806 scFvs were swapped with scFv of our standard CD19-BB-ζ lentiviral vector described previously to generate 806-BB-ζ CAR (17, 18). Briefly, the nucleotide coding sequences of 806 or C225 scFv with the huCD8 leader were synthesized by GeneArt (Thermo Fisher Scientific, Waltham, MA) with 5′ Xba1 and 3′Nhe1 and ligated to Xba1 and Nhe1 sites of CD19-BB-ζ car construct. The C225-BB-ζ CAR was obtained from Dr. Avery Posey’s lab at the University of Pennsylvania. The 2173-BB-ζ CAR T construct was obtained from Dr. Laura Johnson’s Lab at the University of Pennsylvania (19, 20).

Human primary total T cells (CD4 and CD8) were isolated from normal healthy donors following leukapheresis by negative selection using RosetteSep kits (STEMCELL Technologies, Vancouver, CA, Canada). All specimens were collected with protocol approved by the University Review Board, and written informed consent was obtained from each donor. T cells were cultured in RPMI 1640 (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS) (VWR, Radnor, PA, USA), 10 mM HEPES (Thermo Fisher Scientific), 100 U/mL penicillin (Thermo Fisher Scientific), and 100 g/ml streptomycin sulfate (Thermo Fisher Scientific) and stimulated with magnetic beads coated with anti-CD3/anti-CD28 (Thermo Fisher Scientific) at the 1:3 T cell-to-bead ratio. Approximately 24 h after activation, T cells were transduced with lentiviral vectors encoding the CAR transgene at an MOI of 3 to 6. On day 5, beads were removed and thereafter cells were counted and fed every 2 days, supplemented with IL 2 150 U/ml until they were either used for functional assays or cryopreserved for future use.

The human cell line U87MG was purchased from the American Type Culture Collection (ATCC) and maintained in MEM (Richter’s modification) (Thermo Fisher Scientific) with components GlutaMAX-1 (Thermo Fisher Scientific), HEPES pyruvate, and penicillin/streptomycin supplemented with 10% FBS. Primary human keratinocytes were purchased from the Dermatology Core Facility at the University of Pennsylvania. K562 cells were purchased from ATCC and maintained in RPMI media (Invitrogen, Carlsbad, CA, USA) supplemented with 10% FBS, 20 mM HEPES, and 1% penicillin/streptomycin. Primary astrocytes were purchased (ScienCell Research Laboratories, Carlsbad, CA, USA) and cultured according to the manufacturer’s instructions. The cells from early passages were used for cytotoxicity and cytokine experiments. GSC cell lines were cultured in DMEM F-12 media (Sigma-Aldrich, St. Louis, MO, USA) supplemented with 2% B27 without vitamin A (Thermo Fisher Scientific), 20 mM HEPES, and penicillin/streptomycin.

To produce the overexpressing EGFR cell line (designated as U87MG-EGFR), the lentivirus co-expressing wtEGFR and Cyan Fluorescent Protein (CFP) under the control of the EF-1α promoter was transduced into the U87MG cell line. On post-transduction day 4, cells were sorted on an Influx cell sorter (BD, Franklin Lakes, NJ, USA) on the basis of high EGFR expression and subsequently expanded. The lentivirus co-expressing CFP and EGFR mutants EGFR^R108K/G, EGFR^A289D/T/V, or EGFRvIII was transduced into U87MG-EGFR, GSC5077 neurosphere cells (21), and K562 cell lines. CFP-positive cells were sorted by fluorescence-activated cell sorting (FACS). For luciferase killing assays and in vivo tracking studies, U87MG and U87MG-EGFR mutant cell lines were transduced with lentivirus click beetle green (CBG) luciferase and green fluorescent protein (GFP). Anti-GFP-positive cells were sorted by FACS.

CAR T cells and K562 targets expressing EGFR and its variants were cocultured in 1:2 ratio in the R10 medium in a 96-well plate, in triplicate. Plates were incubated at 37°C with 5% CO₂. After 48 h, supernatants were collected and cytokine levels were assessed by ELISA kit (R&D Systems, Minneapolis, MN, USA) for IFN-γ, TNF-α, and IL2 production, according to the manufacturer’s instructions.

The cytolytic efficacy of CAR T cells against K562 cells was evaluated by 4-h chromium release assays using E:T ratios of 5:1, 2.5:1, and 1:1. 51Cr-labeled target cells were incubated with CAR T cells in complete medium or 0.1% Triton X-100, to determine spontaneous and maximum 51Cr release respectively, in a V-bottomed 96-well plate. The mean percentage of specific cytolysis of triplicate wells was calculated from the release of ⁵¹Cr using a TopCount NXT (Perkin-Elmer Life and Analytical Sciences, Inc., Waltham, MA) as:

Data was reported as mean ± SD.

CBG+ target cell lines (U87 variants and GSC5077 variants) were cocultured with CAR T cells at E:T ratios of 10:1, 5:1, and 2.5:1, for 24 h at 37°C. One hundred microliters of the mixture was transferred to a 96-well black luminometer plate, 100 μl of 66 μg/ml D-luciferin (GoldBio, St. Louis, MO, USA) was added, and the luminescence was immediately determined. Results were reported as percent killing based on luciferase activity in wells with tumor cells alone.

To assess CD107a degranulation, we plated 1 × 10⁵ T cells and 5 × 10⁵ stimulator target cells per well in round-bottom 96-well plates, to a final volume of 200 μl in complete R10 medium, in triplicates. The CD107a-PE antibody (BD) was added into each well and incubated at 37°C for 4 h, along with surface staining for CD8 (BioLegend, San Diego, CA, USA) and CD3 and then analyzed by flow cytometry.

For CAR detection, cells were stained with biotinylated protein L (GenScript, Piscataway, NJ, USA), goat anti-mouse IgG, and anti-human IgG (Jackson ImmunoResearch Laboratories, West Grove, PA), followed by streptavidin-conjugated allophycocyanin (APC) (BD). The surface expression of EGFR and its mutants was detected by CFP and APC-conjugated cetuximab antibody (Novus Biologicals, Centennial, CO, USA). EGFRvIII expression was detected by anti-EGFRvIII antibody, clone DH8.3 (Santa Cruz Biotechnology, Dallas, TX, USA). Flow analysis done by LSRFortessa (BD) and data were analyzed by FlowJo software (BD).

All mouse experiments were conducted according to Institutional Animal Care and Use Committee (IACUC)–approved protocols. NSG mice were injected with 2.5 × 10⁵ U87MG-EGFR/EGFRvIII/Luc+ tumors subcutaneously in 100 μl of PBS on day 0, seven animals per cohort. Tumor progression was evaluated by luminescence emission on an IVIS Lumina III In Vivo Imaging System (Caliper Life Sciences, Hopkinton, MA, USA) after intraperitoneal D-luciferin injection according to the manufacturer’s instructions (GoldBio). Tumor size was measured by calipers in two dimensions and approximated to volume using the following calculation:

Seven days after tumor implantation, mice were treated with 3 × 10⁶ CAR T cells intravenously via the tail vein, in 100 μl of PBS. Survival was followed over time until predetermined IACUC-approved endpoints were reached.

GBM organoids (GBOs) were established from primary patient tissue, under a University of Pennsylvania Institutional Review Board-approved protocol and with patient written informed consent, and cocultured with CAR T cells as described previously (22, 23). GBOs were fixed and stained after coculture, using anti-CD3 (BioLegend), anti-cleaved caspase 3 (Cell Signaling Technology, Danvers, MA), anti-EGFR (Thermo Fisher Scientific), anti-EGFRvIII (Cell Signaling Technology), and DAPI (Sigma). To control for tumor heterogeneity, four GBOs per condition were used. Mutational data and variant allele fractions (VAF) were obtained from the Center for Personalized Diagnostics at the University of Pennsylvania, as described previously (24).

All in vitro experiments were performed at least in triplicate. GraphPad Prism 6 software (GraphPad Software, San Diego, CA, USA) was used for statistical analyses. Data were presented as mean ± standard deviation. The differences between means were tested by appropriate tests. For the mouse experiments, changes in tumor radiance from baseline at each time point were calculated and compared between groups using the t-test or Wilcoxon rank-sum test, as appropriate. Survival determined from the time of T cell injection was analyzed by the Kaplan–Meier method, and differences in survival between groups were compared by the log-rank Mantel–Cox test.

In the present study, we have generated CARs that target EGFR and EGFR mutants by fusing the scFv derived from mAb806 to a second-generation CAR construct containing 4-1BB-CD3ζ signaling 806 CAR, the design of which is shown schematically in Figure 1A . The EGFRvIII-specific 4-1BB-CD3ζ-based 2173 CAR used in our clinical trials (NCT02209376 and NCT03726515) was generated for comparative evaluation with 806 CAR. 4-1BB-based cetuximab (C225) and CD19 CARs were used as positive and negative controls. Lentiviral vectors encoding CARs were transduced into a mixture of CD4 and CD8 T cells, and surface expression was confirmed by flow cytometry ( Figure 1B ). We next turned to generating target-positive tumor cell lines, expressing the mutations EGFR^R108K/G, EGFR^A289D/T/V, EGFR^G598V, and EGFRvIII, for testing of our CAR constructs ( Figure 1C ). In order to more faithfully model the EGFR mutations, which are almost always co-expressed with amplified wtEGFR, we transduced the GBM cell line U87MG and patient-derived glioma stem cell line GSC5077 (21), both of which express low levels of wtEGFR, with a lentiviral vector encoding wtEGFR ( Figure 1D ) (resultant lines referred to as U87MG-EGFR and GSC5077-EGFR), as well as K562 chronic myelogenous leukemia (CML) cells that lack endogenous expression of EGFR, with wtEGFR ( Figure 1E ). U87MG-EGFR, GSC5077-EGFR, and K562 cells were also transduced with EGFRvIII lentivirus and expression was then analyzed by an EGFRvIII-specific antibody ( Figure 1F ). A lentiviral vector co-expressing CFP and the targeted EGFR extracellular mutants ( Figure 1D ) was transduced into U87MG-EGFR, GSC5077-EGFR, and K562 cells. The resulting CFP-positive cells were sorted by fluorescence-activated cell sorting to obtain a positively transduced cell population ( Figure 1G ).

To determine the specificity of the 806 and 2173 CARs for overexpressed wtEGFR, EGFRvIII, and the EGFR-ECD mutants, 2173 and 806 EGFR BB-ζ CAR T cells were cocultured with U87MG-EGFR and GSC5077-EGFR cell lines expressing EGFRvIII and extracellular mutants EGFR^R108K/G, EGFR^A289D/T/V, and EGFR^G598V, in 24-h bioluminescence-luciferase based killing assays ( Figures 2A, B ). While 2173 CAR T cells demonstrated specificity for EGFRvIII alone, 806 CAR T cells efficiently lysed all targets and exhibited similar cytolytic potential as C225 CAR T ( Figures 2A, B ). Notably, 806 CAR T cells were able to kill U87MG cells, despite expressing only low levels of wtEGFR, at an equal level when compared to overexpressed wtEGFR and EGFRvIII. Since U87MG-EGFR mutants expressed endogenous and ectopic EGFR, we could not distinguish if the 806 scFv-binding specificity was restricted to the mutant or wtEGFR. To test the exclusive specificity to the mutants, we cocultured 806 and 2173 CAR T cells with the CML cell line K562, transduced to express wtEGFR, EGFRvIII, or EGFR mutants, as K562 does not have any endogenous EGFR ( Figure 2B ). 806 CAR T cells did not lyse untransduced K562 cells, confirming the lack of EGFR on the parental line. The 806 CAR T cells selectively targeted K562 cells expressing EGFR, EGFRvIII, or EGFR-ECD mutants and demonstrated similar efficacy as C225 CAR T cells. 2173 CAR T cells lysed K562-EGFRvIII cells but did not show any activity against either wtEGFR or the ECD mutants, as expected ( Figure 2B ). T cell activation was assessed by induction of surface CD107a expression after coculture of CAR T cells with target-expressing cells ( Figure 2C ). Antigen-specific effector cytokine production was assessed by coculturing K562 target cells transduced with EGFR and its variants with CAR T cells. The resulting supernatants were analyzed for IFN-γ, TNF-α, and IL2 production ( Figure 2D ). Untransduced K562 and Nalm6 cells were used as negative controls. 806 and C225 CAR T cells produced similar levels of CD107 degranulation ( Figure 2C ) and IFN-γ, TNF-α, and IL2 ( Figure 2D ) in response to EGFRvIII, EGFR-ECD mutants, and EGFR overexpressing cells, while 2173 CAR T cells responded to EGFRvIII alone.

Having confirmed the function of the 806 CARs, we next sought to compare the reactivity of 806 and 2173 CAR T cells in response to endogenous levels of EGFR in normal cells, in vitro. We cultured primary human keratinocytes and astrocytes, as those cell types express wtEGFR ( Figure 3A ) and used them to stimulate CAR T cells. We observed production of IFN-γ by C225 CAR T cells, in response to EGFR presented by either astrocytes or keratinocytes, as well as U87MG-EGFR ( Figure 3B ). In contrast, 2173 CAR T cells produced IFN-γ in response to EGFRvIII antigen alone. 806 CAR T cells exhibited low or no cytotoxicity when cocultured with astrocytes or keratinocytes ( Figure 3C ), with corresponding low IFN-γ production ( Figure 3B ).

Having compared the antigen-specific effector function of 806 CAR with 2173 CARs, we next sought to confirm its in vivo antitumor effects, using immunodeficient NSG mice bearing human GBM tumors ( Figure 4A ). On Day 0, U87MG-EGFR/EGFRvIII tumors were implanted subcutaneously, and on Day 5, tumor engraftment was confirmed by bioluminescence imaging (BLI). On Day 7, a single dose of 3 × 10⁶ CAR-positive T cells were infused intravenously (n = 7 per cohort). Total bioluminescence ( Figure 4B ) and individual bioluminescence ( Figure 4C ) were assessed in 806 and 2173 CAR T cell-treated groups. Animals in the negative control cohort, receiving CD19 CAR T cells, demonstrated rapid tumor growth, with all mice reaching a predetermined humane experimental endpoint by 42 days after initial tumor engraftment. To be noted, all mice in the CD19, 2173, and 806 groups reached experimental endpoint by day 42, 63, and 91, respectively ( Figure 4A ).

Given the ability of the 806 CAR to target EGFR alterations beyond EGFRvIII, we turned to patient-derived GBM organoids (GBOs) to demonstrate activity in a heterogeneous model previously characterized to be of high fidelity to human tumors (22, 23). GBOs retain the originating tumor heterogeneity to a high degree out beyond 12 weeks of culturing and maintain the expression of endogenous EGFR and its alterations, providing a valuable model platform for testing therapies aimed at addressing tumor escape. The GBOs selected for coculture experiments contained multiple EGFR mutations ( Figure 5A ). GBO 9057 had EGFR copy number gain, EGFRvIII, and two missense mutations, EGFR^G598V and EGFR^C595Y. The missense mutation was found to have a VAF of 24%, while EGFRvIII was identified in less than 10% of the reads, based on next-generation sequencing (NGS). GBO 9066 had EGFR copy number gain, EGFR^A289V, and EGFR^G598V. Both EGFR^A289V and EGFR^G598V had a VAF of less than 15%, making determination of co-occurrence impossible through NGS.

GBOs were cocultured with 806, 2173, and CD19 CAR T cells, at a 1:10 E:T ratio, for 72 h before fixation and evaluation. CAR T cell infiltration, as quantified by CD3 staining, was more significant in the 806 CAR T cell population than either the 2173 or CD19 CAR T cell population ( Figure 5B ). Cleaved caspase 3 (CC3) was used as a measure of cell death and antitumor activity. As with the CD3⁺ cell infiltration, the 806 CAR coculture resulted in higher CC3 levels than either the 2173 or CD19 CAR cocultures ( Figure 5C ). These results highlighted the broad cross-reactivity of the 806 CAR in a heterogeneous, high-fidelity GBM model. wtEGFR staining in both GBO lines provided additional evidence of the cross-reactive nature of the 806 CAR ( Figure 5D ). Staining intensity, normalized to CD19 CAR-treated GBOs, showed consistent decreases in 806 CAR-treated GBOs, to a greater degree than the 2173 CAR-treated GBOs ( Figure 5E ).