Transcription of HIV-1 is heterogenous among authentic latent CD4+ T cell clones

Abstract

Antiretroviral therapy suppresses HIV-1 infection but fails to eliminate a reservoir of intact latent proviruses that reside primarily in CD4+ T cells. The lack of precise understanding of the latent compartment has made it challenging to develop curative strategies for HIV-1 infection. Here we report on the properties of CD4+ T cell clones carrying intact latent proviruses, expanded in vitro from single cells obtained from the reservoir of people living with HIV-1. The latent proviruses in the clones were integrated into ZNF genes, non-genic satellite and centromeric regions, frequently associated with latency. Despite their descent from single cells, only a fraction of the cells (0.4–14%) expressed relatively low levels of HIV-1 that did not measurably alter host gene transcriptome. Latency reversing agents (LRAs) variably increased expression, but the effects were modest and clone and LRA specific. The results suggest that pharmacologic and immunologic approaches to clear the reservoir should be optimized to accommodate intra- and inter-clonal diversity.

Introduction

HIV-1 is an RNA virus whose life cycle requires reverse transcription and integration into the host genome. High level viral transcription is associated with cell death (Ruelas and Greene, 2013), and immunopathology. Anti-retroviral therapy (ART) prevents the spread of infection and suppresses viremia in people living with HIV-1 (PWH) but fails to cure the disease because intact proviruses persist in the genome of rare CD4⁺ T cells (Chun et al., 1998a; Finzi et al., 1997; Wong et al., 1997). This proviral reservoir is long-lived and responsible for rapid rebound viremia in individuals that interrupt therapy and it represents the primary barrier to HIV-1 cure (Cohn et al., 2020; Davey et al., 1999; Margolis and Archin, 2017; Siliciano and Siliciano, 2022).

Understanding the nature of the reservoir has been challenging because cells carrying intact latent proviruses are rare and there are no definitive markers that can be used to isolate them (Cohn et al., 2018; Collora et al., 2022; Crooks et al., 2015; Ho et al., 2013; Sun et al., 2023; Wei et al., 2023; Weymar et al., 2022; Wong et al., 2023; Wu et al., 2023). Moreover, defective proviruses are far more abundant than intact making it difficult to interpret simple proviral DNA measurements (Cho et al., 2022; Peluso et al., 2020). Quantitative estimates of the reservoir were initially made using in vitro outgrowth assays that depend on the ability of a given stimulus to induce proviral expression (Chun et al., 1998b; Finzi et al., 1997). These measurements were later shown to be underestimates because only a fraction of the latent cells were induced to produce virus after one round of stimulation (Ho et al., 2013; Hosmane et al., 2017). Nevertheless, outgrowth assays established that the reservoir decays rapidly in the first year after ART, but the rate of decay decreases to an estimated half-life of 4–7 years which explains why this therapy fails to eliminate the infection (Bachmann et al., 2019; Cho et al., 2022; Crooks et al., 2015; McMyn et al., 2023; Peluso et al., 2020; Siliciano et al., 2003).

Newer direct nucleic acid-based assays revealed that the proviral reservoir is primarily found in expanded clones of CD4⁺ T cells that wax and wane over time (Bui et al., 2017b; Cohn et al., 2015; Lorenzi et al., 2016; Lu et al., 2018; Simonetti et al., 2016). Sequencing revealed that the intact reservoir decays at an accelerated rate compared to the defective suggesting that there is either direct counterselection against intact proviruses by viral cytopathic effects, or immune-mediated elimination (Cho et al., 2022; Peluso et al., 2020; Reeves et al., 2023). Consistent with this idea, qualitative measurements of proviral mRNA expression showed that some latent proviruses are transcriptionally active (Dube et al., 2023; Einkauf et al., 2022; Procopio et al., 2015; Wiegand et al., 2017). Moreover, the relative rate of proviral transcription appears to be dependent on the position of the provirus in the genome (Battivelli et al., 2018; Collora and Ho, 2023; Jordan et al., 2001; Teixeira et al., 2024). Finally, prolonged therapy and elite control enrich for proviruses integrated into transcriptionally silent regions emphasizing the importance of proviral transcription on selection (Einkauf et al., 2019; Einkauf et al., 2022; Huang et al., 2021; Jiang et al., 2020).

One approach to accelerate reservoir elimination is to pharmacologically increase proviral transcription, by latency reversing agents (LRAs) (Archin et al., 2012; Armani-Tourret et al., 2024; Olesen et al., 2015; Rasmussen et al., 2014; Sogaard et al., 2015) Human trials using these agents can produce measurable increases in HIV-1 in circulation and changes in the population of CD4⁺ T cells carrying integrated proviruses. However, overall changes to reservoir size have been difficult to document possibly because of limited exposure or response heterogeneity within and between expanded clones of infected CD4⁺ T cells (Gunst et al., 2020; Tanaka et al., 2022).

Combinations of outgrowth and sequencing assays suggested that provirus transcription within a clone of CD4⁺ T cells can be heterogeneous (Einkauf et al., 2022; Hosmane et al., 2017). However, precise understanding of how transcription varies among cells within clones of CD4⁺ T cells bearing intact proviruses is limited. To further understand this phenomenon, we developed methods to identify and expand CD4⁺ T cells harboring intact latent proviruses in tissue culture (Teixeira et al., 2024; Weymar et al., 2022). Preliminary results indicated that the transcriptional profiles of the cultured cells resembled their ex vivo counterparts and that they expressed relatively low levels of HIV-1. Here we report on 6 cell lines derived from primary CD4⁺ T cells harboring intact proviruses, their heterogeneous levels of proviral expression, the effects of LRAs on the provirus and the impact of the provirus on the host cell transcriptome.

Results

To obtain clones of CD4⁺ T cells harboring intact latent proviruses CD4⁺ CD45RA- T cells from PWH were enriched for the proviral clone of interest by cell sorting based on T cell receptor ß-chain (TCRb) expression (Weymar et al., 2022). The TCR associated with clones obtained from participants 5104, 5125, 603, 9247 and B207 were known. We determined the TCRb associated with clones from participants 5203 and P301 by flow cytometry using combinations of anti-Vb and Cb antibodies (Fig. S1A). Sorted cells were plated at a density of 5 cells/well in the presence of feeder cells and stimulated with anti-CD3 and -CD28 plus IL-2 in the presence of anti-retroviral drugs to prevent spread of infection (Fig. 1A). After 3 weeks, cultured cells reverted to the resting state and downregulated the CD25 activation marker (resting, Fig. S1B). Cultures were then screened by ddPCR and the presence of intact proviruses confirmed by sequencing. Positive cultures underwent a further 2–4 rounds of expansion. In total we obtained 6 expanded clones of CD4⁺ T cells harboring intact replication competent proviruses, and one carrying a defective provirus (P301^d), each from a different donor (Figs. 1A and B, Table S1). The fraction of HIV-infected cells in the final cultures was determined by ddPCR (Bruner et al., 2019) ranging from 16% to 100% (Fig. S1C).

Figure 1. HIV-1⁺ CD4⁺ T cells from PWH.

(A) Graphic summary of the enrichment, screening, culture and analysis of the HIV-1⁺ CD4⁺ T cell clones. (B) Provirus integration site for each clone. Arrows represent transcription orientation. Transcription from HIV-1 intact provirus is represented in red. The defective provirus is represented in black, with transcription arrows showing the location of the inversion. Transcription from host genes is represented in green. (C) Flow cytometry intracellular p24 detection in HIV-1⁺ CD4⁺ T cell clones in resting and activated conditions Data from 5104 and 603 were previously presented in Teixeira et al. (2024).

The proviruses in 603, B207 and 5203 are integrated in the introns of genic regions of ZNF genes which are preferred sites of intact proviral integration (Einkauf et al., 2019; Einkauf et al., 2022; Huang et al., 2021; Jiang et al., 2020). In contrast, 5104 (Huang et al., 2021), 5125 and 9247 are in non-genic centromeric or pericentromeric satellite regions (Fig. 1B, Table S1). P301^d is defective due to an inversion in its 5’ LTR, it is otherwise intact, and it too is integrated in a pericentromeric satellite region (Fig. 1B, Table S1).

Primary cells corresponding to 4 of the clones (B207, 603, 9247, 5104) carry infectious proviruses (Cohn et al., 2018; Huang et al., 2021; Mendoza et al., 2020). To determine whether the 5125, 5203 and P301^d were able to produce infectious virus we used the supernatants of resting and anti-CD3 and -CD28 plus IL-2 stimulated (activated) cultures to infect HIV-1 negative donor CD4⁺ T cells. Six days after infection supernatants were assayed for HIV-1-Gag production by ELISA. HIV-1 Gag p24 was detected in the supernatants of 5125 and 5203 but not P301^d, which is defective (Fig. S1D). Thus, all but the defective clone can produce HIV-1. However, p24 expression as measured by flow cytometry in resting, or activated HIV-1⁺ CD4⁺ cells was heterogenous (Fig. 1C).

To document HIV-1 transcription in the cultured cells we performed quantitative PCR (qPCR) for HIV-1 LTR, gag and env. We determined the number of mRNA copies using an HIV-1 standard and compared the amount of expression to cells productively infected with HIV-1_YU2. The percentage of HIV⁺ cells within the culture was taken into consideration for all calculations. The 6 clones that carry intact proviruses expressed different quantities of HIV-1, that were orders of magnitude lower than productively infected cells (Fig. 2A). LTR transcripts were barely detectable in 5203, 603 and B207 all of which are integrated into ZNF genes (Figs. 1B and 2A). Notably, higher levels of HIV-1 LTR expression were associated with intact proviruses integrated into pericentromeric satellite and centromeric regions (5104, 9247, 5125 Figs. 1B and 2A). Gag and env were proportional to LTR transcripts for all clones except P301^d that carries a defective provirus with an inverted 5’ LTR (Fig. 2A). Activation with anti-CD3 and -CD28 and IL-2 increased proviral expression in all clones carrying proviruses integrated into ZNF genes and 5125 which is integrated in a centromeric region (Fig. 2A, Table S1). In conclusion mRNA expression data suggested that expanded clones of CD4⁺ T cells derived from a single provirally infected cell express low levels of HIV-1 (Einkauf et al., 2022) (Fig. 2A).

Figure 2. HIV-1 provirus expression in resting and activated conditions.

(A) Histograms show average of HIV-1 transcripts per cell in resting (blue) and activated (red) conditions for the indicated clones and CD4⁺ T cells infected with HIV-1_YU2 for comparison (black). Copies of LTR, gag and env transcripts were determined based on a standard curve. Average copy number per cell was calculated by dividing the estimated number of transcripts by the number of HIV-1⁺ cells in the assay (corrected based on culture enrichment) (B) Graphic representation of the limiting dilution assay. (C) Frequency of cells expressing HIV-1 LTR, Gag and Env transcripts in resting (blue) and activated (red) conditions. The number of expressing cells was estimated by Poisson distribution, based on limiting dilution data, calculated using the Most Probable Number (MPN) method. (D) Average number of transcripts per expressing cell in resting (blue) and activated (red) conditions, estimated based on the frequency of cells expressing HIV-1 transcripts as shown in “C”. Average transcripts per cell of an active infection by HIV-1_YU2 is shown for comparison (black). All experiments were performed in triplicates. Frequency of cells expressing and copies of transcripts could not be calculated for 603 due to the low number of transcripts.

Flow cytometry and qPCR expression data suggest that expression of HIV-1 is heterogenous at the single cell level. To determine the fraction of cells within a clone that express proviral mRNA we performed limiting dilution experiments and measured HIV-1 transcripts by qPCR (Fig. 2B). Under resting conditions, the fraction of cells in a clone that expressed HIV-1 ranged from 4 to 140 per thousand (Fig. 2C). Activation with anti-CD3 and -CD28 increased the frequency of expressing cells in all cases tested except for P301^d, that harbors a defective provirus (Fig. 2C). Notably, for clone 9247, because the number of cells that expressed the provirus was low, the actual amount of HIV-1 mRNA per expressing cell could reach levels that were nearly comparable to that of a productive infection (Fig. 2D).

Proviral transcription can be modified by LRAs but little is known of their effects on single latent proviruses. To determine how proviral expression is modified by LRAs we exposed resting cells to these agents and measured HIV-1 LTR, gag and env expression in limiting dilution experiments by qPCR. Cultures were exposed to each of six different LRAs: Romidepsin, Panobinostat and Suberoylanilide hydroxamic acid (SAHA) which are histone deacetylase (HDAC) inhibitors; Prostratin, an NF-kB activator; JQ1, an inhibitor of bromodomain and extra-terminal motif (BET) proteins which competes with the transcriptional inhibitor P-TEFb (Rodari et al., 2021); and the SMAC mimetic AZD5582 (mSMAC) that activates a non-canonical NFkB pathway (Gavin C. Sampey, 2018). The effect of LRAs on HIV-1 gene expression was compared to resting and activated cells.

The response to the different LRAs varied and was clone specific with modest increases in the number of cells expressing HIV-1 (Fig. 3A, Fig. S2A). Preliminary analysis with cells in bulk revealed that none of the LRAs can activate transcription for clone 603. In no instance were all cells in a clone induced to express HIV-1. The greatest fraction of expressing cells, 27%, was achieved with Panobinostat in clone 5104 in which 14% of the cells expressed HIV-1 under resting conditions. However, Panobinostat was relatively ineffective in most of the other clones. Prostratin was unique in that it measurably increased the percentage of expressing cells in all clones tested but the increases were modest (Fig. 3A). Based on the frequency of cells expressing HIV-1, qPCR data was used to determine whether the LRAs increased the amount of expression per cell (Fig. 3B, Fig. S2B). For all clones, at least one of the LRAs increased the amount of proviral expression per cell over baseline (Fig. 3B, Fig. S2B). However, their effects varied among the clones with HDAC inhibitors, Romidepsin, Panobinostat and mSMAC being the most consistent activators among the LRAs. Supernatant from cells treated with mSMAC for 24h was used to infect CD4⁺ T cells from healthy donors and compared to infection with supernatant from the clones in a resting state. Virus production was measured 5 days post infection by Gag p24 Elisa (Fig. S2C). Despite the increase in HIV-1 transcription observed when cells were treated with mSMAC, p24 was only detected in cells infected with supernatants from clone 5104, and at lower levels than the resting control. In conclusion, the relative proportion of cells expressing HIV-1 and the amount expressed per cell in response to LRAs is heterogeneous and clone and stimulus specific.

Figure 3. HIV-1 provirus expression after LRA treatment.

(A) Frequency of cells expressing HIV-1 LTR, gag and env transcripts after treatment with LRAs estimated by poison distribution, based on limiting dilution data and calculated using the Most Probable Number (MPN) method. (B) Average copies of transcripts per expressing cell. The number of LTR, gag and env transcripts per reaction was determined based on a standard curve. Average copies of transcripts per expressing cell was estimated based on the frequency of cells expressing HIV-1 transcripts as shown in “A”. LRAs had no effect in HIV-1 provirus activation (data not shown) for clone 603. All experiments were performed in triplicates.

To examine the transcriptional profile of the cultured clones and how it compares to their primary ex vivo counterparts (Weymar et al., 2022) we performed single cell mRNA sequencing using the 10X Genomics platform on resting and activated cells. A total of 96,547 cells that carried an integrated provirus were analyzed (Fig. S3A and Table S2). TCR sequences were used to identify members of the CD4⁺ T cell clone of interest. Gene expression profiles were mapped onto primary cell data obtained from 5 of the 7 individuals from which clones were derived using a cut-off of 95% confidence (Fig. 4A). Although there was some heterogeneity in the primary cells and corresponding clones of CD4⁺ T cells, most of the clones mapped to similar or directly overlapping clusters (Fig. 4B). To determine which specific subset of T cells the clones were most like, we projected their gene expression profiles onto a reference multimodal single-cell data set of peripheral blood mononuclear cells (PBMCs) from HIV-1-negative individuals (Fig. S3B). Consistent with the primary ex vivo data the clones were heterogeneous despite originating from single cells (Weymar et al., 2022). Most of the cultured clones showed a T-central memory phenotype, but others were composed of mixtures of CD4⁺ cytotoxic T lymphocytes (CTL) and a variety of different T-central and T-effector memory phenotypes (Figs. S3C and D). Transcriptionally active and transcriptionally silent cells were similarly distributed among CD4⁺ T cell subtypes.

Figure 4. Uniform manifold approximation and projection (UMAP) of single-cell RNA sequencing data.

(A) Uniform manifold projections (UMAPs) of gene expression profiles mapped on primary cell data based on 141,646 resting ex vivo cells from Weymar et al. (2022) and 96,547 in vitro cells (45,297 resting and 51,250 activated) from this study. Cells with similar transcriptional profile were divided into 15 different clusters. The locations of cells from the HIV-1⁺ clone of interest are presented as red dot. (B) Bar graph representing the fraction of HIV-1⁺ cells that share gene expression profile with each of the 15 clusters from the UMAP. Ex vivo resting cells are presented in gray, in vitro resting cells in blue and in vitro activated cells in red.

To determine whether there are unique transcriptional features shared by CD4⁺ T cell clones that harbor latent proviruses we compared their transcriptomes to non-infected cells within the same cultures (Table S3). Each of the clones differed significantly from the uninfected co-cultured controls (Table S3). To determine whether the differentially expressed genes are shared by different clones we compared the 6 intact and one defective clone. We found 41 and 20 differentially upregulated genes that were shared among at least 4 clones in resting and activated conditions respectively (Fig. 5, Table S4). Among these 17/41 and 13/20 differentially upregulated genes from resting and activated cells respectively were previously identified as enriched among CD4⁺ T cells harboring intact latent proviruses (Figs. 5A and 5B). The differentially expressed genes included HLADR genes, CCL5 and genes expressed by cytotoxic T cells, consistent with previous reports (Cohn et al., 2018; Collora et al., 2022; Weymar et al., 2022). In conclusion the cultured clones resemble their primary ex vivo counterparts.

Figure 5. Overlapping DEGs between HIV-positive and HIV-negative CD4⁺ T cells.

Single-cell heatmaps showing mRNA expression of overlapping differentially expressed genes that are upregulated in HIV-positive cells compared to HIV-negative cells (rows), detected in a minimum of 10% of cells from at least four clones under resting (A) or activated (B) conditions. A subset of cells from each clone (columns) was selected for visualization.

Heterogeneous HIV-1 expression was observed by flow cytometry and qPCR and confirmed by single cell RNA sequencing which showed that despite their single cell origin only a fraction of cells in a clone express HIV-1 (Table S5). To determine whether HIV-1 expression by members of a clone alter cellular transcription we compared the transcriptome of individual cells within a clone that express the LTR and those that do not (Fig. 6A and Fig. S4). Notably, HIV-1 transcripts were the only differentially expressed mRNAs shared by all clones (Fig. 6B). We conclude that the HIV-1 expression in clones of CD4⁺ T cells harboring latent proviruses do not induce a common cellular transcriptional response.

Figure 6. Differential expression between HIV-infected cells positive and negative for viral transcripts.

(A) Volcano plots showing differentially expressed genes between HIV-expressing and non-expressing cells from clones 5104 and 5125 under resting or activated conditions. (B) Heatmap showing expression of overlapping differentially expressed genes between HIV-expressing and non-expressing cells under resting or activated conditions. A subset of cells from each clone (columns) was selected for visualization.

Discussion

Experiments in animal models, observational studies in humans that spontaneously control infection, and small interventional clinical studies indicate that the immune system can durably control infection. In HIV-1 controllers and PWH on long term therapy the proviruses that remain in the reservoir tend to be integrated into transcriptionally silent parts of the genome suggesting that strategies that increase proviral transcription and enhance immune responses would help eliminate the reservoir (Einkauf et al., 2019; Einkauf et al., 2022; Huang et al., 2021; Jiang et al., 2020; Lian et al., 2023). However, the limited clinical studies performed to date show little or no measurable effect of LRAs on the HIV-1 reservoir despite documented effects in proviral transcription (Debrabander et al., 2023). In this study we examined intact and defective proviral transcription by clones of CD4⁺ T cells derived from single cells from 7 different ART suppressed individuals. Our observation that baseline and LRA stimulated levels of proviral transcription vary within cells in a latent clone and between clones suggests that combinations of these agents would need to be used to optimize clinical efficacy.

The cells that we selected for expansion in vitro, were derived from PWH suppressed on ART. As might be expected, the proviral integration sites were biased to ZNF genes and non-genic centromeric and satellite regions both of which are poorly transcribed (Cano-Gamez et al., 2020; Einkauf et al., 2022; Jordan et al., 2003; Lewinski et al., 2005). The somewhat higher levels of HIV-1 transcription among proviruses integrated in centromeric regions than in ZNF genes may be due to constitutive low level RNA Pol II dependent transcription of those parts of the genome (Perea-Resa and Blower, 2018; Zhu et al., 2023).

Although not entirely identical, the cells that we expanded in vitro are closely related to their ex vivo counterparts and typically show a T cell memory or effector memory phenotype (Weymar et al., 2022). Moreover, the genes that were differentially expressed between provirus-containing cells and their uninfected co-cultured counterparts were many of the same genes differentially expressed by latently infected primary cells (Cohn et al., 2018; Collora et al., 2022; Horsburgh et al., 2020; Sun et al., 2023; Weymar et al., 2022). Our experiments demonstrate that the difference between infected and uninfected cells cannot be attributed to cellular response to proviral expression because the only reproducible difference between HIV-1 expressing resting or activated cells within a latent clone is HIV-1. Thus, the differences are likely due to yet to be determined factors that impact T cell infectivity and/or fate decisions at the time of or shortly after exposure to the virus.

Over 50% of intact latent proviruses found in PWH reside within expanded clones of CD4⁺ T cells suggesting that their proliferative expansion occurs in the absence of dominant cytopathic effects (Bui et al., 2017b; Cohn et al., 2015; Hosmane et al., 2017; Lorenzi et al., 2016; Simonetti et al., 2016). Like uninfected cells, CD4⁺ T cells bearing intact latent proviruses expand in response to antigen including CMV, EBV and HIV-1, antigens associated with chronic infections (Demoustier et al., 2002; Douek et al., 2002; Niessl et al., 2020; Simonetti et al., 2021). These clones survive long term and proliferate in vivo, and their component cells can produce infectious virus particles when stimulated in vitro. Anti-CD3 and CD28 plus IL-2 stimulation mimics strong T cell activation signals. Notably stimulation of mixtures of primary cells in vitro only induces virus production by some latent cells (Bui et al., 2017a; Hosmane et al., 2017). One potential explanation for these results is that transcription of the provirus within expanded clones obtained from PWH is heterogeneous with some cells producing virus while others remain silent after stimulation. Our experiments support this idea and demonstrate that even after extensive rounds of division in vitro under strong stimulating conditions intact latent proviral transcription is heterogeneous within a clone of CD4⁺ T cells derived from a single primary cell.

The heterogeneity of HIV-1 expression in clones reflects a variety of stochastic mechanisms that normally regulate cellular transcription. These include binding kinetics of transcription factors, enhancer interactions with gene promoters and chromatin architecture (Brouwer and Lenstra, 2019). Thus, intra-clone variability in HIV expression can be attributed to stochastic cellular transcription dynamics in the region of proviral integration combined with complex regulation of the HIV-1 promoter (Damour et al., 2023; Tantale et al., 2021; Weinberger et al., 2005). Moreover, the observation that HIV-1 transcription in expanded clones of CD4⁺ T cells carrying authentic latent proviruses is stochastic is entirely in keeping with experiments in cell lines carrying randomly integrated indicator proviruses (Jordan et al., 2003; Lewinski et al., 2005). The expression of proviral transcripts that we observe in the absence of cellular activation in some of the authentic latent clones may be attributed in part to the stochastic nature of transcription.

Expanded clones of CD4⁺ T cells dominate the circulating intact proviral reservoir, but the bigger the clone the less likely it is that the virus contributes to rebound viremia when ART is interrupted (Lorenzi et al., 2016). There are several possible explanations for this phenomenon including immune selection against clones that produce high levels of virus, and suppression by autologous antibodies (Bertagnolli et al., 2020; Esmaeilzadeh et al., 2023; Lorenzi et al., 2016). An additional possibility suggested by our quantitative data is that although clones can be very large, the number of cells in a clone that produce virus and the amount of virus they produce at any time is small and therefore unlikely to contribute to rebound.

Targeting the HIV-1 reservoir for elimination by pharmacological methods or immune intervention has been challenging. Within clone and between clone heterogeneity in proviral expression at baseline and in response to LRAs represent important temporal hurdles to HIV-1 remission or cure. Our experiments help understand the nature of some of these impediments and suggest how pharmacologic and immunologic approaches to cure or remission should be optimized including longer lasting and combination interventions to maximally impact diverse clones and clonal members.

Limitations of the study

Limitations of this study include the focus on expanded clones found in circulation and relatively low sample size. Additional clones could provide a more complete picture including expression of proviral transcripts in more permissive genomic sites. In addition, although we do not see substantial transcriptional differences when comparing cultured HIV-1 positive cells with their ex vivo counterparts, it is possible that multiple stimulations and long time in culture influenced the results. An additional limitation is that we cannot be certain that clones in which a fraction of the cells express the provirus in the resting state in culture also do the same in vivo. Lastly, the depth of sequencing by 10X is limiting possibly obscuring subtle but important transcriptomic features of the latent HIV-1 clones.

Methods

Study participants and samples

Samples used in this study originate from people living with HIV (PWH) who enrolled in clinical trials or a sample collection protocol at the Rockefeller University Hospital, New York, NY (NCT03526848: 5104, 5125, 5203, NCT02825797: 9247, NCT02588586: 603, protocol MCA-0966: B207) and at the University of Pennsylvania, Philadelphia, PA (NCT05245292: P301) (Cohn et al., 2018; Gaebler et al., 2022; Mendoza et al., 2018). Informed consent was obtained and samples were investigated under the Rockefeller University IRB-approved protocols TSC-0910 and MCA-0966. Total peripheral blood mononuclear cells (PBMCs) were isolated from leukapheresis by Ficoll separation and frozen in aliquots. All samples used in this study were collected at baseline time points of the clinical trials, except for sample 9247, for which the 12-week time point was used.

Cell sorting and culture

CD4⁺CD45RA⁻ T cells were isolated from PBMCs by negative selection using magnetic separation (Miltenyi cat. 130–096-533; cat. 130–045-901). Fc receptor was blocked by incubating cells with Fc-blocking reagent (Miltenyi, cat. 130–059-901). CD4⁺CD45RA⁻ T cells were stained for live/dead detection at 1:1000 with Fixable Viability Dye eFluor 780 (Invitrogen, cat. 65–0865-14) and Brilliant Violet 605 anti-human TCR Cb1 (BD, cat. 747979), PerCP/Cy5.5 anti-human CD4 (BioLegend, cat. 317428), Pacific Blue anti-human CD3 (BioLegend, cat.300431) 1:100. Cells were also stained, at 1:100, with antibodies binding to TCRβ variable chains specific to the clone: FITC anti-human TCR Vβ17 (603) (Beckman Coulter, cat. IM1234); FITC anti-human TCR Vb7.2 (9247) (Beckman Coulter, cat. B06666); FITC anti-human TCR Vβ22 (5125) (Beckman Coulter, cat. IM1484) (Weymar et al., 2022). Beta Mark TCR Vbeta Repertoire Kit (Beckman Coulter, cat. IM3497) is a panel of 24 different anti-TRBV antibodies that are divided into 8 vials, with a mix of three 3 antibodies each (A to H), that are conjugated with PE, or FITC, or PE-FITC. Since there is no available commercial antibody for the TCRβ variable chains for clones 5104 (TRBV5–4) and B207 (TRBV7–8), we enriched by labeling with all antibodies from Beta Mark TCR Vbeta Repertoire Kit and sorted for the negative population. Individual vials from this panel were also used to stain cells from 5203 and P301, mix D (FITC+ TRBV 12–3+) and mix C (FITC-PE+ TRBV5–1⁺), respectively. In summary, latent cells’ enrichment was done by sorting the following CD4⁺CD3⁺ cell populations: B207 and 5104 - TRBC1⁺TRBV⁻; 603 - TRBC1⁺TRBV19⁺; 5125 - TRBC1⁺TRBV2⁺; 9247 - TRBC1⁻TRBV4–3⁺; 5203 - TRBV12–3⁺; and P301 - TRBC1⁻TRBV5–1⁺.

Five cells per well were sorted by FACSymphony S6 using FACSDiva software (BD Biosciences version 9.5.1) in 96 well U-bottom culture plates. The plates contained 200μL/well of activation media composed of R10 (RPMI-1640 supplemented with 10% heat inactivated FCS, 10 mM HEPES, 100 U/ml penicillin–streptomycin and 2mM L-glutamine - Gibco) supplemented with 50 U/ml of recombinant human IL-2 (Roche cat. 10799068001), anti-human CD3 and CD28 antibodies (Biolegend cat. 317325; 302933), both at 100 ng/mL, and feeder cells (5000 rads irradiated PBMCs, depleted of NK and CD8+ T cells) at a concentration of 1 × 10⁶ cells/mL. To prevent the infection of new cells four antiretroviral (ARV) drugs were added to the media: Tenofovir (1 μM), Emtricitabine (1 μM), Nevirapine (1 μM), and Enfuvirtide (10 μM). The number of sorted plates for each clone was determined based on the predicted frequency in the enriched population (Weymar et al., 2022). Cells were cultured in a humidified incubator at 37°C with 5% CO² for 3 to 4 weeks, with media exchange twice a week (maintenance media - R10 + IL-2 + ARVs). Cultures containing latent cells were further expanded in activation media.

HIV provirus screening and enrichment

The presence of HIV-1 positive cells in the cultures was determined by digital droplet PCR (ddPCR) as previously described (Bruner et al., 2019). For DNA extraction, 2μL of the cultures were collected and transferred to a 96 wells PCR plate containing 8μL of RLT buffer (Qiagen cat 79216). Agencourt RNA Clean XP magnetic beads (Beckman Coulter cat. A63987) was used for DNA purification, according to manufacturer’s instructions. ddPCR reaction droplets were generated using the Automated Droplet Generator (Bio-Rad), reactions were read by QX 200 Droplet Reader (Bio-Rad) and data was analyzed using QX Manager Standard Edition software (version 2.0.0.665 Bio-Rad).

Fraction of latent cells within the culture was determined by using the cellular gene RPP30 to estimate the total number of cells, as previously described (Bruner et al., 2019).

Env and near full length proviral PCR (NFL-PCR), followed by sanger sequencing (GeneWiz, Azenta Life Sciences) and next-generation-sequencing (NGS) (MiSeq Nano V2 300 cycles, Illumina) respectively, were used to identify and characterize the provirus integrated in the CD4⁺ T cell clones, as described previously (Weymar et al., 2022).

Integration site

Integration sites from the clones 603, 5104, 5125, B207, 9247 and P301 were determined or confirmed using the integration site loop amplification (ISLA) method, as previously described (Wagner et al., 2014). Genomic DNA was pre-amplified using multiple displacement amplification (Qiagen REPLI-g single cell advanced kit) prior to the ISLA reaction. Integration site from clone 5203 was confirmed by amplification of the integration site using HIVLTR1 primer and ZNF721Int primer based on available information (Table S6) (Huang et al., 2021).

CD4⁺ T cell activation and latency reversal agents (LRA)

HIV-1 expression by CD4⁺ T cell clones was analyzed in resting (R10) and 24 hours after activation (50U/mL IL-2 + 100 ng/mL anti-CD3 and CD28 antibodies) and treatment with: romidepsin (5 nM, Active Motif cat. 14083) (Gunst et al., 2019), panobinostat (50 nM, Sigma-Aldrich cat. SML3060–10MG) (Fiona Wightman a, 2013) prostratin (2.5 μM, Sigma-Aldrich cat. P0077–1MG), JQ1 (1 μM, Sigma-Aldrich cat. SML1524–5MG) (Darcis et al., 2015), SAHA (Sigma-Aldrich 0.5 μM, cat. SML0061–5MG) (Fiona Wightman a, 2013) and the SMAC mimetic (mSMAC) AZD5582 (100nM, Selleck Chemicals cat. S7362) (Gavin C. Sampey, 2018; Nixon et al., 2020). All assays were performed in the presence of ARVs.

Flow cytometry

The expression of HIV-1 Gag p24 protein was analyzed, in resting and activated cells, by flow cytometry. Cells were fixed and permeabilized with BD Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences cat. 554714) and stained with Gag p24-FITC-conjugated antibody (HIV-1 core antigen-FITC, clone KC57, Beckman Coulter cat. 6604665) diluted 1:100. Readings were performed in FACS Symphony A5 flow cytometer (BD Biosciences) running FACS Diva software version 8.5. Data were analyzed using using FlowJo software (version 10.10.0, BD Biosciences)

HIV-1 Expression

To evaluate HIV-1 expression in bulk cultures cells were plated at a density of 2 × 10⁵ per well in a 96 wells plate. 24h after stimulation cells were counted, spun at 500g for 5 minutes to remove the supernatant and lysed in 350μL of RLT buffer (Qiagen cat 79216). Lysed samples were stored at −80°C. RNA was extracted using the RNeasy Plus Micro kit (Qiagen cat. 74034) according to manufacturer’s instruction. cDNA was synthesized using SuperScript III Reverse Transcriptase (Invitrogen cat. 18080–093) according to manufacturer’s instructions and, 2.5 ng/μL random primers (Invitrogen cat. 48190011) along with primers specific (0.1μM each) for long LTR, gag, pol, tat-rev, nef, poly-A viral transcripts (Einkauf et al., 2022; Yukl et al., 2018) (Table S6).

To estimate the frequency of cells expressing HIV-1 and the level of expression per cell, we performed limiting dilution assays. Defined numbers of cells were sorted into 96 well PCR plates, containing 10μL to 20μL of RLT Buffer (Qiagen cat. 79216) per well in triplicate. The number of cells sorted into individual wells was determined based on prior PCR estimates. RNA was purified with Agencourt RNACleanXP magnetic beads (Beckman Coulter cat. A63987), as above. To eliminate DNA, samples were eluted in 10μL DNAseI mix (Invitrogen cat. 18068015), and the reaction was carried out following manufacturer’s protocol. cDNA was synthesized using SuperScript III Reverse Transcriptase as described above.

qPCR

Multiplex qPCR was used to measure LTR, gag and env expression. The reactions were performed using TaqMan^™ Fast Advanced Master Mix (Thermo Fischer Scientific cat. 4444555). Primers and probes for the HIV-1 transcripts LTR, gag and env are described in Table S6. Primers and probes were used at a final concentration of 0.25 μM. A standard curve for quantification was based on 10-fold serial dilutions of a plasmid from HIV-1 molecular clone HIV-1 NL4–3 NIH (HIV Reagent Program cat. ARP-114) calculated by Design and Analysis software (version 2.6.0 Thermo Fischer Scientific). To calculate the number of transcripts per cell, the percentage of HIV-1 positive CD4⁺ T cells within the culture as determined by ddPCR, was taken into consideration.

For relative expression analysis, host gene PPiA was used for normalization, quantified using a predesigned PrimeTime qPCR Assay (Hs.PT.58v.38887593.g, Integrated DNA Technologies). Expression was calculated using the formula $2^{-\left(C{t}_{\mathit{target}}-C{t}_{\mathit{ref}}\right)}$.

The frequency of cells expressing HIV-1 transcripts was determined using Most Probable Number (MPN) analysis based on the proportion of wells testing positive at each cell dilution in the limiting dilution assay. Statistical analysis employed a custom qPCR Poisson Distribution Analysis Shiny application (qpcrpoissondist), which uses the MPN R package (v0.4.0) to perform maximum likelihood estimation of microbial densities from serial dilution data. This method calculates the most probable number of target templates per sample assuming a Poisson distribution, maximizing the likelihood function to estimate λ (lambda), the mean number of template copies per reaction. The distribution of positive and negative wells across dilution series informs this estimation, where the probability of detection is given by $\mathrm{P}\left(\mathrm{d}\mathrm{e}\mathrm{t}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\right)=1-{\mathrm{e}}^{-\mathrm{\lambda }}$, representing the chance of detecting at least one template copy per well. For wells with detectable HIV-1 RNA, copy numbers per expressing cell were calculated by dividing the total RNA copies detected by the estimated number of expressing cells in that well, as established from the MPN-based frequency analysis. This approach accounts for the stochastic distribution of expressing cells across dilution series and provides quantitative estimates with confidence intervals both for the proportion of cells capable of HIV-1 expression and for the transcriptional output per expressing cell under different treatment conditions. The percentage of the HIV-1 positive CD4⁺ T cells in each culture was considered.

Infection of healthy donor cells

The presence of infectious particles in the supernatant of the cultured cells was evaluated by incubation with healthy donor CD4⁺ T cells. CD4⁺ T cells were isolated from PBMCs from healthy donors using magnetic separation (Miltenyi cat. 130–096-533) and cultured in 96 well U bottom plates, at 5 × 10⁵ cells/well, in activation media. After 24 hours, the culture supernatants were exchanged for supernatants of resting or activated HIV⁺ CD4⁺ T cell clones cultured in the absence of ARVs for 24 hours. Cells were incubated for 6 days at 37°C with 5% CO₂ and supernatants were tested for Gag p24 by Elisa using Lenti-X^™ p24 Rapid Titer Kit (Takara cat. 631476), according to manufacturer’s instructions.

Single-cell RNA sequencing

Single-cell mRNA sequencing was performed using the 10X genomics platform. Chromium Single Cell Library & Gel Bead Kit (10X Genomics cat. PN-1000014) and Chromium Single Cell V(D)J Enrichment Kit, Human T Cell (10X Genomics cat. PN-1000005) were used to create the gene expression and V(D)J libraries, respectively. Libraries were prepared following the 10X genomics protocol and sequenced in the NovaSeq 6000 System (Illumina) using a SP Flow cell (500 cycles) (Illumina cat. 20028402).

Binary base call (BCL) files were demultiplexed and BCLtoFastq software (Illumina) was used to transform the files to FASTQ format. Cellranger multi (v8.0.1 Illumina) was used to align the reads with a modified version of human genome hg38 (Perez et al., 2024), that includes HIV specific sequences from each HIV-1 provirus clone. Seurat (v5.3.0) was used for the analysis using Rstudio server (2024.12.0 Build 467). Cells outside the 200 to 2,500 range and/or with mitochondrial content higher then 10% were filtered out. SCTransform was used to merge sample batches and also to normalize and scale. Cellranger multi was also used to assemble and annotate TCR sequences using 10x VDJ human reference (GRCh38-alts-ensembl-7.1.0). Filtering and analysis of the resulting contig annotations was done using R studio. Latent clones were determined based on identical combined TRA CDR3 and TRB CDR3 nucleotide sequences, utilizing TCR sequence information obtained either previously (603, 5104, 5125, 9247 and B207) (Weymar et al., 2022) or during this study (5203 and P301). Sequences were deposited in National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database under accession number GSE310195.

Mapping scRNA-seq data from cultured cells to a CD4⁺ T cell reference

The CD4⁺ T cell population was extracted from a publicly available, multimodally annotated human peripheral blood reference dataset (Hao et al., 2021). A UMAP reference was reconstructed using the first 50 principal components derived from the RNA expression slot. Cultured cells from each individual were then anchored and mapped onto this reference using Seurat’s FindTransferAnchors and MapQuery functions (reference.reduction = “pca”, dims = 1:50, reduction.model = umap).

Mapping scRNA-seq data from cultured cells onto the published UMAP reference

Single-cell RNA-seq data from cultured cells were mapped onto a previously published UMAP of the same participants (Weymar et al., 2022), which served as the reference. Mapping was performed using the FindTransferAnchors and MapQuery functions from Seurat, with parameters reference.reduction = “pca”, dims = 1:30, and reduction.model = umap. Cell type labels were transferred using the refdata = clusters argument. Only cells with a label prediction score ≥ 0.95 were retained for downstream analysis.

Differential expression analysis

We tested for differential expression using the Seurat function FindMarkers() with default parameters, except min.pct = 0. Genes with an absolute average log2 fold change ≥ 1 and adjusted p-value < 0.05 were considered differentially expressed (DEGs). For comparisons between HIV-positive and HIV-negative cells, latent cells were stratified based on 5’LTR expression. Cells with SCT-normalized expression > 0.01 were labeled as HIV-positive, while those below this threshold were classified as HIV-negative. Upregulated overlapping DEGs across different clones were selected if detected in at least two clones under resting or activated conditions. For comparisons between HIV-infected and non-infected cells, clone-specific HIV-infected cells were compared to non-infected cells pooled across all clones. In this case, upregulated overlapping DEGs were selected if detected in a minimum of 10% of cells from at least four clones under resting or activated conditions.

Supplementary Material

Supplementary information

Fig. S1 contains flow cytometry data showing the enrichment strategy for each latent clone, quantification of secreted virus by p24 ELISA and fraction of latent cells within culture. Fig. S2 presents HIV-1 provirus expression and production of infectious virus particles following mSMAC treatment. Fig. S3 shows 10X data, including UMAPs of latent cells within different T cell subsets. Fig. S4 shows differential gene expression analysis comparing cells transcriptioanly active and silent for HIV-1. Table S1 shows genomic integration site and response to activation for each latent clone. Table S2 shows total number of cells analyzed by single-cell RNA sequencing. Table S3 catalogs differentially expressed genes between HIV-positive and HIV-negative CD4+ T cell.Table S4 presents overlapping DEGs between HIV-positive and HIV-negative CD4+ T cells. Table S5 shows proportion of cells expressing HIV-1 LTR based on single-cell RNA sequencing. Table S6 lists primers and probes used in assays.

Abbreviations list:

BET

Bromodomain and extra-terminal motif

BCL

Binary base call

ddPCR

Digital droplet polymerase chain reaction

DEG

Differentially expressed gene

ISLA

Integration site loop amplification

LRA

Latency reversing agent

mSMAC

SMAC mimetic

MPN

Most probable number

P-TEFb

Positive transcription elongation factor b

Pol II

RNA polymerase II

SAHA

Suberoylanilide hydroxamic acid

scRNA-seq

Single-cell RNA sequencing

SCT

SCTransform

TCRβ (TCRb)

T cell receptor beta chain

TRBV

T cell receptor beta variable region

UMAP

Uniform manifold approximation and projection

V(D)J

Variable, diversity, and joining gene segments

ZNF

Zinc finger

Data availability

The 10X genomics data generated in this project is publicly available and stored at the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database under accession number GSE310195.