In the current study, we employed the EGFRvIII CAR T-cell construct based on a second-generation backbone containing 4-1BB and CD3ζ intracellular signaling domains, but this time cloned into an AAV6 vector backbone instead of a lentiviral vector (Fig. 1a), the former allowing for integration of the CAR sequence into a specific locus rather than relying on random genomic integration. Briefly, the strategy for multiplexed gene-editing consists of in vitro stimulation of primary human T cells, followed by electroporation with respective Cas9 ribonucleoproteins (RNPs) and subsequent adeno-associated virus (AAV)-mediated transduction of the CAR (Fig. 1b). CRISPR-Cas9 gene-editing using RNP electroporation for TRAC and B2M genetic loci was efficient, yielding populations of greater than 80% double knock-out surface expression by flow cytometry (Fig. 1c,d). In a separate experimental group, RNP electroporation was multiplexed to generate T cells also edited for PDCD1, in addition to TRAC and B2M. This was followed by AAV6 transduction, which resulted in CAR T cells with either endogenous or deleted PD-1 (i.e., CART-EGFRvIII and CART-EGFRvIIIΔPD-1) (Fig. 1e). Following stimulation with EGFRvIII-expressing glioma, we demonstrated that both control (i.e., T cells edited for TRAC and B2M, without CAR) and CART-EGFRvIII cells (i.e., T cells edited for TRAC and B2M, with CAR) were positive for surface PD-1 by flow cytometry. By contrast, PD-1 was not detected on the surface of CART-EGFRvIIIΔPD-1 cells, confirming effective knock-out at the level of surface protein expression in the entire population (Fig. 1f).Fig1Fig. 1

Multiplexed CRISPR-Cas9 gene-editing is efficient in primary human T cells. a Schematic representation of the EGFRvIII targeted CAR construct. b Primary human T cells were stimulated, RNP electroporated and transduced to produce CAR T cells. c Following expansion, cells were subjected to flow cytometry for TCR and B2M expression. d Bivariate plot displays frequency of cells with both TCR and B2M deletion. e EGFRvIII CAR T cells that have been gene-edited for PD-1 (CART-EGFRvIIIΔPD-1) do not have the ability to interact with PD-L1 expressed on target cells. f Effector cells were incubated with irradiated U87vIII for 1 week and subjected to flow cytometric analysis for surface PD-1 expression. The control group contains cells gene-edited for both TCR and B2M, and mock transduced with AAV

We next sought to assess the levels of PD-L1 expression on commonly used brain tumor cell lines. Importantly, PD-L1 has been shown to be frequently found on the surface of GBMs [9] and upregulated in patients treated with EGFRvIII CAR T cells [6]. To demonstrate proof-of-concept, we selected a well-characterized EGFRvIII-positive glioma line, U87vIII, as a canonical target cell for our study. Compared to its parental line, U87, and another commonly used glioma cell line, U251, we demonstrated that U87vIII naturally expresses PD-L1; however, this appeared to be decreased relative to U87 and U251 by flow cytometric analysis (Additional file 1: Figure S1).

We then proceeded to assess the impact of CRISPR-Cas9 gene-editing of the PDCD1 locus in CAR T cells specific for EGFRvIII. CAR T cells are known to exist in various states of differentiation, with less differentiated stem cell memory (T_SCM) or central memory (T_CM) subtypes preferred over well-differentiated effector memory cells (T_EM), specifically regarding characteristics such as expansion, persistence, and the capacity for self-renewal [10]. Moreover, loss of PD-1 has been shown to alter memory T-cell content and generation in other settings [11]. At baseline, both CART-EGFRvIII and CART-EGFRvIIIΔPD-1 demonstrated similar T-cell differentiation patterns compared to control T cells that had also been gene-edited for TRAC and B2M, in addition to undergoing mock transduction with AAV6 (Fig. 2, left column). By contrast, prolonged stimulation of CART-EGFRvIIIΔPD-1 led to a selective enrichment of T_CM, while CART-EGFRvIII cells expressing native PD-1 appeared to enrich for the more differentiated T_EM compartment (Fig. 2, right column).Fig2Fig. 2

PD-1 disruption promotes favorable differentiation of CAR T cells targeting PD-L1 expressing glioma. Effector cells were cocultured with irradiated target U87vIII at and E:T of 1:1. The phenotype of T cells were assessed at Day 1 (prior to stimulation) and at Day 21 by flow cytometry. Cells were grouped by flow cytometry according to T-cell phenotype as follows: naïve (T_N) CCR7⁺CD45RO⁻, central memory (T_CM) CCR7⁺CD45RO⁺, effector memory (T_EM) CCR7⁻CD45RO⁺, and effector (T_E) CCR7⁻CD45RO⁻

Next, we turned our attention to the functional capacity of gene-edited CAR T cells in mediating antitumor immune responses in vitro. In experiments using primary human T cells, CART-EGFRvIIIΔPD-1 cells were found to produce significantly greater amounts of Th1 proinflammatory cytokines (e.g., IFN-γ and TNF-α) when cultured with EGFRvIII-expressing glioma compared to CAR T cells expressing endogenous PD-1 (Fig. 3a). We also compared each construct for the ability to initiate and maintain T-cell proliferation. Following serial stimulation with irradiated target cells, repeated antigen stimulation through EGFRvIII maintained proliferation of both CART-EGFRvIII cells and CART-EGFRvIIIΔPD-1 cells for over 1 month (Fig. 3b). Impedance-based, microelectronic platforms were then used to capture real-time kinetics of antitumor cytotoxicity as measured by target cell index (e.g., viability). Using this system, we found that CART-EGFRvIIIΔPD-1 cells were significantly more efficacious against U87vIII than those expressing PD-1, but that this difference was observed only after an extended period of time (Fig. 3c).Fig3Fig. 3

PD-1 disruption enhances EGFRvIII CAR T cells. a Cytokine production by CAR-transduced primary human T cells when cocultured for 18 h at an E:T of 1:1. b Proliferation assessment of effector cells stimulated weekly with irradiated U87vIII. c Impedance-based cytotoxicity assay measuring activity of effector cells against U87vIII at an E:T of 1:3, with cell index serving as an inverse measure of target cell viability. Assays were performed in triplicate (mean ± SEM is depicted; unpaired, two-tailed t-test, * = P < 0.05, *** = p < 0.001)

Based on our observations in vitro, we proceeded to evaluate the function of CART-EGFRvIIIΔPD-1 in animal models of human glioma. First, we implanted tumors with stereotactic assistance into the brains of NSG (NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ) mice. This was followed by intravenous infusion of control, CART-EGFRvIII or CART-EGFRvIIIΔPD-1 cells via tail vein. Results did not demonstrate significantly prolonged survival in mice treated with EGFRvIII-specific CAR T cells systemically compared to the control (Fig. 4-c).Fig4Fig. 4

Intravenous delivery of CAR T cells does not significantly prolong survival in mice. a U87vIII cells (5 × 10³) were implanted orthotopically into NSG mice and treated post-implantation with intravenous (IV) effector cells. b Antitumor responses produced by CART-EGFRvIIIΔPD-1 in vivo. Survival curves were estimated for each group using Kaplan–Meier product-limit estimation. Primary comparative analyses of the curves for each group were performed using the log-rank test. c Bioluminescence imaging of U87vIII tumor growth over time, n = 5 mice

In this study, we sought to apply CRISPR-Cas9 technology to EGFRvIII CAR T cells in order to address extant barriers to achieving maximal therapeutic efficacy for patients with GBM. Specifically, we created EGFRvIII-specific CAR T cells with targeted deletion of PD-1 in order to make them resistant to immune checkpoint signaling through this pathway. In addition, we used this approach to simultaneously disrupt loci corresponding to genes for both endogenous T-cell receptor (TCR) and beta-2 microglobulin (B2M). We used several preclinical modeling systems to test our hypotheses, including in vitro and in vivo platforms. These consisted of phenotypic and functional assays. Direct antitumor activity was tested against human glioma cell line targets transduced to express EGFRvIII. In this manuscript, T cells from a single healthy donor batch preparation were used, as would be used in a trial setting. Cells were not purified after genetic manipulation. The CAR T cells used in vitro were isolated from the same T-cell expansion as those used in vivo. Experiments were performed multiple times with representative data shown.

Immune compromised NSG mice were originally purchased from Jackson Laboratory and bred under pathogen-free conditions, according to protocols approved by the Institutional Animal Care and Use Committee. Human glioma cell lines U87 and U251 were obtained from American Type Culture Collection (ATCC) and cultured under conditions as outlined by the supplier. The U87vIII cell line was generated by lentiviral transduction.

CAR T cell constructs were synthesized and cloned into an AAV6 plasmid backbone. All constructs included a CD8 transmembrane domain in tandem with an intracellular 4-1BB costimulatory and CD3ζ signaling domain. Gene-editing and cell preparation was performed using standard techniques as described in detail elsewhere [29]. Briefly, human peripheral blood mononuclear cells (PBMCs) were thawed and the T cells were activated with conjugated CD3/CD28 agonists for 3 days in T-cell media containing human serum, IL-2 and IL-7. After activation, the T cells were electroporated with Cas9 protein and sgRNAs targeting the TRAC and B2M loci or TRAC, B2M, and PDCD1 loci and subsequently transduced with a recombinant AAV6 vector containing donor template DNA for insertion of the EGFRvIII CAR construct, with a typical transduction efficiency of 35%. Following electroporation and transduction, the CAR T cells were expanded for 7 days in T-cell media containing human serum, IL-2, and IL-7. These cells were subsequently transferred to storage in liquid nitrogen prior to assays.

T-cell assays for activity, proliferation and cytotoxicity have been described in detail elsewhere [30]. Briefly, in coculture experiments, T cells were incubated with irradiated U87vIII target cells at an E:T of 1:1 for time periods as described. Cell-free supernatants from cells were also analyzed for cytokine expression using a Luminex array (Luminex Corp, FLEXMAP 3D) according to manufacturers instructions. Expression of surface markers were either taken at baseline or after a period of coculture, and then subjected to flow cytometric analysis. Antigens were stained for using the following antibody clones for flow cytometry where indicated: CCR7 (3E12, BD Bioscience); CD45RO (UCHL1, BD Biosciences), PD-1 (EH12287, Biolegend). For proliferation assays, cells were stimulated with irradiated target cells at an E:T of 1:1. Cells were counted every 7 days and plated again with stimulation at 7 day intervals. In experiments when real-time cytotoxicity was measured against U87vIII, cell index was recorded as a measure of cell impedance using the xCELLigence RTCA SP instrument (ACEA Biosciences, Inc.) according to manufacturer instructions. Percent specific lysis may be calculated from these data using the following equation: % = ((cell index of UTDs - cell index of CAR T cells) / cell index of UTDs) × 100.

Tumor cells were harvested in logarithmic growth phase, counted, and loaded in a 50 μL syringe with an attached 25-gauge needle. Mice were anesthetized and placed in a stereotactic frame to assist in tumor implantation. Tumor cells were implanted at 2 mm to the right of the bregma at a depth of 4 mm from the surface of the skull, in a total volume of 5 μL. Effector cells were then infused systemically by tail vein infusion in a total volume of 100 μL or administered intraventricularly in a total volume of 30 μL. Intraventricular delivery was at 2 mm to the left of and 0.3 mm anterior to the bregma at a depth of 3 mm from the surface of the skull. Effector cell populations were normalized to contain 1 × 10⁶ cells per infusion for all experiments. Tumor progression was assessed over time by bioluminescence emission using the Ami HT optical imaging system (Spectral Instruments) following intraperitoneal luciferin injection. Survival was determined by mice found expired or otherwise sacrificed by a blinded technician at predetermined humane endpoints.

All analyses were performed with GraphPad Prism 7.0c software. Data was presented as means ± SEM with statistically significant differences determined by tests as indicated in figure legends.