Selective persistence of HIV-1-infected T cell clones can occur through immune reprogramming driven by defective, transcriptionally active proviruses

HIV-LTR exaptation drives BACH2 protein expression in a cellular model that replicates in vivo-selected BACH2 proviral clones

In order to establish a comprehensive view of BACH2 proviral integrations and the effects of in vivo selection, we first undertook comparative profiling of proviral integration sites (ISs) detected in PBMCs of aviremic PLWH on ART and integration sites from in vitro infected primary CD4 T cells. Using data from the Retrovirus Integration Database v2.0 [51], we catalogued 11,579 ISs and 3,898 integration genes (IGs) reported in vivo [12,29,30,35,36,48,52–56], and 463,206 ISs and 14,192 IGs in vitro [26,53,57,58] from a total of 15 studies. Consistent with prior reports [26,29,30,34,35], we detected the BACH2 locus as the most frequent and most in vivo selected integration gene (Fig 1a, S1 Fig). We also observed as previously described that in vivo BACH2-integrations showed significant enrichment towards intron 5 and congruent transcriptional orientation with respect to the gene (Fig 1b).

Because at present direct phenotyping of clones with BACH2-integrated proviruses from PLWH is not technically feasible, we generated a cellular model to study how proviral sequences interact with the BACH2 integration locus. To replicate integration features of in vivo-selected BACH2-proviruses, we used CRISPR/Cas9-mediated genome engineering of a leukemia cell line and targeted a minimal HIV-1 reporter into intron 5 of the BACH2 gene, the precise region that BACH2 integrants are selected for in PLWH (Fig 1c, S2A Fig2). Two clonal lines, HIV_BACH2^D5 and HIV_BACH2^E10 were generated, that carry a monoallelic (clone D5) or biallelic (clone E10) insertion of the HIV-1 reporter in congruent transcriptional orientation at the targeted site (Fig 1c). The reporter consists of an expression cassette for blue fluorescent protein (bfp) and HIV-1 transcriptional activator tat, flanked by two HIV-1 long terminal repeat (LTR) sequences (Fig 1d). Of note, the gag leader sequence containing HIV-1 major splice donor site (MSD) is included in the reporter upstream of bfp, allowing for splicing of LTR-driven transcripts. Clonal lines were validated for correct reporter insertion using Southern blotting and Sanger sequencing (S2B, S2C Fig).

In both clonal lines, transcriptional activity of the LTR (as read out by bfp expression) was minimal but could be induced by exposure to phorbol myristate acetate (PMA) PMA/Ionomycin (Iono), a potent HIV-1 latency reversing treatment (Fig 1e). This is consistent with our previous findings on HIV-1 reporter integrations into BACH2 intron 5 [34]. We next tested whether PMA/Iono stimulation can lead to generation of chimeric LTR-BACH2 transcripts. Using primers targeting the reporter LTR and exon 6 of BACH2, we could indeed amplify LTR-BACH2 transcripts in cDNA derived from PMA/Iono-treated HIV_BACH2^D5/E10 cells (Fig 1f). Sanger sequencing confirmed that chimeric transcripts were generated through aberrant splicing between the reporter MSD and the BACH2 exon 6 splice acceptor site. To assess whether chimeric transcripts were translated to full length BACH2 protein, we carried out Western blot analysis of whole cell lysates of uninduced and PMA/Iono-treated HIV_BACH2^D5/E10 cells. As shown in Fig 1g, transcriptional induction by PMA/Iono indeed resulted in induction of BACH2 protein expression. These results demonstrate that the HIV_BACH2^D5/E10 lines provide a suitable in vitro model to study exaptation of proviral sequences, and insertional activation at the BACH2 RdIG.

BACH2 integrated proviral sequences activate the BACH2 gene expression program

Using our HIV_BACH2 model, we next went on to study downstream effects of LTR exaptation at the BACH2 locus. To do so, we employed a tripartite CRISPR activator (CRISPRa) system, termed synergistic activation mediator (SAM), which allows for guide (g) RNA-directed recruitment of the VP64, p65 and HSF1 transcriptional activator domains to specified targets [59]. Leveraging our previously validated gRNAs for efficient activation of the HIV-1 LTR and the BACH2 gene promoter [34,60], we set out to assess transcriptomic changes induced by SAM-mediated activation of the BACH2-integrated HIV reporter and contrast these to the transcriptomic program induced by activation of the endogenous BACH2 promoter. In brief, HIV_BACH2^E10 cells were transiently transfected with LTR-targeting (LTR-SAM) or BACH2-targeting (BACH2-SAM) SAM. Expression of SAM components was linked to a green fluorescent protein (GFP) marker to allow for selection of successfully transfected cells. Wild-type (wt) cells, which harbor the endogenous BACH2 locus without HIV-1 reporter integrant were used as controls. 72 hours post transfection SAM-expressing cells were enriched by flow cytometry and subjected to transcriptomic profiling (Fig 2a).

We first assessed effects of BACH2-SAM treatment and observed that BACH2 gene transcription was induced in HIV_BACH2^E10 and wt control cells showing the same transcript splicing pattern (Fig 2b). Only minimal transcriptional induction of the BACH2-integrated reporter was observed in BACH2-SAM HIV_BACH2^E10 cells, which confirms our previous results that activity of integrated LTRs within the BACH2 intron 5 is independent of BACH2 locus activity [34]. Next, we examined the impact of LTR-SAM treatment: In HIV_BACH2^E10 cells, LTR-SAM induced HIV reporter transcripts as expected, but additionally a prominent splicing activity from the reporter-inherent MSD to BACH2 exon 6 was observed, giving rise to chimeric LTR/BACH2 transcripts. The splicing pattern of these chimeric transcripts downstream of exon 6 was reminiscent of endogenous BACH2 transcripts (Fig 2b).

In order to assess the extent to which BACH2 transcription induced by the LTR versus the endogenous promoter differentially affects the overall transcriptional profile of the cells, we performed RNA-seq on the SAM-induced HIV_BACH2^E10 cells. As shown in Fig 2c, in principal component space, LTR-SAM and BACH2-SAM-treated HIV-BACH2^E10 samples segregate close to one another together with BACH2-SAM wt cells and away from LTR-SAM wt cells and non-treated wt or HIV_BACH2^E10 cells. This indicates that transcriptomic changes induced by LTR-driven insertional activation of BACH2 mirror changes driven by endogenous BACH2. This was confirmed by differential gene expression analysis, which showed a highly similar pattern of differentially transcribed genes driven by LTR-driven versus endogenous BACH2 activation (Fig 2d, 2e). Gene set enrichment analysis of differentially expressed genes (DEGs) in both comparisons further demonstrated that near identical cellular pathways were affected (S3A, S3B Fig). Hence, these data indicate that transcriptional activity of in vivo selected BACH2-integrated proviruses can result in transcriptomic reprogramming driven by provirus-induced, aberrant BACH2 expression.

Aberrant BACH2 expression in CD4 + lymphocytes drives cells toward a proliferative precursor memory-like fate

We next set out to explore how aberrant activation of the BACH2 gene program, as observed via provirus promoter exaptation in our cell line models, could impact primary CD4 + T cells, as main HIV target cells in vivo. In this context it is noteworthy that dysregulated BACH2 expression has previously been reported to cause disease: BACH2 haploinsufficiency can lead to hypogammaglobulinemia and immune dysregulation, manifesting clinically as recurrent infections and chronic intestinal inflammation [61]. The BACH2 gene locus has also been prominently associated with conditions linked to immune imbalance in genome-wide association studies (GWAS) (S4 Fig) [62]. All characterized single nucleotide polymorphisms in these studies mapped to non-coding, intronic or upstream regions of BACH2 with highest occurrence at the 5’ end of the locus and thus coinciding with regions that likely regulate BACH2 transcription and expression levels.

In order to investigate how elevated BACH2 expression as a result of LTR exaptation could alter functions and differentiation states of primary CD4 + T cells, we used a lentiviral overexpression system (Fig 3a). In short, CD4 + lymphocytes isolated from PBMCs of HIV-seronegative donors were TCR-stimulated for 24hs and transduced with lentiviruses coding for cDNAs of either BACH2 (LV-BACH2) or green fluorescent protein (LV-GFP) as negative control. By linking BACH2- and GFP-cDNA expression to expression of a surface tag (SBP-dLNGFR), transduced cells could be enriched using magnetic beads (Fig 3b). Transduced cells showed robust expression of BACH2 or GFP RNA and protein over a time course of 25 days as assessed by RT-qPCR and flow cytometry (S5A Fig).

We first assessed whether constitutive BACH2 expression affects CD4 + T cell proliferation in this system. LV-BACH2 and LV-GFP-transduced cultures were stained with a marker dye 12 days post-transduction (13 days post TCR stimulation), and cell-division-dependent dilution of the dye was measured for up to five days. We observed a significant increase in proliferation of BACH2-transduced cells compared to GFP-transduced or non-transduced control cells as indicated by raised division index (Fig 3c). In agreement, LV-BACH2-transduced CD4 + T cells outgrew LV-GFP transduced cells in a growth competition co-culture assay (Fig 3d). These data suggest that BACH2 overexpression drives expansion of primary CD4 + T cells.

We next determined BACH2-induced changes in the transcriptomic profile of primary CD4 + T cells. BACH2- and GFP-transduced cells were enriched at days 4, 9, and 25 post-transduction and analyzed by RNA-seq. In principal component space, LV-BACH2- and LV-GFP cells increasingly segregate away from one another during the 25 days of analysis (Fig 3e). In agreement, differentially expressed gene (DEG) analysis revealed that BACH2 overexpression progressively drives a new transcriptional program: at day 4 post transduction, 14 genes were differentially expressed, and this number increased to 183 at day 9 and 323 at day 25 (S1 Table, Fig 3f). As expected, BACH2 transcripts were detected as significantly upregulated at all time points in LV-BACH2 cells.

We went on to assay transcriptomic changes using gene set enrichment analysis (GSEA) and STRING functional network profiling based on DEGs between LV-BACH2 and LV-GFP cells. At day 4, no particular differentially regulated pathway was highlighted, likely due to the low numbers of DEGs at this time point. However, we observe that LV-BACH2 cells show upregulation of the RAS protein activator like 2 (RASAL2) gene, coding for a Ras GTPase-activating protein that has been associated with inhibition of T cell activation and cytokine production [63], as well as downregulation of adhesion g protein-coupled receptor G1 (ADGRG1; also known as GPR56) and heme oxygenase 1 (HMOX1), two genes linked to effector fates, namely cytotoxic T cells and T regs, respectively [64,65] (S1 Table).

At d9, GSEA shows downregulation of cellular pathways associated with chemokine signaling, cellular migration and immune response (Fig 3g). Most downregulated DEGs belong to two main networks linked to effector T cell differentiation and function (e.g., pro-inflammatory cytokine IFN gamma and chemokine receptor IL17RB [66]) as well as arachidonic acid metabolism, necessary for inflammatory signaling [67–69] (Fig 3h). In addition, we observe downregulation of PRDM1 (encoding B-lymphocyte-induced maturation protein-1 (Blimp1)), a direct target of BACH2. PRDM1 is a master transcriptional regulator that promotes terminal effector differentiation of T cells and reduced PRDM1 expression levels have been shown to support an early memory T cell phenotype [70,71].

At d25, LV-BACH2 cells show further downregulation of a large subset of genes associated with effector T cells and inflammatory cytokine signaling, including CXC-chemokine receptor 6 (CXCR6), interleukins 5 and 13 (IL5/13) and granzyme B and H (GZMB/H) (Fig 4a). Furthermore, we observe upregulation of TCF7 (encoding the T cell factor-1 (TCF-1)) and lymphoid enhancer-binding factor 1 (LEF-1), that have been reported to act synergistically as transcription factors that support formation of T cell memory precursors and acquisition of stemness properties for long-term persistence [72,73]. We also find positive enrichment for pathways linked to mitotic activity, DNA repair and ribonucleotide complex binding in LV-BACH2 cells at d25, driven for example by upregulation of the mitotic regulators NIMA-related kinase 2 (NEK2) and cytoskeleton associated protein 2 (CKAP2) (Fig 4a, S5B Fig). This proliferative transcriptomic signature matches our results using cell tracer assays (Fig 3c).

To further characterize LV-BACH2 cells, we undertook multi-cytokine profiling in supernatants from LV-BACH2 cells at d24 post transduction. Consistent with observed BACH2-induced transcriptomic changes, we detected reduced levels of several cytokines typically secreted by effector CD4 + T cells, including IFN gamma and TNF alpha (Fig 4b). In summary, these data suggest that aberrant BACH2 expression induces a gradual immune reprogramming, marked by initial inhibition of CD4 + T cell activation and effector type differentiation, followed by initiation of a memory-precursor-like program that features a signature of persistence and proliferation.

Finally, we asked whether BACH2-induced reprogramming in primary CD4 + T cells in vitro correlates with characteristics of clonally expanded HIV-1 infected cells in PLWH on ART. These cells demonstrate a signature of immunoregulatory markers and immune checkpoint molecules that confer T-cell functional inhibition and resistance to immune-mediated killing [50,74], coupled with a pro-survival gene expression profile [75–77]. As observed for expanded clones in vivo, LV-BACH2 cells at d25 show upregulation of genes coding for cell-surface proteins that protect from CTL or NK cell-mediated killing, such as non-classical MHC class I molecule HLA-E, cell adhesion mediator poliovirus receptor (PVR), cell surface receptors TNFRSF14 (encoding herpes virus entry mediator (HVEM)) and programmed death ligand 1 (PDL-1) (Fig 4c). We also detected induction of immune checkpoint genes, similar to expanded clones in vivo, such as T-cell immunoreceptor with Ig and ITM domains (TIGIT), programmed cell death protein 1 (PD1) and B- and T-lymphocyte attenuator (BTLA) (Fig 4c). Furthermore, BACH2 overexpression activates an intrinsic anti-apoptotic state in CD4 + T cells: as reported for expanded HIV-1 infected clones in vivo, LV-BACH2 cells were marked for example by upregulation of apoptotic inhibitor BCL2L1 (encoding for BCL-XL protein), inhibitor of apoptosis (IAP) family member BIRC5 and IAP regulator USP19 (deubiquitinase that stabilizes IAP proteins) (Fig 4d). Thus, these data indicate that BACH2-overexpression drives a CD4 + T cell phenotype that overall mimics the clonal expansion immune signature that is selected for in PLWH [50,74–77].

Aberrant expression of the STAT5B RdIG induces an activated, inflammatory CD4 + T cell state

Aside from BACH2 as a highly prominent RdIG in PLWH on ART, a selection of further proviral integration genes have been proposed to be subject to insertional activation that could be driving cellular expansion [20,25,26,28–30,78]. We asked whether insertional activation of these genes by proviral promoter exaptation could redirect CD4 + T cell fate in a similar way as here-demonstrated for BACH2. Signal transducer and activator of transcription 5B (STAT5B) is - like BACH2 - a highly selected gene for proviral integrations in PLWH on ART (S1 Fig) [26,28–30]. STAT5B is a central transcription factor in CD4 + T cells that regulates homeostasis, survival, and differentiation [79–82], and has been associated with CD4 + T cell differentiation into Th1, Th2, Th9 and Treg cells [83]. In PLWH the STAT5B RdIG shows significant clustering bias towards congruent proviral integrations in intron 1 (Fig 5a) [26,28] and chimeric provirus/STAT5B transcripts (spliced into STAT5B exon 2 which contains the start of the open reading frame) have been detected in PBMCs of PLWH off and on ART [44], providing evidence for provirus promoter exaptation at this locus. Indeed, using a cell line clone transduced with a near full length HIV reporter that has integrated in transcriptional congruent orientation within intron one of the STAT5B gene locus, i.e., the region selected for proviral integrations in PLWH on ART (S6A, S6B Fig), we observe that activation of LTR transcriptional activity though treatment with PMA/Iono resulted in generation of LTR-STAT5B chimeric transcripts, thereby raising overall levels of STAT5B-coding transcripts (S6C, S6D Fig). These results mimic above-described findings on HIV_BACH2 cell models and provide further evidence for HIV provirus-triggered insertional activation of STAT5B gene expression.

We went on to address whether exaptation-driven, aberrant STAT5B activation could redirect CD4 + T cell fate in a similar way as BACH2. To do so, we used above-described overexpression assay in primary CD4 + T cells. We transduced activated CD4 + T cells from HIV-seronegative donors with lentiviral SBP-deltaLNGFR bi-cistronic expression construct for STAT5B (LV-STAT5B) (Fig 5b). LV-STAT5B cells showed robust expression of STAT5B at RNA and protein level (Fig 5c).

Contrary to BACH2, aberrant STAT5B expression did not alter proliferation rates of CD4 + lymphocytes in comparison to controls in a marker dye-based assay (Fig 5d). Further, transcriptomic profiling and DEG analysis at 12 day post-transduction revealed that STAT5B overexpression led to upregulation of 222 and downregulation of 145 genes (S1 Table). Differing from LV-BACH2-treated cells, these transcriptomic changes positively highlighted pathways of cell activation, immune response and MHC class II antigen presentation as shown by GSEA and STRING analysis (Fig 5e, 5f). Functional networks promoting inflammatory signaling and cell survival were also found enriched (Fig 5f). Profiling of cytokines secreted by LV-STAT5B cells indeed showed increased secretion of several cytokines associated with effector CD4 + T cell function, such as TNF alpha, interleukin 4, interleukin 9 and interleukin 17F (Fig 5g). Taken together, aberrant STAT5B expression promotes an activated, effector CD4 + T cell state with inflammatory signature. This contrasts BACH2-driven CD4 + T cell reprogramming and suggests that insertional activation of RdIGs through LTR exaptation likely results in clones with different T cell states depending on the proviral integration site.

A negative feedback loop autoregulates endogenous BACH2 expression

Our results indicate that an aberrant increase in BACH2 expression leads to changes in CD4 + T cell state and drives longevity and cell proliferation. Our data also suggest that transcriptional activity of BACH2-integrated proviruses can cause such aberrant, LTR-driven BACH2 expression. In context of murine leukemia virus (MLV)-associated B cell lymphomas, LTR-driven BACH2 expression has indeed been associated with transformation and malignant expansion [84]. However, outside the infection context, BACH2 has to our knowledge not been associated with malignant cell transformation. We therefore asked whether there might be differences in the regulation of endogenous versus LTR-driven BACH2 expression. In particular, we were interested in a potential autoregulatory feedback loop for endogenous BACH2 that had been previously hypothesized [45]. We undertook investigation of the BACH2 promoter region to gain further understanding. Motif scanning of a 40kb region around BACH2 exon1 identified four T-MARE consensus motifs (TGCTGA(G/C)TCA; denoted binding site (BS) 1–4, Fig 6a), which are motifs that have been found to be targeted by BACH2 [85]. To understand whether these motifs could have functional relevance, we assessed their epigenetic signature in CD4 + T cells. Specifically, we profiled for chromatin accessibility and distribution of two transcription-regulatory histone post-translational modifications, namely trimethylation of H3K4 and acetylation of H3K27, using assay for transposase-accessible chromatin (ATAC) sequencing and chromatin immunoprecipitation (ChIP) sequencing, respectively. We found that two motifs, BS1 and BS2, had an accessible chromatin profile and enriched for both histone marks, indicating that these sites may act as cis-regulatory elements for the proximal endogenous BACH2 gene promoter in CD4 + T cells (Fig 6a). Next, we assessed whether BACH2 protein can directly bind BS1 and BS2 motifs. CD4 + T cells, transduced with LV-BACH2, were submitted to ChIP using an antibody targeted to the BACH2-fused V5 protein tag (see Fig 3b) and enrichment for BS1, BS2 and BS3/4 on ChIP-ed DNA was assessed by quantitative PCR. We observed significant enrichment for BS1 and BS2, while BS3/4 failed to enrich over a negative control region 1.2kb upstream of BS2 in BACH2 intron 1 or the HOXC3 gene promoter, which was equally tested as negative control region (Fig 6b). These results demonstrate that BACH2 can target cis-regulatory elements in proximity of its own gene promoter and suggest that endogenous BACH2 expression is subject to autoregulation in primary CD4 + T cells.

Since BACH2 has been primarily described as negative regulator of target gene transcription through competitive occupation of cis-regulatory elements [38,86,87], we hypothesized that BACH2 binding at BS1 and/or BS2 negatively impacts BACH2 promoter activity. To further investigate this hypothesis, we undertook a luciferase (GLuc) reporter assay that has been previously used to assess cis-regulatory activity of selected DNA elements [88]. We co-transfected HEK293T cells with BACH2 or control (ctrl) overexpression vectors and GLuc reporter vectors that either included no additional sequences (GLuc), the BACH2 BS1 site (BS1-GLuc) or a BACH2 intronic negative control site (NC-GLuc) upstream of the minimal TK promoter driving GLuc reporter gene expression. Notably, HEK293T cells do not endogenously express BACH2. In agreement with our hypothesis, we observed that presence of BACH2 BS1 reduced reporter expression, and this effect was seen only upon BACH2 co-expression (Fig 6c).

Finally, to further investigate whether BACH2 may limit its own expression via autoregulatory feedback, we used our aberrant BACH2 expression system in primary CD4 + T cells transduced with LV-BACH2. We assessed whether transcriptional activity from the endogenous BACH2 locus could be affected by raised BACH2 protein levels. To do so, we quantified BACH2 transcripts that originated from the non-coding region in exons 1–5 and reflect endogenous BACH2 transcriptional activity, in presence or absence of lentiviral-driven BACH2. Notably the LV-BACH2 overexpression construct only contains coding BACH2 sequences. Indeed, CD4 + T cells that were transduced with lentiviral LV-BACH2 demonstrated significantly reduced transcription of the endogenous BACH2 locus compared to controls (Fig 6d). Total and endogenous BACH2 transcript levels were negatively correlated (Fig 6e).

Altogether, these data suggest that BACH2 expression levels are usually restricted through a negative autoregulatory feedback loop involving binding of BACH2 to BS1 and/or BS2 in its own promoter region. Outside the HIV-1 infection context, this feedback loop could thus prohibit a continued raise of BACH2 protein levels, and prevent reprogramming toward the proliferative CD4 + T cell phenotype that we observe upon overexpression. On the other hand, in PLWH on ART the prominent selection of cells harboring BACH2 proviral integrants could thus, at least in part, be a result of LTR-driven aberrant BACH2 expression that escapes endogenous regulatory mechanisms.

Aberrant BACH2 expression through LTR exaptation can be controlled by HIV-1 latency promoting agents

Our data suggest that LTR exaptation can affect CD4 + T cell profiles, which may contribute to immune modulation and dysregulation observed in PLWH on ART. The immune modulatory activity is dependent on promoter activity of the integrated HIV LTR. To determine whether agents that interfere with HIV transcriptional activity can halt exaptation-dependent effects on the integration site locus, we tested how latency promoting agents (LPAs) impact on LTR-driven BACH2 expression within our HIV_BACH2 cellular model. We focused on two previously reported LPAs, Spironolactone, an aldosterone receptor antagonist that has been shown to interfere with proviral transcription through degrading components of the TFIIH transcription factor complex [89,90]; and KRIBB11, a heat shock factor 1-pTEFb complex inhibitor that has been shown to inhibit HIV-1 latency reversal [91]. HIV_BACH2^D5/E10 cells were pre-treated with Spironolactone, KRIBB11, or DMSO control and then stimulated with PMA/Ionomycin to induce transcription of the BACH2 integrated proviral reporter. Samples were processed for flow cytometry to detect reporter expression and quantitative RT-PCR for LTR/BACH2 chimeric transcript detection. Both LPAs reduced expression of the HIV_BACH2 bfp reporter (Fig 7a). In addition, both LPAs also significantly reduced induction of the chimeric LTR/BACH2 transcripts (Fig 7b). Thus, in the model system used, treatment with LPAs is able to reduce LTR exaptation activity and block insertional activation of the provirus integration gene.

Generation and analysis of provirus integration site datasets

Data on provirus integration sites was extracted from the Retrovirus Integration Database (https://rid.cancer.gov) [51]. Datasets of integration sites were generated combining integration site data from 11 studies on in vivo integrations mapped in PBMC samples of PLWH on ART [12,29,30,35,36,48,52–56], and from 4 studies on in vitro integrations mapped in in vitro HIV-1 infected CD4 + T cells [26,53,57,58]. The IS were annotated using the ‘closest’ subcommand of BEDTools v2.27.1. For gene annotation, the RefSeq GRCh37 Release 105.20220307 reference was used, selecting only the “RefSeq Select” isoform for each protein-coding transcript. For visualizing of genome-wide integration site frequency, Circos plots were generated using Circos Builder (v069-815) of Galaxy Server, based on human hg19 karyotype ideogram [105]. Bar plots depicting integration site distribution with regards to intronic position and provirus orientation were generated using Ggplot2. The integration site (IS) frequency for both in vivo and in vitro datasets was calculated for each gene using the formula: IS_freq = (IS count per gene / total IS count) × 100. To compare the two datasets, a scatter plot was generated where the x-axis represents individual genes of interest, and the y-axis displays the log₁₀-transformed difference in IS frequency between in vivo and in vitro.

Genome-wide association study analysis

Genome-wide association study (GWAS) data including SNP location and trait information in context with BACH2 was downloaded from https://www.ebi.ac.uk/gwas/genes/BACH2 (accessed May 2024). SNPs in BACH2 were visualized in JBrowser2 [106]. Trait information was categorized, counted and plotted using Microsoft Excel.

Cell culture

HAP1 cells (Horizon Discovery) were used for generation of HIV_BACH2 reporter lines and cultured in IMDM medium (Thermo Fisher Scientific) with 10% FBS and 1% Pen/Strep. MOLT4 cells (incl HIV_STAT5B cells) were cultured in X-VIVO 15 cell medium (Lonza) with 10% FCS and 0,5% Pen/Strep. HEK293T cells were used for the GLuc reporter assay as well as lentivirus production and were cultured in DMEM (PAN BIOTECH) with 10% FCS and 1% Pen/Strep. Buffy coats for PBMC isolation were obtained from anonymized healthy donors through collaboration with the Institute of Transfusion Medicine, University Medical Center Hamburg-Eppendorf. PBMCs were isolated using density gradient centrifugation. CD4 + T cells were then isolated using the EasySep Human CD4 + T Cell Isolation Kit (stem cell technologies). CD4 + T cells were cultured in X-VIVO 15 cell medium (Lonza) with 10% FCS, 0,5% Pen/Strep and 200U/mL IL-2 at a density of 1x10^⁶ cells/ml. Upon isolation, cells were activated with CD3/CD28 (ImmunoCult Human CD3/CD28 T Cell Activator, Stemcell).

Generation of HIV_BACH2 reporter lines

An HIV-1 reporter construct was cloned in pMK backbone (Thermo Fischer Scientific) using sequential Gibson assembly (NEBuilder HiFi DNA assembly). 5’LTR, gag leader region and 3’LTR were derived from pNL4–3 (BEI Resources), bfp 2A sequence was amplified from pNLT2DenvBLB [107]. To generate the transfer plasmid for CRISPR/Cas9-mediated knock-in, 5’ and 3’ homology arms of 1000 bp length were amplified from genomic DNA extracted from Jurkat cell line and cloned flanking the HIV reporter. An 8 bp long barcode sequence was introduced upstream the 5’LTR. Guide RNA for CRISPR/Cas9 targeting of the BACH2_3710 locus (5’-ggagattgtgtgaacccagc-3’) was cloned into pX330-U6-Chimeric_BB-cBh-hSpCas9 (Addgene #42230) utilizing BbsI restriction sites (yielding pX330-U6-Chimeric_3710_BACH2-cBh-hSpCas9). E-CRISP tool was used for gRNA design (http://www.e-crisp.org).

CRIPSR/Cas9-mediated insertion of the HIV reporter into BACH2 intron 5 in HAP1 cells was carried out following a workflow previously described [108,109]. In brief, HAP1 cells were co-transfected using Lipofectamine 3000 (Thermo Fisher Scientific) with transfer plasmid and pX330-U6-Chimeric_3710_BACH2-cBh-hSpCas9 plasmid. Transfected cells were plated in limiting dilution for single cell clone generation at a density of 0.5-1 cells/well in flat bottom 96-well format. Screening for targeted clones was carried out in two rounds. First, bfp-expressing clones were selected after induction with 50 ng/ml phorbol 12-myristate 13-acetate (PMA; Sigma Aldrich) and 1 µM Ionomycin (Sigma Aldrich) for 18 hours. Positive clones based on flow cytometry were subsequently screened by PCR on whole cell lysate for presence genomic insertion of the reporter at the targeted locus. To exclude cell clones with an integrated vector backbone, clones were screened by PCR for presence of pMK backbone (negative selection). Site-specific mono- or biallelic BACH2 targeting events were assessed by Southern blotting using bfp-specific and genomic probes as described previously [108]. In brief, high molecular weight genomic DNA from 3710_BACH2_positive clones was extracted from whole cell lysate by isopropanol precipitation, digested with ExoRI and blotted onto Amersham Hybond-N⁺ blotting membrane (GE Healthcare). The corresponding probes were PCR amplified from genomic DNA or targeting plasmid DNA and labelled with ³²dCTP using High Prime DNA labeling (Roche). Membranes were hybridized with labelled probes in hybridization buffer at 65°C over night. After stringent washing, membranes were exposed to Amersham Hyperfilm MP (GE Healthcare).

Generation of HIV_STAT5B reporter line

Clones were generated by limiting dilution following transduction of MOLT-4 cells with pV1/SBP-P2A-GFP [110], a V1-derived minimal reporter virus carrying authentic NL43 proviral 5′ and 3′ sequences, at an MOI of 0.3. Successful proviral integration in each single-cell clone was confirmed by amplification of host–virus junctions using ligation-mediated PCR [48], followed by visualization of PCR products via agarose gel electrophoresis. Integration sites of positive clones were determined by sequencing of LM-PCR products. The HIV_STAT5B single-cell clone was selected based on its integration site in STAT5B.

PMA/Iono and LPA treatment

For PMA/Iono treatment, cells were seeded at 1 x 10⁶ cells/ml and treated with 50 ng/ml phorbol 12-myristate 13-acetate (PMA; Sigma Aldrich) and 1 µM Ionomycin (Sigma Aldrich) or DMSO as a control. Reporter expression was analysed by flow cytometry 16 hours post treatment. The effect of latency promoting agents (LPAs) KRIBB11 and Spironolactone was assayed in HIV_BACH2 cells as follows: 1 x 10⁶ cells/6-well were seeded and simultaneously treated with KRIBB11 (final concentration 5 uM, Absource Diagnostic) or Spironolactone (final concentration 10 uM, Sigma-Aldrich). 18 h post seeding, cells were induced with PMA/Iono. One day post seeding RNAzol samples were collected and reporter [50] expression was analyzed by flow cytometry. Samples labeled as no P/I where neither treated with PMA/Iono nor with an LPA compound.

Western Blot analysis of HIV_BACH2 cells

HIV_BACH2 cells were activated with PMA and Ionomycin for 16h as described before. Cells were lysed in RIPA buffer (150 mM NaCl, 1% Triton X-100, 0.25% NP40, 0.1% SDS, 50 mM Tris HCl pH8.0 in H2O) supplemented with cOmplete protease inhibitor cocktail (Millipore) and protein concentration determined by Pierce 660 reagent according to manufacturer’s instructions (Thermo Fisher). Twenty ug of total protein were separated by SDS-PAGE and transferred to nitrocellulose membranes (Whatman Protran). Primary antibodies against BACH2 (clone D3T3G, Cell Signaling) and LaminA/C as control (sc-376248, Santa Cruz) were used in combination with secondary HRP-conjugated antibodies (Jackson Laboratories) for detection.

SAM-mediated targeted transcriptional activation

SAM-mediated LTR (LTR-SAM) and BACH2 (BACH2-SAM) targeted transcriptional activation of HIV_BACH2 and wt cells was carried out as previously described [34]. In brief, 5*10⁵ cells were seeded in a six-well plate and co-transfected with 1.4 µg sgRNA(MS2)-gRNA for either LTR or BACH2 activation, 0,8 µg dCas9-VP64_GFP (Addgene #61422) and 1 µg MS2-P65-HSF1_GFP (Addgene #61423) per well. Transfection was performed using TransIT LT1 reagent (Mirus Bio) according to the manufacturer’s protocol. Three days post transfection cells were harvested, washed with PBS, sorted for BFP/GFP-positive cells (FACS Aria, BD Biosiences), washed once more and stored in RNAzol (Sigma Aldrich) at -80°C upon further processing. Total RNA was isolated following the RNAzol manufacturer’s protocol. All experiments were conducted in n = 3 independent replicates.

Production of lentivirus and transduction of CD4 + T cells

Lentiviruses were produced by plasmid co-transfection of the following plasmids into LentiX cells (Takara Bio) using TransIT-LT1 (Mirus Bio) according to manufacturer’s instructions: psPAX2 (Addgene #12260), pCMV-VSV-G (Addgene #8454) and lentiviral transfer vectors. Supernatants were collected 72 hours post transfection and filtered through a 0.22 µm filter (Millipore). Virus concentration was conducted by ultracentrifugation at 12,300 × g for 1.5h at 4 °C (Beckman Coulter) and resuspended in X-VIVO 15 medium (Lonza). CD4 + T cells from healthy donors were activated by CD3/CD28 stimulation 24 h prior transduction. Spin-inoculation was used to transduce the cells with lentivirus stocks in X-VIVO 15 medium (Lonza) containing 0.2% protamine sulfate (Sigma) (850 x g, 1 h, 30 °C). Following centrifugation, the plates were incubated for 24 h at 37 °C before medium was replaced by X-VIVO 15 cell with 10% FBS, 0.5% Pen/Strep and 200 U/mL IL-2 (Biomol).

Magnetic bead-mediated enrichment for SBP-LNGFR + cells

5x10⁶ transduced CD4 + T cells were enriched using magnetic selection of the SBP-∆LNGFR surface marker with Dynabeads Biotin Binder (Invitrogen). In brief, cells were washed three times with incubation buffer (0.5 M EDTA, 1 mg/L BSA in PBS). Beads were added at a 10:1 ratio (beads:cells) and the mixture was incubated at 4 °C for 30 min. After incubation, the samples were placed on a magnetic stand for 3 minutes before carefully removing the supernatant. Bead-bound cells were washed with cold incubation buffer, placed into the magnetic stand again for 3 minutes before decanting the supernatant. Bead-bound cells were resuspended in pre-warmed release buffer (1 M HEPES, 20 mM biotin in medium) and incubated at room temperature for 15 min. The tube was placed on the magnetic stand for 3 minutes before collecting the released cells in the supernatant.

Analysis of cell proliferation and co-culture assay

Division index of transduced cells was measured using CellTrace Far Red Cell Proliferation Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. In brief, transduced cell populations and non-transduced control cells (n = 3) were stained at d12 after transduction. Expression of the Far Red marker dye was measured at d12 and at three following time points (d14, d15 and d17 after transduction) by flow cytometry. For growth competition assays, LV-BACH2 and LV-GFP transduced CD4 + T cells of n = 3 healthy donors were enriched for SBP-LNGFR using magnetic beads at day 8 post transduction and mixed at a 1:4 ratio in co-culture. Percentages of LV-BACH2 and LV-GFP cells in co-culture were determined at days 0, 8, 11 and 16 post initiation of co-culture through flow cytometric measurement (LNGRF, GFP).

Quantitative RT PCR assays

RNA was extracted using RNAzol (Sigma Aldrich) according to manufacturer’s instructions. cDNA was generated from 1ug total RNA using SuperScript III Reverse Transcriptase (Invitrogen). Droplet digital PCR with the Bio-Rad QX200 system was used for quantification of i) endogenous BACH2 (amplicon spanning exon 4/5, upstream of region of recurrently-detected proviral integrations in PLWH) and total BACH2 (exon 6, downstream of region of recurrently-detected proviral integrations in PLWH) transcript levels; and ii) endogenous (exon 1 + exon 2), total (exon 10 + exon 11) and chimeric (LTR + exon 2) STAT5B transcript levels (utilized oligonucleotides are provided in S2 Table). Each 20 µL reaction contained 10 µL of 2 × ddPCR Supermix for probes (No dUTP) (Bio-Rad), 2 µL of each primer (10 µM stock), 1 µL probe (5 µM stock), 2 µL of cDNA template, 1µL of RPL13A primer/probe mastermix for normalization (Bio-Rad #10031255) and 2 µL nuclease-free water. The reaction mix was loaded into the QX200 Droplet Generator (Bio-Rad) with 70 µL of Droplet Generation Oil, and droplets were generated according to the manufacturer’s guidelines. PCR amplification was performed using a T100 Thermal Cycler (Bio-Rad) under the following cycling conditions: 95 °C for 10 min, 40 cycles of 94 °C for 30 sec and 57 °C for 1 min, 98 °C for 10 min, hold at 4 °C. Following amplification, droplets were analyzed using the QX200 Droplet Reader (Bio-Rad). Resulting data was visualized using GraphPad Prism v10 software.

RNA sequencing and RNAseq dataset analysis

RNA sequencing was carried out on total RNA, isolated from SAM-treated HAP1_BACH2_LTRbfp2Atat cells and LV-transduced CD4 + T lymphocytes. For HAP1 samples, 1 ug of RNA was subjected to library preparation utilizing the NEBNext Ultra II Directional RNA mRNA kit in combination with NEBNext Multiplex Oligos Unique Dual Index UMI Adaptors and sequenced at ~100M clusters 2x100bp per sample (Illumina NextSeq500). For primary T lymphocytes, about 5 million transduced CD4 + T cells were enriched using magnetic selection based on SBP-∆LNGFR surface marker expression (see above). Total RNA was isolated from selected primary CD4 T cells using RNAzol (Sigma Aldrich) and following manufacturer’s instructions. Isolated RNA was subjected to library preparation using CORALL mRNA-Seq Library Prep Kit (Lexogen) and subsequent sequenced at ~20M reads, 2x50bp per sample (Illumina NextSeq500).

Sequence reads were mapped using HISAT2 (HAP1 RNAseq) or Salmon (CD4 T lymphocyte RNAseq) aligners against the human reference genome (hg38) after adapter trimming and quality filtering using TrimGalore [111–113]. Gene count matrices were further analysed in R in terms of differential gene expression (DEG) and gene set enrichment analysis (GSEA) utilizing packages DESeq2 (v1.44.0), EnhancedVolcano (v1.22.0), pheatmap (v1.0.12), ClusterProfiler (v4.12.6) and ggplot2 [114–118]. In CD4 + T lymphocyte over expression experiments contrasting LV-GFP vs LV-BACH2 samples, ‘donor_ID’ was considered as covariate. In PCA plots this was achieved utilizing DESeq2 vst transformation of gene counts and limma::removeBatchEffect() (v3.60.3) for ‘donor_ID’ [116,119]. Treeplots of significant gene ontology (GO) terms were generated by hierarchical clustering based on Jaccard’s similarity index using ClusterProfiler v(4.12.6) [117]. Analysis of enrichment of functional gene networks using MCL clustering (inflation parameter 3) was done using STRING (v12.0) [120].

Single BACH2 exon analysis in the HAP1 reporters was conducted with a filtered genome annotation file (GTF) for BACH2 reference transcript ENST00000257749 containing 9 exons. HTSeq function htseq_count (v2.0.5) with parameters -m union -i exon_number -t exon was used to get read counts for individual BACH2 exons [121]. Downstream analysis and plotting in R utilizing packages ggplot2 (v3.5.1) and ggpubr (v0.6.0) [118,122]. Statistical method: unpaired t-test.

HIV-to-host transcript splicing in HAP1 reporters was analysed by inserting the HIV reporter sequence into the BACH2 locus (position chr6:90012553–90015110) of the reference genome version GRCh38. Genome annotations (GTF) were accordingly updated using reform [123]. Subsequently raw FASTQ files were aligned to the updated reference genome and annotation using STAR [124]. Splicing pattern plots were generated using the modified GTF annotation file and the sashimi_plot command of miso with scaling parameters for intron and exon at 50 and 3 respectively [125,126].

Cytokine profiling using LEGENDplex assay

CD4 + T cells transduced with lentiviral vectors encoding BACH2, STAT5B, or GFP control were enriched for transduced, SBP-LNGFR + cells using magnetic beads (see above) at day 7 post transduction and supernatants were collected at day 24 (LV-BACH2 and LV-GFP controls) and day 18 (LV-STAT5B and LV-GFP controls), respectively, for quantification of secreted cytokines using the LEGENDplex Human T-Helper Cytokine Panel Version 2 (BioLegend). Sample processing and data acquisition via flow cytometry was done according to the manufacturer’s instructions Data was processed using the LEGENDplex analysis software. Cytokine levels for LV-BACH2- or LV-STAT5B cells were normalized to those of LV-GFP control cells.

BACH2 binding site and chromatin landscape analysis

The datasets supporting the conclusions of this analysis are publicly available and can be accessed in NCBI repositories under the accession numbers GSE102460 [BACH2 ChIP-seq] and GSE337423 and GSE337424 [H3 ChIP-seq & ATAC-seq]. To identify putative BACH2 binding sites across the human genome, the BACH2 motif was obtained from the JASPAR database (2024 expansion, ID: MA1101) [127]. The hg38 human genome assembly was scanned for BACH2 motifs using FIMO, a tool within the MEME suite (v5.5.5) [128]. FIMO was run using the MEME suite’s precomputed hg38 background model based on the genome’s nucleotide frequencies. Significant BACH2 motif hits were defined as those with p < 1e-4. Raw ChIP-seq and ATAC-seq data were first assessed for quality using FastQC (v0.12.1) (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads were trimmed with Trimmomatic (v0.39) to remove adapter sequences and low-quality bases [129]. Next, trimmed reads were aligned to the hg38 human reference genome using Bowtie2 (v2.5.1) with the --very-sensitive parameter [130]. For ATAC-seq, a maximum fragment size of 1000 bp was permitted to allow for the capture of multi-nucleosomal fragments. All unmapped reads and reads with a mapping quality score below 10 were excluded from further analysis. PCR duplicates were removed using Picard MarkDuplicates (v2.26.10) (http://broadinstitute.github.io/picard). ATAC-seq reads were corrected for Tn5 transposase insertion bias using the alignmentSieve tool within deepTools (v3.5.5) with the --ATACshift option [131]. MACS2 (v2.2.9.1) was used for peak calling with parameters tailored for each assay – broad peak calling for H3K27ac ChIP to capture large acetylation domains, while narrow peak calling was used for BACH2 ChIP, H3K4me3 ChIP, and ATAC-seq open chromatin regions [132]. Peak annotation to hg38 genes was performed using Homer (v4.11) (https://homer.ucsd.edu/). For H3K4me3 ChIP-seq and ATAC-seq, individual samples were integrated post-mapping using samtools (v1.9) [133]. CPM-normalized bigwig tracks were generated for both individual and merged BAM files using the bamCoverage function in deepTools (v3.5.5) with a bin size of 10 bp [131,134]. These tracks were then used to visualize the BACH2 locus in JBrowse2 (v4.3.0) [135].

Chromatin Immunoprecipitation (ChIP), ChIP-qPCR and ChIP-seq

Chromatin immunoprecipitation for V5-tagged BACH2 was performed utilizing dual crosslinking with 2mM disuccinimidyl glutarate (DSG) and 1% formaldehyde as described previously with minor adaptations to CD4 + T cells [136]. In brief, sheared chromatin of 4x10^6 CD4 + T cells was used per ChIP utilizing 4 ug anti-V5 (TCM5, Invitrogen) antibody. Following DNA capture with magnetic beads, de-crosslinking and proteinase K digest steps, DNA was purified utilizing the DNA ChIP DNA Clean & Concentrator kit (Zymo) according to manufacturer’s instructions. BACH2 binding was assayed by qPCR at indicated target sites and plotted as percent of input values. Primer sets for each target region are provided in S2 Table. A qPCR reaction consisted of 5 µl PowerUp SYBR Green Master Mix (2X, Applied Biosystems), 0.4 µl of forward and reverse primer each (10 µM stock) in a final volume of 10 µl. The PCR program was set to 50 °C for 2 minutes, 95 °C for 2 minutes and subsequently 40 cycles of 95 °C for 15 seconds and 60 °C for 45 seconds and run on a 7500 Fast qPCR cycler (Applied Biosystems).

ChIP-seq for histone modifications has been described previously [137] and was conducted with minor adaptations. Briefly, chromatin from 1x10^6 cross-linked cells was sonicated to 200–500 bp fragments (Bioruptor, Diagenode) and immuno-precipitated overnight using 4 µl of anti-H3K4me3 (Merck Millipore, #04–745) or anti-H3K27ac (Abcam, #ab4729) antibodies. Following stringent washing, chromatin was eluted from magnetic beads at 65 °C for 30 min and DNA was de-crosslinked in presence of 200 mM NaCl at 65 °C overnight. DNA was purified utilizing the DNA ChIP DNA Clean & Concentrator kit (Zymo) according to manufacturer’s instructions. ChIP-seq libraries were prepared with the NEBNext Ultra II DNA Kit (NEB) and sequenced on an Illumina NextSeq 2000 platform (75 bp paired-end).

Assay for transposase-accessible chromatin using sequencing (ATAC-seq)

ATAC-seq was performed following the Omni-ATAC-seq protocol as described before with minor adaptation [137,138]. In brief, 5x10^4 CD4 + T cells were DNaseI-treated (200 U/ml, Worthington) at 37 °C for 30 min and subsequently washed 2x with cold PBS. Cell pellets were transposed with 50 µl of transposition mix containing 25 µl 2x TD Buffer (20 mM Tris-HCl pH 7.6, 10 mM MgCl₂, 20% Dimethyl Formamide), 2 µl transposase (custom made, 100 nM final), 16.5 µl PBS, 0.5 µL 1% digitonin and 6.5 µl H2O at 37 °C and 1000 rpm on a thermo block for 30 min. The reaction was stopped by adding 250 μl DNA Binding Buffer and DNA was isolated using the Clean and Concentrator-5 kit (Zymo) following manufacturers instructions. Libraries were produced by PCR amplification of tagmented DNA and sequenced 2x150bp on a NextSeq 2000 sequencer (Illumina).

Luciferase reporter assay

To evaluate the functional impact of BACH2 binding sites on proximal transcriptional activity a Gaussia luciferase (GLuc) reporter vector was used as previously described [88]. A 120 bp fragment containing BACH2 BS 1 (BS1-GLuc) or a selected negative control (NC) region of equal length from the BACH2 intron 1 locus (NC-GLuc), were cloned upstream of the minimal TK promoter using Acc65I and XhoI restriction enzymes (oligonucleotide sequences provided in S2 Table). The GLuc assay was performed in HEK293T cells, which were seeded at 50,000 cells per well in 12-well plates one day prior to transfection. Using TransIT-LT1 reagent (Mirus Bio) according to the manufacturer’s instructions, the GLuc, BS1-GLuc or NC-GLuc was co-transfected with either a BACH2.2A.mCherry overexpression vector or a mCherry-only control vector (0.625 µg of each plasmid, 1.25 µg total DNA per well). Three days post-transfection, cells were trypsinized, counted, and analyzed for mCherry expression by flow cytometry (Fortessa, BD Biosciences). Gaussia luciferase activity was subsequently measured via chemiluminescence readout (Infinite 200 Pro, Tecan Life Sciences). Raw GLuc signals were normalized to the number of viable transfected cells.

Graphics

Schematic graphics were self-drawn or created in Biorender (www.biorender.com). Brauckmann, L. (2026) https://BioRender.com/1vu1fc4 (Fig 3a), Brauckmann, L. (2026) https://BioRender.com/s96zj7i (Fig 3b).

Data sharing