Robust CAR-T memory formation and function via hematopoietic stem cell delivery

Due to the durability and persistence of reservoirs of HIV-1-infected cells, combination antiretroviral therapy (ART) is insufficient in eradicating infection. Achieving HIV-1 cure or sustained remission without ART treatment will require the enhanced and persistent effective antiviral immune responses. Chimeric Antigen Receptor (CAR) T-cells have emerged as a powerful immunotherapy and show promise in treating HIV-1 infection. Persistence, trafficking, and maintenance of function remain to be a challenge in many of these approaches, which are based on peripheral T cell modification. To overcome many of these issues, we have previously demonstrated successful long-term engraftment and production of anti-HIV CAR T cells in modified hematopoietic stem cells (HSCs) in vivo. Here we report the development and in vivo testing of second generation CD4-based CARs (CD4CAR) against HIV-1 infection using a HSCs-based approach. We found that a modified, truncated CD4-based CAR (D1D2CAR) allows better CAR-T cell differentiation from gene modified HSCs, and maintains similar CTL activity as compared to the full length CD4-based CAR. In addition, D1D2CAR does not mediate HIV infection or stimulation mediated by IL-16, suggesting lower risk of off-target effects. Interestingly, stimulatory domains of 4-1BB but not CD28 allowed successful hematopoietic differentiation and improved anti-viral function of CAR T cells from CAR modified HSCs. Addition of 4-1BB to CD4 based CARs led to faster suppression of viremia during early untreated HIV-1 infection. D1D2CAR 4-1BB mice had faster viral suppression in combination with ART and better persistence of CAR T cells during ART. In summary, our data indicate that the D1D2CAR-41BB is a superior CAR, showing better HSC differentiation, viral suppression and persistence, and less deleterious functions compared to the original CD4CAR, and should continue to be pursued as a candidate for clinical study.


Introduction
Virus-specific T cell adaptive immunity is key to the elimination of HIV-1-infected cells and is crucial for any strategic approach to achieve cure of infection. Latently infected cells persist even after decades of cART-dependent suppression of plasma viremia, thus precluding viral eradication by cART treatment alone [1]. Attempts to purge HIV-1 reservoirs also have had limited to little success in clinical trials [2,3]. To date, there have been three documented therapeutic cures, after allogeneic bone marrow transplantation from CCR5Δ32 homozygous donors [4][5][6]. However, allogeneic bone marrow transplantation is unlikely to be broadly translatable due to high mortality risk of this procedure, failures to repeat prior successes, and difficulties in finding matching donor bone marrow with homozygous CCR5Δ32 mutation [7,8]. These studies highlight the fact that although CCR5 elimination or other genetic strategies are essential for HIV-1 eradication, allogeneic HSC transplantation is not broadly applicable, underscoring the need to enhance host anti-viral responses in order to persistently suppress viral replication arising from residual, latently-infected cells.
The use of chimeric antigen receptor (CARs) to redirect T cell immunity against HIV-1 represents a highly promising gene therapy approach that can be used in HIV-1 infected individuals irrespective of Human Leukocyte Antigen (HLA) type. CARs recognize target cells through direct binding to specific cell surface antigens and thus are HLA-unrestricted and redirect both CD4+ and CD8+ T cells, bypassing a major limitation for T cell immunotherapies [9]. While recently widely applied for cancer, the first CAR clinical trials were for HIV-1 infection, testing a CAR composed of the CD4 extracellular domain linked to the intracellular CD3-z chain signaling domain (CD4CAR), utilizing CD4 binding to HIV-1 Envelope (Env) for targeting and killing of HIV-1 infected cells by redirected peripheral T cells [9]. Initial clinical trials with CD4CAR modified T cells showed excellent safety but revealed limited antiviral efficacy in vivo [10,11]. Follow-up over a decade indicated only low-level persistence of transduced cells but continued clinical safety [12]. The limited efficacy was likely due to factors such as: suboptimal T-cell processing/handling, low levels of HIV-1 antigen during ART treatment, CAR T cell dependence on antigenic stimulation for proliferation/maturation, and potential enhancement of cellular HIV-1 infection by the CD4 portion of the CAR [13][14][15]. We and others have shown that protection of CD4CAR modified cells from viral entry facilitated by the CAR is essential in order to ensure T-cell functionality and survival [13,14,16].
The success of peripheral T cell CAR gene therapy to target HIV-1 and various malignancies has been limited by inadequate function, persistence, and trafficking of transduced T cells [17]. Hematopoietic stem cell (HSC) gene therapy has the potential to overcome many of these issues and offers the lifelong generation of physiologically functional CAR T cells in vivo. Autologous HSCs are capable of long-term engraftment with substantially lower morbidity and mortality than allogeneic transplantation. HSC based gene therapy allows for transduced cells to undergo normal immune developmental mechanisms, including thymic selection, which eliminates potentially self-reactive T cells and increases the potential for the development of immunological memory [16,18,19]. We have previously demonstrated that autologous HSCs modified with a gene for a T cell receptor (TCR) against HIV-1 or melanoma antigen [19,20], or an anti-HIV-1 CAR [18], resulted in successful differentiation of modified T cells that resulted in significant suppression of HIV-1 replication/cancer growth in humanized bone marrow-thymus-liver (BLT) mice. Most recently, we demonstrated the feasibility, safety, and in vivo SHIV suppression of the HSC-based CAR approach in non-human primates (NHPs) [16]. Importantly, we found CAR-HSCs transplanted animals have lower viral rebound after release of cART as compared to control animals and CAR expressing cells were found in multiple lymphoid tissues, resulting in decreased tissue viral RNA levels and protection of CD4+ T cells in the gut, which is one of the primary replication and reservoir sites for HIV-1.
Addition of costimulatory domains in tandem with CD3z, such as the signaling domains of CD28 or CD137 (4-1BB), has been shown to enhance in vivo CAR T cell function and persistence [21][22][23]. These second-generation CAR T cells have been confirmed to mediate potent anti-leukemia responses of peripheral T cells in clinical trials [24,25]. While CAR T cells containing the CD28 costimulatory domain generally undergo more intense proliferation immediately following stimulation, peripheral CAR T cells with the 4-1BB costimulatory domain show longer persistence and more sustained anti-tumor effect [21,26,27]. However, it is completely unknown if addition of costimulatory domains would improve HSC-based CAR therapy. In this report, we examine ways to effectively enhance HSC-based HIV-targeting CAR cell function. We developed an optimized CD4-based CAR that is resistant to non-specific receptor activation as well as does not mediate viral entry. We examined the effects of the costimulatory domains CD28 and 41BB on the engraftment and anti-HIV-1 efficacy of HSCs based CAR therapy in vivo. We found that the resultant CARs facilitated greater engraftment, T cell production, antiviral function compared to the original CD4CAR, as well as uniquely generated memory T cell responses in vivo, suggesting potentially greater clinical efficacy with this improved stem cell-based approach.

Truncated form of CD4CAR does not mediate HIV-1 infection
One undesirable feature of the CD4CAR used in our previous studies and clinical trials is that the receptor itself can mediate HIV-1 infection [14,15,28]. The CD4 molecule is comprised of four extracellular domains, denoted D1-D4 [29]. The D1 domain, the most distal from the transmembrane domain, comprises the HIV-1 envelope binding region (Fig 1A) [30,31]. HLA class II binding sites are found in primarily in D1, D2 domain [32], and the D4 domain mediates interleukin-16 (IL-16) binding and CD4-CD4 dimerization [33]. IL-16 is an immune-modulatory cytokine that primarily functions as a chemoattractant for CD4-bearing cells at sites of inflammation [34]. In the context of a CD4-based CAR, IL-16 binding in the D4 domain can lead to possible non-specific signaling of CD4CAR via IL-16 [33]. The D3 domain of CD4 receptor also plays an important role in TCR:CD4 complex formation and TCR stimulation [35]. To address these issues and create a CAR that does not cross-react with IL-16, has decreased interaction with HLA class II and the endogenous TCR, and potentially increases the safety of the molecule, we created a truncated CD4 CAR molecule by deleting the D3 and D4 domains, leaving only the D1D2 domains of CD4 that will allow HIV-1 envelope recognition (D1D2CAR) (Fig 1A).
To elucidate CD4CAR mediated HIV-1 infection, we transduced HOS.CCR5 or HOS. CXCR4 cells that are not susceptible to HIV-1 due to the lack of CD4 expression with lentiviral vectors expressing the CD4CAR or D1D2CAR and then infected them with either R5 tropic HIV NFNSXSL9 or X4 tropic HIV NL4-3 . Infection of CAR+GFP+ cells were analyzed by intracellular staining of Gag followed by flow cytometry. As shown in Fig 1B, expression of CD4CAR resulted in successful infection of HOS CCR5 or HOS CXCR4 cells by R-5 tropic HIV NFNSXSL9 or X-4 tropic HIV NL4-3 . Expression of the D1D2CAR did not render cells permissive to HIV-1 infection (Fig 1B), indicating that this truncated CAR does not allow viral entry despite its ability to bind to HIV-1 Envelope. These results suggest that the D1D2CAR has an added safety aspect in its use, not facilitating HIV-1 infection like the full length CD4CAR does when expressed on the surface of a cell.

The D1D2CAR functions similarly to full length CD4CAR in T cells
Previously, to prevent CD4CAR mediated HIV-1 infection, we have included anti-HIV-1 reagents in our lentiviral vectors to protect CD4CAR expressing cells from HIV-1 infection PLOS PATHOGENS [13,16]. One lentiviral vector contained the CD4CAR molecule as well as two antiviral genes: a short hairpin RNA (shRNA) targeting CCR5 coreceptor expression and a shRNA that targets HIV-1 RNA for degradation [13]. We have shown previously that the presence of these antiviral shRNAs protected CAR T cells from infection [13]. To compare the function of D1D2CAR T cells with CD4CAR T cells, we constructed lentiviral vectors that contain both the anti-HIV-1 reagents as well as the D1D2CAR (which itself does not permit infection) (Fig 2A). This will allow protection of all CAR-modified progeny cell lineages expressing endogenous CD4, including mature CD4+ T lineage cells. Although D1D2CAR does not mediate HIV infection on cells that do not express endogenous CD4, it does not have protective effects for CD4+ progenitor and T cells against HIV infection; therefore, anti-HIV protection with shRNAs is still necessary. In this study, for fair comparison, all CAR lentiviruses used in in vitro or in vivo studies contain anti-CCR5 and anti-HIV shRNAs. To assess if the D1D2CAR transduced T cells respond to HIV-1 infection, we co-incubated D1D2CAR transduced CD8+ T cells with HIV-infected or uninfected T1 cells and measured production of intracellular cytokines interferon-γ (IFN-γ) and tumor necrosis factor alpha (TNF-α). As shown in Fig 2B, both CARs exhibit similar expression levels of IFN-γ and TNF-α production in response to HIV-1 infected target cells. We further examined specific cytotoxic activity of CD4CAR-expressing T

Fig 2. D1D2CAR CAR T cells has similar in vitro function compared to CD4CAR T cells.
A. Schematic review of protective CD4CAR and D1D2CAR vector. B. Cytokine production of CD8 CAR T cells co-incubated with uninfected or HIV infected T1 cells. Sorted primary CD8 T cells were stimulated with anti-CD3/CD28 for 2 days and transduced with either protective CD4CAR or D1D2CAR lentiviral vector. Two days after transduction, CAR T cells were co-incubated at with either uninfected or HIV-1 infected T1 cells at 1:10 ratio overnight. Afterwards, cells were treated with GolgiPlug for 6hours before intracellular staining of cytokines and flow cytometry analysis. C. Specific killing of unstimulated (Env-) and stimulated U1 cells (Env+) by CD4CAR or D1D2CAR CD8 T cells in vitro. Primary CD8 T cells were transduced with protective CD4CAR or D1D2CAR lentiviruses. 2 days after transduction, CAR T cells were co-incubated with pre-stained PMA stimulated (Env+) or unstimulated (Env-) U1 cells at 1:1, 3:1, 5:1 effector to target ratio for 16 hours. Afterwards, cells were stained with zombie violet fixable viability kit and analyzed by flow cytometry. % specific killing was calculated by (%live gag+ U1 cells without CAR cell-%live gag+ U1 cells with CAR cells)/ %live gag+ U1 cells without CAR cells.
https://doi.org/10.1371/journal.ppat.1009404.g002 Fig 2C, both CD4CAR and D1D2CAR show comparable cytotoxic activity at multiple effector to target ratios. These results indicate that the D1D2CAR is functionally capable of inducing antiviral responses at a level similar to the full length CD4CAR.

D1D2CAR does not affect T cell differentiation and TCR expression
We have previously demonstrated that CAR modified HSCs can successfully differentiate into functional T cells that recognize HIV-1 infected cells in both a humanized mouse model [13] and in NHPs [16]. Interestingly, we also observed that expression of full length CD4CAR led to shut down of endogenous TCR rearrangement and surface expression in roughly half of the differentiated CAR-expressing T cells [13]. This appears to be caused by the extracellular CD4 domain's interaction of the CD4CAR with HLA class II expressed by the thymic stroma, which permits positive selection and shut down of endogenous TCR rearrangement.
To examine if D1D2CAR also suppresses endogenous TCR expression, we constructed humanized bone marrow-liver-thymus (BLT) mice with HSCs transduced with lentiviral vectors expressing either the CD4CAR or D1D2CAR as described above. Mice were infected with HIV-1 10 weeks after surgery, following confirmation of humanization ( Fig 3A). Percentages of CAR+GFP+ cells among "classic" CD45+CD3+ T cells and CD45+CD19+ B cells from human peripheral blood mononuclear cells (PBMCs) were assessed before HIV-1 infection. Humanized mice receiving the full length CD4CAR transduced HSCs developed a higher percentage of CAR+ B cells as compared to CAR+ T cells in the peripheral blood, suggesting a preferential differentiation of CAR B cells or impaired CAR T cell differentiation (Fig 3B). In contrast, humanized mice with D1D2CAR transduced HSCs showed similar levels of CAR expressing T cells and B cells within the same mouse, thus did not show a skewing of hematopoiesis towards B cells. Interestingly, when we examined TCR expression among thymocytes and T cells in the spleen (CD45+CD2+CD5+CD20-cells), CD4CAR transduced mice had significantly reduced expression of TCR among CAR+ thymocytes and spleen T cells (Fig 3C), similar to what we have reported previously [13]. In contrast, D1D2CAR expression did not affect endogenous TCR expression on CAR T cells. In addition, similar to our previous report, high level expression of the CD4CAR on T cells led to shut down of TCR and CD3 expression ( Fig 3D). In contrast, expression of D1D2CAR did not affect TCR and CD3 expression on T cells regardless of the expression level. These results strongly suggest that the truncated D1D2CAR does not affect endogenous TCR rearrangement during thymopoiesis, preventing hematopoietic skewing towards the B cell lineage and leading to normal CAR T cell development. This may due to the deletion of the D3 and D4 regions that are required for signaling via HLA-II binding, which in turn affects positive and/or negative selection in the thymus [13,35].

HSCs based D1D2CAR therapy suppresses HIV-1 replication in vivo
To examine whether the D1D2CAR expressing cells derived from HSCs can suppress HIV-1 replication in vivo, humanized BLT mice transplanted with unmodified HSCs or HSCs modified with CD4CAR or D1D2CAR were infected with an R5-tropic HIV NFNSXSL9 for 10 weeks. Plasma viral load was monitored bi-weekly after HIV-1 infection. Mice that received HSCs transduced with either the CD4CAR or the D1D2CAR vectors showed lower levels of viral load for 10 weeks compared to untreated control animals (1 log lower than unmodified control mice) (Fig 4A). To determine if CAR T cells (defined as CD45+CD2+CD56-GFP+ to include CD3-/TCR-CAR T cells for the rest of the study), can respond to Env+ target cells or IL-16 nonspecific stimulation ex vivo, splenocytes from CD4CAR or D1D2CAR mice were exposed to Env+ ACH2 cells, Env-parental CEM cells, or soluble IL-16 cytokine. Afterwards, cells were analyzed for intracellular expression of IFN-γ and the degranulation marker CD107a. Full length CD4CAR T cells responded to both Env+ and IL-16 cytokine stimulation, suggesting non-specific signaling via IL-16 mediated CD4 dimerization (Fig 4B and 4C). In contrast, D1D2CAR T cells responded only to Env expressing target cells, but not to soluble IL-16. These results suggest that relative to the full length CD4CAR, D1D2CAR has comparable function, does not mediate HIV-1 infection or IL-16 signaling, and has a favorable expression pattern on T cells in vivo.

Costimulatory domain 4-1BB, but not CD28, allows for successful CAR T cell development from HSCs
We have demonstrated that HSCs modified with a CD4CAR can differentiate into naïve CAR T cells 36 . Similar to primary responses of endogenous naïve T cells, our previous study

PLOS PATHOGENS
suggested that upon first contact with antigen (HIV-1 env/SHIV-1 infected cells), these cells are delayed in their ability to become fully activated and functional [13,15,36]. Addition of co-stimulatory domains to the CAR molecule may facilitate more rapid primary responses of CAR+ cells, but their impact on HSC engraftment and thymopoiesis is unknown. To determine whether the costimulatory domains 4-1BB or CD28 enhance in vivo functions of CAR T cells, we constructed CD4CAR or D1D2CAR molecules (in a lentiviral vector containing CCR5 shRNA and sh516, as shown in Fig 2A) with either the 4-1BB or CD28 costimulatory domains (see S1 Fig for the schematic representation of the lentiviral plasmids). Humanized BLT mice were transplanted with donor matched fetal thymus and liver derived HSC (CD34+) transduced with either CD4CAR, D1D2CAR, CD4CAR-41BB, D1D2CAR-41BB, CD4CAR-CD28 or D1D2CAR-CD28. Transduction efficiency of CD34+ by these vectors was evaluated following in vitro culture of CD34+ cells and the levels did not appear to correlate to observed qualitative or quantitative differences between groups (S2 Fig). 20 weeks after transplantation,

PLOS PATHOGENS
we examined HSC engraftment in the bone marrow and found that the percentage of CARexpressing cells among CD34+ HSC is similar across different CARs, suggesting successful engraftment of all gene modified HSC (Fig 5A). However, addition of the CD28 co-stimulatory domain appeared to have detrimental effect on T cell and B cell differentiation: we observed significantly reduced levels of mature CAR T (Fig 5B) or CAR B cells (Fig 5C) in peripheral blood for CD4CAR-CD28 and D1D2CAR-CD28 as compared to mice that were transplanted with CD4CAR, D1D2CAR, CD4CAR 4-1BB and D1D2CAR 4-1BB modified HSCs. Interestingly, both CD4CAR-CD28 and D1D2CAR-CD28 have significantly lower level of CAR+ cells among both TCR+ and TCR-thymocytes as compared to CARs without costimulatory molecule or with 4-1BB (Fig 5D-5E), suggesting that the impairment of CAR T cell development by addition of CD28 happened prior to TCR rearrangement. Notably, 4-1BB does not affect T cell or B cell differentiation of either CD4CAR or D1D2CAR modified HSCs.

The costimulatory domain 4-1BB enhances CAR T cell's function in vivo
To examine if the costimulatory molecule 4-1BB improves CAR T cell function in vivo, we infected humanized BLT mice transplanted with untransduced HSCs, or HSCs transduced with protective CD4CAR, D1D2CAR, CD4CAR-41BB or D1D2CAR-41BB with R5-tropic HIV NFNSXSL9 after immune reconstitution. All of the CAR mice showed significant suppression of viral load as compared to control mice (Fig 6A) at 2 and 4 weeks after HIV-1 infection. Mice that received CD4CAR-41BB and D1D2CAR-41BB HSPC transplants demonstrated significantly lower viral loads than mice that received CD4CAR and D1D2CAR HSC transplants during early infection (week 2), suggesting that the costimulatory domain triggers a faster and more effective response. In peripheral CAR T cell based therapies, co-stimulatory signals are required to achieve persistence and robust anti-tumor/anti-HIV-1 activity [14,37]. To examine how the 4-1BB costimulatory domain affects CAR T cell differentiation into naïve, effector, memory and terminally differentiated effector memory T cells in vivo in response to antigen

PLOS PATHOGENS
stimulation, we performed flow cytometry on peripheral PBMCs collected 4 weeks after HIV-1 infection. We found that compared to CD4CAR, CD4CAR 4-1BB CAR T had increased central memory (CM) differentiation (p<0.1) and significantly reduced terminally differentiated effector memory cells (EMRA), similar to what was previously reported for CD19CAR T cells [37] (Fig 6B and 6C). Interestingly, we found that the D1D2CAR had reduced effector memory (EM) differentiation as compared to the CD4CAR (Fig 6C). D1D2CAR 4-1BB had increased EM differentiation as compared to D1D2CAR, to the level similar to CD4CAR and CD4CAR 4-1BB. Importantly, CD4CAR-41BB, D1D2CAR 4-1BB and D1D2CAR T cells all have significantly reduced EMRA differentiation, which are terminally differentiated T cells that display senescent phenotypes and are thus less desirable [38] (Fig 6C). While EM differentiation is critical in antiviral responses, the skewing of the 4-1BB containing CD4CAR and D1D2CAR towards central memory type responses is a desirable aspect of the response as central memory (CM) T cells have superior persistence and proliferation and can support sustained response for longer periods and maintain immune memory [39].

D1D2CAR 4-1BB cells show faster viral suppression in combination with ART and better CAR T cell persistence during ART treatment
To examine the effects of CD4-CAR HSC-based therapy on HIV-1 suppression, persistence, and viral rebound, we constructed BLT mice with HSCs modified with CD4CAR, CD4CAR 4-1BB, D1D2CAR and D1D2CAR 4-1BB containing lentiviral vectors and infected with R-5 tropic HIV NFNSXSL9 after immune reconstitution. 5 weeks after HIV-1 infection, mice were put on daily ART (TDF, FTC and Elvitegravir) for 6 weeks (allowing for full viral suppression) followed by ART withdrawal. Plasma viral loads were measured bi-weekly prior to, during, and after ART (Fig 7A). With the exception of one mouse in the mock group, we observed full viral suppression by ART to undetectable levels in each mouse by week 11 (Fig 7A). Survival analysis of time to undetectable viral loads shows that D1D2CAR-41BB mice had significantly faster viral suppression as compared to mock mice (Log rank test, p = 0.0375) (Fig 7B). 3 weeks following ART release, all mice rebounded except one mouse in D1D2CAR 4-1BB group. All CAR-expressing mice maintained lower viral load as compared to mock mice after viral rebound (Fig 7A) at a point lower than pre ART levels. D1D2CAR-41BB mice exhibited lower average viral loads as compared to D1D2CAR but did not reach statistical significance.
We then closely examined the change of CAR+ T cells over time (Fig 8). As expected, all CAR T cells expanded after HIV-1 infection (Fig 8A). Interestingly, during ART treatment, numbers of CAR+ T cells in both CD4CAR and D1D2CAR containing mice declined after ART, likely due to reduction of viral antigen and consequent stimulation. Interestingly, percentages of CAR+ T cells in CD4CAR 4-1BB and D1D2CAR 4-1BB containing mice were maintained in the peripheral blood during ART treatment (Fig 8B). After ART withdrawal, D1D2CAR 4-1BB CAR T cells further expanded and had significantly higher percentages of CAR+ T cells in the peripheral blood as compared to other groups (Fig 8A). CM populations of CAR+ T cells developed in each CAR-expressing group of animals, albeit to different degrees (Fig 8C-8D). Compared to CD4CAR, CD4CAR 4-1BB, D1D2CAR and D1D2CAR 4-1BB mice had significantly higher percentages of CM CAR T cells and lower percentages of EMRA CAR T cells (Fig 8D). Interestingly, while D1D2CAR 4-1BB mice had higher percentages of EM phenotype during early HIV-1 infection as compared to D1D2CAR mice (Fig 6C), during ART viral suppression, D1D2CAR 4-1BB mice had lower level of EM CAR T phenotype as compared to D1D2CAR mice (Fig 8D). This strongly suggests a skewing of D1D2CAR 41BB to develop into CM phenotype during ART mediated viral suppression, which may explain the better persistence of D1D2CAR 4-1BB T cells during ART treatment when HIV-1 viral antigen is scarce.

Discussion
Recent years have seen rapid developments in CAR-T therapy. Our study further contributes a superior CD4CAR-based molecule, D1D2CAR-4-1BB, that is suitable for hematopoietic stem cell based CAR therapy and can achieve potent, long lasting anti-viral responses in vivo. Numerous studies have shown that the second and third generation of CAR molecules that incorporate the signaling domains of co-stimulatory molecules such as CD28 or 4-1BB into the CAR molecule demonstrate enhanced CAR T cell function and survival [40]. For anti-HIV-1 CARs, recent studies showed that both CD28 and 4-1BB enhanced the anti-HIV-1 function of CD4-based CAR-transduced peripheral T cells in vitro [14], and a 4-1BB

PLOS PATHOGENS
containing CAR allowed better expansion and control of HIV-1 spread in vivo [14,41]. However, most anti-HIV-1 CAR studies focused on peripheral based T cells and their short term effects on HIV-1 infection in vivo [14,[41][42][43]. It is unclear if this peripheral anti-HIV-1 CAR T therapy can achieve long term persistence.
For successful long-term immune containment of HIV-1 infection, it is critical to have transduced T cells that normally survive and proliferate in vivo over decades, to suppress viral reactivations from the long-lived latent reservoir. Previously, using a CD4-based first generation CAR without a co-stimulatory domain, we demonstrated the feasibility and efficacy of using HSC based CAR therapy to achieve lifelong engraftment, development, and persistence of functional CAR T cells that trafficked to multiple lymphoid tissues and suppressed viral replication in vivo [13,16,36]. To further enhance the safety and efficacy of HSC-based CAR therapy, here we tested a novel, truncated D1D2CAR designed to eliminate IL-16 mediated signaling and CD4CAR mediated HIV-1 infection. We also tested the addition of co-stimulatory domains CD28 and 4-1BB. We found that D1D2CAR did not mediate HIV-1 infection PLOS PATHOGENS (Fig 1B), did not react to IL-16 stimulation (Fig 4B and 4C) and had similar anti-HIV-1 activity as compared to CD4CAR both in vitro and in vivo (Figs 2B, 2C and 4A). Interestingly, unlike CD4CAR, which binds to MHC-II and affects TCR gene arrangement and T cell differentiation, we found that D1D2CAR had no effect on T cell differentiation and TCR expression (Fig 3B-3D). As a result, while CD4CAR modified HSCs had biased development towards CAR B cells, D1D2CAR HSCs had unbiased development of T and B lymphocytes (Fig 3B). Importantly, we found that while the 4-1BB costimulatory domain did not affect CAR T cell development for both CD4CAR and D1D2CAR, addition of CD28 severely impacted T cell and B cell development (Fig 5B-5E), suggesting that adding the CD28 co-stimulatory domain is deleterious for HSC based CAR therapies. As CAR therapy advances, third or fourth generations of CAR designs have emerged by combining multiple co-stimulatory domains and/or inclusion of cytokine receptor stimulatory domains for IL-7, IL12, or IL-15 to improve CAR T cell expansion and persistence [44,45]. However, incorporating additional co-stimulatory domains may affect thymus-mediated T cell differentiation. Therefore, the effects of additional signaling domains and cytokines on HSC-based CAR therapy would need to be tested in the context of each individual CAR molecule.
Recent studies suggest that incorporation of 4-1BB in certain CARs could improve memory formation and ameliorate exhaustion [21,26,46]. However, the clinical impact of choosing and combining different costimulatory domains for different CAR molecules and diseases has yet to be clearly established 38 . In our study, we found that addition of the 4-1BB costimulatory domain resulted in faster viral suppression in CD4CAR 4-1BB and D1D2CAR 4-1BB mice as compared to CD4CAR and D1D2CAR mice (Fig 6A without ART, Fig 7A and 7B with ART). 4-1BB also promoted better CAR T cell persistence during ART treatment when viral antigen was drastically reduced (Fig 8A and 8B), particularly for D1D2CAR 4-1BB T cells. In general, we observed skewing of CM and reduction in EMRA cells for CD4CAR-41BB, D1D2CAR 4-1BB, and D1D2CAR 4-1BB T cells during early infection and during ART treatment. Compared to CD4CAR, D1D2CAR T cells have reduced EM formation, improved formation of CM cells and reduction in EMRA cells, potentially due to reduction in nonspecific signaling via HLA class II and IL-16. Interestingly, the impact of 4-1BB on D1D2CAR T cells versus on CD4CAR T cells differ slightly: CD4CAR 4-1BB T cells have improved CM formation and reduced formation of EMRA as compared to CD4CAR T cells during untreated primary infection and ART treatment; in contrast, D1D2CAR 4-1BB T cells showed increased EM differentiation as compared to D1D2CAR T cells during untreated primary HIV-1 infection and reduced EM phenotype during ART treatment. 4-1BB may have promoted naïve to EM differentiation for D1D2CAR 4-1BB T cells during early infection and promoted transition of D1D2CAR 4-1BB T cells from EM to CM during ART infection. The detailed mechanism of 4-1BB impact on CD4CAR and D1D2CAR T cells is yet to be studied. The fact that 4-1BB has differential impact on two closely related CD4-based CARs strongly suggests that the utilization and selection of costimulatory domains should be carefully selected based on the context of the CAR molecule and the disease.
To achieve ART-free remission or "HIV-1 cure", it is crucial to have long term anti-HIV-1 immunity to control reactivation of latently infected cells. The critical role of HIV-1 Env interaction with CD4 for viral replication limit potential viral escape from a CD4-based CAR, since loss of Env binding to CD4 is associated with marked loss of viral fitness [47].Our HLA-independent, CD4 based HSC-derived CAR therapy could provide a long term, enhanced anti-HIV-1 immunity that is also resistant to HIV-1 immune evasion. Given the persistence and functionality of CD4 based CAR T cells derived from HSCs, our HSCs based CAR therapy could also be combined with multiple rounds of latency reversal agents as an effective 'kill' part for the 'kick-and-kill' strategy to reduce and eliminate HIV-1 reservoir [48].

Ethics statement
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health ("The Guide"), and was approved by the Institutional Animal Care and Use Committees of the University of California, Los Angeles, protocol #2010-038-31J. For humanized mice, all surgeries were performed under ketamine/xylazine and isofluorane anesthesia and all efforts were made to minimize animal pain and discomfort.
Humanized mice construction. CAR BLT mice were constructed similarly to previously reported HIV-1 Triple CAR BLT humanized mice (1,2). Briefly, human fetal liver derived CD34+ cells were purified by immunomagnetic separation. Cells were then transduced overnight with control or protective CAR lentiviruses with retronectin-coated plates. To evaluate transduction efficiency, 0.1 million transduced CD34+ cells was set aside from surgery and were cultured in extension media (IL-3 100 ng/ml, IL-6 100 ng/ml, SCF 100 ng/ml in 10% FCS RPMI 1640) for 7 days. Afterwards, cells were analyzed by flow cytometry. On day of transplant, NOD.Cg-PrkdcscidIl2rgtm1Wjl/SZJ (NOD/SCID/IL2Rγ−/− or NSG; The Jackson Laboratory) mice received 2.7 Gy total body sublethal irradiation and then were transplanted with transduced CD34+ in Matrigel (corning Life Sciences), liver and thymus tissue under the kidney capsule same donor as the CD34+ cells. Afterward, mice were injected with *0.5×10 6 lentivirus-based CAR vector transduced CD34+ cells. At 8-10 weeks post-transplantation, each mouse was bled retro-orbitally and peripheral blood mononuclear cells analyzed by flow cytometry to check human immune cell engraftment. Upon stable human leukocyte reconstitution efficiency more than 50%, mice were used for HIV-1 infection and further experiments. To supplement for low level of dendritic cells in NSG-BLT mice [13], all mice were injected with 1 million unmodified dendritic cells cultured from autologous HSCs (prepared as detailed in [49]) prior to HIV infection or before ART withdrawal. Lentiviral production. The lentivirus based GFP control vector and CAR vectors were produced in 293FT cells using the Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA) as previously described [13]. Briefly, 293FT cells were co-transfected simultaneously with CAR vectors with pCMV.ΔR8.2.Δvpr packaging construct and the pCMV-VSV-G envelope protein plasmid as previously described. Supernatant was collected from transfected 293FT cells 48 hours following transfection, filtered using a 0.45 μm sterile filter, and concentrated by ultracentrifugation using a Beckman SW32 rotor at 30,000 rpm at 4˚C. Medium was aspirated and pellet was resuspended with PBS and stored at -80˚C.
Cytokine assay. Primary CD8 T cells were purified from healthy PBMCs using CD8 microbeads (Miltenyi #130-097-057). Afterwards, CD8 cells were stimulated with plate bound anti-CD3 and soluble anti-CD28 antibodies for 2 days and transduced with protective CD4CAR or D1D2CAR lentiviral vector for 2 days. Infected T1 cells were prepared by infection of T1 cells with 500ng p24 of HIV NL4-3 /million cells for 2 days. CAR T cells were co-incubated at with either uninfected or HIV-1 infected T1 cells at 1:10 (Effector: Target) ratio for 16 hours (overnight). Afterwards, cells were treated with GolgiPlug for 6hours before intracellular staining of cytokines and flow cytometry analysis.
HIV-1 infection and ART treatment. The R5 tropic strain of HIV-1 (NFNSXSL9) was generated by transfection of 293T cells with plasmid containing full-length HIV-1 (NFNSXSL9) genome. Humanized mice were infected with NFNSXSL9 (200 ng p24 per mouse) through retro-orbital injection while under inhalant general anesthesia. Infected mice with demonstrable viral infection were treated for 6 weeks with ART drugs. The ART regimen is consisted of tenofovir disproxil-fumarate (TDF, 80mg/kg), emtricitabine (FTC, 120mg/kg), and Elvitegravir (ELV160mg/kg) given by food. TDF, FTC and ELV were generously supplied by Gilead Sciences. TDF, FTC and ELV were dissolved in DMSO and mixed with sweetened moist gel meal (DietGel Boost, ClearH2O; Medidrop Sucralose).

Statistical analysis
Statistical analysis was performed using software Prism. Man-Whitney U test is used for nonparametric testing of independent groups. Wilcoxon matched pairs signed rank test was used for nonparametric testing of paired groups. Log rank test was used for survival analysis.