CD32⁺CD4⁺ memory T cells are enriched for total HIV-1 DNA in tissues from humanized mice

Summary

CD32 has raised conflicting results as a putative marker of the HIV-1 reservoir. We measured CD32 expression in tissues from viremic and virally suppressed humanized mice treated relatively early or late after HIV-1 infection with combined antiretroviral therapy. CD32 was expressed in a small fraction of the memory CD4⁺ T-cell subsets from different tissues in viremic and aviremic mice, regardless of treatment initiation time. CD32⁺ memory CD4⁺ T cells were enriched in cell-associated (CA) HIV-1 DNA but not in CA HIV-1 RNA as compared to the CD32⁻CD4⁺ fraction. Using multidimensional reduction analysis, several memory CD4⁺CD32⁺ T-cell clusters were identified expressing HLA-DR, TIGIT, or PD-1. Importantly, although tissue-resident CD32⁺CD4⁺ memory cells were enriched with translation-competent reservoirs, most of it was detected in memory CD32^-CD4⁺ T cells. Our findings support that CD32 labels highly activated/exhausted memory CD4⁺ T-cell subsets that contain only a small proportion of the translation-competent reservoir.

Graphical abstract

Highlights

• •CD32 is rarely expressed in memory CD4⁺T cells in humanized mice infected with HIV-1
• •Tissue-resident CD32⁺CD4⁺ memory T cells are enriched for HIV-1 DNA but not for RNA
• •CD32⁺CD4⁺ memory cells are enriched for translation-competent reservoirs
• •CD32 labels highly activated/exhausted memory T-cell subsets in tissues

Molecular Biology; Immunology ; Virology; Cell Biology

Introduction

The advent of combined antiretroviral therapy (cART) has dramatically improved the clinical care of HIV-1 infection. Since the START trial demonstrated that immediate initiation of cART is superior in reducing serious acquired immunodeficiency syndrome (AIDS)- and non-AIDS-related events (Lundgren et al., 2015), treatment guidelines advise to start therapy as early as possible (Panel on Antiretroviral Guidelines for Adults and Adolescents, 2017). Although beneficial in reduction of morbidities, early treatment started even during acute infection is not curative due to the establishment of long-lived viral reservoirs (Chun et al., 1998; Archin et al., 2012; Deng and Siliciano, 2014; Colby et al., 2018). Viral reservoirs, quantified by total HIV-1 DNA, are generally described as all cells infected with HIV-1, irrespective of their replication status or competence (Avettand-Fenoel et al., 2016). Viral persistence of replication-competent HIV-1 under cART, on a timescale of years, is however attributed to the latent portion of viral reservoirs. Therefore, using a practical definition (Eisele and Siliciano, 2012), latently infected cells that contain dormant but replication-competent provirus and are capable to resume viral production at a future time point represent the HIV-1 reservoir.

Latently infected cells have a half-life of 43.9 months, which makes them extremely resistant to antiretroviral therapy (ART). Therefore, discontinuation of treatment will, with great certainty, be followed by a viral rebound. The establishment of HIV-1 reservoirs occurs very rapidly upon primary infection. In simian immunodeficiency virus, reservoirs are found as early as three days post-infection in various tissues, even before virus detectability in circulation (Whitney et al., 2014). Likewise, in humans, initiation of cART during the acute phase of infection does not impede reservoir formation (Chun et al., 1998; Colby et al., 2018). Latent reservoirs are extremely stable with an estimated half-life of 43.9 months. This significantly hinders cART ability to eliminate HIV-1 within a patient's lifetime, with viral rebound highly likely to happen after treatment interruption (Siliciano et al., 2003b; Besson et al., 2014; Buzon et al., 2015; Crooks et al., 2015). The frequency of latently infected cells is very low (approximately 1 in 10⁵-10⁸ of CD4⁺ T cells) (Finzi et al., 1997; Siliciano et al., 2003a; Strain et al., 2005), resulting in enormous technical challenges to precisely determine and study HIV-1 latency. Originally, the highest frequency of latently infected cells was thought to be localized in resting central memory CD4⁺ T cells (Chomont et al., 2009; Descours et al., 2012; Sáez-Cirión et al., 2013), but its heterogeneous dissemination over various CD4⁺ T-cell memory subsets has been now recognized (Buzon et al., 2014; Banga et al., 2016; Zerbato et al., 2019; Kwon et al., 2020). In addition, the tissue distribution of reservoirs is diverse and poorly reflected by recirculating CD4⁺ T cells in peripheral blood (Blum and Pabst, 2007; De Scheerder et al., 2019). Against this background, the characterization of cellular biomarkers of latently infected cells would enormously facilitate the development of strategies for their targeted depletion.

Various surface markers (CXCR3, CCR6, CD2, PD-1, LAG-3, TIGIT) were proposed to identify latently infected CD4⁺ T cells but studies have not converged (Iglesias-Ussel et al., 2013; Anderson et al., 2014; Banga et al., 2018; Fromentin et al., 2016; Gosselin et al., 2017). The Fc-γ-RII receptor CD32 was recently proposed as a putative marker of cells highly enriched for total HIV-1 DNA and for replication-competent provirus (Descours et al., 2017). These findings have attracted a lot of attention but some recent works have relinquished inconclusive results (Abdel-Mohsen et al., 2018; Bertagnolli et al., 2018; Martin et al., 2018; Noto et al., 2018). CD32⁺ CD4⁺ T cells were shown to harbor a higher activation profile and to correlate with general immune activation. Overall, data from several patient cohorts did not substantiate the relevance of CD32 expression in latent HIV-1 reservoir probably due to the heterogeneity of cART initiation's timing, cART duration, and levels of immune activation. Recently, the different purity of CD32⁺CD4⁺ T cells across various studies was proposed as a possible explanation for these divergent findings (Abdel-Mohsen et al., 2018; Darcis et al., 2019). Furthermore, only few studies have scrutinized the role of CD32 in tissues (Noto et al., 2018; Cantero-Pérez et al., 2019; Thornhill et al., 2019).

Molecular mechanisms of HIV-1 latency have been mainly studied in cell lines, but these data have very limited relevance within clinical settings (Han et al., 2007; Van Lint, Bouchat and Marcello, 2013; Whitney and Brad Jones, 2018). Instead, humanized mice models have earned their place in HIV-1 research as versatile tools to study HIV-1 pathogenesis in vivo (Marsden and Zack, 2017). To date, cART treatment has reproduced the features of HIV-1 latency in these small animals (Marsden et al., 2017; Satheesan et al., 2018). The advantages of such models are the easy access to all anatomical sites and the precise control over timing and duration of treatment allowing a detailed understanding of HIV-1 latency under various conditions.

We have developed an HIV-1 latency model in humanized Nonobese diabetic/Severe Combined ImmunoDeficient gamma (NSG) mice to investigate the early establishment of the viral reservoir in untreated, early (one week after HIV-1 infection), and late treated mice (seven weeks after HIV-1 infection) with cART. Our study examined the expression of CD32 as a marker of viral reservoirs in memory CD4+T cells with capabilities of long-term persistence and its relationship to immune activation. Using a strict flow cytometry gating strategy, we show here that CD32 expression was low in CD4⁺ T cells from different tissues in humanized mice treated early or late and was mostly due to the expression of the CD32a isoform. CD32⁺CD4⁺ memory T cells were enriched in cell-associated HIV-1 DNA (CA HIV-1 DNA) but not in cell-associated HIV-1 RNA (CA HIV-1 RNA), indicating HIV latency in these cells. CD32⁺CD4⁺ memory subsets were associated with their activation/exhaustion status under cART, and cluster analysis revealed that CD32⁺CD4⁺ memory cells are phenotypically related to cells expressing HLA-DR, PD-1, or TIGIT. Importantly, most of the p24 productive reservoir upon latency reversal was situated in the CD32^- memory CD4⁺ T cell fraction. Taken together, our data indicate that CD32 expression describes activated CD4⁺ T-cell memory subsets enriched for proviral DNA independently of treatment initiation. Ultimately, we showed here that CD32 does not clearly label the translation-competent HIV-1 reservoir that needs to be eliminated in the perspective of HIV-1 cure.

Results

Early cART treatment leads to faster viral suppression than late treatment in humanized mice

In this study, NSG mice were “humanized” with CD34+ hematopoietic stem cells over 24 weeks and then infected with HIV-1 JR-CSF Infectious Molecular Clone (pYK-JRCSF). One group was left untreated, a second group was treated early (one week after HIV-1 infection), and a third group was treated late (seven weeks after HIV-1 infection) (Figure 1).

Figure 1.

Design of the Study

One hundred twenty humanized NSG mice were infected with HIV-1 JRCSF. At week one post-infection, the cART regimen was initiated (n = 58) in the early treated group and sustained for two months. Late treatment was started at seven weeks post-infection (n = 42) and continued for two months. In each group, a number of animals were sacrificed at the end of treatment, corresponding to week nine for early treated (n = 43) and to week fifteen for late treated (n = 34) animals. The untreated group (n = 20) never received cART and was euthanized correspondingly to the sacrifices of both treatment groups (week nine for early treated and week fifteen for late treated mice). Viral rebound after treatment interruption was monitored for the remaining early (n = 15) and late treated (n = 8) animals for a duration of six weeks or longer if animals were not detectable for HIV-1 viral load.

In each group, oral treatment was maintained for eight weeks. As expected, in all conditions, a sharp increase of plasma HIV-1 RNA in the range of 10⁴-10⁵ copies per ml was detected during the first week after infection (Figure 2A). Subsequently, plasma viremia was strongly reduced within two weeks of either early or late treatment. Interestingly, early treatment led to a significantly quicker viral suppression (i.e. undetectability of the viral load) (p < 0.0001, Figure 2C). Conversely, the kinetics of viral rebound after treatment interruption were similar for both treatment groups (Figure 2D). Importantly, we observed that a portion of early treated mice rebounded only transiently (four out of fifteen or 26.7%) with “spontaneous” viral suppression up to six weeks after treatment cessation, whereas this “secondary control” phenomenon was not observed in the late treated group (Figures S1A and S1B). We measured HIV-1 RNA in the bone marrow, spleen, and lymph nodes of early treated mice as compared to untreated viremic mice after one week of HIV-1 infection. Viral loads were almost undetectable in the bone marrow of the treated mice, whereas some ongoing HIV-1 replication still occurs in the spleen and in the lymph nodes as shown by a detectable viral load in these two tissues (Figures S1C and S1D). Taken together, these observations indicate the establishment of replication-competent viral reservoirs within one week of infection in this model.

Figure 2.

Early Treatment Significantly Decreased Time to Viral Suppression

(A and B) (A) Plasma viral load of early treated (n = 43), late treated (n = 34), and untreated (n = 20) mice and (B) percentage of CD4⁺ T cells were determined every two weeks in peripheral blood. Background coloring of graphs highlights treatment periods: light green for early treatment, violet for late treatment, dark green is obtained by a two-week overlap (week seven to week nine) of early and late treatment periods (figure A and B).

(C and D) (C) Time to viral suppression (median two and four weeks, respectively) (p < 0.0001) and (D) viral rebound (p = 0.107) for the early and late group. Statistical analysis was executed using Mantel COX rank test. (∗∗∗p < 0.005; ns, nonsignificant test p > 0.05).

The immunological hallmark of HIV-1 infection is the depletion of CD4⁺ T cells. As expected, we observed a rapid decline in circulating CD4⁺ T cells upon infection. At week three post-infection, untreated mice lost around one third of their human CD4⁺ T cells, whereas early treatment prevented CD4⁺ T-cell loss (Figure 2B). Late treatment onset allowed a gradual recovery of CD4⁺ T cells. At the endpoint (week sixteen), the late treated mice showed peripheral blood CD4⁺ T-cell counts restored to similar levels when compared to the early treated mice (Figure 1). In agreement with our findings from peripheral blood (Figure 2B), the relative number of human CD4⁺ T cells was also significantly increased by cART in the spleen for both treatment groups (Figures S2 and 2A). These data suggest recovery of the human CD4⁺ T-cell compartment even after late cART onset in our model.

Total HIV-1 DNA and T-cell exhaustion are significantly reduced by early cART treatment

The stable persistence of provirus during therapy is a defining factor of HIV-1 latency. Therefore, we analyzed three different organ sites (bone marrow, lymph nodes, and spleen) for the presence of CA HIV-1 DNA by droplet digital Polymerase Chain Reaction (ddPCR), either at the end of cART treatment or corresponding time points for the non-treated animals (see experimental design in Figure 1). cART led to a significant reduction of CA HIV-1 DNA compared to the untreated control group (Figure 3A). Furthermore, early initiation of cART significantly reduced CA HIV-1 DNA as compared to late initiation in bone marrow (1.82 and 3.85 median log₁₀ copies/million cells_, respectively) and lymph nodes (2.45 and 3.7 median log₁₀ copies/million cells, respectively), with a similar but non-significant trend observed in the spleen. These data underline that humanized mice mimic reservoir establishment and persistence during cART treatment similar to humans.

Figure 3.

cART Significantly Decreased Total HIV-1 DNA in Peripheral Organs and HIV-1-Related Immune Exhaustion and Activation on Human T Cells

(A) Total HIV-1 DNA quantified from the bone marrow, lymph nodes, and spleen is represented for early treated (n = 9), late treated (n = 11), and untreated mice (n = 6). Percentages of cells expressing the respective markers PD-1, TIGIT, and HLA-DR on CD4+ T cells (B) and CD8+ T cells (C) extracted from the spleen at experimental endpoint (median age of nine months, eight weeks total cART duration for both treatment groups) are depicted for early treated mice (n = 17), late treated mice (n = 23), and untreated mice (n = 16). Statistical testing was performed using non-parametric Kruskal-Wallis tests with multiple comparison Dunn's correction. (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.005, ∗∗∗p < 0.001, ns, nonsignificant test, p > 0.05).

Infection with HIV-1 induces not only loss of CD4⁺ T cells in humans but also long-term alterations of the activation status in CD4⁺ and CD8⁺ T cells which persist during cART (Hazenberg et al., 2000; Hunt et al., 2003). In our humanized mice model, we assessed T-cell phenotypes in the tissues at the end of treatment and corresponding time points for untreated animals by flow cytometry. The expression of the exhaustion markers PD-1 and TIGIT was decreased by cART treatment on both CD4+ and CD8+ T cells, but the decrease was only significant between untreated and early treated groups (Figures 3B and 3C). Interestingly, the activation marker HLA-DR showed no difference among groups (Figures 3B and 3C). In addition, we observed a normalization of CD27 expression on CD4+ T cells by both treatment regimens (Figures S2B), suggesting a recovery in T-cell differentiation in our humanized mice model.

CD32⁺CD4⁺ memory T-cells do not contain T-B cell conjugates and harbor higher amounts of CA HIV-1 DNA but not of HIV-1 CA RNA

Next, we investigated the expression of CD32, a proposed marker of the HIV-1 reservoir. Given the previously reported technical difficulties to obtain pure CD32⁺CD4⁺ T cells (Martin et al., 2018; Osuna et al., 2018; Pérez et al., 2018), we applied an elaborated flow cytometry gating strategy to isolate CD4⁺ memory T cells (defined as CD45RA⁻) excluding doublets. To preclude contaminations by doublets, B cells or monocytes, strict singlet gating (on area versus width for forward and side scatter), as well as exclusion of CD19 and CD14 signals, were performed to analyze and sort CD32⁺CD4⁺ T cells by flow cytometry (Figure S3). The mean post-sorting purity of the CD32⁺CD4⁺ T cells was 98.8% of CD4⁺ T-cell singlets (Figure S4). We observed also that human CD4⁺ T cells contained a CD32 ^dim population with low expression of CD32 in contrast to a CD32 ^bright population expressed in CD19⁺ B cells and T-B cell doublets or conjugates (Figure S5A). We next investigated whether T-B cell trogocytosis was involved as previously reported (Martin et al., 2018; Osuna et al., 2018; Thornhill et al., 2019) by using image cytometry on CD32⁺ CD4⁺ CD45RA⁻ memory sorted T cells of humanized mice and of HIV-1-infected patients. CD32⁺CD4⁺ memory T cells contained no doublets, T-B cell conjugates, or B cell/monocyte contaminations (manual viewing of sorted cell images, data not shown). CD32 expression was localized in small, concentrated areas of CD4⁺ T cells (Figures 4A and 4C) in contrast to sorted B cells on which CD32 was distributed homogeneously over the cell membrane (Figures 4B and 4D). Of note, no signal of CD19 or CD20 expression was detected on CD32⁺CD4⁺ T cells (Figures 4A and 4C). Importantly, overlay images emphasize that CD32 expression is rooted in small, non-CD4⁺ T-cell compartments attached to CD4⁺ T cells (Figures 4E and 4F). To explore if other cell types known to closely interact with CD4+ T cells might be the source of these CD32 expressing fragments attached to the CD4+ T cells, we used flow cytometry in peripheral blood mononuclear cells (PBMCs) of two HIV-1-infected patients treated with cART for at least two years and we did not exclude B or CD14+ cells as previously done. We detected different lineage markers coexpressed on CD32+ CD4+ T cells that could explain the origin of the attached fragments (potential trogocytosis): between 30 and 39% of HLADR (Human Leukocyte Antigen – DR isotype) + CD19+ B cells, around 1% of NK cells (CD4+CD16 + HLADR+), 4% of monocytes (all CD14+ cells), and up to 1% of dendritic cells (CD11c+). Around 7% of CD32+ CD4+ T cells were expressing HLA-DR, and this expression was not provided by CD19+ B cells, CD11c+, or CD16+ cells. Therefore, as we excluded B cells and monocytes in our strict flow cytometry strategy in this study, the fragments attached to the CD4+ T cells might come from other antigen-presenting cells (APCs) or NK cells. Furthermore, we measured the mRNA expression of CD32a and CD32b by real-time PCR in sorted CD32⁺CD4⁺CD45RA⁻T cells and in a pool of CD14⁺ and CD19⁺ cells that were excluded by our flow cytometry gating strategy. CD32⁺ CD4⁺ CD45RA⁻ memory T cells expressed mainly CD32a and low levels of CD32b mRNA, whereas both CD32a and CD32b mRNA were expressed in the pool of CD14⁺ and CD19⁺ cells (Table 1), confirming the exclusion of B cells in our purified CD32⁺CD4⁺ memory T cells.

Figure 4.

CD32 Is Expressed on Cell Fragments Attached to CD4⁺ T Cells with No Signs of T-B Cell Trogocytosis

To assess whether the expression of CD32 on CD4⁺ T cell comes from T cells or T-B cell conjugates, imaging flow cytometry was performed on sorted CD32⁺CD4⁺ T cells (A, C, E, and F) or sorted CD19⁺ B cells (B and D) isolated from the spleen of cART suppressed HIV-1-infected humanized mice (A, B, E) or peripheral blood mononuclear cells (PBMCs) from cART suppressed HIV-1-infected patients (C, D, and F). Flow sorted B cells from humanized mice or patients were stained, acquired, and represented with the same settings and experimental procedures and,hence, serve as reference population with regard to CD32, CD19 ,and CD20 signal on images (B, D). Overlay shots of CD4 and CD32 signals of splenocytes from a humanized mouse (E) and of PBMCs from one HIV-1-infected patient (F).

Table 1.

Gene expression of CD32a and CD32b mRNA isoforms of sorted CD45RA⁻CD32⁻CD4⁺ T cells and a pool of CD19⁺ and CD14⁺ cells isolated from 3 single virally suppressed humanized mice (S789, S803, S804) and sorted CD45RA⁺CD32⁺CD4⁺ T cells pooled from the 3 virally suppressed humanized mice

Mouse ID	BM	Spleen	Sorted cell fraction	Ct values			ΔCT value against GAPDH
				CD32a	CD32b	GAPDH	CD32a	CD32b
S789	X		Pool of CD19+ and CD14+ cells	32,36	31,92	24,65	7,71	7,28
S803	X		Pool of CD19+ and CD14+ cells	28,86	30,4	21,76	7,1	8,63
S804	X		Pool of CD19+ and CD14+ cells	34,45	33,7	25,18	9,27	8,51
S789		x	Pool of CD19+ and CD14+ cells	34,51	32,91	27,92	9,59	4,98
S803		x	Pool of CD19+ and CD14+ cells	28,99	30,04	25,25	3,73	4,79
S804		x	Pool of CD19+ and CD14+ cells	32,35	30,9	26	6,35	4,9
S789	x		CD32-CD4+ memory T cells	36,58	37,14	25,07	11,51	12,7
S803	x		CD32-CD4+ memory T cells	36,67	36,09	25,36	11,3	10,73
S804	x		CD32-CD4+ memory T cells	36,23	37,61	25,97	10,26	11,64
S789		x	CD32-CD4+ memory T cells	39,48	38,31	27,73	11,75	10,58
S803		x	CD32-CD4+ memory T cells	36,75	36,79	25,37	11,38	11,41
S804		x	CD32-CD4+ memory T cells	36,67	36,69	24,81	11,86	11,88
S789, S803, and S804 pooled	x		CD32 + CD4+ memory T cells	32,14	36,72	26,69	5,45	10,02
S789, S803, and S804 pooled		x	CD32 + CD4+ memory T cells	34,21	38,82	28,74	5,47	10,08
PCR-negative controls				UND	UND	UND	NA	NA

Results were analyzed against the GADPH Ct value. UND, undetectable; N/A, not available.

When assessing the expression pattern of the memory CD32⁺CD4⁺ T cells in the different groups of mice, we found that overall CD32 cell surface expression did not differ substantially across various tissues nor between treatment groups (Figures S6A–S6D). Only lymph node T cells from healthy mice had significantly lower CD32 expression when compared to the untreated group (Figures S6 and 6D). Importantly, we did not find any difference during longitudinal sampling in the spleen (Figure S5B). To determine whether CD32 expressing cells were enriched for total CA HIV-1 DNA or RNA in cART-suppressed mice, CD32^- and CD32⁺ were flow sorted from the CD4⁺ memory (i.e. CD45-) T cells. A sufficient quantity of CD32⁺CD4⁺ memory T cells were only obtained from the bone marrow and spleen. Memory CD32⁺CD4⁺ T cells harbored significantly higher levels of HIV-1 DNA as compared to their CD32 negative counterparts (Figures 5A and 5B) in the spleen (4.35 vs 3.51 median log₁₀ copies of HIV-1 DNA per million cells, respectively, p < 0.01) and in the bone marrow (4.42 vs 3.55 median log₁₀ HIV-1 DNA copies per million cells, respectively, p < 0.05). Next, we quantified CA HIV-1 RNA but only 14 spleen-derived samples could be included due to low total RNA yield from CD32⁺CD4⁺ T cells. Although we used a limited set of sampling, we observed no difference for CA HIV-1 RNA between the CD32⁺CD4⁺ and the CD32⁻CD4⁺ memory subsets of the spleen (Figure 5C) (11.52 versus 45.16 median copies of CA RNA per million cells) for all mice with a detectable viral load.

Figure 5.

Total HIV-1 DNA Is Higher in CD32⁺CD4⁺ Memory T Cells in the Spleen and the Bone Marrow

Total HIV-1 DNA from flow sorted CD3⁺CD4⁺CD45RA⁻ T cells according to CD32 expression in bone marrow (A) and spleen (B) in virally suppressed mice (mixed from early and late treatment conditions) (bone marrow n = 11, spleen n = 10) depicted as aligned dot plots of individual measurements (CD32- in black circles and CD32 + in red squares). The flow gating strategy is depicted in Figure S3. (C) HIV-1 CA RNA in flow sorted CD4⁺ memory T cells (CD32^- (n = 9) CD32⁺ (n = 5)) from virally suppressed mice (mixed from early and late treatment conditions). Statistical testing was done using non-parametric paired Wilcoxon rank tests (A and B) or non-paired tests (C). Exact p values for HIV-1 DNA: bone marrow, 0.0278; spleen, 0.0039, (∗p < 0.05. ∗∗p < 0.01).

CD32 is expressed on highly activated memory CD4⁺ T cells and its level of expression does not correlate with the HIV-1 total DNA

Next, we assessed whether CD32⁺CD4⁺ T cells from tissue lymphocytes showed a particular activation/exhaustion profile as recently suggested (Badia et al., 2018). We found a significant upregulation of HLA-DR and PD-1 expression but not of the immune checkpoint marker TIGIT in CD32⁺CD4⁺ memory T cells (Figures 6A–6C), as compared to their CD32^- counterparts. This trend was observed in both early (p < 0.05) and late treated mice (p < 0.001). We also measured the expression of the early and late activation markers CD69, HLA-DR, and CD38 (all combined in one channel) of the flow sorted CD32⁺ and CD32^-CD4⁺ memory T cells. A median of 80% CD32⁺CD4⁺ memory T cells expressed at least one of these activation markers as compared to a median of only 7% CD32⁻CD4⁺ memory T cells (Figure 6D). Moreover, the geometric mean fluorescence intensity (GMFI) of CD32 correlated moderately with the cumulated GMFI of the three activation markers (Spearman correlation, p value = 0.0004, r = 0.6872) (Figure 6E). In contrast, there was no correlation between CD32 expression and the total HIV-1 DNA content in CD4⁺ T cells (data not shown, p value = 0.5809, r = −0.2167).

Figure 6.

CD32 Expression Was Associated with High Expression Levels of Activation/Exhaustion Markers and Did Not Correlate with HIV-1 DNA

(A–C) Expression of HLA-DR (A), PD-1 (B), and TIGIT (C) in CD32-positive and negative CD4⁺ memory T cells from the spleen.

(D) Expression level of three combined activation markers (HLA-DR, CD38, and CD69) in flow sorted CD4⁺ memory T cells according to their CD32 expression. Geometric mean fluorescence intensity (GMFI) of activation markers plotted against the GMFI of CD32 from virally suppressed mice (mixed early and late treatment conditions).

(E) Spearman correlation with fitted linear regression line (p value: 0.0004, R value: 0.6872). Statistical testing by non-parametric Mann-Whitney or Kruskal-Wallis tests with multiple comparison Dunn's correction. Two-tailed spearman correlation (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.005, ∗∗∗∗p < 0.001, ns, nonsignificant test, p > 0.05).

We next performed multidimensional single-cell analysis using flow cytometry to investigate the activation/exhaustion co-expression patterns of the CD32⁺CD4⁺ memory T cells. CD32⁺CD4⁺ memory T cells were analyzed in untreated viremic mice (untreated [UT] mice, after four and eight weeks of HIV-1 infection) and in virally suppressed treated mice after four and eight weeks of cART treatment for both early (early treated [ET] mice) and late treated (late treated [LT] mice) groups. Co-expression of the markers was analyzed using the unsupervised algorithmic technique Flow Self-Organizing Map (FlowSOM). The clusters defined by FlowSOM (Figure 7A) were organized in a two-way hierarchical heatmap to determine their relationship (Figure 7B). Statistical analysis of the differential enrichment of clusters at week 4 and 8 revealed major changes in mice undergoing late treatment. Six clusters belonging to the central effector memory CD4 metacluster (metacluster number 8) showed an increase at week 8. The majority of these clusters expressed HLA-DR and in one we observed the expression of CD32 (cluster 6). Overall, clustering identified three CD4⁺ T-cell populations harboring CD32 expression that showed a significant increase at week 8 in early treated mice (ET clusters 3, 11, 21) and one CD4⁺CD32⁺ cluster demonstrating a similar increase in late treated mice (LT clusters 6). Cluster 9 and 47 in LT mice were the unique CD4+CD32+ clusters showing a decrease at week 9. No significant changes were observed in UT mice. These memory CD4⁺ clusters expressed either CD32^dim alone (ET cluster 21) or CD32^dim HLA-DR^high expression (ET cluster 3) or CD32^dim HLA-DR^high and PD-1^low expression (LT cluster 6) or CD32^dim HLA-DR^dim TIGIT^high PD-1^dim expression (ET cluster 11) or CD32^high HLA-DR^dim TIGIT^dim PD-1^high expression (LT cluster 9) (Figure 7B). Only one naive CD8⁺CD32⁺ cluster harboring HLA-DR^high expression was detected in CD8⁺ T cells in late treated mice (LT cluster 47). These data showed the presence of phenotypically related activated or exhausted subsets in the memory CD4⁺ CD32⁺ T cells in the cART-treated mice when compared to their untreated counterpart. Overall, the expression of CD32 in these clusters was always associated with another activation (HLA-DR) or exhaustion marker (PD1, TIGIT), demonstrating that CD32 labeled a highly activated part of the memory CD4⁺ T cells.

Figure 7.

High-Dimensional Analyses Highlighted Different Activated/Exhausted Clusters Expressing CD32 in Memory CD4 T Cells from the Spleen of Early and Late Treated Mice with cART

(A) FlowSOM hierarchical tree colored according to metaclusters.

(B) Metaclusters were annotated according to the heatmap shown in. Two-way hierarchical clustering of the 49 clusters defined by FlowSOM and the median expression of the 9 markers indicated on the right of the heatmap. Median expression across the three groups of mice is shown. Three viremic untreated mice (untreated mice, UT)) were kept four weeks without treatment and four viremic mice were kept eight weeks; twenty-two mice were virally suppressed by receiving early cART treatment (n = 12, ET mice) or receiving late cART treatment (n = 10, LT). The clusters significantly different (p < 0.05) between four weeks and eight weeks of treatment are depicted on the bottom of the heatmap. A significant increase or decrease is indicated in red or cyan, respectively.

Tissue-resident memory CD32⁻CD4⁺ T cells contain most of the translation-competent virus producing p24

Finally, we sought to answer whether CD32⁺CD4⁺memory T cells contained the inducible translation-competent HIV-1 by assessing p24 production after phorbol myristate acetate (PMA)/ionomycin activation in the presence of raltegravir and lamivudine to prevent viral spread. Interestingly, more than 95% of p24 producing cells were detected in the CD32-negative fraction from the spleen in both viremic and virally suppressed mice (Figure 8B). Nevertheless, CD32⁺CD4⁺ memory T cells harbored a higher frequency of p24-positive cells when compared to CD32⁻CD4⁺ memory T cells suggesting a relative enrichment, as reported for total HIV-1 DNA (Figure 8C). Furthermore, we found higher levels of activation/exhaustion markers on CD32⁺CD4⁺ memory T cells producing p24 than their CD32^- counterpart, but this was significant only for PD1 (Figure 8D).

Figure 8.

P24 Production after Latency Reversal Is Predominantly Found in CD32⁻CD4⁺ Memory T Cells and Not in the CD32⁺ Fraction

(A) Representative fluorescence activated cell sorting (FACS) plots of p24 staining in an untreated viremic mouse after five weeks of HIV-1 infection and one ART-treated mouse, as well as a non-infected control mouse after PMA/ionomycin reactivation. The strict gating assured no false positive signal.

(B) Intracellular p24 staining after PMA/ionomycin activation in ART-treated animals (n = 4) revealed that the highest proportion of productive cells is CD32^- and not CD32 ⁺.

(C) The frequency of p24 producing cells is higher within CD32⁺CD4⁺ memory T cells than in CD32^- cells.

(D) Expression of HLA-DR, CD38, and PD-1 at the surface of the reactivated cells producing p24 from graph B. Statistics were performed using non-parametric testing. (∗p < 0.05, ns, nonsignificant test, p > 0.05).

Discussion

HIV-1 persists in a pool of latently infected CD4+ T cells, which constitutes a major barrier to curing HIV-1 (Katlama et al., 2013). Early studies found that this reservoir is mainly localized in CD4⁺ resting T-cell memory subsets (Chomont et al., 2009) but recent evidence has substantiated on its wide spread distribution over various compartments of CD4⁺ T cells (Buzon et al., 2014; Khoury et al., 2016; Hiener et al., 2017; Zerbato et al., 2019; Kwon et al., 2020). A number of strategies aiming at reducing the size of the HIV-1 reservoir are currently under investigation. Therefore, the discovery of a specific latently infected cell's marker could facilitate tailored interventions. CD32 was proposed as such by Descours et al. (Descours et al., 2017) but has been heavily debated ever since (Bruel and Schwartz, 2018; Darcis et al., 2019). Here, we scrutinized the relationship between the timing of ART initiation, T-cell immune activation/exhaustion, and reservoir characteristics with regard to CD32 expression in tissues from humanized mice in order to delineate its use as a surrogate marker of the viral reservoir.

Several groups have reported on humanized mice models of HIV-1 latency and their utility (Nischang et al., 2012; Halper-Stromberg et al., 2014; Iordanskiy et al., 2015; Satheesan et al., 2018; Llewellyn et al., 2019). We validated our humanized mice model of HIV-1 latency by first showing that the time to viral suppression was shortened by early cART treatment as compared to late therapy which is similar to clinical observation in humans (Ssebunya et al., 2017). Multiple studies have reported the benefits of early cART on immune activation. Persistent chronic infections trigger increased expression of immune checkpoints such as PD-1 and TIGIT on CD4⁺ and CD8⁺ T cells (Day et al., 2006; Trautmann et al., 2006), whereas cART decreases PD-1 expression in parallel with reducing plasma VL after treatment onset (Cockerham et al., 2014b). In our study, we observed a consistent decline of PD-1 and TIGIT expression that was only significant by early cART initiation when compared to untreated mice. The T-cell activation marker HLA-DR was not significantly decreased by early or late cART regimen. Nevertheless, the recovery of total CD4⁺ T-cell counts was rapid and successful in both treatment groups. These findings highlight that key aspects of HIV-1 pathogenesis and cART treatment are recapitulated in the presented model.

It is accepted that early cART effectively reduces the size of viral reservoirs and preserves the functionality of CD8 T cells, two conditions that might allow some individuals to achieve functional HIV-1 cure (Conway and Perelson, 2015). Our findings underline the impact of treatment initiation's timing on the level of total HIV-1 DNA spread in tissues as described in humans (Ananworanich et al., 2012; Laanani et al., 2015). Conversely, time to viral rebound did not differ between early and late treated conditions, similarly reported by two clinical studies that found also no prolonged control upon treatment cessation despite early ART initiation (Gianella et al., 2015; Colby et al., 2018). Consequently, it could be argued that total HIV-1 DNA, when sampled in tissues, is not a surrogate marker for time to viral rebound in contrast to human studies from peripheral blood (Azzoni et al., 2013; Williams et al., 2014). Importantly, we observed that a portion of early treated mice rebounded only transiently (26,67%) with subsequent “spontaneous” viral suppression, whereas this “secondary control” phenomenon was not observed in the late treated group. This suggests that early treatment increases the odds for delayed rebound, similar to a conclusion reached by the CHAMP cohort (Namazi et al., 2018) that found higher percentage of post-treatment control (PTC) in early treated individuals. The antiviral CD8⁺ T-cell response could play a role during PTC, but this was not explored in greater detail because NSG mice lack antigen-specific T cells due to the absence of HLA-I presentation (Policicchio et al., 2016). Clearly, future studies should investigate the mechanisms of delayed rebound in this model. Taken together, our data showed however that the viral reservoir is already established and functional within one week after infection in most of the mice.

A number of cellular HIV-1 reservoir markers have been proposed over the last years. Among those markers the activation and immune checkpoint molecules HLA-DR, PD-1, TIGIT, but also CD32a (Cockerham et al., 2014a; Fromentin et al., 2016; Descours et al., 2017; Gosselin et al., 2017; Hogan et al., 2018). According to Descours et al., CD32a exhibited the promising trait to contain the replication-competent reservoir and to enrich for about 10³ fold enrichment for total HIV-1 DNA (Descours et al., 2017). These results have been challenged by others, reporting either no (Abdel-Mohsen et al., 2018; Martin et al., 2018) or only marginal enrichment (Bertagnolli et al., 2018; Pérez et al., 2018), for both HIV-1 DNA and replication-competent proviruses. Such divergent findings were proposed to be due to technical tribulations during purification of CD32⁺CD4⁺ T cells (Darcis et al., 2019), as several studies showed that CD32⁺CD4⁺ T cells contained residual contamination by APCs which express high levels of CD32 (Martin et al., 2018; Osuna et al., 2018; Pérez et al., 2018; Thornhill et al., 2019). Therefore, we established a flow cytometry strategy with strict exclusion of B cells and monocytes (CD19 and CD14, respectively) as both cell types had been identified as sources for non-T-cell contamination before (Martin et al., 2018; Osuna et al., 2018). This approach enabled the identification of a CD32^dim and not the CD32^high population in a small fraction of CD4⁺ T cells. Using imaging cytometry, in both humanized mice and in HIV-1-infected patients, we found that purified CD32⁺CD4⁺ T cells expressed CD32 on small, concentrated areas as previously reported by Osuna et al. (Osuna et al., 2018) but without co-localization of monocytes or B cell markers (CD19 or CD20), thus excluding B-cell trogocytosis. Still, we detected a polarized and concentrated localization of the CD32 protein on the membrane of CD4⁺ T cells that could come from other APCs or NK cells. We detected CD32a mRNA expression on CD4⁺ memory T cells and a weak expression of CD32b mRNA as well, whereas a pool of CD19⁺ and CD14⁺ population of B cells and monocytes expressed both CD32a and CD32b mRNA. This suggests endogeneous expression of CD32a by CD4⁺ T cells similar to previous findings (Holgado et al., 2018; Darcis et al., 2020). However, the latter results have to be interpreted with caution as imaging cytometry data of ours and others (Osuna et al., 2018; Thornhill et al., 2019) unanimously challenge the origin of CD32 protein on CD4⁺ T cells. We assume that CD32b mRNA might originate from the attached CD32⁺ non-CD4⁺ T-cell entities or from trogocytosis-positive T cells expressing higher level of activation markers, such as CD32a, due to more intensive T-cell/APC interaction. Taken together, we clearly identified a non-monocyte or non-B-cell-induced portion, potentially due to trogocytosis from other APC types or NK cells of CD32⁺CD4⁺ T cells or even from residual binding with platelets known to express CD32a. These findings are in line with a recent study working on highly pure CD32⁺CD4⁺ T cells (Darcis et al., 2020) and might advance the debate opened by many reports on the source for the CD32 signal on CD4⁺ T cells (Martin et al., 2018; Osuna et al., 2018; Pérez et al., 2018; Thornhill et al., 2019). Ultimately, our results open new questions about the exact origin of the CD32 protein on CD4⁺ T cells.

In our study, we further focused on CD32 expression in memory CD4⁺ T cells (CD45RA⁻) as this subset is comprises central and effector memory T cells which are well known to contain the highest frequencies of latently infected cells (Chomont et al., 2009; Bosque et al., 2011; Imamichi et al., 2014; Soriano-Sarabia et al., 2014; Murray et al., 2016). As compared to other studies, we included both resting (HLA-DR- or CD38- or CD69-) and non-resting (HLA-DR+ or CD38+ or CD69+) memory T cells. We found a significant increase of CD32 expression between HIV-1-infected and uninfected mice in human lymph nodes only as previously described (Bertagnolli et al., 2018). Importantly, a significant increase of about 0.85 log₁₀ copies of CA HIV-1 DNA per millions cells in CD32⁺ when compared to CD32^- memory T cells was observed in virally suppressed mice. These findings are in agreement with results obtained from human lymph nodes (Noto et al., 2018) but not with data derived from peripheral blood of HIV-1-infected patients (Abdel-Mohsen et al., 2018; Badia et al., 2018; Osuna et al., 2018; Pérez et al., 2018) unless the CD32⁺ cells underwent multiple, consecutive rounds of purification to exclude B cells, monocytes, and T-B cell doublets (Darcis et al., 2020). Despite CA HIV-1 DNA enrichment, there was no difference in CA HIV-1 RNA between CD32⁺ and CD32⁻CD4⁺ memory T cells from the spleen, suggesting that CD32 labels cells with proviruses that are transcriptionally silent in tissues. However, the sample size of our data is limited because the total RNA yield from sorted subsets was low, and this represents a limitation of our study. Several studies from the lymph node and gut tissue reported the contrary (Abdel-Mohsen et al., 2018; Badia et al., 2018; Noto et al., 2018; Vasquez et al., 2019), but these findings are in accordance with a recent study from highly purified blood-derived lymphocytes (Darcis et al., 2020). Low transcription levels could be counterintuitive because we confirmed here, as others, that CD32⁺CD4⁺ T cells express HLA-DR activation marker. Nevertheless, activated subsets can indeed establish latent HIV-1 infection (Chavez, 2015; Lee et al., 2019).

CD32 is a low affinity receptor for the Fc portion of IgG (FcγR) which represents a link between humoral and cellular immunity and is highly expressed on myeloid cells and B cells. The functional role of CD32 on CD4⁺ T cells is not well understood yet. However, it has been suggested that immune complexes might contribute to the continuation of chronic inflammatory responses by enhancing the activation of T cells through CD32 (Holgado et al., 2018) that might maintain persistence of the latent HIV-1 reservoir. Importantly, the expression of HLA-DR, CD38, and CD69 was strongly increased in CD4⁺CD32⁺ memory T cells from the tissues of cART-treated mice similar to recent studies from peripheral blood lymphocytes in humans (Badia et al., 2018; Wittner et al., 2018). Therefore, CD32⁺ CD4⁺ T cells might be more susceptible to HIV-1 infection because they are activated in a first place. The increased levels of HIV-1 co-receptors CXCR4 and CCR5 on CD4⁺CD32⁺ T cells on healthy individuals argue in favor of this concept (Hogan et al., 2018; Wittner et al., 2018). In addition, a recent study described tissue resident memory T cells from the cervical mucosa expressing CD32 as enriched for several proteins associated to HIV-1 susceptibility (Cantero-Pérez et al., 2019). CD32⁺CD4⁺ T cells could represent a pool of recently infected cells as drug penetration in tissues can be insufficient to suppress viral replication (Lorenzo-Redondo et al., 2016; Estes et al., 2017; Rothenberger et al., 2019). Accordingly, CD32 expression is induced upon infection with HIV-1 in vitro (Descours et al., 2017; Abdel-Mohsen et al., 2018; Badia et al., 2018). Furthermore, in human lymphoid and gut tissues, CD32 mRNA expression was associated with persistent HIV-1 mRNA replication (Hogan et al., 2018; Noto et al., 2018) which could indeed be explained by poor cART penetration unable to block residual HIV-1 replication. A relationship between active HIV-1 transcription and the CD32 gene has been already suggested (Abdel-Mohsen et al., 2018; Vasquez et al., 2019). Whether up-regulation of CD32 mRNA is a direct result of HIV-1 transcription or T-cell activation is still unknown. In this regard, our multidimensional reduction analysis revealed that CD32⁺CD4⁺ memory T cells have a high activation/exhaustion profile and clustered phenotypically with distinct activated/exhausted memory T-cell subsets expressing HLA-DR, PD-1, or TIGIT during suppressed viremia. This is in agreement with Badia et al. (2018) who has shown, following ex vivo HIV-1 infection in activated and resting cells or in PBMCs from HIV-1+ individuals, that CD32 expression is a marker of activation in a subset of CD4+ T cells both in uninfected controls and in HIV-1+ individuals. These authors state that establishment of HIV-1 latency may be the consequence of infection in CD4+T cells within a narrow window of time after activation. Thus, CD32 expression may signal a transition state to or from a fully susceptible phenotype. Our data provide further evidence that CD32 labels a highly activated/exhausted subset of memory CD4⁺ T cells that might favor subsequently an HIV-1 enrichment in these cells.

Indispensable for the characterization of CD4⁺CD32⁺ T cells as latent HIV-1 reservoirs is to determine if they contain replication-competent virus and can resume to viral production. The current gold standard, the quantitative viral outgrowth assay (qVOA), requires large amounts of cells. Given the rarity of CD32⁺ CD4⁺ T cells in mice, we used intracellular p24 staining after in vitro latency reversal as reported previously. We found a relative enrichment of p24 production in CD4⁺CD32⁺ memory T cells similar in magnitude (about ten times) than the increase in CA HIV-1 DNA. However, overall CD32⁺CD4⁺ T cells constituted only a small portion of the total p24 inducible cellular reservoir (<5% of total p24 producing cells). Similar findings were reported in human lymph nodes (Abdel-Mohsen et al., 2018). Clearly, the presence of p24 alone, although produced in an advanced stage of the HIV-1 replication cycle, does not prove that the resulting viruses are correctly assembled, released, and infectious. Furthermore, a single round of stimulation with PMA/ionomycin for 24 hr only does not suffice to induce all replication-competent viruses (Ho et al., 2013; Laird et al., 2013). Unfortunately, there is currently no assay that can accurately activate the entire replication-competent fraction. Near full-length DNA sequencing does not take inducibility of viral gene transcription into account (Josefsson et al., 2013), and qVOA largely underestimates the replication-competent fraction (Ho et al., 2013). Taken together, our data underline that CD4⁺CD32⁺ memory T cells are enriched for p24 production after latency reversal but represent a small and negligible portion of the total inducible p24 producing cells.

Finally, the clinical relevance of the memory CD32⁺ CD4⁺ T cells as long-term viral reservoir remains difficult to estimate at this point in time. We observed a correlation of CD32 expression with the combined expression of all activation markers. Highly activated T cells are generally regarded as rather short lived (Chun et al., 2005; Noto et al., 2018) but their contribution to HIV-1 persistence is widely accepted (Hatano et al., 2013; Chavez, 2015; Deeks et al., 2016; Khoury et al., 2016; Lee et al., 2019). Consequently, for HIV-1 infection to persist, activated reservoirs are required for continous replenishment either by ongoing viral replication (Martinez-Picado and Deeks, 2016) or by cellular proliferation (Chomont et al., 2009). Since CD32 was shown to promote immune activation in CD4⁺ T cells in vitro (Holgado et al., 2018), it is tempting to speculate that CD32 itself could induce and maintain activation leading to preferential vulnerability to infection. The results of the cluster analysis reinforces this hypothesis by showing phenotypic similarities and a high number of clusters in the memory CD32⁺ CD4⁺ T cells as compared to the CD32⁺ CD8⁺ T cells. This in turn would question its role as a long-lived reservoir. Finally, there was a significant upregulation of PD-1 in CD32⁺ vs CD32^- cells producing the inducible translation-competent virus after PMA stimulation. Interestingly, Noto et al. have associated CD32⁺PD1⁺CD4⁺ T cells from lymph nodes with the highest levels of HIV-1 transcription (Noto et al., 2018).

In conclusion, the present work highlights that tissue-resident CD32⁺CD4⁺ memory T cells represent different cellular subsets that are highly activated, enriched for total HIV-1 DNA, and produce P24 upon latency reversal. Despite a relative enrichment, CD32⁺CD4⁺ T cells form an overall negligible portion of the inducible P24 producing cells. Our findings support that CD32 expression is linked to immune activation in tissues which may explain the enrichment of CA HIV-1 DNA in CD32⁺CD4⁺ memory T cells. More distinctive markers to target the silent and replication-competent HIV-1 reservoir are still needed.

Limitations of the study

We acknowledge several limitations to our study. The limited life span of humanized mice allowed cART therapy for a maximum duration of eight weeks in our study design. This duration clearly differs from patient cohorts investigating HIV-1 latency after several years of suppressed viremia. Therefore, we cannot exclude the impact of short-lived reservoirs with decay rates above eight weeks (Palmer et al., 2011) and that the reservoir dynamics in CD32 + T cells derived from our latency model could be different than in patients with a longer treatment duration. This could potentially explain the lower relative enrichment in our humanized mice as compared to human PBMCs (Darcis et al., 2020). In addition, the quantity of blood available for longitudinal sampling and of human cells in tissues is also extremely limited in humanized mice. Lastly, PMA/ionomycin's activation used here to measure P24 producing cells might be able to induce only a fraction of the cells harboring translation-competent viruses in the humanized mice. A significant fraction of p24 producing cells does not also produce intact viral particles that are replication competent. Therefore, we induce probably only one part of the translation-competent virus in our study. However, the HIV-1 flow technique, by using 2 different P24 antibodies, provides a specific staining of the P24+ cells and allows us to compare accurately between CD32+ CD4+ memory T cells and their CD32-CD4+ counterpart, the translation-competent virus induced by PMA/ionomycin.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the Lead Contact, Carole Devaux (caroledevaux@lih.lu).

Materials availability

The study did not generate any unique reagents.

Data and code availability

This published article includes all data sets generated or analyzed during this study.

Methods

All methods can be found in the accompanying Transparent methods supplemental file.

Published: January 22, 2021