Targeting PI3K decreases trogocytosis-mediated CAR T cell fratricide
CAR T cell-mediated trogocytosis (CMT) is a proposed mechanism of tumor immune evasion, in which the cognate interaction between CAR T cells and tumor cells results in the transfer of target antigen from the tumor cell surface to the CAR T cell. In addition to the emergence of antigen-negative target cells, antigen-positive CAR T cells become the targets of other CAR T cells in a process called fratricide (Fig. 1a ). Multiple strategies to overcome CMT have been pursued by different research groups. Hamieh et al. [ 1 ], who initially described CMT, used a strategy that simultaneously targets multiple antigens to reduce the risk of tumor escape. Lu et al. demonstrated that overexpressing 25-hydroxycholesterol (25HC) in mesothelin-targeting CAR T cells decreased trogocytosis and improved CAR T cell efficacy. They also discovered that activating transcription factor-3 (ATF3) downregulated cholesterol 25-hydroxylase expression, thereby reducing 25HC expression and increasing CMT. Li et al. [ 2 ] employed a dual CAR system, one that binds CD19 on tumor cells and activates the NK cells, and an inhibitory CAR targeting CS1 to prevent fratricide. Olson et al. [ 3 ] demonstrated that using the low-affinity CD19 scFv, CAT, instead of FMC63 in CAR T cells decreased CMT. Koh et al. [ 4 ] simultaneously administered anti-CD19 monoclonal antibody along with CD19 CAR T/NK cells and showed that the combination treatment reduced trogocytosis resulting in better tumor control and survival in vivo. However, the mechanism driving CMT remains poorly understood. Fig. 1: PI3K interaction with CD3ζ regulates CAR trogocytosis. Full size image a Schema representing CAR T cell trogocytosis, CAR-antigen (Ag) interaction leads to loss of Ag from tumor cells and Ag presentation on CAR T cells leads to fratricide. b Schema representing components of wild type (WT) CAR and mutant form that lacks signaling domain CD3ζ and 4-IBB (ΔSig CAR T cells). Components of the wild-type CAR include an antigen-binding domain, a single-chain variable fragment (scFv), a hemagglutinin tag (HA), a CD8 transmembrane domain (TMD), a co-stimulatory domain 4-1BB, and a signaling domain CD3ζ. c–e WT or ΔSig CD19 CAR T cells were co-cultured with B-ALL cell line NALM6 at a 1:1 ratio for one hour, and surface expression of CD19 was measured by flow cytometry. c Panel showing a representative histogram of CD19 expression on NALM6 cells. CD19 fluorescence intensity represented as median on NALM6 cells d and CAR T cells e of four replicates and statistical significance assessed by two-sided Student’s t test. f PI3K p110δ was knocked down using short hairpin RNA (shRNA) in T cell line Jurkat cells stably expressing CD19 CAR. Immunoblotting of cell lysate with antibodies specific for PI3K p110δ and beta actin. g PI3K p110δ knocked down Jurkat CD19 CAR T cells were co-cultured with NALM6 cells at a 1:1 effector: target ratio, and surface expression of CD19 on NALM6 was measured by flow cytometry and represented as a histogram. h Equal numbers of CD19 CAR T cells were incubated with different concentrations of ME401 in DMSO and cultured for 20 h; live cells were counted by flow cytometry using DAPI as a live-dead marker. Data represented as mean ± SD of triplicates, and statistical significance assessed by one-way ANOVA with Sidak’s multiple comparison test. i CD19 CAR T cells were incubated with specific PI3K p110δ inhibitor ME401at different concentrations for 20 h and then co-cultured with NALM6 cells for one hour at a 1:1 ratio, and the levels of phosphorylated and pan AKT were measured by immunoblotting. j–l CD19 CAR T cells were incubated with ME401 for 20 h and then co-cultured with NALM6 cells for 15 min at a 5:1 E:T ratio j, k or 1:1 ratio for 30 min l , and CD19 expression on NALM6 and CAR T cells was measured by flow cytometry. Data shown in triplicate and statistical significance by an unpaired two-tailed Student’s t-test. m, n ME401 or DMSO-treated CD19 CAR T cells were co-cultured with NALM6 cells for 1 h at a 1:2 E:T ratio, and then fixed, permeabilized, and stained with Phalloidin AF-594, and fluorescence was measured by flow cytometry. o WT or ΔSig CD19 CAR T cells were co-cultured with NALM6 cells at a 1:2 E:T ratio for one hour, and then CAR was pulled down using anti-HA antibodies from cell lysates and probed for the presence of P13K p110δ and CAR protein by western blotting. p BCMA CAR T cells were co-cultured with or without myeloma cell line U266, and CAR was pulled down using anti-HA antibodies and probed for the presence of P13K p110δ and CAR protein by western blotting. q ME401 or DMSO-treated CD19 CAR T cells were co-cultured with NALM6 cells at a 1:2 E:T ratio for one hour, and the CAR protein was pulled down using anti-HA antibody and probed for P13K p110δ and CAR protein by western blotting. r Schema representing the proposed function of PI3K in regulating CAR T cell trogocytosis. All experiments were repeated at least 3 times except for Figs. 1 h , 1 i , and 1 p which were repeated twice. Here, we aimed to elucidate the proximal molecular mechanism driving CMT. We generated CD19 CAR T cells that lack intracellular CD3ζ and co-stimulatory signaling domains (Δsig) (Fig. 1b ) and co-cultured them or wildtype (WT) CD19 CAR T cells with the CD19 + B cell acute lymphoblastic leukemia (B-ALL) cell line, NALM6. We observed significantly greater CD19 loss in NALM6 cells co-cultured with WT compared to Δsig CAR T cells (Fig. 1c, d ). In addition, we found increased transfer of CD19 to WT CAR T cells compared to Δsig CAR T cells (Fig. 1e ). Previous studies on T cell receptor (TCR) trogocytosis identified phosphoinositide-3-kinase (PI3K) as an important mediator of trogocytosis [ 5 ]. CARs and TCRs share CD3ζ, which is critical for PI3K pathway activation. While the PI3K family comprises many isoforms, PI3K p110δ (PI3Kδ), is considered the main PI3K isoform involved in proximal TCR signaling [ 6 ]. We generated CD19 CAR-expressing Jurkat T cells and then downregulated PI3K/p110δ using shRNA (shPI3Kδ, Fig. 1f ). CD19 expression was significantly reduced on NALM6 cells co-cultured with WT compared to shPI3Kδ CD19 CAR Jurkat cells (Fig. 1g ), suggesting that PI3K is an important contributor to CMT. To inhibit PI3K in primary CAR T cells, we used the clinically validated PI3Kδ-specific inhibitor ME401 (zandelisib) [ 7 ]. We first determined the highest concentration of ME401 that does not affect CAR T cell viability to be 2 µM by incubating CD19 CAR T cells with ME401 for 20 h (Fig. 1h ). At the same concentration, we observed that the phosphorylation of AKT, a downstream target of PI3K, was substantially reduced in ME401-treated CAR T cells compared to DMSO-treated CAR T cells confirming PI3K inhibition (Fig. 1i ). Furthermore, PI3K inhibition by ME401 was transient with AKT phosphorylation returning to baseline 72-h post treatment (Supplementary Fig. 1 ) We next co-cultured ME401- or DMSO-treated primary human CD19 CAR T cells with NALM6 cells and observed that CD19 loss on tumor cells (Fig. 1j, k ) and antigen transfer to CAR T cells (Fig. 1l ) was reduced following ME401 treatment. We hypothesized that PI3K may mediate CMT by increasing actin polymerization, which is known to be critical for trogocytosis [ 1 , 8 ]. Indeed, we found that ME401-treated CD19 CAR T cells co-cultured with NALM6 cells exhibited reduced actin cross-linking (Fig. 1m,n ). PI3K-mediated actin polymerization has previously been shown to mediate TCR internalization following PI3K binding to CD3ζ [ 5 ]. To study whether a similar interaction occurs between PI3Kδ and CAR molecules, we co-cultured ∆sig or WT CD19 CAR T cells with NALM6 cells. Following co-culture, CAR-containing complexes were co-immunoprecipitated using an anti-hemagglutinin tag (HA) antibody (Fig. 1b ) and probed for PI3Kδ/p110δ. We found that PI3Kδ was only complexed with CAR in WT CD19 CAR T cells but not Δsig CAR T cells and required the presence of CD19 + target cells, suggesting a specific interaction between CAR and PI3K upon activation (Fig. 1o ). PI3K is known to interact with CD19 in B cells [ 9 ], and we hypothesized that, instead of PI3K binding to the CAR, pull-down of PI3K could be the result of the CAR binding to CD19, which could itself be associated with PI3K in the target cells. To investigate this possibility, we studied whether using a CAR construct targeting an antigen not associated with PI3K would still result in PI3K pull-down. BCMA is not known to bind to PI3K, and co-immunoprecipitation using lysates from BCMA CAR T cells co-cultured with BCMA + myeloma cell line U266B1 (U266) again resulted in pulldown of PI3K (Fig. 1p ). This finding indicates that the CAR-PI3K interaction is agnostic of the CAR target and not the result of PI3K binding to the antigen. We further found that the interaction between CAR and PI3K could be inhibited by pretreating the CAR T cells with ME401 (Fig. 1q ). To test whether PI3K inhibition would affect CMT in the solid tumor setting, we co-cultured ME401 or DMSO pretreated folate receptor-1(FOLR1)-specific CAR T cells with ovarian cell line OVCAR3 and found reduced FOLR1 antigen transfer in the ME401 treated group (Supplementary Fig. 2 ). We suggest a working model of CMT that involves CAR T cell activation-induced PI3K interaction with the CAR, resulting in actin polymerization and subsequent transfer of target antigen, leading to loss of antigen on the tumor cell surface and presentation of antigen on the CAR T cell surface (Fig. 1r ). We next investigated whether this decrease in antigen transfer could result solely from reduced T cell function, specifically anti-tumor activity and proliferation, rather than from a specific effect on CMT. However, ME401-treated CD19 and BCMA CAR T cells showed increased tumor cell killing (Fig. 2a, b , Supplementary Fig. 5a ) and expansion (Fig. 2c, d ) when co-cultured with target cells. This increased CAR T cell expansion was also associated with reduced apoptosis, possibly due to less fratricide (Fig. 2e , Supplementary Fig. 1 ). ME401-treated CD19 and BCMA CAR T cells also exhibited reduced expression of exhaustion markers LAG3, PD1, TIM3, and TIGIT following co-culture (Fig. 2f , Supplementary Fig. 3 ) and no significant changes in CAR T cell phenotype following short PI3K inhibitor pretreatment (Supplementary Fig. 2 , 5b ). Interestingly, despite the increased cytotoxicity of ME401-treated CD19 and BCMA CAR T cells, we observed lower levels of inflammatory cytokines (Fig. 2g , Supplementary Fig. 4 ). Next, we determined whether pretreatment with ME401 could have beneficial effects on CAR T cell fratricide and T cell function in vivo (Fig. 2h ). ME401-pretreated CD19 CAR T cells demonstrated increased in vivo tumor control (Fig. 2i ) and substantially increased CD19 CAR T cell numbers (Fig. 2j,k , Supplementary Fig. 5c ). In addition, ME401-pretreated CAR T cells showed lower expression of LAG3 and PD1 (Fig. 2l, m ) in line with our in vitro results. Fig. 2: PI3K inhibition reduces CAR T cell fratricide and improves efficacy. Full size image a–d CAR T cells are pretreated with ME401 or DMSO (carrier) for 20 h and washed prior to experiments. ME401 (1μM) or DMSO-treated CD19 or BCMA CAR T cells were co-cultured with NALM6 or U266 cells labeled with cell trace far red (CTFR) for 90 min at different E:T ratios. To measure the killing of CAR T cells, an equal number of CAR T cells corresponding to each E:T ratio was cultured without target cells as a control for the same time. Similarly, NALM6 or U266 cells were cultured in the absence of CAR T cells as a control to measure target cell killing. Cell numbers were measured in a fixed-volume co-culture using a Cytek Aurora flow cytometer. Number of viable NALM6 cells or U266 cells (DAPI-) in co-culture with corresponding CAR T cells, normalized to control wells. a, b Number of viable CD19 or BCMA CAR T cells (DAPI-) in the