Distinct gene expression by expanded clones of quiescent memory CD4+ T cells harboring intact latent HIV-1 proviruses.

Antiretroviral therapy controls, but does not cure, HIV-1 infection due to a reservoir of rare CD4+ T cells harboring latent proviruses. Little is known about the transcriptional program of latent cells. Here, we report a strategy to enrich clones of latent cells carrying intact, replication-competent HIV-1 proviruses from blood based on their expression of unique T cell receptors. Latent cell enrichment enabled single-cell transcriptomic analysis of 1,050 CD4+ T cells belonging to expanded clones harboring intact HIV-1 proviruses from 6 different individuals. The analysis reveals that most of these cells are T effector memory cells that are enriched for expression of HLA-DR, HLA-DP, CD74, CCL5, granzymes A and K, cystatin F, LYAR, and DUSP2. We conclude that expanded clones of latent cells carrying intact HIV-1 proviruses persist preferentially in a distinct CD4+ T cell population, opening possibilities for eradication.

Introduction

Antiretroviral therapy prevents HIV-1 viral replication but does not impact latent proviruses that are integrated into the genome of host CD4+ T cells. The reservoir of latent proviruses is responsible for rapid rebound viremia in most individuals undergoing treatment interruption and is the key impediment to HIV-1 cure (Bachmann et al., 2019; Chun et al., 1997, 2010; Crooks et al., 2015; Finzi et al., 1999; Kearney et al., 2016; Peluso et al., 2020; Siliciano et al., 2003; Wong et al., 1997).

Although the precise composition of the latent compartment is not known, the relative representation of expanded clones of CD4+ T cells harboring intact and defective latent proviruses increases over time such that they account for at least 50% of the reservoir in chronically infected individuals (Antar et al., 2020; Bui et al., 2017; Cho et al., 2022; Cohn et al., 2015; De Scheerder et al., 2019; Einkauf et al., 2022; Hosmane et al., 2017; Lorenzi et al., 2016). Members of infected clones express the same unique T cell receptor (TCR) and have a single distinctive proviral integration site, each of which can serve as a molecular identifier for the latent clone (Cohn et al., 2018; Einkauf et al., 2019, 2022; Huang et al., 2021; Simonetti et al., 2021). Hypotheses about how the latent reservoir is maintained include proviral integration sites that enable cell division (Maldarelli et al., 2014; Wagner et al., 2014) and homeostatic (Chomont et al., 2009) and antigen-driven proliferation (Douek et al., 2002; Gantner et al., 2020; Henrich et al., 2017; Mendoza et al., 2020; Simonetti et al., 2021).

Because cells harboring intact latent proviruses are rare (Bachmann et al., 2019; Crooks et al., 2015; Ho et al., 2013; Peluso et al., 2020; Sengupta and Siliciano, 2018) and have no well-defined markers that distinguish them from other CD4+ T cells (Cohn et al., 2020; Darcis et al., 2019), characterizing their transcriptional program has not been possible to date. Intact proviruses are enriched among CD4+CD45RA−HLA-DR+ memory T cells (Chomont et al., 2009; Cockerham et al., 2014; Hiener et al., 2017; Horsburgh et al., 2020; Lee et al., 2017, 2019), but other surface markers, such as CD2 (Iglesias-Ussel et al., 2013), remain controversial. Latent cells can be identified after re-activation of HIV-1 transcription in vitro (Baxter et al., 2016; Cohn et al., 2018; Grau-Exposito et al., 2019; Liu et al., 2020; Neidleman et al., 2020; Pardons et al., 2019), and there are numerous cell line- or tissue culture-based models of latency (Cameron et al., 2010; Iglesias-Ussel et al., 2013; Krishnan and Zeichner, 2004; Lassen et al., 2012; Marini et al., 2008; Telwatte et al., 2019), but how well these experimental conditions and latency models reflect the physiology of latent cells in circulation is not known.

Here, we present a strategy to enrich clones of quiescent latent cells from samples that were assayed directly ex vivo from 6 individuals living with HIV-1 based on cell-surface expression of their unique TCRs. The enrichment strategy enabled analysis of the transcriptional landscape of these rare cells and identification of distinct features of this population.

Results

During development, T and B lymphocytes assemble unique cell-surface receptors by variable, diversity, and joining gene (V(D)J) recombination. This process is under feedback regulation by the receptor such that each lymphocyte expresses a single specificity, a phenomenon referred to as allelic exclusion (Nussenzweig et al., 1987). Because antigen receptors are fixed early in development, naive T cells that become activated and expand produce clones of CD4+ T cells that are defined by expression of a singular TCR. Thus, the TCR expressed by a clone of latent cells is a unique molecular identifier for members of that clone, and because it is a cell-surface protein, the TCR can be used to enrich members of the latent clone.

People living with HIV-1 frequently harbor large, expanded clones of latent CD4+ T cells. We studied 6 chronically infected individuals controlled on antiretroviral therapy that were aviremic at the time of sample collection (Table S1). Near-full-length sequencing, envelope (env) gene sequencing, and/or viral outgrowth assays showed that the latent reservoir of each of these individuals was dominated by a single expanded intact latent proviral clone (Table S2) (Cohn et al., 2018; Gaebler et al., 2019, 2021, 2022; Huang et al., 2021; Lorenzi et al., 2016; Mendoza et al., 2020). In all cases, the members of these clones could be identified by the sequence of their env gene (Figure S1). The frequency of the clonally expanded latent provirus of interest in the 6 individuals ranged from 13–431/106 total CD4+ T cells based on the frequency of the specific HIV-1 env sequence (Table S2).

Latent cells are predominantly found in the CD45RA− memory T cell compartment (Chomont et al., 2009; Hiener et al., 2017; Morcilla et al., 2021). To determine whether the CD4+ T cells harboring latent proviral clones of interest are in the memory compartment, we purified CD45RA+ and CD45RA− cells and analyzed proviral DNA by sequencing the env gene and producing maximum-likelihood phylogenetic trees. As expected, in all cases, the latent clone of interest was predominantly found in the CD45RA− population (Figures 1A and S2A). Based on staining with an anti-CD45RA antibody and flow cytometric analysis of CD4+ T cells from five individuals, CD45RA− cells accounted for 42%–68% of all CD4+ T cells and therefore purification of CD45RA− cells results in a 1.5- to 2.4-fold enrichment of the latent clone (Figure S2B).

Figure 1

Screening for intact latently infected cell enrichment by sorting for CD45RA and TRBC

Red outlines indicate the population containing the clone of interest. Maximum-likelihood phylogenetic trees show env gene of the latent clone of interest marked in red. Each dot represents a recovered env sequence from the respective subpopulation of CD4+ T cells. The scale bars indicate the number of substitutions per site. Each sort was performed once. See also Figures S2 and S3 and Table S3.

(A) CD4+ T cells from individual 5125 were magnetically sorted into CD45RA+ and CD45RA− populations. Subsequent gDNA extraction, limiting dilution, and env sequencing revealed that the infected clone of interest in was enriched in the CD45RA− memory compartment.

(B) CD4+ T cells from individual 5125 were stained with anti-TRBC1 antibody and sorted into TRBC1+ and TRBC1− populations. The latent clone was enriched in the TRBC1+ compartment.

(C) CD4+ T cells from individual 9247 were stained with anti-TRBC1 antibody and sorted into TRBC1+ and TRBC1− populations. The latent clone was enriched in the TRBC1− compartment.

The TCRβ locus encodes 2 different constant region genes (TRBC1 and TRBC2). Allelic exclusion ensures that all members of a CD4+ T cell clone express the same TCRβ constant region. To determine which of the 2 different TCRβ constant regions is expressed by each of the latent clones of interest, we performed flow cytometry experiments to purify TRBC1+ and TRBC1− CD4+ T cells. In each case, env sequencing revealed that the latent provirus was found among CD4+ T cells expressing one of the two TRBC domains, resulting in 1.5- to 4.5-fold enrichment (Figures 1B, 1C, and S3).

The TCRβ locus contains 48 functional variable domains (TRBV). To determine whether the TRBV can be used to enrich quiescent clones of latent cells, we made use of a collection of 24 different anti-TRBV monoclonal antibodies. The antibodies were divided into 8 groups of 3 each that were conjugated with either phycoerythrin (PE), fluorescein isothiocyanate (FITC), or both (PEFITC).

The TRBV expressed by CD4+ T cells harboring the latent HIV-1 proviral clones of interest in individuals 603, 605, and B207 were known (TRBV-19, TRBV11-2, and TRBV7-8 respectively) (Cohn et al., 2018). However, the TRBV expressed by CD4+ T cells harboring the latent HIV-1 proviral clones of interest in individuals 5104, 5125, and 9247 were not known. To identify the TRBV expressed by CD4+ T cells that harbor the latent clone in these individuals, we combined limiting dilution cell sorting with env sequencing. As a first screening step, CD4+ T cells were stained with the 24 anti-TRBV antibodies and sorted into TRBV+ and TRBV− populations. The latent clone in individual 5104 was present in the TRBV− population, yielding a 2-fold enrichment (Figure 2A). For individuals 5125 and 9247, the latent clone was found in the TRBV+ population (Figures 2B and 2C). To identify the precise TRBV expressed by the latent clone in individual 9247, the 24 anti-TRBV antibodies were split into 2 groups of 12 antibodies, and CD4+ T cells were stained with either one of the 2 groups (Figure 2D). The latent clone was found in one of the 2 groups, and the 12 antibodies split again into 4 groups of 3 anti-TRBV antibodies, with each group containing one FITC-, one PE-, and one PEFITC-labeled antibody (Figure 2E). The latent clone was only found in one of the 4 groups, and the 3 anti-TRBV antibodies in that group were then used to stain and sort for each of the anti-TRBVs individually (Figure 2F). These experiments revealed that in individual 9247, the latent clone expresses TRBV4-3 (Figure 2F). For individual 5125, the 24 TRBV antibodies were split into 8 groups of 3, and the latent clone was found in only one of the groups (Figure 2G). The 3 antibodies in that group were then used to purify the individual TRBV-expressing cells, which showed that the latent clone of interest was found among TRBV2+ cells (Figure 2H).

Figure 2

Screening for the TRBV expressed by the latent clone in individuals 5104, 5125, and 9247

Red outlines indicate the population containing the clone of interest. Maximum-likelihood phylogenetic trees of env sequences derived from sorted population, as indicated. Env gene of the clone of interest marked in red. The scale bars indicate the number of substitutions per site. TRBVA−H indicate the grouping of monoclonal antibodies used in the cocktails (see STAR Methods).

(A–C) Flow cytometry plots show TRBV staining using a combination of 24 anti-TRBV antibodies. Percentage of TRBV− cells is indicated for individual 5104 and TRBV+ cells for 9247 and 5125.

(D) Flow cytometry plots show TRBV staining with one of 2 groups of 12 antibodies for individual 9247. Percentages of TRBV+ cells are indicated. The sorted population is indicated. The positive group was further subdivided in (E).

(E) Flow cytometry plots show TRBV staining with one of 4 groups of 3 TRBV antibodies from individual 9247. Percentages indicate the fraction of TRBV+ cells in each group. The sorted population is indicated. The clone of interest was only found in cells stained with the TRBVH cocktail indicated in red and was further subdivided in (F).

(F) Flow cytometry plots show TRBV staining with 3 different monoclonal antibodies from individual 9247. Percentage indicates the fraction of TRBV4-3+ cells.

(G) Flow cytometry plots show TRBV staining with antibody cocktails A–H from individual 5125. The clone of interest was only found in the TRBVG cocktail, which was further subdivided in (H). Percentage indicates the fraction of TRBV+ cells for each cocktail.

(H) Flow cytometry plots show TRBV staining with the 3 different antibodies in the TRBVG cocktail from individual 5125. The clone of interest was found in the TRBV2+ population. Percentage indicates the fraction of TRBV2+ cells.

To determine the amount of enrichment that could be achieved for each individual, we combined anti-CD45RA, -TRBC, and -TRBV enrichment and performed limiting dilution, env amplification, and sequencing on genomic DNA (Figure 3; Table S3).

Figure 3

Combined enrichment based on CD45RA, TRBC, and TRBV expression in individuals 603, 605, B207, 5104, 5125, and 9247

Red outlines indicate the population containing the clone of interest. Maximum-likelihood phylogenetic trees of env sequences derived from sorted population, as indicated. Env gene of the clone of interest marked in red. The scale bars indicate the number of substitutions per site. Each sort was performed once for limiting dilution Env PCR amplification and sequencing and confirmed the previous screening for the respective marker of the latent clone of interest. See also Tables S3 and S4.

(A) Enrichment in individual 603 after purification of CD45RA− memory cells followed by staining and sorting for TRBC1+ and TRBV19+ cells. Flow cytometry plots show TRBC1 and TRBV19 staining. Percentages of CD45RA−TRBC1+ and CD45RA−TRBC1+TRBV19+ cells are indicated.

(B) Enrichment in individual 605 after purification CD45RA− memory cells followed by staining TRBV11-2+ cells, which were >90% TRBC1+ (Figure S3B). Flow cytometry plot shows TRBV11-2 staining. Percentage of CD45RA−TRBV11-2+ cells is indicated.

(C) Enrichment in individual B207 after purification of CD45RA− memory cells followed by staining and sorting for TRBC1+ and TRBV− cells. Flow cytometry plots show TRBC1 and combined 24 TRBV staining. Percentages of CD45RA−TRBC1+ and CD45RA−TRBC1+TRBV− cells are indicated.

(D) Enrichment in individual 5104 after purification CD45RA− memory cells followed by staining and sorting for TRBC1+ and TRBV− cells. Flow cytometry plots show TRBC1 and TRBV staining. Percentages of CD45RA−TRBC1+ and CD45RA−TRBC1+TRBV− cells are indicated.

(E) Enrichment in individual 5125 after purification CD45RA− memory cells followed by staining and sorting for TRBC1+ and TRBV2+ cells. Flow cytometry plots show TRBC1 and TRBV2 staining. Percentages of CD45RA−TRBC1+ and CD45RA−TRBC1+TRBV2+ cells are indicated.

(F) Enrichment in individual 9247 after purification of CD45RA− memory cells followed by staining and sorting for TRBC1- and TRBV4-3+ cells. Flow cytometry plots show TRBC1 and TRBV4-3 staining. Percentages of CD45RA−TRBC1− and CD45RA−TRBC1−TRBV4-3+ cells are indicated.

To enrich the latent clone in individual 603, CD4+ T cells were magnetically sorted for CD45RA− cells and then stained with antibodies to TRBC1 and TRBV19. The latent clone was found in the CD45RA−TRBC1+TRBV19+ population, resulting in an overall 47-fold relative enrichment (Figure 3A; Table S3). The latent clone in individual 605 was enriched 40-fold in the CD45RA− TRBV11-2+ population (Figures 3B and S3B; Table S3). Antibodies to TRBV7-8 expressed by the latent clone in individual B207 were not available. Therefore, B207 CD4+ T cells were stained with the 24 anti-TRBV antibodies and sorted into a CD45RA−TRBC1+TRBV− population, resulting in only a 9-fold overall enrichment (Figure 3C; Table S3). The latent clone in individual 5104 was found in the CD45RA−TRBC1+TRBV− population with an 11-fold enrichment (Figure 3D; Table S3). The latent clone in individual 5125 was found in the CD45RA−TRBC1+TRBV2+ population with a 54-fold enrichment (Figure 3E; Table S3). Lastly, the latent clone in individual 9247 was found in the CD45RA−TRBC1−TRBV4-3+ population with a 675-fold enrichment (Figure 3F; Table S3).

After combined enrichment, the latent provirus of interest was found in 20, 163, 6, 14, 15, and 18 in 104 CD4+ T cells in individuals 603, 605, B207, 5104, 5125, and 9247, respectively (Table S4).

Although we were able to identify the TRBC and TRBV expressed by expanded clones of interest in individuals 5104, 5125, and 9247, the precise TCRαβ sequence remained unknown. Candidate TCRs were initially identified among clones of CD4+ T cells in the enriched populations by single-cell TCR sequencing by 10x Genomics (Figure 4A). To definitively determine the TCR expressed by the clone of CD4+ T cells harboring the latent provirus, we combined anti-CD45RA, -TRBC, and -TRBV staining and sorted 5 candidate cells into multi-well plates, followed by DNA and RNA extraction. Env amplification and sequencing from genomic DNA were used to identify wells containing latent cells. cDNA from wells containing the latent provirus or negative controls was used to amplify and sequence TCRα and β chains (Figure 4A). CD4+ T cells from individuals 603 and 605 that express known TCRs were used as positive controls to validate the method (Figures 4B and 4C). For individual 603, the TCR expressed by the latent clone was found in 63% of Env+ wells and only in 4% of Env− wells (Figure 4B). In 605, the specific TCR was found in 90% of the Env+ wells and only in 6% of Env− wells. In both cases, there was no enrichment of irrelevant TCRs in Env+ wells. All 7 Env+ wells obtained from individual 5104 contained TCRα TRAV12-3/J9 and/or TCRβ TRBV5-4/J1-1, representing a unique TCR clone in the 10x Genomics sequencing data (Figures 4B and 4C). This TCR was absent in the random selection of Env− wells, and no other TCR was enriched in Env+ wells (Table S5). Similarly, all 4 Env+ wells from individual 5125 contained TRAV26-2/J32 and/or TRBV2/J1-1, which was not found among Env− wells. Finally, in individual 9247, TRAV38-1/J33 and/or TRBV4-3/J2-3 was present in 8 out of 9 of Env+ wells but absent in the random selection of Env− wells. We conclude that the latent clone of interest in 5104, 5125, and 9247 express TRAV12-3/J9/TRBV5-4/J1-1, TRAV26-2/J32/TRBV2/J1-1, and TRAV38-1/J33/TRBV4-3/J2-3, respectively (Figure 4C).

Figure 4

Identification of latent clone TCR sequences

(A) Limiting dilution sorting strategy to identify the TCR expressed by the CD4+ T cells harboring the clone of interest. Each sample was enriched based on CD45RA, TRBC, TRBV expression, and 5 cells sorted per well (c/w) into microwell plates. Env sequencing identified wells that contained a cell of the latent clone of interest. TCRs were amplified and sequenced from all Env+ and a selection of Env− wells in technical duplicates. Each sort was performed once.

(B) Bar graph shows the relevant TCR clonotypes identified in Env+ and Env− wells. 603: Env+ n = 8 and Env− n = 77; 605: Env+ n = 21 and Env− n = 62; 5104: Env+ n = 7 and Env− n = 36; 5125: Env+ n = 4 and Env− n = 31; 9247: Env+ n = 9 and Env− n = 42. See also Table S5.

(C) Pie charts show the relative size of TCR clones as slices. The areas indicated in white represent unique TCR sequences. The number on the left above the pie chart is the donor ID for each individual. The number in the center of the pie chart represents the number of cells assayed for each individual. The clone of interest is indicated by a red arrow and pie slice. See also Table S4.

HIV-1 proviruses can integrate into CD4+ T cells undergoing clonal expansion at the time they start dividing or sometime thereafter. Proviral integration in early stages of clonal expansion would yield a homogenous group of cells, the vast majority of which would harbor an HIV-1 provirus in the same genomic location. Integration at a later time would produce a heterogeneous CD4+ T cell clone wherein only some of the cells in the expanded clone harbor the HIV-1 provirus (Simonetti et al., 2021). To estimate the fraction of infected cells within a particular clone based on its representation by TCR, we performed 10x Genomics single-cell TCR sequencing on samples enriched using the antibody methods described above. We compared the TCR frequencies with the relative frequency of the specific env from proviral DNA in similarly enriched samples (Table S4). For individuals 5104, 5125, and 9247, the frequency of the specific provirus was similar to the frequency of corresponding TCR, but in B207, 603, and 605, the number of cells expressing the specific TCR of interest was 2–3 times higher than the frequency of proviral copies (Table S4). Thus, there is heterogeneity among clones of expanded CD4+ T cells that harbor latent HIV-1 proviruses. In half of our samples, most clonally expanded CD4+ T cells harbor latent HIV-1, and in the others, the provirus is found in a fraction of the clone.

To determine whether CD4+ T cell clones harboring latent proviruses share a transcriptional profile, we combined the 10x Genomics mRNA and TCR sequencing data obtained from the enriched populations of latent cells. TCR sequencing data were used to identify CD4+ T cell clones harboring the latent HIV-1 provirus but were omitted for gene-expression analysis to prevent TCR-biased clustering. Uniform manifold approximation and projection (UMAP) analysis of the transcriptional profile of all 109,217 cells from the six individuals produced 15 unique clusters of CD4+ memory T cells (Figure 5A). Visual inspection revealed that CD4+ T cells expressing the specific TCR associated with latent proviruses are found predominantly in gene expression cluster 7 and neighboring clusters (Figure 5A). On average, 57% of all cells expressing the TCR associated with the expanded latent clone in the 6 individuals were found in cluster 7. Except for individual 9247, the fraction of CD4+ T cells belonging to the latent clone in cluster 7 was greater than all other clusters ranging from 48% to 73% (Figure 5B; Table S6). In individual 9247, the largest fraction of latent cells was in cluster 6 (35%), which is closely related to cluster 7, and the second largest fraction of latent cells was in cluster 7 (29%).

Figure 5

Uniform manifold approximation and projection (UMAP) of 10x gene-expression data

Data representing mRNA expression by 109,217 individual cells is shown. The latent clone of interest as well as the next biggest and next smallest clone in size were located in the UMAP by their TCR sequence. Enrichment and 10x Genomics gene expression and TCR sequencing were performed once for individuals 603, 605, and 5125 and twice for individuals B207, 5104, and 9247. See also Tables S6 and S7.

(A) UMAPs show the position of the cells expressing the latent clone TCR for each of the 6 individuals as red dots. Underneath, UMAPs show the position of the next biggest (yellow triangles) and the next smallest (blue squares) clones in size to the clone of interest. For individual 605, the latent clone of interest was the biggest clone, and only the next smallest clone is shown.

(B) The bar graphs show the fraction of cells in the latent clone (red bars) in each of the 15 UMAP clusters, and the fraction of cells of the next biggest (yellow bars) and the next smallest (blue bars) clones in size to the clone of interest in each of the 15 UMAP clusters (Table S6).

Cluster 7 was enriched in clonally expanded cells which made up 58%–89% of all cells in the cluster (Table S7). To determine whether residence in cluster 7 is a general property of expanded clones of memory CD4+ T cells, we determined the position of the cells in the next largest and/or smallest clones of CD4+ memory T cells in the UMAP (Figure 5A). None of the 11 neighboring CD4+ T cell clones examined were predominantly found in gene expression cluster 7. In addition, some TCR clones clustered by gene expression, whereas others did not. For example, in individual B207, cells in the next largest clone to the one containing latent proviruses were found in 8 of the 15 clusters. In contrast, cells in the next smaller clone in B207 were found primarily in gene expression cluster 10 (63%), which contains cells expressing Foxp3 (Figure 5B; Table S6). We conclude that clonal expansion per se is not sufficient for CD4+ memory T cell accumulation in gene expression cluster 7.

To evaluate proviral transcription, HIV reads in the 10x gene-expression dataset were analyzed. HIV reads were detected in 58 cells that were found in all individuals except 5125. Six of those cells belonged to the latent clone of interest in individual 605. The remaining cells with HIV reads were not associated with any of the latent clones of interest. The number of HIV reads per cell ranged from 1 to 34 (mean 4.9). Those reads were, based on the unique molecular identifiers (UMIs), derived from 1 to 4 HIV transcripts per cell (mean 1.2). The 10x Genomics platform preferentially captures highly expressed genes, and proviral transcription is only found at a relatively low level in resting latent cells (Wiegand et al., 2017). Thus, HIV-expressing cells were difficult to identify reliably, and a comparison of HIV-expressing and transcriptionally silent latent cells was not possible.

Examination of the top 120 genes that define the UMAP clusters revealed that clusters 7 and 8 are closely related (Figure 6A). Cluster 7 is enriched in genes that encode antigen-presenting molecules or their chaperones such as HLA-DR and HLA-DP, as well as CD74, the invariant chain for major histocompatibility complex class II (MHC class II) molecules (Stumptner-Cuvelette and Benaroch, 2002) (Figure 6A; Table S8). In addition, cluster 7 is distinguished by expression of CCL5, granzymes A and K (GZMA, GZMK), cystatin F (CST7), and the nuclear proteins LYAR and DUSP2. Differential expression of some of these genes has been reported in activated latent CD4+ T cells (Cohn et al., 2018; Horsburgh et al., 2020; Lee et al., 2019). However, granzyme A and K proteins do not accumulate specifically in latent cells because flow cytometry-based cell sorting for these markers did not enrich latent cells.

Figure 6

Heatmap of cluster-defining, differentially expressed genes

(A) Heatmap shows up to 10 of the most upregulated genes per UMAP cluster compared with all other clusters. Genes are indicated on the left and clusters above. Yellow shows relatively highly expressed genes and purple relatively downregulated genes. See also Table S8.

(B) Heatmap shows unsupervised clustering based on the expression of differentially expressed genes in cytotoxic CD4+ T cells (Appay et al., 2002; Cachot et al., 2021; Hashimoto et al., 2019; Juno et al., 2017; Takeuchi and Saito, 2017; Zaunders et al., 2004). Genes are indicated on the left and clusters below. Yellow shows relatively highly expressed genes, and black shows relatively downregulated genes.

Cluster 7 shares many upregulated genes with cluster 8, namely CCL5, CST7, and GZMA (Figure 6A). The two closely related clusters differ in that cluster 8 cells also express GZMB, GZMH, perforin 1 (PRF1), natural killer cell granule protein 7 (NKG7), and granulysin (GNLY), each of which is associated with cytotoxic CD4+ T cells (Hashimoto et al., 2019; Turman et al., 1993),. This population of cells is frequently expanded in infection and chronic inflammation (Juno et al., 2017) and is also found enriched among tumor-infiltrating lymphocytes (Oh and Fong, 2021). Moreover, the relative proportion of cytotoxic CD4+ T cells among memory cells is expanded in HIV-1-infected individuals, including those on suppressive antiretroviral therapy (ART) (Appay et al., 2002; Zaunders et al., 2004).

To examine the relationship between the cells in UMAP clusters and genes associated with CD4+ T cell identity, we performed unsupervised clustering analysis with a collection of genes that are up- or downregulated in cytotoxic CD4+ T cells (Appay et al., 2002; Cachot et al., 2021; Hashimoto et al., 2019; Juno et al., 2017; Takeuchi and Saito, 2017; Zaunders et al., 2004) (Figure 6B). In agreement with the UMAP, cluster 8 stands out as most closely related to CD4+ cytotoxic cells, and cluster 7 is its closest relative. Nonetheless, cluster 7 differs from cytotoxic CD4+ T cells in cluster 8 in several respects, including expression of higher levels of CD27, CD28, and CCR7 (Appay et al., 2002) and lower levels of the transcription factors eomesodermin (EOMES) and RUNX3 (Cachot et al., 2021; Juno et al., 2017; Takeuchi and Saito, 2017).

To further characterize the expanded clones of latent cells harboring intact HIV-1 proviruses, the 10x gene-expression data were projected on a reference dataset of a multimodal single-cell analysis of peripheral blood mononuclear cells (PBMCs) from HIV-1-negative individuals (Hao et al., 2021). Cluster 8 falls into the cytotoxic CD4+ T cell (CTL) population, and cluster 7 falls into the CD4+ T effector memory (TEM) population that is marked by the expression of granzymes A and K, DUSP2, CST7, LYAR, and HLA-DRB1 (Figure 7) (Hao et al., 2021). In four individuals, the overall fraction of cells in the clones harboring intact HIV-1 proviruses was greatest in the TEM compartment (Figure 7). In two other individuals, 5104 and 9247, the clones were enriched in the central memory (TCM) compartment. However, the TCM population is larger than the TEM compartment, and when corrected for compartment size, all individuals but individual 605 showed relative enrichment of the latent clone among TEM cells (Table S9).

Figure 7

Projection of 10x gene-expression data on UMAP of CD4+ T cells from multimodal single-cell sequencing

Projection of data representing mRNA expression by 109,217 individual cells on a multimodal UMAP of CD4+ T cells from HIV-negative individuals (Hao et al., 2021). The latent clone of interest in each individual was identified by its TCR sequence and is represented as red dots. Underneath each UMAP, the bar graph shows the fraction of cells in the latent clone in each T cell subpopulation as indicated by the number above each bar. See also Table S9.

In conclusion, gene expression cluster 7 is characteristic of CD4+ TEM cells. This cluster harbors some of the features of CTLs but differs from cytotoxic cells in the expression in key transcriptional factors and components of the cytotoxic machinery (Juno et al., 2017; Takeuchi and Saito, 2017).

Discussion

CD4+ T cells harboring intact latent HIV-1 proviruses represent only a tiny fraction of all CD4+ T cells. They are usually found among subsets of memory cells (Chomont et al., 2009; Hiener et al., 2017; Morcilla et al., 2021), but there are no specific markers that facilitate the purification of these cells (Cohn et al., 2020; Darcis et al., 2019). Consequently, their transcriptional program has only been studied after activation of HIV-1 expression, which thereby allows their identification (Baxter et al., 2016; Cohn et al., 2018; Liu et al., 2020; Neidleman et al., 2020; Pardons et al., 2019). We have devised a method to enrich quiescent latent memory CD4+ T cells by means of their specific antigen receptors. We find that expanded clones of memory CD4+ T cells that carry intact integrated HIV-1 proviruses are enriched among cells that express a transcriptional program that is found in CD4+ TEM cells. Our data are in accordance with studies showing enrichment of genetically intact proviruses in the CD4+ TEM compartment but extends previous observations by revealing the transcriptional program of resting latent cells (Duette et al., 2022; Hiener et al., 2017).

TCM cells are antigen-experienced T cells that are CD45RA−CD45RO+CD27+CCR7+CD62L+ and circulate through secondary lymphoid organs (Chomont et al., 2009; Chtanova et al., 2005; Riou et al., 2007; Sallusto et al., 2004; Schiott et al., 2004; Weng et al., 2012). Upon re-stimulation with cognate antigen, TCM cells secrete interleukin-2 (IL-2) and can differentiate further into TEM cells that are polarized to secrete specific effector cytokines (Sallusto et al., 2004). TEM cells are CD45RA−CD45RO+CD27−CCR7- and express chemokine receptors that enable them to home to inflamed tissues (Chomont et al., 2009; Sallusto et al., 2004; Weng et al., 2012). These cells are more committed to specific T helper (TH) differentiation lineages than TCM cells and secrete effector cytokines or function as CTLs upon cognate antigen challenge (Farber et al., 2014; Riou et al., 2007; Sallusto et al., 2004). The observation that expanded clones of latent cells are frequently found in the CD4+ TEM compartment is consistent with the finding that clones of latent cells and TEM cells develop in response to chronic viral infection (Farber et al., 2014; Harari et al., 2004, 2005; Mendoza et al., 2020; Simonetti et al., 2021; Stubbe et al., 2006; Weng et al., 2012).

CD4+ cytotoxic T cells are a subset of TEM cells whose effector function is killing target cells. Like other TEM cells, they develop in response to chronic antigenic stimulation. They are prominently expanded in HIV-1, cytomegalovirus (CMV), Epstein-Barr virus (EBV) infection, chronic inflammatory diseases, and virally induced malignancies (Abana et al., 2017; Appay et al., 2002; Juno et al., 2017; Oh and Fong, 2021; Zaunders et al., 2004). Their polyfunctional phenotype is most closely associated with the Th1 phenotype, but they can also develop from other T cell lineages. The mechanism that regulates their development is not entirely defined but is associated with expression of T-bet, EOMES, Runx3, and Blimp1 and downregulation of ThPOK, Bcl6, and TCF1 (Cachot et al., 2021; Juno et al., 2017; Takeuchi and Saito, 2017). In keeping with the finding that clones of CD4+ T cells harboring intact latent proviruses cluster in close proximity to CD4+ cytotoxic cells, this subset of TEM cells can respond to antigens found in chronic viral infections such as HIV-1, CMV, and EBV (Cameron et al., 2010; Gantner et al., 2020; Mendoza et al., 2020; Simonetti et al., 2021).

CD4+ T cell-containing integrated proviruses in individuals on suppressive ART can express HIV-1 RNA, but the majority of these cells harbor defective proviruses (Bruner et al., 2016; Cho et al., 2022; Cohn et al., 2015; Einkauf et al., 2022; Eriksson et al., 2013; Hiener et al., 2017; Ho et al., 2013; Imamichi et al., 2016; Kearney et al., 2016; Lee et al., 2017; Peluso et al., 2020; Pollack et al., 2017). When examined based on HIV-1 RNA expression alone, irrespective of whether the provirus is intact, CD4+ T cells containing integrated proviruses are enriched in granzyme B expression, suggestive of residence in the CD4+ cytotoxic T cell compartment (Collora et al., 2022). Although we find a fraction of latent cells containing intact proviruses in the CTL compartment, this is a minority in 5 out of 6 individuals.

Members of a T cell clone expressing the same TCR can adopt different fates depending on several different factors including affinity, antigen concentration, and the cytokine milieu (Chang et al., 2007; Hale and Ahmed, 2015; Lee et al., 2017). Consistent with this idea, CD4+ T cells expressing the TCR associated with latent proviruses are not entirely limited to a single gene-expression cluster. Nevertheless, the observation that a large fraction of the cells in the expanded clones we studied can be found in one specific transcriptional cluster among all memory CD4+ T cells stands in contrast to other similarly sized expanded clones obtained from the same individuals. The latter are found in several different clusters that diverge between clones and individuals, and many of the cells in these clones are widely dispersed among clusters with different transcriptional signatures. For example, among the 11 random memory CD4+ T cell clones of similar size examined in the 6 individuals, only 1 in individual 603 showed enrichment above 5% in cluster 7.

In chronically infected individuals, at least 50% of the cells carrying intact proviruses belong to expanded clones (Bui et al., 2017; Lorenzi et al., 2016), each of which can be distinguished by expression of a specific TCR that is associated with a unique proviral integration site (Cohn et al., 2015, 2018; Einkauf et al., 2019, 2022; Maldarelli et al., 2014; Simonetti et al., 2021; Wagner et al., 2014). The intact reservoir is dynamic, and while the absolute number of cells in the reservoir decreases slowly with a half-life of 4–5 years, clonality increases with time after infection in people on suppressive ART (Antar et al., 2020; Cho et al., 2022; Cohn et al., 2015; Crooks et al., 2015; Hosmane et al., 2017; Lee et al., 2017; Siliciano et al., 2003). However, clonal expansion is not a unique feature of CD4+ T cells harboring intact proviruses. Defective proviruses are also found predominantly in expanded clones (Cho et al., 2022; Cohn et al., 2015; Imamichi et al., 2014), and clones of CD4+ T cells are prominent among non-HIV-1-infected individuals (Qi et al., 2014).

Some of the genes that help define quiescent CD4+ T cells in cluster 7 have also been reported to be expressed in latent cells, such as HLA-DR (Cohn et al., 2018; Horsburgh et al., 2020; Lee et al., 2017), or, like CD2 or LYAR, upregulated in in vitro models of HIV infection and latency (DeBoer et al., 2014; Iglesias-Ussel et al., 2013). In contrast, CCL5 is downregulated upon latent cell reactivation in vitro (Cohn et al., 2018; Iglesias-Ussel et al., 2013) but was upregulated in gene expression cluster 7. Notably, these genes had not been linked to a specific transcriptional program that differentiates latent cells from other CD4+ T cells.

Cells expressing the genes that define cluster 7 are not a unique feature of HIV-1 infection and can be found in non-infected individuals (Hao et al., 2021; Hashimoto et al., 2019). Therefore, HIV-1 proviral integration and latency per se are not required for T cells to acquire the cluster 7 transcriptional program. Accordingly, clones in which all cells were latently infected (individuals 5104, 5125, and 9247) did not display different cluster patterns than clones in which only a fraction of cells was latently infected (individuals B207, 603, and 605). Thus, latent proviral integration does not induce a specific transcriptional profile.

Why latent proviruses are enriched in cells expressing this particular program is not known. One possibility among many is that the cluster 7 program favors suppression of HIV-1 gene expression during T cell activation, which would permit cell division in the absence of HIV-1 virion production and cell death. An alternative, but non-exclusive, possibility is that these cells are among the most likely to respond to a chronic infection and therefore the most likely to be infected and become latent and subsequently undergo clonal expansion in response to a persistent antigen. Cluster 7 shows increased expression of CST7, which is associated with reduced natural killer (NK) cytotoxicity and could help such cells evade elimination (Perisic Nanut et al., 2017).

Limitations of the study

Our analysis is limited to large, expanded clones of latent cells in 6 individuals and did not include non-circulating CD4+ T cell subsets such as tissue-resident CD4+ T cells (Sasson et al., 2020). Whether these observations also apply to less-expanded or tissue-resident populations of latent cells remains to be determined. Moreover, the depth of sequencing available on the 10x Genomics platform is also limiting, and therefore additional elements of the cluster 7 transcriptional program remain to be defined. Despite these caveats, the observation that latent cells preferentially display a specific transcriptional program suggests that these cells could be specifically targeted for elimination.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Michel Nussenzweig (nussen@rockefeller.edu).

Materials availability

This study did not generate new unique reagents.

Experimental model and subject details

Participant cohort

Study participants were recruited at the Rockefeller University hospital and gave informed written consent before participation in the study. The study protocols and procedures (MCA-0966, TSC-0910) met the standards of Good Clinical Practice and were approved by the institutional review board of the Rockefeller University. The participants’ characteristics, including age and sex, are available in Table S1. An expanded intact latent clone was present in the latent reservoir of each individual (Figure S1, Table S2).

After leukapheresis, peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll separation and stored in aliquots in liquid nitrogen. At the time of sample collection, all study participants were receiving ART, were aviremic and had not undergone experimental treatment regimes.

Method details

Cell sorting

All procedures were performed while maintaining cells at 4°C. CD4+ T cells were negatively selected from PBMCs with the CD4+ T cell isolation kit (Miltenyi, cat. 130-059-901). For memory cell separation, CD45RA− T cells were negatively selected using magnetic CD45RA MicroBeads (Miltenyi, cat. 130-045-901).

CD4+CD45RA− T cells were incubated with Fc-blocking reagent for 10 min (Miltenyi, cat. 130-059-901). Fixable Viability Dye eFluor 780 (Invitrogen, cat. 65-0865-14) was used for live/dead cell staining (dilution 1:1,000). The following antibodies were used for surface staining (dilution 1:100): PerCP/Cy5.5 anti-human CD4 (BioLegend, cat. 317428), PacificBlue anti-human CD3 (BioLegend, cat. 300431), Brilliant Violet 605 anti-human TCR Cβ1 (BD, cat. 747979), FITC anti-human TCR Vβ17 (Beckman Coulter, cat. IM1234), FITC anti-human TCR Vβ21.3 (Backman Coulter, cat. IM1483), FITC anti-human TCR Vβ7.2 (Beckman Coulter, cat. B06666), and Beta Mark TCR Vbeta Repertoire Kit (Beckman Coulter, cat. IM3497). The Vbeta Repertoire Kit contains 24 different anti-TRBV antibodies that come in 8 vials (A – H). In each vial, there are three antibodies conjugated with either PE, or FITC, or PE-FITC.

For enrichment, the following population was sorted: individuals B207 and 5104 CD3+CD4+TRBC1+TRBV− lymphocytes; individual 603 CD4+TRBC1+TRBV19+ lymphocytes; individual 605 CD4+TRBV11-2+ lymphocytes; individual 5125 CD4+TRBC1+TRBV2+ lymphocytes; individual 9247 CD4+TRBC1−TRBV4-3+ lymphocytes. Sorts were performed on BD FACS Aria IIu and BD FACS Aria III using FACSDiva software (version 2.0.2, BD). Flowcytometry data was analyzed with FlowJo (version 10.8.1, BD)

gDNA extraction and quantification

Sorted cells were incubated with Proteinase K buffer (10 mM Tris, pH 8, 0.5% SDS, 100 mM NaCl, and 1 mM EDTA) and 250 μg/mL Proteinase K (ThermoFisher, cat. AM2548) at 50°C for 6 h. An equal volume of phenol-chloroform-isoamyl (Sigma-Aldrich, cat. 77617) was added and gDNA was extracted with Phase Lock Gel light tubes (QuantaBio, cat. 10847-800). gDNA was precipitated with ethanol and GlycoBlue Coprecipitant (ThermoFisher, cat. AM9515), washed twice with 70% ethanol and resuspended in TE buffer. gDNA concentration was measured by Qubit dsDNA BR Assay Kit (ThermoFisher, cat. Q32853).

Near full-length proviral amplification and Q4PCR

Near full-length proviral (NFL) amplification was performed as previously described (Gaebler et al., 2019).

The first round of near full-length proviral amplification (NFL1 PCR) was run in a 20 μL reaction volume with 0.5 U of Platinum Taq DNA Polymerase High Fidelity (Fisher Scientific, cat. 11-304-029) and outer PCR primers BLOuterF (5′-AAATCTCTAGCAGTGGCGCCCGAACAG-3′) and BLOuterR (5′-TGAGGGATCTCTAGTTACCAGAGTC-3′) (0.2 μM each) (Li et al., 2007), 2 mM MgSO4 and 0.2 mM dNTPs at 94°C for 2 min; (94°C for 30 s, 64°C for 30 s, and 68°C for 10 min) for 3 cycles; (94°C for 30 s, 61°C for 30 s, and 68°C for 10 min) for 3 cycles; (94°C for 30 s, 58°C for 30 s, and 68°C for 10 min) for 3 cycles; (94°C for 30 s, 55°C for 30 s, and 68°C for 10 min) for 41 cycles; and then 68°C for 10 min as a touch down PCR.

Next, a quadruplex qPCR in a total reaction volume of 10 μL was performed. To 5 μL of TaqMan Universal PCR Master Mix, (ThermoFisher, cat. 4304437), 1 μL of the NFL1 PCR product and 4 μL of a master mix containing the following primer pairs and internal probes were added (final concentration in 10 μL reaction volume in brackets): PS F (5′-TCTCTCGACGCAGGACTC-3′) and PS R (5′-TCTAGCCTCCGCTAGTCAAA-3′) (675 nM each), PS probe (5′/Cy5/TTTGGCGTA/TAO/CTCACCAGTCGCC/3′/IAbRQSp (187.5 nM) (Bruner and Robert, 2018); env F (5′-AGTGGTGCAGAGAGAAAAAAGAGC-3′) and env R (5′-GTCTGGCCTGTACCGTCAGC-3′) (90 nM each), env probe (5′/VIC/CCTTGGGTTCTTGGGA/3′/MGB) (25 nM) (Bruner et al., 2019); gag F (5′-ATGTTTTCAGCATTATCAGAAGGA-3′) and gag R (5′-TGCTTGATGTCCCCCCACT-3′) (337.5 nM each), gag probe (5′/6-FAM/CCACCCCAC/ZEN/AAGATTTAAACACCATGCTAA/3′/IABkFQ) (93.75 nM) (Palmer et al., 2003); pol F (5′-GCACTTTAAATTTTCCCATTAGTCCTA-3′) and pol R (5′-CAAATTTCTACTAATGCTTTTATTTTTTC-3′) (675 nM each), pol probe (5′/NED/AAGCCAGGAATGGATGGCC/3′/MGB) (187.5 nM) (Schmid et al., 2010). qPCR was performed using the Applied Biosystem QuantStudio 6 Flex Real-Time PCR System with the following conditions:

94°C for 10 min; (94°C for 15 s, 60°C for 1 min) for 40 cycles.

QuantStudio Real-Time PCR Software version 1.3 (ThermoFisher) was used for qPCR data analysis. Baseline correction (start cycle 3, end cycle 10) and normalized reported signal threshold (ΔRn threshold = 0.025) was set for all targets/probes. A fluorescent signal above the threshold between cycle value 10 and 40 of any two or more probes was identified and the corresponding NFL1 PCR samples were selected for NFL2 PCR.

NFL2 PCR was run in a 20 μL reaction volume with 1 μL of NFL1 PCR product, 0.5 U of Platinum Taq DNA Polymerase High Fidelity (Fisher Scientific, cat. 11-304-029) and inner PCR primers 275F (5′-ACAGGGACCTGAAAGCGAAAG-3′) and 280R (5′-CTAGTTACCAGAGTCACACAACAGACG-3) (0.2 μM each) (Ho et al., 2013), 2 mM MgSO4 and 0.2 mM dNTPs at 94°C for 2 min; (94 °C for 30 s, 64 °C for 30 s, and 68 °C for 10 min) for 3 cycles; (94 °C for 30 s, 61 °C for 30 s, and 68 °C for 10 min) for 3 cycles; (94°C for 30 s, 58°C for 30 s, and 68°C for 10 min) for 3 cycles; (94°C for 30 s, 55°C for 30 s, and 68°C for 10 min) for 41 cycles; and then 68°C for 10 min as a touch down PCR. NFL2 PCR products were then sequenced.

For each sample, a serial dilution of gDNA was performed to achieve a fluorescent signal above the threshold between cycle value 10 and 40 of any two or more probes in <30% of wells.

Env PCR

Envelope gene amplification was performed as previously described (Salazar-Gonzalez et al., 2008, 2009).

The first round of env amplification (Env1 PCR) for individuals 603, B207, 5104, 5125, and 9247 was run in a 20 μL reaction volume with 0.5 U of Platinum Taq DNA Polymerase High Fidelity (Fisher Scientific, cat. 11-304-029) and outer PCR primers Env5out (5′-TAGAGCCCTGGAAGCATCCAGGAAG-3′) and Env3out (5′-TTGCTACTTGTGATTGCTCCATGT-3′) (0.2 μM each) (Salazar-Gonzalez et al., 2008), 2 mM MgSO4 and 0.2 mM dNTPs at 94°C for 2 min; (94°C for 15 s, 58.5°C for 30 s, and 68°C for 3 min) for 35 cycles; then 68°C for 15 min. Due to a primer mismatch, the first round of env amplification (Env1 PCR) for individual 605 was run in a 20 μL reaction volume with 0.5 U of Platinum Taq DNA Polymerase High Fidelity (Fisher Scientific, cat. 11-304-029) and outer PCR primers B3F3 (5′-TGGAAAGGTGAAGGGGCAGT-AGTAATAC-3′) (Salazar-Gonzalez et al., 2009) and Env3out (5′-TTGCTACTTGTGATTGCTCCATGT-3′) (0.2 μM each), 2 mM MgSO4 and 0.2 mM dNTPs at 94°C for 2 min; (94°C for 15 s, 60.4°C for 30 s, and 68°C for 6 min) for 40 cycles; then 68°C for 15 min.

The second round of env amplification (Env2 PCR) for all individuals was run in a 20 μL reaction volume with 1 μL of Env1 PCR product, 0.5 U of Platinum Taq DNA Polymerase High Fidelity (Fisher Scientific, cat. 11-304-029) and inner PCR primers Env5in (5′-TTAGGCATCTCCTATGGCAGGAAGAAG-3′) and Env3in (5′-GTCTCGAGATACTGCTCCCACCC-3′) (0.2 μM each) (Salazar-Gonzalez et al., 2008), 2 mM MgSO4 and 0.2 mM dNTPs at 94°C for 2 min; (94°C for 15 s, 61°C for 30 s, and 68°C for 3 min) for 35 cycles; then 68°C for 15 min.

4 μL aliquots of Env2 PCR products were added to 16 μL aliquots of nuclease-free water and run on E-Gel 96 Agarose Gels with SYBR Safe DNA Gel Stain, 1% (ThermoFisher, cat. G720841) for visualization. Samples with a band size of ∼2.5 kb were selected for sequencing. For each sample, a serial dilution of gDNA was performed to achieve a PCR product band on agarose gel for <30% of wells.

For env gene sequencing for latent clone TCR identification in individual 605, a shortened nested PCR was performed to amplify a ∼500 bp amplicon of the env gene in which the expanded intact latent clone differed from other proviruses that were enriched. The first round of env amplification (Env1short PCR) for individual 605 was run in a 20 μL reaction volume with 0.5 U of Platinum Taq DNA Polymerase High Fidelity (Fisher Scientific, cat. 11-304-029) and outer PCR primers Env 1,133 F1 (5′-GAGGGGAATTTTTCTACTGTAACAC-3′) and Env 1,956 R1 (5′- GTTCTGCGAATTTT-CAATTAAGGTG -3′) (0.2 μM each), 2 mM MgSO4 and 0.2 mM dNTPs at 94°C for 2min; (94°C for 15 s, 56°C for 30 s, and 68°C for 1 min) for 35 cycles; and 68°C for 15 min. The second round of env amplification (Env2short PCR) for individual 605 was run in a 20 μL reaction volume with 1 μL of Env1short PCR product, 0.5 U of Platinum Taq DNA Polymerase High Fidelity (Fisher Scientific, cat. 11-304-029) and inner PCR primers Env 1,240 F2 (5′- ATCACACTCCGATGC-AGAATAAAAC-3′) and Env 1,765 R2 (5′-TTAGGTATCT-TTCCACAGCCAGTAC-3′) (0.2 μM each), 2 mM MgSO4 and 0.2 mM dNTPs at 94°C for 2 min; (94°C for 15 s, 58°C for 30 s, and 68°C for 1 min) for 35 cycles; and 68°C for 15 min. Env2short PCR products were either sequenced by Sanger sequencing (GeneWiz, Azenta Life Sciences) with primers 1,240 F2 (5′- ATCACACTCCGATGC-AGAATAAAAC-3′) and Env 1,765 R2 (5′-TTAGGTATCT-TTCCACAGCCAGTAC-3′) or on the Illumina MiSeq platform as described below.

Env2 PCR product and NFL2 PCR product sequencing

DNA concentrations of second round PCR products were measured with Qubit dsDNA BR Assay Kit (ThermoFisher, cat. Q32853). Samples were diluted to a concentration of 10–20 ng/μL. For tagmentation, 1 μL of diluted second round PCR product was added to 0.25 μL Nextera TDE1 Tagment DNA enzyme and 1.25 μL of TD Tagmentation buffer (Illumina Tagment DNA TDE1 Enzyme and Buffer, Illumina cat. 20034198). Subsequently, DNA fragments were ligated to i5/i7 barcoded primers from the Illumina Nextera XT Index Kit v2 A – D (Nextera XT Index Kit v2 Set A, Illumina, cat. FC-131-2001 – FC-131-2004) using KAPA HiFi HotStart ReadyMix PCR Kit (Roche, cat. 07958935001). The ligated DNA fragments were pooled and purified with Agencourt RNAClean XP magnetic beads (Beckman Coulter, cat. A63987) for paired-end sequencing at a concentration of 12 pM with MiSeq Reagent Kit v2 (300-cycles) (Illumina, cat. MS-102-2002).

Sequences were assembled with the Defective and Intact HIV Genome Assembler.

Phylogenetic trees

Env sequences were aligned to the HIV HXB2CG env sequence using Geneious Prime software (version 11.0.12, Biomatters). Maximum-likelihood phylogenetic trees were built with PHYML, substitution model HKY85, without bootstrapping, to identify identical env sequences by clustering.

10× genomics

10x Genomics gene expression and V(D)J libraries were generated with the Chromium Single Cell 5′ 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) as described in the 10x Genomics protocol. The 5’ expression library was sequenced with NovaSeq 6000 S1 (100 cycles) (Illumina, cat. 20012865) and the V(D)J library was sequenced with NextSeq 500/550 Mid Output Kit v2.5 (300 Cycles) (Illumina, cat. 20024905).

Latent clone TCR identification

Quiescent latent cells were enriched based on CD45RA, TRBC, and TRBV, and 5 cells per well were sorted into 96-well plates containing TCL buffer (Qiagen, cat. 1031576) with 1% beta-mercaptethanol and snap frozen on dry ice. Plates were stored at - 80°C until further use. After thawing on ice, magnetic bead clean-up was performed with Agencourt RNAClean XP magnetic beads (Beckman Coulter, cat. A63987). TCR mRNA was reverse transcribed into cDNA in a reaction volume of 20 μL with 200 U of SuperScript III Reverse Transcriptase (Invitrogen, cat. 18080044), 40 U of RNasin Plus Ribonuclease Inhibitor (Promega, cat. N2615), 0.5 mM dNTPs, 5 mM DTT, and the following primers: AC1R (5′-ACACATCAGAATCCTTACTTTG-3′), BC1R (5′-CAGTATCTGGAGTCATTGA-3′) (0.125 μM each) (Mamedov et al., 2013). After a second magnetic bead clean-up, binding TCR cDNA and gDNA, the first round of env amplification PCR from gDNA was performed as described above. After this, each well contained multiple copies of TCR cDNA and, if a latent cell was present, multiple copies of the env gene. A 1 μL aliquot from each well was taken for the second round of env amplification and sequencing to identify wells that contained a latent cell of the clone of interest (Env+ wells). From all Env+ wells and a random selection of Env− wells, an aliquot was taken to amplify the TCRα and TCRβ chain separately in duplicates with primers as previously described (Han et al., 2014). The first round of PCR amplification was performed with 2 μL aliquots of cDNA in a total reaction volume of 25 μL, using 0.75 U of HotStarTaq DNA polymerase (Qiagen, cat. 203209), the forward primers TRAV Ph1 (0.06 μM each) and the reverse primer AlphaPhase1 (0.03 μM) for TCRα amplification; the forward primers TRBV Ph1 (0.06 μM each) and the reverse primer BetaPhase1 (0.03 μM) for TCRβ amplification, and 0.2 mM dNTPs at 95°C for 15 min; (94°C for 30 s, 62°C for 1 min, and 72°C for 1 min) for 25 cycles; and 72°C for 5 min. The second round of PCR amplification was performed with 1 μL aliquots of first round PCR amplification products in a total reaction volume of 25 μL, using 0.75 U of HotStarTaq DNA polymerase (Qiagen, cat. 203209), the forward primers hTRAV (0.06 μM each) and the reverse primer AlphaPhase2 (0.03 μM) for TCRα amplification, the forward primers hTRBV (each at 0.06 μM) and the reverse primer BetaPhase2 (0.03 μM) for TCRβ amplification, and 0.2 mM dNTPs at 95°C for 15 min; (94°C for 30 s, 64°C for 1 min, and 72°C for 1 min) for 25 cycles; and 72°C for 5 min. The third round of PCR amplification was performed with 1 μL aliquots of second round PCR amplification products in a total reaction volume of 25 μL, using 0.75 U of HotStarTaq DNA polymerase (Qiagen, cat. 203,209), the AlphaBC primers (0.05 μM) for TCRα amplification, the BetaBC primers (0.05 μM) for TCRβ amplification, PlateNN primers (0.05 μM), PE1 and PE2 primers (0.05 μM each), and 0.2 mM dNTPs at 95°C for 15 min; (94°C for 30 s, 64°C for 1 min, and 72°C for 1 min) for 36 cycles; and 72°C for 5 min. The barcoded PCR amplification products were pooled, concentrated with Agencourt RNAClean XP magnetic beads (Beckman Coulter, cat. A63987), and run on an 1.5% agarose gel. The band between 300 and 400 bp was excised and purified with the NucleoSpin Gel and PCR Clean-up kit (Quiagen, cat. 740609.50). The purified product was subjected to paired-end sequencing at a concentration of 20 pM with MiSeq Reagent Kit v3 (600 cycle) (Illumina, cat. MS-102-3003).

PhiX-derived reads were removed from downstream analysis using a k-mer based approach implemented by bbduk.sh from BBTools v38.72 (https://sourceforge.net/projects/bbmap/). Samples were demultiplexed according to the combination of oligos used to uniquely identify the plate, row, and column they were placed (Han et al., 2014). The quality control check was performed with Trim Galore package v0.6.7 (https://github.com/FelixKrueger/TrimGalore) to trim Illumina adapters and low-quality bases. Paired-overlapping reads were exported into a single read by BBMerge. Subsequently, TRUST4 (Song et al., 2021) was used to reconstruct and annotate the T cell receptor (TCR) sequences. There is a chance of incorrect TCR assignment if the sequencing errors occur at the barcodes present in reads. A threshold was established based on the number of reads used to assemble a specific TCR contig assigned to an empty well. In most empty wells, less than 20 reads for a single TCR chain sequence were detected. An individual threshold was established for each sequencing run. We obtained the putative clonotypes, defined by the 10x Genomics V(D)J single-cell sequencing, that shared either the α or β chain found in that well. The most frequent clonotype associated with Env+ wells in each individual was selected as the clonotype associated with proviral integration.

The bottleneck in this method is the env amplification from one single copy. The efficiency of the TCR amplification is documented as 88% for TCR α and 93% for TCR β based on published literature (Han et al., 2014). This is because TCR is abundantly expressed in T cells. Since either a unique TCR α or β chain is sufficient to identify a TCR of interest from the 10× TCR sequencing data, the detection rate for the specific TCR in Env+ wells is rather high.

Quantification and statistical analysis

Single-cell RNA-seq and single-cell TCR-seq processing

Single-cell RNA-seq binary base call (BCL) files were demultiplexed and converted into FASTQ files using BCLtoFastq prior to alignment to hg38 with CellRanger (v4.0.0) and analyzed in R studio with Seurat (v4). Cells with a mitochondrial proportion greater than 5% and/or a feature count <200 or >2,500 were discarded. Sample batches were combined, normalized and scaled with SCTransform. Uniform Manifold Approximation and Projection (UMAP) clustering was performed selecting the first thirty principal components. Single-cell TCR-seq FASTQs were aligned with CellRanger (v4.0.0) to the default CellRanger VDJ reference. Output contig annotations were filtered and analyzed in R studio with Seurat.

Mapping scRNA-seq to CD4+ T cells reference

CD4+ T cell population was extracted from published human peripheral blood cells multimodal annotated reference (Hao et al., 2021). The UMAP reference from extracted CD4+ T population was recreated using the first 50 principal components of the RNA expression slot and the cells from each individual were anchored and mapped utilizing the FindTransferAnchors and MapQuery functions from Seurat (reference.reduction = "pca", dims = 1:50, reduction.model = umap).

HIV-1 transcript detection using 10× data

Samtools was used to extract reads that cellranger failed to align to the human reference. Sequences extracted from the BAM file generated by cellranger contain identified cell barcodes and UMI in the sequence header. We used bbduk.sh from the BBtools package to search for reads containing HIV-1 k-mers (k = 31) from intact HIV-1 genome sequences obtained from the Los Alamos HIV database. Cell barcode and UMI were extracted from sequences containing HIV-1 k-mers and used to calculate HIV-1 expression.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-human CD3 PB	BioLegend	cat. 300431;RRID AB_1595437
Anti-human CD4 PerCP-Cy5.5	BioLegend	cat. 317428;RRID AB_1186122
Anti-human CD45RA PE-TR	ThermoFisher	cat. MHCD45RA17;RRID AB_10372222
Anti-human TCR Cβ1 BV605	BD	cat. 747979;RRID AB_2872440
Anti-human TCR Vβ17 FITC	Beckman Coulter	cat. IM1234;RRID AB_131007
Anti-human TCR Vβ21.3 FITC	Beckman Coulter	cat. IM1483;RRID AB_131021
Anti-human TCR Vβ7.2 FITC	Beckman Coulter	cat. B06666
Beta Mark TCR Vbeta Repertoire Kit	Beckman Coulter	cat. IM3497
Chemicals, peptides, and recombinant proteins
Recombinant Proteinase K Solution	ThermoFisher	cat. AM2548
Phenol-chloroform-isoamyl alcohol mixture	Sigma-Aldrich	cat. 77617
GlycoBlue™ Coprecipitant	ThermoFisher	cat. AM9515
1× TE Buffer	ThermoFisher	cat. 12090015
Platinum™ Taq DNA Polymerase High Fidelity	Fisher Scientific	cat. 11-304-029
TaqMan Universal PCR Master Mix	ThermoFisher	cat. 4304437
Buffer TCL	Qiagen	cat. 1031576
2-Mercaptoethanol	Sigma-Aldrich	cat. 63689–25ML-F
Agencourt RNAClean XP	Beckman Coulter	cat. A63987
SuperScript™ III Reverse Transcriptase	ThermoFisher	cat. 18080044
RNasin® Plus Ribonuclease Inhibitor	Promega	cat. N2615
HotStarTaq DNA polymerase	Qiagen	cat. 203209
E-Gel™ 96 Agarose Gels with SYBR™ Safe DNA Gel Stain, 1%	ThermoFisher	cat. G720841
Illumina Tagment DNA TDE1 Enzyme and Buffer	Illumina	cat. 20034198
KAPA HiFi HotStart ReadyMix PCR Kit	Roche	cat. 07958935001
Critical commercial assays
CD4+ T cell isolation kit	Miltenyi	cat. 130-059-901
CD45RA MicroBeads	Miltenyi	cat. 130-045-901
MiSeq Reagent Kit v2 (300-cycles)	Illumina	cat. MS-102-2002
Miseq Reagent Kit v3 (600 cycle)	Illumina	cat. MS-102-3003
Qubit™ dsDNA BR Assay Kit	ThermoFisher	cat. Q32853
NucleoSpin Gel and PCR Clean-up kit	Quiagen	cat. 740609.50
Chromium Single Cell 5′ Library & Gel Bead Kit	10x Genomics	cat. PN-1000014
Chromium Single Cell V(D)J Enrichment Kit, Human T Cell	10x Genomics	cat. PN-1000005
NovaSeq 6000 S1 (100 cycles)	Illumina	cat. 20012865
NextSeq 500/550 Mid Output Kit v2.5 (300 Cycles)	Illumina	cat. 20024905
Deposited data
Single cell RNA-Seq and TCR data	This paper	NCBI GEO: GSE204756
HIV-1 Envelope sequences	This paper	GenBank: ON662322 - ON664914
Raw FASTQ TCR sequences	This paper	NCBI SRA: SRR19524296 - SRR19524298
Oligonucleotides
See Table S10
Software and algorithms
Code for single-cell analysis	This paper	https://doi.org/10.5281/zenodo.6950427
Code to identify the TCR of the latent clone of interest	This paper	https://doi.org/10.5281/zenodo.6954076
FACSDiva software (version 2.0.2)	BD	https://www.bdbiosciences.com/en-us/products/software/instrument-software/bd-facsdiva-software
FlowJo (version 10.8.1)	BD	https://www.flowjo.com/solutions/flowjo/downloads
Defective and Intact HIV Genome Assembler	(Gaebler et al., 2019)	https://github.com/stratust/DIHIVA
QuantStudio Real-Time PCR Software version 1.3	ThermoFisher	https://www.thermofisher.com/us/en/home/global/forms/life-science/quantstudio-3-5-software.html
Geneious Prime software (version 11.0.12)	Biomatters	https://www.geneious.com/prime/
GraphPad Prism 9 (version 9.3.1)	GraphPad Software, LLC.	https://www.graphpad.com
Other
Fixable Viability Dye eFluor 780	Invitrogen	cat. 65-0865-14
FcR Blocking Reagent, human	Miltenyi	cat. 130-059-901RRID AB_2892112
Phase Lock Gel light	QuantaBio	cat. 10847–800
MiSeq System	Illumina	cat. SY-410-1003
FACS Aria IIu	BD	N/A
FACS Aria III	BD	N/A