard, we have previously validated that patients receiving cART with CD4+ T-cell counts persistently below 250 cells/µl are poor immunological responders and are associated with worse clinical outcomes [10,12,15]. On the other hand, to avoid CD4+ T cells as a confounding factor, HIV-infected subjects were matched by pre-cART CD4+ T-cell count and other confounding factors, making our results more consistent in the search of a predictive metabolomic signature. However, lower median values in INR subjects can be observed compared with IR subjects, probably due to the relevance of the CD4+ T-cell count before cART onset as an intrinsic risk factor of immune failure to cART.

In conclusion, the present study presents novel and relevant data that corroborate our previous observation that inflammation and CD4+ T-cell turnover are strongly linked in this context [12] and contributes to a better understanding of the molecular mechanisms preceding a discordant response and immunological progression under suppressive cART. NMR and MS-based metabolomics can be prognostic immunological progression tools, and their combination with traditional soluble parameters increases the metabolomic biomarker value. The knowledge of how these metabolic pathways are interconnected and regulated provides new targets for future therapeutic interventions not only in HIV infection but also to other metabolic disorders.

Cell cycle regulation and T-cell turnover seem to promote the expression of pro-inflammatory molecules and down-regulate the characteristic pathways of quiescent cells in favor of glutaminolysis in an attempt to offset the CD4+ T-cell exhaustion in INR subjects even after 96 weeks of cART. A specifically reduced thymic output in INR subjects could be the driver for the high requirement of their profoundly depleted CD4+ T-cells to consequently switch on glutaminolysis, an alternative pathway of rapid energy production previously related to the proliferation of tumor cells. Additionally, metabolomic profiling supports previous studies suggesting that damage of GI tract and microbial translocation are associated with poor immune recovery in HIV-infected subjects under suppressive cART and highlight the diverse functionality of HDL particles, especially by their relation to glucose metabolism and the regulation of pro-inflammatory cytokines/chemokines in INR subjects.

Clinical perspectives

  • INRs are associated with worse long-term prognosis, including a higher risk of progression toward AIDS and non-AIDS-defining clinical events and death. Metabolomics profiling offers an additional value to traditional soluble markers.

  • Here, we show novel and relevant data that corroborate our previous observation that inflammation and CD4+ T-cell turnover are strongly linked in immune recovery.

  • The metabolomic signature of ART-naïve HIV subjects with a subsequent late immune recovery is the expression of pro-inflammatory molecules and glutaminolysis. The knowledge about how these metabolic pathways are interconnected and regulated provides new targets for future therapeutic interventions not only in HIV infection but also in other metabolic disorders such as human cancers.

Acknowledgements

The present study would not have been possible without the collaboration of all the patients, medical and nursing staff, and data managers who have taken part in the project. We thank Veronica Alba for her technical support, Montserrat Vargas and Alfonso J. Castellano for nursing support at the HIV Unit at Hospital Universitari Joan XXIII (Tarragona), and Lluis Gallart and Miriam Campos from IISPV-Biobank platform at Hospital Universitari Joan XXIII (Tarragona). We also thank the HIV BioBank integrated in the Spanish AIDS Research Network and collaborating Centres for the generous gifts of clinical samples used in this work. The HIV Biobank, integrated in the Spanish AIDS Research Network, is supported by Instituto de Salud Carlos III, Spanish Health Ministry (RD06/0006/0035, RD12/0017/0037 and RD16/0025/0019) as a part of the Plan National R+D+I and co-financed by ISCIII- Subdirección General de Evaluación y el Fondo Europeo de Desarrollo Regional (FEDER). The RIS Cohort (CoRIS) is funded by Instituto de Salud Carlos III through the Red Temática de Investigación Cooperativa en SIDA (RIS C03/173, RD12/0017/0018 and RD16/0002/0006) as part of the Plan National R+D+I and co-financed by ISCIII- Subdirección General de Evaluación y el Fondo Europeo de Desarrollo Regional (FEDER). We also thank the comments and criticisms of the anonymous reviewers that greatly helped to improve the manuscript.

Competing interests

The authors declare that there are no competing interests associated with the manuscript.

Funding

This work was supported by the Fondo de Investigacion Sanitaria [grant numbers PI10/02635, PI13/0796, PI16/00503, PI18/01216, PI19/01337]-ISCIII-FEDER; the Programa de Suport als Grups de Recerca AGAUR [grant numbers 2014SGR250, 2017SGR948]; the Gilead Fellowship Program [grant number GLD14/293]; the SPANISH AIDS Research Network [grant numbers RD12/0017/0005, RD16/0025/0006, RD16/0025/0019]-ISCIII-FEDER (Spain); the Agencia Estatal de Investigación (Acciones de carácter internacional ‘Europa Investigación’) [grant number EUIN2017-89297]; the Consejería de Economía, Innovación, Ciencia y Empleo, Junta de Andalucía (Proyecto de Investigación de Excelencia) [grant number CTS2593]; the Programa de Intensificación de Investigadores (ISCIII) [grant number INT15/226 (to F.V.)]; the Servicio Andaluz de Salud through Programa Nicolás Monardes [grant number C-0013-2017 (to Y.M.P.)]; the Acció Instrumental d’incorporació de científics i tecnòlegs (Departament de Salut, Generalitat de Catalunya) [grant number PERIS SLT002/16/00101 (to A.R.)].

Author contribution

All authors have seen and approved the submitted version of the manuscript. The authors’ contributions are as follows: experimental design (I.R.-S., E.R.-G., P.H., and A.R.) and intellectual guidance (J.P. and C.V.); recruitment of subjects (J.P., C.V., M.L., S.V., M.L.-D., J.B., F.F., F.G., E.B., R.P.) and sample procurement (I.R.-S. and E.R.-G.); data collection (I.R.-S., E.R.-G. and P.H.); data analysis and interpretation (I.R.-S., Y.M.P. and A.R.); manuscript preparation (I.R.-S., E.R.-G., Y.M.P. and A.R.). I.R.-S., E.R.-G., J.B., F.V., Y.M.P. and A.R. were responsible for the study design, data analysis, and article development. F.V., Y.M.P. and A.R. reviewed and edited the manuscript.

Abbreviations

     
  • ABCA1

    ATP-binding cassette transporter sub-family A member 1

  •  
  • AMPK

    5′ adenosine monophosphate-activated protein kinase

  •  
  • cART

    combined antiretroviral therapy

  •  
  • CoRIS

    cohort of the Spanish AIDS Research Network

  •  
  • CCR2-CCL2

    chemokine (c-c) motif receptor 2 - chemokine (c-c) motif ligand 2

  •  
  • DDP-4

    dipeptidyl peptidase-4

  •  
  • GI

    gastrointestinal

  •  
  • HDL-C

    high-density lipoprotein cholesterol

  •  
  • HIF-1

    hypoxia-inducible factor 1

  •  
  • HILIC

    hydrophilic interaction liquid chromatography

  •  
  • hsCRP

    high-sensitivity C-reactive protein

  •  
  • IL-6

    interleukin 6

  •  
  • INR

    immunological non-responder

  •  
  • IR

    immunological responder

  •  
  • Jak-STAT

    janus kinase/signal transducers and activators of transcription

  •  
  • KEGG

    kyoto encyclopedia of genes and genomes

  •  
  • LC

    liquid chromatography

  •  
  • LDL-C

    low-density lipoprotein cholesterol

  •  
  • LPS

    lipopolysaccharide

  •  
  • MDA

    mean decrease in accuracy

  •  
  • MS

    mass spectrometry

  •  
  • NALFD

    non-alcoholic fatty liver disease

  •  
  • NMR

    nuclear magnetic resonance

  •  
  • NRTI

    nucleoside reverse transcriptase inhibitor

  •  
  • PalC

    palmitoylcarnitine

  •  
  • PC

    phosphatidylcholine

  •  
  • PI

    protease inhibitor

  •  
  • PLS-DA

    partial least squares discriminant analysis

  •  
  • RF

    random forest

  •  
  • RORγt

    RAR-related orphan receptor gamma 2

  •  
  • TGF-β

    transforming growth factor-β

  •  
  • TH

    tyrosine hydroxylase

  •  
  • Th17

    T helper 17 cells

  •  
  • Treg

    regulatory T-cells

  •  
  • VLDL-C

    very-low density lipoprotein cholesterol

  •  
  • W

    Wilcoxon ttest

References

References
1.
Hernando
V.
,
Alejos
B.
,
Monge
S.
,
Berenguer
J.
,
Anta
L.
,
Vinuesa
D.
et al.
(
2013
)
All-cause mortality in the cohorts of the Spanish AIDS Research Network (RIS) compared with the general population: 1997-2010
.
BMC Infect. Dis.
13
,
382
2.
Battegay
M.
,
Nüesch
R.
,
Hirschel
B.
and
Kaufmann
G.R.
(
2006
)
Immunological recovery and antiretroviral therapy in HIV-1 infection
.
Lancet Infect. Dis.
6
,
280
287
[PubMed]
3.
Goicoechea
M.
,
Smith
D.M.
,
Liu
L.
,
May
S.
,
Tenorio
A.R.
,
Ignacio
C.C.
et al.
(
2006
)
Determinants of CD4+ T cell recovery during suppressive antiretroviral therapy: association of immune activation, T cell maturation markers, and cellular HIV-1 DNA
.
J. Infect. Dis.
194
,
29
37
[PubMed]
4.
Pacheco
Y.M.
,
Jarrín
I.
,
Del Amo
J.
,
Moreno
S.
,
Iribarren
J.A.
,
Viciana
P.
et al.
(
2009
)
Risk factors, CD4 long-term evolution and mortality of HIV-infected patients who persistently maintain low CD4 counts, despite virological response to HAART
.
Curr. HIV Res.
7
,
612
619
[PubMed]
5.
May
M.T.
,
Gompels
M.
,
Delpech
V.
,
Porter
K.
,
Orkin
C.
,
Kegg
S.
et al.
(
2014
)
Impact on life expectancy of HIV-1 positive individuals of CD4R cell count and viral load response to antiretroviral therapy
.
AIDS
28
,
1193
1202
[PubMed]
6.
Masiá
M.
,
Padilla
S.
,
Álvarez
D.
,
López
J.C.
,
Santos
I.
,
Soriano
V.
et al.
(
2013
)
Risk, predictors, and mortality associated with non-AIDS events in newly diagnosed HIV-infected patients: Role of antiretroviral therapy
.
AIDS
27
,
181
189
[PubMed]
7.
Gazzola
L.
,
Tincati
C.
,
Bellistrì
G.M.
,
d’Arminio Monforte
A.
and
Marchetti
G.
(
2009
)
The absence of CD4+ T cell count recovery despite receipt of virologically suppressive highly active antiretroviral therapy: clinical risk, immunological gaps, and therapeutic options
.
CID
48
,
328
337
8.
Casotti
J.A.S.
,
Passos
L.N.
,
Oliveira
F.J.P.
and
Cerutti
C.
(
2011
)
Factors associated with paradoxical immune response to antiretroviral therapy in HIV infected patients: a case control study
.
BMC Infect. Dis.
11
,
306
9.
Engsig
F.N.
,
Zangerle
R.
,
Katsarou
O.
,
Dabis
F.
,
Reiss
P.
,
Gill
J.
et al.
(
2014
)
Long-term mortality in HIV-positive individuals virally suppressed for >3 years with incomplete CD4 recovery
.
CID
58
,
1312
1321
10.
Pacheco
Y.M.
,
Jarrin
I.
,
Rosado
I.
,
Campins
A.A.
,
Berenguer
J.
,
Iribarren
J.A.
et al.
(
2015
)
Increased risk of non-AIDS-related events in HIV subjects with persistent low CD4 counts despite cART in the CoRIS cohort
.
Antiviral Res.
117
,
69
74
11.
Cenderello
G.
and
De Maria
A.
(
2016
)
Discordant responses to cART in HIV-1 patients in the era of high potency antiretroviral drugs: Clinical evaluation, classification, management prospects
.
Expert Rev. Anti. Infect. Ther.
14
,
29
40
[PubMed]
12.
Rosado-Sánchez
I.
,
Jarrín
I.
,
Pozo-Balado
M.M.
,
de Pablo-Bernal
R.S.
,
Herrero-Fernández
I.
,
Alvarez-Ríos
A.I.
et al.
(
2017
)
Higher levels of IL-6, CD4 turnover and Treg frequency are already present before cART in HIV-infected subjects with later low CD4 recovery
.
Anti. Res.
142
,
76
82
[PubMed]
13.
Rosado-Sánchez
I.
,
Herrero-Fernández
I.
,
Tarancon-Diez
L.
,
Moreno
S.
,
Iribarren
J.A.
,
Dalmau
D.
et al.
(
2018
)
Increased frequencies of Th17 cells and IL17a-producing regulatory T-cells preceding the immunodiscordant response to antiretroviral treatment
.
J. Infect.
76
,
86
92
14.
Sitole
L.J.
,
Williams
A.A.
and
Meyer
D.
(
2013
)
Metabonomic analysis of HIV-infected biofluids
.
Mol. Biosyst.
9
,
18
28
[PubMed]
15.
Rodríguez-Gallego
E.
,
Gómez
J.
,
Pacheco
Y.M.
,
Peraire
J.
,
Viladés
C.
,
Beltrán-Debón
R.
et al.
(
2018
)
A baseline metabolomic signature is associated with immunological CD4+ T-cell recovery after 36 months of ART in HIV-infected patients
.
AIDS
32
,
565
573
[PubMed]
16.
Rodríguez-Gallego
E.
,
Gómez
J.
,
Domigo
P.
,
Ferrando-Martínez
S.
,
Peraire
J.
,
Viladés
C.
et al.
(
2018
)
Circulating metabolomic profile can predict dyslipidemia in HIV patients undergoing antiretroviral therapy
.
Atherosclerosis
273
,
28
36
[PubMed]
17.
Cassol
E.
,
Misra
V.
,
Holman
A.
,
Kamat
A.
,
Morgello
S.
and
Gabuzda
D.
(
2013
)
Plasma metabolomics identifies lipid abnormalities linked to markers of inflammation, microbial translocation, and hepatic function in HIV patients receiving protease inhibitors
.
BMC Infect. Dis.
13
,
203
[PubMed]
18.
Hollenbaugh
J.A.
,
Munger
J.
and
Kim
B.
(
2011
)
Metabolite profiles of human immunodeficiency virus infected CD4+ T cells and macrophages using LC-MS/MS analysis
.
Virology
415
,
153
159
[PubMed]
19.
Scarpelini
B.
,
Zanoni
M.
,
Sucupira
M.C.A.
,
Truong
H.H.M.
,
Janini
L.M.R.
,
Segurado
I.D.C.
et al.
(
2016
)
Plasma metabolomics biosignature according to HIV stage of infection, pace of disease progression, viremia level and immunological response to treatment
.
PLoS ONE
11
,
e0161920
[PubMed]
20.
Caro-Murillo
A.M.
,
Castilla
J.
,
Perez-Hoyos
S.
,
Miro
J.M.
,
Podzamczer
D.
,
Rubio
R.
et al.
(
2007
)
Spanish cohort of naïve HIV-infected patients (CoRIS): Rationale, organization and initial results | Cohorte RIS de pacientes con infección por VIH sin tratamiento antirretroviral previo (CoRIS): Metodología y primeros resultados
.
Enferm. Infecc. Microbiol. Clin.
25
,
23
31
[PubMed]
21.
García-Merino
I.
,
de las Cuevas
N.
,
Jiménez
J.
,
Gallego
J.
,
Gómez
C.
,
Prieto
C.
et al.
(
2009
)
The Spanish HIV BioBank: a model of cooperative HIV research
.
Retrovirology
6
,
27
[PubMed]
22.
Mallol
R.
,
Amigó
N.
,
Rodríguez
M.A.
,
Heras
M.
,
Vinaixa
M.
,
Plana
N.
et al.
(
2015
)
Liposcale: a novel advanced lipoprotein test based on 2D diffusion-ordered 1H NMR spectroscopy
.
J. Lipid Res.
56
,
737
746
[PubMed]
23.
Wishart
D.S.
,
Jewison
T.
,
Guo
A.C.
,
Wilson
M.
,
Knox
C.
,
Liu
Y.
et al.
(
2013
)
HMDB 3.0—The Human Metabolome Database in 2013
.
Nucleic Acids Res.
41
,
D801
7
[PubMed]
24.
Szklarczyk
D.
,
Santos
A.
,
von Mering
C.
,
Jensen
L.J.
,
Bork
P.
and
Kuhn
M.
(
2016
)
STITCH 5: augmenting protein-chemical interaction networks with tissue and affinity data
.
Nucleic Acids Res.
44
,
D380
D384
[PubMed]
25.
Hulme
H.E.
,
Meikle
L.M.
,
Wessel
H.
,
Strittmatter
N.
,
Swales
J.
,
Thomson
C.
et al.
(
2017
)
Mass spectrometry imaging identifies palmitoylcarnitine as an immunological mediator during Salmonella typhimurium infection
.
Sci. Rep.
7
,
2786
[PubMed]
26.
Mutomba
M.C.
,
Yuan
H.
,
Konyavko
M.
,
Adachi
S.
,
Yokoyama
C.B.
,
Esser
V.
et al.
(
2000
)
Regulation of the activity of caspases by -carnitine and palmitoylcarnitine
.
FEBS Lett.
478
,
19
25
[PubMed]
27.
Tao
J.-H.
,
Barbi
J.
and
Pan
F.
(
2015
)
Hypoxia-inducible factors in T lymphocyte differentiation and function. A review in the theme: cellular responses to hypoxia
.
Am. J. Physiol. Cell Physiol.
309
,
C580
9
[PubMed]
28.
Dang E
V.
,
Barbi
J.
,
Yang
H.
,
Jinasena
D.
,
Yu
H.
,
Bordman
Z.
et al.
(
2011
)
Control of TH 17/Treg balance by hypoxia-inducible factor 1
.
Cell
146
,
772
784
[PubMed]
29.
Shi
L.Z.
,
Wang
R.
,
Huang
G.
,
Vogel
P.
,
Neale
G.
,
Green
D.R.
et al.
(
2011
)
HIF1α-dependent glycolytic pathway orchestrates a metabolic checkpoint for the differentiation of T H 17 and T reg cells
.
J. Exp. Med.
208
,
1367
1376
[PubMed]
30.
Palmer
C.S.
,
Hussain
T.
,
Duette
G.
,
Weller
T.J.
,
Ostrowski
M.
,
Sada-Ovalle
I.
et al.
(
2016
)
Regulators of glucose metabolism in CD4+ and CD8+ T cells
.
Int. Rev. Immunol.
35
,
477
488
[PubMed]
31.
Gerriets
V.A.
and
Rathmell
J.C.
(
2012
)
Metabolic pathways in T cell fate and function
.
Trends Immunol.
33
,
168
173
32.
Palmer
C.S.
,
Ostrowski
M.
,
Gouillou
M.
,
Tsai
L.
,
Yu
D.
,
Zhou
J.
et al.
(
2014
)
Increased glucose metabolic activity is associated with CD4+ T-cell activation and depletion during chronic HIV infection
.
AIDS
28
,
297
309
[PubMed]
33.
Arenzana-Seisdedos
F.
,
Fernandez
B.
,
Dominguez
I.
,
Jacqué
J.M.
,
Thomas
D.
,
Diaz-Meco
M.T.
et al.
(
1993
)
Phosphatidylcholine hydrolysis activates NF-kappa B and increases human immunodeficiency virus replication in human monocytes and T lymphocytes
.
J. Virol.
67
,
6596
6604
[PubMed]
34.
Palmer
C.S.
,
Duette
G.A.
,
Wagner
M.C.E.
,
Henstridge
D.C.
,
Saleh
S.
,
Pereira
C.
et al.
(
2017
)
Metabolically active CD4+ T cells expressing Glut1 and OX40 preferentially harbor HIV during in vitro infection
.
FEBS Lett.
591
,
3319
3332
[PubMed]
35.
Rosado-Sánchez
I.
,
Herrero-Fernández
I.
,
Genebat
M.
,
Del Romero
J.
,
Riera
M.
,
Podzamczer
D.
et al.
(
2018
)
HIV-infected subjects with poor CD4 T-cell recovery despite effective therapy express high levels of OX40 and α4β7 on CD4 T-cells prior therapy initiation
.
Front. Immunol.
9
,
1
6
[PubMed]
36.
Kipnis
J.
,
Cardon
M.
,
Avidan
H.
,
Lewitus
G.M.
,
Mordechay
S.
,
Rolls
A.
et al.
(
2004
)
Dopamine, through the extracellular signal-regulated kinase pathway, downregulates CD4+CD25+ regulatory T-cell activity: implications for neurodegeneration
.
J. Neurosci.
24
,
6133
6143
[PubMed]
37.
Huang
H.-W.
,
Zuo
C.
,
Chen
X.
,
Peng
Y.-P.
and
Qiu
Y.-H.
(
2016
)
Effect of tyrosine hydroxylase overexpression in lymphocytes on the differentiation and function of T helper cells
.
Int. J. Mol. Med.
38
,
635
642
[PubMed]
38.
Bergquist
J.
,
Tarkowski
A.
,
Ekman
R.
and
Ewing
A.
(
1994
)
Discovery of endogenous catecholamines in lymphocytes and evidence for catecholamine regulation of lymphocyte function via an autocrine loop
.
Proc. Natl. Acad. Sci. U.S.A.
91
,
12912
12916
[PubMed]
39.
Blachier
F.
,
Boutry
C.
,
Bos
C.
and
Tomé
D.
(
2009
)
Metabolism and functions of L-glutamate in the epithelial cells of the small and large intestines
.
Am. J. Clin. Nutr.
90
,
814S
821S
[PubMed]
40.
Crenn
P.
,
De Truchis
P.
,
Neveux
N.
,
Galpérine
T.
,
Cynober
L.
and
Melchior
J.C.
(
2009
)
Plasma citrulline is a biomarker of enterocyte mass and an indicator of parenteral nutrition in HIV-infected patients
.
Am. J. Clin. Nutr.
90
,
587
594
[PubMed]
41.
Palmer
C.S.
,
Anzinger
J.J.
,
Zhou
J.
,
Gouillou
M.
,
Landay
A.
,
Jaworowski
A.
et al.
(
2014
)
Glucose transporter 1-expressing proinflammatory monocytes are elevated in combination antiretroviral therapy-treated and untreated HIV+ subjects
.
J. Immunol.
193
,
5595
5603
[PubMed]
42.
Fujita
T.
,
Hada
T.
and
Higashino
K.
(
1999
)
Origin of D- and L-pipecolic acid in human physiological fluids: a study of the catabolic mechanism to pipecolic acid using the lysine loading test
.
Clin. Chim. Acta
287
,
145
156
[PubMed]
43.
Pols
T.W.H.
,
Puchner
T.
,
Korkmaz
H.I.
,
Vos
M.
,
Soeters
M.R.
and
De Vries
C.J.M.
(
2017
)
Lithocholic acid controls adaptive immune responses by inhibition of Th1 activation through the Vitamin D receptor
.
PLoS ONE
12
,
e0176715
[PubMed]
44.
Matsumoto
S.
,
Yamamoto
S.
,
Sai
K.
,
Maruo
K.
,
Adachi
M.
,
Saitoh
M.
et al.
(
2003
)
Pipecolic acid induces apoptosis in neuronal cells
.
Brain Res.
980
,
179
184
[PubMed]
45.
Merlini
E.
,
Bai
F.
,
Bellistrì
G.M.
,
Tincati
C.
,
d’Arminio Monforte
A.
and
Marchetti
G.
(
2011
)
Evidence for polymicrobic flora translocating in peripheral blood of HIV-infected patients with poor immune response to antiretroviral therapy
.
PLoS ONE
6
,
e18580
[PubMed]
46.
Marchetti
G.
,
Bellistrì
G.M.
,
Borghi
E.
,
Tincati
C.
,
Ferramosca
S.
,
La Francesca
M.
et al.
(
2008
)
Microbial translocation is associated with sustained failure in CD4+ T-cell reconstitution in HIV-infected patients on long-term highly active antiretroviral therapy
.
AIDS
22
,
2035
2044
[PubMed]
47.
Siegel
M.O.
,
Borkowska
A.G.
,
Dubrovsky
L.
,
Roth
M.
,
Welti
R.
,
Roberts
A.D.
et al.
(
2015
)
HIV infection induces structural and functional changes in high density lipoproteins
.
Atherosclerosis
243
,
19
29
[PubMed]
48.
Siebel
A.L.
,
Heywood
S.E.
and
Kingwell
B.A.
(
2015
)
HDL and glucose metabolism: current evidence and therapeutic potential
.
Front. Pharmacol.
6
,
258
[PubMed]
49.
van der Vorst
E.P.C.
,
Theodorou
K.
,
Wu
Y.
,
Hoeksema
M.A.
,
Goossens
P.
,
Bursill
C.A.
et al.
(
2017
)
High-density lipoproteins exert pro-inflammatory effects on macrophages via passive cholesterol depletion and PKC-NF-κB/STAT1-IRF1 signaling
.
Cell Metab.
25
,
197
207
50.
Spillmann
F.
,
Miteva
K.
,
Pieske
B.
,
Tschöpe
C.
and
Van Linthout
S.
(
2015
)
High-density lipoproteins reduce endothelial-to-mesenchymal transition
.
Arterioscler. Thromb. Vasc. Biol.
35
,
1774
1777
[PubMed]

Author notes

*

These authors contributed equally to this work.

Senior authors that contributed equally to this work.