In vivo killing of primary HIV-infected cells by peripheral-injected early memory–enriched anti-HIV duoCAR T cells

HIV-specific chimeric antigen receptor–T cell (CAR T cell) therapies are candidates to functionally cure HIV infection in people with HIV (PWH) by eliminating reactivated HIV-infected cells derived from latently infected cells within the HIV reservoir. Paramount to translating such therapeutic candidates successfully into the clinic will require anti-HIV CAR T cells to localize to lymphoid tissues in the body and eliminate reactivated HIV-infected cells such as CD4+ T cells and monocytes/macrophages. Here we show that i.v. injected anti-HIV duoCAR T cells, generated using a clinical-grade anti-HIV duoCAR lentiviral vector, localized to the site of active HIV infection in the spleen of humanized mice and eliminated HIV-infected PBMCs. CyTOF analysis of preinfusion duoCAR T cells revealed an early memory phenotype composed predominantly of CCR7+ stem cell–like/central memory T cells (TSCM/TCM) with expression of some effector-like molecules. In addition, we show that anti-HIV duoCAR T cells effectively sense and kill HIV-infected CD4+ T cells and monocytes/macrophages. Furthermore, we demonstrate efficient genetic modification of T cells from PWH on suppressive ART into anti-HIV duoCAR T cells that subsequently kill autologous PBMCs superinfected with HIV. These studies support the safety and efficacy of anti-HIV duoCAR T cell therapy in our presently open phase I/IIa clinical trial (NCT04648046).


Generation of anti-HIV duoCAR-T cell products on the CliniMACS Prodigy 28
For clinical-scale CAR-T cell manufacturing, leukopaks were obtained from Key Biologics (HIV -) 29 or Vitalant (HIV + ) and shipped overnight priority to Lentigen Technology, Inc. (LTI)  were washed twice as described above. Stained cells were resuspended in wash buffer (300 µL) 94 and analyzed using the MQ10 flow cytometry analyzer (Miltenyi Biotec). Prior to start of the 95 analysis, the instrument was properly calibrated using flow cytometry calibration beads (Miltenyi 96 Biotec) and compensated using single stain controls. For T cell memory phenotyping, the pre-97 infusion (Day 8, harvest) anti-HIV duoCAR-T cell product was stained with an antibody cocktail 98 (T cell memory panel) consisting of APC-Vio770-conjugated anti-CD45RA (Miltenyi Biotec,, PE-Vio770-conjugated anti-CCR7 (Miltenyi Biotec, 100 Catalog No. 130-118-488/Clone: REA108) and FITC-conjugated anti-CD95 (Miltenyi Biotec,. Within the live lymphocyte gate (>95% CD3 + T cells), the 102 T cell memory markers CD45RA and CCR7 were used to stratify T cell populations into naive 103 appropriate biological (e.g., UTD T cell control) and/or fluorescence minus one (FMO) control 107 prior to start of the run. At minimum, 30,000 events were acquired for each sample. FlowJo 108 version 10 software was used to analyze cell populations. 109 110 Long-term co-culture cytotoxicity assay using HEK 293T-Env + GFP + target cells 111 The long-term killing efficacy of anti-HIV duoCAR-T cells was evaluated by challenging anti-HIV 112 duoCAR-T cells with HEK293T-Env + Luc + GFP + (Env + GFP + ) target cells on a weekly basis over a 113 17-day period. Briefly, on day 0, the anti-HIV duoCAR-T cells (0.6 × 10 6 cells/well) were co-114 cultured with Env + GFP + target cells (2 × 10 6 cells/well) at an effector to target (E:T) ratio of 0.3:1 115 in TexMACS medium without cytokine supplementation. To determine CAR-mediated killing of 116 Env + GFP + target cells, a portion of the co-cultured cells was evaluated by flow cytometry at 117 baseline (day 0) and on days 1, 4, and 7 post-challenge via detection of residual target cell 118 numbers using Absolute Counting Beads (Thermo Fisher Scientific). The co-cultured cells were 119 stained with the 7-AAD cell viability dye to determine cytotoxicity-induced target cell apoptosis in 120 the Env + GFP + cell population. Flow cytometry data was acquired on the MQ10 flow cytometer 121 (Miltenyi Biotec) and analyzed by Flow Jo version 10 software (Tree Star). At the end of the first 122 round of co-culture, cells in the culture were counted and fresh Env + GFP + target cells were added 123 at an approximate E:T ratio of 0.3:1. The second and third challenge round were conducted in the 124 same way as the first round. 125

Infectious molecular clones of HIV-1 Renilla luciferase (LucR) reporter viruses 127
We utilized HIV infectious molecular clones (IMC) which encode the Renilla luciferase (LucR) 128 reporter as a surrogate marker to quantify productive HIV infection. As previously described (67), pNL4-3, to express, in cis, the LucR reporter gene and heterologous env gene sequences of selected HIV-1 strains. Virus stocks were produced by transfection of 293T cells with the indicated 132 proviral plasmid DNAs, and the infectious titers (IU/mL) were determined on TZM-bl cells as

In vitro CAR-T cell mediated killing of HIV-infected PBMCs assay 139
The capacity of duoCAR-T cells derived from seronegative donors to eliminate acute HIV-infected 140 cells in vitro was quantified as we previously described (6). For studies using PWH donors, CD8 + 141 T cells were depleted from donor PBMCs by incubating with anti-CD8 magnetic beads (Miltenyi 142

Biotec) followed by T cell activation with phytohemagglutinin A (PHA, 4 µg/mL). Activated T cells 143
were cultured for 2 days in complete RPMI media supplemented with 10% fetal bovine serum and 144 recombinant human IL2 (100 IU/mL), washed, plated in a 96-well U bottom plate (100,000 145 cells/well) and spinfected with the indicated HIV-LucR IMC virus (~1x10 6 IU/mL). Autologous 146 duoCAR T cells or untransduced (UTD) control T cells were added at an effector to target (E:T) 147 ratio of 1:1 and cultured for 3 days (acute infection). Alternatively, spinfected cells were cultured 148 for 3 days to establish HIV infection followed by addition of syngeneic CAR T cells or UTD control 149 T cells at an effector to target (E:T) ratio of 1:1 and cultured for an additional 3 days. The cells 150 were then lysed, and the magnitude of HIV infection was quantified by measuring LucR levels 151 using the Renilla Luciferase System (Promega) as previously described (36,67). 152 153

In vitro CAR-mediated killing of HIV-infected monocytes assay 154
Monocytes were isolated and cultured as previously described (69). Briefly, uninfected PBMCs 155 from leukopaks obtained from the New York Blood Center were isolated by Ficoll density gradient centrifugation using Ficoll-Paque Plus (GE Healthcare). CD14 + monocytes were isolated by isolated monocytes were cultured (2 x 10 6 cells/mL) for 3 days non-adherently in Teflon flasks 159 with complete RPMI media containing macrophage colony-stimulating factor (M-CSF, 10 ng/mL) 160 (Peprotech) to increase the mature CD14 + CD16 + monocyte subpopulation (70, 71). After 3 days, 161 monocytes (10 x 10 6 cells/mL) were either uninfected or infected with HIV BaL-LucR virus (~1x10 6 162 IU/mL). After 8 hours, the cells were centrifuged, and the excess virus was removed, resuspended Cell-associated total HIV DNA qPCR was performed by BioQual, Inc. as previously described (6). 175 Briefly, genomic DNA was isolated from one million total CD4 + and CD8 + T cells at baseline and 176 at various time points during the CAR-T cell manufacturing process using the DNeasy Blood and 177 Tissue kit (Qiagen) as per the manufacturer's instruction. Eluted DNA was subjected to real-time 178 qPCR using primers and probes specific for wild type Gag HIV-1 and normalized to human β-179 actin gene. Total HIV DNA results were expressed as copies of total HIV DNA per 10 6 β-actin 180 equivalents. 181 previously described by our group (6). Results were expressed as average vector copy number 185 per cell, per transduced cell, or per 10 6 PTBP2 copies as indicated in the study. injected into the tail vein and the mice were monitored for the duration of the study for behavioral 209 changes, overt phenotypic abnormalities, and weight loss as a potential consequence of the 210 injected CAR-T cells. Mice were euthanized after 17-18 days and the spleens from one group of 211 mice were harvested, homogenized into single cells suspensions, red blood cells were lysed, and 212 total splenocytes were partitioned to analyze luciferase levels, CD4 + and CD8 + T cell populations, 213 total cell-associated (CA)-HIV DNA, and the CAR-T cell population. Splenocyte luciferase levels 214 were measured using the Renilla Luciferase Assay System (Promega) as previously described 215 (6, 67). 216 217

Flow cytometric analysis of CD4 + and CD8 + T cell populations in the mouse spleens 218
After the mouse spleens were harvested, they were dissociated to generate a single-cell 219 suspension which was treated with red cell lysis buffer to remove red blood cells, pelleted, and 220

Determination of anti-HIV duoCAR-T cell persistence and biodistribution in mice 231
Blood and major organs (brain, lungs, liver, heart, blood, spleen, and intestinal/gut) were 232 harvested from humanized NSG mice intravenously injected with CAR-T cells (2 x 10 6 cells) and 233 flash frozen for the detection of CAR-specific DNA in blood and tissues. CAR-T biodistribution Results in the study were expressed as cFrag copies per million PTBP2 copies. 236 237

Mass cytometry by Time-Of-Flight (CyTOF) analysis 238
Cryopreserved untransduced (d0) and duoCAR-T cell (d8) enriched CD4 + and CD8 + T cells from 239 uninfected and PWH donors were thawed, and 1-2 million cells were stained for mass cytometry 240 following a previously published protocol (72, 73). A list of CYTOF antibodies is provided in 241 Supplemental data file S1. Briefly, dead cells were removed by incubating the samples for one 242 minute with 25mM Cisplatin (Sigma-Aldrich) in phosphate buffered saline (PBS) containing EDTA. 243 Surface staining was performed using metal-tagged antibodies in PBS with 0.5% bovine serum 244 albumin (BSA) for 30 minutes at room temperature followed by fixation and cells were 245 permeabilized following manufacturer's instructions for the eBioscience Foxp3/Transcription 246

Data quality control for CyTOF analysis 256
Following CyTOF data acquisition, the FCS files were normalized using bead standards and the 257 data normalization algorithm using the R package 'premessa' 258 (https://github.com/ParkerICI/premessa). The live cell events were debarcoded using a single-cell 259 debarcoding algorithm (75) and we analyzed >25,000 (mostly >50,000) cells per sample.  Supplemental Figure 3. Gating strategy for long-term cytotoxicity assay. Anti-HIV duoCAR-T cells were challenged with HEK293T-engineered Env + GFP + cells and stained with VioBlue-conjugated anti-CD3 to discriminate total effector T cells (CD3 + GFPcells) from GFP + Env-expressing target cells (CD3 -GFP + cells). Dead cells were excluded from the analysis using 7-AAD live/dead cell exclusion dye. Total cells in the co-culture samples were gated on forward (FSC-A) and side scatter (SSC-A) plots followed by exclusion of cell doublets from the analysis (FSC-H versus FSC-A) and discrimination of live cells (7-AAD negative) versus dead cells (7-AAD positive) via 7-AAD staining (PerCP-Vio700-A channel). Gates were set using biological or FMO controls to identify total effector T cells containing duoCAR-T cells (V1-A::VioBlue-A channel) and Env + GFP + (B1-A::FITC-A channel). Counting beads were used to enumerate absolute cell numbers.

Supplemental Figure 4. Representative plot of the T SCM /T CM anti-HIV duoCAR-T cell memory phenotype.
The pre-infusion anti-HIV duoCAR-T cell product (Day 8, harvest) was stained with an antibody cocktail (T cell memory panel) consisting of APC-Vio770-conjugated anti-CD45RA, PE-Vio770-conjugated anti-CCR7 and FITC-conjugated anti-CD95. Within the live lymphocyte gate (not shown), the T cell memory markers CD45RA, CCR7, and CD95 were used to stratify T cell populations into naive (T N ), central memory (T CM ), effector memory (T EM ), and effector memory T cells re-expressing CD45RA (T EMRA ). Early-memory stem cell or stemcell like (T SCM ) cells were differentiated from T N cells based on co-expression of CCR7 and CD95. Anti-HIV duoCAR-T cells were manufactured using the CliniMACS Prodigy (clinical process).