Dual CD4-based CAR T cells with distinct costimulatory domains mitigate HIV pathogenesis in vivo

Abstract

An effective strategy to cure HIV will likely require a potent and sustained antiviral T cell response. Here we explored the utility of chimeric antigen receptor (CAR) T cells, expressing the CD4 ectodomain to confer specificity for the HIV envelope, to mitigate HIV-induced pathogenesis in bone marrow, liver, thymus (BLT) humanized mice. CAR T cells expressing the 4-1BB/CD3-ζ endodomain were insufficient to prevent viral rebound and CD4+ T cell loss after the discontinuation of antiretroviral therapy. Through iterative improvements to the CAR T cell product, we developed Dual-CAR T cells that simultaneously expressed both 4-1BB/CD3-ζ and CD28/CD3-ζ endodomains. Dual-CAR T cells exhibited expansion kinetics that exceeded 4-1BB-, CD28- and third-generation costimulated CAR T cells, elicited effector functions equivalent to CD28-costimulated CAR T cells and prevented HIV-induced CD4+ T cell loss despite persistent viremia. Moreover, when Dual-CAR T cells were protected from HIV infection through expression of the C34-CXCR4 fusion inhibitor, these cells significantly reduced acute-phase viremia, as well as accelerated HIV suppression in the presence of antiretroviral therapy and reduced tissue viral burden. Collectively, these studies demonstrate the enhanced therapeutic potency of a novel Dual-CAR T cell product with the potential to effectively treat HIV infection.

Conflict of interest statement

Competing interests

C.R.M. and J.L.R. have filed an institution-owned patent (20180265565: Method of Redirecting T Cells to Treat HIV Infection) describing the construction of these HIV-specific CARs specific to Figs. 1 and 2. This patent application has been published but has not yet been granted. C.R.M. and J.L.R. have also filed an institution-owned patent (Dual CAR Expressing T Cells Individually Linked to CD28 and 4–1BB) specific to Figs. 4–6. G.J.L., J.A.H. and J.L.R. have also filed an institution-owned patent (Non-Signaling HIV Fusion Inhibitors And Methods Of Use Thereof) specific to Figs. 5 and 6. J.L.R. cofounded a company called Tmunity Therapeutics that has the rights to license the technology described in this paper. J.L.R. holds an equity interest in Tmunity. C.R.M. and J.L.R. declare no other competing financial interests. No other authors declare any competing financial interests. No authors declare any nonfinancial interests.

Figures

Extended Data Fig. 1 |. CD28 costimulation enhances the ex vivo effector function of CAR T cells.

HIV-uninfected mice were infused with an equal mixture of CD4-based CAR T cells expressing either CD3-ζ, 4–1BB/CD3-ζ and CD28/CD3-ζ costimulatory domains linked to unique fluorescent proteins to facilitate identification in vivo as described in Fig. 2 legend. Cumulative data indicating the frequency of TNF+, IL-2+ and MIP-1β+ CAR.BBζ and CAR.28ζ T cells within the same mice after ex vivo stimulation with K.Env (+) or K.WT (−) cells. Data represents the aggregate of cytokine producing cells from liver and terminal blood (n = 8). CAR.ζ T cells were too infrequent for analysis. Data shows box and whisker plots where the middle line indicates median, bounds of the box show 25th to 75th percentiles, and bars extend to min and max values. Symbols represent biologically independent animals. Significance was calculated using two-sided Wilcoxon matched-pairs signed rank test.

Extended Data Fig. 2 |. CAR.BBζ T cells accumulate multiple inhibitory receptors as disease progresses.

a, Frequency of CD4+ and (b) CD8+ CAR.BBζ T cells (G1; n = 6) and CAR.BBΔζ T cells (G2; n = 6) co-expressing TIGIT and PD-1 after infusion. Shaded box indicates the window of ART. Symbols and error bars indicate mean ± SEM. c, Frequency of CD4+ and (d) CD8+ CAR.BBζ T cells (G1) and CAR.BBΔζ T cells (G2) co-expressing TIGIT, PD-1 and 2B4 in tissues 12 weeks post-infusion. e, Cumulative data indicating the frequency of 2B4+, PD-1+ and TIGIT+ CD4+ CAR.BBζ T cells (G1) compared to CAR− CD4+ T cells (G1) within the spleens of the same mice, and (f) CD8+ CAR.BBζ T cells (G1) compared to CAR− CD8+ T cells (G1) within the spleens of the same mice. c–f, Bars indicate mean, error bars show ± SEM and symbols represent individual mice. Significance was calculated using two-sided Wilcoxon rank-sum test. Sample sizes in these studies represent biologically independent animals.

Extended Data Fig. 3 |. eomeshiT-betdim CAR.BBζ T cells accumulate from acute to chronic phases of infection.

BLT mice were infected with HIVJRCSF and infused 48 h later with either 2×107 CAR.BBζ T cells (n = 5) or inactive control CAR.BBΔζ T cells (n = 3). a, FACS plots show the change in Eomes and T-bet expression within the different CAR T cell types over time. b, Summary data indicating the longitudinal frequency of EomeshiT-betdim CD8+ (left panel) and CD4+ (right panel) CAR T cells (left y-axis), and mean log plasma HIV RNA (copies mL−1) (right y-axis). Thin dotted line denotes limit of viral load quantification. Symbols and error bars indicate mean ± SEM. c, Spearman correlation analysis of frequency of EomeshiT-betdim CD8+ CAR.BBζ T cells compared with viral burden measured as the frequency of HIVGAG+ CD8− T cells in various tissues 10 weeks post-infection. Sample sizes in these studies indicate biologically independent animals.

Extended Data Fig. 4 |. Dual-CAR T cell product transiently delays CD4+ T cell loss despite persistent HIVJRCSF infection.

BLT mice received Dual-CAR T cell product (TCP) (n = 6) 48 h after HIVJRCSF challenge, while control mice were untreated (Untx) (n = 5). a, Concentration of peripheral total memory CD4+ T cells (CAR−). b, Concentration of peripheral central memory (CD45RA−CD27+CCR7+; left panel), transitional memory (CD45RA−CD27+CCR7−; middle panel), and effector memory (CD45RA−CD27−CCR7−; right panel) CD4+ T cells (CAR−). c, Frequency of memory CD4+ T cell (CAR−) subsets in tissues 8 weeks post-infection. a, b, Significance was calculated using a two-sided Wilcoxon rank-sum test. Symbols and bars indicate mean, and error bars show ± SEM. Sample sizes indicate biologically independent animals.

Extended Data Fig. 5 |. Dual-CAR T cell product prevents CD4+ T cell loss despite persistent HIVMJ4 infection.

BLT mice were infused with Dual-CAR T cell product (TCP) (n = 6) 48 h post-HIVMJ4 challenge, while control mice were untreated (Untx) (n = 6). a, Concentration of peripheral total memory CD4+ T cells (CAR−). b, Concentration of peripheral central memory (CD45RA−CD27+CCR7+; right panel), transitional memory (CD45RA−CD27+CCR7−; middle panel), and effector memory (CD45RA−CD27−CCR7−; left panel) CD4+ T cells (CAR−). c, Frequency of memory CD4+ T cell (CAR−) subsets in tissues 8 weeks post-infection. a, b, Significance was calculated using a two-sided Wilcoxon rank-sum test. Symbols and bars indicate mean, while error bars show ± SEM. Sample sizes indicate biologically independent animals.

Extended Data Fig. 6 |. Dual-CAR T cells exhibit superior in vivo expansion compared to 4–1BB-costimulated, CD28-costimulated, and 3rd-generation CAR T cells.

a, BLT mice were challenged with either HIVJRCSF (n = 6) or HIVMJ4 (n = 6) and infused with 2×107 Dual-CAR T cell product (TCP). Fold-change in CAR T cell concentration from baseline to peak levels in peripheral blood. Data is the aggregate of both infection cohorts. b, Schematic shows the structural components of the 3rd-generation (3 G) CD4-based CAR construct. c–e, Dual-CAR T cell product and 3G-CAR T cells were combined at an equal frequency prior to infusion into uninfected mice (n = 9). c, FACS plots indicate the frequency of Dual-CAR and 3G-CAR T cells present within the pre-infusion T cell product. d, Longitudinal concentration of peripheral CAR T cells following adoptive transfer into HIV-negative mice. Symbols and error bars indicate mean ± SEM. e, At 2 weeks post-infusion, mice received either 107 irradiated K.Env cells (n = 6) or 107 irradiated K.WT cells (n = 3). Fold change in the concentration of peripheral CAR T cells 1-week post-K562 infusion from baseline concentration prior to K562 infusion. a, e, Bar and error bars indicate mean ± SEM, and symbols represent individual mice. a, d, e, Two-sided Wilcoxon rank-sum test was used to calculate significance. Sample sizes in these studies indicate biologically independent animals.

Extended Data Fig. 7 |. C34-CXCR4+ CAR T cells are selected for during chronic infection and exhibit superior ex vivo effector functions.

a, BLT mice were infected with HIVJRCSF and 48 h later infused with 107 C34-CXCR4+ Dual-CAR T cell product (TCP). FACS plots indicate the frequency of C34-CXCR4+ throughout infection. b, Mice were infected with HIVMJ4 and 48 h later were infused with 106 C34-CXCR4+ CAR.BBζ (n = 5), CAR.28ζ (n = 5), or purified Dual-CAR (n = 4) T cells. Frequency of C34-CXCR4+ CAR T cells in tissue 8 weeks post-infection. Thin dotted line indicates the frequency of C34-CXCR4+ CAR T cells in the pre-infusion TCP for the indicated CAR T cell type. Line and error bars indicate mean ± SEM. c, d, Mice were infected with HIVMJ4 and 48 h later received 106 C34-CXCR4+, purified CAR.BBζ.BBζ (n = 3), CAR.28ζ.28ζ (n = 4), or Dual-CAR (n = 3) T cells. c, FACS plots and (d) cumulative data show the frequency of each CD8+ CAR T cell population expressing MIP-1β and CD107a, and the frequency of CAR T cells with cytotoxic potential (granzyme B+ perforin+ CD107a+). CAR T cells were isolated from the spleen and bone marrow of mice 8 weeks post-infection and ex vivo stimulated. Significance was calculated using two-sided Wilcoxon matched-pairs signed rank test. For all data, symbols represent individual mice. Sample sizes in these studies indicate biologically independent animals.

Extended Data Fig. 8 |. CAR T cells from HIV-infected mice exhibit ex vivo cytotoxic function.

HIVJRCSF-infected mice (n = 3) treated with the Dual-CAR TCP were euthanized and the bone marrow cells were ex vivo stimulated with K.Env or K.WT cells for 24 h at the indicated E:T ratios. a, Representative FACS plots and (b) cumulative data shows the induction of active caspase-3 within the different target cell populations. Symbols and error bars indicate mean ± SEM. Sample size indicates biologically independent animals.

Extended Data Fig. 9 |. Dual-CAR and CAR.28ζ T cells exhibit similar ex vivo functional profiles.

BLT mice were challenged with HIVJRCSF (n = 5) and infused with 2×107 Dual-CAR T cell product (TCP) 48-hours post infection. a, Frequency of CD8+ and (b) CD4+ CAR T cell populations from tissue at necropsy (8-weeks post-infection) within the same mice expressing CD107a, MIP-1β, IL-2 and TNF after ex vivo stimulation. Bars and error bars indicate mean ± SEM, and symbols represent individual mice. Significance was calculated using two-sided Wilcoxon rank-sum test. c, Principle Components Analysis (PCA) of IL-2, TNF, MIP-1β, and CD107a expression in ex vivo stimulated CD8+ and CD4+ CAR T cells from PBMCs of HIVJRCSF-infected mice (n = 5). Sample sizes in these studies represent biologically independent animals.

Extended Data Fig. 10 |. HIV-resistant Dual-CAR TCP reduces virus replication in vivo.

a, Frequency of HIVGAG+ CD8− T cells (CAR−) within the bone marrow and spleen of HIVJRCSF-infected mice (n = 6) and (b) HIVMJ4-infected mice (n = 6) that were treated 48 h post-challenge with the Dual-CAR T cell product (TCP) or were untreated (Untx). c, Mean log plasma HIVMJ4 RNA (copies mL−1) after ART discontinuation of mice infused at ART initiation with 107 protected >98% C34-CXCR4+ (n = 5) or partially-protected <20% C34-CXCR4+ (n = 7) Dual-CAR TCP, or were untreated (n = 9). d, e, HIVBAL-infected mice were ART-treated and simultaneously infused with 107 HIV-resistant Dual-CAR TCP (n = 6) or were untreated (n = 6). d, Mean log plasma HIV RNA (copies mL−1). Shaded box indicates ART and arrow indicates CAR TCP infusion. e, Percent log reduction in plasma HIV RNA from pre-ART (week 3) and 0.5 and 1 week post-ART. For all data, bars and error bars indicate mean ± SEM, and symbols represent individual mice. Significance was calculated for (a–d) by two-sided Wilcoxon rank-sum test and (e) two-sided Kolmogorov-Smirnov test. Sample sizes in these studies indicate biologically independent animals.

Fig. 1 |. BLT mouse-derived HIV-specific CAR T cells are functionally indistinguishable from human-derived CAR T cells in vitro.

αCD3/CD28 Dynabeads were used to activate purified human T cells from a BLT mouse and healthy human donor, and then cells were transduced with the CD4-based CAR.ζ construct coexpressing GFP. a, FACS plots identify CAR.ζ T cells from each T cell source as GFP+ and CD4+. b, Following 10 d of culture, CD8+ CAR.ζ T cells were mixed with HIVYU2 Envelope+ K562 cells (K.Env) and upregulation of human cytokines was measured. c, Polyfunctionality profiles for combinatorial subsets of CD4+ and CD8+ CAR.ζ T cells producing 0 to 5 of the human cytokines GM-CSF, IFN-γ, IL-2, MIP-1β and TNF. Average of three unique donors per T cell source. d,e, HIV suppression assay as described in the Methods. d, FACS plots indicating the frequency of HIV-infected cells 6 d after coculturing with BLT mouse- or human-derived CAR.ζ T cells at denoted E/T ratios. e, Summary of frequency of HIV-infected cells (live CAR− CD8− cells) at 2, 4 and 6 d after coculture with BLT mouse- or human-derived CAR.ζ and untransduced (UTD) T cells at indicated E/T ratios. f,g, HIV elimination assay as described in the Methods. FACS plots (f) and summarized data (g) for frequency of active caspase-3 within HIV-infected cells (live CTV+ HIVGAG+ cells) 24 h after coculture with BLT mouse- or human-derived CAR.ζ and UTD T cells at 1:1 E/T ratio. Each symbol represents the average of technical duplicates per donor (n = 3 biologically independent donors). e,g, Symbols and lines reflect mean and error bars indicate ±s.e.m. RD1 is another name for phycoerythrin (PE).

Fig. 2 |. CAR T cells expressing the 4–1BB costimulatory domain exhibit a proliferative advantage and induce B cell aplasia in vivo.

a–e, BLT mouse-derived T cells were transduced with mCherry.T2A.CAR.ζ, iRFP670.T2A.CAR.BBζ or GFP.T2A.CAR.28ζ. In total, 5 × 106 CAR-transduced T cells of each type were mixed before infusion into syngeneic mice (n = 8). a, Frequency of each CAR T cell type within the preinfusion TCP. b, Frequency of peripheral CAR T cells within the same mouse at 5 weeks postinfusion. c,d, Peripheral concentration (c) and cumulative persistence (d) of each CAR T cell type over 5 weeks of engraftment. e, Relative tissue frequency of each CAR T cell type at 7 weeks postinfusion. f, In a separate study, 2 weeks after infusion of the CAR T cell mixture described in a, BLT mice received 107 irradiated K.WT (n = 8) or HIVYU2 Envelope+ (K.Env; n = 8) K562 cells. Peripheral concentration of each CAR T cell type following K.WT or K.Env infusion. g, FACS plots indicating frequency of MIP-1β+ and TNF+ CAR.BBζ and CAR.28ζ T cells within the same mouse after ex vivo stimulation. CAR.ζ T cells were too infrequent for analysis. h, Frequency of granzyme B+ perforin+ CD8+ CAR T cells within the same mice ex vivo. i–k, Mice were infused with 5 × 106 CD19-specific CAR.BBζ (n = 4) or control CD4-based CAR.BBζ T cells (n = 3). i, Concentration of peripheral CD19+ cells following infusion. j,k, FACS plots showing frequency of CD19+ cells (j), and number of CD19+ cells (k) in tissues at 7 weeks postinfusion. For all data, symbols and bars reflect mean and error bars show ±s.e.m., except d and k where symbols represent individual mice. d, Friedman’s test with Dunn’s multiple corrections test. f,h, Two-sided Wilcoxon matched-pairs signed rank test was performed to calculate significance. Sample sizes for all mouse groups indicate biologically independent animals. AUC, area under the curve; ctrl, control.

Fig. 3 |. HIV-specific CAR.BBζ T cells display features of T cell exhaustion after failing to control viral rebound.

a, Mean log plasma viral RNA (copies per milliliter) in HIVJRCSF-infected mice treated with ART from week 3 to 5 (G1 and G2 mice; gray box) or from week 3 to 8 (G3 and G4 mice; brown box). At 5 weeks post-infection, mice in G1 (n = 6) and G3 (n = 10) received 107 CAR.BBζ T cells, and mice in G2 (n = 6) and G4 (n = 9) received 107 inactive control CAR.BBΔζ T cells. Thin, dotted line denotes limit of quantification. Sample sizes in these studies indicate biologically independent animals. b,c, FACS plots (b) and summary data (c) illustrate the frequency of total memory CD4+ T cells (CAR−) following ART cessation in CAR.BBζ and CAR.BBΔζ T cell–treated mice. d,e, Concentration of peripheral CAR T cells for G1/G2 (d) and G3/G4 (e). f, Frequency of CAR T cells in tissues at 12 weeks post-CAR T cell infusion for G1/G3 and G2/G4. g, PD-1 and TIGIT expression on peripheral CAR.BBζ or CAR.BBΔζ T cells from G1/G2 after ART discontinuation. h–l, FACS analysis of splenic tissue of BLT mice (G1 and G2) 12 weeks after ART cessation. h, Coexpression of TOX and 2B4, PD-1 or TIGIT on peripheral CAR.BBζ or CAR.BBΔζ T cells. i, Frequency of TOX− and TOX+ CAR.BBζ T cells positive for indicated inhibitory receptors. j, Frequency of T-bet- and Eomes-expressing CAR.BBζ and CAR.BBΔζ T cells. k, Frequency of TOX expression within T-bet+ and Eomes+ CAR.BBζ and CAR.BBΔζ T cells. l, Memory distribution of 2B4+PD-1+TIGIT+ and EomeshiT-betdim CAR.BBζ T cells. For all data, symbols and bars reflect mean and error bars indicate ±s.e.m., except f and i–l where symbols represent individual mice. In a, f and i–k, significance was calculated using a two-sided Wilcoxon rank-sum test. Cerv., cervical; Mes., mesenteric; NS, not significant.

Fig. 4 |. Dual-CAR TCP mitigates CD4+ T cell loss and exhibits superior proliferative capacity.

a–i, Mice were challenged with HIVJRCSF (n = 12) or HIVMJ4 (n = 12) and 48 h later six mice from each group were infused with Dual-CAR TCP (green lines) or were untreated (Untx, black lines). a, Dual-CAR TCP comprises CAR.BBζ, CAR.28ζ and Dual-CAR T cells. b, Concentration of total peripheral CAR T cells in individual mice (green dotted lines; left y axis) and mean log plasma viral RNA (copies per milliliter) (solid lines; right y axis) in HIVJRCSF-infected mice. Thin black dotted line denotes limit of quantification. c, Frequency of peripheral memory CD4+ T cells (CAR-) in HIVJRCSF-infected mice. d, Concentration of total peripheral CAR T cells in individual mice (green dotted lines; left y axis) and mean log plasma viral RNA (copies per milliliter) (solid lines; right y axis) in HIVMJ4-infected mice. Thin black dotted line denotes limit of quantification. e, Frequency of peripheral memory CD4+ T cells (CAR-) in HIVMJ4-infected mice. f, Frequency of CD4+ T cell (CAR−) memory subsets in tissue from HIVMJ4- (left) HIVJRCSF-infected mice (right) at 8 weeks post-CAR T cell infusion. Six distinct tissues were analyzed from three biologically independent animals per infection cohort. g, Peripheral longitudinal frequency of each CAR T cell type present in the Dual-CAR TCP. h,i, Peak peripheral frequency (h) and cumulative persistence (i) of CAR T cells. Data are aggregated from both infection cohorts (n = 12). j–m, Equal frequencies of Dual-CAR TCP and 3G CD4-based CAR T cells were combined (Extended Data Fig. 6c) before infusion into HIVMJ4-infected mice (n = 6). j, Overlaid FACS plots showing frequency of peripheral Dual-CAR (iRFP670+NGFR+) and 3G-CAR (GFP+) T cells within the same mouse. k, Concentration of peripheral CAR T cells. l,m, Total number of splenic CAR T cells (l) and cumulative CAR T cell persistence (m) at 5 weeks post-infection. For all data, bars and error bars show mean ± s.e.m., and symbols represent individual mice, except c and e where symbols represent mean. Two-sided Wilcoxon rank-sum test was used to calculate significance, except h and i where Friedman’s test with Dunn’s multiple corrections test was used. Sample sizes in these studies indicate biologically independent animals.

Fig. 5 |. HIV-resistant Dual-CAR T cells mediate superior virus-specific immune responses.

a, Schematic of HIV-resistant (C34-CXCR4+) Dual-CAR T cells. b, HIVJRCSF-infected BLT mice received 107 CAR T cells at 48 h postchallenge. HIV DNA load in sorted CAR T cells from individual mouse splenic tissue (n = 8). c,d, HIVMJ4-infected mice were infused at 48 h postchallenge with 106 C34-CXCR4+, CAR.BBζ (n = 6), CAR.28ζ (n = 5) or purified Dual-CAR (n = 4) T cells. Longitudinal peripheral concentration (c) and peak peripheral CAR T cell concentration (d). e–n, HIVMJ4-infected mice were infused at 48 h postchallenge with 106 C34-CXCR4+, purified CAR.BBζ.BBζ (n = 5), CAR.28ζ.28ζ (n = 5) or Dual-CAR (n = 5) T cells, or were untreated (n = 4). Purification strategy is described in Supplementary Fig. 10. e, Frequency of CAR T cell populations out of total human CD45+ cells 2 and 3 weeks post-infection. f,g, Longitudinal concentration (f) and cumulative peripheral CAR T cell persistence (g). h, FACS plots showing CCR5 expression within peripheral memory CD4+ T cells (CAR−). i,j, Concentration of total memory (i) and CCR5+ (j) CD4+ T cells (CAR−) at 6 weeks post-infection. k,l, FACS plots (k) and frequency (l) of MIP-1β+ and CD107a+ CD8+ CAR T cells from tissue at 8 weeks post-infection after ex vivo stimulation. m,n, Distribution (m) and frequency (n) of granzyme B+ perforin+ cells within CD107a+ CAR T cells from tissues after ex vivo stimulation. b, Two-sided Wilcoxon matched-pairs signed rank test was used to calculate significance. For remaining analyses, significance was calculated using a two-sided Wilcoxon rank-sum test. Bars and error bars indicate mean ± s.e.m., and symbols represent individual mice, except for c where symbols indicate mean. Sample sizes in these studies indicate biologically independent animals.

Fig. 6 |. Mitigating CAR T cell infection improves control over HIV replication.

a, Mean log plasma viral RNA (copies per milliliter) of active, unprotected CAR T cell–treated mice (n = 38), and untreated/inactive CAR T cell–treated mice (n = 36). Data are aggregated across six independent studies. Thin dotted line denotes limit of quantification. b, Mean log plasma viral RNA (copies per milliliter) in mice infused at 48 h post-HIVMJ4 challenge with 107 protected (>98% C34-CXCR4+) Dual-CAR TCP (n = 12) or untreated mice (n = 12). c,d, Frequency of splenic HIV-infected CD8− T cells (CAR−) (c) and cell-associated HIV DNA load in lymph nodes (d) at 6–8 weeks post-infection. e–k, HIVJRCSF-infected mice were ART-treated and simultaneously infused with 107 HIV-resistant Dual-CAR TCP (n = 12) or inactive Dual-ΔCAR TCP (n = 5), or were untreated (n = 7). e, Mean log plasma HIV RNA (copies per milliliter). Shaded box indicates ART and arrow indicates CAR TCP infusion. f, Percentage log reduction in plasma HIV RNA from pre-ART (week 3) to 1 and 1.5 weeks post-ART. g,h, Data are aggregated from HIVJRCSF- and HIVBAL-infected cohorts. g, Correlation between percentage viral load reduction at first post-ART time-point and contemporaneous peripheral CAR T cell concentration. h, Kaplan–Meier curve of time to viral suppression after treatment initiation for Dual-CAR TCP versus control mice. i,j, Frequency of HIV-infected CD8− T cells (CAR−) (i) and HIV-infected CD14+ macrophages (j) aggregated from various tissues of plasma viremia suppressed mice. k, Cell-associated HIV DNA load in sorted central memory (CAR−CD45RA−CCR7+) CD4+ T cells. Statistical significance was calculated by one-sided unpaired Student’s t-test (a and b), two-sided Wilcoxon rank-sum test (c–f and i–k), Spearman correlation (g) and log-rank test (h). Bars and error bars indicate mean ± s.e.m. Symbols represent individual mice. Sample sizes in these studies indicate biologically independent animals. r, coefficient of correlation.