Bispecific Antibody Armed Metabolically Enhanced Headless CAR T Cells

Adoptive T cell therapies for solid tumors is challenging. We generated metabolically enhanced co-activated-T cells by transducing intracellular co-stimulatory (41BB, ICOS or ICOS-27) and CD3ζ T cell receptor signaling domains followed by arming with bispecific antibodies (BiAbs) to produce armed “Headless CAR T cells” (hCART). Various hCART armed with BiAb directed at CD3ϵ and various tumor associated antigens were tested for: 1) specific cytotoxicity against solid tumors targets; 2) repeated and dual sequential cytotoxicity; 3) survival and cytotoxicity under in vitro hypoxic condition; and 4) cytokine secretion. The 41BBζ transduced hCART (hCART41BBζ) armed with HER2 BiAb (HER2 hCART41BBζ) or armed with EGFR BiAb (EGFR hCART41BBζ) killed multiple tumor lines significantly better than control T cells and secreted Th1 cytokines/chemokines upon tumor engagement at effector to target ratio (E:T) of 2:1 or 1:1. HER2 hCART serially killed tumor targets up to 14 days. Sequential targeting of EGFR or HER2 positive tumors with HER2 hCART41BBζ followed by EGFR hCART41BBζ showed significantly increased cytotoxicity compared single antigen targeting and continue to kill under in vitro hypoxic conditions. In summary, metabolically enhanced headless CAR T cells are effective serial killers of tumor targets, secrete cytokines and chemokines, and continue to kill under in vitro hypoxic condition.


INTRODUCTION
Chimeric antigen receptor (CAR) T cells (CAR-T) immunotherapy has been shown to be clinically effective in inducing lasting remissions in hematological malignancies (1)(2)(3), however, early phase clinical trials in solid tumors using CAR-T have met with limited success (4)(5)(6)(7)(8). The major challenges of CAR-T in solid tumors have been the lack of efficacy, T cell exhaustion, off-tumor toxicity, cytokine release syndrome (CRS), and metabolic insufficiency of CAR-T to persist and provide effector functions in the immunosuppressive tumor microenvironment (TME) (9)(10)(11)(12)(13)(14). To address these challenges, we combined the potent intracellular signaling domains of CAR-T with the bispecific antibody (BiAb) arming strategy to redirect the non-MHC restricted cytotoxicity of coactivated T cells. The genetically engineered signaling domains of "Headless CAR-T cells" (hCART) contain the transmembrane, the intracellular domain (ICD) of the co-stimulatory receptors, and the T cell receptor signaling-CD3z domains of a CAR-T except the extracellular scFv CAR domain.
The rationale for combining hCART with the BiAb arming approach is based on in vitro and in vivo clinical evidence that shows: 1) Bispecific antibody Armed T cells (BATs) have shown encouraging clinical results in breast (15,16), prostate (17), and pancreatic cancer (18); 2) adapting the BATs platform strategy for arming hCART permits choice of any BiAb for redirecting specific cytotoxicity; 3) BATs release cytokines/chemokines upon tumor engagement and block the suppressive properties of myeloid-derived suppressor cell (MDSC) and T regulatory cells (T REGs ) in the TME (19,20). Infusions of BATs armed with anti-CD3 x anti-HER2 BiAb (HER2 BATs) in metastatic breast cancer patients were safe without CRS and induced endogenous cellular and humoral immune responses that persisted up to 4 months post infusion (15).
This flexible BiAb arming approach will facilitate the adjustment of arming dose of BiAb and dose and frequency of armed hCART infusions. In addition, BiAb loaded on hCART would be diluted with each cell division and may provide a selfbraking mechanism on the hCART to decrease the risk of CRS seen in CAR-T therapies. This study addresses the following questions: 1) Can the adaptable BiAb armed hCART engineering platform target different tumor antigens? 2) Can BiAb armed hCART target multiple tumor associated antigens? 3) Will BiAb armed hCART remain functionally active under an in vitro hypoxic environment? 4) Can BiAb armed hCART proliferate and mediate serial killing of tumors?
This study shows that hCART transduced with 41BBz (BTC 41BBz ) armed with BiAbs are superior to non-genetically modified T cells armed with BiAbs (BATs) and mediate superior levels of cytotoxicity directed at all tested solid tumor lines as well as kill in a hypoxic environment.

T Cell Activation and Expansion
Activated T cells (ATC) were generated from peripheral blood mononuclear cells (PBMCs) by activation with 20 ng/million cells of anti-CD3 monoclonal antibody (OKT3). ATC were expanded by adding 100 IU/million cells of IL-2 every other day for 14 days in RPMI-1640 supplemented with 10% FBS. Harvested ATC were armed with bispecific antibody anti-CD3 x anti-EGFR [EGFRBi] or anti-CD3 x anti-HER2 [HER2Bi] at a pre-optimized concentration of 50 ng/10 6 ATC. Similar to ATC, co-activated T cells (COATC) were generated from PBMCs by activating T cells with anti-CD3/anti-CD28 beads (Dynabeads, Thermofisher) at the 3:1 bead:cell ratio and transduced with lentivirus (LV) at day 2 (multiplicity of infection, MOI: 5). Beads were removed by magnetic separation at day 6 and cells were expanded till day 14 by adding 100 IU/million cells of IL-2 every other day. Harvested COATC were armed with EGFRBi or HER2Bi at pre optimized concentration of 50 ng/10 6 COATC.
Design of "Headless" CAR Construct and Production of Headless CAR T Cells (hCART) Various combinations of second and third generation selfinactivating lentiviral vector (LV) expressing CAR-less TCR signal-transduction domain CD3 zeta (z) alone or in combination with co-stimulatory molecules were generated as shown in Figure 1A. The intracellular co-stimulatory signaling domains include CD28z, 41BBz, ICOSz, and ICOS-CD27z constructed in tandem were selected to enhance the metabolic function of hCART. Table 1 shows the effector cells, their respective acronyms, type of stimulation, the ICD construct in the lentiviral vector, and the BiAb used to engineer the various effector populations. The effector populations evaluated in this study can be divided into 2 groups: 1) anti-CD3 activated T cells (ATC) or anti-CD3/anti-CD28 co-activated T cells (COATC) armed with HER2Bi or EGFRBi; 2) Headless CAR T cells (hCART) are COATC transduced with different CAR-less constructs [CD28z, 41BBz, ICOSz and ICOS-CD27z] e.g. hCART transduced with 41BBz and armed with HER2Bi are referred as HER2 hCART 41BBz or armed with EGFRBi referred as EGFR hCART 41BBz . The LV construct containing CD8 leader, Tag, CD8 Hinge, transmembrane domain, and ICD is shown in Figure 1A.
The human breast cancer cell lines (MDA-MB-231 [MB231], SK-BR-3 [SKBR3], MCF-7) were obtained from ATCC and were maintained in DMEM culture media (Lonza Inc., Allendale, NJ) supplemented with 10% FBS (Lonza Inc.), 2 mM L-glutamine (Invitrogen, Carlsbad, CA), 50 units/ml penicillin, and 50 µg/ml streptomycin (Invitrogen). These cell lines were used in experiments only up to 10 passages and after 10 passages a new original ATCC stock was used to initiate the cell line culture and cultured up to 3 passages before using them the experiments. Non-ATCC stocks were not used for the study. To target adherent tumor cell lines, target cells were plated in 96well flat-bottom microtiter plates at 4x10 4 cells/well, allowed to adhere overnight at 37°C, and labeled with 51 Cr at 20 µCi/mL in the labeling media (50% fetal bovine serum (FBS) in complete RPMI-1640 medium supplemented with 10% FBS, 2% penicillin- Vector Construct and Bispecific Antibody Loading of Headless CAR T cells. Shows a sequence and basic structure of an intracellular signaling domain of the construct. The intracellular signaling domain includes a fusion protein with a detectable tag, hinge/transmembrane domain, and intracellular domain inserted in lentivirus vector. T cells are co-activated with anti-CD3/anti-CD28 antibody coated beads, transduced with lentivirus vector and expanded for 10-14 days in IL-2. Co-activated T cells (COATC) containing the intracellular signaling domain are defined as "Headless CAR T cells" (hCART) and targeting ability of hCART is facilitated by loading or arming them with bispecific antibodies (BiAb) that are produced either by chemical heteroconjugation of whole IgG molecules with Fc-Fc permanent covalent linker or by recombinant technology. hCART can be armed with single, dual, or triple loading with one, two, or three different BiAbs to target multiple tumor antigens. (B) T cell Subpopulations and Stimulatory/Inhibitory Co-Receptor Expression on hCART. (B) shows the CD4 + /CD8 + T cell ratios of hCART 41BBz , hCART ICOSz , hCART ICOS-CD27z , COATC and ATC after one month in culture (n=3). (C) Shows the percent positive CD4 + and CD8 + T cell subpopulations of hCART 41BBz , hCART ICOSz , hCART ICOS-CD27z , COATC and ATC (n=3). (D) Shows the expression pattern of co-stimulatory receptors on hCART 41BBz , hCART ICOSz , hCART ICOS-CD27z , COATC and ATC after long term culture [1 month; (n=3)]. (E) Shows the expression of co-inhibitory receptors of hCART 41BBz , hCART ICOSz , hCART ICOS-CD27z , COATC and ATC on CD4 + and CD8 + cells derived from the same donors (n=3). streptomycin, and 1% L-glutamine as described (21). Effectors, unarmed or armed hCART, unarmed or armed COATC or unarmed or armed ATC were then added to achieve effector: target (E:T) ratios of 10:1. Co-cultures were incubated for 18 hours (adherent cell lines) and the supernatants were collected for liquid scintillation counting in order to quantitate the amount of 51 Cr released. Percent specific cytotoxicity was calculated as follows: (experimental counts per minute (cpm)spontaneous cpm)/(maximum cpmspontaneous cpm) × 100. Means and standard errors were calculated from four to six replicates per sample.

Real Time Cytotoxicity
In the Real Time Cell Analysis (RTCA) system, cytotoxicity is measured by cellular impedance readout as Cell Index (CI) to monitor real-time changes in cell number. Cell attachment was monitored using the RTCA software until the plateau phase was reached, which was usually after approximately 22-24 hours before adding effector cells as described previously (23). We used breast cancer cell lines (MB231, MCF-7, SKBR3) as targets (n=4, each condition had 3-4 replicates). The target cells (10-20,000 cells/well was optimal for each cell line) were plated in 96-well E-Plates, cells were allowed to adhere overnight or longer until reaching the CI of 1.0, followed by adding effectors (unarmed or armed hCART, unarmed or armed COATC or unarmed or armed ATC) at E:T of 2:1 or 1:1. For sequential killing, tumor cells were incubated with HER2 hCART for 24 hours followed by adding EGFR hCART or vice versa at 2:1 or 1:1 E:T ratios, target cells' impedance signals were continuously monitored for 72-120 hours in 15 minute intervals. Untreated targets or effectors without targets served as controls. To analyze the acquired data, CI values were exported and percentage of lysis was calculated in relation to the tumor cell impedance without effector cells.

Serial Killing Assay
Serial killing assays by armed and unarmed hCART, COATC, and ATC were performed as described previously (23). Briefly, armed and unarmed hCART, COATC and ATC effectors were co-cultured with MCF-7 breast cancer cell line (n=3, each condition had 4-6 replicates) at 2:1 E:T ratio for 72 hours (Initial co-culture). Effector cells from the initial co-cultures were serially transferred to new plates with new target cells for subsequent rounds of killing (2 nd , 3 rd , and 4 th ). With each transfer, effectors were counted and re-suspended in fresh media supplemented with IL-2 to adjust the E:T ratio. Longterm serial killing of MCF-7 by different type of effectors with without HER2Bi arming were sequentially monitored in a RTCA using xCELLigence.

Effect of Hypoxia on the Functional Activities of Headless CAR T Cells
Before determining the effect of hypoxia on the functional activities of effector cells, we determined the effects of hypoxia on the survival of MCF-7 target cells and on effectors using hypoxia-mimetic Cobalt Chloride (CoCl 2 ) by RTCA and flow cytometry, respectively. Tumor cells were plated at 10,000 cells/ well in 96-well plates and incubated overnight at 37°C in 5% CO 2 . Once adherent cells reached a CI of 1.0, cells were treated with 0 µM -400 µM concentrations of CoCl 2 to induce hypoxia tracked by RTCA up to 120 hours and the effect of hypoxia on the T cell viability was performed using the same doses of CoCl 2 by flow cytometry. A non-toxic CoCl 2 100 µM dose was used to determine the effect of hypoxia on the functional activity of hCART (n=3). Cytokines and Chemokines Profile of Target and Effector Co-Cultures COATC, hCART, and BATs (n=2, each pooled from three replicates) either left unarmed or armed with HER2Bi or EGFRBi were co-cultured overnight in the absence or presence of targets followed by collection of the cell-free supernatants to quantitated multipanel (R&D Systems) 45 cytokines, chemokines and growth factors using a Bio-Plex 200 system (BIO-RAD, Hercules, CA). The values are reported in pg/ml of culture supernatants.

Statistical Analysis
All experiments were repeated at least 2-3 times and each condition had at least three replicates. The data are expressed as the means ± SD. Comparisons among groups were performed using ANOVA, and the comparisons within groups were performed using the Bonferroni and the Dunnett's method. Multiple comparison were performed using Two-way ANOVA and Turkey's multiple comparisons test. All statistical analyses were performed using GraphPad Prism version 8.0 software (GraphPad Software, Inc.). P<0.05 was considered as a statistically significant difference.

Memory Phenotyping of hCART in Extended Cultures
After 1 month of culture, phenotyping of 3 sets of hCART generated from 2 donors showed (Figures 1B-E) very low CD4/ CD8 ratios of 0.038 in COATC, 0.037 in hCART ICOSz , and 0.062 in hCART ICOS-27z compared to CD4/CD8 ratios of 0.38 for hCART 41BBz and 0.87 for ATC ( Figure 1B). Further analysis of CD4 + and CD8 + T cells in hCART and COATC showed a higher proportion of effector memory T cells that re-express CD45RA (T EMRA cells) in CD8 + T cells in contrast to ATC that have a dominant population of T effector memory cells (T EM ) ( Figure 1C, right panel). The CD4 + T cells had equal proportions of T EMRA and T EM in all three hCART and COATC while the dominant population in ATC was T EM ( Figure 1C, left panel). Central memory (T CM ) in CD4 + ranged between~1-8% among hCART and COATC with 8% T CM in ATC.
Since the BiAb loaded on the surface of the effectors is diluted by cell divisions, therefore, tumor engagement capacity of the BiAb is concomitantly diluted and should result in an expected decrease in specific cytotoxicity. The unexpected increased cytotoxicity at fourth round of killing by unarmed effector cells may be due to expansion of memory clones as a result of repeated antigenic exposure (in vitro immunization) or to the emergence of innate recognition molecules such as NKG2C capable of T cell receptor (TCR)-dependent and TCR-independent release of cytotoxic granule proteins. Serial killing by hCART directed at SKBR3 cell line showed similar repeat killing pattern (data not shown). Mean cytotoxicity mediated by other unarmed effectors (hCART ICOSz , hCART ICOS-27z , COATC, and ATC) and armed effectors (HER2 hCART ICOSz , HER2 hCART ICOS-27z , HER2 COATC and HER2 BATs) was also measured by RTCA in a long-term serial killing assay directed at MCF-7, data is presented in Figure S1. Specific cytotoxicity reached 80-100% on the first killing cycle within 24 hours at an E:T of 10:1 against MCF-7 ( Figure S1, upper left panel) and 100% within 10 hours in second killing cycle for all armed effector cells ( Figure S1, upper right panel). By the third round, cytotoxicity reached~70 -100% at 72 hours for all armed effector cells but unarmed effector cells showed delayed cytotoxicity (40-80%) at 72 hours ( Figure S1, lower left panel). Intriguingly, by the fourth round of killing, specific cytotoxicity was~80% for both unarmed and HER2 hCART 41BBz , HER2 hCART ICOSz , COATC and ATC by 72 hours ( Figure S1, lower right panel). While armed ATC (HER2 BATs) showed transient killing up to 30 hours and declined thereafter.

Engagement of HER2 hCART or EGFR hCART With Tumor Induces Cytokines and Chemokines
HER2Bi or EGFRBi armed hCART 41BBz , hCART ICOS-27z and BATs induced robust levels of Th 1 cytokines-IFN-g and TNF-a upon tumor engagement at 1:1 E:T ratio compared to unarmed counterparts except for IL-2 that did not show any change in levels between armed hCART 41BBz and BATs compared to unarmed hCART 41BBz and BATs. Background levels produced by effectors without targets were subtracted from the data shown in Figure 2C.
Likewise, high levels of T cell recruiting and activating IFN-g induced IP-10/CXCL10 chemokine was induced by EGFR hCART 41BBz, during overnight co-culture with tumor cells at 1:1 E:T ratio compared other hCART, COATC and BATs. The levels of other T cell recruiting chemokines, MIP-1b and RANTES were higher in the supernatants from HER2-or EGFR hCART 41BBz , hCART ICOSz and hCART ICOS-27z , COATC and BATs co-cultured with tumor targets at 1:1 E:T ratio compared to their corresponding unarmed hCART, COATC and ATC ( Figure 2D).

hCART 41BBz and hCART ICOSz Are Resistant to In Vitro Hypoxic Condition
To determine the effect of hypoxia on the cytotoxicity activity of hCART 41BBz , hCART ICOSz and hCART ICOS-27z , hypoxiamimetic cobalt chloride (CoCl 2 ) was pretitrated on MCF-7 breast cancer cell line and COATC at doses from 50 µM to 400 µM ( Figure S1). Hypoxia induced death of MCF-7 cells ( Figure S1, upper panel) was measured by RTCA and hypoxia induced cytotoxicity of COATC was measured by flow cytometry using Annexin-V and 7-AAD staining ( Figure S1, lower panel). Cell death at 100 µM CoCl 2 was <10% for MCF-7 targets at 24 hours and cell death for COATC at 100 µM CoCl 2 was <10% after 24 hours. Therefore, a dose of 100 µM dose of CoCl 2 was selected for inducing hypoxia to determine whether hCART are functionally active under hypoxic stress. Interestingly, HER2 hCART 41BBz and HER2 hCART ICOSz showed no significant functional difference between normoxia and hypoxia in long term (120 hours) cytotoxicity assays, cytotoxicity ranged between 80-100% against MCF-7 cells. However, cytotoxicity was significantly lower under hypoxic conditions for HER2 hCART ICOS-27 (p<0.005), HER2 COATC (p<0.009) and HER2 BATs (p<0.0001) against MCF-7 cells compared to normoxia. These data suggest that 41BBz and ICOSz intracellular domains are significantly better at resisting in vitro hypoxic stress than COATC and BATs. Representative data is shown in Figure 2E.

Focus on Functional Characterization of hCART 41BBz
Based on phenotyping of T cell receptors and co-receptors especially in CD4 + T cells (T EMTA ); functional screening for the optimal ICD for cytotoxic effector function; the resistance to hypoxia induced cell death; ability to function under hypoxic conditions and provide an effective in vitro immunization effect, we chose hCART 41BBz for the next set of experiments.

High Expression Levels of Transgene in hCART 41BBz
Flow cytometry confirmed that T cells transduced with 41BBz construct show high expression of 41BBz in~75% of CD4 + / CD8 + T cells on days five and eight by detecting FLAG tag ( Figure 3A). Proportion of CD4 + and CD8 + T cell populations of 41BBz transduced T cells (hCART 41BBz ) were comparable to the untransduced COATC and CD19-41BBz CAR-T cells (CD19-BBz). On day five, CD4 + and CD8 + T cells were 63.1% and 27.8% positive for FLAG positive hCART 41BBz , respectively; on day eight, 63.7% CD4 + and 32% CD8 + T cells were FLAG positive , respectively ( Figure 3B).

Resistance to Hypoxia and Exhaustion Is a Function of the 41BBz Co-Stimulatory Endodomain of hCART
We next determined whether the hCART 41BBz can survive in stringent hypoxic conditions using hCART z or BTC 28z as controls. hCART z or BTC 28z were 72% and 61% apoptotic under 0.5% oxygen, respectively, whereas only 13% of the hCART 41BBz became apoptotic under hypoxic conditions ( Figure 3C). These data suggest that survival in hypoxia was differentially affected by the co-stimulatory 28z or 41BBz intracellular domains. Upper panel of Figure 3C shows the representative data from one experiment, bar graph below shows the composite data for sensitivity of hCART containing different constructs for hypoxia induced apoptosis (n=3). Data show significantly less (p<0.0003) apoptosis of hCART 41BBz compared to hCART z or hCART 28z under hypoxic condition.
The differential effects of hypoxia on hCART with 28z and 41BBz are reflective of differential metabolic activity of hCART transduced with 28z or 41BBz. The metabolic reprogramming of hCART expressing either 28z or 41BBz intracellular domains on day 0 or day 7 are shown in Figure S3, demonstrating enhanced basal oxygen consumption rates (OCR) with augmented spare respiratory capacities of hCART 41BBz than hCART 28z on day 7 of culture. The metabolic features of hCART were measured by a sea-horse assay.

Enhanced Cytotoxicity by HER2 hCART 41BBz and EGFR hCART 41BBz in Double Sequential Targeting of MCF-7 Targets
We posited that sequential killing by HER2 hCART 41BBz followed by (f/b) EGFR hCART 41BBz or EGFR hCART 41BBz f/ b HER2 hCART 41BBz would decrease antigen escape and enhance cytotoxicity above that is mediated by EGFR hCART 41BBz or HER2 hCART 41BBz alone. We selected MCF-7 cells as targets since these cells express low levels of HER2 and EGFR to provide proof of concept for targeting of tumors that express low levels of antigen ( Figure 4B, upper panel). The sequential addition of HER2 hCART 41BB z f/b EGFR hCART 41BBz or EGFR hCART 41BBz f/b HER2 hCART 41BBz exhibited significantly higher (p<0.007 and p<0.003) specific cytotoxicity in two normal donors (ND1 and ND2) compared to killing mediated by single antigen targeting at a low E:T of 1:1 (n=2) in the presence of 100 IU/mL IL-2 at 72 hours ( Figure 4B, lower panel).
Similar to MCF-7, significantly higher cytotoxicity was found after sequential targeting of two antigens by HER2 hCART 41BBz f/b EGFR hCART 41BBz (p<0.03) or EGFR f/b HER2, p<0.02) compared to single antigen targeting on SKBR3 cell line. COATC also showed significantly enhanced killing in the sequential targeting of EGFR f/b HER2 compared to single targeting of EGFR antigen (p<0.03); similarly, targeting with HER2 BATs f/b EGFR showed significantly enhanced killing (p<0.004) compared to single targeting of HER2 expressed on SKBR3 cell line ( Figures 4C, D, left and lower panels). Comparable patterns of significantly higher cytotoxicity were seen in MB231 cells after sequential targeting of EGFR hCART 41BBz f/b HER2 hCART 41BBz compared to single antigen targeting of HER2 (p<0.001) or EGFR (p<0.0002). Interestingly, hCART 41BBz mediated sequential or single targeting cytotoxicity was significantly higher compared to COATC (p<0.0005) or BATs (p<0.004) against SKBR3 and MB231 (BATs, p<0.05) ( Figures 4C, D, right and lower panels).

DISCUSSION
The solid TME is a formidable challenge that includes the physical barrier of cancer stroma, an altered metabolic landscape, nutrient insufficiency, regulatory cells and cytokines and chemokines derived from tumors that inhibit adoptively transferred T cells to proliferate, persist and function in the TME (24)(25)(26). Metabolic profiling of T cells from tumors reveals markedly depressed oxidative function and upregulation of coinhibitory molecules such as PD-1, TIM-3, and LAG-3 which correlate with both T cell dysfunction and metabolic insufficiency (27)(28)(29). We designed the constructs to enhance metabolic activity of ex vivo expanded hCART.
HER2 hCART 41BBz or EGFR hCART 41BBz exhibit high levels of specific cytotoxicity directed at multiple tumor targets that express high and low antigen levels. It is reassuring to confirm that sequential targeting of two distinct antigens led to significantly higher cytotoxicity than targeting a single antigen. These data support the strategy of dual targeting to avoid the clonal escape. In long-term cultures involving repetitive exposure Showing that hypoxia differentially affected survival of hCART depending on the co-stimulatory endodomain. Headless CAR T cells with z alone (hCART z ) or CD28z endodomain (hCART 28z ) showed 72% and 61% apoptosis under hypoxia, respectively, whereas hCART 41BBz showed just 13% apoptosis under hypoxic (0.5% oxygen) condition. Lower panel shows significantly less (p<0.0003) apoptosis of hCART 41BBz compared to hCART z or hCART 28z under hypoxic condition. Data is presented as mean ± SD of 3 experiments and analyzed using Two-way ANOVA and Turkey's multiple comparisons test (***p < 0.0003).
to tumor, unarmed hCART, COATC and ATC developed specific cytotoxicity that was equal to or comparable to that exhibited by HER2 hCART 41BBz, HER2 hCART ICOSz , and HER2 BATs after three rounds of exposure to antigen and repeat killing. These results show that repeated exposure to tumor antigens in vitro can lead to primary immunization and expansion of T effector memory clones against tumor associated antigens (TAA) in the culture (30).
The BiAb armed non-MHC restricted killing in hypoxic condition mediated by metabolically enhanced hCART provides a novel and highly adaptable platform for serial or sequential targeting of multiple antigens on tumors and the ability to select different BiAbs targeting multiple TAAs. These features provide the platform for the next generation of more potent targeted effector T cells for immunotherapy in solid tumors. Data is presented as mean ± SD (n=3) and analyzed using Two-way ANOVA and Turkey's multiple comparisons test (****p < 0.00001). (B) Sequential Targeting of MCF-7 with HER2 hCART 41BBz and EGFR hCART 41BBz . Upper panel shows the sequential cytotoxicity against MCF-7 cell line with either HER2 hCART 41BBz followed by (f/b) EGFR hCART 41BBz or EGFR 41BBz (f/b) HER2 hCART 41BBz using RTCA up to 96 hours at E:T of 1:1. The dashed lines denote single antigen targeting by HER2 hCART 41BBz or EGFR hCART 41BBz and solid lines show the double sequential killing mediated by effectors from 2 normal donors (ND1 =Normal Donor 1 and ND2 = Normal Donor 2). Lower panel shows the significant killing at 72 hours with sequential antigen targeting by HER2 hCART 41BBz f/b EGFR hCART 41BBz or EGFR hCART 41BBz f/b HER2 hCART 41BBz of MCF-7 compared to single antigen targeting by HER2 hCART 41BBz (p<0.007) and EGFR hCART 41BBz (p<0.003). Data is presented as mean ± SD (n=3) and analyzed using Two-way ANOVA and Turkey's multiple comparisons test (**p < 0.001*). Right panel shows the sequential targeting schema. (C) Show Sequential Targeting of Two Antigens in SKBR3 and MB231 Cell Lines. The hCART 41BBz induced significantly higher cytotoxicity against SKBR3 cells by sequential targeting of HER2 f/b EGFR (p<0.03) or EGFR f/b HER2 (p<0.02) compared to single antigen targeting. COATC and BATs also showed significantly enhanced killing by sequential targeting of EGFR f/b HER2 (p<0.03), HER2 f/b EGFR (p<0.004) compared single antigen targeting on SKBR3 cell line (Left panels of C and D). Significantly higher cytotoxicity after sequential targeting of EGFR f/b HER2 by hCART 41BBz compared to single antigen targeting of HER2 (p<0.001) or EGFR (p<0.0002) was observed for MB231 cells (Right panels of C, D). Data is presented as mean ± SD (n=3) and analyzed using Two-way ANOVA and Turkey's multiple comparisons test (***p < 0.0001, **p < 0.001, *p < 0.01). While co-stimulatory signaling endodomains play key roles in persistence and effector functions of CAR-T cells (31)(32)(33), the composition and length of the non-signaling, non-targeting flexible hinge can influence the strength of the CAR response (34). CAR-T cells with CD28z domain show rapid effector functions but decreased persistence while CAR-T 41BBz cells show sustained effector functions with increased persistence and enriched T CM (32). Similar to CAR-T 41BBz cells, hCART 41BBz show greater resistance to hypoxia induced apoptosis compared to hCART 28z (13% vs. 61% apoptosis under 0.5% O 2 ). The TME is often hypoxic, O 2 ranges 0.3% to 2.1% O 2 depending on tumor type (35)(36)(37) compared to the normoxic condition of 20% O 2 in in vitro 2D cultures (38). This study shows that the cytotoxicity mediated by 2 out 3 armed hCART (HER2 hCART 41BBz and HER2 hCART ICOSz ) under hypoxia had no effect on their cytotoxicity compared to normoxic condition. However, HER2 hCART ICOS-27z , COATC and ATC all had significantly lower cytotoxicity under hypoxic conditions compared to normoxic conditions. These data suggest that hCART transduced with 41BBz and ICOSz ICD are able to survive and kill tumor cells in vitro under hypoxic condition.
It is not surprising to see induction of Th 1 (IL-2 and IFN-g) cytokines by unarmed (UA) activated T cells (ATC) co-cultured with tumor cells. The fact that UA ATC exhibit non-MHC restricted cytotoxicity against tumor cells indicate that variable levels of IL-2 and IFN-g are likely to be produced for their proliferation and anti-tumor activity. Although the cytotoxicity by UA ATC against tumor cells is lower compared to armed T cells, but may induce comparable levels of IL-2 (IL-2 in the culture supernatants of UA ATC and tumor cells ranges: 47-170 pg/ml). Whereas levels of IFN-g are usually low (IFN-g in the culture supernatants of UA ATC and tumor cells ranges: 0-92 pg/ml) compared to armed T cells, which is consistent with our previous studies showing induction of IFN-g in UA ATC + tumor cells co-cultures (39,40). Likewise studies have shown that unarmed ATC exhibit cytotoxicity directed at hematologic and solid tumors (41)(42)(43). We translated this product in the clinic by infusing UA ATC after stem cell transplant (SCT) in women with metastatic breast cancer (44) and UA ATC infusions after immunization with BATs can transfer antigen specific killer cell activity into patients after myeloablative therapy and autologous SCT (16).
The activation mechanism of "headless CAR" construct is likely through TCR signaling that may transcostimulate intracellular "headless CAR" construct either through intracellular signaling molecules or through external stimuli such as binding of IFN-g to its receptor (positive feedback mechanism), inducing intracellular signaling such as JAK/ STAT pathway. Other studies have also shown this transcostimulation phenomenon delivered by small moleculemediated aggregation of the transmembrane and cytoplasmic MyD88-CD40 domains, which, imparts critical functionality to the various CAR constructs (45). The likelihood of transcostimulation with engagement of the T cells bearing a myriad of co-stimulatory receptors and "forced" engagement by highly crosslinked TCR with the BiAb binding not only with target antigen but also number of ligands on tumor cells to co-stimulatory receptors on T cells providing additional transcostimulation to the "CAR-less" construct.
We deleted the CAR of the construct to minimize the possibility of cytokine release syndrome by eliminating tonic stimulation by tumor antigen and establish an adaptable and flexible targeting platform. The metabolically enhanced hCART could be armed with off-the-shelf BiAb(s) to target one or more antigens on solid tumors sequentially or simultaneously. This exogenous targeting approach avoids the necessity for engineering a different or multiple CAR(s) for each tumor type. This approach has an "autobrake" built-in to the effector cells since the concentration of BiAb loaded onto the surface hCART is fixed and dilutes out with each cell division; after multiple rounds of repeated killing (typically 2-3 weeks),thus avoiding tonic cytokine/chemokine stimulation leading to CRS. This self-limiting governance of antigen-engagement would not only limit toxicity but will also allow for multiple infusions. An advantage of the BiAb loading of hCART is that commercial antibodies (Herceptin ® , Erbitux ® , etc) can be quickly derivatized into BiAbs with OKT3 or existing recombinant BiAbs can be loaded. Thus, BiAb armed hCART offer enhanced survivability and anti-tumor cytotoxicity with metabolic fitness and flexibility for targeting one or more tumor antigens in the TME with built-in CRS regulator.
In a phase I trial in 23 metastatic breast cancer patients and 7 hormone refractory prostate cancer patients (15,17,18), HER2 BATs were safe without dose limiting toxicities in the outpatient setting, and induced a partial remission in a HER2 negative patient for 7.5 months (15). For the entire group, HER2 negative group, and HER2 positive group of breast cancer patients, the median overall survival was 36, 27, and 57 months, respectively. In a phase I/II in a series of seven patients with unresectable or metastatic pancreatic cancer, EGFR BATs infusions induced clinical responses with stable disease for 7.4 months, two complete remissions to chemotherapy given after EGFR BATs, and one complete response survivor at 60 months (18). These encouraging clinical signals in metastatic breast and pancreatic cancer provide a strong rationale for engineering armed hCART for a phase I trial for solid tumors.
In summary, our preclinical studies show that HER2 or EGFR hCART 41BBz can effectively kill multiple tumor targets, kill tumor targets in sequential and serial killing assays, secrete Th 1 cytokines and chemokines and exhibit specific target killing in in vitro hypoxic condition.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

AUTHOR CONTRIBUTIONS
LL, CJ, and AT conceived the idea. AT and LL share lead authorship. AT, LL, CJ, and JS designed the study and performed the statistical analysis. AT, JS, EK, EB, and DS performed the experiments and participated in the data analysis. All authors contributed to the article and approved the submitted version.

FUNDING
This study was primarily supported by funding from in part by R01-CA92344, R01-CA140314, R0-CA182526, and P30-CA022453 (Microscopy, Imaging, and Cytometry Resources Core) and startup funds from the University of Virginia Cancer Center, R01-CA226983 (CJ) from the University of Pennsylvania.