Novel Fas-TNFR chimeras that prevent Fas ligand-mediated kill and signal synergistically to enhance CAR T cell efficacy

The hostile tumor microenvironment limits the efficacy of adoptive cell therapies. Activation of the Fas death receptor initiates apoptosis and disrupting these receptors could be key to increasing CAR T cell efficacy. We screened a library of Fas-TNFR proteins identifying several novel chimeras that not only prevented Fas ligand-mediated kill, but also enhanced CAR T cell efficacy by signaling synergistically with the CAR. Upon binding Fas ligand, Fas-CD40 activated the NF-κB pathway, inducing greatest proliferation and IFN-γ release out of all Fas-TNFRs tested. Fas-CD40 induced profound transcriptional modifications, particularly genes relating to the cell cycle, metabolism, and chemokine signaling. Co-expression of Fas-CD40 with either 4-1BB- or CD28-containing CARs increased in vitro efficacy by augmenting CAR T cell proliferation and cancer target cytotoxicity, and enhanced tumor killing and overall mouse survival in vivo. Functional activity of the Fas-TNFRs were dependent on the co-stimulatory domain within the CAR, highlighting crosstalk between signaling pathways. Furthermore, we show that a major source for Fas-TNFR activation derives from CAR T cells themselves via activation-induced Fas ligand upregulation, highlighting a universal role of Fas-TNFRs in augmenting CAR T cell responses. We have identified Fas-CD40 as the optimal chimera for overcoming Fas ligand-mediated kill and enhancing CAR T cell efficacy.


INTRODUCTION
Adoptive transfer of chimeric antigen receptor (CAR) T cells has seen remarkable success in the treatment of relapsed/refractory hematological cancers; however, approximately 60% of patients eventually relapse, partly due to the hostile tumor microenvironment (TME). 1 Extending these clinical successes to solid tumor indications is more challenging due to an even more complex and immunosuppressive TME. [1][2][3] The Fas/Fas ligand (FasL) pathway is a key inhibitory checkpoint contributing to the immunosuppressive TME. [4][5][6][7] Fas is a member of the tumor necrosis factor receptor (TNFR) superfamily and comprises one of eight TNFR death receptors. 8 Upon binding FasL, Fas trimerizes allowing for binding of the adaptor protein, Fas-associated death domain (FADD), to the intracellular death domains of Fas via homotypic interactions. 9 Pro-caspase-8 then binds FADD via death effector domains, creating the death-inducing signaling complex, and is then cleaved to activate downstream executioner caspases, initiating apoptosis ( Figure 1A).
T cells constitutively express Fas and are consequently vulnerable to FasL-mediated apoptosis. The FASLG gene and FasL protein are overexpressed in many cancers, either by cancer cells themselves or by cells constituting the TME, such as regulatory T cells (Tregs), myeloid-derived suppressor cells (MDSCs), cancer-associated fibroblasts (CAFs), and tumor endothelial cells. 5,6 Moreover, T cells upregulate FasL upon activation, inducing fratricide, an effect particularly observed with third-generation CARs. 10,11 Therefore, the Fas/FasL checkpoint can limit the efficacy of adoptive T cell therapy.
Several strategies to overcome FasL in immunotherapy have been explored. Therapeutic monoclonal antibodies that block Fas or FasL effectively prevent FasL-mediated T cell loss; however, FasLmediated killing of tumor is concomitantly compromised. [12][13][14] Adoptive immunotherapy with engineered immune cells affords more discrete methods: disruption of Fas expression by small interfering RNAs or CRISPR-Cas9 is effective. 15,16 An alternative strategy is expression of non-functional Fas, which competes with native Fas. This latter strategy includes a truncated Fas receptor lacking the death domain (FasDDD) or a chimeric Fas-41BB protein. 5,[17][18][19] Expression of FasDDD or Fas-41BB rescues FasL-mediated apoptosis. The Fas-41BB chimera additionally converts the death signal into a pro-survival 4-1BB signal by activating NF-kB and mitogen-activated protein kinase (MAPK) pathways via TNFR-associated factors (TRAFs). 20 There are many other members of the TNFR superfamily apart from 4-1BB that provide co-stimulatory signals that, due to differential TRAF activation, may be qualitatively different. In this paper we Molecular Therapy: Nucleic Acids perform a functional assessment of Fas-TNFR chimeric proteins in the context of human T cells. We identify several novel Fas-TNFR chimeras that co-stimulate CAR T cells, delivering enhanced target cytotoxicity and CAR T cell persistence/proliferation compared with FasDDD and the Fas-41BB chimera. In particular, Fas-CD40 optimally enhanced CAR T cell efficacy when co-expressed with either 4-1BB-or CD28-containing CARs. Moreover, we demonstrate that a major source of FasL for Fas-TNFR activation derives from T cells themselves, highlighting a universal role of Fas-TNFRs to augment CAR T cell therapy.

Screening of Fas-TNFRs reveals Fas-CD40 as a potent inducer of proliferation upon binding FasL
We first created a set of Fas-TNFR chimeras comprising the ectodomain and transmembrane domain of Fas fused to the endodomains of pro-survival TRAF-interacting TNFRs, as well as the endodomain of the TRAIL decoy receptor, TRAIL-R4/DcR2 21 ( Figures 1A and 1B). The Fas-TNFR chimeras (highlighted in gold in Figure 1B) were co-expressed in primary human T cells with the RQR8 suicide/sort marker 22 and a first-generation CD19-targeting CAR Figure 1C), and were screened for their ability to resist FasL-mediated cell death and to alter T cell activity upon binding FasL. Screening was performed in two separate experiments due to the large numbers of Fas-TNFR chimeras.

Fas-TNFR chimeras protect from FasL-mediated kill
We next assessed how efficiently the Fas-TNFRs could rescue FasLmediated kill. We co-expressed the Fas-TNFRs with RQR8 and a second-generation CD19-targeting CAR (19-BBz; Figure 2A). FasDDD, Fas-CD27, and Fas-CD40 had the highest protein expression, respectively, followed by Fas-Fn14 then Fas-BCMA with Fas-41BB having the lowest ( Figures 2B and 2C). Upon co-culture with SupT1 cells engineered to express FasL ( Figure S4A), Fas-41BB could only partially rescue cell death as measured by cell survival; however, the percentage of apoptotic cells was indistinguishable from SupT1 control cells ( Figure 2G). Fas-TNFR-19-BBz CAR T cells did not proliferate autonomously when co-cultured with CD19 À SupT1 cells ( Figure 2E), nor did they increase tonic cytotoxicity relative to 19-BBz alone (Figure 2H). The level of cytotoxicity slightly increased against SupT1-FasL cells ( Figure 2H).
To stress test the Fas-TNFR chimeras we set up an in vitro restimulation cytotoxicity assay, where 19-BBz cells were serially challenged with Nalm6 cells and Nalm6 cells engineered to express FasL (Figure S5D). Expression of FasDDD and the Fas-TNFRs enhanced 19-BBz-mediated Nalm6 cytotoxicity, killing targets for all 10 stimulations ( Figure 3B), with Fas-CD40 inducing greatest proliferation, a 113-fold increase from the initial 1:8 effector to target (E:T) seeding ratio ( Figures 3B and 3C). Fas-CD40, Fas-BCMA, Fas-CD27, and Fas-Fn14 enhanced serial cytotoxicity against Nalm6-FasL cells compared T cells were treated as described in Figure 1F. TCR diversity genes included in the nCounter CAR-T Characterization Panel have been removed from this list. FC, fold change.
Fas-CD40 also enhances 19-28z CAR efficacy We next investigated whether the co-stimulatory activity of the Fas-TNFRs would be affected if we replaced the co-stimulatory endodomain within the CAR from 4-1BB to CD28 (19-28z; Figure 4A). 19-28z was co-expressed in human T cells with RQR8 and the Fas-TNFRs ( Figure S6A). As seen with 19-BBz co-expression, FasDDD displayed highest protein expression followed by Fas-CD27 and Fas-CD40, with Fas-41BB having the lowest expression ( Figure 4B), which again correlated with the ability to rescue FasL-mediated cell death ( Figures S6B and S6C). Fas-CD40, Fas-Fn14, and Fas-BCMA induced a slightly higher level of basal proliferation upon CD19 À SupT1 stimulation ( Figure S6B), which correlated with increased tonic cytotoxicity; however, this was not sustained (Figure S6D). This is different to what was seen with 19-BBz implying a difference in signal transduction. 19-28z cells co-expressing Fas-CD40, Fas-Fn14, and Fas-BCMA displayed greater cytotoxicity against SupT1-FasL cells, which was sustained with Fas-CD40 ( Figure S6D).
Fas-CD40-19-28z cells exhibited greatest Nalm6 and Nalm6-FasL serial cytotoxicity, completely killing targets for all 10 stimulations ( Figures 4C and 4D), which correlated with the level of CAR T cell proliferation (a 138-fold increase from the initial 1:8 E:T seeding ratio) and IFN-g and IL-2 secretion ( Figure 4E). The proliferative capacity of Fas-CD40-19-28z cells was not limitless, however, as CAR T cell proliferation decreased after the last two stimulations, as did the other Fas-TNFRs. Fas-Fn14, Fas-BCMA, and Fas-CD27 trended to augment target cytotoxicity similar to Fas-CD40; however, an outlier precluded definitive statistical analyses. Throughout the stimulations, Fas-CD40-19-28z cells maintained an earlier memory profile compared with the other Fas-TNFRs by having fewer CD45RA + CD62L À TEMRA cells across both CD8 and CD4 populations ( Figure 4F).

CAR T cell-derived FasL may be an additional source for Fas-TNFR activation
We observed from the in vitro restimulation experiments that coexpression of the Fas-TNFRs augmented both CD19-CAR and GD2-CAR T cell proliferation against target cells not exogenously expressing FasL (Figures 3B, 4C, and S7G). Staining for surface FasL expression in SupT1 and Nalm6 cells revealed that they did not express FasL, even in the presence of IFN-g ( Figures 5A and S8). Moreover, SupT1 cells cultured with 19-BBz or GD2-28z cells did not induce FasL-mediated CAR T cell apoptosis, further evidence that SupT1 cells did not express FasL ( Figure 5B). This indicated that tumor-derived FasL was not the only source of FasL in these co-cultures. We therefore hypothesized that T cell-derived FasL upregulated upon activation could be responsible. To address this, Fas-CD40-GD2-28z cells were serially stimulated with an anti-CAR idiotype antibody in the presence of anti-Fas and anti-FasL blocking antibodies. As expected, addition of anti-Fas/FasL antibodies significantly  Figure S6A. Ten independent donors tested, bars indicate means, ***p < 0.001, ****p < 0.0001, one-way ANOVA (Dunnett's multiple comparisons test relative to FasDDD). (C) 19-28z cells from four independent donors were stimulated up to 10 times with either Nalm6 FasKO or Nalm6 FasKO -FasL cells at a starting 1:8 E:T ratio, measuring for target survival and 19-28z cell counts after each stimulation. Effectors were stimulated with 50,000 targets for the first five stimulations and 100,000 targets for the final five stimulations, error bars are SEM. (D) Relative target survival of Nalm6 FasKO (left) and Nalm6 FasKO -FasL (right) cells after the ninth or fourth rounds of stimulation, respectively, as described in (C). *p < 0.05, **p < 0.01; ns, non-significant, two-way ANOVA, error bars are SEM. (E) Cell culture supernatants after the first round of target stimulation from the experiment described in (C) were analyzed for IFN-g and IL-2. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, non-significant, two-way ANOVA, error bars are SEM. (F) T cell memory phenotypes were analyzed for CD8 (top) and CD4 (bottom) cells after the fifth, seventh, and ninth stimulations from the restimulation experiment described in (C). Error bars are SEM, an "X" denotes where too few cells were present to accurately determine memory phenotype.

The Fas receptor is ubiquitously expressed in T cells and its activation upon binding FasL triggers apoptosis. 4-7 Many cancer cells express
FasL, in addition to TME cells such as MDSCs, CAFs, Tregs, and the tumor endothelium, 5,6 as well as T cells themselves. 10,11 Therefore, the Fas/ FasL checkpoint may inhibit cancer immunotherapeutic approaches such as adoptive cell therapy by limiting the persistence of T cells.
Strategies to overcome the Fas/FasL checkpoint include systemic antibody blockade. [12][13][14] Approaches applicable to adoptive immunotherapy also include genetic manipulation by FAS knockdown and knockout using siRNA and CRISPR-Cas9, respectively. 15,16 Additional approaches include the expression of non-functional Fas, such as FasDDD and Fas-41BB. Both FasDDD and Fas-41BB rescue T cells from FasL-mediated   upregulated DEGs relating to the cell cycle, chemokine and interleukin signaling, JAK-STAT, MAPK/PI3K, and NF-kB pathways, and metabolism. Notably, Fas-CD40 upregulated chemokine receptor/ligand genes: CCR8, CXCR3, CXCR4, CCL1, CXCL10, and CXCL13; which were confirmed at the protein level and have all been implicated in T cell trafficking and could facilitate T cell homing to tumors. [24][25][26][27] We also observed a trend for Fas-CD40 upregulating CCL3, CCL4, and CCL5 transcription; however, this did not reach statistical significance. Interestingly, CCR8 overexpression in CAR T cells enhanced tumor homing, driven by a feedforward loop of activated CAR T cells secreting CCL1 (the cognate ligand for CCR8). 24 Expression of the Fas-TNFRs increased 19-BBz-mediated in vitro serial cytotoxicity over FasDDD, except for Fas-41BB, against FasL-expressing targets, with Fas-CD40 inducing greatest proliferation upon serial target stimulation. Furthermore, we showed that Fas-CD40, Fas-Fn14, Fas-BCMA, and Fas-CD27 enhanced 19-BBz efficacy in vivo compared with FasDDD, with Fas-CD40 demonstrating a significant benefit over Fas-41BB. There was no significant survival advantage in vivo between FasDDD and Fas-41BB, suggesting that complementary transacting signaling domains between chimeras enhance CAR T cell efficacy, rather than increasing the amplitude of one signaling pathway. Incorporation of trans-acting chimeric receptors/signaling domains to enhance CAR T cell activity has been reported previously. [28][29][30][31] Enhanced CAR T cell-mediated serial cytotoxicity and proliferation upon Fas-CD40, Fas-BCMA, and Fas-Fn14 expression were confirmed in the context of a CD28-containing CAR (19-28z) and a CAR targeting a different cognate antigen (GD2-28z). Mechanistically, Fas-CD40 appears to demonstrate an advantage over other Fas-TNFRs by enhancing CAR T cell proliferation and maintaining T cell memory, particularly in the context of a CD28-containing CAR, likely mediated by increased TCF-1 expression 23 10 Importantly, although, this tonic activity did not persist, and rather than this tonic activity inducing functional exhaustion/dysfunction, the opposite was true with Fas-CD40 maintaining the capacity for serial target killing. Importantly, we did not observe any evidence of autonomous proliferation with Fas-CD40, or with any other Fas-TNFR, co-expressed with either 4-1BB-or CD28-containing CARs.
Interestingly, we observed the Fas-TNFRs enhanced CAR T cell proliferation and anti-tumor cytotoxicity even when we did not enforce FasL expression on target cells. We subsequently demonstrated that an additional source of FasL for Fas-TNFR activation derives from CAR T cells themselves, an effect observed with TCR-engineered Fas-41BB cells. 17,18 Expression of the Fas-TNFRs therefore creates a self-regulatable way to augment CAR T cell activation, irrespective of tumor FasL expression, whereby CAR activation (signals one and two) upregulates FasL surface expression, binding the Fas-TNFR on a sister CAR T cell, which delivers an additional third signal to the CAR T cell (Figure 7). This is akin to physiological TCR-mediated activation between a T cell and an antigen-presenting cell (APC), with APCs delivering additional signals to T cells via presentation of TNFR ligands, a concept explored with expression of full-length 4-1BB or OX40 in CAR T cells, which enhances their efficacy. 34,35 However, it remains to be determined whether CAR T cells would have the ability to physically interact with each other within the complex TME to mediate this effect.
As well as the chimeras highlighted above, some chimeras exhibited different effects on T cell function. Fas-LTbR and Fas-CD30 induced constitutive IFN-g secretion; however, they could not rescue FasLmediated kill. Expression of full-length LTbR in T cells has been shown to potentiate TCR-activated IFN-g secretion 36,37 ; however, it did not constitutively induce IFN-g, therefore the constitutive IFNg secretion observed with Fas-LTbR suggests the Fas ecto-and transmembrane domains might be clustering the chimera to form dimers, as has been described previously for Fas prior to ligand binding. 38 Fas-RANK induced strong constitutive activation of NF-kB, an effect observed with overexpressing full-length RANK 39 ; however, this did not correlate with an enhancement of proliferation or IFN-g secretion. Fas-DcR2 remarkably induced very high levels of NF-kB activation upon binding FasL, consistent with the literature that DcR2 activates NF-kB 21 ; however, this did not correlate with increased proliferation or IFN-g release.
Differences in functional activity between Fas-TNFRs are likely due to qualitative differences in TRAF recruitment. For example, CD40 and BCMA recruits TRAFs 1-3, 5, and 6 upon activation; whereas 4-1BB only recruits TRAFs 1-3. 8 However, this cannot solely explain the differences in Fas-TNFR performance, as OX40 and RANK, which did not induce proliferation upon binding FasL, also interact with TRAFs 1-3, 5, and 6. 8 Quantitative differences in the amount of recruited TRAFs to each TNFR will likely also dictate the amplitude of signaling output. TRAF6 could likely be responsible for differentiating between Fas-TNFR function, as TRAF6 appears to be the predominant TRAF for CD40-induced NF-kB activation in dendritic cells, 40 and TRAF6 also binds BCMA and Fn14, the chimeras of which augmented CAR T cell activity akin to Fas-CD40. TRAF6 is unique from the other TRAFs in several ways: having a different binding motif (P-x-E-x-x-[acidic/aromatic residue]); being involved beyond TNFR signaling such as IL-1R and Toll-like receptor signaling 41 ; and being able to activate the Src-family tyrosine kinases resulting in Akt activation via PI3K, in addition to activation of transcription factors NF-kB and AP-1, the latter of which being common among other TRAFs. 41,42 Fas-TNFR co-expression with 4-1BB-or CD28-containing CARs adds a further layer of signaling complexity, particularly because CD28 belongs to the immunoglobulin superfamily and as such recruits www.moleculartherapy.org different signaling proteins to TNFRs, namely SH2 and SH3 domaincontaining proteins such as Grb2, PI3K, and Lck. [43][44][45][46] Protein expression of the Fas-TNFR chimera does not appear to determine its co-stimulatory activity, as Fas-CD27 consistently had the highest expression across multiple CAR architectures; however, its ability to augment CAR T cell activation varied depending on the CAR co-stimulatory domain. Similarly, Fas-BCMA, which had relatively low expression, was able to enhance CAR T cell activation akin to Fas-CD40 with either 4-1BB-or CD28-containing CARs. It is possible that even a low level of Fas-TNFR expression will saturate the amount of available endogenous TRAFs. Fas-TNFR expression does seem particularly important for rescuing FasL-mediated kill, however.
CD40 is typically expressed in APCs such as macrophages, B cells, and dendritic cells, interacting with CD40 ligand on T cells, functioning in a co-stimulatory manner known to activate both canonical and non-canonical NF-kB pathways. 47,48 However, CD40 is also expressed in T cells, similarly functioning in a co-stimulatory manner: activating canonical and non-canonical NF-kB pathways, AP-1, and the AP-1 activator JNK 49 ; generating T cell memory and ameliorating exhaustion. 50,51 CD40 has been identified in several independent screens for enhancing T cell function 37,52 ; and has been synthetically incorporated into CAR T cells, either as a separate module or incorporated into the CAR architecture, displaying superior anti-tumor activity compared with conventional CAR T cells, facilitated by enhanced proliferation and maintaining T cell stemness/memory. 30,53-58 BCMA is expressed in mature B lymphocytes and has been synthetically expressed in CAR T cells, augmenting proliferation 58 ; whereas Fn14, expressed in healthy tissue and particularly in solid tumors such as glioblastoma, 59 has not been previously synthetically expressed in T cells to alter their function. CD27 is a wellknown T cell co-stimulatory protein, enhancing T cell function and generating memory. 37,60-62 CD27 has been incorporated into CARs, displaying equivalent in vivo functionality to 4-1BB and CD28. 63,64 We have extended possibilities of engineering T cells to be resistant to FasL-mediated apoptosis by showing that chimeras of Fas with a  (5), delivering signal 3 to the cell, augmenting CAR T cell activation (6). Upon activation, the CAR T cell induces target cell killing (7).
range of TNFR endodomains can have potentially useful biological functions. Fusion proteins such as Fas-CD40 may enhance anti-tumor activity when co-expressed with CARs.  , T100B), where 3 Â 10 5 activated T cells were seeded and then spun by centrifugation for 1,000 Â g, 40 min at room temperature, and then cultured at 37 C, 5% CO 2 . Transduced cells were identified by measuring for CAR expression using anti-fmc63 and anti-Huk666 idiotypes (both produced in-house). Viral titers were calculated with T cells that were less than 20% transduced.

Flow cytometry and antibodies
Flow cytometry was performed using MACSQuant 10 and X flow cytometers (Miltenyi Biotec). All staining, unless specified otherwise, was performed at room temperature for 10 min, protected from light, with antibodies diluted in either PBS (Sigma, D8537) or cell staining buffer (BioLegend, 420201). Cell viability dyes used were 7-AAD (BioLegend, 420404) or Sytox blue (Thermo Fisher Scientific, S34857). To detect TCF-1 expression, cells were first surface stained for RQR8 as described above and then stained for TCF-1 using the True-Nuclear Transcription Factor Buffer Set (BioLegend, 424401).
Antibodies were from BioLegend, unless otherwise stated. Antibodies

Detection of apoptotic cells
Transduced T cells were treated as described, incubated for 5 h at 37 C, 5% CO 2 , surface stained for CD3 and RQR8, washed once in PBS, washed once in Annexin V binding buffer (BioLegend, 422201), resuspended in Annexin V binding buffer with Annexin V BV421 (BioLegend, 640924), and incubated for 15 min at room temperature protected from the light. Cells were then washed and resuspended in Annexin V binding buffer containing 7-AAD and analyzed by flow cytometry. Apoptotic cells were defined as being Annexin V + 7-AAD À .

Immobilized FasL assays
Recombinant FasL (2 mg) (PeproTech, 310-03H) was immobilized onto 96-well microplates (Starlab, CC7672-7596) overnight at 4 C and the plate was washed several times with PBS. For the proliferation experiments, 5 Â 10 4 CAR T cells was seeded onto the FasL-immobilized microplate and incubated for 5 days. The number of CAR T cells was quantified by flow cytometry, using CountBright Counting Beads. For the measurement of NF-kB activity, 1 Â 10 5 transduced NF-kB Jurkat reporter cells were seeded onto the FasL-immobilized microplate, incubated overnight, and then treated as described.

NF-kB reporter assay
Transduced NF-kB Jurkat reporter cells (1 Â 10 5 ) were cultured with immobilized FasL (20 mg/mL) overnight, at which point cells were analyzed with the Bright-Glo Luciferase Assay System (Promega, E2610) according to the manufacturer's instructions, and then luminescence measured on a Varioskan LUX microplate reader (Thermo Scientific).

Memory and exhaustion phenotyping
For memory phenotyping, CAR T cells were stained for CD62L and CD45RA expression, with CD62L + CD45RA + being naive T cells (TN), CD62L + CD45RA À being central memory T cells (TCM), CD62L À CD45RA À being effector memory T cells (TEM), and CD62L À CD45RA + being effector memory T cells expressing CD45RA (TEMRA). For expression of markers associated with exhaustion, CAR T cells were stained for PD-1, LAG3, and TIM3, using Boolean gating to identify cells expressing one, two, or three of these markers. To get an accurate representation of the cell's phenotype, only cohorts that had at least 2,000 cells acquired in the CAR T gate (CD3 + RQR8 + ) on the flow cytometer were analyzed. Any cohorts below this threshold were excluded from analysis.

Immobilized anti-GD2 CAR restimulation assays
Anti-GD2 CAR ideotype antibody (anti-Huk666) (100 ng) was immobilized onto 96-well microplates overnight at 4 C and the plates were washed several times with PBS prior to seeding with 1 Â 10 5 CAR T cells, which were incubated for 3 or 4 days. CAR T cell numbers were enumerated by flow cytometry using CountBright Counting Beads and were reseeded onto another anti-GD2 CAR immobilized microplate.

In vivo studies
All animal studies were performed under a UK Home Officeapproved project license. Six-to 10-week-old female NSG mice (Charles River Laboratory) were raised under pathogen-free conditions. Nalm6 cells (0.5 Â 10 6 ) engineered to express firefly luciferase and an HA tag were inoculated intravenously into NSG mice 4 days prior to CAR T cell engraftment. Mice were randomized 1 day prior to CAR T cell engraftment, where the following day 3 Â 10 6 CAR T cells were injected intravenously. Tumor engraftment and ongoing tumor growth was measured by bioluminescent imaging using the IVIS Spectrum System (PerkinElmer) after intraperitoneal injection of VivoGlo luciferin (Promega, P1041). Human T cells were transduced at an MOI of 1.5.

Data analysis
Data and statistical analyses were performed on GraphPad Prism 9. Flow cytometry analysis was performed on FlowJo (v.10.8.1). Transcriptomic analysis from the NanoString platform was performed using nSolver 4.0, R, and Python. Quantification of western blot images were performed using ImageJ.

DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.

ACKNOWLEDGMENTS
Graphical abstract created with BioRender.com.