Since the identification of HIV-1 in 1983 (ref.1), there have been major advances in the treatment and prevention of infection using antiretroviral therapy (ART). However, despite the effectiveness of ART, there is currently no cure for HIV-1 infection and life-long treatment is necessary to maintain suppression of HIV-1 replication owing to viral persistence in a latent reservoir2,3,4,5. The latent reservoir is a long-lived population of resting CD4+ T cells that harbour intact proviruses that are transcriptionally silent but retain the ability to produce infectious virions6,7,8. There are various methods used to quantify the size of the latent reservoir and the way in which reservoir size is estimated has implications for the interpretation of results (Box 1). Latently infected cells can persist through clonal expansion resulting from homeostatic, antigen-driven or integration site-driven proliferation9, which is a major barrier to the elimination of latently infected cells. Therefore, for most people living with HIV-1 (PLWH), the cessation of ART leads to viral rebound, typically after approximately 2 weeks10.

There are many proposed strategies for HIV-1 cure. These include approaches to achieve either control of viral replication without ongoing treatment or a complete elimination of infectious virus. The former strategy may involve therapeutic vaccines to promote immune responses against HIV-1 and thus allow for sustained immune-mediated suppression of viraemia. By contrast, the latter strategy will require a significant, multi-log reduction in the size of the latent reservoir to prevent viral rebound11. Recent efforts to eliminate the latent reservoir have focused on the ‘shock and kill’ strategy. This two-step strategy uses latency-reversing agents (LRAs) to ‘shock’ latently infected cells, thereby inducing the transcription of HIV-1 genes. This, in turn, results in the death of infected cells owing to viral cytopathic effects or it allows for immune cells to recognize and kill infected cells9,12 (Fig. 1).

Fig. 1: ‘Shock and kill’ strategies to eliminate CD4+ T cells latently infected with HIV-1.
figure 1

Latently infected CD4+ T cells contain the HIV-1 genome integrated into the host cell genome and remain undetectable by other immune cells owing to the lack of HIV-1 gene expression. Antiretroviral therapy (ART) prevents active viral replication but is unable to eliminate latently infected cells. Toll-like receptor 7 (TLR7) agonists and dendritic cells presenting cognate antigen can help to reactivate (‘shock’) infected cells and induce the transcription of HIV-1 genes, leading to the production of viral proteins. The infected cells can then be recognized and ‘killed’ by cytolytic effectors such as cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells. These effectors release granzymes and perforin to induce apoptotic cell death, leading to the elimination of infected cells. A reduction in the size of the latent reservoir is characterized by a decrease in the number of CD4+ T cells harbouring intact proviruses (Box 1).

Many LRAs have been identified in in vitro studies and several of these have been tested in vivo13. The LRAs tested in vivo have largely focused on reversing epigenetic blocks to HIV-1 gene expression using histone deacetylase inhibitors such as vorinostat, romidepsin and panobinostat14,15,16. Other latency-reversing strategies involve inducing some degree of T cell activation, for example, through protein kinase C agonists17. In resting CD4+ T cells, key host transcription factors necessary for HIV-1 gene expression are absent or sequestered in an inactive form. Activation of T cells allows for nuclear translocation of host transcription factors involved in the initiation of HIV-1 gene transcription, such as NF-κB and NFAT, as well as recruitment of the host transcription elongation factor P-TEFb18,19,20. In addition, recent evidence has shown that some LRAs might mediate their effects through stress response pathways involving the transcription factor HSF1, which can help to recruit transcription elongation factors to the HIV-1 promoter21,22. However, compared with global T cell activation, most candidate LRAs are ineffective at reversing latency ex vivo in cells from PLWH and, so far, they have failed to reduce the size of the latent reservoir in vivo23,24,25,26. Moreover, both HIV-1 infection27,28,29 and treatment with specific LRAs30,31 have been previously shown to impair the functions of CD8+ T cells and natural killer (NK) cells, both of which are involved in the clearance of HIV-1-infected cells. Therefore, novel therapeutic approaches are being explored for both reactivation and elimination of the latent reservoir of HIV-1 (Table 1).

Table 1 Recently completed and ongoing clinical trials for HIV-1 cure targeting innate immunity

The innate immune system is the first line of defence against foreign pathogens. However, the role of innate immunity during acute and chronic HIV-1 infection has not yet been fully explored. Importantly, recent studies show that both transmitted/founder viruses and viruses that are present in the plasma during initial rebound after cessation of ART have high levels of resistance to type I interferon responses32,33 (Box 2), which indicates that interferon-sensitive viral variants are inhibited and highlights the importance of targeting the innate immune system to achieve a cure. A better understanding of how HIV-1 escapes immune control could be achieved through further examination of the resistance of reservoir viruses to key elements of the innate and adaptive immune responses.

As the mechanism of recognition of HIV-1 infection by the innate immune system is not as well understood as recognition by the adaptive immune system, innate immunity has so far received less research attention in terms of HIV-1 cure strategies. However, it is becoming clear that innate immune pathways may contribute to both the reversal of HIV-1 latency and the subsequent elimination of infected cells. Indeed, some current trials of HIV-1 cure strategies are using therapeutics that engage aspects of the innate immune response to reactivate or eliminate latently infected cells (Table 1). In this Review, we discuss various approaches for engaging components of the innate immune system to eradicate the latent reservoir of HIV-1, focusing on the potential roles of dendritic cells (DCs) and NK cells.

Reactivating latently infected cells

DCs are a diverse population of innate immune cells with a unique ability to link innate and adaptive immune responses34. Based on the differential expression of key transcription factors, DCs are classified into three major subsets: plasmacytoid DCs (pDCs) and two types of myeloid conventional DCs (cDC1s and cDC2s)35,36. pDCs recognize viral nucleic acids in the form of single-stranded RNA or unmethylated CpG-rich DNA through their endosomal Toll-like receptors (TLR7 or TLR9, respectively). Activation of these TLRs in pDCs induces the transcription of type I interferon genes as well as of genes encoding other pro-inflammatory cytokines, such as CC-chemokine ligand 4 (CCL4), IL-6 and tumour necrosis factor (TNF), that contribute to the activation of other immune cell types37,38,39 (Fig. 2). By contrast, cDCs are highly efficient at capturing and processing antigens in peripheral tissues, trafficking to draining lymph nodes, and presenting the antigens to naive T cells together with costimulatory and polarizing signals40,41.

Fig. 2: Ex vivo generation of monocyte-derived dendritic cells for therapeutic vaccination.
figure 2

Monocytes from people living with HIV-1 (PLWH) are cultured with granulocyte–macrophage colony-stimulating factor (GM-CSF), IL-4, additional maturation-associated cytokines and sources of HIV-1 antigens to generate IL-12p70-producing mature monocyte-derived dendritic cells (moDCs). These DCs loaded with HIV-1 antigens ex vivo are returned to the donor to induce HIV-1-specific CD8+ T cell responses (cytotoxic T lymphocyte (CTL) responses) and overcome dysfunctional antigen presentation by conventional DCs (cDCs). Recent DC-based vaccination efforts have also involved the transduction of moDCs with CD40L as a strategy to enhance their maturation, increase IL-12p70 production and aid effective antigen presentation to T cells. Emerging studies also suggest that moDCs can induce the activation of CD4+ T cells and implicate moDCs in the induction of latency reversal. One recent study brought attention to canonical antigen presentation by DCs and CD40–CD40L signalling in latency reversal but the mechanism remains unexplained. Other strategies to reactivate latently infected CD4+ T cells involve the use of Toll-like receptor 7 (TLR7) agonists (such as GS-986 and GS-9620) to induce a type I interferon (IFN) response by plasmacytoid dendritic cells (pDCs) and the production of other pro-inflammatory cytokines associated with the antiviral response (such as CCL4 and IL-6) to activate other immune cells. TCR, T cell receptor.

The mechanisms by which HIV-1 avoids detection by DCs at multiple stages of infection have been described in detail elsewhere42,43,44. These reviews offer insights into approaches to enhance the innate sensing pathways of DCs and thereby improve adaptive immunity to HIV-1. Here, we describe strategies that engage innate immunity through manipulating DC responses to reactivate latently infected cells and improve the efficacy of HIV-1 cure efforts.

Toll-like receptor agonists

Of particular interest with respect to HIV-1 latency is TLR7, which recognizes single-stranded RNA in endosomes and can also be activated by small molecule agonists such as imidazoquinoline analogues or guanosine cyclic triphosphates45,46,47. Although, in humans, TLR7 is predominantly expressed by pDCs (Box 2) and B cells48,49, rather than by CD4+ T cells, several studies indicate that TLR7 agonists can affect the latent reservoir of HIV-1 by enhancing innate immune responses. In a study of peripheral blood mononuclear cells isolated from PLWH, treatment with the TLR7 agonist GS-9620 (vesatolimod) induced HIV-1 RNA production, which is suggestive of active transcription of HIV-1 genes from reactivated latently infected cells50. The depletion of pDCs from the culture, which decreased IFNα production in response to GS-9620, resulted in reduced activation of both CD4+ and CD8+ T cells. In addition, blocking the IFNα receptor in combination with GS-9620 treatment of CD4+ T cells did not result in increased HIV-1 RNA expression. Together, the results suggest that the latency reversal mediated by GS-9620 depends on the presence of pDCs and IFNα secretion50.

Rhesus macaques infected with simian immunodeficiency virus (SIV) establish a latent reservoir in resting CD4+ T cells, which makes these non-human primates an accepted in vivo model to study various aspects of the latent reservoir relevant to HIV-1 (ref.51). SIV-infected macaques on suppressive ART given repeated and escalating doses of the TLR7 agonist GS-986 (a close analogue of GS-9620) had transiently increased levels of SIV RNA in plasma into the detectable range52. In addition, treatment with GS-986 in combination with an adenovirus-based therapeutic vaccine (Ad26/MVA) expressing SIV gag, pol and env genes decreased levels of viral DNA in both peripheral blood mononuclear cells and lymph nodes, which suggests that there was a reduction in the size of the latent reservoir53. Furthermore, rhesus macaques that received both the Ad26/MVA vaccine and GS-986 not only had a significantly lower plasma SIV RNA setpoint (the value at which the viral load remains relatively stable) but also a 2.5-fold delay of viral rebound after cessation of ART53. The results suggest that the combination treatment generated SIV-specific cytotoxic T lymphocytes (CTLs) that recognize and kill reactivated latently infected cells. In a following study of macaques infected with simian–human immunodeficiency virus (SHIV) and treated with ART during acute infection54, GS-9620 was given in combination with a broadly neutralizing antibody (bNAb), PGT121, that recognizes N-linked glycans on the V3 loop of HIV-1 gp120 to enable antibody-dependent cellular cytotoxicity (ADCC). Macaques treated with both PGT121 and GS-9620 had significantly lower levels of viral DNA, indicating a potential reduction in the size of the latent reservoir; following discontinuation of ART, 5 of 11 animals receiving this combination treatment experienced a delay in viral rebound54. All three of these studies in rhesus macaques showed that treatment with TLR7 agonists can activate CD4+ T cells, CD8+ T cells and NK cells, and increase plasma levels of various cytokines and chemokines, including IFNα, IL-6 and IL-1β52,53,54. These results suggest that TLR7 agonists can induce a potent innate immune response and, in combination with an HIV-1-specific bNAb or therapeutic vaccine, could improve the elimination of latently infected cells. However, it remains unclear whether the beneficial effects of TLR7 agonists in these studies result from their activity as LRAs, their immunomodulatory capacity as vaccine adjuvants or both.

GS-9620 was also evaluated in two phase Ib clinical trials for its safety and effect on the latent reservoir of HIV-1. Treatment of PLWH on suppressive ART with GS-9620 was well tolerated but had minimal effects on plasma viraemia and reservoir size, with no significant changes from baseline measurements of plasma HIV-1 RNA observed55. However, treatment of ART-suppressed HIV-1 controllers with GS-9620 resulted in a modest delay in viral rebound after analytical treatment interruption compared with the placebo group56. Although there were no differences in total HIV-1 DNA between the placebo group and those receiving GS-9620, the intact proviral DNA assay (Box 1) detected a significant decrease in intact, replication-competent proviruses in the latent reservoir of GS-9620-treated HIV-1 controllers56 (Table 1). Although both studies55,56 observed increased expression of interferon-stimulated genes, serum cytokines and IFNα in response to GS-9620 treatment, only the treatment of HIV-1 controllers resulted in a decrease of intact proviruses in the latent reservoir56. The differing results between these two studies could be due to variations in participant HIV-1 controller status as HIV-1 controllers may have CD8+ T cells with enhanced polyfunctional responses compared with non-controllers. In addition, continued ART administration after GS-9620 treatment in one of the studies55 could have affected virus–host dynamics.

Treatment with a TLR7 agonist caused transient increases in SIV RNA in plasma that were suggestive of latency reversal in one study of non-human primates52; this latency-reversing activity may have been undetected or underestimated in cases where very early treatment with ART after infection resulted in a small latent reservoir53,54. Furthermore, differences in the challenge strain of SIV or SHIV, in dosing intervals and in the immune status of the macaques could affect the ability of TLR7 agonists to mediate latency reversal. The ability of TLR7 agonists to potently stimulate IFNα production by pDCs suggests a potential mechanism for indirectly activating HIV-1-infected CD4+ T cells50,57. In addition, the increased production of IFNα in response to TLR7 agonists may also indirectly enhance the cytolytic activity of NK cells (see later discussion), suggesting a role for DC–NK cell crosstalk in latency reversal56. Overall, TLR7 agonists, either in combination with a therapeutic vaccine to strengthen HIV-1-specific T cell responses or in combination with bNAbs to enable ADCC, have the potential to improve the elimination of HIV-1-infected cells. Agonists of other TLRs, such as TLR1/2, TLR5, TLR8 and TLR9, have also been shown to have modest latency-reversing effects in HIV-1-infected cell lines or HIV-1-infected central memory T cells58,59; these warrant further study, for example given the far broader effects of TLR8 agonists on innate immunity than those of TLR7 agonists. In summary, TLR agonists have proven to be potent activators of multiple immune functions and may aid HIV-1 cure strategies designed to reduce the size of the latent reservoir or maintain long-term virological control.

Using DCs for latency reversal

DCs could also contribute to the reactivation of latently infected cells during ‘shock and kill’ approaches through their antigen presentation function. For example, two vaccine studies using DCs expressing HIV-1 antigens have reported induction of CD4+ T cell activation and increases in viral gene expression in vivo60 and in vitro61. Through their effective presentation of antigen, DCs can contribute to the activation of latently infected CD4+ T cells that are specific for the presented antigen40,41,62. However, this approach of reactivating antigen-specific latently infected cells has several limitations, including the small fraction of infected cells specific for any given antigen and the fact that the antigen specificity of most latently infected cells is unknown.

The addition of autologous immature monocyte-derived DCs (iDCs) to a primary cell model of latent HIV-1 infection caused latency reversal by a contact-dependent mechanism that may have involved factors independent of canonical T cell receptor (TCR) stimulation63. Activation of infected CD4+ T cells through the TCR and subsequent stimulation with iDCs increased extracellular levels of HIV-1 RNA and induced more viral outgrowth compared with TCR stimulation or iDC stimulation alone. This DC-mediated latency reversal may have involved contact-induced activation of the PI3K–AKT–mTOR pathway in CD4+ T cells, which was necessary for potent viral reactivation. Furthermore, a separate study using the same primary cell model of HIV-1 latency to investigate the effects of different DC subsets and other immune cell types on latency reversal showed that tissue-resident and blood-derived myeloid DCs induce viral reactivation with different efficiencies64. However, these findings should be interpreted with caution considering that the latency model used in these studies was generated under conditions that do not reflect naturally occurring establishment of HIV-1 latency. In addition, these studies did not show whether DCs were presenting HIV-1 antigens and therefore affected HIV-1-specifc CD4+ T cells through an antigen-dependent mechanism62,63. Studies using primary cell latency models to investigate DC-induced reversal of HIV-1 latency have so far yielded inconclusive data owing to the fact that these model systems are more efficient to reactivate than latently infected cells from PLWH23,24. Thus, the role of DCs in latency reversal requires further investigation using cells from PLWH. One recent study using cells from ART-suppressed PLWH reported that using monocyte-derived DCs (moDCs) to present peptide pools of HIV-1 Gag or of cytomegalovirus to autologous CD4+ T cells resulted in latency reversal65. This effect was dependent on bidirectional crosstalk between moDCs and CD4+ T cells during their canonical antigen-driven interaction, including CD40–CD40 ligand (CD40L) signalling65. However, the latency reversal observed in this study might be influenced by the fact that prolonged cultures were maintained in the absence of antiretroviral drugs, which enabled viral spread from reactivated cells. Indeed, the authors acknowledge that moDC-mediated viral reactivation might have been an indirect result of a potent global antigen-specific response rather than a direct effect on HIV-1-infected cells in the reservoir; thus, the underlying mechanism of moDC-mediated latency reversal remains unclear. Further investigation of the antigen specificity of the latent reservoir may improve DC-based latency reversal strategies by allowing for the reactivation of latently infected cells through the presentation and recognition of cognate antigen.

DC vaccines to enhance CTL responses

Recent studies suggest that depleting the latent reservoir of HIV-1 may require a combination therapeutic approach with both ‘shock’ and ‘kill’ components. Latency reversal by itself may not necessarily be followed by the clearance of cells expressing HIV-1 antigens12,66,67. Novel candidate vaccines and immunotherapies designed to enhance the killing of infected cells following latency reversal have been tested in various clinical trials without any clear promising results68,69,70. Targeting DCs may offer a path to inducing HIV-1-specific CTLs as DCs naturally prime both CD4+ and CD8+ T cells62. DCs have been incorporated in immunotherapeutic strategies for both cancer and HIV-1 (refs71,72). However, there are multiple challenges associated with the effective targeting of infected cells by CTLs73, including spatial separation of CTLs and target cells74, CTL exhaustion28, impaired recognition of HIV-1 epitopes owing to escape mutations66,75, and intrinsic resistance of latently infected cells to cytolysis76. Another major obstacle for successful DC-based therapeutic vaccination is endogenous DC dysfunction during HIV-1 infection77. ART can restore the function of endogenous DCs although some functions, such as IL-12p70 secretion in response to certain stimuli, are only partially rescued78.

Vaccination with autologous antigen-loaded DCs

One approach to overcoming the problem of endogenous DC dysfunction and improving vaccine-induced immune responses has been to generate mature, antigen-loaded autologous DCs ex vivo and then reintroduce these DCs to PLWH as a form of adoptive immunotherapy. This has involved generating fully matured and functional type-1-polarized moDCs that express both CCR7, which is necessary for their migration to lymph nodes for antigen presentation to T cells, and IL-12p70, which promotes CD4+ T helper 1 cell responses79. Ex vivo generation of large numbers of autologous moDCs requires the culture of monocytes from PLWH with granulocyte–macrophage colony-stimulating factor (GM-CSF) and IL-4 (ref.80). The addition of IFNγ and a combination of maturation-associated cytokines to ex vivo-generated moDCs produces polarized DCs capable of increased IL-12p70 expression81,82. DCs of this phenotype are associated with sustained HIV-1-specific effector T cell responses83.

Reinfusion of autologous moDCs presenting HIV-1 antigens into PLWH has been shown to be safe and immunogenic in several trials68,69,70. For these studies, various HIV-1 antigens, including whole inactivated virus, recombinant envelope glycoproteins, viral peptides or recombinant viral vectors, were used to prime ex vivo-generated moDCs60,68,69,70,84,85 (Fig. 2). The most promising results were obtained by using whole inactivated autologous virus as the antigen source68,69,70. In one study, vaccination with moDCs pulsed with autologous heat-inactivated HIV-1 resulted in a small yet significant decrease in viraemia and increased HIV-1-specific T cell responses in ART-naive individuals at 24 weeks after vaccination69. In a follow-up study, the same DC vaccination strategy was tested in PLWH treated with ART who underwent several analytical treatment interruptions after vaccination70. A log10 decrease in HIV-1 RNA setpoint was achieved70, although viral rebound after stopping ART was detected in all study participants. However, the reservoir size at the time of the second treatment interruption after vaccination was not measured and the authors note that potential replenishment of the latent reservoir during the first analytical treatment interruption is a major problem with the study design.

It is important to also acknowledge the practical limitations of this vaccination approach for a larger scale-up, particularly with regard to obtaining large numbers of moDCs. In addition, the choice of immunogen can affect optimal maturation of moDCs and the propagation of IL-12p70-producing moDCs. Although using whole inactivated autologous virus as a source of antigen for loading DCs may have an advantage over other immunogens in terms of the presentation of relevant, donor-specific epitopes, the practicality of this approach on a large scale is a concern. Furthermore, there is the additional concern of ensuring that the virus preparation has been completely inactivated and is no longer infectious.

Novel approaches to DC-based vaccination

To facilitate the generation of IL-12p70-producing moDCs, new strategies have been developed for the ex vivo genetic manipulation of DCs. One such approach towards improving DC-based vaccines involves genetic delivery of CD40L to DCs86,87. CD40L, which is transiently expressed on the surface of activated T cells and B cells88,89, binds to its cognate CD40 receptor on the surface of DCs90. This CD40–CD40L interaction induces the upregulated expression of costimulatory molecules by DCs and their secretion of cytokines such as IL-12p70, which are important for the maturation of DCs into fully competent antigen-presenting cells91 and the effective priming of T cells. Exploiting the knowledge of this interaction has led to the development of a personalized DC-based immunotherapy, termed AGS-004, which consists of autologous moDCs co-electroporated with RNAs encoding human CD40L and autologous HIV-1 antigens (Gag, Vpr, Rev and Nef)92,93,94. In one clinical trial, this approach was shown to induce HIV-1-specific CD8+ T cell effector responses, although these responses were not effective compared with placebo in improving host control of viral replication as measured during an analytical treatment interruption92. A recent study (Table 1) investigated the combined effect of the LRA vorinostat and AGS-004 on the latent reservoir of HIV-1 (ref.94). Six individuals on suppressive ART received four doses of AGS-004 every 3 weeks over a 12-week period, followed by vorinostat treatment every 72 hours for 30 days. Although the treatment was well tolerated, there was no effect on the latent reservoir as measured using the quantitative viral outgrowth assay (Box 1). Furthermore, no increase in HIV-1-specific CD8+ T cell or NK cell responses was observed. There are several possible explanations for the absence of a treatment effect in this study. First, the study had a small sample size and lacked a control group. Second, although some in vivo studies have shown that vorinostat treatment causes a significant increase in HIV-1 RNA expression indicative of the reactivation of latently infected cells67,95, there is no evidence that treatment with vorinostat can induce sufficient expression of HIV-1 genes from replication-competent proviruses and effective antigen presentation to allow for the elimination of HIV-1-infected cells by autologous CD8+ T cells96. In addition, the absence of detectable HIV-1-specific CD8+ T cell responses could be a result of CTL escape variants of HIV-1 being dominant in the latent reservoir of the participants in this study75.

Overall, although DC-based immunotherapies for HIV-1 infection have proven safe, well tolerated and moderately immunogenic, they have had limited success in clinical trials in terms of reducing the size of the latent reservoir97. Methods of DC generation and preparation, choice of immunogen, immunological parameters to measure viral control, and the assessment of immunological and virological responses have varied between studies, which makes it difficult to compare results60,68,69,70,84,94. One possible explanation for a lack of clinical efficacy for DC-based therapies is the limited capacity of DCs to induce potent HIV-1-specific CD8+ T cell responses owing to CTL exhaustion. In this respect, the combination of DC-based vaccines with immune-checkpoint inhibitors, such as antibodies to PD1 or CTLA4, is currently being investigated in several clinical trials for solid tumours98,99, which could improve the effector functions of CD8+ T cells through reversal of exhaustion100. Furthermore, the incomplete maturation of DCs during vaccine production could result in DCs with a limited capacity to migrate to the lymph nodes and that have a tolerogenic phenotype101. As an alternative to the laborious process of generating DC-based vaccines in vitro, the cross-presentation capacity of native circulating cDC1s could be enhanced by targeted antigen delivery to these cells. For example, some studies have reported that tumour antigen-conjugated antibodies recognizing C-type lectin receptors, such as DEC-205 or CLEC9A, that are highly expressed on cDC1s elicit robust tumour-specific CTL responses99,102,103. Therefore, we anticipate that rather than DC-based vaccine monotherapies, combinatorial strategies that incorporate DC-based vaccination or approaches to improve the antigen presentation of native cDCs will be more effective for HIV-1 cure efforts. A better understanding of the factors leading to impairment of DC responses in PLWH may contribute to the design of effective adjuvants or vectors that can rescue DC function in vivo and improve current DC-based vaccine strategies for the treatment of HIV-1 infection.

Killing of infected cells by NK cells

NK cells are a subset of lymphoid cells that mount innate antiviral responses. In response to infection, NK cells can directly lyse infected cells and can also secrete pro-inflammatory cytokines that enhance the functions of other innate and adaptive immune cells. A protective role for NK cells in HIV-1 infection is suggested by their ability to drive viral escape mutations and by the existence of several conserved mechanisms that HIV-1 has evolved to avoid detection by NK cells. For example, one study proposed that the immune pressure generated by NK cell-mediated ADCC contributed to the selection of mutations in HIV-1 epitopes recognized by antibodies that mediate ADCC104. These mutations could not be attributed to escape from antibody-mediated neutralization as the antibodies involved did not have neutralizing activity105. Similarly, selective pressure may alter the repertoire of viral peptides presented on MHC class I molecules to favour the engagement of inhibitory members of the killer immunoglobulin-like receptor (KIR) family on NK cells105.

Another mechanism by which HIV-1-infected cells might be recognized by NK cells involves the G2 cell-cycle arrest activity of the HIV-1 protein Vpr. Vpr promotes the activation of DNA damage and stress response pathways, resulting in increased surface expression of the stress proteins ULBP1 and ULBP2 by infected cells, which are ligands for the activating NK cell receptor NKG2D and are sufficient to induce NK cell-mediated killing of target cells106,107. However, HIV-1 avoids the recognition of infected cells by NKG2D through Nef-mediated degradation of ULBP1, ULBP2 and an additional NKG2D ligand, MICA108. To avoid the killing of infected cells by CTLs, HIV-1 also downregulates surface expression of HLA-A and HLA-B in a Nef-dependent manner109,110. To counteract the loss of interaction with inhibitory KIRs on NK cells caused by downregulation of HLA-A and HLA-B, HIV-1 is thought to selectively maintain surface expression of HLA-C and HLA-E111,112, which interact with additional inhibitory KIRs and NKG2A on NK cells, respectively. It is, however, worth noting that most studies that have reported the maintenance of surface expression levels of HLA-C and HLA-E by HIV-1-infected cells have used laboratory-adapted HIV-1 strains, which may not reflect the activity of primary isolates. More recent studies have suggested that, in cells infected with primary HIV-1 strains, Vpu and Nef may cause the downregulation of HLA-C and HLA-E, respectively113,114,115.

Further support for a role of NK cells in inhibition of HIV-1 infection comes from evidence linking NK cell-mediated ADCC to protection from contracting HIV-1 infection and suppression of viraemia. Post hoc analysis of the clinical trial results of RV144, the only HIV-1 vaccine trial so far to have resulted in some degree of protective efficacy, showed that one of the strongest positive correlates of protection is a high titre of non-neutralizing antibodies to the V1V2 region of the HIV-1 Env protein that can mediate ADCC116,117. However, additional analysis suggests that such antibodies may have multiple Fc-mediated effector functions, such as antibody-dependent cellular phagocytosis, in addition to ADCC118,119. HIV-1-specific ADCC responses have also been detected in the acute phase of HIV-1 infection and may contribute to determining the setpoint viral load120,121. Finally, high levels of HIV-1-specific ADCC activity correlate with slower rates of disease progression122,123.

Although the data are controversial, there is some evidence that ADCC responses may be increased in HIV-1 controllers. In several cohorts, sera from HIV-1 controllers enhanced ADCC responses to a greater extent than did sera from viraemic progressors124,125,126. Furthermore, the breadth of ADCC-mediating antibodies may also contribute to viral control as HIV-1 controllers in one cohort had greater cross-reactivity of ADCC responses against multiple Env clades than did viraemic progressors125. By contrast, no differences in ADCC responses were detected between HIV-1 controllers and viraemic progressors in additional cohorts127,128,129. Definitive conclusions about the role of ADCC in HIV-1 controllers will likely require standardized definitions of HIV-1 control and standardized methodologies for detecting ADCC activity to allow for better comparison between studies.

Overall, the above studies suggest a role of NK cell-mediated cytolytic clearance of infected cells in inhibiting HIV-1 acquisition and disease progression. The ability of NK cells to potently kill HIV-1-infected CD4+ T cells in diverse biological contexts warrants further study into the design of strategies to augment the innate antiviral properties of NK cells for an HIV-1 cure. Immunotherapies that improve the existing capacity of NK cells to eliminate HIV-1-infected cells in vivo may be effective alone or in combination with CTL-based strategies to reduce the size of the latent reservoir.

Enhancing ADCC activity through passive antibody administration

Passive administration of antibodies to Env that are capable of mediating strong ADCC is a strategy under consideration to enhance the clearance of HIV-1-infected cells (Fig. 3). In vitro studies have shown that several well-characterized monoclonal bNAbs can facilitate the killing of HIV-1-infected cells through NK cell-mediated ADCC130,131,132. Evidence from treatment with bNAbs in models of HIV-1 or SHIV challenge suggests that bNAbs can also mediate the elimination of infected cells in vivo. bNAbs consistently protect from SHIV challenge in macaque models and potently reduce viraemia133,134,135,136,137,138,139. Some of this effect is clearly owing to the direct antibody-mediated neutralization of virus. However, mutation of the bNAb b12 to abrogate Fc receptor binding reduced its ability to prevent SHIV acquisition133,134, which implicates a role for Fc-mediated effector functions, such as ADCC, in this activity. Furthermore, mutation of the bNAb VRC07-523LS139 and the bNAb-derived bispecific antibody 117/1,400 (ref.140) to abrogate Fc receptor binding slows the rates at which these antibodies clear virus from the plasma of SHIV-infected macaques. An additional study in a humanized mouse model of HIV-1 challenge has shown that mutations in bNAbs that enhance Fc receptor binding result in greater protective activity, providing further evidence for the involvement of Fc-dependent mechanisms in bNAb-mediated protection against HIV-1 infection141. In PLWH, bNAb infusions have been well tolerated and have delayed rebound of viraemia following analytical treatment interruption142,143,144,145. In some individuals who receive bNAb monotherapy during analytical treatment interruption, the outgrowth of pre-existing bNAb-resistant escape variants of HIV-1 can occur142,143,144. However, this phenomenon was not observed when a combination of bNAbs targeting distinct epitopes of HIV-1 was administered during treatment interruption145. Disappointingly, however, bNAb treatment during analytical treatment interruption has not resulted in an observable reduction in size of the latent reservoir of HIV-1 (refs146,147). This may be owing to the infrequent reactivation of latently infected cells and, thus, pairing the administration of bNAbs with a strategy to reverse latency may improve the clearance of infected cells. It should also be noted that it is unclear to what extent ADCC or other Fc-dependent antibody effector functions may contribute to the observed delay in viral rebound after analytical treatment interruption in PLWH who received bNAb infusions because Fc-mutated variants of the bNAbs have not been evaluated in this context.

Fig. 3: Strategies to enhance natural killer cell-mediated killing of HIV-1-infected cells.
figure 3

Various broadly neutralizing antibodies (bNAbs) and non-neutralizing antibodies (nNAbs) to HIV-1 Env protein bind the Fc receptor FcγRIIIa (CD16a) on the surface of natural killer (NK) cells through their Fc domains. In addition, bispecific antibodies (BsAbs) can bind both Env on the surface of HIV-1-infected CD4+ T cells and FcγRIIIa on NK cells. This interaction results in NK cell activation, release of cytolytic granules (containing perforin and granzymes) and pro-inflammatory cytokines (such as IFNγ and tumour necrosis factor (TNF)) and apoptosis of the infected CD4+ T cell through a process known as antibody-dependent cellular cytotoxicity. Binding of CD4 mimetic compounds to Env can stabilize it in an ‘open’ conformation, thereby exposing CD4-induced (CD4i) epitopes and rendering the infected cell more vulnerable to binding of serum and monoclonal antibodies to CD4i epitopes and subsequent antibody-dependent cellular cytotoxicity. Cytokine-based therapies, including pegylated IFNα and the IL-15 super-agonist ALT-803, sensitize NK cells to cytolytic killing of HIV-1-infected cells. IFNAR, interferon-α/β receptor; NKR, natural killer cell receptor.

Similarly to bNAbs, monoclonal non-neutralizing antibodies to the HIV-1 Env protein mediate ADCC in vitro, albeit with generally weaker activity130,148. Of particular interest are V1V2-targeting non-neutralizing antibodies, similar to those antibodies that were thought to be protective in the RV144 vaccine trial. One study found that V1V2-targeting non-neutralizing antibodies have highly potent ADCC activity, greater in magnitude than that observed with tested bNAbs149. However, in contrast to bNAbs, the administration of non-neutralizing antibodies failed to protect macaques from SHIV challenge150,151,152,153. Although the non-neutralizing antibodies were not protective against infection, treatment was associated with positive outcomes in some of these studies, including reductions in viral load150,153 and levels of cell-associated HIV-1 DNA153 as well as the clonal restriction of transmitted/founder viruses151. Given that non-neutralizing antibodies cannot inhibit infection by cell-free virions, these treatment effects may be attributed to Fc-mediated effector activity against infected cells.

Engineering antibodies for improved NK cell cytolytic activity

Advances in antibody engineering have made it possible to modify existing monoclonal antibodies to better elicit Fc-mediated effector functions. Mutational analysis of the IgG Fc domain has revealed combinations of point mutations that result in improved Fc receptor binding and ADCC activity compared with the wild-type Fc domain154,155,156. In addition to or as an alternative to modifying the protein sequence of the antibody Fc domain, alterations to the Fc glycan structure can enhance the ability of IgG antibodies to mediate ADCC. Compared with IgG antibodies expressed in commonly used cell lines (such as wild-type CHO or 293T cells), IgG antibodies expressed in cell lines that yield low Fc fucose content (such as Lec13, YB2/0 or FUT8 cells) have increased Fc receptor binding and ADCC activity157,158,159,160. The afucosylated forms of the bNAbs 2G12 and b12 have enhanced Fc receptor binding and ADCC activity compared with the fucosylated forms161,162. Unexpectedly, however, the afucosylated form of b12 did not improve protection from SHIV challenge compared with the fucosylated form162. It would be of considerable interest to compare an ADCC-enhanced bNAb to the wild-type version in the context of analytical treatment interruption for differences in efficacy in delaying viral rebound and reducing levels of cell-associated HIV-1 DNA.

In addition to these efforts to enhance the interaction between the IgG Fc domain and the Fc receptor FcγRIIIa (also known as CD16a), several groups have engineered bispecific antibodies that induce potent NK cell-mediated cytolytic activity (Fig. 3). Bispecific antibodies that engage Env on the surface of HIV-1-infected cells and CD3 on the surface of T cells have previously been shown to promote CTL-mediated killing of infected target cells163,164,165,166. Recent studies have shown that replacing the CD3-specific antibody variable fragment (Fv) of existing bispecific antibodies that engage CD3 and Env with a CD16-specific antibody Fv166 or fusing a CD16-specific antibody Fv to the carboxyl terminus of gp41-specific IgG antibodies167 enabled these molecules to effectively bind NK cells and enhance the killing of HIV-1-infected cells. The design of novel bispecific antibody reagents may contribute to more efficacious clearance of HIV-1-infected cells if incorporated into a ‘shock and kill’ strategy.

Sensitizing infected cells to ADCC

Recent work suggests that a specific class of small molecules may enhance the capacity of autologous antibodies to induce ADCC. Sera from PLWH contain a high concentration of antibodies that are potentially capable of mediating ADCC but that target Env epitopes that are conformationally occluded168. These epitopes, termed CD4-induced (CD4i) epitopes, are exposed only in the ‘open’ conformation of Env that is induced by CD4 binding168,169. The downregulation of CD4 expression in HIV-1-infected cells by Nef and Vpu is thought to minimize the interaction between Env and CD4 on the surface of infected cells, thus resulting in infrequent exposure of these CD4i epitopes and protection of the infected cells from ADCC mediated by CD4i-specific antibodies168,169. CD4 mimetic compounds can bind Env on the surface of infected cells in a similar manner to soluble CD4 and trigger the exposure of CD4i epitopes170, thereby enhancing the ADCC potential of serum antibodies against autologous infected CD4+ T cells171,172,173 (Fig. 3).

Cytokine-based therapies to enhance NK cell-mediated killing of infected cells

The incorporation of cytokine-based therapies into a ‘shock and kill’ regimen could be used to augment the clearance of HIV-1-infected cells by enhancing the cytolytic effector functions of innate immune cells. Of current interest are therapies based on IFNα or IL-15 (Table 1). The administration of pegylated IFNα2 to PLWH on ART has been shown to delay time to viral rebound upon analytical treatment interruption and reduce levels of cell-associated HIV-1 DNA as detected by Alu-Gag PCR174,175, prompting consideration for its use as part of a ‘shock and kill’ strategy. A growing body of evidence suggests that the ability of IFNα to decrease levels of cell-associated HIV-1 DNA results from increased NK cell-mediated cytotoxicity (Fig. 3). Low levels of cell-associated HIV-1 DNA in PLWH on ART who were administered IFNα2 correlated with changes in the expression of NK cell receptors, including an increased frequency of NK cells expressing the activating receptors NKp30 and NKG2D175. In addition, NK cells from individuals who maintained control of viral load throughout a 12-week analytical treatment interruption after receiving IFNα therapy had lower levels of expression of inhibitory KIRs and a superior capacity for cytolytic killing ex vivo176. In vitro treatment of NK cells from viraemic progressors and HIV-1 controllers with IFNα2 results in an increase of both cytokine production and cytolytic potential177.

Recently, there has been growing interest in using the IL-15 super-agonist ALT-803 in ‘shock and kill’ strategies for HIV-1 cure. IL-15 is naturally produced by macrophages and DCs and is presented by these cells in trans to neighbouring NK cells. IL-15 is known to stimulate various functions of NK cells that would be desirable in the context of a ‘shock and kill’ regimen, including cytotoxic activity, IFNγ production, ADCC and inhibition of HIV-1 outgrowth178. ALT-803 consists of a highly potent mutated form of IL-15 complexed with the sushi domain of IL-15 receptor-α fused to an IgG Fc domain179,180 (Fig. 3). In addition to improving the biodistribution and half-life of ALT-803, the Fc domain allows for binding to Fc receptors on the surface of macrophages and DCs, which mimics endogenous trans-presentation of IL-15 and further improves the potency of ALT-803 (ref.180). There is evidence from animal models of HIV-1 infection that ALT-803 enhances the activity of NK cells and CTLs. In a humanized mouse model of acute HIV-1 infection, ALT-803 inhibited viral replication and this inhibition was not observed in NK cell-depleted mice181. Furthermore, ALT-803 transiently reduced viral load in SIV-infected rhesus macaques, which was associated with an increased number of NK cells and CD8+ T cells in peripheral blood182. In SHIV-infected macaques, ALT-803 increased the migration of NK cells and SHIV-specific CD8+ T cells to lymph node follicles, sites that are known to harbour latent HIV-1 during ART and that normally exclude cytolytic effector cells183. In addition to its potential for enhancing the elimination of HIV-1-infected cells by cytolytic effector mechanisms, ALT-803 may also promote latency reversal. In SIV-infected or SHIV-infected macaques and HIV-1-infected humanized mice, the administration of ALT-803 paired with depletion of CD8+ T cells during ART led to detectable plasma viraemia184,185. The potential of this IL-15 super-agonist to function both in latency reversal and by enhancing the clearance of infected cells makes it an attractive candidate for ‘shock and kill’ strategies.

Concluding remarks

Emerging strategies for an HIV-1 cure include combination approaches that use LRAs coupled with methods to enhance immune effector functions. Although current efforts to reactivate latently infected cells are mainly focused on novel pharmacological LRAs, TLR agonists are promising candidates in this regard that may receive more attention in the future. Furthermore, recent studies reporting the potential of moDCs to reverse latency may offer insights into designing novel LRAs based on understanding the underlying mechanism of viral reactivation reported in those studies. The induction of a potent and broad immune response to HIV-1 infection might be achieved through the use of DC-based immunotherapies to enhance the killing of infected cells. However, there are many limitations to the use of DC-based therapies, including their large-scale production and the determination of which immune responses are the best measures of viral control. Future studies might focus, instead, on interventions that both activate and deliver antigen to DCs directly in situ. Another promising strategy to improve immune-mediated clearance of HIV-1-infected cells is to enhance the cytolytic capacity of innate immune effector cells such as NK cells. ADCC is a mechanism of cell lysis mediated by NK cells that could be exploited for the elimination of infected cells through the therapeutic administration of small-molecule pharmacological agents, monoclonal antibodies or recombinant proteins such as ALT-803. Clinical trials are already under way to study and apply innate immune-based therapies for HIV-1 cure (Table 1). With further continuation of clinical trials involving TLR7 agonists and ALT-803, we believe that it is of crucial importance to better understand the molecular mechanisms of latency reversal caused by these reagents and whether their effects are mediated by direct activation of latently infected cells or indirect engagement through an innate immune response. Better mechanistic understanding of these therapeutic agents may lead to more effective targeting of innate immune responses and inform the design of combinatorial approaches to an HIV-1 cure. We remain optimistic that progress will continue to be made in defining the role of innate immunity in shaping the biology of viral rebound and in contributing to spontaneous control of HIV-1. In summary, we expect future studies to focus on enhancing our understanding of the role of innate immunity in the progression and control of HIV-1 infection as well as on investigating innate immunity-enhancing therapeutics as tools for achieving HIV-1 cure.