An HIV-induced mechanism for T cell quiescence and
proviral latency


 HIV persists in infected individuals despite effective antiretroviral therapy due to the rapid establishment of a latent HIV reservoir, mainly composed of quiescent memory CD4+ T cells1–3. The mechanisms governing the formation of the latent reservoir remain poorly understood. It is commonly assumed that entry of HIV into latency is a rare and random event associated with sporadic infection of effector T-cells transitioning to a memory phenotype4–8. Using human primary CD4+ T cell models, we show instead, that HIV infection itself triggers a strong transcriptomic remodeling that results in activation of a quiescence program, including downregulation of cellular proliferation and metabolic pathways. This transcriptional program is initiated by KLF2, a key regulator of quiescence, along with activation of the p53 pathway and downregulation of MYC. Loss and gain of function studies confirmed that KLF2 and p53 signaling are responsible for the downregulation of MYC and proliferation pathways, and consequently, proviral transcriptional silencing. Thus, HIV infection per se, enhances the formation of the latent reservoir in T-cells, ensuring viral persistence in infected individuals. These findings identify a new and unexpected mechanism for the formation of the latent HIV reservoir, and broaden the repertoire of strategies through which viruses can control the host cell to their advantage.

confirmed by differential gene expression studies (Extended Data Fig. 2A). Th1, Th2, Th17 and Treg polarized cells showed virtually identical change in gene expression pattern in the above studies regardless of polarization identity (Fig. 1A, Extended Data Fig. 1B).
The use of QUECEL models allowed, for the first time, the investigation of the poorly understood process of entry of primary human CD4+ T cells into quiescence at high transcriptomic resolution. Distinctive gene signatures for the quiescence state that were shared among uninfected, +vector and + HIV quiescent cells included the strong upregulation of KLF2, TOB1 and CDKN2B, known pro-quiescence regulatory factors and quiescence markers (Fig. 1C). By contrast, multiple pathways, including MYC signaling and MTORC pathways, both of which are known to be key drivers of the T cell proliferative phase 27 , were downregulated, along with pathways related to cell cycle, transcriptional and translational regulation and cellular metabolism (Fig. 1D, E, Extended Data Fig.   3A). Similarly, genes known to be up-or down-regulated during entry into quiescence, based on mSigDB gene sets and published studies 28 , overwhelmingly showed differential regulation in the expected direction during entry of uninfected cells into quiescence in QUECEL models (Extended Data

HIV infection leads to transcriptomic changes identical to those observed in quiescent cells
Unexpectedly, both dimensionality reduction studies and Jensen-Shannon divergence calculations (Fig. 1A, B) showed that +HIV 72 hpi cells (Extended Data The cellular pathways altered after HIV infection ( Fig. 2A, Extended Data Fig. 4B) closely resembled those that were downregulated after entry into quiescence in uninfected and +vector cells ( Fig. 2A, 1D). Among the ~3300 protein-coding genes that were differentially expressed in both uninfected and +vector cells when entering quiescence, over 2650 showed a correlated change in expression in +HIV 72 hpi cells (Extended Data Fig. 4C), further demonstrating strong similarities between quiescence and 72h post-HIV infection gene expression patterns.
Bolstering the above results, the HIV-downregulated pathways correspond to key proliferative pathways, including both MYC and mTORC1 signaling, consistent with reduced transcriptional and translational activity. Metabolic pathways such as glycolysis, adipogenesis and oxidative phosphorylation similarly showed reduced activity ( Fig. 2A, Extended Data Fig. 4B). Pathways regulated by E2F and HIF1A (hypoxia-related pathways), also major players in the T cell proliferative state 29,30 , were likewise downregulated ( Fig

HIV infection leads to a block to MYC signaling, a key proliferative factor in CD4+ T cells
We evaluated the pattern of enrichment of transcription factor binding sites near the promoters of genes differentially expressed after HIV infection, which pointed to strongly significant reduction in expression of genes with binding sites for the two major proliferative transcription factors in CD4+ T cells, MYC and E2F (Fig. 2D, Extended Data Fig. 6A). The temporal pattern of gene expression changes in response to HIV infection and during entry into full quiescence indicated that a number of key proliferative regulatory pathways and genes, most prominently MYC target genes and MYC itself, show a strong and rapid downregulation predominantly in early time points (within 72 hours) after HIV infection (Extended Data Fig. 6B-E). A second group of regulatory pathways and genes, such as cyclins and E2F family genes and pathway, are predominantly downregulated in later time points (days 4-14 after HIV infection), including during entry into full quiescence after the addition of quiescence-inducing cytokines (Extended Data Fig. 6C-F). Among the upregulated pathways, the p53 pathway and its downstream signaling pathway of Wnt/beta catenin 31,32 were among the "early" responding group. Further, several pro-quiescence genes including KLF2, CDKN2B, CDKN1A, SMPD3 and SMAD3 were also predominantly or exclusively induced in early time points after HIV infection (Extended Data Fig. 6E). Altogether, these data suggest that the majority of the upstream signaling events of the quiescence program were already set in motion within 72 hours post HIV infection, with the gene expression changes at the later time points corresponding to the downstream results.
Although the impact of HIV infection on the induction of the transcriptional program of quiescence has been previously overlooked, individual aspects of this phenotype had been noted in literature including reduced expression of genes associated with activated state in CD4+ T cells 31,33 , suppression of transcription, RNA processing 34 , translation 35 , metabolism, proliferation-related pathways and cellular growth rate 36,37 . To determine the reproducibility and generality of our findings, we used multiple independently-performed RNA-seq studies involving early HIV infection from public databases (Table S1) in which high levels of infection were achieved using different strains of replication-competent HIV from LAI and NL4-3 clones 31,33,35,38,39 and primary HIV isolates 34,36 , which were used to infect diverse CD4+ T cell lines and primary CD4 cells (Table S1). Our re-analysis of these independently-performed RNA-seq studies showed changes in key cellular pathways that were 6 120 125 130 135 140 145 nearly identical to those found in our study. These included an upregulation of p53 pathway and downregulation of critical proliferation pathways, most prominently MYC and E2F signaling and cell cycle-related pathways (Extended Data Fig. 6G). To determine whether the HIV-induced quiescence program was also observed in other reservoir cells, we re-analyzed an RNA-seq dataset focused on microglia 40 , the main reservoir cell type in the brain (Extended Data Fig. 7A-C). Interestingly, we observed nearly identical changes after HIV infection in microglia, suggesting a similar induction of the dormant state after HIV infection in these cells, including the downregulation of MYC and E2F signaling (Extended Data Fig. 7A To determine if downregulation of MYC alone could recapitulate the observed HIV-induced changes in gene expression, we identified multiple independently performed, publicly available RNAseq studies of the impact of MYC knockdown (Table S2). Interestingly, in all cases, MYC knock down resulted in downregulation of E2F targets, in addition to multiple other proliferative pathways, resulting in a gene expression pattern that very closely mimicked that observed following HIV infection (Fig.   2E). Consistently, knock down of MYC in proliferative, uninfected T cells led to loss of proliferative markers, similar to what was observed after HIV infection (Fig. 2F). Thus, downregulation of MYC was sufficient to largely recapitulate the gene expression pattern observed after HIV infection.

Early activation of KLF2 and p53 signaling pathways downregulates MYC and initiates the program of quiescence
To define the mechanism of MYC downregulation after HIV infection, we next investigated the potential contribution of HIV-induced early activation of the p53 pathway, a known negative regulator of the expression of MYC 45 . Treating proliferating primary human CD4+ T cells with two known agonists of p53 pathway, RITA and nutlin 46,47 , led to p53 induction along with a concomitant slowdown in cellular proliferation and pronounced downregulation of proliferation markers within 24 hours ( To find potential candidate genes that may play such a role, we identified the HIV-induced genes which showed a negative correlation with MYC level (Fig. 3E). Interestingly, KLF2, a known strong negative regulator of MYC and a key inducer of cellular quiescence in CD4+ T cells, was among the genes showing the strongest negative correlation with MYC in multiple studies both in the presence and absence of HIV (Fig. 3E, Extended Data Fig. 8C-G). Among the ~100 protein-coding genes which were consistently up or down-regulated after HIV infection in datasets shown in Extended Data Fig. 6G, the most upregulated gene was KLF2 (Fig. 3F), whereas MYC was among the genes consistently showing a strong downregulation after HIV infection (Fig. 3F).
Time course studies in primary human CD4+ T cells following infection with the HIV reporter virus indicated that the level of KLF2 was significantly upregulated at 24 hpi and showed further increase at 48 hpi (Fig. 3G). Importantly, this increase was not a passive side effect of cellular stress but depended on integration of the HIV provirus into the genome, as the addition of raltegravir blocked the rise in KLF2 levels (Fig. 3G). KLF2 knock down prevented entry into quiescence, even when the knock down cells were cultured in quiescence-inducing media (Fig. 3H, I). On the other hand, treatment of proliferating primary human CD4+ T cells with simvastatin, a known inducer of KLF2 48-50 , led to a strong reduction in proliferative markers (Extended Data Fig. 8A), consistent with exit from proliferating state.
To complement the above studies, we also performed an unbiased shRNA screen against 15,000 cellular protein-coding genes to identify those involved in positive or negative regulation of HIV transcriptional activity 51 . Since proviral activation is strongly associated with the T cell activated state 8 , loss of function of factors involved in inhibiting T cell activation will yield positive hits in our screen.
Interestingly, ~40 genes among those consistently up or down-regulated following HIV infection (  Fig. 8H). Interestingly, a recent shRNA screen study independently identified KLF2 as a key factor in maintenance of HIV latency 52 . Taken together, these data not only indicate that KLF2 and MYC are among the most differentially regulated genes and pathways after HIV infection, but also that their loss of function is sufficient to induce and prevent HIV latency reversal, respectively, likely through regulation of the balance between the activated and quiescent state in CD4+ T cells. We also performed single cell RNA-seq analysis on primary CD4+ T cells before infection, 96 hpi with our HIV reporter virus, and following 14 days of incubation of the infected cells in the quiescence-inducing media. As with bulk RNA-seq results (Fig. 1B), dimensionality reduction studies indicated that +HIV 96 hpi cells that were maintained in proliferative media were distinct from the identically maintained uninfected cells and occupied a position between the uninfected and quiescent cells (Fig. 4A). Multiple markers of proliferation and metabolic activation were negatively enriched in the +HIV quiescent and 96 hpi cells compared to proliferative uninfected and +HIV αTCR 24h cells Taken together, our data indicate that HIV infection leads to activation of a multi-pronged mechanism involving KLF2 and p53 pathways that leads to suppression of expression of MYC, which is known to play a pivotal role in maintaining the activated state of T cells 53 . It has been shown that shortly after HIV infection, induction of the IFN response and cellular stress result in p53 activation 54,55 .
In addition to its pro-quiescent function, p53 can directly restrict HIV replication and pro-viral gene expression through multiple mechanisms 56-59 , which may contribute to the observed proviral transcriptional shutdown. While the induction of p53 leads to apoptosis in a fraction of HIV infected 9 215 220 225 230 235 240 cells 60 , a steep rise in the expression of pro-quiescence, pro-survival factor KLF2 after HIV infection leads to survival and induction of quiescence in CD4+ T cells 20,21,24 . In contrast to p53, the mechanism of induction of KLF2 after HIV infection is not known, and the absence of KLF2 upregulation after treatment of primary CD4+ T cells with IFN-α did not support the possibility that similar to p53, upregulation of KLF2 is induced by the innate immune response (Extended Data Fig. 10A). FOXO1, a critical factor in homeostasis of CD4+ T cells, has been shown to play a role in induction of KLF2 61 . It is plausible that cellular stress and/or DNA damage response following viral integration leads to activation of FOXO1 62,63 , which in turn induces the expression of KLF2. Our observation that proviral integration is necessary for KLF2 induction after HIV infection is consistent with this possibility.

Methods
For a description of reagents used in this work, please see Supplementary Table 3.

Primary cell culture
Peripheral blood mononuclear cells from HIV negative Caucasian male donors were purchased from Allcells. Naive CD4+ T cells were isolated using EasySep™ Human Naïve CD4+ T Cell Isolation Kit II according to the manufacturer's instructions. Cells were resuspended at 1 x 10 6 cells/ml in primary culture media (RPMI with Normocin, 10% fetal bovine serum and 25mM HEPES). Immediately after isolation, cells were treated with Dynabeads™ Human T-Activator CD3/CD28 at a 1:1 bead to cell ratio and 60 IU/ml IL-2. Cells were maintained at 37 °C and 5% CO2. Dynabeads™ were removed using a magnet after 24 hours. Polarizations into the four major effector T cell subsets Th1, Th2, Th17 and Treg cells were performed as previously described 8 . During subsequent expansion, cell populations were diluted back to 1 x 10 6 cells/ml in primary culture media and supplemented with an additional 60 IU/ml IL-2 every two days.

Cell lines
Jurkat cells were purchased from ATCC and maintained at 37 °C and 5% CO2 in the described primary culture medium.

Reporter virus
Single-round VSV-pseudotyped HIV reporter virus included the tat, rev, env, vpu and nef genes and a CD8-EGFP fusion protein that permits the use of magnetic beads for a gentle, minimally-disturbing purification of infected cells (Extended Data Fig. 1A). The viral preparation was performed as previously described 8 .

Reporter virus infection
Ten million polarized CD4+ T cells at a concentration of 5x10 6 cell/ml were mixed with high titer singleround VSV-pseudotyped HIV reporter virus and centrifuged at 3,000 rpm for 1.5 hours to promote viral fusion. Cells were incubated overnight in the existing media, then diluted back to 1 x 10 6 cells/ml in primary culture media and supplemented with an additional 60 IU/ml IL-2. 72 hours post reactivation, the infected cells were purified in a minimally disturbing manner using magnetic beads. The purified cells were >95% positive for the viral Nef protein and GFP 8 .

Quiescence and reactivation of CD4+ T cells
To induce quiescence, on day 7 post infection, the purified, HIV-infected (+HIV) or identically maintained uninfected or vector-infected cells were switched to media containing a defined cocktail of quiescence-inducing cytokines including 10 ng/ml TGF-β, 50 ng/ml IL-8 and 10 ng/ml IL-10 8 . Every 4 days, these cytokines were re-added along with 3.75 IU/ml IL-2 to maintain cell viability. To reactivate the quiescent cells, they were stimulated through the TCR using CD3/CD28 beads for 24 hours as described above.
Entry into quiescence is confirmed by flow cytometry assays measuring the progressive reduction in proliferation markers and sharply reduced proliferation rate 8 . Once quiescent, the level of HIV expression is reduced to almost undetectable levels (1%), indicative of HIV latency. Upon stimulation through the TCR by exposure to α-CD3/α-CD28 mAbs, the fraction of HIV expressing cells increases dramatically (>80%), indicative of exit from quiescence and proviral latency.

RT-qPCR assays
For RT-qPCR-based assays, total cellular RNA was harvested using TRIzol reagent and preparation of cDNA was performed with PrimeScript RT Reagent Kit as described, using both oligo(dT) and random hexamers 64 . The resulting cDNA was used in qPCR reactions with iQ™ SYBR® Green Supermix on an Eppendorf Mastercycler Realplex 2 system and analyzed as described 64 . For studies with Raltegravir, either vehicle or Raltegravir were added to the proliferating primary CD4+ T cells at a final concentration of 5 mM 48 hours before infection with the vector or HIV reporter viral preparations, followed by harvesting the cellular RNA at the indicated time points after infection.

Bulk RNA-seq library preparation
Preparation of sequencing libraries for bulk RNA-seq studies were performed as previously

scRNA-seq
Aliquots of approximately 300,000 cells were taken for scRNA-seq analysis at each of the following time points: before the infection with HIV reporter viruses (Uninfected), 96 hours after infection (+HIV 96 hpi), fourteen days after the initial quiescence inducing cytokine treatment (+HIV quiescent) and 24 hours after reactivation using CD3/CD28 beads (+HIV αTCR 24h). The Drop-seq protocol was performed according to the McCarrol laboratory guidebook 65 . Barcoded mRNA capture beads were used at a concentration of 120,000 beads/ml in lysis buffer prepared as described. Cells were used at 100 cells/µl in PBS/10% Bovine Serum Albumin. Flow rates were set to 2,000 µl/hr for cells and beads and 7,500 µl/h for oil. The beads were then washed with 6x SSC and the 5x RT buffer supplied with Maxima H minus reverse transcriptase. Reverse transcription, treatment with Exonuclease I, PCR amplifications and tagmentation were performed strictly as described in the Drop-seq guidebook 65 using Maxima H minus reverse transcriptase, Exonuclease I and Nextera XT DNA Library Preparation To identify genes that change in a concordant manner during quiescence in different cell types capable of entry into a quiescent state we selected quiescence-related pathways in the mSigDB which had been derived from non-cancerous cells, along with a related published study (accession number: GSE24739) on quiescence in primary hematopoietic stem cells 28 . The resulting gene list was used as a signature of gene expression pattern in quiescent cells and was compared to the datasets analyzed in this study.
shRNA screen The shRNA screen using Cellecta shRNA library was performed as previously described 51

Flow cytometry analysis
Cells were washed with PBS and treated with Fixable Viability Dye eFluor™ 450 for 15 minutes, then washed with PBS again. Cells were fixed in 4% formaldehyde and permeabilized with Perm/Wash™ buffer. To detect Cyclin D3 and Ki67, cells were incubated with 3 µg/ml AF647 mouse anti-Ki67 and 3 µg/ml PE mouse anti-Cyclin D3 for 15 minutes. Perm/Wash™ buffer was used to wash the cells twice after incubations. Fluorescent signals were measured using a BD LSRFortessa™ Flow Cytometer.

Analysis of cellular proliferation
To assess proliferation of RITA and vehicle treated cell populations, viable cell count was measured at the same time each day using trypan blue stain and a Countess™ II FL Automated Cell Counter.

RITA, simvastatin and Nutlin treatments
To upregulate p53, cells were treated with the small-molecule MDM2 inhibitors RITA and Nutlin. RITA was used at a concentration of 10 µM for Th17 cells (Fig. 3A,D) or 1.5 µM for unpolarized CD4+ T cells which were more susceptible to RITA toxicity (Fig. 3B,C). Nutlin was used at 10 µM concentration. To upregulate KLF2, cells were treated with 10 µM Simvastatin. As a control for each drug treatment, a parallel population was treated with an equivalent volume of DMSO, the vehicle used for drug delivery.

Knock down studies of MYC and KLF2
One million cells were washed with PBS and resuspended in 60 µl MaxCyte electroporation buffer.
Dicer-substrate short interfering RNAs (dsiRNAs) targeting MYC, KLF2 or a scrambled sequence were added at 3 µM concentration. For MYC and KLF2, two different dsiRNA constructs were used as a mixture. A TEX615-labeled non-targeting dsiRNA was applied at the same concentration as a transfection efficiency control. Electroporation was performed using the MaxCyte STX® Scalable Transfection System. Cells were incubated at 37˚C for 15 mins in the existing buffer, then resuspended at 1 x 10 6 cells/ml in primary culture media with 60 IU/ml IL-2 or the quiescence inducing cytokine cocktail described above. Fluorescent signal from transfection efficiency control dsiRNA was detected using a BD LSRFortessa™ Flow Cytometer.

Immunofluorescence cytochemistry assays
For immunofluorescent detection of p53 following RITA treatment, 200,000 cells were fixed in 4% formaldehyde at 8 hours post-treatment, then permeabilized with Perm/Wash™ buffer. To detect p53, cells were incubated with rabbit anti-p53 primary antibody at 7 µg/ml, then AF647 goat anti-rabbit secondary antibody at 7 µg/ml for 15 minutes each. DAPI was also applied at 1 µM concentration for 5 minutes. Perm/Wash™ buffer was used to wash the cells twice after each incubation. Cells were transferred to 0.17 mm poly-L lysine coated cover slips and mounted onto microscope slides with ProLong™ Diamond Antifade Mountant. Images were collected using a DeltaVision Deconvolution Microscope.

Quantification and Statistical Analysis
All experiments were performed with at least three technical replicates and two or more biological replicates. All error bars shown in figures correspond to standard error of the mean (SEM).

RT-qPCR data analysis
At least two technical repeats and two biological repeats were performed for each RT-qPCR

Analysis of bulk RNA-seq data
An average of ~50 million single end (replicates 1 and 2) or paired-end reads (replicate 3) were obtained for each sample. RNA-seq reads were quality controlled using Fastqc and trimmed for any leftover adaptor-derived sequences, and sequences with Phred score less than 30 with Trim Galore, which is a wrapper based on Cutadapt and FastQC. Any reads shorter than 40 nucleotides after the trimming were not used in alignment. The pre-processed reads were aligned to the human genome (hg38/GRCh38) with the Gencode release 27 as the reference annotations using STAR version 2.7.2b 66 , followed by gene-level quantitation using htseq-count 67 . In parallel, the pre-processed reads were pseudoaligned using Kallisto version 0.43.1 68 with 100 rounds of bootstrapping to the Gencode release 27 of the human transcriptome to which the sequence of the transfected HIV genome and the deduced HIV spliced transcripts were added. The resulting quantitations were normalized using Sleuth. The two pipelines yielded concordant results. Pairwise differential expression tests were performed using edgeR (QL) 69 and false discovery rate (FDR) values were calculated for each differential expression value.
Protein-coding genes that were expressed at a minimum abundance of 5 transcripts per million (TPM) were used for pathway analysis with fold change values as the ranking parameter while controlling false discovery rate at 0.05. Gene Set Enrichment Analysis (GSEA) package was used to identify the enriched pathway and promoter elements using datasets C2, C3, C5 and Hallmark in the mSigDB.
Pathways that showed an FDR q-value <= 0.25 were considered significantly enriched, per the GSEA package guidelines. The number of genes contributing the enrichment score was calculated using the leading edge output of GSEA (tag multiplied by size).
Publicly available datasets focusing on early time points after HIV infection were analyzed as described above and the results were compared to the published manuscript associated with the dataset, when applicable, which showed complete agreement between our analysis and the results reported in the related manuscripts. Datasets that showed >50% infection rate were included in the study in an effort to capture the transcriptomic pattern of the infected, rather than bystander, cells.
Genes which showed concordant differential expression in all datasets were selected and the average of differential expression values was used to identify the top shared differentially expressed genes.

MYC knock down transcriptomic studies
Analysis of the publicly available MYC knockdown datasets (Table S2)

Declaration of Interests
The authors declare no competing interests.           Cell number (x10^6)