Genetic Ablation of HLA Class I, Class II, and the T-cell Receptor Enables Allogeneic T Cells to Be Used for Adoptive T-cell Therapy

Adoptive cell therapy (ACT) demonstrates objective clinical responses in recent clinical trials (1). Antitumor T cells genetically engineered with a CD19-specific chimeric antigen receptor (CAR) induce impressive therapeutic effects in B-cell malignancies (2–4). ACT using tumor antigen–specific T-cell receptor (TCR)–transduced T cells is tested in multiple clinical trials (5–7). To prepare antitumor T-cell grafts, T cells are often obtained from the patient, genetically engineered, and expanded in vitro before adoptive transfer. The therapeutic effects of ACT are significantly affected by the quality of the antitumor T cells, such as the CD4/8 T-cell ratio and T-cell differentiation status (8–11). In addition, production of antitumor T-cell grafts is sometimes challenging for immunocompromised patients. To overcome these limitations, the use of T-cell sources other than autologous T cells has been under active investigation. Manufacturing of antitumor T cells from third-party donors using a standardized protocol may allow for a more widespread use of this therapy, provided that the allogeneic immune responses are appropriately controlled (12).

Donor-derived allogeneic T cells recognize host-derived normal tissues through their endogenous TCR, which can manifest clinically as graft-versus-host disease (GVHD). Strategies to curtail this issue for off-the-shelf ACT includes depletion of alloreactive T cells after stimulating with recipient antigen-presenting cells (13–16); inducing allo-anergization in donor CAR-T cells by culturing with allo-stimulators and CD80/86 blocking antibodies (17); or using third-party T cells with defined antigen-specificity against viral antigens (18). Conversely, unmatched donor T cells can be recognized and rejected by the host immune system (19). Therapies used for allogeneic transplants, such as alemtuzumab (20) and fingolimod (21), limit the rejection of allogeneic CAR-T cells. However, these strategies would require prolonged suppression of host immunity, which may not be ideal in the context of cancer immunotherapy.

With the advent of efficient gene-editing techniques in primary human T cells (22–24), knockout of endogenous TCR and HLA represents a feasible approach, with the potential to completely and permanently abrogate these allogeneic immune reactions. TCR expression is efficiently abolished by targeting the constant region of the TCR α (TRAC) or β chain (TRBC; refs. 25–29). Abrogation of HLA class I expression by targeting the B2M gene significantly reduces allogeneic responses (28). However, these studies did not address the impact of HLA class II expression on allogeneic responses. Although not expressed in naïve T cells, HLA class II is upregulated in activated T cells by the class II major histocompatibility complex transactivator (CIITA) gene in humans (30). This process is absent in murine T cells, which fail to express CIITA and MHC class II after stimulation (31). Because antitumor T-cell grafts undergo activation during in vitro expansion and upon antigen recognition in vivo, ablation of HLA class II in addition to class I may further reduce allogeneic responses by host T cells.

In this study, we disrupted both HLA class I and class II as well as endogenous TCR expression by genetically ablating B2M, CIITA, and TRAC through the Cas9 system. We demonstrated that the triple-knockout (tKO) T cells persisted better than HLA-sufficient T cells without inducing GVHD in vivo.

Artificial antigen-presenting cells (aAPC/mOKT3) that express a membranous form of a CD3 mAb (clone OKT3), CD80, and CD83 was derived from the human erythroleukemia cell line K562 (32). The A375 melanoma and PG13 retroviral packaging cell lines were obtained from the American Type Culture Collection (ATCC), and the CD19⁺ B-cell leukemia cell line NALM6 was obtained from DSMZ. The 293GPG cell line was a kind gift from Drs. Dranoff and Mulligan (33). 293GPG, A375, and PG13 cells were received in 2008, and NALM6 in 2015. To generate transgenic cell lines by retrovirus, 293GPG packaging cells were transfected with a transfer plasmid using TransIT-293 (Mirus Bio) according to the manufacturer's protocol. Culture media were changed 1 day after transfection. Virus supernatant was collected 2 days after the media change and used for infection. To infect NALM6, 1 × 10⁵ cells were cultured in the presence of 293GPG virus supernatant and centrifuged at 1,000 ×g, 32°C for 1 hour. NALM6 transduced with the EGFP-luciferase fusion gene was purified by flow cytometry sorting using FACSAria (BD Biosciences) to generate NALM6-GL. K562, NALM6, and their derivatives were cultured in RPMI-1640 (Thermo Fisher Scientific) with 10% FBS (Thermo Fisher Scientific). A375, 293GPG, and PG13 were cultured in DMEM (Thermo Fisher Scientific) with 10% FBS. Cells were used for experiments within 2 weeks after thawing and under 30 passages. Authentication of the cell lines was not performed. Mycoplasma testing was not conducted in the past year.

Healthy donor–derived peripheral blood mononuclear cells (PBMC) were isolated by Ficoll-Paque PLUS density gradient centrifugation (GE Healthcare). CD3⁺ T cells were purified through negative magnetic selection by a Pan T Cell Isolation Kit (catalog 130-096-535, Miltenyi Biotec) according to the manufacturer's protocol. The isolated T cells were stimulated with aAPC/mOKT3, irradiated for 200 Gy using a cesium-137 source, at an effector:target (E:T) ratio of 5:1 and cultured in the presence of 100 IU/mL IL2 and 10 ng/mL IL15 (PeproTech). Natural killer (NK) cells were purified from PBMCs by the Human NK Cell Isolation Kit (catalog 130-092-657, Miltenyi Biotec) according to the manufacturer's protocol, and cultured in 500 IU/mL IL2 and 50 ng/mL IL15 for 1 week prior to use. Cytokines were supplemented every 3 days. All primary lymphocytes were cultured in RPMI-1640 with 10% human AB serum (Gemini Bio-Products). Lymphocytes were never cultured for more than 12 days after the start of the experiment. Details of culture time can be found in respective figure legends. All cultures were supplemented with 50 μg/mL gentamicin (Thermo Fisher Scientific).

The antibodies used for flow cytometry are shown in Supplementary Table S1. For surface staining, cells were pelleted and resuspended in a mAb master mix with 7-aminoactinomycin D (7-AAD, BioLegend), diluted in a PBS/2% FBS buffer. Cells were incubated at 4°C for 15 minutes and washed prior to analysis. For intracellular staining, pelleted cells were first washed with PBS, following staining with Aqua fixable viability dye (Thermo Fisher Scientific) according to the manufacturer's protocol. Aqua stained cells were washed by PBS and stained for surface markers as described above. Cells were washed after surface staining, followed by fixation and permeabilization with the BD Cytofix/Cytoperm kit (BD Biosciences) according to the manufacturer's protocol. Permeabilized cells were stained with cytokine-specific antibodies for 30 minutes at room temperature and washed prior to analysis. Stained cells were analyzed with a FACSCantoII or LSRFortessa flow cytometer (BD Biosciences). Data analysis was performed with FlowJo software (Tree Star). All data shown were gated on singlets and live cells as shown in Supplementary Fig. S1A.

Single guide RNAs (sgRNA) targeting B2M, CIITA, and TRAC were generated using the Guide-it sgRNA In Vitro Transcription Kit (Takara Bio) according to the manufacturer's instructions. The target sequences are provided in Supplementary Table S2. Guide RNAs targeting TRAC and B2M have been described previously (28, 34). CIITA-targeting sgRNAs were designed by CHOPCHOP (http://chopchop.cbu.uib.no; refs. 35, 36). To generate a ribonucleoprotein (RNP) complex, 10 μg/sgRNA of the recombinant Streptococcus pyogenes Cas9 protein (Thermo Fisher Scientific) and sgRNAs (5 μg each) were mixed and incubated at 37°C for 10 to 15 minutes. The RNP complex was chilled on ice for no more than 30 minutes prior to electroporation by using the program T-007 (37) on the Nucleofector IIb Device and Amaxa Human T cells Nucleofector Kit (Lonza). The knockout efficiency based on surface marker expression was estimated by dividing positivity of the marker in the electroporated, by that of control T cells, and subtracting from 100%.

Tracking of Indels by Decomposition (TIDE; ref. 38) was used to directly assess the gene-editing efficiency of CIITA sgRNA #6, which provided the highest efficiency. Genomic DNA was extracted from one million electroporated and control T cells using the QIAamp DNA Mini Kit (Qiagen) according to the manufacturer's protocol. The region surrounding the target site was amplified by forward 5′-CACCAGCTGGGAGTTGTTGTAGGTGTCA-3′ and reverse 5′ CCCAGCTCCTTAGCCAAGCTACTCTAGT-3′ primers, then sequenced using the 5′-GTGTCAATTTTCTGCCTCTT-3′ primer. Sanger sequencing results were analyzed using the web tool (https://tide.nki.nl/), with the settings: 80 bp left boundary, 80–600 bp decomposition window, 2–20 bp indel size range, and P value threshold of 0.001 for all samples.

The CD19-specific CAR construct, with FMC63-derived antigen binding domain and the cytoplasmic domains of CD28, truncated IL2 receptor β and CD3ζ tagged with truncated nerve growth factor receptor (ΔNGFR) via a P2A sequence, and cloned into the pMX retrovirus plasmid (39). Stable retrovirus packaging cells were generating by culturing 1 × 10⁵ PG13 cells in the presence of 293GPG-derived virus. PG13-derived retrovirus was then used to infect T cells. CD3⁺ T cells were stimulated with aAPC/mOKT3 and transduced with the CAR gene 2 days after stimulation by using Retronectin (Takara Bio). On the day of infection, Retronectin-coated 24-well non-tissue plates were centrifuged with the PG13 virus supernatant at 2,000 × g, 32°C for 2 hours. The virus was then aspirated, and 1 × 10⁵ stimulated T cells were seeded per well and expanded in 100 IU/mL IL2 and 10 ng/mL IL15 (PeproTech).

HLA and TCR-knockout CAR-T cells were isolated by using flow cytometry or magnetic bead sorting. For flow cytometry sorting by FACSAria (BD Biosciences), CAR-T cells were stained by FITC anti-CD3 (clone UCTH1, BioLegend), PE anti-β₂m (clone 2M2, BioLegend), and Alexa Fluor 647 anti-HLA-DR, DP, DQ (clone Tü39, BioLegend) as described under Flow Cytometry of Materials and Methods. For magnetic bead sorting, cells were labeled with biotin-conjugated anti-CD3 (clone UCHT1, BioLegend), anti-β₂m (clone 2M2, BioLegend), and anti-human HLA-DR (clone L243, BioLegend) 3 to 4 days after electroporation. Five to 10 μg/mL of each antibody was used. Labeled cells were depleted or enriched with Anti-Biotin Microbeads (catalog 130-090-485, Miltenyi Biotec) according to the manufacturer's protocol.

Cytokine production and CD107a upregulation was measured by stimulating 1 × 10⁵ T or NK cells at an E:T ratio of 1:1 for 5 hours. The CD107a mAb was added at the beginning of coculture. Brefeldin A (BioLegend) was added 1-hour post-stimulation at 1,000 × dilution. In NK cell assays, target cells were labeled with 5 μmol/L carboxyfluorescein succinimidyl ester (CFSE, Thermo Fisher Scientific) in PBS at 37°C for 5 minutes, and cultured with NK cells at an E:T ratio of 1:1 for 5 hours. Cytolytic activity of T cells was analyzed by culturing 0.3 × 10⁴ CAR-T cells with 1 × 10⁵ NALM6-GL cells. The frequency of GFP⁺ cells was tracked over 3 days by flow cytometry. Surface CD137 and HLA class II expression was measured by flow cytometry at the 24- and 72-hour time point after coculture, respectively.

Six- to 12-week-old male nod/scid gamma (NSG) mice (The Jackson Laboratory) were used. Experimental details are described in respective figures and legends. Five million NALM6-GL cells were injected intravenously via tail vein for tumor-bearing models. The leukemia burden was analyzed with Xenogen IVIS Spectrum (PerkinElmer). Mice were injected intraperitoneally with 200 μL of 15 mg/mL D-luciferin in PBS (Caymen Chemical), and imaged 20 to 30 minutes post-injection. To evaluate the development of xenogeneic GVHD, T cell–infused mice were monitored for clinical symptoms at least three times a week. The following parameters were scored 0 (absent) or 1 (present) according to the previous reports (40): more than 10% weight loss compared with the weight on day 0, hunched posture, skin lesions, dull fur, and diarrhea. Endpoint was defined by morbidity or 20% weight loss based on institutional guidelines.

The significance of the differences between two groups was assessed with a two-tailed t test. Comparisons between more than two groups were performed by analysis of variance (ANOVA) with Tukey multiple comparisons test. Survival was compared by the log-rank test. Differences were considered significant at a P value of less than 0.05. All statistical analyses were performed using GraphPad Prism 7.

This study was performed in accordance with the Helsinki Declaration and approved by the Research Ethics Board of the University Health Network, Toronto, Canada. Written informed consent was obtained from all of the healthy donors. Peripheral blood was collected from 12 donors by phlebotomy. PBMC was isolated within 24 hours and cryopreserved in liquid nitrogen until use. All animal experiments were approved by the Ontario Cancer Institute/Princess Margaret Cancer Centre Animal Care Committee at the University Health Network and the Animal Care and Use Committee of Aichi Cancer Center Research Institute (Nagoya, Japan).

First, we analyzed HLA class II expression in human primary and cultured T cells. In peripheral T cells, the frequency of HLA class II⁺ cells was 5% to 20% in both CD4⁺ and CD8⁺ T cells (Fig. 1A; Supplementary Fig. S1B). HLA class II expression was substantially induced in cultured T cells compared with primary T cells, especially in the CD8⁺ T-cell population. Human T cells can be classified into different subsets according to their phenotypic markers: naïve/stem-cell-like memory (CD45RA⁺ CD62L⁺ CCR7⁺), central memory (CD45RA⁻ CD62L⁺ CCR7⁺), and effector memory (CD45RA⁻ CD62L⁻ CCR7⁻) T cells (41). The frequency of HLA class II⁺ cells was significantly higher in the effector memory T-cell population than in the other populations (Fig. 1B; Supplementary Fig. S1C). Naïve and stem-cell–like memory T cells progressively differentiate into central memory T cells and then effector memory T cells upon stimulation and proliferation in vitro (41). In contrast to freshly isolated primary T cells, HLA class II⁺ cells were also detected in a subset of cultured T cells with a stem cell–like memory or central memory phenotype (Fig. 1C). HLA class II expression appeared to be inversely correlated with the expression of multiple memory T-cell markers, such as CD27, CD28, and CD127 (Supplementary Fig. S1D). These molecules are preferentially expressed in less differentiated T cells and downregulated during T-cell differentiation (41, 42). These results suggested that HLA class II expression was progressively induced in human T cells in parallel with their expansion and differentiation.

To disrupt HLA class II expression, we targeted CIITA by electroporation of Cas9/sgRNA RNP complexes. We tested multiple sgRNA targets for their efficiency to ablate HLA class II expression. Target sequences were explored within exons 2 and 3, regions essential for the transcriptional activity of CIITA (Supplementary Fig. S2A; ref. 43). Recombinant Cas9 protein mixed with CIITA-targeting sgRNAs was transfected into the A375 melanoma cell line, which constitutively expresses HLA class II (44). Pairs of sgRNAs targeting regions in close proximity were tested to enhance the knockout efficiency. Among the tested combinations, sgRNAs #6 and #8 ablated HLA class II expression most efficiently (Supplementary Fig. S2B). To simplify the gene-editing process, we also tested the efficiencies of using sgRNA #6 or #8 alone. HLA class II ablation with sgRNA #6 alone was similarly efficient as some of the other pairs tested, such as #2 and #5 (Supplementary Fig. S2B). All of the subsequent experiments were performed using CIITA sgRNA #6 alone to knockout class II expression, unless otherwise indicated in the figure caption.

We next studied whether transfection of Cas9/sgRNAs targeting TRAC, B2M, and CIITA could efficiently knockout TCR and HLA class I/class II expression in human primary T cells concurrently. CD3⁺ T cells were stimulated with aAPC/mOKT3, transduced with a CD19-specific CAR (39), and electroporated 2 days later with the RNP complexes. Both TCR and class I expression were reproducibly disrupted in approximately 50% of cells. Reduction in HLA class II expression was more varied (Fig. 2A and B). Triple RNP electroporation generally produced cells negative for surface TCR, β₂m, and HLA class II in approximately 25% of the sample (Fig. 2C). Total cell yield was decreased in electroporated cells compared with control T cells, probably due to electroporation-related toxicity (Supplementary Fig. S2C). Although previous studies show that CIITA negatively regulates Fas ligand (FasL) expression in T cells, which modulates T-cell proliferation (45), we did not detect a significant difference in FasL expression between CIITA-disrupted and control T cells (Supplementary Fig. S2D). There were no consistent differences between the proportion of CD4 and CD8 single positive subsets between electroporated and control CAR-T cells (Fig. 2D).

To directly assess the editing efficiency of the CIITA locus, we amplified the genomic sequence surrounding the target site of sgRNA #6 in control and electroporated T cells. Knockout efficiency was quantified by TIDE analysis (38), and by measuring surface HLA class II expression in the same samples. Quantification by measuring surface HLA class II provided conservative estimations of CIITA editing efficiency, as it consistently produced lower values than the TIDE method (Supplementary Fig. S2E). These data indicated that the targeted CIITA locus was edited in the cells with reduced HLA class II expression, and tKO CAR-T cells could be reliably generated through multiplexed gene knockout.

We interrogated whether T cells ablated of HLA class I/II expression evaded allogeneic lymphocyte recognition in vivo. The bulk electroporated cells, consisting of a mixed population of HLA class I/II positive or negative T cells (target), were transplanted into irradiated NSG mice with or without activated allogeneic PBMCs (responder; Fig. 3A). Responder cells could be distinguished from targets based on CD3 positivity (Supplementary Fig. S3A). When HLA class I and class II expression was analyzed in the peripheral blood CD3⁻CD8⁺ target T-cell population, HLA class I/II double-negative T cells gradually predominated compared with the T cells expressing class I and/or class II (Fig. 3B). The majority of the persisting target T cells in the blood or spleen by the end of the experiment were negative for both HLA class I and class II, only when allogeneic PBMCs were coinfused into the mice (Fig. 3B and C; Supplementary Fig. S3B and S3C). These results suggested that concurrent ablation of both HLA class I and class II was required for efficient in vivo T-cell persistence in the presence of allogeneic T cells.

Next, to investigate persistence of HLA-deficient CAR-T cells in a therapeutic model, we separated HLA class I– and class II–negative cells from HLA-positive T cells. HLA-negative or -positive CAR-T cells were infused with allogeneic PBMCs into NSG mice bearing NALM6 leukemia (Fig. 3D; Supplementary Fig. S4A). In this model, we also observed improved persistence of HLA-negative CAR-T cells compared with HLA-positive T cells in the presence of allogeneic PBMCs (Fig. 3E), which translated to moderately better tumor control 2 weeks after CAR-T infusion (Fig. 3F; Supplementary Fig. S4B).

To evaluate the functional properties of purified tKO CAR-T cells in more detail, we measured their in vitro responses to NALM6 stimulation in comparison with unelectroporated control CAR-T lymphocytes. Purification consistently depleted CD3-, HLA class I–, and class II–positive cells (Fig. 4A and B). There were no substantial differences in the activation status of CD4 or CD8 subset, determined by intracellular cytokine production and CD137 upregulation (Fig. 4C–F; Supplementary Fig. S4D). Except in 1 donor tested, tKO and control CAR-T cells were similarly capable in controlling NALM6 growth in vitro when cocultured at an initial limiting E:T ratio. Meanwhile, NALM6 in the presence of untransduced T cells nearly saturated the culture by the last time point (Fig. 4G). After 3 days of coculture, HLA class II expression was consistently lower in tKO compared with control CAR-T cells, and was significantly lower among the CD8 subset (Fig. 4H). Therefore, CIITA was likely knocked out efficiently among the purified triple-negative cells. By simply testing 3 donors, we were able to generate tKO CAR-T cells that effectively eliminated leukemia in vitro with minimal HLA class II upregulation (Fig. 4I), demonstrating the feasibility in selecting appropriate donors from which to establish universal T-cell lines.

In the in vivo model shown in Fig. 3D, where allogeneic PBMCs were coinfused with CAR-T cells, GVHD developed from engraftment of allogeneic cells (Supplementary Fig. S4C). To assess the ability of engineered lymphocytes to elicit GVHD, we injected leukemia-bearing mice with purified tKO or control CAR-T cells (Fig. 5A). Both control and tKO CAR-T cells efficiently controlled leukemia progression in most of the treated mice (Fig. 5B). Some of the mice receiving control CAR-T cells developed xenogeneic GVHD, marked by clinical symptoms, and T-cell expansion on autopsy analyses (Fig. 5C–E; Supplementary Fig. S5). No animals from the tKO CAR-T cohort developed signs of GVHD (Fig. 5C and D). At autopsy, tKO CAR-T cells were detected at lower levels compared with unedited CAR-T cells in the bone marrow, liver, and spleen, indicating minimal persistence of the infused tKO CAR-T cells after eradicating leukemia cells (Supplementary Fig. S5). We also tested another GVHD model where tumor-free animals were infused with control or tKO CAR-T cells. To promote T-cell engraftment in the absence of antigen stimulation, each mouse was injected with IL15 (46) after CAR-T treatment (Fig. 5F). Similar to the tumor-bearing model, only mice receiving control CAR-T cells experienced severe weight loss (Fig. 5G). After supplementing IL15 and prior to signs of GVHD, control and tKO CAR-T cells were found in peripheral blood at comparably low frequencies based on human CD4 or CD8 positivity (Fig. 5H). When reassessed at disease onset, control T cells were significantly enriched in the periphery compared with tKO cells and control cells at day 6; the frequency of tKO CAR-T cells over this period did not significantly increase (Fig. 5H and I). Overall, these data indicated that tKO CAR-T cells were unlikely to induce GVHD.

Although β₂m deletion protects CAR-T cells from allogeneic T-cell recognition, lack of surface HLA class I should activate NK cells (47). To empirically address this assumption, we cocultured purified NK cells with CFSE-labeled control or tKO CAR-T, and K562 cells as a positive control, at an equal E:T ratio. Activation of NK cells was measured by CD107a upregulation and cytokine production within the CFSE⁻CD56⁺ population. Triple-negative CAR-T cells consistently induced, albeit weak, NK responses in all donors tested, whereas control CAR-T cells did not. K562 robustly activated NK cells under the same conditions, demonstrating tKO CAR-T were relatively poor stimulators (Fig. 6A and B). However, over the course of the 5-hour coculture, NK cells eliminated some of the tKO CAR-T. Control CAR-T cells remained at 50% of the total cells (Fig. 6C). These data demonstrated a potential limitation of HLA knockout CAR-T platforms.

Lastly, we sought to determine the additional benefit of CIITA-knockout and HLA class II depletion to TCR and β₂m deletion, in terms of resisting elimination by allogeneic PBMCs. We generated double-knockout (dKO) CAR-T cells by knocking out and depleting TCR and β₂m expression (Fig. 7A). Among the analyzed markers, dKO CAR-T cells differed from the tKO only by HLA class II expression (Fig. 7A). Both forms of universal CAR-T cells were cocultured with allogeneic PBMC and NALM6-GL. The fold expansion of CD3 and β₂m-negative, CAR-positive cells within the cultures was measured after 4 days. Across independent replicates and different allogeneic PBMCs, tKO CAR-T cells expanded better compared with dKO cells (Fig. 7B; Supplementary Fig. S6A). Presence of allogeneic PBMCs limited the proliferation of dKO more than tKO CAR-T cells by a median of 3-fold; and in certain allogeneic cocultures, tKO expanded more than dKO by greater than 10-fold following tumor-specific stimulation. No advantage for tKO cells was observed if allogenic responders were absent (Fig. 7C), with near identical proliferation across independent experiments (Fig. 7B). NALM6 growth was similarly controlled by dKO and tKO CAR-T cells (Fig. 7D; Supplementary Fig. S6B). Overall, these data highlighted the benefit of HLA class II ablation in generating universal T cells.

Advances in primary human T-cell gene-editing technology have permitted efficient single-step multiplexed gene knockout using Cas9/sgRNA RNP complexes (22–24). Nevertheless, CIITA editing efficiency, and thus HLA class II deletion, was more variable and generally less efficient compared with editing at the TRAC or B2M loci. This could be due to the accessibility of the respective genomic sites, where TCR and β₂m are constitutively expressed and CIITA is repressed at the steady state. Transcriptional activity and chromatin influence the efficiency of Cas9 nuclease (48). Further testing for when to electroporate with Cas9/CIITA sgRNA after CD3 stimulation will be required to improve the yield of universal HLA class II–negative CAR-T cells. This is important for reducing the in vitro expansion required to produce the necessary doses, which should improve the therapeutic quality of the product by retaining less-differentiated T cells (49). Finally, scaled up manufacturing needs to be investigated to fully assess the clinical utility of this system.

A major challenge of HLA class I–deficient universal T cells will be to avoid recipient NK cell killing. Although the frequency at which universal CAR-T cells will encounter NK cells in vivo is difficult to predict, we observed HLA-negative T-cell elimination in as short as 5 hours in vitro. Therefore, it is likely universal CAR-T cells will eventually be rejected by endogenous NK cells after infusion, which may limit their therapeutic efficacy. To overcome this problem in generating universal stem cells, Gornalusse and colleagues knocked-in a covalently linked peptide–HLA-E–β₂m single-chain trimer construct into the endogenous B2M gene. This simultaneously abolished endogenous HLA class I molecules and prevented NK targeting via the inhibitory effect of HLA-E (50). Deuse and colleagues demonstrated that constitutive overexpression of CD47 can also reduce NK response against HLA-deficient universal stem cells (51). Both systems are potentially useful in establishing more optimized universal T cells and should be feasible, since the HLA-E trimer or CD47 can fit within the packaging capacity of retro or lentiviral constructs along with a CAR gene. Additional investigations will be required to determine the optimal mechanism for universal T cells to avoid NK targeting.

In summary, we generated antitumor T cells devoid of both HLA class I/II and TCR molecules through CRISPR/Cas9-mediated tKO of B2M, CIITA, and TRAC. In multiple in vivo and in vitro models, we demonstrated that deleting both HLA class I and II expression in human T cells was required to enhance their resistance against elimination by allogeneic PBMCs. This platform may pave a way to convert allogeneic T-cell grafts derived from third-party donors into off-the-shelf products for adoptive T-cell therapy.

N. Hirano is a consultant for Takara Bio, Otsuka Pharmaceutical, and F. Hoffmann-La Roche; reports receiving a commercial research grant from Takara Bio; and reports receiving speakers bureau honoraria from Genentech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Kagoya, T. Guo, M.O. Butler, N. Hirano

Development of methodology: Y. Kagoya, T. Guo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Kagoya, T. Guo, B. Yeung, M. Anczurowski, K. Murata, K. Sugata, H. Saijo, Y. Matsunaga, Y. Ohashi, M.O. Butler

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Kagoya, T. Guo

Writing, review, and/or revision of the manuscript: Y. Kagoya, T. Guo, C.-H. Wang, M.O. Butler, N. Hirano

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Saso, C.-H. Wang

Study supervision: N. Hirano

This work was supported by the following grants and fellowships: the Princess Margaret Cancer Foundation (M.O. Butler and N. Hirano), Ontario Institute for Cancer Research Clinical Investigator Award IA-039 (N. Hirano), a Takara Bio Sponsored Research Agreement (N. Hirano), a Japan Society for the Promotion of Science Postdoctoral Fellowship for Overseas Researchers (Y. Kagoya), Princess Margaret Postdoctoral Fellowships–Guglietti Fellowship Award (Y. Kagoya), Bennett Family Foundation (T. Guo), and the Province of Ontario (M. Anczurowski).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.