Genome-scale metabolic models for natural and long-term drug-induced viral control in HIV infection

Abstract

Genome-scale metabolic models (GSMMs) can provide novel insights into metabolic reprogramming during disease progression and therapeutic interventions. We developed a context-specific system-level GSMM of people living with HIV (PLWH) using global RNA sequencing data from PBMCs with suppressive viremia either by natural (elite controllers, PLWH_EC) or drug-induced (PLWH_ART) control. This GSMM was compared with HIV-negative controls (HC) to provide a comprehensive systems-level metabo-transcriptomic characterization. Transcriptomic analysis identified up-regulation of oxidative phosphorylation as a characteristic of PLWH_ART, differentiating them from PLWH_EC with dysregulated complexes I, III, and IV. The flux balance analysis identified altered flux in several intermediates of glycolysis including pyruvate, α-ketoglutarate, and glutamate, among others, in PLWH_ART The in vitro pharmacological inhibition of OXPHOS complexes in a latent lymphocytic cell model (J-Lat 10.6) suggested a role for complex IV in latency reversal and immunosenescence. Furthermore, inhibition of complexes I/III/IV induced apoptosis, collectively indicating their contribution to reservoir dynamics.

Figure 1.. System-level transcriptomics signature in PLWH_ART.

(A) Digital cell-type quantification using Estimating the Proportions of Immune and Cancer cell (EPIC) methodology. Mean cell proportions estimated from the samples of each of the four cohorts are visualized in the bar graph. (B) Visualization of sample distribution using expression of combination antiretroviral therapy–specific genes and dimensionality reduction by UMAP. (C) Network visualization of pathways identified as significantly enriched by combination antiretroviral therapy–specific genes. Nodes are genes and edges represent association with pathways. Node size is relative to the mean expression of the genes among the PLWH_ART. Genes overlapping between pathways and high abundance genes are labeled.

Figure S1.. Box plot of the proportion of various cell types estimated using Estimating the Proportions of Immune and Cancer cell.

Mann–Whitney U test results are displayed between each cohort.

Figure 2.. Comparative analysis of PLWH_ART and PLWH_EC.

(A) Relative reservoir quantification using total HIV-1 DNA in PLWH_ART (n = 17) and PLWH_EC (n = 14). (B) MA plot showing differential gene expression results of PLWH_ART versus PLWH_EC. Negative log₂-fold change values represent down-regulation and positive values represent up-regulation in PLWH_ART. Grey-colored dots denote non-significant genes (adjusted P > 0.05). (C) Heatmap showing the expression pattern of significantly regulated genes between PLWH_ART and PLWH_EC (adjusted P < 0.05). Column annotation denotes cohort, age, gender, and duration of combination antiretroviral therapy of the corresponding samples. Row and column clustering was performed using Euclidian distance. (D) Gene set enrichment analysis results using MSigDB hallmark gene set between PLWH_ART versus PLWH_EC. A positive enrichment score represents up-regulation and negative score represents down-regulation in PLWH_ART. Statistically significant pathways are labeled and highlighted by asterisk. Bubble size is relative to the adjusted P-values of the pathways. *Indicates FDR < 0.2. (E) Schematic visualization of the five complexes of OXPHOS pathway. The heat-map shows expression pattern of genes belonging to OXPHOS pathway in PLWH_ART and PLWH_EC. Column annotation denotes OXPHOS pathway complexes and row annotation denotes the cohort. The bottom annotation shows the log₂ fold change values of the genes. Red color represents up-regulation and green color represents down-regulation of the gene in PLWH_ART compared with PLWH_EC.

Figure 3.. Context-specific genome-scale metabolic modeling and flux balance analysis.

(A) Network visualization of significant reporter metabolites (adjusted P < 0.2) identified in PLWH_ART versus PLWH_EC. Red-colored nodes represent up-regulated reporter metabolites and steel-blue colored nodes represent dysregulated (non-directional) reporter metabolites. (B) Workflow diagram of context-specific genome-scale metabolic model reconstruction. (C) Reaction diagram showing flux balance analysis results. Reactions show specific flux changes in PLWH_ART compared with PLWH_EC and HC cohorts highlighted with colored arrows. The direction of the arrow represents the flux change of the corresponding reaction in the cohort. (D) Communities identified from the topology analysis of the metabolic network in PLWH_ART, PLWH_EC, and HC. Node size is relative to betweenness centrality measurement. The top five ranked genes and metabolites based on betweenness centrality are labeled.

Figure S2.. Heat map visualizing expression pattern of enzymatic genes belonging to the reactions involving the significant reporter metabolites identified between PLWH_ART and PLWH_EC.

Figure S3.. Flow cytometry analysis of reactive oxygen species (ROS) production in HC (n = 18), PLWH_EC (n = 16), and PLWH_ART (n = 18).

Related to Fig 4. (A) Gating strategy for CD4⁺ T cells, CD8⁺ T cells, classical monocytes (CM), intermediate monocytes (IM), and non-classical monocytes (NCM) from total PBMCs. (B) Proportion of CD4⁺ and CD8⁺ T cells in HC, PLWH_EC, and PLWH_ART. (C) Proportion of CM, IM, and NCM in HC, PLWH_EC, and PLWH_ART. (D) UMAP shows the heat map distribution of ROS in HC, PLWH_EC, and PLWH_ART in lymphocytic and monocytic cell populations. (E) MFI of ROS in monocytic cell populations from PLWH_EC, short-term ART (<10 yr, n = 8), and long-term ART (>10 yr, n = 8). The histogram shows a representative sample from HC, PLWH_EC, and PLWH_ART exhibiting the median expression in each group. Statistical significance was determined using Mann–Whitney U test (P < 0.05 with * < 0.05, ** < 0.033, *** < 0.002) and represented with the median.

Figure 4.. Redox homeostasis during suppressive viremia.

(A) Reactive oxygen species (ROS) detection in lymphocytic and monocytic cell populations from HC (n = 18), PLWH_EC (n = 16), and PLWH_ART (n = 18). UMAP representation showing the distribution of lymphocytic (CD4⁺ and CD8⁺ T cells) and monocytic (classical monocytes [CM], intermediate monocytes [IM], and non-classical monocytes [NCM]) cell populations. (B) Median fluorescence intensity (MFI) of ROS in CD4⁺, CD8⁺, CM, IM, and NCM in the cohort. Histograms show a representative sample from HC, PLWH_EC, and PLWH_ART exhibiting the median expression in each group. (C) Graphs showing MFI of ROS in each individual from HC, PLWH_EC, and PLWH_ART. (D) MFI of ROS in PLWH_EC (n = 16), short-term ART (sART, n = 8), and long-term ART (lnART, n = 8). Histograms show a representative sample from PLWH_EC, sART, and lnART exhibiting the median expression in each group. Statistical significance was determined using Mann–Whitney U test (P < 0.05 with *<0.05, **<0.03, ***<0.002) and represented with median. See also Fig S3.

Figure 5.. Pharmacological inhibition of OXPHOS in lymphocytic HIV-1 latency cell model.

(A) Schematic representation of inhibition of OXPHOS complexes with metformin (complex I), aTOS (complex II), antimycin (complex III), arsenic trioxide (complex IV), and oligomycin (complex V). (B) Drug toxicity for 24 h treatment of OXPHOS inhibitors. (C) Annexin V positive cells after treatment with OXPHOS inhibitors and respective controls. (D) Activation from latency in J-Lat 10.6 cells after treatment with OXPHOS inhibitors and respective controls. (E) Percentage Ki-67 negative cells after treatment with OXPHOS inhibitors or respective controls. (F) Percentage PCNA negative cells after treating with Antimycin or DMSO control. (G) Western blot detection of H2A.X (S139) and β-Actin in Jurkat and J-Lat 10.6 after treatment with OXPHOS inhibitors or respective controls. (H) Quantification of H2A.X (S139). The graph shows fold change (Fc) of protein expression in relation to respective control after normalization to β-Actin. Experiments were performed in three biological replicates. Significance was determined using two-tailed t test (P < 0.05 with * < 0.05, ** < 0.033, *** < 0.002) and represented with mean and SD. Significance for each drug is compared with respective control. See also Figs S4 and S5. Source data are available online for this figure.

Figure S4.. Effect of OXPHOS inhibition in Jurkat and J-Lat 10.6 cell lines.

Related to Fig 5. (A) Cytotoxicity curves of metformin, aTOS, antimycin, arsenic trioxide, and oligomycin in Jurkat and J-Lat 10.6 treated for 24 h. The experiment was performed in technical triplicates. (B) Representative plots showing Annexin V/Viability staining of Jurkat and J-Lat 10.6 cells during OXPHOS inhibition for 24 h. (C) Representative plots showing activation from HIV latency (GFP) in J-Lat 10.6 cells during OXPHOS inhibition for 24 h.

Figure S5.. The effect of OXPHOS inhibition on cellular senescence markers in Jurkat and J-Lat 10.6 with metformin (complex I), aTOS (complex II), antimycin (complex III), arsenic trioxide (complex IV), and oligomycin (complex V).

Related to Fig 5. (A) Histogram of CD57 expression in Jurkat and J-Lat 10.6 cell lines. (B) Expression of CD57 after inhibition of OXPHOS complexes I-V in Jurkat and J-Lat 10.6. (C) Comparison of CD57 expression in Jurkat and J-Lat 10.6 cells. (D, E) Flow cytometry detection of Ki-67 (D) and PCNA (E) negative cell populations after inhibition of OXPHOS complex I-V in Jurkat and J-Lat 10.6. (F) Western blot detection of H2A.X (S139) and β-Actin in Jurkat and J-Lat 10.6 cells after inhibition of OXPHOS complex I-V. The graph shows the fold change (Fc) of protein expression in relation to respective control after normalization to β-Actin. Experiments were performed in three biological replicates. Significance was determined using a two-tailed t test (P < 0.05 with * < 0.05, ** < 0.03, *** < 0.002) and represented with mean and SD.