Barriers for HIV Cure: The Latent Reservoir

Integration site selection and epigenetic regulation

The transcription of integrated HIV-1 proviruses, as well as the expression of host genes, is strongly regulated by variations in the condensation state of the chromatin. Genes located in high compacted DNA regions, or heterochromatin, remain silenced due to inaccessibility of transcription factors to their regulatory elements. By contrast, highly expressed genes are found in less condensed DNA areas (euchromatin).¹⁰⁶ Contrary to what it was previously thought, the process that leads to HIV-1 proviral DNA integration does not occur in a random manner, but in a more selective way: by targeting euchromatin regions through a direct interaction between the viral integrase and the host protein LEDGF/p75, which in turn interacts with RNA-binding proteins found in areas where DNA is being actively transcribed.^107,108 In addition to LEDGF/p75, other host proteins are involved in the selective integration of HIV-1 proviruses into actively transcribed chromatin areas, such as CPSF6 (through its direct interaction with capsid molecules found in the preintegration complex), and the host splicing machinery (in this case, through their association with LEDGF/p75).^{107,109–111} Consistent with this, a high-throughput study of the integration landscape of HIV-1, in which primary cells from HIV-1 viremic progressors, viremic controllers, and patients under cART were analyzed, revealed that HIV-1 proviruses frequently integrate within intronic regions of highly expressed genes, irrespective of the type of patient (controllers or progressors).¹¹² However, in patients controlling viremia with cART, the majority of the proviruses analyzed corresponded to latent HIV-1, since actively infected cells are negatively selected by the antiretroviral drugs. In this context, these proviruses were rarely found within genes, and in cases where they integrated within coding regions, these genes were surrounded by heterochromatin.¹¹² This is in agreement with findings using in vitro T cell line models for latency, where HIV-1 proviruses are frequently located in/or near heterochromatin,¹¹³ suggesting that the condensation state of the chromatin at the integration sites influences the outcome of the infection.¹¹¹ However, studies in primary cells showed that HIV-1 proviruses have an analogous integration pattern in both resting memory CD4⁺ T cells and actively infected CD4⁺ T cells,^114,115 reflecting that there must be factors other than chromatin condensation state and integration site selection involved in the establishment and maintenance of latency in vivo.

Although discrepancies have been found when analyzing the proviral integration sites in T cell lines and primary cells,^116,117 even when using the same cell line models, common features have been defined, such as the observation that integration favors Alu repeats both in genic and intergenic regions, implying that the integration sites may harbor some consensus sequence, although unknown yet.^112,115 Furthermore, the inconsistent results on the chromatin status surrounding latent proviruses may simply reflect versatile epigenetic modifications. As stated earlier, besides the HIV-1 integrase, cellular cofactors aid in the integration of the provirus, and these molecules preferentially bind to less condensed chromatin areas. Therefore, it is tempting to speculate that events occurring during the activated-to-resting transition of CD4⁺ T cells may influence the overall chromatin structure, causing chromatin remodeling at the locations where HIV-1 proviruses are integrated.¹¹⁸

The condensation state of the chromatin, and therefore, its related transcriptional levels, is regulated by a highly dynamic process, which involves posttranslational modifications (PTMs) of different DNA-associated proteins. Chromatin is wrapped around histone proteins, forming nucleosomes. However, unlike euchromatin, heterochromatin is characterized by a highly condensed status, where the chromatin–histone interactions are more rigid. Differential combinations of PTMs on these histones will determine the condensation state of surrounding chromatin, since PTMs can alter the ability of histones to recruit Mi-2/NuRD and SWI/SNF complexes, which are directly related to DNA condensation and unwinding processes, respectively. Histone acetyltransferases (HATs) promote histone acetylation and DNA relaxation, increasing transcriptional activity, whereas HDACs promote chromatin condensation, which represses gene expression.¹⁰⁶ Similar to cellular DNA, the 5′ long-terminal repeat (LTR) of HIV-1 proviruses contains two nucleosomes: (1) nuc-0, located at the 5′ end of the LTR (between nucleotides −405 and −245, from the transcription start site), and (2) nuc-1, located after the provirus promoter (between nucleotides +20 and +165)^119–121 (Fig. 1). These nucleosomes are present even when proviral DNA integration occurs within euchromatin regions.^122,123 Therefore, their condensation status will directly influence proviral transcription, mainly because both nuc-0 and nuc-1 partially overlap with the transcriptional control regions in the LTR, and thus, can alter the association of transcriptional regulators to the HIV-1 promoter as well as interfere with RNA elongation¹²³ (Fig. 1A). For instance, viral reactivation and nuc-1 remodeling was observed in different studies when HATs, such as CBP or GCN5, were recruited to the provirus integration site.^124–126 Similar results were obtained when resting CD4⁺ T cells were exposed to HDAC inhibitors, which led to nuc-1 displacement and facilitated the recruitment of the RNA Pol II to the LTR.^6,127 These results indicate that the acetylation and, in turn, condensation state of the HIV-1 nucleosomes, play an important role in the establishment and maintenance of latency.

In addition to histone acetylation status, epigenetic control can also occur by directly modifying DNA through methylation, a reversible modification that has a profound impact on gene expression. The methylation of cytosines on CpG-rich regions causes transcriptional repression. This has also been observed for HIV-1 proviruses, particularly in areas of the proviral DNA containing the transcriptional control regions. Methylation of cytosines in this location causes proviral DNA silencing and promotes latency^6,82,102,128 (Fig. 1A, top). Several reports have shown a high methylation pattern on the 5′ LTRs of latent proviruses from latently infected CD4⁺ T cells obtained in vitro.^128–131 These epigenetic modifications prevent transcription factors from binding to their cognate enhancers and thus, repress provirus transcription.¹³² In agreement with these results, demethylation of the provirus LTRs has been associated with transcriptional reactivation and virus reemergence from latency in vitro.^128,133,134 However, DNA methylation does not seem to play a major role in HIV-1 latency in vivo, since no significant levels of CpG methylation are found within the 5′ LTR in resting CD4⁺ T cells from HIV-1-infected individuals on suppressive ART.¹³⁵ These observations suggest that the contribution of CpG methylation in the establishment of the latent reservoir in patients is minimal compared with the other epigenetic factors mentioned above.

Transcription factors

As indicated earlier, the largest reservoir for HIV-1 is found in resting memory CD4⁺ T cells, after they transition from an activated status to a resting memory phenotype. Because HIV-1 RNA synthesis relies on the presence of cellular transcription factors that are only temporarily available upon antigen exposure, HIV-1 transcription declines in this activated-to-resting transition.^6,118,136 Therefore, the availability of these transcription factors is a major factor governing the establishment of latency. The HIV-1 proviral DNA contains several regulatory regions on its 5′ LTR with responsive elements for different host transcription factors, which, upon binding, alter viral RNA synthesis by modulating RNA Pol II access and recruitment to the TATA box in the promoter.^102,137,138 Among these factors, NF-κB, NFAT, and Sp-1 play a key role in this transactivation process.^{6,102,139–141} The presence of these three transcription factors in the nucleus, and their ability to interact with their enhancer elements at the LTR, is required to maintain basal HIV-1 transcriptional levels,^140,142 whereas their cytoplasmic sequestration, or inability to bind to the regulatory elements due to mutations, has been linked to latency in CD4⁺ T cells.^{6,102,140,142,143} In addition to NF-κB, NFAT, and Sp-1, there are other transcription regulators, AP-1,¹⁴⁴ C/EBPb,¹⁴⁵ USF-1, and Ets-1,¹⁴⁶ with the ability to cooperate in HIV-1 transcription.^6,147

Above all, NF-κB seems to play a central role in the establishment and maintenance of latency. In response to pathogens and proinflammatory cytokines, NF-κB monomers, p50 and RelA (p65), dimerize and translocate into the nucleus, where they interact with the NF-κB regulatory elements in their target cellular genes, but also in the proviral LTR, strongly activating HIV-1 transcription.^139,148 In resting CD4⁺ T cells, RelA is found sequestered in the cytoplasm by the inhibitory protein IκBα.^148–150 Moreover, while IκBα retains RelA, p50/p50 homodimers are able to translocate to the nucleus and further reinforce proviral latency by recruiting HDACs to the HIV-1 provirus, promoting its heterochromatinization (Fig. 1A, top).^102,127 Interestingly, a recent study uncovered that the cellular protein breast cancer associated-gene 2 (BCA2, also known as Rabring7) is an important negative effector of NF-κB, and in turn, detrimentally affects HIV-1 transcription and replication.¹⁵¹ Specifically, BCA2 promotes the SUMOylation of IκBα, rendering this molecule resistant to degradation and a stronger NF-κB inhibitor.¹⁵¹ Ongoing studies in our laboratory have revealed that BCA2 plays a pivotal role in fine-tuning NF-κB in primary CD4⁺ T cells, and that its depletion from naive CD4⁺ T cells makes HIV-1 less prone to undergo latency. In line with these observations, upregulation of BCA2 enforces and prolongs proviral silencing while the selective knockdown of this gene in latently infected cells results in spontaneous viral reactivation, highlighting the relevance of BCA2 and its NF-κB regulatory activity to proviral latency (Colomer-Lluch et al., unpublished data).

In addition to BCA2, there are other transcriptional regulators that negatively impact HIV-1 transactivation and thus, enhance the establishment of latency. Among these molecules we find CYLD and Naf1, which also inhibit NF-κB-driven viral transcription,^152,153 LSF-1 and YY-1, which recruit HDAC1¹⁵⁴; CBF-1, which reduces the amount of RNA Pol II in the LTR¹⁵⁵; Blimp-1, which interferes with the Tat-trans-activation response (TAR) association¹⁵⁶; or Bcl-11b, which represses HIV-1 transcription by promoting heterochromatinization.¹⁰⁰ Therefore, besides the epigenetic elements considered above, the relative abundance of cellular transcription activators, such as NF-κB, and transcriptional repressors will dictate the fate of the HIV-1 infection in terms of latency.

Tat and RNA elongation

One of the main obstacles during HIV-1 transcription is the inefficient synthesis of genome-length transcripts due to RNA Pol II pausing, an event caused by secondary structures in the nascent RNA that halts RNA elongation.^157,158 To overcome this block, HIV-1 uses the viral transactivation protein Tat, which increases the elongation efficiency during transcription by interacting with a secondary structure in the nascent HIV-1 RNA (TAR element), allowing Tat to associate with positive and negative host regulatory protein complexes. To release the RNA Pol II pause, HIV-1 requires the presence of the transcription elongation factor (P-TEFb).^159,160 P-TEFb is a protein complex containing cyclin T1 and CDK9 kinase, which phosphorylates the carboxyl-terminal domain of the RNA Pol II recruited at the 5′ LTR, enhancing in turn its transcription elongation activity.^159,161 However, most P-TEFb molecules are held catalytically inactive through their association with the 7SK small nuclear ribonucleoprotein (snRNP) complex. Tat promotes P-TEFb dissociation from 7SK, allowing P-TEFb to become catalytically active, hyperphosphorylate RNA Pol II, and release the transcriptional pause.^162,163 By contrast, the negative elongation factor (NELF) complex plays the opposite role in the regulation of RNA Pol II activity.^160,164 Specifically, NELF associates with RNA pol II, limiting its escape from the HIV-1 promoter region.¹⁶⁵ Besides this primary role, NELF may also influence the overall structure of the chromatin. In latent proviruses, RNA pol II complexes accumulate near the transcription start site at the 5′ LTR (nucleotide +11), and at nucleotide +50^165,166—the latter is analogous to the major pause observed in host genes.^167–169 However, in the absence of NELF, the +50 pause site is shifted 20 nucleotides downstream, reflecting repositioning of nucleosomes.¹⁶⁵ Therefore, NELF seems to impact HIV-1 transcription at several levels. In productive infections, however, P-TEFb counteracts this repressing effect by phosphorylating NELF and rescinding NELF's association with the nascent HIV-1 RNA.¹⁷⁰ Consistent with this, in vitro as well as in vivo studies have associated low levels of P-TEFb with HIV-1 latency in resting CD4⁺ T cells.^118,136

Despite the fact that Tat significantly enhances the efficiency of HIV-1 RNA synthesis, the first round(s) of viral transcription after the integration of the proviral DNA need to occur in a Tat-independent manner, since Tat is a nonstructural protein. Although these initial Tat-independent transcriptional events are very inefficient, some genome-length HIV-1 transcripts are successfully synthesized, mainly due to the action of NF-κB. Similar to Tat, RelA (a monomer of the NF-κB heterodimer) also has the ability to recruit P-TEFb to the 5′ LTR¹⁷¹ (Fig. 1A, bottom), increasing the efficiency of RNA elongation and generating in turn full-length transcripts that will subsequently be translated into viral proteins, among which, we find Tat. From this point on, and due to a positive feedback, a highly efficient Tat-dependent HIV-1 transcription phase begins by the mechanisms described above.^{82,159,161,162,172} Additionally, Tat recruits chromatin-remodeling factors, such as SWI/SNF, which facilitate transcription by promoting DNA relaxation and consequently nuc-1 remodeling.¹⁷³

Tat's role in latency has been extensively documented.^159,174 For instance, Tat protein alone was found to be sufficient to successfully reactivate HIV-1 transcription in latently infected CD4⁺ T cells isolated from patients undergoing cART.^159,175,176 This was demonstrated by delivering tat either by transduction with non-HIV vectors,¹⁷⁶ or by ectopically delivering recombinant Tat protein to these cells.¹⁷⁵ In this particular case, a soluble Tat-GFP chimeric protein was produced from pancreatic cells of transgenic mice, and delivered to primary, latently infected CD4⁺ T cells of patients undergoing cART. Remarkably, ectopic delivery of Tat or T cell receptor (TCR)-mediated cell activation caused similar levels of reemergence from latency.¹⁷⁵ Likewise, the ectopic expression of this chimeric Tat protein, or tat's stable expression through transduction, in Jurkat and THP-1 cells prevented the establishment of latency and reactivated latent proviruses.^{175,177–179} Consistent with these findings, HIV-1 variants encoding attenuated tat genes are more prone to undergo latency, as evidenced by studies by the Karn laboratory using primary CD4⁺ T cells infected either with wild-type HIV-1 NL4-3 or a NL4-3 variant carrying the attenuating H13L mutation in Tat.^118,178,180 Therefore, despite the presence of cellular transcription factors that aid in HIV-1 RNA synthesis, defects in Tat will make HIV-1 more prone to latency due to the inability to relieve the RNA Pol II pause block.

Interference

Not only does the integration site of the HIV-1 DNA influence virus transcription (due to the local chromatin state), but also the presence of active host genes surrounding the provirus, and their relative orientation, have a significant impact on HIV-1 replication.^181–184 As mentioned earlier, HIV-1 shows a preference of integration within transcriptionally active regions in the latently infected J-Lat cell line¹⁸¹ as well as in resting CD4⁺ T cells isolated from patients on cART.¹¹⁴ Despite the fact that these proviruses can easily gain access to transcription factors and RNA Pol II, the presence of actively transcribed genes nearby the proviral DNA drastically affects HIV-1 RNA synthesis. In fact, the elongation of RNAs from neighboring host genes can have both a repressive and activating effect on HIV-1 transcription through different mechanisms: (1) when the integrated provirus and the nearby host gene share the same polarity, RNA Pol II elongation may prevent transcription factors from binding to their cognate sequences on the 5′ LTR, and thus, promote proviral silencing and latency^181,183 (Fig. 1B, top); (2) alternatively, the activation of host genes upstream the HIV-1 5′ LTR can cause read-through transcription of the provirus, which is an important limitation to a productive infection. This phenomenon consists of adjacent promoters initiating transcription of a host gene and continuing through the HIV-1 coding sequence, therefore impacting HIV-1 gene expression.^114,181,185 Although these HIV-1 transcripts are generally nonproductive, the effect of host gene read-through transcription might prevent latency and up-regulate proviral expression if the provirus is integrated in the same orientation as the host gene. (3) On the other hand, proviral integration in the opposite orientation of an actively expressed host gene leads to (i) a premature termination of transcription due to the collision between RNA Pol II complexes moving in different directions, which inevitably reduces the amounts of HIV-1 transcripts^181,184 (Fig. 1B, bottom), or (ii) the generation of antisense HIV-1 RNAs. Although this event has mainly been observed from the HIV-1 3′ LTR,^103–105 one can envision a similar phenomenon occurring from host genes that are located downstream, and in the opposite direction of the HIV-1 provirus, leading to antisense HIV-1 genomic transcripts. Nevertheless, upon cellular activation, the repressive effect of these interfering mechanisms is diminished by the strong binding affinity that NF-κB exhibits for the HIV-1 5′ LTR, enhancing transcription from the virus promoter rather than from the host gene.^6,186 Although these observations indicate that the proviral integration site can affect HIV-1 transcription due to (1) the chromatin condensation status, which may hinder the recruitment of the cellular transcriptional machinery, and (2) the interference of cellular genes surrounding the HIV-1 provirus, most of these studies examined only few clones of latently infected cells. Therefore, a comprehensive interrogation of more clones is needed to ascertain how the orientation of HIV-1 proviruses relative to host genes affects the outcome of the infection.

MicroRNAs

It is well established that the presence of cellular miRNAs can regulate gene expression through the RNA-induced silencing complex (RISC), either by impairing translation or by promoting the degradation of transcripts.¹⁸⁷ This is also true for HIV-1, since it has been reported that cellular miRNAs present in memory CD4⁺ T cells can directly affect HIV-1 RNA synthesis.^188–191 Using computational analysis, miR-29a, miR-29b, miR-149, miR-324-5p, and miR-378 were identified as miRNAs with a regulatory effect on HIV-1 transcription.¹⁸⁹ Ahluwalia et al. confirmed that miR-29a inhibits HIV-1 transcription in Jurkat cells by targeting the nef region on HIV-1 transcripts, although Frattari et al. reported that this miRNA affects HIV-1 RNA levels by targeting the 3′ untranslated region (UTR).^192,193 miR-28, miR-125b, miR-150, miR-223, and miR-382 have also been reported to reduce HIV-1 gene expression and infectivity by targeting the 3′ UTR of the HIV-1 RNA.¹⁹⁰ In addition, miRNAs can also affect the overall expression levels of HIV-1 dependency factors, playing a role in HIV-1 latency.¹⁹⁴ This has been observed for cyclin T1¹⁹⁵ and Pur-α, a co-factor of Tat.¹⁹⁶

The study of miRNAs is still an emerging field, and so is the investigation of their effects on HIV-1 infection. Since the amount and species of these small RNA molecules depend on multiple intracellular and extracellular factors, research on how miRNAs affecting HIV-1 are regulated, as well as their mechanisms of action, will not only increase our understanding on their role in HIV-1 latency, but also could open new avenues to neutralize this pool of cells.

RNA export

Even if the epigenetic signature, cellular environment, and the presence of cellular transcription factors is optimal to support HIV-1 RNA synthesis, HIV-1 might, in theory, still establish a latent infection if viral transcripts are not exported, translated, or packaged into nascent virions. An important mechanism directly involved in the regulation of HIV-1 gene expression is the transport of viral transcripts from the nucleus to the cytoplasm.^197,198 The efficiency of this process relies on the presence and activity of both host and viral proteins. Rev is an auxiliary HIV-1 protein that binds to a specific nucleotide region (rev-responsive element or RRE) within viral transcripts (nonspliced and singly spliced), facilitating their cytosolic translocation.^199–202 In fact, the presence of low levels of Rev is required for a productive HIV-1 infection, whereas its absence appears to play a role in proviral latency.²⁰³ In addition, Rev-independent events related to RNA transport also impact the establishment of latency. For instance, nuclear retention of HIV-1 transcripts, particularly multiply spliced HIV-1 RNAs, was observed in primary, latently infected CD4⁺ T cells obtained from patients undergoing cART.¹⁷⁶ Although the mechanism by which this retention only occurs in resting, latently infected cells is unknown, the overexpression of the RNA-binding protein PTB circumvented this posttranscriptional block, allowing for the production of virions without requiring CD4⁺ T cell activation.¹⁷⁶ Hence, factors that affect HIV-1 gene expression at a posttranscriptional level also influence the outcome of the infection, and can enforce latency.

Implication of these mechanisms in the establishment of latency in vivo

The relevance of each of the mechanisms presented above in the formation of latent reservoirs is widely accepted. However, how their combination affects the outcome of the HIV-1 infection is not well understood. In addition, the fact that many of the studies discussed in this review were performed in the context of cell lines or ex vivo models may not accurately reflect how latency is established in vivo, which complicates our understanding of the molecular bases influencing latency in patients. Despite these limitations, few aspects stand out for their convergence in vitro, ex vivo and in vivo, such as the roles of Tat and host transcription factors. In particular, Tat and NF-κB displace the effect of regulatory elements, such as epigenetic modulation and transcriptional interference, allowing HIV-1 RNA synthesis to proceed.^{138,150,159,162,163,171,173,175} Nevertheless, we need to stress that our knowledge of the roles of less investigated elements, such as miRNAs and HIV-1 RNA export in the outcome of HIV-1 infection is incomplete. Therefore, additional studies on physiologically relevant systems are needed to better comprehend their influence over the other mechanisms, and their overall contribution to HIV-1 latency.

Shock and kill strategies

Multiple therapeutic strategies aimed at achieving a functional cure of the HIV-1 infection are focused on the “shock and kill” strategy, using LRAs which, in combination to cART, disrupt HIV-1 latently infected cells, so they can be eliminated by immune effectors (cytotoxic T cells or CTLs) or cleared by the cytopathic effect of the reactivated virus.^208–211 Optimally, effective and safe LRAs should reactivate the latent provirus and promote viral transcription followed by viral protein production. This approach should enhance HIV-1-specific immune responses (CTL-mediated responses) to recognize the presentation of viral antigens and clear the reactivated cells. So far, several assays in ex vivo latency models, as well as in clinical trials, have been performed to evaluate the feasibility of small-molecule LRA candidates to effectively reverse HIV-1 latency, which have recently been reviewed elsewhere^210,211 (Fig. 2, left panel). These include:

• (1)

  Protein kinase C (PKC) agonists (e.g., PMA, prostratin, bryostatin, and ingenol molecules). PKC agonists activate PKC cellular isoforms, which then initiate NF-κB signaling to induce HIV-1 transcription, thereby, potently reversing latency.^88,212,213 According to recent studies, only bryostatin-1 has shown efficacy when tested in cells derived from patients on suppressive cART.^214,215 However, one of the main concerns when applying PKC agonists relates to the associated side effects resulting from continuous activation and inflammation.²¹⁴ To address this, alternative approaches have been proposed to reduce the PKC agonist-mediated off-target effects of T cell activation. For example, the use of PKC agonists in combination with adjuvant agents, such as the JAK inhibitor ruxolitinib, has shown a decrease in proinflammatory cytokine production while maintaining the capacity to reverse latency.²¹⁰ Also, an attractive option to restrict PKC agonist-derived toxicities is the recently proposed oral therapy based on the PKC agonist ingenol, a plant-derived natural compound, which reactivates latent HIV-1 and when combined with other LRAs shows synergistic effects, facilitating a reduction in the dosing of both agents.²¹⁶ Another possibility to reduce toxic effects associated with polyclonal T cell activation might be the combination of PKC agonists with Rapamycin. In a recent study by the Siliciano laboratory, the combination of LRAs with Rapamycin, a drug that triggers autophagy but also works as an immunosuppressant, showed efficient viral reactivation with no toxic effects from the systemic release of proinflammatory cytokines.²¹⁷

• (2)

  Histone deacetylase inhibitors (HDACi; e.g., vorinostat/SAHA, panobinostat, and romidepsin). As previously stated, histone deacetylation (HDAC) is an epigenetic modification of chromatin that results in the repression of HIV-1 RNA synthesis.²¹⁸ Small-molecule inhibitors targeting HDAC enzymes have shown to safely reverse proviral latency by promoting HIV-1 transcription both in vitro and in vivo, although with different success. For example, the HDACi vorinostat/SAHA modestly reactivates latent HIV-1 in resting CD4⁺ T cells.^{208,209,219–221} However, this proviral reactivation is not associated with an effective killing of latently infected cells in vivo,²²¹ suggesting that CTL activation is a prerequisite for HDACi treatment success. By contrast, treatment with panobinostat not only reactivates proviruses, but also increases innate immune activity of natural killer (NK) cells, suggesting that innate immune responses can critically modulate the efficacy of LRAs.²²² In the recent past, HDACi have been the pillars of latency reversal in clinical trials, evidencing that after HDACi treatment an enhancement in HIV-1 RNA and subsequent increase in viral production was observed.^{208,223–225} Unfortunately, to date, none of the HDACi-based treatments tested in clinical trials has succeeded at significantly reducing the size of the reservoir (reviewed in Spivak and Planelles²¹⁰).

• (3)

  Disulfiram, an FDA-approved drug, reactivates latent HIV-1 in a primary CD4⁺ T cell model by decreasing the levels of the PTEN cytoplasmic protein with the subsequent activation of the Akt signaling pathway.^90,226 Two clinical trials have recently assessed disulfiram as a LRA, with no significant reduction in the size of the latent reservoir being observed.^227,228

• (4)

  Bromo and extraterminal (BET) bromodomain inhibitors, together with toll-like receptor agonists, have shown promising results as LRAs, but the efficacy of many of these compounds in vivo remains to be examined.

• (5)

  Remarkably, in a search for novel LRAs, maraviroc, a drug currently used in cART as a CCR5 antagonist, was found to reverse HIV-1 latency in infected individuals under cART in a phase II clinical trial.

• (6)

  The authors of this study suggested that the interaction of maraviroc with CCR5 induces the activation of NF-κB and the subsequent enhancement of HIV-1 transcription in resting CD4⁺ T cells. Therefore, maraviroc shows potential as a LRA to reduce the HIV-1 viral reservoir in ART-treated patients.²³⁵

Despite all this progress, the shock and kill strategy is beset by significant limitations before effective elimination of the reservoir can be achieved. Although LRAs may disrupt latency in CD4⁺ T cells, LRAs are poorly efficient in other HIV-1 target cell populations, such as macrophages and microglial cells. Thus, their potential to reactivate all the cell types that constitute the latent reservoir is partial at best. Even the most effective LRAs have been reported to fail at reactivating all replication-competent proviruses.³ An important caveat to LRAs is that they modulate gene expression in a general manner; thus, their potential for off-target or toxic effects is a major concern. For example, HDACi, such as SAHA act by opening the chromatin structure around the promoter of the HIV-1 integrated provirus, and therefore, may lead to the expression of other host genes, and potentially induce reactivation of other integrated proviruses alongside HIV-1. Additionally, some LRAs may activate uninfected HIV-1 target cells, which then may become susceptible to HIV-1 infection. In some cases, despite their combination with ART, LRA implementation might result in reinfection of target cells in anatomical viral sanctuary tissues, where ART has less penetration, including the CNS.²³⁶

Another main challenge of this approach is that LRA compounds fail at inducing sufficient viral cytopathic effects and potent HIV-1-specific CTL responses to purge HIV-1 reservoirs, showing so far a moderate impact on the reservoir size. Obstacles, such as the presence of CD8⁺ T cell escape mutations in the proviral genome of latently infected CD4⁺ T cells,²³⁷ LRAs limited success at reactivating latent proviruses to allow antigen presentation for CTL-mediated elimination,²³⁸ or the presence of HIV-1 reservoirs in lymph nodes and the CNS, where CTL infiltration is more difficult, seem to prevent the CTL-mediated clearance of latently infected cells, enhancing even further the establishment of HIV-1 reservoirs.

Additional considerations include the window opportunity of the LRAs and the fact that these compounds could alter the host immune function. In this regard, a recent study explored the effect of LRA combinations in HIV-1-specific CD8⁺ T cell responses and found that although the administration of romidepsin (HDACi) and bryostatin-1 (PKC agonist) effectively results in the reversal of the latent sate, this dual treatment inhibits HIV-1-specific CD8⁺ T cell function to recognize and eliminate HIV-1-reactivated cells.²³⁹ Interestingly, another study suggests that natural latent HIV-1 reservoirs in ex vivo CD4⁺ T cells from ART-treated patients present intrinsic resistance to CD8⁺ T cell-mediated clearance, which may complicate even further their eradication.²⁴⁰ The authors speculate that the observed resilience may be attributed to (i) HIV-1 Nef downmodulation of MHC-I on the surface of HIV-1-infected cells, resulting in a lack of recognition by CD8⁺ T cells, and/or (ii) HIV-1 latently infected cells with inherent resistance may be selected to conform a large fraction of the reservoir. These limitations underscore the need for more effective and specific LRAs for viral reactivation accompanied by potent immunotherapeutic interventions to ensure stronger HIV-1-specific immune responses, and urge the design of alternative therapeutic strategies to successfully eliminate the latent reservoir.

Block and lock strategy

The block and lock approach has recently emerged as an alternative to the shock and kill strategy to functionally cure HIV-1. This approach consists of the use of small-molecule inhibitors, HIV-1 latency-promoting agents (LPAs), to effectively silence (block) the HIV-1 promoter region in latently infected cells, and therefore, restrict (lock) viral transcription of HIV-1 integrated proviruses, impairing in turn viral rebound, and further intensifying their latent state. Several compounds are being investigated for their LPA potential, some of them with promising results (Fig. 2, central panel). These include:

• (1)

  Levosimendan, which specifically blocks the Tat-enhanced HIV-1 transcription in a PI3K pathway-dependent manner.²⁴¹

• (2)

  The small-molecule Curaxin 100 (CBL0100), which impairs the access of the RNA Pol II and FACT at the HIV-1 promoter region, blocking the elongation process.²⁴²

• (3)

  The G1V001 telomerase-derived peptide, which inhibits HIV-1 transcription and viral reactivation by blocking the NF-κB pathway in an Hsp90-dependent fashion.²⁴³

• (4)

  Didehydro-Cortistatin A (dCA), a Tat inhibitor characterized by Mousseau et al.^96,244 that potently silences HIV-1 transcription.

Altogether, with this study, other compunds affecting Tat biogenesis have already been reported to cause proviral silencing.²⁴⁵ However, to our knowledge, dCA is the only compound used in vivo that provided the first proof of concept for the potential use of a Tat inhibitor to durably suppress viral transcription.²⁴⁶ With the exception of dCA, the effectiveness of the other compounds needs further investigation, particularly in vivo, before their ultimate therapeutic use as anti-HIV-1 agents.

Despite some limitations, the encouraging results obtained in these studies indicate that the block and lock strategy is feasible and thus, urge to explore the potential of other molecules as LPAs. One promising locking approach is to target the HIV-1 Rev protein. Rev is critical for viral replication, since it regulates the export of nonspliced and singly spliced viral transcripts from the nucleus to the cytoplasm and, as mentioned above, defects in RNA export are thought to play an important role in HIV-1 latency. Campos and collaborators recently characterized ABX464, a Rev inhibitor that blocks viral replication in vitro as well as in humanized mice, showing long-lasting effects and no associated adverse effects in normal cellular splicing.²⁴⁷ Specifically, ABX464 binds to the Cap-Binding Complex (CBC) and impairs Rev-dependent export of nonspliced and singly spliced HIV-1 messenger RNAs (mRNAs), preventing their translation and neutralizing the expression of the viral genome. This molecule holds promise to affect the viral reservoir and encourages future studies to move further into clinical trials. Other attractive molecules as LPAs are miRNAs. In the past decade, more studies have revealed that miRNAs play a role in modulating latency in HIV-1-infected, quiescent cells.²⁴⁸ As mentioned above, miR-29a, which is one of the best-characterized miRNAs affecting HIV-1, efficiently suppresses the expression of viral transcripts. This observation led the authors to consider miR-29a as a potential tool for suppressing the HIV-1 reservoir by potentiating the endogenous levels of miR-29a, but also through miR-29a agonists. Either strategy could lead to a permanently silent state of HIV-1 proviruses as part of the block and lock strategy.¹⁹³

Although still an emerging line of investigation, permanently repressing these latently infected cells appears a promising approach to keep dormant proviruses in check, given the limited success of the shock and kill strategy. A major caveat to this approach is the need to identify cellular markers that will allow to specifically target LPAs into cells harboring replication-competent proviruses and minimize off-target effects.

Proviral DNA scission

In the past few years, the use of gene-editing technologies, such as CRISPR/Cas9 (clustered, regularly interspaced short palindromic repeats—CRISPR-associated proteins) has gained momentum, including in the HIV-1 field. Genetic modulation by CRISPR/Cas9 is based on the use of a guide RNA (gRNA) that recognizes and mutates target DNA in the HIV-1 genome, offering sequence-dependent specificity and high cleavage efficiency.^249,250 These properties make CRISPR/Cas9 a very attractive tool to permanently inactivate or even remove the proviral DNA from HIV-1 latently infected cells, and also to protect cells from future HIV-1 infections due to the continuous expression of the CRISPR/Cas9 system, thus, minimizing any potential harm to the host. To date, this technique has been tested in several cell types targeting critical HIV-1 genes as well as the LTR regions, since cleavage of both LTRs will result in the complete excision of the proviral DNA²⁴⁹ (Fig. 2, right panel, B). For example, Ebina et al. reported the successful removal of HIV-1 proviruses in infected Jurkat T cells with CRISPR/Cas9 by targeting the HIV-1 LTRs.²⁵¹ In another study, by targeting the HIV-1 Rev region in the provirus, latently HIV-1-infected Jurkat cells were efficiently inactivated.²⁵² The CRISPR/Cas9 system has also been successfully implemented in human pluripotent stem cell-derived HIV-1 reservoir cell types, where this technique was used not only to excise latent proviruses, but also to neutralize incoming infections by specifically targeting the incoming viral cDNA.²⁵³ Hu and collaborators further analyzed the capacity of the CRISPR system to eradicate the HIV-1 proviral DNA from myeloid lineage cells, the main HIV-1 cell reservoir in the brain. By targeting the U3 region in the HIV-1 LTRs, they effectively blocked viral gene expression, replication, and completely removed the proviral integrated genome that was flanked by the two LTRs, with no detectable off-target effects.²⁵⁴ Despite these encouraging results, some studies have reported undesired effects when using CRISPR/Cas9 to edit the human genome.²⁵⁵ This is mainly due to the cellular machinery that naturally repairs the cleaved DNA: nonhomologous end joining (NHEJ) or double-strand break (DSB) homologous recombination. Whereas homologous recombination has high fidelity, NHEJ is highly error prone, and the choice of one over the other is not controllable. To overcome this limitation, other cutting-edge gene-editing tools have been proposed to safely excise HIV-1 proviruses, such as a modified version of the Cre-Lox recombinase.²⁵⁶ Although an attractive alternative, the recent advances and improvement of the CRISPR/Cas9 system, together with the fact that errors introduced by NHEJ will mainly affect HIV-1 proviruses rather than the host genome, has allowed to reliably clear integrated HIV-1 proviral DNA. For instance, the study developed by Kaminski et al. demonstrated efficient cleavage and removal of the HIV-1 genome in latently HIV-1-infected CD4⁺ T cells by targeting the LTR with no associated harm to the DNA of host cells, and no undesired effects. Remarkably, in their ex vivo assays they found that their lentivirus-delivered CRISPR/Cas9 significantly reduced HIV-1 DNA and protein levels in CD4⁺ T cells from HIV-1-infected patients.²⁵⁷ The same team also provided proof of concept in an in vivo study that the CRISPR/Cas9 system can cleave HIV-1 DNA in transgenic mice and rats.²⁵⁸ Overall, these data support the solid potential of the CRISPR/Cas9-based tools to efficiently and safely cleave the HIV-1 integrated genome in latently infected cells, and therefore, block viral gene expression and replication as well as immunize against future HIV-1 infections.

Besides its use to inactivate proviruses, CRISPR/Cas9 can also be exploited to activate the transcriptional activity of the HIV-1 integrated provirus in latently infected cells in a site-specific manner, in line with the shock and kill strategy (Fig. 2, right panel, A). There is increasing evidence supporting the use of CRISPR-based transcriptional activation systems, alone or in combination with LRAs, to reverse HIV-1 latency and achieve the complete eradication of the latent infection. These CRISPR-based activation systems present substantial advantages over current LRA treatments, such as enhanced LTR activity compared with traditional shock and kill strategies, and minimal toxic and off-target effects,^259–261 triggering fully infectious viral production.²⁶²

One of the main concerns when using CRISPR/Cas9 to edit DNA in quiescent cells, such as the HIV-1 reservoir, is the reliance on the cellular enzymatic machinery to express the gRNA and the Cas9 protein. Most CRISPR/Cas9 tools involve the use of lentiviral or adenoviral vectors that encode for these components. Therefore, it is easy to envision that the success rate for efficiently editing the HIV-1 proviral genome would significantly increase by preactivating these latently infected cells with LRAs, so the cells are metabolically active by the time they are transduced with CRISPR/Cas9. However, this does not seem to be the case. According to a recent publication by Zhu et al., latently infected cells can be efficiently edited without preactivation.²⁵² Nevertheless, aspects such as the cell cycling status may interfere with the success of this technique, and should be investigated in detail.

There is no doubt that CRISPR/Cas9, as well as other gene-modifying techniques, has revolutionized the landscape of editing tools. Although very promising, there are some limitations and concerns that should be addressed before its general implementation as anti-HIV-1 treatments. HIV-1 viral escape is one of the main hurdles for a broad application of CRISPR/Cas9, and several studies have recently addressed whether HIV-1 could evade the CRISPR/Cas9-mediated suppression. Interestingly, Deng and collaborators found that the latent reservoir can also contain escape mutants for CRISPR/Cas9.²³⁷ However, the use of multiple gRNAs targeting different HIV-1 conserved regions could circumvent HIV-1 escape. In fact, recent reports proved that HIV-1-infected cells can be cured by a combination of two gRNAs targeting various sites in the proviral DNA, effectively blocking HIV-1 replication and preventing viral escape.^263,264 Unfortunately, other studies evidenced HIV-1 evasion of the CRISPR/Cas9 suppression when mutations arose in the gRNA target sequence.^265,266 In addition, due to the limited success in the delivery of these CRISPR/Cas9 systems, alternative delivery methods such as nanoparticles should be tested to enhance the applicability of this approach. Another caveat is that most of the data we have on CRISPR/Cas9 and HIV-1 is derived from in vitro studies. CRISPR/Cas9 effectiveness and safety should be analyzed in actual resting, latently infected CD4⁺ T cells obtained from HIV-1-infected patients to design customized CRISPR/Cas9 pairs, and this can be costly and time consuming.

Quantitative PCR-based methods

Different quantitative PCR (qPCR)-based methods are currently used to measure the relative amount of HIV-1 cDNA or HIV-1 RNA present in a certain pool of peripheral blood mononuclear cells or plasma obtained from infected patients.^283–285 For the DNA-based qPCR methods, the most widely used is the AluPCR, which measures the total integrated proviral DNA. By combining primers that target gag or the LTR together with the highly repetitive elements Alu in the human genome, this technique exclusively detects integrated proviral DNA. However, it does not discriminate between actively or latently infected cells, nor if the detected proviruses are replication competent or defective.^271,286 For the RNA-based qPCR, the most common methods aim at measuring the relative levels of extracellular RNA (plasma viremia) over cell-associated RNA.^{271,280,283,287} However, these methods also fail at pinpointing latently infected cells. Although PCR-based assays are among the most sensitive and simple available tools, their main caveat is the unwarranted detection of defective integrated proviruses.^271,280,288 In an attempt to address this limitation, Procopio et al. developed the Tat/Rev-induced limiting dilution assay (TILDA), which measures by qPCR the relative amount of cell-associated multiply-spliced HIV-1 RNA upon T cell stimulation, which in turns correlates with the presence of replication-competent proviruses.²⁸⁰

Quantitative virus outgrowth assays

The Q-VOA, unlike PCR-based methods, can single out resting CD4⁺ T cells capable of supporting a productive infection by measuring the relative release of infectious virions upon cellular stimulation. Hence, it is considered the gold standard method to quantify drug-resistant infected cells from resting CD4⁺ T cells isolated from patients on cART.^271,281,282 However, compared with the qPCR assays, Q-VOA is more demanding, since this method is lengthy, requires large amounts of samples from donors, and seems to underestimate the actual size of the reservoir.^3,27,35,281 Unfortunately, comparative experiments using both Q-VOA and PCR-based assays have shown clear discrepancies between these two techniques. In addition, they are limited to the quantitative nature of the reservoirs and they are incapable of identifying which subpopulation of the pooled cells has the potential to sustain a productive infection.²⁸⁸ Therefore, new improved methodologies to measure the latent reservoir in a simple, reproducible and effective manner are urgently needed. For instance, approaches based on proviral DNA sequencing are now gaining more attention, since they can accurately elucidate the composition of the proviral landscape within an individual, and discriminate between defective and replication-competent proviruses. Although they can be extremely expensive and labor intensive, the use of next-generation sequencing seems to streamline this approach significantly.³⁵

In addition to the methods described above, several in vitro, ex vivo, as well as animal models have been developed to facilitate the study of HIV-1 latency. The in vitro systems include immortalized cell lines such as the T lymphocyte cell models, ACH2, J1.1, and Ca5, or the premonocytic U1 clone.^83,289–292 These cell lines remain chronically and nonproductively infected, but retain the potential to support a productive infection upon stimulation with mitogens and cytokines such as αCD3/CD28 antibodies or the tumor necrosis factor alpha (TNFα), respectively.^290,291 Jordan et al. generated the widely used J-Lat cell line, which was produced by inducing the stable and latent infection of Jurkat cells with a Δenv/Δnef HIV-1 NL4-3 carrying a GFP reporter.¹¹³ One of the weaknesses of these cell lines is that, unlike resting CD4⁺ T cells, they remain in a constant proliferation state and thus, they may not entirely reflect latency in vivo. To overcome this, several groups have developed ex vivo systems of latency using primary CD4⁺ T cells.^{118,293–297} Among these studies, the method developed by Bosque and Planelles is characterized by the isolation of naive CD4⁺ T cells to near purity, and inducing their differentiation through TCR engagement into a central memory-like phenotype. While activated, the cells are infected with a molecular clone of HIV-1, and are subsequently allowed to transition to this resting phenotype in a nonpolarized manner,^296,298 which closely resembles the characteristics of T_CM in vivo. Ex vivo systems have been extensively used to dissect signaling pathways that are involved in HIV-1 latency in CD4⁺ T cells, but fail at addressing other aspects of proviral latency, such as the role of non-T cells and the contribution of viral sanctuary tissues to the size of the reservoir. In this regard, humanized mouse models represent a more physiological scenario for the study of latency. These animal models are generated by injecting HIV-1 into immunodeficient recipient mice transplanted with either (1) fetal human thymus and liver, which procures human T cell lymphopoiesis (Thy/Liv mice model),^299–301 (2) human bone marrow CD34⁺ hematopoietic stem cells (HSCs; huCD4 mice model),³⁰² or (3) a combination of both, which generates the commonly used BLT mice model.³⁰³ However, many investigators are skeptical to use humanized mice as a model for the study of HIV-1 latency. Although they virtually have a human immune system, the roles of other organs and tissues, such as viral sanctuaries, may not accurately reflect the overall picture of the HIV-1 reservoir in humans. Animal models that better resemble human physiology and lentiviral latency are nonhuman primates infected with SIV (simian immunodeficiency virus) or recombinant viruses of SIV and HIV (SHIV), such as rhesus macaques or pig-tailed macaques.^{38,234,304–308} These models allow for (1) a more accurate study of latency and the roles of organs such as the CNS and the gut in the size and distribution of the reservoir,^307–309 and (2) testing the efficiency of antiretrovirals,³¹⁰ and possibly LRAs and LPAs. Despite these advantages, the use of nonhuman primates in AIDS research is expensive and requires the use of specialized facilities. Moreover, unlike the humanized mice models, this system requires long periods of experimental preparation.^234,308 Therefore, in spite of the improvement of the experimental systems in the study of latency^{234,280,311,312} much work is still needed to develop cost and time-efficient models that would closely resemble HIV-1 latency in vivo.³¹³