This non-interventional study was approved by the institutional review board of the “Comité de Protection des Personnes” (CPP) of Ile-de-France (VIH-PLAQUETTES, ID-RCB: 2020-A00307-32) and conforms to the principles outlined in the Declaration of Helsinki. Accordingly, all participants were informed in writing about the study and allowed not to participate.

Platelet-rich plasma (PRP) and peripheral blood mononuclear cell (PBMC) samples were obtained during routine blood testing of 32 HIV-infected individuals on cART from the French Hospital Database on HIV receiving care at the Ambroise Paré (Boulogne-Billancourt, France) and Raymond Poincaré (Garches, France) Hospitals. The enrollment criterions for the French Hospital Database on HIV were confirmed HIV-1 and cART initiated for at least 1 year before the time of blood sampling. Human PBMC and PRP samples from healthy HIV-seronegative donors used as a negative control for all experiments were obtained from the French blood collection center [Etablissement Francais du Sang (EFS) Paris, France]. PRP and PBMCs were prepared as previously described (5).

InR subjects were defined as having sustained CD4⁺ T-cell counts below 350 cells/μl for more than 6 months during viral suppression, i.e., plasma viral load copies below limit of detection (LOD) as detected by the Abbott RealTime HIV-1 assay on an automated m2000 system. IR subjects were defined as having sustained CD4⁺ T-cell counts above 350 cells/μl during virological suppression during this time. Clinical information on InR and IR subjects is presented in Table 1 .

To measure comparatively infectious viral production from CD4⁺ T cells and macrophages, we adapted our previously described VOA using CD4⁺ T cells from healthy donors as reporter cells.

Real-time cell metabolic analysis was carried out by Seahorse XFp Analyzer using the ATP Rate Assay Kit (Agilent) and following manufacturer’s instructions. Platelets from InRs, IRs, or healthy donors were cultivated at 37°C overnight with healthy donor’s CD4⁺ T cells or macrophages (pool of three donors, 5 × 10⁵ cells/well) in 48-well plates or in Seahorse XFp cell culture plates, respectively. After thorough wash out of unconjugated platelets, cell metabolism was further analyzed. The platelet:leukocyte ratio employed was 2:1 and 5:1 for CD4⁺ T cells and macrophages, respectively. For CD4⁺ T cells that had interacted with platelets, T cells were centrifuged at 300 g for 10 min, resuspended in Seahorse XF RPMI medium, and placed into Seahorse XFp cell culture plates pretreated with Cell-Tak (Corning, Fisher Scientific Inc.) following manufacturer’s instructions. For macrophages that interacted with platelets directly in the Seahorse XFp cell culture plates, complete medium was replaced by Seahorse XF RPMI medium. The ATP Rate assay protocol established for the Seahorse XFp Analyzer by the manufacturer provides dynamic information on the total ATP energy production by mitochondrial respiration via oxidative phosphorylation (OXPHOS) and/or by glycolysis in living cells. Basic oxygen consumption rate (OCR), which indicates aerobic respiration by OXPHOS, and extracellular acidification rate (ECAR), which indicates glycolysis, were measured over 1 h at indicated time points before and after successive addition of OXPHOS inhibitor oligomycin and the mitochondrial respiratory inhibitors rotenone/antimycin. Results were analyzed by the Agilent Seahorse Analytics online software (seahorseanalytics.agilent.com) to assess the levels of OCR (pmol/min), ECAR (mpH/min), and ATP production rate (pmol/min). A metabolic cell energy map was built by correlating OCR vs. ECAR values, displaying data in a spectrum of metabolic profiles comprising aerobic (predominant ATP production via mitochondrial respiration), glycolytic (predominant ATP production via glycolysis), quiescent (low ATP production rate by both pathways), or energetic (high ATP production rate by both pathways). Results were displayed as mean with standard errors of values obtained from biological replicates of platelet samples (InR n = 6; IR n = 5; and healthy donors n = 6).

The public microarray dataset GSE106792 from Gene Expression Omnibus (GEO) comparing the gene expression of CD4⁺ T cells from InRs, IRs, and healthy donors (17) was reassessed using the GEO2R tool provided by the GEO website. Datasets were grouped in InRs, IRs, and healthy donors according to the Immune Non-Responders, Immune Responders, and Healthy Donors donor class metadata, respectively, but disregarding the source name metadata related to CD71 expression. Dimensional reduction was performed by Uniform Manifold Approximation and Projection (UMAP) of all individual datasets (from InRs, IRs, and healthy donors) as provided by GEO2R software. Confidence ellipses were plotted on UMAP in R software using UMAP coordinates. Volcano plots were generated in SPSS software (IBM) using the data output of the volcano plots comparing InR vs. IR gene expression generated by GEO2R software. The genes most significantly upregulated in InRs vs. IRs in the differentially expressed gene (DEG) analysis were selected based on Log₂ fold change >0.5 and the -Log₁₀ p value >3. The lists of DEG in InRs vs. IRs were used for biological pathway enrichment analysis and visualization of the biological processes enriched in InRs vs. IRs as described (18).

These lists were also submitted to g:Pprofiler (https://biit.cs.ut.ee/gprofiler/gost) (19). The statistical tests were performed by the g:GOSt functional profiling tool, and p values were calculated as described in https://biit.cs.ut.ee/gprofiler/page/docs, “section significance threshold”. Briefly, the tool uses an in-house algorithm called g:SCS that performs multiple testing corrections outperforming the commonly used Bonferroni correction (BC) or Benjamini–Hochberg False Discovery Rate (FDR).

The graphical representation of the pathway enrichment analysis was performed by applying g:Profiler output into a Cytoscape Enrichment Map as described (18). Nodes represent gene sets that participate in a given function, and their size represents the number of genes composing this functional cluster. Lines connecting nodes indicate an overlap between gene sets composing both connected nodes. The line width represents the number of genes common to both nodes.

Next, the list of DEGs in InRs vs. IRs was submitted to HumanCyc metabolic map software (20) to visualize the contribution of DEGs on different metabolic pathways in InRs vs. IRs.

All statistical analyses were performed using SPSS software (IBM) and parametric and non-parametric tests for normal and non-normal data distributions, respectively. Statistical significance is indicated by asterisks in figures and was established by p values <0.05.

To address the immunomodulatory role of platelets on CD4⁺ T cells in immunological failure, we first investigated the presence of platelet–CD4⁺ T-cell conjugates in PBMCs from InRs, IRs, and healthy donors by flow cytometry. The median frequency of platelet–CD4⁺ T-cell conjugates among CD4⁺ T cells was 11% in InRs as compared with 6% in IRs and 3.5% in healthy controls ( Figure 1A ), indicating that platelets from InRs form more conjugates with CD4⁺ T cells than platelets from IRs in vivo that could result in deleterious effects to T cell functions.

The InR subjects tested for platelet–CD4⁺ T-cell conjugates harbor HIV in platelets as quantified by FISH-flow cytometry (see Materials and Methods), in contrast to IR subjects whose platelets were negative for viral components. A consequence of the intimate contact between HIV-containing platelets from InR and T cells when conjugates formed could be the transfer of HIV productive infection from platelet to T cells and in turn in their production of infectious viruses. To measure these infectious viruses, we established a viral outgrowth assay (VOA) in which HIV-containing platelets from InRs were allowed to interact directly with healthy donor CD4⁺ T cells or macrophages. Resulting productive infection of CD4⁺ T cells or macrophages was then quantified ( Figure 1B ). No outgrowth was detected in healthy donor’s CD4⁺ T cells (serving also as infection reporter cells) after interaction with HIV-containing platelets from InRs, even in the presence of polybrene, a facilitator of membrane fusion ( Figure 1C , left). In contrast, we observed that the transfer of HIV sheltered in InR platelets to macrophages is productive, as infection spreads replication-competent virus to healthy donor’s reporter CD4⁺ T cells ( Figure 1C , right). Adding the anti-GPIIbIIIa platelet drug abciximab to the platelet–macrophage coculture blocked this platelet-mediated HIV transfer to macrophages in vitro, as we demonstrated previously (5). Thus, unlike macrophages, infectious HIV enclosed in InR platelets does not target CD4⁺ T cells, although HIV-containing platelets might immunomodulate CD4⁺ T-cell functions, in turn triggering the immunological failure observed in non-responders.

Different studies have demonstrated the role of cellular metabolism in the innate and adaptive host responses to infection (21), especially in T-cell immunity (22, 23). To get insight on the CD4⁺ T cell functional changes induced by InR platelets, we reassessed a transcriptome dataset obtained by Younes et al. (17), reporting on different transcriptome signatures presented by InR, IR, and healthy donor CD4⁺ T cells. We now used an improved analytical pipeline comprising also metabolic pathway networks (18, 20). Our revised grouping of InR, IR, and healthy donor samples relied only on the provided metadata of clinical status but discarding the CD71 marker used to substratify InR samples as performed in the original analysis (17). Transcriptomic data of InR, IR, and healthy donor as we grouped and compared in a dimensional reduction strategy showed that differences between the three groups remained conserved ( Figure 2A ).

We then thus focused our differential analysis on two major metabolic pathways implicated in T-cell activated vs. resting states: OXPHOS and glycolysis (24). Among the genes upregulated in InR CD4⁺ T cells as compared with IR counterparts in a DEG analysis, a significant number was implicated in glycolytic pathways ( Figure 2B ). Upregulated genes in InRs included Tyrosyl-DNA phosphodiesterase 1 (TDP1) that decreases phosphoglycolate formation, and phosphoglycolate is known to have an effect on two steps in glycolysis (25); Leucine Rich Repeat Neuronal 3 (LRRN3), a clinically relevant biomarker of immune status in HIV-1 infection whose upregulation correlates with a less senescent T-cell phenotype (26), that binds to LEDGF/p75 or LEDGF/p52 involved in HIV-1 integration (27) and associates with glycolysis canonical gene products ENO2, HK2, PFKFB3; Phosphofructokinase muscle type (PFKM) that is a subunit of phosphofructokinase enzyme, a central controller of the mammalian glycolytic pathway (28); Homeodomain-interacting protein kinase-2 (HIPK-2) that is a member of a family of proteins upregulated in highly proliferative tumors whose expression correlates with upregulation of genes involved in aerobic glycolysis (29); Progesterone Receptor Membrane Component protein-2 (PGRMC2) that is increased in glycolysis (30); and finally, STK36 that plays an important role in the Sonic hedgehog (SHH) pathway that regulates the activity of GLI transcription factors and drives glucose metabolism (31, 32). When analyzed together for functional enrichment using g:Profiler (19), genes specifically upregulated in InRs contributed to two main biological pathways about metabolism, namely, cellular metabolic processes (GO:0044237) and macromolecule metabolic pathways (GO:0043170), whereas those increased in IRs concern macromolecule catabolic processes (GO:0044265) and mRNA metabolic processes (GO:0016071) ( Figure 2C ).

Next, to get a better insight into how these biological pathways interact together, we built a network of biological processes most significantly enriched in CD4⁺ T cells from InRs as compared with those from IRs. Three main networks of biological processes emerged: “vesicle mediated localization”, “processing peptide antigen”, and “regulation metabolic process” ( Figure 2D ). The “regulation metabolic process” cluster connects biological pathways enriched in both InR and IR CD4⁺ T cells. However and in agreement with g:Profiler biological pathway analyses, whereas the nodes connecting biological pathway in IR CD4⁺ T cells relate mainly to mRNA processing, those connecting biological pathways in InR CD4⁺ T cells are mainly composed of macromolecule metabolic processes connected with pathways involved in ATP production. To better define the genes upregulated in InR CD4⁺ T cells directly implicated in ATP production, we submitted the list of DEGs in InR vs. IR T cells to the HumanCyc database (20) to build a metabolic map. This map showed which enzymes that are implicated in biochemical reactions leading to ATP production would be differentially regulated in IR and InR CD4⁺ T cells. Although both sets of cells harbor differentially upregulated genes for enzymes involved in glycolysis, pyruvate oxidation, and Tricarboxylic Acid TCA cycle, InR CD4⁺ T cells show a remarkable upregulation of genes coding for enzymes involved in specific reactions of the glycolytic pathway ( Figure 2E ). Notably, Hexokinase 2 (HK2) is responsible for breaking glucose in D-glucopyranose-6-phosphate, ATP-dependent 6-phosphofructokinase muscle type (PFKM) is responsible for breaking fructose-6P in fructose-1,6-diphosphate, and, very importantly, Lactate Dehydrogenase B (LDHB) is responsible for breaking pyruvate (the end product of glycolysis) into L-lactate. This last reaction is a hallmark of aerobic glycolysis, a metabolic reaction performed by activated effector T cells (33). The upregulation of LDHB gene in InR as compared with IR CD4⁺ T cells indicates that InR CD4⁺ T cells have increased aerobic glycolysis or Warburg effect, a characteristic of T-cell activation states (34). Altogether, the analysis of CD4⁺ T-cell transcriptomic data indicates that InR CD4⁺ T cells are more prone to aerobic glycolysis than their IR counterparts, but the potential role of increased platelet–T-cell conjugates we observed in InRs in this process could not be assessed.

To approach this question experimentally, we cocultured InR, IR, or healthy donor platelets with healthy donor CD4⁺ T cells and evaluated the platelet-induced metabolic profile of these T cells. Accordingly, glycolysis and mitochondrial respiration in CD4⁺ T cells after coculture with platelets were evaluated simultaneously and in live cells by measuring ECAR and OCR, respectively. CD4⁺ T cells were sequentially treated with oligomycin (ATP synthase and OXPHOS inhibitor) and antimycin A/rotenone (mitochondrial complex III and complex I inhibitors) to functionally assess OXPHOS and mitochondrial respiration capacities. InR platelets positive for the presence of HIV in platelets induced in CD4⁺ T cells a glycolysis baseline increase compared with InR lacking HIV in platelets, IR (positive or negative of HIV in platelets), or healthy donor platelets ( Figure 3A , upper graph). No difference was observed in mitochondrial respiration baseline induced by platelets from the five studied conditions with the exception of InR platelets positive for the presence of HIV in platelets, capable of inducing an increased baseline mitochondrial respiration ( Figure 3A , lower graph). As expected in this system, when the mitochondrial inhibitors were added sequentially, the mitochondrial respiration was drastically inhibited in all groups, and a compensatory increase in glycolysis was observed in CD4⁺ T cells that was remarkable and more prominent after interaction with InR platelets positive for the presence of HIV as compared with all the other conditions ( Figure 3A ).

By correlating mitochondrial respiration capacities with the glycolytic compensatory increase, we built a bioenergetics phenotype map for CD4⁺ T cells that interacted with InR or IR platelets, either positive or negative for HIV in platelets, or healthy donor platelets. In contrast with InR platelets without HIV, IR platelets positive or negative for HIV or healthy donor platelets, in which ATP production relies mostly on mitochondrial respiration (aerobic profile), InR platelets positive for HIV induce a more energetic phenotype in CD4⁺ T cells, in which the ATP production is promoted by both mitochondrial respiration and glycolysis, indicating aerobic glycolysis ( Figure 3B ).

We then compared the contribution of mitochondrial respiration and glycolysis to ATP production in CD4⁺ T cells vs. macrophages after interaction with InR platelets positive for HIV, IR platelets negative for HIV, or healthy donor platelets using the same dynamic metabolic measurements. Macrophages inherently produce more ATP than CD4⁺ T cells upon interaction with all platelets tested, with macrophage ATP being produced mostly by mitochondrial respiration. In contrast, an augmented ATP production resulting mainly from an increase in ATP generated by glycolysis was observed in CD4⁺ T cells that interacted with InR compared with IR or healthy donor platelets ( Figure 3C ).

We next compared specifically the ATP production via glycolysis induced in CD4⁺ T cells after interaction with InR or IR platelets, positive or negative for HIV, or healthy donor platelets. In contrast to the other experimental conditions in which ATP production via glycolysis is similar whatever the type of platelets they had interacted with, interaction between InR platelets positive for HIV and CD4⁺ T cells resulted in higher ATP production via glycolysis compared with interaction with IR platelets, independently of the presence of HIV, or healthy donor platelets ( Figure 3D ).