Current Status of Latency Reversing Agents Facing the Heterogeneity of HIV-1 Cellular and Tissue Reservoirs

1Approximately, 36.9 million people in the world are living with HIV, while there were 33.3 million in 2010, showing that HIV is still a global public health problem (UNAIDS 2018).

2The development of ART has allowed the suppression of virus replication to undetectable levels. Thus, ART reduces HIV-associated morbidity, prolongs survival, prevents transmission, and results in a partial reconstitution of the immune system, as measured by an increase in circulating CD4+ T cells. However, ART can cause cumulative toxic effects with the emergence of HIV-1 drug-resistant variants. Moreover, ART is unable to achieve complete virus eradication since it targets only actively replicating virus, and as such, treatment needs to be taken life-long. The main barrier to an HIV cure is the formation of stable reservoirs of latent HIV-1 that are defined as a cell type or an anatomical site in which integrated proviruses persist. The latent reservoirs are highly heterogeneous and composed of multiple cell types, such as macrophages, but the best-characterized ones are a small population of HIV-1-infected memory CD4+ T cells. In addition to cellular reservoirs, GALT, CNS, genital tract, and lymph nodes are the main anatomical reservoirs of HIV-1. The role of the respiratory tract, liver, kidney, and adipose tissue as HIV-1 reservoirs is also gaining importance. Moreover, an anatomical site in which ART penetration is limited (called a sanctuary) or a site which is immune-privileged (such as B cell follicular centers within lymph nodes, testis, and the brain) may allow for residual replication contributing to viral persistence. Together, these reservoirs can be induced to actively produce viruses by various cellular stimuli and therefore represent one potential source of rebound of viremia after ART interruption.

3Viruses have developed different mechanisms to escape the host immune system. In addition to the rapid evolution of viral variants, the establishment of viral latency in the infected cells is one of these mechanisms. Latently-infected cells contain stably-integrated, replication-competent proviruses repressed by a plethora of silencing mechanisms operating at the transcriptional and post-transcriptional levels. The number of latently-infected cells carrying replication-competent proviruses is extremely low with a majority of defective proviruses. Significant progress has been made in the development of various therapeutic approaches that target HIV-1 and prevent disease progression. Currently, two strategies are under investigation in order to reach long-term control of viral replication in the absence of ART: the first strategy is aiming at achieving a sterilizing cure (i.e., a total elimination of the virus from the human body), whereas the second is a functional cure (i.e., a remission or long-term control of HIV-1 in the absence of ART, without loss of CD4+ T cells, no clinical progression, lack of HIV-1 transmission, and a reduction of the size of the reservoirs). Given the difficulty of achieving a sterilizing cure, one specific approach for a functional cure, called the “shock and kill” strategy, has become one of the major focus of attention. However, it now appears that the main barrier to reaching success with this “shock and kill” strategy is the heterogeneity of the latent HIV-1 reservoirs that is reflected by the diversity of infected cell types residing in the blood and in the different tissues and by the complexity of the molecular mechanisms governing latency and most likely differing from one cell to the other.

4The present review will discuss the heterogeneity of the HIV-1 reservoirs that has been highlighted by studies using LRAs. We will describe the determinants responsible for these heterogeneous responses to LRAs, including the diversity of cell types composing the reservoirs (their origin, their differentiation, and their activation state). We will briefly present the multiple molecular mechanisms governing latency, the HIV-1 diversity within the latent and reactivated reservoirs, the patient-to-patient and cellular variations in response to LRAs. Finally, we will discuss why a better understanding of these elements is crucial for reaching an HIV-cure.

5HIV-1 infects subsets of CD4+ T lymphocytes, leading either to productively-infected cells or, rarely, to latently-infected cells. HIV-1 infected CD4+ T cells can be distinguished either by their state of differentiation, by their function, or by the markers they express on their surface.

6CD4+ T cells infected with HIV-1 can be grouped by their state of differentiation into naive T cells (TN, CD45RA+, CD62L+, and CD95–) and memory CD4+ T cells, which can be further divided into four subpopulations: central memory T cells (TCM, CD45RA–, CCR7+, and CD27+), effector memory T cells (TEM, CD45RA–, CCR7–, and CD27–), transitional memory T cells (TTM, CD45RA–, CCR7–, and CD27+), and stem cell-like memory T cells (TSCM, CD45RA+, CCR7+, CD27+, and CD95+). TN cells that are more resistant to HIV-1 infection than memory CD4+ T cells and as such contain lower levels of HIV-1 DNA, produce as many virions or greater amount of replication-competent HIV-1 than the memory CD4+ T cells following latency reversal with LRAs, indicating that TN cells are an important reservoir of latent HIV-1. Moreover, the total proportion of TN increases under ART regimen compared to memory T cells. However, clonal expansion is more frequently observed in memory than in TN cells, thus contributing to HIV-1 persistence. Indeed, homeostatic proliferation driving clonal expansion has been evidenced by pioneering work of that identified TCM and TTM as the main reservoir for HIV-1 infection. These two major subgroups are a more stable reservoir for HIV-1 than TEM which are short-lived and, unlike TCM and TTM, express markers of activation (HLA-DR) and might be more sensitive to programmed cell death. It has been previously shown by the group of Sarah Palmer that TEM which are in an activation state carry the highest number of intact proviruses compared to TCM, TTM, and TN. This finding suggests that the resting phenotype of CD4+ T cells (i.e., characterized by no expression of activation markers) is not a prerequisite for latent infection. These differences could be explained by patient-specific variations related to the duration of infection before ART initiation or to the duration of ART. Moreover, some evidence supports the theory that HIV-1 can establish latent infection in actively replicating CD4+ T cells, suggesting that HIV-1 infection of both resting and activated primary CD4+ T cells could result in latency. The TTD, corresponding to aged T-cell populations that reflect HIV-1 disease progression and in which integrated HIV-1 DNA is also detected, are a very small reservoir with a reduced frequency in HIV-1 individuals under suppressive ART and improved CD4+ T cell counts. Finally, TSCM are permissive to HIV-1 infection and contribute to its persistence by their capacity of self-renewal and prolonged survival rate. The proportion of viral DNA associated with these TSCM cells is higher than in TCM. Although TSCM are latently-infected, they represent only a small fraction of the total reservoir.

7Different CD4+ T cell subsets that are generated from TN were shown to be HIV-1 infected. For instance, resting CD4+CD25+ Treg were found to be sensitive to HIV-1 infection acting as a viral reservoir in patients under long ART. These cells are characterized by a lesser response to T cell activation which limits virus expression and inhibits CD8+ T-cell cytotoxic function. Additionally, the gamma-delta T cells (γδ T) which represent approximately 2 to 10% of total circulating CD3+ T lymphocytes and whose majority express TCR from Vδ2 variable regions (hereafter referred as Vδ2 cells), are classified as memory cells according to the expression of CD45RO and CD27 markers. These Vδ2 cells have been documented to be productively infected and depleted upon HIV-1 infection. More recently, it has been demonstrated that Vδ2 cells in ART-treated patients with complete suppression of HIV-1 plasma viremia harbor latent HIV-1 that can replicate following ex vivo stimulation indicating that peripheral Vδ2 T cells are a potential HIV-1 reservoir. Also, Th17 CCR6+ memory CD4+ T-cell subsets in the blood and colon are long-lived cells that act as HIV-1 reservoirs during ART. In addition, T follicular helper cells (Tfh) from the germinal center and peripheral blood (pTfh) are highly susceptible to HIV-1 infection holding replication-competent virus and serve as reservoirs during ART. These cells are characterized by surface expression of CXCR5 and PD-1, reside in the lymph node follicles in immediate anatomical proximity to B cells, and support the germinal center reaction essential for the generation of effective humoral immunity. Notably, the group of Matthieu Perreau, by investigating lymph node Tfh (expressing CXCR5 and PD-1) and pTfh (expressing CXCR3), has shown that these subpopulations are the major sources of infectious replication-competent HIV-1.

8Very recently, resident memory CD4+ T cells (TRM), present in tissues such as the lower female genital tract has been described as a critical HIV-1 reservoir in cervical mucosa. Interestingly, cervical tissues from aviremic ART-treated HIV-1 infected woman contained higher viral DNA content compared to blood samples and showed that CD4+ TRM harboring viral DNA and viral RNA are the main contributors to this reservoir.

9Studies investigating the role in latency of activation markers such as HLA-DR and immune checkpoint molecules (i.e., PD-1, LAG-3, TIGIT and Tim-3) have indicated that these markers are preferentially expressed at the surface of memory CD4+ T cells (TCM and TTM) harboring latent HIV-1 provirus. Although several studies, including those carried on SIV-infected macaques, have demonstrated that cells expressing these markers carry latent, replication-competent integrated viral DNA, the replication competence of the integrated proviruses and the contribution of the cells bearing these markers to the latent reservoir still need to be fully elucidated.

10Recently, the expression of CD32a has been reported as a potential marker of memory CD4+ T cells harboring a replication-competent latent virus in aviremic patients under ART. The role of CD32a as a cellular marker of HIV-1 reservoirs has been the subject of several works. A complete study presented at CROI by Darcis et al. (CROI 2019, Poster 346 - CD32+ CD4+ T cells are enriched in HIV-1 DNA) showed that active CD4+ T cells co-expressing HLA-DR and CD32a are highly enriched with HIV-1 DNA.

11The integrin α4β7 has been shown on a T cell subset that is highly susceptible to HIV-1 infection. Moreover, the integrin α4β1 was shown to be expressed by more than 70% of infected cells both in untreated and ART-suppressed individuals. Integrins mediate the adhesion and transendothelial migration of lymphocytes facilitating their homing to GALT (for α4β7) and to the inflamed central nervous system/bone marrow (for α4β1) suggesting a role in HIV-1 persistence by enhancing T cell expansion.

12In addition, high levels of CD2 receptor expression on latently infected resting memory CD4+ T cells in virally suppressed HIV-1-infected subjects has been also identified. Moreover, CD30 receptor was identified on transcriptionally active CD4+ lymphocytes in individuals on suppressive ART suggesting that it might be a marker of residual, transcriptionally active HIV-1 infected cells in the setting of suppressive ART. Last but not least, very recently the B lymphocyte antigen CD20 has also been identified as a marker of transcriptionally-active HIV-infected cells.

13Even though memory CD4+ T cells are a long-term cellular reservoir for HIV-1, they are not the only source of viral rebound during treatment interruption. Macrophages, DCs, and tissue macrophages, such as microglial cells, are part of the viral reservoir.

14Indeed, overwhelming evidence supports the notion that tissues, such as the CNS, lymph nodes, testes, gut, genital tract, and lungs, serve as HIV-1 sanctuaries that counteract viral eradication. In this regard, the group of Morgane Bomsel has recently reported that urethral tissue macrophages expressing IL-1 receptor, CD206, and IL-4 receptor, but not CD163, constitute a replication-competent HIV-1 reservoir. Importantly, lipopolysaccharides specifically reactivate the production of replication-competent infectious proviruses from these tissue macrophages.

15In the case of circulating monocytes, they are more resistant to HIV-1 infection and their contribution as a viral reservoir is controversial and remains debatable. Monocytes have been proposed as a vehicle of HIV-1 dissemination throughout the body upon their entry in tissues where they differentiate into macrophages. Especially, due to their ability to cross the blood-tissue barrier, monocytes consequently, spread the infection into sanctuaries such as the brain. These characteristics indicated the monocytes as an important viral reservoir. However, even if several studies have reported persistent infection of monocytes in ART treated individuals, others failed to detect HIV-1 in circulating monocytes. A recent work from the group of Nicolas Chomont indicated that monocyte infection is infrequent, and they highlighted the importance of using flow cytometry cell-sorting to minimize contamination by CD4+ T cells.

16Follicular dendritic cells (fDCs) in lymphoid tissues specialized in the trapping and retention of antigens in the form of immune complexes on their surface, including HIV-1 virions, can serve as a potential viral reservoir. Using specific and sensitive next-generation in situ hybridization approach, documented the importance of B cell follicles in active, latent, and persistent infections by analyzing lymphoid tissues from macaques prior to and during ART. These fDCs could thus transfer the virus to T cells present in the follicles of secondary lymphoid organs. In addition, myeloid dendritic cells (mDCs) located in the lymph nodes may support a very low level of viral replication and have a role in HIV-1 latency. However, the mechanism of viral persistence in these cells is not yet clearly understood [reviewed in ].

17Viral DNA has also been detected in hematopoietic stem cells (HSCs) from ART-treated patients, which could demonstrate their involvement in HIV-1 persistence. Despite the limited infection and detection of HSCs in a subset of patients, their role as a viral reservoir may be crucial. Finally, HIV-1 can infect kidney allografts after transplantation despite undetectable viremia thereby suggesting that podocytes can serve as viral reservoirs and revealing once again the heterogeneous composition of the HIV-1 reservoir.

18In conclusion, multiple cell subsets being present at various cellular differentiation states, either resting or activated, can serve as HIV-1 reservoir. All these cells contribute to the complexity and heterogeneity of the reservoirs of latent HIV-1. It is very likely that different molecular mechanisms of latency establishment and persistence are involved in different latently-infected cell types and even may vary among cells of the same lineage.

19It is well established that repression of HIV-1 expression is mediated through a plethora of molecular mechanisms both for the establishment and for the maintenance of post-integration latency (summarized in Figure 1). In fact, HIV-1 latency is a multifactorial phenomenon, controlled by several interlinked mechanisms operating at the transcriptional and post-transcriptional levels [reviewed in )] and influenced by the transcriptional program of the host cell. For instance, some studies have suggested that HIV-1 viral latency is related to the integration site and orientation of the provirus. HIV-1 integration seems random but favors the introns of transcriptionally-active genes located in gene dense regions in the outer shell of the nucleus close to the nuclear pores. The importance of transcriptional interference as a mechanism suppressing the expression of the integrated provirus (including steric hindrance, convergent transcription, and enhancer trapping) has also been proposed [reviewed in ].

20Moreover, during latency, the HIV-1 promoter is heavily controlled by epigenetic mechanisms, including DNA methylation and histone post-translational modifications, such as histone acetylation, methylation, and crotonylation. The degree of DNA methylation on the HIV-1 promoter was long considered controversial due to conflicting observations in patients. However, this heterogeneity in promoter methylation profile between patients has now been explained by the duration of the infection and/or the duration of ART therapy which impacts the accumulation of cytosine methylation on the 5′LTR. Still, the exact mechanisms of this DNA methylation accumulation in the latent reservoir of HIV-1-infected individuals remain unclear.

21Additionally, an antisense transcript originating from the 3′LTR (named the ASP RNA) was recently shown to recruit the PRC2 repressor complex to the HIV-1 5′LTR, increasing the repressive epigenetic mark H3K27me3 while reducing RNAPII occupancy at the viral promoter and promoting the establishment and maintenance of HIV-1 latency at the epigenetic level. An HIV-encoded antisense long non-coding RNA directs the same epigenetic silencing mechanisms by recruiting and guiding chromatin-remodeling complex consisting of proteins such as DNMT3a, EZH2, and HDAC-1 to the viral promoter leading to transcriptional HIV-1 latency.

22HIV-1 latency is also associated with poor availability of transcriptional activating factors including NF-κB, NFAT, and STAT5 due to their cytoplasmic sequestration and with a low expression level of the viral transactivator Tat. The interaction of the cellular transcriptional cofactor CTIP2 with HP1α is involved in the relocation of Tat to transcriptionally inactive regions of the chromatin. However, HIV-1 counteracts CTIP2 mediated-repression by promoting its degradation by the HIV-1 accessory protein Vpr. Another recent finding has revealed that APOBEC3A, the restriction factor that suppresses HIV-1 infection in macrophages, maintains HIV-1 latency by recruiting KAP1 and HP1 inducing repressive histone marks. In addition, the importance of TRIM22, known to inhibit the binding of Sp1 to the 5′LTR promoter and to colocalize with HIV-1 transcriptional repressors TRIM19/PML, contributes to the maintenance of HIV-1 latency.

23During latency and even after transcription initiation, the RNAPII is paused during early elongation, constraining its processivity and resulting in repression of HIV-1 gene expression. RNAPII terminates prematurely and accumulates at specific positions of the HIV-1 promoter that overlap with the position of the repressive nucleosome nuc-1, positioned in the 5′LTR just after the TAR region, inducing a block of transcriptional elongation. This inhibition of HIV-1 transcription elongation is regulated by the binding of negative factors such as NELF and DSIF to the RNAPII and by the limited availability of the cellular transcription factor P-TEFb due to its sequestration by the 7SK snRNP complex. P-TEFb is composed of CyclinT1 and CDK9 and requires the CDK9 Thr-186 phosphorylation for its activity to mediate the phosphorylation of the C-terminal domain of the paused RNAPII antagonizing its negatives factors. The level of CDK9 phosphorylation has been found to be lower in resting CD4+ T cells harboring latent HIV-1. Furthermore, CTIP2 binds to the 7SK snRNP complex to inhibit P-TEFb and HIV-1 transcription in microglial cells.

24Abortive TAR transcripts have been reported even during transcriptional elongation block. In a very elegant manner, the group of Steven Yukl has evidenced that multiple blocks to transcription and even to splicing and export exist in patient cells. They developed a panel of RT-ddPCR assays to measure at the same time different HIV-1 transcripts in patient-derived cells. They quantified the transcripts suggestive of transcriptional interference (U3-U5), initiation (TAR), 5′ elongation (R-U5-pre-Gag), distal transcription (Nef), completion (U3-polyA), and multiple splicing (Tat-Rev). Using blood CD4+ T cells from HIV-1 individuals under ART, they showed a greater block to HIV-1 transcriptional elongation, distal HIV-1 transcription, completion, and multiple splicing than to transcriptional initiation. This transcription profiling approach highlighted the blocks at distinct stages of HIV-1 transcription and splicing (which may be governed by different mechanisms), thereby underlying the heterogeneity of HIV-1 latency in CD4+ T cells.

25The work of and and also of have highlighted the importance of post-transcriptional blocks in HIV-1 latency. Post-transcriptional blocks of the nuclear export of various viral transcripts including unspliced, partially spliced, multiply-spliced, and translation of HIV-1 RNA by miRNAs have also been reported. Interestingly, a recent study from the group of Alessandro Marcello has demonstrated that the expression level of MATR3 and PSF, two known post-transcriptional cofactors of the HIV-1 protein Rev required for Rev-mediated export of RRE-containing HIV-1 RNAs, are poorly expressed in latently-infected patients cells, suggesting a novel post-transcriptional block linked to RNA export. This latter block was reversed by ectopic overexpression of MATR3 which boosted the action of the HDAC inhibitor SAHA in the J-Lat cell line model for HIV-1 latency. These results demonstrated the importance of post-transcriptional blocks, especially at the level of viral RNAs export, that need to be relieved to reach full viral reactivation by LRAs. Similarly, the work of Andrew Mouland and colleagues has demonstrated that UPF-1, a known RNA surveillance protein, acts as a positive post-transcriptional regulator of viral reactivation from latency.

26Additional post-transcriptional mechanisms including the involvement of cellular miRNA and non-coding RNAs in latency have been described. More specifically, the regulation of viral expression and production by miRNAs targeting PCAF and CyclinT1 have been related to latency. Many of these miRNAs involved in this process are expressed in resting cells but are downregulated during T-cell activation. For instance, the lower susceptibility to HIV-1 infection of monocytes in comparison to macrophages has been shown to be correlated with high and low expression of these miRNAs, respectively. Finally, lncRNA called NRON which is strongly expressed in resting CD4+ T cells was shown to be involved in HIV-1 latency by inducing Tat degradation through the proteasome pathway. The combination of SAHA with NRON knockdown significantly reactivates viral production from latently-infected primary CD4+ T cells.

27Altogether, the complexity of the molecular mechanisms underlying HIV-1 latency marks the heterogeneous nature of the HIV-1 reservoirs.

28Mechanistic insights into HIV-1 transcriptional repression during latency have allowed to develop the “shock and kill” strategy. This approach is based on the use of small-size chemical compounds called LRAs which reactivate HIV-1 from latency (the “shock” phase) while maintaining ART in order to prevent new infection events. This kind of strategy would allow the “kill” phase during which latently-infected cells would then die from viral cytopathic effects or host cytolytic effector mechanisms following viral reactivation. Of note, this strategy does not discriminate between replication-competent and defective proviruses. Several classes of LRAs have been developed and studied (reviewed in;; ). These include: PKC agonists, MAPK agonists, CCR5 antagonist, Tat vaccine, SMAC mimetics, inducers of P-TEFb release, activators of Akt pathway, benzotriazole derivatives, epigenetic modifiers (including HDACis, HMTis, and DNMTis), and immunomodulatory LRAs (including TLR agonists, IL-15 agonist and immune checkpoint inhibitors) (summarized in Table 1).

29It has been further demonstrated by several groups, including ours, that targeting a single mechanism might not be efficient enough to reactivate the majority of latent proviruses and that combinations of LRAs acting on several HIV-1 silencing molecular mechanisms are needed to obtain synergistic viral reactivations and achieve a more significant decrease in the size of the reservoirs. Moreover, when LRAs are used in combination, lower concentrations are effective, thereby reducing the toxicity of each LRA. However, even if these LRA combinations reverse latency, they could inhibit multiple CD8+ T cell function [reviewed in ]. Importantly, despite the potent effect of a single LRA and of LRA combinations in vitro and ex vivo, multiple clinical trials have failed to show a decrease in the size of the latent reservoir in vivo. A clear change in plasma HIV-1 RNA with a subsequent decrease in the reservoir size in vivo has been seen only for the immune check point inhibitor nivolumab and for romidepsin (combined with an immunovaccine) but without prolonging time to viral rebound after ART interruption in this latter study. In the case of nivolumab, the study was conducted only on one patient who experienced a drastic and sustained decrease of the HIV-1 reservoir. However, other controversial studies found no consistent changes in the size of the latent reservoir nor in HIV-specific CD8+ T-cell responses in HIV-1-infected individuals treated with anti-PD1 antibodies (including nivolumab). Success to completely eradicate latent viruses with LRAs is hampered by the heterogeneous nature of the latent reservoir and by the diversity of the silencing mechanisms governing latency that make the current “shock and kill” strategy inefficient.

30All studies investigating LRAs have demonstrated the heterogeneous nature of the cellular and tissue reservoirs of latent HIV-1 and their diverse reactivation capacity, highlighting the different determinant responsible for the heterogeneous responses to LRAs. All these studies are summarized in Table 2.

31Few studies so far have focused on investigating the contribution of the viral diversity, the compartmentalization, the intact or defective nature of the viral reservoir, and the origin of the rebounding virus in latency reversal. A previous study in patients who initiated ART during acute infection showed that proviral sequences from PBMCs and GALT presented low level of genomic diversity and divergence and remained unchanged after treatment interruption. Moreover, there was no phylogenetic link between the rebounded plasma viral sequences and those from the GALT proviral DNA, indicating that HIV-1 cellular reservoirs in the GALT may be different from those circulating in peripheral blood and might not contribute to the rebounded plasma viremia. Other studies have supported the compartmentalization idea of the viral population in the gut with divergent opinions. Depending on the stage of HIV-1 infection, the diversity of HIV-1 RNA appears lower in patients with early infection versus chronic infection, and thus, compartmentalization is lost during chronic infection. However, other works have supported other conflicting ideas with findings showing absence of compartmentalization of HIV-1 between the gut and blood, providing evidence for cross infection between these two compartments.

32HIV-1 sequence diversity has been reported to be either higher or similar to genital tract compared to blood. Viral compartmentalization between the blood and the male genital tract has been reported by multiple studies including SIV-infected macaques. More recently, patients under suppressive ART exhibited a significant positive correlation between viral diversity and genetic compartmentalization in the blood and testes, but it was attributable to differential frequencies of identical HIV-1 sequences between the two sites. However, there was no evidence of compartmentalization when only unique sequences per sites are considered, suggesting that compartmentalization between blood and testes is linked to clonal expansion.

33HIV-1 phylogenetic analysis of post-mortem CSF, brain, and spleen from HIV-1 patients under ART and presenting dementia symptoms showed that HIV-1 strains from the blood and spleen are different from those in the brain and CSF. A more recent study of env and nef phylogeny on ART suppressed individuals confirmed the presence of considerable viral diversity in the spleen and lymph nodes. However, there is no viral compartmentalization between spleen and PBMCs in SIV-infected and suppressed rhesus macaques.

34The comparison of latent HIV-1 from the blood and lymph node CD4+ T cells from HIV-1 individuals undergoing ART interruption, after TLR9 coadministration with ART, suggests the same frequencies of intact proviruses in the blood and lymph nodes and the fact that CD4+ T cells carrying latent viruses circulate between the blood and lymphoid tissues. However, there is no overlap between latent reservoirs and rebounded virus, thereby supporting the idea that recombinations may play a role in the emergence of the rebounded viremia.

35Time to ART from estimated date of infection as an early ART initiation is associated with less molecular diversity in CSF without impacting HIV-1 DNA provirus compartmentalization in the CNS which occurs very early after infection. In the study of, a higher diversity in PBMCs than in CSF is reported. The compartmentalized HIV-1 RNA in CNS is found to contribute to viral rebound within the CSF in patients undergoing treatment interruption but is phylogenetically distinct from those present in the paired blood plasma.

36Most of the previous studies estimated that blood memory CD4+ T cells are the source of viral rebound after ART interruption or viral reactivation from latency. The first work of Timothy Schacker and colleagues examined HIV-1 variants by single-genome amplification and phylogenetic analyses in matched lymph nodes, GALT biopsies and blood from HIV-1 suppressed individuals under longstanding ART after treatment interruption. The rebounded virus after treatment interruption was found to be detectable in latently-infected cells at multiple sites with a highly complex and genetically diverse population of virions which bring out the challenges facing the heterogeneity of HIV-1 reservoir. Another study of the effect of a brief treatment interruption on the HIV-1 latent reservoir of individuals who initiated ART during chronic infection showed no alteration either in the size or in the diversity of the peripheral reservoir and highlighted the substantial variability and the prevalence of clonally-expanded viral populations. Interestingly, the group of Sarah Palmer indicated that analytical treatment interruption (ATI) activated proviruses with similar sequence between plasma and intestinal lamina propria mononuclear cells, indicating that intestinal HIV-1 reservoir is contributing to viremia following ATI. A recent study by the group of Linos Vandekerckhove confirmed the heterogeneous source of viral rebound from distinct anatomical reservoirs in HIV-1 individuals undergoing treatment interruption, showing that genetically-identical viral expansions play a significant role in viral rebound. In a study presented by Oliveira et al. (CROI 2019, Poster 327—Characterizing the HIV DNA reservoirs in whole-body tissues in the “Last Gift” cohort), HIV-1 DNA was detected in most body tissues with a nice distribution and compartmentalization of HIV-1 reservoir between tissues from the Last Gift cohort enrolling altruistic, terminally-ill persons living with HIV-1. They have successfully sequenced env from 10 different tissues with many identical HIV-env sequences sampled in multiple body tissues.

37Some LRA interventions analyzed whether the increase of CA-US HIV-1 RNA is related to limited or to broad activation of HIV-1 proviruses. The single genome sequencing of viral RNA transcripts showed that panobinostat and vorinostat activate genetically diverse HIV-1 proviruses that are similar to that observed during ATI but with a high percentage of defective viral sequences. More precisely, TCM contributed to the rebounded viremia, indicating once again the important role of this subset in the persistence of latent HIV-1. Finally, romidepsin administration to HIV-1 individuals under suppressive ART activates transcription from blood CD4+ T cells latent HIV-1 proviruses. Importantly, the viremia induced by romidepsin contained few defective mutations and is characterized by low genetic diversity.

38Together, these studies illustrate that cure strategies should consider the complex and variable composition of the different viral reservoirs, the replication-competent capacity, the diversity and compartmentalization of HIV-1 reservoir, and the role of cellular clonal expansion and cellular proliferation in promoting HIV-1 persistence.

39An important issue for the development of a more effective “shock and kill” approach is to determine how effective LRAs are in terms of full viral reactivation from all latently-infected cells composing the reservoirs. Indeed, most of the in vivo reactivation studies demonstrated a rather weak reactivation effect of LRAs, i.e., an effect only on HIV-1 transcription with much less or no effect on plasma HIV-1 RNA. For instance, a recent clinical trial using escalating doses of disulfiram demonstrated that even if all doses produced an increase in the level of intracellular HIV-1 RNA, only the highest dose increased plasma HIV-1 RNA level albeit with very low effect. Similar observations have been reported for SAHA: i.e., weak effect at the level of viral particle production in vivo as assessed by measurement of plasma HIV-1 RNA. have shown in their primary CD4+ T cell models that disulfiram and SAHA treatments increased viral transcription, but failed to effectively enhance viral translation. In addition, the group of Maria Buzon has very recently demonstrated that a median of 16.28% of the whole HIV-reservoir exhibited HIV-1 transcripts induction after viral reactivation using various LRAs and their combinations, but only 10.10% of these HIV-1 RNA+ cells produced viral p24 proteins. Recent work by the group of Andrew Lever has demonstrated that inefficient splicing regulation may also influence the action of LRAs. They studied the polymorphisms occurring in a recently identified viral mRNA splicing regulatory element (ESEtat) regulating tat mRNA splicing which results in more silent phenotypes of the virus. Indeed, higher doses of LRAs (PMA, JQ1, and panobinostat) were required to reactivate silent viruses bearing the polymorphisms in ESEtat, reflecting their lower rate of inducibility as compared to wild-type HIV-1. Therefore, different post-transcriptional mechanisms including blocks to export and alteration of mRNA splicing may be considered as druggable targets for a combined approach of more potent latency reversal.

40Latency reversing agents were shown to be cell model-specific exhibiting diverse reactivation profiles across multiple HIV-1 latency model systems. A comprehensive study by tested the potency of a panel of thirteen LRAs for their ability to reactivate HIV-1 in several broadly used HIV-1 latency models (primary T-cell models, multiple J-Lat cell lines, and ex vivo T-cell cultures derived from the blood of HIV-1+ individuals). They showed that PHA was the only stimulus that uniformly reactivated latent HIV-1 in all these cell models, although other LRAs exerted largely heterogeneous responses among the various models. Importantly, following LRA treatment, none of the in vitro cell model systems could accurately capture the ex vivo response characteristics of latently-infected T cells from patients.

41In addition to cell model-specific effects of LRAs, recent studies demonstrated the cell type-specific effects of LRAs. For instance, highlighted heterogeneous responses of CD4+ populations to bryostatin and ingenol. The authors showed that bryostatin induced HIV-1 expression in TEM cells but had limited effect in TCM and TTM cells. While ingenol, on the other hand, exhibited more similar reactivation effects among the different memory T-cell subpopulations. Similarly, the group of Maria Buzon has very recently demonstrated heterogeneous responses to LRAs of the latent reservoirs present in different CD4+ T-cell subpopulations. Romidepsin and ingenol and their combination were the most potent LRAs at reactivating HIV-1 in almost all the subsets of CD4+ T cells by increasing, respectively, the proportion of TCM, TEM, TTM, TNA, and TNA, TSCM, TCM, TTM cells expressing HIV-1 RNA. Panobinostat successfully reactivated HIV-1 only in TCM cells. Bryostatin-1 reactivated very modestly some T-cell subsets, including TNA, TTD, and TCM. Therefore, T-cell differentiation status may impact the action of LRAs. Indeed, show in ex vivo and in vitro models that differentiated phenotype of TEM cells from that of quiescent TCM cells is associated with a potentiated response to LRAs and to a highest level of inducible HIV-1 reservoir. Additionally, the effects may also be cell type-specific. We have also recently demonstrated that disulfiram exhibited limited reactivation spectra, being active only in myeloid-derived HIV-1 latently infected cell lines (U1, THP89GFP monocytic, and CHME-5/HIV-1 microglial cells) but not in Jurkat-based T-cell lines. These heterogeneous cellular responses to LRAs indicate that distinct and cell-type dependent molecular mechanisms contribute to HIV-1 latency in diverse reservoirs.

42It has recently been demonstrated by, using an elegant transcriptional profiling approach, that LRAs exhibit silencing mechanisms-specific effects. These authors found that HDACis (panobinostat and romidepsin) and PKC agonists (ingenol 3,20-dibenzoate and ingenol mebutate) exert differential effects on the latency blocks in the blood latently-infected CD4+ T cells. More specifically, HDACis increased total and elongated transcripts but had less or no effect on polyadenylated and multiply spliced transcripts, whereas ingenol mebutate strongly induced polyadenylated and multiply spliced transcripts but had lesser effects on transcription initiation and elongation. These latter results are in agreement with another study showing that romidepsin administration after six doses of the therapeutic vaccine Vacc-4x, in HIV-1 individual under suppressive ART, increased early events in HIV-1 transcription (initiation and elongation) but had less effect on later stages (completion, multiple splicing). The differential effects of these LRAs suggest that the mechanisms underlying the blocks to completion and splicing may differ from those that mediate the blocks to initiation and elongation. The group of Yukl compared CD4+ T cells from the blood and rectum tissue reservoirs using a similar transcriptional profiling approach and found a much greater block to HIV-1 transcription initiation in the rectum compared to blood. Indeed, the ratio of total to elongated transcripts was 6-fold lower in the rectum CD4+ T cells, suggesting less of a block to HIV-1 transcriptional elongation in rectal CD4+ T cells. In fact, a multi-dose trial of SAHA has evidenced that the cell-associated HIV-1 RNA in latently infected CD4+ T cells from the blood was 5-fold higher compared to CD4+ T cells from the rectal tissue, suggesting that the LRA-driven effects may also be tissue-specific. Thus, further studies should investigate whether gut cells differ from blood cells in their response to LRAs in terms of HIV-1 transcript production and of cellular gene expression.

43The group of Guillaume Fillion demonstrated that different LRAs reactivate different subsets of latent proviruses. Using a method called B-HIVE to map the chromosomal locations of individual proviruses, these authors revealed in Jurkat cells that responses to LRAs are also viral integration site-specific. They found that PHA and SAHA reactivated proviruses inserted at distinct genomic locations, suggesting that the insertion context of HIV-1 is a critical determinant of the viral response to LRAs. Using the same B-HIVE technology, the groups of Albert Jordan and Guillaume Fillion demonstrated that MMQO (an LRA which acts as an I-BET) and JQ1 reactivate latent HIV-1 proviruses integrated at distinct sites from those proviruses targeted by SAHA and prostratin. Interestingly, demonstrated that only less than 5% of latently-infected primary CD4+ T cells are reactivated by LRAs. By further sequencing analysis, these authors showed the preference toward integration events in active chromatin sites for the reactivable cell population, while these regions were significantly disfavored in the non-reactivable group, highlighting that the role of chromatin environment is an important determinant of LRA effectiveness.

44Many other determinants may also be responsible for the heterogeneous reactivation profile of LRAs. Patient-dependent effects of LRAs were reported. Indeed, viral productions evident in some patients but not in others were observed in all reactivation clinical trials. Additionally, in our previous reactivation studies, we demonstrated patient-specific variations in terms of reactivation capacity of their ex vivo cell cultures following treatments with various LRAs. Indeed, we established a positive correlation between the size of the HIV-1 reservoirs and the ex vivo capacity of HIV-1-infected patients’ cell cultures to be reactivated by LRAs, but we identified HIV-1+ patients who deviated from this linearity relative to their corresponding HIV reservoir size, indicating that the reservoirs size is one determinant of the cell capacity to produce virus but that this parameter alone is not sufficient. The patient-dependent heterogeneity in the responses to LRAs could be explained by patient characteristics such as genetic background, time to treatment initiation, duration and type of therapy and also by the gender-specificity, as recently proposed by the team of Jonathan Karn. Sex-based differences in HIV-1 reservoir activity is characterized by a higher cell-associated HIV-1 RNA, higher plasma HIV-1 RNA, higher T-cell activation, and PD-1 expression in men compared to women. The group of Jonathan Karn has shown that the estrogen receptor-1 (ESR-1) is a key regulator of HIV-1 latency. More specifically, antagonists of ESR-1 activate latent HIV-1 proviruses and potentiate HIV-1 reactivation by LRAs such as SAHA, TNFα, and IL-15, while ESR-1 agonists potently block HIV-1 reactivation. Despite a reduced inducible reservoir compared to men, women showed much higher levels of inhibition in response to TCR stimulation in the presence of ESR-1 agonists but exhibited a higher reactivation in response to ESR-1 antagonists when combined with SAHA than the group of male HIV+ individuals. The circadian rhythm is an additional biological process that can affect HIV-1 transcription and reactivation. The circadian rhythm has been first suggested in the clinical trial testing short term administration of disulfiram as a parameter influencing HIV-1 transcription. The authors find an unexpected large variation in pre-dosing CA-US HIV-1 RNA which was statistically significantly higher immediately prior to the first dose of disulfiram than at the two previous time points without changes in HIV-1 DNA or plasma HIV-1 RNA. Indeed, Sharon Lewin and colleagues have subsequently demonstrated a significant time-dependent variation in CA-US HIV-1 RNA in CD4+ T cells from HIV+ individuals on suppressive ART, a variation which is modulated by circadian regulator factors driving transcription from the viral LTR. Thus, in ex vivo studies, the time of blood collection could affect LRA reactivation potency and should be considered to improve latency reversal.

45The multiplicity of mechanisms that regulate HIV-1 latency and the diversity of factors responsible for the heterogeneity of the latent HIV-1 reservoir most likely vary from one patient to the other and even from one cell to the other in a single patient. Indeed, several single-cell studies reported cell-to-cell variability of the latent reservoir and the heterogeneity of cellular response to LRAs. A recent single-cell transcriptome profiling study from Angela Ciuffi laboratory has demonstrated that latently-infected cells are transcriptionally heterogeneous and can be separated into two different cell clusters based on their cellular states. These distinct states correlated with the susceptibility to cellular activation and HIV-1 reactivation, highlighting that the cellular environment could also contribute to the success of HIV-1 reactivation strategies.

46In the frame of the “shock and kill” strategy, clinical trials using LRAs have so far produced unconvincing results. This strategy faces multiple barriers which prevent the complete eradication of replication competent viruses of the HIV-1 reservoir and must, therefore, be optimized. Targeting and reactivating latent cells is challenging due to the heterogeneous nature of the viral reservoirs. Recent studies demonstrating diverse responses of infected cells to LRAs point to their weak effect and highlight the diversity of determinants responsible for the reservoirs’ heterogeneity that were demonstrated so far to be virus genetic background-, cell model-, cell type-, silencing mechanism-, tissue reservoir-, integration site-, patient-, and gender- specific. In addition, some studies demonstrate a heterogeneous effect of LRAs on NK cells and cytotoxic T-cell lymphocyte activity with conflicting observations, suggesting either an immunosuppressive effect or a reduced impact of LRA activity on cells sensing HIV-1 reactivation. Moreover, prolonged ART treatment is associated with a significant reduction in the frequency of HIV-1-specific CD8+ T-cells. Thus, as demonstrated by the group of Robert Siliciano, stimulating HIV-1-specific CTLs prior to reactivating latent HIV-1 should be considered for a successful eradication in future clinical trials. Another determinant of the effectiveness of a given LRA in reactivating and purging the viral reservoirs is its ability to efficiently induce latent HIV-1 by targeting not only transcriptional but also post-transcriptional mechanisms that need to be considered for a combined approach of more potent latency reversal. Additionally, an effective LRA needs to penetrate the multitude of HIV-1 tissue reservoirs and sanctuary sites. For instance, it has been shown that panobinostat did not sufficiently penetrate the central nervous system and romidepsin’s concentration in CSF of non-human primate was only 2% of the level found in plasma. Consequently, ensuring a better tissue penetration of LRAs by enhancing the drug delivery system, and most importantly strengthening the killing of the LRA-reactivated cells by stimulating CD8+ T responses are essential for the eradication strategy. Rational design of LRAs considering all these determinants is not possible at the moment, due to the lack of knowledge of all the cellular factors and pathways impacting HIV-1 gene expression and leading to productive viral replication. New approaches including single-cell technologies should be considered to understand why some cells respond to LRAs while others do not as this will be essential for improving the “shock and kill” strategy and hopefully reaching a cure. Moreover, due to the challenges that hinder the effectiveness of the “shock and kill” approach, some attention is also given to strategies aimed at completely suppressing HIV-1 transcription named the “block and lock.” In this context, the Tat inhibitor didehydro-cortistatin A (dCA) has shown its ability to inhibit ex vivo residual viral replication under ART and to prolong the time to viral reactivation after treatment interruption, but some in vitro resistance mutations to dCA were reported. Currently, several drugs are identified for their ability to be used as LPA (Latency Promoting Agents) to ensure a functional cure. Undeniably several strategies must be exploited in order to reach a functional cure. Dealing with the residual viremia and the contribution of ongoing viral replication in the reservoir’s replenishment is one of the major issues. If one of these strategies is promising its efficacy in clinic will be a long process and should, therefore, lead to the formation of a deep latency state preventing viral rebound after ART interruption.

47AA-A, AK, GD, RV, and CV wrote the manuscript. SD, VG, PM, AM, and OR gave advice and suggestions for the writing of the manuscript.

48The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.