Toll-like receptor 3 activation selectively reverses HIV latency in microglial cells

Multiple toll-like receptors (TLRs) are expressed in cells of the monocytic lineage, including microglia, which constitute the major reservoir for human immunodeficiency virus (HIV) infection in the brain. We hypothesized that TLR receptor mediated responses to inflammatory conditions by microglial cells in the central nervous system (CNS) are able to induce latent HIV proviruses, and contribute to the etiology of HIV-associated neurocognitive disorders. Newly developed human microglial cell lines (hµglia), obtained by immortalizing human primary microglia with simian virus-40 (SV40) large T antigen and the human telomerase reverse transcriptase, were used to generate latently infected cells using a single-round HIV virus carrying a green fluorescence protein reporter (hµglia/HIV, clones HC01 and HC69). Treatment of these cells with a panel of TLR ligands showed surprisingly that two potent TLR3 agonists, poly (I:C) and bacterial ribosomal RNA potently reactivated HIV in hμglia/HIV cells. LPS (TLR4 agonist), flagellin (TLR5 agonist), and FSL-1 (TLR6 agonist) reactivated HIV to a lesser extent, while Pam3CSK4 (TLR2/1 agonist) and HKLM (TLR2 agonist) only weakly reversed HIV latency in these cells. While agonists for TLR2/1, 4, 5 and 6 reactivated HIV through transient NF-κB induction, poly (I:C), the TLR3 agonist, did not activate NF-κB, and instead induced the virus by a previously unreported mechanism mediated by IRF3. The selective induction of IRF3 by poly (I:C) was confirmed by chromatin immunoprecipitation (ChIP) analysis. In comparison, in latently infected rat-derived microglial cells (hT-CHME-5/HIV, clone HC14), poly (I:C), LPS and flagellin were only partially active. The TLR response profile in human microglial cells is also distinct from that shown by latently infected monocyte cell lines (THP-1/HIV, clone HA3, U937/HIV, clone HUC5, and SC/HIV, clone HSCC4), where TLR2/1, 4, 5, 6 or 8, but not for TLR3, 7 or 9, reactivated HIV. TLR signaling, in particular TLR3 activation, can efficiently reactivate HIV transcription in infected microglia, but not in monocytes or T cells. The unique response profile of microglial cells to TLR3 is fundamental to understanding how the virus responds to continuous microbial exposure, especially during inflammatory episodes, that characterizes HIV infection in the CNS.


Background
Although highly-active anti-retroviral therapies (HAART) can reduce circulating virus to below the levels of detection, these regimens are unable to eliminate residual viral infections due to the creation of long-lived reservoirs of latently infected cells [1]. The best characterized of these reservoirs are the resting memory CD4 + T cells found in the peripheral circulation [2,3]. However, additional cell types, including peripheral blood monocytes, dendritic cells and macrophages in the lymph node, and astrocytes, perivascular macrophages and microglial cells in the brain, can also be infected with HIV and can potentially contribute to viral persistence [4][5][6][7][8].
Unlike T cells, cells of the monocyte-macrophage lineage are partially resistant to HIV infection due to the activity of the SAMHD1 restriction factor [9]. Less than 1% of the HIV-1 DNA in the peripheral circulation is found in circulating monocytes. Nonetheless, there is increasing evidence that myeloid cells facilitate the dissemination of the virus and contribute to the persistence of viral reservoirs in the CNS [10]. It seems likely that perivascular macrophages initiate brain infection, which then spreads to resident microglial cell populations [10][11][12]. Analysis of the sequence and phenotypes of viruses recovered from the CNS demonstrates that HIV in HAND patients represents a distinct, macrophage tropic virus population [13][14][15]. It has been speculated that once infected, monocyte-macrophage lineage cells are more resistant to certain anti-retroviral drugs and the cytopathic and apoptotic effects of HIV than T cells, and therefore can harbor actively replicating viruses for longer periods even under conditions of effective suppression by HAART. Detection of viral escape mutations in the cerebrospinal fluid of individuals under HAART with undetectable viral presence in blood supports this hypothesis [16][17][18].
One of the main drivers of HIV infections, and the reason why there is chronic inflammation during HIV disease, is that damage to the gastrointestinal track leads to release of microbial products into the circulation [19]. These potent inducers of inflammatory responses are recognized by pattern recognition receptors, including the ten TLRs (TLR1-10), which are responsible for the innate recognition of viruses, bacteria, fungi, and parasites [20]. Monocytic cells characteristically express most, if not all, of the TLR family members, while T cells can only express a more limited set of these pattern recognition receptors. In response to the presence of pathogens invading the CNS, microglial cells activate their TLRs and initiate CNS innate and adaptive immune responses [21]. Excessive and/or chronic TLR activation of microglial cells in the brain [22] is believed to be directly responsible for a number of CNS diseases, including chronic HIV encephalitis, and the mounting pro-inflammatory reactions in response to HIV infections that lead to neurocognitive disorders [23,24].
The signaling pathways following TLR activation are complex, and involve a series of protein adaptors (MyD88, TIRAP, TRIF, and TRAM) and downstream effectors (IRAKs, TRAF6, and TRAF3). In general, TLR signaling cascades derived from receptors 2/1, 4, 5, 6, and 8 eventually lead to the activation of the IKK complex and the translocation of the transcription factors NF-κB and AP-1 to the nucleus. Alternative pathways to this central signaling mechanism result in the production of type I interferons (IFNs) and IFN-inducible genes by activation of IRF7 (TLR7 and 9) and IRF3 (TLR3) [25]. Since reactivation of latent HIV proviruses can be potently stimulated by NF-κB in many cell types [26][27][28][29], TLR signaling can potentially lead to enhanced viral replication in cells that carry the appropriate receptors.
Even though TLR pathways have been relatively well characterized, investigations of TLR activation on HIV expression in various target cell types have been scarce, and largely limited to studies of transformed monocytic and T cell lines. Pomerantz et al. [30] initially reported that LPS, a bacterial TLR4 ligand, potently stimulated HIV-1 long terminal repeats (LTR) CAT constructs transfected into monocyte/macrophage-like cell lines, but not a T cell line. Subsequently, LPS-mediated activation in macrophages was linked to the activity of the PU.1 and NF-κB transcription factors [31][32][33], and the direct participation of TLR4 in mediating LPS-induced NF-κB and HIV-LTR activation was established [32,33]. Later studies using the monocytic THP-1 cells showed that TLR2 ligands can also stimulate HIV transcription [33].
An inconsistent set of results has been obtained for the other TLR ligands. TLR9 signaling results in increased HIV expression in monocytes obtained from transgenic mouse spleen cells [33]. However, in latently infected ACH-2 T cells, a TLR9 agonist, ODN2006, strongly activated HIV transcription, but inhibited productive HIV infection in MT4 T cells and primary T cells [34]. In a model of latently infected mature mast cells, stimulation of TLR2, 4 or 9 triggered HIV-1 replication [35]. Thibault et al. [36] reported that TLR5 stimulation is sufficient to trigger reactivation of latent HIV-1 provirus in Jurkat T cells and to also activate viral gene expression in central memory CD4 + T cells. Consistent with these results, Brichacek et al. [37] found that activation of TLR5, but not TLR9, triggers HIV-1 transcription in lymphoid tissue ex vivo. The differential effects of these TLR ligands on HIV-1 replication correlated with changes in production of CC and CXC chemokines in the ligand-treated HIV-1-infected tissues [37]. More recently, Novis et al. reported that Pam3CSK4 (TLR2/1 agonist) leads to viral reactivation from latency in cultured central memory T (T CM ) cells, suggesting a unique pattern of TLR-mediated HIV reactivation in primary T cells [38]. It has been known that TLR3 specifically recognizes bacterial ribosomal RNA [39], but its role in HIV transcriptional regulation in known.
In this study we investigated the effects of a wide range of TLR ligands, including potent TLR2 agonist molecules purified from Mycobacterium tuberculosis (Mtb), on the reactivation of pro-viral HIV in a novel human model of latently-infected microglial cells. Surprisingly, we found that TLR3 ligation potently reactivated HIV in microglial cells, but not in monocytes or T cells, using an NF-κBindependent pathway that involves activation of the IRF3 transcription factor. Additionally, HIV was modestly reactivated in human microglial cells by engagement of TLRs 4, 5, and 6. More restricted reactivation patterns were observed after stimulation of TLRs 1 and 2, whereas no significant reactivation was obtained by activation of TLRs 7,8,or 9. The unique responses of microglial cells to TLR ligands provide an important framework for understanding how circulating bacterial antigens and nucleic acids can enhance HIV replication in microglial cells and potentially exacerbate HAND.

Characterization of hµglia/HIV (HC01) and (HC69) cells
The HIV latently-infected cells (hµglia/HIV) used in this study were derived from simian virus 40 large T antigen (SV40Tag)/human telomerase reverse transcriptase (hTERT)-immortalized human microglial cell lines prepared as described by Garcia-Mesa et al. [40]. These novel cell lines, which have characteristic microglial cell phenotypes, have superseded CHME-5 cells in our laboratory. CHME-5 cells were originally believed to be of human origin [41], and have been widely used as models for human microglial cells even until very recently [42,43]. Unfortunately, our detailed molecular analyses of CHME-5 proved that they are actually rat cells, and were probably derived from a contaminating rat glioma [40].
To prepare latently infected cells, we followed a strategy used successfully to create latently infected Jurkat T cell lines [44,45] and CHME-5 cell lines [46]. Briefly, as depicted in Fig. 1a, and as extensively explained in Garcia-Mesa et al. [40], commercially-available primary microglial cells were immortalized through infection with vesicular stomatitis virus G (VSV-G) envelope pseudotyped viruses expressing SV40Tag and hTERT. Immortalized cells (hµglia) were either used to isolate clonal populations, such as C20 and C06 [40] or infected with a VSV-G pseudotyped lentivirus vector (PHR1′/d2EGFP) expressing the shortlived green fluorescence protein (d2EGFP) as a reporter construct (Fig. 2a) to develop clonal populations of hµglia/HIV, such as HC01 and HC69. In general, cells expressing GFP then were selected by cell sorting. In approximately 3-4 weeks, latently infected (GFP − ) cells begin to outgrow the GFP + cell population. The two clones hµglia/HIV (HC69) and hµglia/HIV (HC01), carrying latent proviruses, were then isolated from the GFP − population.
The proviral integration site for each of these two clones has been sequenced and located within the host genome (Table 1). Both HC01 and HC69 are single integrants and, as expected from the extensive studies characterizing HIV proviral integration sites [44,[47][48][49], the provirus was located in the introns of host genes. The uninfected cell line C20 was used as a negative control, where no HIV sequence was detected.
As shown in Fig. 1b, hµglia/HIV cells, exemplified by the clone HC01, express the well-established microglial surface markers CD11b and P2RY12 [50]. The cells also express the macrophage lineage marker CD14, suggesting that they display an activated phenotype, which is consistent with RNA-seq analyses of the hµglia (C20) cells [40].

Induction of HIV expression in hµglia/HIV cells
The presence of latent HIV proviruses (Fig. 2a) in individual hµglia/HIV clones was confirmed by viral reactivation in the overnight presence of 500 pg/mL of tissue necrosis factor α (TNF-α) or 30 µM of the histone deacetylase inhibitor 4b (HDACi 4b) [51]. Induction of GFP was monitored by immunofluorescence ( Fig. 2b) as well as flow cytometry results (Fig. 2c). In these cells, basal HIV expression was extremely low, with 1-6% of the cells expressing GFP. However, after exposure to TNF-α for 16 h, HIV was induced in approximately 90% of the HC69 cells, or 25% of the HC01 cells. Similarly, exposure to HDACi 4b resulted in a strong induction of HIV in nearly 81% of the HC69 cells, or 61% of the HC01 cells. In general, HC01 cells displayed a somewhat more restricted HIV reactivation profile than HC69 cells (Fig. 2b, c).

TLR-mediated HIV reactivation of latently-infected human microglial
Microglial cells are known to express all TLRs at low, but detectable, levels [22]. Because engagement of TLRs can lead to NF-κB pathway activation, and in turn, HIV expression, we assumed that treatment of latentlyinfected microglial cells with selective TLR agonists would lead to viral reactivation provided there was sufficient receptor expression. However, as described below, microglial cells show specific restrictions in their responses to TLR agonists.

Microglial cells respond poorly to potent TLR2 agonists derived from Mtb
Because of the poor responses of microglial cells to classical TLR2/1 ligands, we decided to also test a set of highly potent TLR2 agonists derived from Mtb. Mtbderived lipoproteins such as LprG [52] as well as the glycolipid lipomannan (LM; [53][54][55]), but not mannosylated lipoarabinomannan (ManLAM; [53]), are potent TLR2 agonists that trigger pro-inflammatory and microbicidal innate immune responses. In addition, the Mtb-derived polar glycolipid phosphatidylinositol mannoside (PIM) 6 (PIM 6 ), but not PIM 1,2 , triggers TLR2 signaling [56]. containing Tat, Rev, Env, Vpu and Nef with the reporting gene d2EGFP, is cloned into the pHR' backbone. The resulted plasmid was used to produce the VSVG HIV particles, as described previously [112]. b Fluorescence microscopy analysis of TNF-α-and HDACi 4b-mediated reactivation of HIV in latently-infected microglial cells [hµglia/HIV (HC69) and (HC01)]. Cells treated with TNF-α (500 pg/mL) or HDACi 4b (30 µM). c FACS analysis 16 h post-treatment. In these, and subsequent FACS profiles, GFP + cell populations are indicated in green, and the proportion of GFP-expressing cells is indicated in % In contrast to microglial cells, HIV was strongly reactivated by PIM 6

Bacterial rRNA, like poly (I:C) agonists, reactivate HIV in hμglia/HIV cells
It is well established that TLR3 is able to response to dsRNA, particularly dsRNA of viral origin. The effect on TLR3 activation by rRNA from bacteria is less documented, although some evidence indicates that dsRNA isolated from bacterial cultures engages TLR3 signaling [39]. Here, we show that rRNA purified from bacteria ( Fig. 4a) reactivated HIV in HC69 cells up to ~50%, which is comparable to the reactivation obtained with poly (I:C) (both LMW and HMW; ~60-65%) (Fig. 4b).
MNase digestion of bacterial rRNA at these concentrations resulted in accumulation of small dsRNA fragments (Fig. 4a), which still reactivate HIV. By contrast the poly (I:C) was completely digested (Fig. 4a) and could not reactive HIV (Fig. 4b).

TLR expression on microglial/HIV and monocytic cells does not correlate with HIV reactivation
As described above, the HIV-latently infected monocytic cell lines responded more robustly to TLR stimulation than microglial cells, with the exception of TLR3 activation by poly (I:C) (Figs. 3, 5; Additional file 3: Fig.  S3). Surprisingly, this did not strictly correlate with TLR expression patterns.
Despite the relatively weak responses to LPS (Figs. 3a, 5a), almost the entire microglial cell population was positive for TLR4 expression (Fig. 6b). One important observation when comparing the LPS-responses of the three monocytic cell models was that the THP-1 response was stronger than U937, which in turn was stronger than SC. THP-1 cells were ~98% positive for TLR4 (~96% reactivated HIV by LPS), whereas only ~62% of U937 cells (~53% reactivated HIV by LPS) and ~25% of SC cells (~22% reactivated HIV by LPS) expressed the receptor (Additional file 4: Fig. S4), paralleling the hierarchy of responses of these cell types to LPS (Fig. 5b).
hµglia/HIV cells are positive for CD14, as exemplified by the HC01 staining (Fig. 1b). CD14 is the TLR4 co-receptor needed for LPS-mediated activation, ruling out the possibility that the relatively weak TLR4-mediated HIV transcriptional response was due to the absence of CD14. Therefore, the expression profiles of TLR4 on hµglia/HIV cells did not correlate with the more restricted HIV reactivation induced by LPS compared to monocytes.
A similar situation exists for the TLR5 receptor. A greater proportion of the microglial line expressed TLR5 (~32-53%) compared to the monocytic line (~12%; Fig. 6b), but the monocytes, nonetheless, responded more robustly to flagellin than the microglial cells.
Both cell types also expressed relatively high levels of TLR7 (Fig. 6b), but neither cell type responded to imiquimod (Fig. 5a, b). On the other hand, the high level of TLR8 expression on the monocytes (~82%) relative to the microglia (~9-45%) (Fig. 6b) can explain the higher response of HIV reactivation on monocytes by ssRNA40 (Fig. 5b). Finally, the differential TLR9 expression observed between the two microglial cell lines (~1 vs. ~95%), and compared with the monocytes (~87%) (Fig. 6b), does not explain that ODN2009 was not able to reactivate HIV in either of these cell lines (Fig. 5a, b). and compared it to that on infected cells (HC69.1 and HC01) as well as monocytic and T cells (Jurkat) without and with HIV. Cells were serum-starved to reduce the overall levels of TLR expression, since they were at maximal levels under normal growing conditions (1-5% FBS). Poly (I:C) treatment increased the expression of TLR3 (from ~40 to 98%) and TLR4 (from ~65 to 97%), decreased the expression of TLR7 (from ~70 to 40%), and did not affect the expression of TLR5 (Fig. 6c).
In complementary control experiments, the reporter THP1-XBlue ™ cells (Invitrogen), which carry an embryonic alkaline phosphatase under the control of a synthetic NF-κB-inducible promoter, were used to measure NF-κB induction by TLR agonists. The results (Additional file 6: Fig. S6a) confirmed that only the TLR agonists that reactivated HIV in THP-1/HIV (HA3) cells induced NF-κB activation in a time-dependent manner in THP1-XBlue reporter cells; this reactivation was blocked in the presence of DRB and flavopiridol, inhibitors of P-TEFb (Additional file 6: Fig. S6b; please, read further below). Consistently, Western blot analysis of nuclear p65 levels in THP-1/ HIV (HA3) cells treated for 30 min with TNF-α or indicated TLR agonists demonstrated that the TLR agonists that reactivated HIV were those that were able to induce NF-κB nuclear recruitment (Additional file 6: Fig. S6c).

Induction of IRF3 in microglial cells by TLR3 ligands
Surprisingly poly (I:C), which reactivated HIV in hµglia/ HIV, but not in THP-1/HIV cells, did not induce p65 nuclear translocation in any of these cell lines (Fig. 7), indicating that an NF-κB-independent pathway is involved in TLR3-mediated HIV reactivation in microglial cells. We therefore tested whether poly (I:C) was capable of inducing IRF3, which could potentially trigger HIV reactivation either directly or indirectly. As shown in Fig. 8a, poly (I:C) induced IRF3 nuclear recruitment in hµglia/HIV (HC69) cells, but not in THP-1/HIV (HA3) cells. In contrast, LPS did not induce IRF3 nuclear translocation in either hµglia/HIV (HC69) or THP-1/HIV (HA3) cells (Fig. 8a).
To confirm the involvement of IRF3 in HIV induction in hµglia/HIV cells, we performed studies with the IRF3 modulator bufalin. Bufalin is a cardiotonic steroid that has been found to potently prevent IRF3 dimerization and nuclear localization [58]. We also tested the PI3kinase inhibitors LY294002 and wortmannin, since both of these drugs can have indirect effects on IRF3 signaling. For example, LY294002 was found to inhibit LPS-and poly (I:C)-mediated IFN-β transcription and secretion, as well as IRF3 transcriptional activation and binding to the IFN-β promoter in fibroblasts [59]. By contrast, wortmannin did not inhibit IFN-β production in these studies [59]. Treatment of monocyte-derived dendritic cells (DC) with wortmannin or LY294002 enhanced IFN-β expression upon TLR3 or TLR4 engagement [60]. In the same study, it was reported that wortmannin-treated DC cells exhibited enhanced levels of IKK-α/β phosphorylation and IκB-α degradation with a concomitant increase in NF-κB nuclear translocation, as well as enhanced NF-κB activity induced by TRIF overexpression in HEK 293T cells [60].
As shown in Fig. 8a (right), the nuclear recruitment of IRF3 induced by poly (I:C) was blocked by bufalin. Bufalin also progressively inhibited HIV reactivation in response to poly (I:C) in HC69 cells (Fig. 8b). This effect can be seen when either the drug titrated and poly (I:C) and LPS are used at fixed concentrations (Fig. 8b, left-hand graphs) or when the inducers are titrated and bufalin was used at a concentration of 25 nM (Fig. 8b, right-hand graphs). As expected, there was no significant inhibition of LPS-mediated viral reactivation by bufalin (Fig. 8b). Treatment of the cells with the negative control, wortmannin, slightly enhanced responses to poly (I:C), but had no effect on LPS-mediated HIV reactivation (Fig. 8b). Similarly, LY294002 did not significantly inhibit or enhance HIV reactivation after poly (I:C) or LPS treatment (Fig. 8b).

IRF3 is recruited to the HIV promoter upon activation with poly (I:C)
In light of the involvement of IRF3 in regulating poly (I:C)-mediated HIV reactivation in microglial cells, we performed chromatin-immunoprecipitation (ChIP) analysis on HC69 cells untreated or treated with either TNF-α (positive control), poly (I:C) or LPS for 30 min (Fig. 9). To obtain objective information about the distribution of the transcription factors on the HIV proviral DNA we utilized a new ChIP-Seq protocol in which we selected for HIV sequences by hybridization prior to sequencing. The reads were then mapped to the HIV LTR.
Isolated crosslinked DNA-protein samples were subjected to immunoprecipitation with anti-IRF3 and anti-NF-κB p65 antibodies. Anti-RNAP II pSer5 antibody was used as control. There was minimal RNAP II pSer5 detected on the HIV LTR in the latently infected cells. As expected, RNAP II (Fig. 9b) was recruited to HIV promoter in response to TNF-α, poly (I:C) or LPS. The RNAP II accumulated at the promoter and promoter proximal sites downstream of TAR, which mirrors the distribution patterns seen in induced T cells [61].
Upon treatment with poly (I:C) (Fig. 9d), IRF3 is recruited to the HIV promoter region (Fig. 9a). IRF3 accumulated preferentially in the −100 to +200 region, suggesting the presence of a binding site in this region. By contrast, NF-κB p65 (Fig. 9c) was practically absence at the HIV promoter in the presence of poly (I:C).
Treatment with LPS yielded opposite results: the abundance of NF-κB p65 at the HIV promoter, especially in the −100 to +1 region (Fig. 9c), which overlaps the two tandem NF-κB binding site in the HIV enhancer, was strongly increased after LPS treatment. There was also some accumulation NF-κB p65 in downstream regions, consistent with previous reports of transcription factor binding in the promoter proximal region in myeloid lineage cells [62,63]. As expected from our Western blot analyses (Figs. 7, 8), LPS treatment did not lead to IRF-3 recruitment.
Treatment with TNF-α strongly induced NF-κB p65 recruitment (Fig. 9c) to the HIV promoter region but recruitment of IRF3 by TNF-α was considerably weaker than by poly (I:C) (Fig. 9d).

P-TEFb components are constitutively expressed in microglial cells
Tat-dependent HIV expression is strictly dependent upon P-TEFb [27,64]. Since CycT1, a subunit of P-TEFb, is known to be up-regulated during early monocyte differentiation, and then subsequently down-regulated in mature macrophages [65], we were concerned that poor responses to TLR agonists could be associated with limiting P-TEFb levels in microglial cells. There could also potentially be upregulation of CycT1 levels in response to TLR activation.
We measured CycT1 levels in hµglia/HIV HC69 and HC01 cells, before and after TLR activation. As a control, we measured CDK9 levels, the second P-TEFb subunit, whose expression is maintained at high levels during and after macrophage differentiation [65]. Western blot analysis of both cell lines (Fig. 10a) untreated or treated with TLR ligands [Pam3CSK4, HKLM, poly (I:C), LPS, and flagellin] demonstrated that both CycT1 and CDK9 were stably and constitutively expressed in these cells. In HC69 cells, a slight induction of CycT1 was observed (Fig. 10b), especially with Pam3CSK4 (from ~0.4 to 1 arbitrary units), but in general, not significant augmentation of P-TEFb by the tested TLR agonists was detected. As expected, HIV reactivation in hµglia/HIV (HC01) and (HC69) and hT-CHME-5/HIV (HC03) and (HC14) cells was impaired by the P-TEFb inhibitors DRB or flavopiridol (Additional file 7: Fig. S7b).
Because hµglia/HIV (HC01) cells showed a more restrictive phenotype than hµglia/HIV (HC69) cells (Fig. 2), we wanted to determine whether the combination of pro-inflammatory stimuli with TLR ligands could potentiate HIV reactivation. Surprisingly, none of the TLR ligands tested, poly (I:C) and imiquimod (Fig. 11a,  b), or Pam3CSK4 and LPS (Fig. 11c) showed additive effects with TNF-α or IL-1β. However, additive effects were seen using HDACi 4b in combination with poly (I:C) and, to a lesser extent, LPS (Fig. 11). By contrast, the negative control, imiquimod, did not increase the

HIV can establish latency in a wide variety of myeloid cell types
In order to understand HIV expression regulation in the context of pro-inflammatory conditions occurring in CNS, we have developed a number of ex vivo models of HIV latency in microglial cells and, for comparison, in monocytic cell lines. Taking advantage of lentiviral reporters, we developed these cellular models representing three different cell types: hµglia/HIV (HC69) and (HC01) (Fig. 5; Additional file 3: Fig. S3) (human microglia); CHME-5/HIV (H1F3) (Additional file 5: Fig. S5), hT-CHME-5/HIV (HC03) and (HC14) (Additional file 7: In previous studies of HIV latency in microglial cells, we employed the CHME-5 cells [46], which were originally believed to be of human origin [73]. Unfortunately, CHME-5 cells are actually rat cells [40]. To study latency in bona fide human microglial cell lines, we have developed microglial cell lines from human primary glial cultures using SV40Tag or SV40Tag/hTERT-immortalization (hµglia; Fig. 1a, and [40]). Subsequently, we used the hµglia cells to create HIV-latently infected microglial cells (hµglia/HIV). The hµglia cell lines demonstrate a microglial phenotype and display surface markers and functional properties of primary microglia [40]. In particular, even after infection with HIV, these cells continue to express CD11b and P2RY12 [50] (Fig. 1b).
One important observation is that these microglial cell lines showed high level expression of CD14, which should be low in resting microglia, suggesting that the cells have an activated phenotype [72]. We speculate that one unexpected consequence of culturing microglial ex vivo is that it induces and permits sustained expression of CD14 even in the absent of classical stimuli such as TNF-α or IL-1β.

HIV transcriptional control in microglial cells
The HIV promoter is an auto-regulated promoter, which requires the Tat protein to stimulate efficient transcriptional elongation. As a result of this feedback mechanism, HIV silencing occurs whenever Tat availability falls below certain threshold levels needed to sustain transcription [27,[74][75][76]. The decline in Tat levels is typically the result of multiple complementary inhibitory mechanisms. First, in latently infected cells, the transcription initiation   factor NF-κB (or in T cells, NFAT) is sequestered in the cytoplasm. Second, the LTR acquires heterochromatic structures that block the transcription start site [44,77]. Typically, latent proviruses accumulate high levels of histone deacetylases and deactylated histones [78][79][80], and methylated histones [44,[81][82][83]. Finally, in quiescent T cells and monocytic precursors, but not in microglial cells, the essential Tat-associated transcription elongation factor, P-TEFb, is severely reduced because of degradation of CycT1 [65,84,85]. Induction of HIV transcription requires reversal of the chromatin restrictions, which is typically the result of NF-κB induction and binding to its recognition sites in the HIV LTR region. Chromatin restrictions appear to be particularly important for maintaining HIV latency in microglial cells [44,[77][78][79][80][81][82][83]. Broad spectrum inhibitors of histone deacetylases or histone methylase transferases are usually potent inducers of latent proviruses [86,87]. We have shown here that HDACi 4b [51], which is a chemically novel histone deacetylase inhibitor, is capable of reactivating HIV in the absence of NF-κB induction and is one of the most potent activators of HIV transcription in the microglial cells that we have evaluated ( Fig. 2; Additional file 3: Fig. S3 and Additional file 8: Fig. S8). In view of these molecular mechanisms, it is not surprising that TLRs, which are potent inducers of NF-κB and AP-1, are also able to activate HIV transcription in latently infected microglial cells.

Unique mechanisms of TLR-mediated HIV reactivation in microglial cells
We tested the ability of agonists for TLR1-9 to reactivate HIV in microglia as well as in monocytic and in T cells. Surprisingly, only the agonist for TLR3, and to a lesser extent for TLR4, 5 and 6, reactivated significantly HIV in latently-infected microglial cells (Figs. 3, 4, 5; Additional file 2: Fig. S2; Additional file 3: Fig. S3). The agonists for TLR4, 5 and 6 induced NF-κB p65 nuclear translocation, but the effects were transient and therefore proviral reactivation was inefficient.
The engagement of TLR3 resulted in the strongest HIV reactivation in human microglial cells of any of the TLRs (Fig. 5). We confirmed that in microglial cells, poly (I:C) did not induce the nuclear translocation of p65 (Fig. 7), but rather uniquely activated the IRF3 transcription factor (Figs. 8,9). HIV reactivation by poly (I:C), but not by LPS, was significantly inhibited by the IRF3 activation inhibitor bufalin [58] (Fig. 8b), which precluded poly (I:C)-mediated IRF3 nuclear recruitment. This is consistent with earlier reports that LPS did not induce IRF3 nuclear translocation [88]. This suggests that an NF-κBindependent pathway, utilizing IRF3 activation pathway [89], may be responsible for mediating HIV reactivation in the presence of poly (I:C). There have been no previous reports of IRF3 activation of latent HIV proviruses. Indeed, the only previous reports of IRF3 effects on HIV replication come from Suh et al. [90], who found that ligands for both TLR3 and TLR4 inhibited HIV replication in microglia in an IRF3-dependent manner due to the induction of host restriction factors. Another report by Sang et al. [91] also indicated that activation of TLR3 inhibited SIV infection and replication in macaque macrophages through induction of viral restriction factors. More detailed studies will be needed to distinguish between the effects of TLR3 activation on pre-integration events, subject to host restriction, and post-integration proviral reactivation events of the type we have investigated.
These results were confirmed using ChIP experiments (Fig. 9). Treatment with poly (I:C), but not with LPS ( Fig. 9) led to the selective recruitment of IRF3, but not of NF-κB, at the HIV LTR. Transcriptional regulation by IRF3 of NF-κB responsive genes has been observed in other contexts. For example, Wang et al. [92] showed that the expression of the ZAP gene was directly regulated by IRF3 following virus infection or stimulation of cells with dsRNA or dsDNA, and that interaction with ZAP promoter was not dependent on NF-κB. Similarly, Freaney et al. conducted a comprehensive ChIP-seq study, which provided a detailed and quantitative genome-wide analysis of transcriptional regulation of the cellular antiviral response, and revealed extensive colocalization of IRF3 and NF-κB during virus infection [93].
TLR3 is activated by double-stranded RNA (dsRNA) [94], an intermediate formed by most viruses during their replication phase, and functions as a signal to activate inflammatory cells [95,96]. For example, in vitro transcribed HIV gag mRNA complexed with lipofectin activates TLR3 [97].
In the context of HIV infections, it is important to note that TLR3 responses can also be triggered by bacterial and other dsRNAs [97][98][99]. High levels of circulating bacterial rRNA, LPS, and other bacterial antigens is a hallmark of the chronic immune activation seen in HIVinfected patients, due to long-term damage to the gut. Since LPS and dsRNA can reach the brain, it seems likely that they can act in unison to reactivate HIV in infected microglial cells and thereby induce neuronal damage.
The signaling pathways initiated by dsRNA/TLR3 differ between human cells of separate lineages, as well as between mouse and human. Lundberg et al. [100] showed that TLR3-induced mechanisms of human primary dendritic cells, macrophages, endothelial cells, and synovial fibroblasts, while expressing TLR3 at comparable levels, differ substantially. For example, poly (I:C) induced IP-10 secretion by all cell types, while dendritic cells and macrophages failed to produce TNF-α and IL-6 and, unexpectedly, TNF-α was secreted only by synovial fibroblasts. Interestingly, these findings were specific for human cells, and not for murine cells.
Microglial cells are extraordinarily sensitive to activation by the pro-inflammatory cytokines TNFα or IL-1β, and the presence of these cytokines in the CNS correlates strongly with the development of HAND [69]. At low concentrations, these cytokines are capable of eliciting strong HIV responses. Surprisingly, we found no additive effects, or even cross-talk, between TNF-α or IL-1β, and TLR ligands (Fig. 11). However, when we used the chromatin-modifying agent HDACi 4b in combination with Pam3CSK4, poly (I:C), or LPS, but not imiquimod, HIV reactivation was stronger than each of the elements individually (Fig. 9). This strongly implies that chromatin remodeling is required for potent TLR-mediated responses.

Attenuation of TLR-mediated activation of latent HIV
The level of TLR-mediated HIV reactivation varies widely between the different myeloid and T cell types (Fig. 5). Even though microglial cells can express the entire spectrum of TLR receptors ( Fig. 6; [22]), they are relatively inefficient for the reactivation of HIV compared to THP-1 cells. In most cases, the poor responses in microglial cells do not correlate with levels of receptor expression. For example, hT-hµglia/HIV (HC01) cells express very high levels of CD14 (Fig. 1) and TLR4 (Fig. 6), but respond poorly to LPS (Fig. 5). In hµglia/HIV cells, engagement of TLR4 is able to transiently induce NF-κB, as exemplified by clone HC69 (Fig. 7), suggesting that specific feedback mechanisms dampen down NF-κB responses. By contrast, in monocytic cells the NF-κB response is sustained and HIV reactivation is enhanced.
In microglial cells (and astrocytes), unlike in the rest of the cells of the immune system, activation of the NF-κB pathway not only leads to expression of pro-inflammatory genes, but also to the recruitment of repressor complexes to the promoter of pro-inflammatory genes to prevent the mounting of exacerbated inflammatory responses in the brain, which can damage adjacent neurons [101,102]. We are currently investigating whether specific co-repressors are also used to attenuate HIV responses to NF-κB in microglial cells.
We have also examined the effect of TLR ligands on HIV reactivation in latently-infected T cells. Our results demonstrate that activation of TLR5 weakly, but significantly, induce HIV expression in latently-infected Jurkat/ HIV (2D10) cells as well as in primary Th17/HIV (Fig. 5). This is consistent with previous observations by Thibault et al. [36], that TLR5 stimulation is a potent activator of latent HIV-1 provirus in Jurkat T cells, and also activates virus gene expression in T CM . Novis et al. also reported that the TLR2/1 agonist Pam3CSK4 leads to viral reactivation from latency in cultured T CM , but in contrast to our results using both Jurkat and primary Th17 cells, there is no significant activation of these cells by TLR4 or TLR5 agonists [103].

Conclusions
HIV does not infect neurons, but it is frequently found in perivascular macrophages and microglia [104]. HIV encephalitis correlates with the number of activated brain mononuclear phagocytes (both perivascular macrophages and microglia), but not with the amount of virus or the number of infected cells [105,106]. Therefore, the major role of microglial cells in HAND development appears to be neuroinflammatory and neurotoxic, which is greatly potentiated by viral proteins shed by viral CNS sanctuaries [107].
Our data demonstrate that unique patterns of TLR expression and novel signaling cascades create unique responses to microbial products by microglial, monocytic, and T cell lineages. In particular, TLR3-mediated HIV reactivation by IRF3 in microglial cells is a novel pathway, which allows HIV to emerge from latency in infected microglia and potentially cause neuronal damage.
The TLR pathways may also exacerbate neuronal damage in response to viral proteins and drugs of abuse. Recently, El-Hage et al. [108] have shown that exposure to HIV-1 Tat and/or gp120 altered TLR expression in astrocytes, providing a clue on how viral proteins may interfere with the innate immune response of the CNS to HIV-1. Similarly, Dutta et al. [109] showed that morphine and HIV Tat, together, can lead to up-regulation of TLR2, 4 and 9, enhanced pro-inflammatory cytokines (IL-6, TNF-α) levels, and neuronal damage.
Since HIV patients characteristically have chronic inflammation due to the release of microbial components into the circulation, TLR responses in each of these cell types is likely to contribute to disease progression. Manipulation of TLR signaling pathways is likely to find applications in strategies for viral eradication and/ or silencing [110], since these receptors are differentially engaged in cells of the CNS, the monocyte/macrophage lineage, and T cell subsets.

Development of latently-infected cells
The human immortalized microglial cells (hµglia) used to generate HIV-latently infected cells are described in Garcia-Mesa et al. [40]. HIV infection was then conducted essentially as previously described for CHME-5/HIV [46] to obtain mixed and clonal populations of hµglia/HIV cells. Briefly, infection by spinoculation was carried out with vesicular stomatitis virus G-(VSVG) pseudotyped lentiviral vectors bearing a fragment of HIV-1 pNL4-3 , containing Tat, Rev, Env, Vpu, and Nef (some cell lines contain an older HIV construct carrying no Nef [44]) cloned into the pHR' backbone together with the short-lived green fluorescence protein (d2EGFP), as previously shown [44,111]; (Fig. 2a). mNF-κB HIVs, bearing a fragment carrying mutations in the NF-κB binding sites on the LTR region [29], were used as negative controls. The viral particles were produced by the triple transfection of 293T cells using lipofectamine, and the vector titer was determined as described previously [46,112]. GFP + cells (mixed and clonal populations) were isolated 48 h post-infection by fluorescence-activated cell sorting (FACS), further cultured, expanded, and allowed to enter into a latent state (stable, low GFP expression) for four weeks or more, depending on the cell type. hT-CHME-5 cells, described in Garcia-Mesa et al. [40], was used to obtain hT-CHME-5/HIV cells by superinfection with HIVs, as above. Routine evaluation of HIV latency was performed by treatment with TNF-α (Sigma-Aldrich, Cat. # T6674) or HDACi 4b [51], or any other appropriate stimulator for 16 h prior to quantification of GFP + cells by flow cytometry analysis (see below). To keep HIV basal expression low (below 5%), latent cells were maintained in 1% FBS (in DMEM supplemented with 1× normocin) instead of 5% FBS.

Integration site analysis
We basically followed the protocol described in Wu et al. [113] with some modifications. Briefly, we digested 250 ng of isolated genomic DNA (using the Qiagen DNeasy Blood and Tissue kit) with MseI and BglII Fast Digest restriction endonucleases (ThermoFisher Scientific) in Fast Digest buffer for 5 min at 37 °C, followed by heat-inactivation for degree for 10 min at 70 °C. 50 ng of the digest was then ligated with 100 ng of phosphorylated and annealed MseI linker using Quick Ligation (Ther-moFisher Scientific) for 5 min at room temperature (per manufactures instructions), and then heat-inactivated for 10 min at 70 °C. The product was then subjected to a first PCR round using the Phusion Flash High Fidelity Taq master mix (ThermoFisher Scientific), and following a protocol consisting of initial denaturing at 98 °C for 10 sec (1 cycle), and denaturing at 95 °C for 1 s, annealing at 56 °C for 5 s and extension at 72 °C for 15 s (25 cycles). In the reaction mix (20 µL total volume), we added 10 µL of 2× Phusion Flash PCR Master mix, 1 µL of Fwd HIV-1 3′ LTR Primer (10 µM; AGTGCTTCAA-GTAGTGTGTGCC), 1 µL of Rvs Linker Primer (10 µM; GTAATACGACTCACTATAGGGC), and 5 ng of ligated DNA. A second PCR round, using the same master mix, was carried out by initial denaturing at 98 °C for 10 s (1 cycle), and denaturing at 95 °C for 1 s, annealing at 60 °C for 5 s and extension at 72 °C for 15 s (25 cycles). In the reaction mix (50 µL total volume), we added 25 µL of 2× Phusion Flash PCR Master mix, 0.5 µL of Gex-Barcode-A Ion Adapter (50 µm), 1 µL of Fwd HIV-1 3′ LTR Primer, 1 µL of Rvs Linker Primer, and 1 µL of the first round PCR product. Finally, the second round PCR product was run on a 1.5% agarose gel and sizes from 200 to 350 were selected for sequencing. 300 pg of DNA was used for Ion Torrent sequencing following manufactures protocol, and the flanking sequences of genomic DNA were analyzed using the BLAT alignment tool (http://genome. ucsc.edu).

Reagents, cell culture, and treatments
TLR ligands (Human TLR1-9 Agonist kit, Invivogen, Cat. # tlrl-kithw) were prepared as recommended by the manufacturer prior to addition to cell cultures. Safe, non-toxic doses of these TLR agonists were chosen by the propidium iodide (PI) exclusion method [described elsewhere, counting stained (dead) cells with a Cellometer ® Vision automatic cell counter (Nexcelom Bioscience, MA)], after experimentation with THP-1/HIV (HA3) cells for further treatment with the rest of the cell lines tested.
For testing NF-κB dependence of TLR ligands-mediated HIV reactivation, cells were pre-treated for 2 h with either 100 µM of IKKγ NEMO binding domain inhibitory peptide, or equivalent amount of the control peptide (Imgenex), prior to incubation with indicated doses of TLR ligands for 16 h. Quantification of GFP + cells by flow cytometry followed.
In general, assays on suspension cells (THP-1, U937, SC, and Jurkat) were carried out at a density of 1 × 10 6 cells per mL, in 96-well plates in a volume of 100 µL. Assays on microglial cells (attached) were carried out in 24-well plates containing 1 × 10 5 cells per well plated at least 8 h prior to treatments. Cell culture maintenance was carried out at 37 °C in 5% CO 2 , and treatments were performed under the same conditions for 16 h prior to evaluation of viral reactivation by flow cytometry and/or fluorescence microscopy.

Flow cytometry and microscopy
Quantification of GFP-expressing cells was carried out by fluorescence-activated cell sorting (FACS or flow cytometry) analysis using the LSRFortessa instrument for cell sorting, the FACSDiva software (BD, NJ) for data collection, and the WinList 3D software for data analysis. For counting GFP + cells, treated and untreated cells were collected and resuspended in 300 µL of cold PBS.
Further characterization of the newly-developed hµglia/HIV cells included measuring the surface expression of CD11b (BD 553310), P2RY12 (Abcam ab86195) and CD14 (eBioscience 12-0149), as well as the expression of TLR3 (SCBT sc-10740) by fluorescence microscopy. For fluorescence microscopy, microglial cells were cultured on glass coverslips, treated, and subjected to immunofluorescence with anti-CD11b-FITC, anti-CD14-PE, anti-P2RY12-PE antibodies, or secondary PE-conjugated antibody for anti-TLR3 primary antibody to detect expression of the target receptors. Briefly, treated cells were washed, fixed with 4% paraformaldehyde, and permeabilized with 0.1% Triton X-100 prior to incubation with antibodies for 2 h followed by washing with DAPI-containing washing solution for nuclear staining. For FACS analysis, used to evaluate surface expression of TLR1 (SCBT sc-130896), TLR2 (BD 558318), TLR3 (SCBT sc-10740), TLR4 (eBioscience 12-9917), TLR5 (SCBT sc-130897), TLR6 (SCBT sc-30001), TLR7 (SCBT sc-30004), TLR8 (SCBT sc-25467), or TLR9 (SCBT sc-25468), we used 1 × 10 5 cells resuspended in 1 mL of cold PBS in the presence of 0.5 µg of the antibody or equivalent amount of species-specific isotype control antibody for 20 min on ice. For comparative profile of TLR expression, we used serum-starved HIV-infected or uninfected microglia, monocytic, and T cells, in the absence or presence of poly (I:C) (100 ng/mL). TLR binding to cognate antibody was carried out as described above. Appropriate secondary antibodies were used in the absence of fluorophore-conjugated primary antibody. Cell-antibody complexes were centrifuged, and the pellet resuspended in 300 µL of PBS before FACS analysis.
Where applicable, cells were brightfield-photographed and imaged for GFP + fluorescence emission with a Nikon TE2000 inverted scope equipped with a DS-QiMc camera and controlled by NIS Elements software (Nikon). Images were produced by using the ImageJ software (NIH).

Detection of TLR stimulation in THP1-XBlue ™ cells
THP1-XBlue ™ cells (Invivogen, Cat. # thpx-sp) were cultured as recommended by the Invivogen protocol. After the treatment with the TLR ligands or pre-treatment with the CDK inhibitors DRB (10 µM) or flavopiridol (30 nM) prior to treatment with TLR ligands, at the doses indicated in the Fig. legend, cells were centrifuged at 1500 rpm for 5 min, while preparing the QUANTI-Blue ™ following the instructions on the pouch (Invivogen). 180 µL of resuspended QUANTI-Blue ™ was mixed with 20 µL of cell supernatant in a well of a flat-bottom 96-well plate, and incubated at 37 °C for the periods of time indicated (for 4 h in the experiments involving the inhibitors) prior to determining the SEAP levels using a spectrophotometer at 620-655 nm.

SDS-PAGE/Western blot analysis
Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE)/Western blot analysis (carried out as described elsewhere) of NF-κB p65 or IRF3 was used to assess pathway activation of NF-κB or IRF3, respectively. Briefly, cells were treated with TNF-α or shown TLR ligand at the indicated concentration for the indicated time period. For this, 7 × 10 6 cells per treatment were cultured or plated (microglia), and treated as described above. After two washings with cold PBS, cells were collected in 500 µL of Buffer A [10 mM Hepes/KOH, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 1 mM ethylenediaminetetraacetic acid (EDTA)] in the presence of phenylmethylsulfonyl fluoride (PMSF; 1 mM), dithiothreitol (DTT; 1 mM), 0.5% nonyl phenoxypolyethoxylethanol (NP-40), and 1× Halt ® protease/phosphatase inhibitors cocktail (Pierce), and cytoplasmic extracts were recovered in the supernatant after centrifugation at 1500×g for 10 min at 4 °C. Nuclei were then washed three times in Buffer A, and re-suspended in 100 µL of Buffer B (20 mM Hepes/KOH, pH 7.9, 25% glycerol, 420 mM NaCl, 1 mM EDTA, 1.5 mM MgCl 2 ) containing PMSF (1 mM), DTT (1 mM) and the inhibitors cocktail, and centrifuged at 20,000×g for 15 min at 4 °C to recover nuclear extracts (NE). Protein concentration in NE was measured by Bradford assay, and protein solutions were subjected to SDS-PAGE/Western blot using the Santa Cruz Biotechnology antibodies against NF-κB p65 (sc-372), IRF3 (BD Biosciences #550428), or SPT-5 (sc-28678) or TBP (sc-273), as loading controls. These primary antibodies were bound by the appropriate IRDye 800CW or 680LT secondary antibody, and the membranes were scanned and analyzed using the Odyssey ® Infrared Imaging System (LI-COR Biosciences, NE). Similarly, expression of CycT1 and CDK9 in whole cell extracts (WCE) from untreated or treated hµglia/HIV (HC01) and (HC69) cells was detected by SDS-PAGE/Western blot using anti-CycT1 antibody (SCBT sc-10750) or anti-CDK9 antibody (SCBT sc-8338), respectively, and anti-tubulin antibody as loading control.

Chromatin immunoprecipitation (ChIP)
Sample preparation for ChIP experiments were carried out essentially as previously described [44,61] with minor modifications. For each experimental condition, 7 × 10 6 HC69 cells were plated on a 150-mm diameter plate and incubated overnight. Cells were then left untreated or treated with either TNF-α (10 ng/mL), poly (I:C) (100 ng/mL) or LPS (10 ng/mL) for 30 min. The cells were then cross-linked in 1% formaldehyde, incubated for 10 min at ambient temperature, and the reaction quenched by adding glycine 1 mM, and further incubated for 5 min. After a double wash with cold phosphate buffer saline (PBS), cells were collected in PBS/Halt ® cocktail solution by centrifugation for 5 min at 3000×g. The cell pellets were resuspended in 500 µL of CE buffer vortexed, and incubated for 10 min on ice. Nuclei were collected by centrifugation for 3 min at 9000×g, and resuspended 250 µL of SDS lysis buffer(1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1), incubated on ice for 15 min with periodic vortexing, and sonicated in a water bath using a Biorupter Plus water bath sonicator for 20 min (30″ ON/30″ OFF cycles). Under these conditions the chromatin was cleaved to between 100 and 500 bp.
Fragmented chromatin was recovered in the supernatant after centrifugation for 5 min at 9000×g. 5 µg of control IgG or anti-RNAP II pSer5 (ab5131, Abcam), anti-NF-κB p65 (C-20, SCB), or anti-IRF3 (BD Biosciences #550428) antibodies were incubated in a blocked protein A/G coated plate for 30 min. For each immunoprecipitation, 45 µL of the chromatin fractions were diluted in 100 µL of IP dilution buffer and added to antibody coated well. Antibody binding reactions were carried out for 1 h at ambient temperature with 500 rpm shaking. After two washes with RIPA buffer (25 mM Tris, pH 7-8, 150 mM Na, 0.1% SDS, 0.5% sodium deoxycholate, 1% Triton X-100) and one wash with TE buffer, chromatin-IgG complexes were eluted and digested in elution/Proteinase K buffer for 30 min at 65 °C. Freed DNA was purified with PCR magnetic clean up beads (PCR cleanup beads, Axygen).
Samples were pooled and HIV was enriched using hybridization to biotinylated HIV probes. The probes were produced by sonicating the HIV-1 pNL4-3 construct to generate 500 bp fragments. The probes were then end repairing/daTailed using the protocol as described above and Biotin-14-dATP was added using terminal transferase. To enrich for HIV sequences, the pooled barcoded library was added to 6× SSC buffer, 500 nM blocking oligos (pool of all barcodes), 5 µg Cot-1 DNA and water to bring up to 98 µL and incubated for 10 min at 95 °C. Immediately after, 100 ng of biotinylated HIV probes were added to the sample, vortexed, and the temperature was reduced to 65 °C and incubated for 1 h. After hybridization, the biotinylated probe/HIV library was bound to 13 µL of washed myOne streptavidin Dynabeads (Thermofisher). Beads were then pelleted on a magnet and washed 3 times with 200 µL of Ion Torrent wash buffer. Each wash was performed by pipetting up and down 10 times followed by vortexing for 10 s. The last wash was done with 200 µL of ddH 2 O and then the beads were resuspended in 10 µL of fresh ddH 2 O.
The DNA on the beads were amplified using a BesTaq master mix (ABMgood) and primers to the Ion Torrent adapters A and trP1 for 35 cycles. The library was size selected between 200 and 500 bp using gel electrophoresis followed by 1:4 water diluted PCR clean up bead isolation. The enriched library was sequenced using the S5 Torrent Sequencer using a 540 chip following the manufactures protocol.