The first-generation anti-HIV CAR-T cells linked singular CD3ζ intracellular signalling domains to antigen recognition domains directed against HIV Env using either CD4 or scFvs derived from anti-HIV Env bNAbs ( Figure 1B ). These CAR-T cells were very specific in recognizing and lysing cells expressing HIV Env in vitro ( Table 2A ). Using a CD8 cell line (WH3) expressing CD4-CD3ζ CARs, Romeo and Seed demonstrated that after binding to gp120 expressing HeLa cells, CD4-CD3ζ CAR-T cells exhibited extremely high cytolytic activities by chromium-release assay (91). These data were replicated in human PBMC derived CD8 T cells expressing CD4 or scFv derived from anti-HIV mAb 98.6 -CD3ζ based CARs (92). After binding to HIV Env, these CAR-T cells exhibited extremely high cytolytic activities by JAM assay, a cell lysis assay which measures the DNA retained by living cells rather than the cellular components lost by dying cells, against the CEM cell line infected with the HIV-1 IIIB strain and HEK293 expressing HIV env. Furthermore, using Jurkat cells expressing CD4 or scFv derived from anti-HIV mAb 98.6 -CD3ζ based CARs, Roberts et al. demonstrated that HEK293 env target cells stimulated Jurkat CAR-T cells to secrete IL-2 at high levels (92). Subsequently, Yang et al. confirmed that primary CD8 T cells expressing CD4 ligand or scFv derived from anti-HIV mAb 98.6 -CD3ζ based CARs lysed HIV-1 IIIB infected T1 and H9-B14 cells and inhibited HIV-1 replication in vitro (93). The kinetics of lysis and efficiency of inhibition were comparable to that of naturally occurring HIV-1-specific CTL clones isolated from infected individuals (93). Such preclinical studies provided clear evidence that first generation CD4 or scFv derived from anti-HIV mAb 98.6 -CD3ζ based CAR T cells exhibited significant anti-HIV activity in vitro.

First generation CD4 ligand-based CAR-T cells also demonstrated stable engraftment and long-term safety in the clinical setting ( Table 2B ). However, despite these observations and the previously demonstrated anti-HIV activity in vitro, the CD4-CD3ζ CAR-T cells were not capable of reducing viral burden permanently in most clinical studies (87, 88). Specifically, Mitsuyasu et al. noted CAR expression in 1% - 3% of circulating T cells at 8 weeks and 0.1% at 1 year following administration of autologous CD4 and CD8 T cells expressing CD4-CD3ζ CARs with or without IL-2 to HIV-infected patients (n=24) with a detectable level of virus (1000 copies/mL). However, no significant reduction in HIV RNA or DNA was observed (88). Moreover, administration of exogenous IL-2 with CAR T cells was not shown to enhance survival. Another study showed greater persistence of CD8 T cells expressing CD4-CD3ζ CARs at 1 year after multiple infusions when co-administered with CD4 T cells expressing CD4-CD3ζ CARs compared to CD8 CAR-T cells infused without CD4 CAR-T cells in identical twins (n=33) sero-different for HIV infection (87). Furthermore, a reduction in mean HIV RNA levels was observed in patients who received repeated infusions of CD8 CAR T cells (87). Subsequently, it has been demonstrated that infusion of CD4-CD3ζ expressing CD4 and CD8 T cells led to a decline from baseline in HIV after CAR T cell treatment, but no meaningful difference in HIV burden was observed between treatment and placebo groups (89). Notably, both groups experienced a treatment-related increase in CD4 T-cell counts (89).

The failure of these clinical trials to generate robust cell lysis of their targets may be attributed to several factors: (1) low transduction efficiency and/or dose of CD4-CD3ζ CAR-T cells leading to lower efficacy (94); (2) T cell exhaustion during ex vivo CD4-CD3ζ CAR-T cell expansion by excessive IL-2 stimulation (95); (3) CAR design leading to low expansion rate and effector functions (96); (4) Initial ART administration prior to CD4-CD3ζ CAR T cell infusion may block the efficacy of CAR-T cells by reducing HIV Env antigen required for optimal expansion of a CD4-CD3ζ CAR-T cells in vivo and the production of a robust anti-HIV response (97); (5) CD4-CD3ζ CAR T cells may induce selective pressure on HIV Env to select for escape mutants (98); however, (6) the most common explanation for the failure of the CD4-CD3ζ CAR T cell clinical trials is the inclusion of CD4 as the antigen recognition domain, rendering the CD8 T cells expressing CD4 susceptible to HIV infection (97). Overall, the application of first-generation anti-HIV CAR T cell therapy in HIV patients has demonstrated a lack of adverse effects, and the generation of compartmental immunity to HIV (87–89). However, the lack of efficacy of these CAR T cells in vivo despite previously demonstrated significant anti-HIV activity in vitro led to the development of second and third generation CARs incorporating the signals require for full T-cell activation within the CAR structure to enhance their function.

Based on the success of second generation anti-CD19 CAR-T cells for hematologic malignancies, various laboratories have developed second generation anti-HIV CAR T cells and analysed them in preclinical studies ( Figure 1B ). Landmark studies have been summarised in Table 3 . Second-generation CD8 CAR T cells incorporating the CD4 ligand or scFv derived from anti-HIV bnAbs as antigen recognition domains co-stimulated by 41BB were directly compared with first generation CARs and shown to be at least 50-fold more potent at suppressing HIV replication in vitro (94). Furthermore, second generation CARs demonstrated superior antigen dependent expansion, CD4 T cell protection via reduced cell loss in vivo when compared to HIV-specific CARs without co-stimulatory molecules (101). In the same study, a direct comparison of 41BB and CD28 co-stimulatory domains showed that 41BB costimulatory domains promote antigen independent T cell persistence in vivo (101) and that they are superior to the CD28 domain in reducing viral rebound following cessation of ART treatment. This is in accordance with other 4-1BB co-stimulatory domain containing CARs showing long term persistence (96). Primary CD4 and CD8 T cells expressing scFv derived from anti-HIV bnAbs including PGT128, PGT145, VRC07-523, or 10E8 in second generation CARs with 41BB co-stimulatory domains were developed and compared (99). All CARs incorporating scFv from each bnAb were shown to be efficacious as anti-HIV designs using HIV infected cell lines and Env transfected cells (99). However, PGT145- and VRC07-523-CARs showed the most consistent potency in cytotoxic activity (99). Perhaps most importantly, this study demonstrated the success of homology-directed recombination of CAR transgene into the CCR5 (HIV co-receptor) locus to suppress entry of HIV into these cells (99). By showing the feasibility of CAR integration to disrupt CCR5 in T cells, Hale et al. demonstrated an important consideration when developing a CAR T cell immunotherapy for HIV. The efficacy of CAR expressing CD4 T cells in the context of HIV has also been demonstrated in other studies. Focussing on HIV-resistant CD4 CAR-T cells modified by introduction of a single Asp mutation (D97N) and expressing co-stimulatory domains derived from 41BB, CD28, CD27, OX40 or ICOS, Maldini et al. showed direct elimination of env expressing cell lines in vitro. They also demonstrated improved proliferation and persistence of CD4 ligand based CAR expressing CD8 T cells, especially when co-injected with CD4 ligand based CAR expressing CD4 T cells in vivo (101). Furthermore, HIV-resistant CAR expressing CD4 T cells containing costimulatory domains from the CD28 receptor family exhibited the greatest production of effector cytokines in vitro. whilst 41BB co-stimulated CAR expressing CD4 T cells showed superior expansion and reduced HIV pathogenesis in vivo using a humanized mouse model of HIV infection (100).

These data demonstrate that co-stimulatory domains along with helper functions supplied from CAR expressing CD4 T cells differentially enhance HIV-specific CAR expressing CD8 T cell function in vivo. Future studies may benefit from adopting strategies to develop HIV-resistant CAR expressing CD4 T cells and preferentially incorporating 41BB co-stimulatory domains.

Third-generation HIV-specific CARs ( Figure 1B ) have not been extensively studied in HIV. A single study combining multiple intracellular signalling domains from 41BB, CD28 and CD3ζ linked to the scFv of anti-HIV bnAb VRC01 displayed increased potency in lysing env expressing cell lines in vitro compared to HIV recognition via CD4-based third generation CAR (102). Unfortunately, this study failed to directly compare the potency of their third generation CAR to an appropriately designed second generation CAR counterpart. However, an important finding of the study was that the VRC01-based third generation CAR-T cell effectively eliminated LRA-reactivated HIV-1-infected CD4 T cells isolated from infected individuals receiving ART, thus validating the development of CAR-T cell immunotherapies against latent HIV (102). The fourth generation of CAR-T cells are known as T cells redirected for universal cytokine-mediated killing (TRUCKs) and have yet to be employed in the context of HIV infection ( Figure 1B ). TRUCKs can utilise the additional secretion of IL-12 or other inflammatory mediators to attract innate immune cells and eliminate antigen negative cancer cells. CAR designs are constantly evolving with novel features being added to provide multiple functions, especially to overcome the challenges of the tissue microenvironment and future anti-HIV CARs will no doubt incorporate such features.

Antigen escape, including antigen loss or downregulation, is a major limitation of CARs designed with single antigen recognition domains, particularly for scFv-based CARs. Immune escape was demonstrated in recent clinical trials where a decrease in plasma viremia followed by viral rebound using monotherapy of bNAb VRC01 (103) was observed. In contrast, treatment with combinations of two or more bNAbs significantly reduced the viral reservoir and demonstrated long-term viral suppression (104). Therefore, single bnAb scFv-based CARs may be insufficient for long term HIV suppression due to the emergence of escape mutants. Thus, combination of multiple antigen recognition domain by CARs, using duo or Tandem designs may provide long-term suppression and are currently under investigation. Duo CARs expressing complimentary antigen recognition domains directed against different antigens via transfection as a single bicistronic vector ( Figure 2A ) have been used successfully in cancer to overcome immune escape (106). Tandem CAR-T cells consists of a CAR where distinct antigen recognition domains are fused to generate smaller transgenes compared to dual CARs whilst retaining capacity to target multiple antigens can prevent immune escape ( Figure 2B ) (107). More recently, Hajduczki et al. further improved a bispecific CAR design based on the CRD of human mannose-binding lectin (MBL) recognizing the highly conserved oligomannose patch on gp120 and CD4 ligand by addition of a third antigen recognition domain against a distinctly conserved region on Env. The group employed a polypeptide sequence derived from the N-terminus of the HIV coreceptor CCR5 to develop the trispecific CAR to demonstrate enhanced in vitro anti-HIV potency compared to the bispecific CAR (108). Binding of either CAR in dual or tandem systems to its associated antigen is sufficient to stimulate full T cell activation. Moreover, cancer studies have demonstrated that enhanced T-cell function is observed when both targets are present (106). A preliminary study of anti-HIV dual CARs designed with CD4 antigen recognition domains linked to bnAb 17b scFvs or carbohydrate recognition domains of a human C-type lectin (109) transduced into CD8 T cells showed enhanced potency against genetically diverse strains of HIV when compared to single CD4 antigen recognition domain-based CARs in vitro. Furthermore, Liu et al. demonstrated that bispecific CAR function is influenced by the length between various antigen-binding domains to inform future studies incorporating dual antigen recognition domain designs (110). Subsequent studies have included CD4-based duo CARs linked to various bnAb scFvs or the carbohydrate recognition domain of a human C-type lectin receptor and have all shown long -term suppression of HIV in vitro (109, 111). Most recently, a study describes duo CARs targeting two or three antigens simultaneously via HIV neutralising antibody fragments in transduced primary human T cells in vitro (111). Multi-specific CAR - T cells were able to eliminate PBMCs infected with single bnAb-resistant HIV strains and displayed long-term control of HIV infection in vivo and the prevention of CD4 T cell loss (111). Further, the group demonstrated long-term control of HIV infection in vivo and prevented the loss of CD4+ T cells during HIV infection using a humanized NSG mouse model (111). Using a bicistronic lentiviral vector to allow for simultaneous expression of both CARs, the group targeted dual antigens on HIV env, first the mD1.22 domain to its CD4 binding site, and subsequently the m36.4 domain to its conformationally exposed binding site (111). Collectively, these data demonstrate that dual-targeting antigen recognition domains are important in overcoming immune escape in the context of an anti-HIV immunotherapy.

HIV specific CAR-T cells have demonstrated long-term safety in the clinical setting (87–89). However, the primary concern with CD4-ligand based anti-HIV CAR-T cell therapies is the capacity for CD4-MHC II interactions resulting in non-specific T cell activation. Fortunately, cytolysis of MHC II-expressing cells has not been demonstrated in CD4-based anti-HIV CAR-T cell studies (91, 110). Development of anti-HIV bnAb-based CAR T cell therapies specifically targeting the HIV envelope glycoproteins is safer than targeting the CD4 binding partners. It is well established that the non-covalent interaction of the gp120 and gp41 envelope glycoproteins of HIV is disrupted by soluble CD4 binding, resulting in irreversible gp120 release (112). Since CAR-T cell recognition of its specific antigen is an MHC I independent process, another important consideration is the recognition of cell-free or virion-associated HIV env by HIV specific CAR-T cells leading to activation and potentially lethal cytokine release. Although this area has not been extensively explored in HIV CAR-T cell therapies, dual targeting of HIV env with cell surface markers may provide a novel strategy to mitigate off-target cytotoxicity resulting from non-specific T cell responses to cell-free or virion-associated HIV env.

However, targeting two antigens may lead to increased risk for off-target toxicity following CAR-T cell treatment. Strategies to mitigate off-target cytotoxicity include synNotch, LOCKR and convertible systems where CARs are activated upon recognition of a specific combination of antigens. The primary signal can be provided following stimulation of the CAR linked to the CD3ζ chain whilst co-stimulation is split and provided by another CAR to limit non-specific activation ( Figure 2C ) (113, 114). A more recent strategy employs a two-step process where activation of the first receptor known as the synthetic Notch (synNotch) induces the expression of a secondary antigen specific receptor ( Figure 2D ) (115). This AND-gate recognition system ensures T cell activation only occurs when both antigens are expressed on target cells thus enabling precision killing and sparing cells expressing either of the single antigens in isolation. Other AND-gating strategies involve dimerising protein switches, which can serve as lock and keys that are activated only upon the two independent antigens (lock-bound and key-bound) present together. An interesting example of this is described in Lajoie et al. (116) ( Figure 2F ).

Another strategy to reduce toxicity is by allowing for controlled expression of CAR T cells. For example, CARs are directed against an inert molecule which is conjugated to an antigen binding antibody. By infusing various antibodies, the universal CAR can be redirected against various antigens and CAR activity can be controlled by regulating the times and amount of antibody infusions. Most recently Herzig et al. engineered a novel version of universal CAR known as convertibleCAR ( Figure 2E ) where they combine the cytotoxic T cell with many antibodies (117). This is crucial for controlling HIV as different variants exist and there would be no long-term success with a CAR-T cell carrying a single antibody to fight HIV. In this scenario, Herzig et al. used a ligand from the MIC/ULBP family, expressed on stressed cells, and its receptor NKG2D, expressed on CTL and NK cells. They engineered the convertibleCAR cell T cell to express NKG2D thus converting them into a potent killer, but only when bound to its partner, a protein called MIC-A with A being any antibody of interest. Consequently, neither the CAR-T cells nor the MicAbody can kill target cells until they have bound to each other and the MicAbody has specifically engaged its epitope on an HIV-infected cell to create an immunologic synapse. This strategy compensated for the cleavage of soluble ligands (MICA and ULBP) for the natural NKG2D receptor from the surface of infected cells, thereby compromising the ability of CTL and NK cells to kill infected cells (118).

The possibility of multiplexing a single convertibleCAR cell with several MicAbodies makes this platform quite promising for tackling multiple diseases or pathogen variants and for avoiding the universal problem of drug resistance. Another advantage of this type of CAR T cells is the exclusive ligand-receptor interaction where CAR can be introduced into patients in an inert state. Only after the administration of the specific MicAbody or MicAbody mix will the CAR-T get activated. This increases safety and controls dependent on both the dose and timing of administered HIV-specific MicAbody. By contrast, classical CAR T cells are ‘‘on’’ all the time. Using a platform that is inert unless both parts are present can potentially enhance safety.

A major shared challenge to target cell clearance in chronic HIV infection and cancer is the need for T cell localisation and exertion of effector functions within the immunosuppressive tissue microenvironment. Anti-HIV CAR T cells that over-express the chemokine receptor CXCR5, to promote trafficking into lymphoid tissues have been developed (119). This was based on the identification of follicular CD8 T cells expressing CXCR5 and their control of HIV replication within lymphoid tissue (120), in addition to a proof-of-concept study demonstrating CXCR5 overexpression in CD8 T cells led to localisation within the germinal centres of lymphoid tissue in macaques (121). These CD4-based CXCR5 expressing CAR T cells demonstrated suppression of SIV infection in vitro and trafficking in response to CXCL13 in migration assays using transwells and lymph node organoid cultures (122). Most recently, Barber-Anxthelm et al. employed hematopoietic stem cells modified with CD4 ligand based CARs to demonstrate trafficking and multilineage engraftment at lymphoid germinal centres, gut associated lymphoid tissue and the central nervous system (123). Furthermore, CAR expressing cells persisted for nearly two years at these sites of tissue-associated viral reservoirs (123). Since lymph nodes are important sites of HIV reservoirs, these data highlight important innovations that can be included in the development of a dual-targeting HIV-specific CAR-T cell with increased capacities to traffic to anatomical sites that contain large fractions of the latent HIV reservoir. Both the solid tumour microenvironment and lymph nodes can be characterised by the secretion of various cytokines and growth factors by stromal and immune cells to remodel the extracellular matrix of these tissues and contribute to the suppression of T cell responses. To overcome such barriers, CAR T cells have been engineered to secrete extracellular matrix modifying enzymes such as heparinase to degrade heparin sulphate proteoglycans in the extracellular matrix (124). These CAR-T cells demonstrated an improved capacity to infiltrate xenograft tumours in mice and prolonged survival compared with CAR T cells lacking heparinase expression (124). Such strategies could be included in future HIV CAR designs.

After overcoming the physical barriers and localising to the tumour microenvironment, upregulation of inhibitory ligands such as PDL-1 that bind to inhibitory receptor PD1 to suppress effector T cell responses have been observed (125). Several approaches to overcome the effect of PD-1 using CAR T cells have been reported and are succinctly described in Yoon et al. (126). Engineered CAR-T cells co-transduced with truncated PD-1 receptors lacking intracellular signalling functions which still bind PDL-1 but are unable to propagate inhibitory signals have been developed and demonstrated resistance to T cell exhaustion mediated by PDL-1 and prolonged survival in mice bearing xenograft tumours, compared with CAR T cells lacking these truncated PD-1 receptors (127). More recently, Jiang et al. demonstrated that expression of PD-1 dominant negative receptor (DNR) in a 3BNC117 bnAb scFv based CAR T cell resulted in superior lytic and functional responses in vitro and in vivo when compared to 3BNC117 bnAb scFv CAR expression alone (128). PD-1 receptor expression has also been disrupted completely via CRISPR/Cas9 resulting in augmentation of CAR T cell mediated killing of tumor cells in vitro and enhanced clearance of PDL-1 expressing tumor xenografts in vivo (129). The complexity of the tissue microenvironment continues to drive novel CAR design as previously outlined by our group (130).

The risk of cytokine release syndrome (CRS) is most acute during the first infusion when the cancer tumour burden is at its peak (131). However, the number of HIV-infected target cells in patients on ART is dramatically less than target cell numbers in individuals with leukaemia and HIV reactivation of latently infected cells is characterised by low surface antigen density (132). Tumour specific antigens provoke a potent immune response relative to reactivated latently infected resting memory T cells. For this reason, CRS has not been extensively studied in the context of HIV specific CAR – T cells in vivo. However, should life threatening CRS ensue, it would be beneficial to incorporate safety switches to “turn off” the CAR T cells. The popular switches currently being trialled are based on inducible caspase-9 (133) and truncated epidermal growth factor receptor (tEGFR) (elimination induced by the infusion of clinically approved antibody Cetuximab) (134).

The low number of HIV-infected target cells comprising the latent HIV reservoir and their low antigen density upon reactivation leads to challenges surrounding the expansion and persistence of HIV specific CAR – T cells. One strategy involving orthogonal IL-2 cytokine-receptor pairs may overcome poor engineered T cell expansion and persistence upon patient infusion (135). It is known that transcription factor STAT5 is phosphorylated upon IL-2 engagement of its receptor (136). After translocation to the nucleus STAT5 promotes T cell proliferation and cell cycle progression. By engineering orthogonal IL-2 cytokine-receptors pairs, Sockolosky et al. demonstrated that orthogonal IL-2 potently activated STAT5 on orthogonal IL-2 receptor transduced primary mouse CAR – T cells when compared to wild-type resulting in specific expansion of primary mouse CAR – T cells with negligible toxicity whilst retaining anti-tumour effector functions (135). As mentioned previously, 41BB co-stimulated CARs induce a delayed T cell effector response and increased persistence whilst CD28 co-stimulated CARs lead to quicker T cell activation, proliferation, cytolysis. Dual-CARs incorporating distinct CD28 and 4-1BB costimulatory domains can combine both antigen-driven expansion and long-term persistence with potent effector functions (101). Using LRAs to stimulate robust reactivation of latent reservoirs, may help in antigen dependent expansion and persistence of HIV-specific CAR T cells during clinical translation of HIV specific T cell therapies. Thus, a combinatorial approach of LRA and CAR T cells can be expected to aid in increased in vivo expansion and persistence.

CD4 ligand-based HIV specific CARs have illustrated significant barriers to HIV cure due to their susceptibility to HIV infection (137). To overcome this limitation, one strategy describes co-expressing short hairpin RNA (shRNA) sequences targeting CCR5 and HIV LTR to silence target genes (138). It was shown that expressing CCR5 and HIV LTR shRNA decreased the risk of CAR T cell infection leading to overall increased levels and persistence whilst maintaining HIV specific effector functions in vitro (138). Targeted CAR integration at the CCR5 locus in primary human T cells has been also demonstrated by RNA-based nuclease expression coupled with adeno-associated virus-mediated delivery of a CCR5-targeting donor template and was shown to effectively supress viral replication in primary HIV infected human T cells (99). Thus, targeted integration of the CAR into the CCR5 locus is an important milestone in developing robust T cell therapies against HIV. More recently, Anthony-Gonda et al. showed infection resistance when the C46 viral fusion inhibitor peptide was co-expressed in anti- HIV CAR T cells without interfering with their effector function (111). Various studies have shown that anti-HIV bnAbs may be used as an alternative to CD4 antigen recognition domains in HIV specific CARs to stimulate specific T cell activation and killing of HIV infected cells without HIV infection of CAR T cells (99, 102, 111). These studies demonstrate the feasibility of utilising bnAb scFv in CAR constructs as immunotherapies for HIV infection.