CRISPR-mediated TCR replacement generates superior anticancer transgenic T cells

Adoptive transfer of genetically engineered T cells has become one of the most promising avenues of cancer immunotherapy. Numerous trials have shown objective clinical responses, and even complete remissions, after adoptive cell transfer in patients with cancers resistant to other therapeutic interventions.^1-6  The genetic retargeting of T cells to cancer can be achieved either by transduction with a chimeric antigen receptor (CAR) or a T-cell receptor (TCR) specific for an antigen of choice. Although CAR-based therapy has proven extremely successful in hematological malignancies positive for CD19,⁷  CARs can only target surface-expressed molecules. In contrast, use of cancer-specific TCRs allows targeting of intracellular proteome and/or metabolome.⁸ 

Vertebrate TCRs exist as heterodimers composed of either αβ or γδ TCR chains. Conventional αβ TCRs recognize short antigenic peptides presented by major histocompatibility complex (MHC) I or II molecules (by CD8⁺ and CD4⁺ T cells, respectively). The targets recognized by human γδ T cells tend to be predominantly proteins expressed on the cell surface in the context of a generalized cellular stress, including malignant transformation.⁹  A notable exception to this rule is recognition of pyrophosphate metabolites from the mevalonate pathway (henceforth referred to as phosphoantigens) by the predominant peripheral blood subset of γδ T cells that express TCRs composed of the Vγ9 and Vδ2 chains.¹⁰  Because there is no evidence for MHC restriction of γδ T cells, and their targets are expressed on a broad range of cancers, γδ TCRs offer an exciting potential for pan-population immunotherapy.¹¹ 

The use of a transgenic TCR in primary, patient-autologous T cells is hampered by the presence of preexisting, endogenous TCRs within these cells. Expression of TCRs at the cell surface requires the formation of a ternary complex with the CD3 components of this receptor that constitute a limiting factor for surface expression of the antigen-binding chains of the TCR. As a result, successful expression of transduced TCRs at the cell surface requires that it must successfully compete with the endogenous TCR chains for CD3 association.¹²  In addition, there is also potential for the formation of hybrid TCRs due to mispairing of endogenous and transduced TCR chains (so-called mixed TCR dimers). Thus, a transduced T cell has potential to express 4 distinct TCRs, only 1 of which is desired. Mixed TCR dimers can also exhibit unpredictable, and potentially dangerous, target specificities, and have been shown to cause fatal autoimmunity.¹³ 

Several methodologies have been explored to overcome the issue of TCR competition and mispairing. These approaches include generation of affinity-enhanced TCRs,¹⁴  engineering of mutations to improve the pairing of transgenic TCRs,¹⁵  or overexpression of CD3 components.¹²  Affinity-enhanced TCRs have shown high rates of objective clinical response because even a small number of functional TCR molecules is sufficient to convey antigen-specific signaling due to superphysiological activity.¹⁶  However, affinity-enhanced, engineered TCRs have bypassed the rigors of thymic selection and have the potential to react to self-antigens. Indeed, unanticipated cross-reactivity by an affinity-enhanced MAGE A3-specific TCR with an epitope from titin caused fatal autoreactivity in both patients who were treated with T cells expressing this TCR.^17,18 

Here, we aimed to enhance the functionality of natural TCRs during TCR gene transfer of primary CD8⁺ and CD4⁺ T cells by simultaneous knockout of the endogenous αβ TCR during transfer of a TCR of choice. This approach enhanced the expression of the transduced TCR at the T-cell surface and resulted in TCR transductants that displayed substantially improved antigen sensitivity. In particular, we focused on leveraging broadly cancer-reactive γδ TCRs in the TCR transfer system as this approach can be used irrespective of patient HLA type. T cells transduced with this system were shown to have superior in vitro and ex vivo reactivity to primary hematological malignancies compared with T cells expressing both endogenous and transgenic TCRs.

The following cell lines were purchased from ATCC and cultured according to manufacturer’s recommendations: Jurkat E6.1, Molt-3, KBM7, K562, THP-1, U266, TK6. The primary B-cell acute lymphoblastic leukemia (B-ALL) cells (HP, VB, BV, KÖ, CM, PH) were cultured in defined serum-free media as described previously.^19,20  The B lymphoblastoid cell line (LCL) 146 was generated by Epstein-Barr virus infection of peripheral mononuclear cells²¹  (PBMCs) obtained from a healthy donor. Primary B cells and T cells were isolated from PBMCs based on CD19 or CD4 expression, respectively, and used for functional assays 1 day after isolation. The HLA-A2⁺ melanoma cell line was cultured in RPMI 1640 medium supplemented with 10% fetal calf serum, penicillin/streptomycin, and l-glutamine (all from Gibco, Paisley, United Kingdom). T-cell clone γδ20 was generated by single-cell cloning from PBMCs as described.²²  We also made use of an αβ T-cell clone Mel13, specific for a Melan-A epitope EAAGIGILTV presented in the context of HLA-A2.²³  T-cell clones and lines were expanded in presence of 1 µg/mL phytohemagglutinin (PHA) and allogeneic irradiated feeders from at least 3 donors.²⁴ 

A TCR from clone γδ20 was sequenced in-house using the SMARTer RACE kit (Clontech) and 2-step polymerase chain reaction using universal forward primers and reverse primers specific for constant regions of TCR-γ and TCR-δ. The γδ20 TCR was found to be composed of a Vγ9 and Vδ2 chain. Mel13 is a sister clone of Mel5 and the TCR sequence has been published before.²³  We have also produced a TCR-peptide-HLA A2 cocomplex structure of this TCR with analog²⁵  and natural²⁶  antigens. Codon-optimized, full-length TCR chains, separated by a self-cleaving 2A sequence,²⁷  were synthesized (Genewiz) and cloned into the third-generation lentiviral transfer vector pELNS (kindly provided by James Riley, University of Pennsylvania, Philadelphia, PA). The pELNS vector contains a rat CD2 (rCD2) marker gene separated from the TCR by another self-cleaving 2A sequence. For clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein-9 (Cas9) mediated knockout of both TCR-β constant regions (trbc1 and trbc2; International Immunogenetics Information System [IMGT] Web site), 4 guide RNAs (gRNAs) targeting the first exon of the trbc gene segments were designed using, and cloned into, pLentiCRISPR v2 plasmid²⁸  (Addgene plasmid 52961, kindly provided by Feng Zhang, McGovern Institute for Brain Research, Massachusetts Institute of Technology). pLentiCRISPR v2 plasmid encodes SpCas9 protein and a puromycin-resistance marker gene (puromycin N-acetyltransferase [pac]). The sequence alignments of gRNAs are summarized in supplemental Figure 1 (available on the Blood Web site).

Lentiviral particles were generated by calcium chloride transfection of HEK 293T cells. TCR transfer vectors were cotransfected with packaging and envelope plasmids pMD2.G, pRSV-Rev, and pMDLg/pRRE, whereas CRISPR/Cas9 vectors were cotransfected with packaging and envelope plasmids pMD2.G and psPAX2 (all from Addgene). Lentiviral particles were concentrated by ultracentrifugation prior to transduction of T cells.

PBMCs were obtained from healthy donors obtained via the Welsh Blood Service. Primary T cells were purified by Ficoll separation followed by magnetic enrichment for either CD8⁺ or CD4⁺ T cells (Miltenyi Biotec). T cells were subsequently activated overnight by incubation with CD3/CD28 beads (Dynabeads; Life Technologies) at a 3:1 bead-to-T-cell ratio. After activation, the T cells were transduced with lentiviral particles encoding either a TCR only or both a TCR and CRISPR/Cas9, in the presence of 5 µg/mL polybrene (Santa Cruz Biotechnology). T cells that had taken up the virus were selected by incubation with 2 µg/mL puromycin (Life Technologies) and magnetic enrichment with α-rCD2 phycoerythrin (PE) antibody (clone OX-34, Biolegend) followed by α-PE magnetic beads (Miltenyi Biotec). Fourteen days posttransduction, T cells were expanded with allogeneic feeders.²²  For all functional experiments, transduced T cells were >95% rCD2⁺.

For surface staining, 50 000 cells were stained with Fixable Live/Dead Violet Dye (Life Technologies) and the following antibodies: rCD2 fluorescein isothiocyanate (Biolegend), pan-αβ TCR PE, pan-γδ TCR allophycocyanin (APC), and CD4 PE-Vio770 and CD8 APC-Vio770 (where applicable; all obtained from Miltenyi Biotec). Mel13-transduced cells were also stained with a cognate tetramer (HLA-A2 refolded in-house with the EAAGIGILTV epitope) according to the optimized tetramer-staining protocol.²⁹  For characterization of the differentiation phenotype of the transduced T cells, the following antibodies were used: programmed death 1 (PD-1) PE, CCR7 peridinin chlorophyll protein–Vio770, CD45RA PE-Vio770, CD45RO fluorescein isothiocyanate, and CD27 APC (all obtained from Miltenyi Biotec). All cell lines tested were stained with BTN3 PE antibody (Biolegend), with or without zoledronate pretreatment. For the Jurkat activation assay, cells were incubated with antigen for 16 hours and subsequently stained for CD69. For intracellular cytokine staining, T cells were incubated for 5 hours with target cell lines, and stained for CD107a (BD Biosciences), tumor necrosis factor α (TNFα), and interferon γ (IFNγ), according to the manufacturer’s recommendation (all obtained from Miltenyi Biotec). Cells were simultaneously stained for combinations of surface markers rCD2, CD3, CD4, and CD8 as required. Events were acquired on FACSCanto II (BD Biosciences) and analyzed using FlowJo software (TreeStar). Polyfunctionality plots were generated using SPICE software.³⁰  A minimum of 10 000 viable events were collected per sample.

For the assessment of cytotoxicity, target cells were preincubated with Chromium-51 (Perkin Elmer) and then coincubated with T cells at various effector-to-target (E:T) ratios for 4 hours, as described before.²⁹  Cell lysis was calculated according to the formula: % lysis = (experimental ⁵¹Cr release − spontaneous ⁵¹Cr release)/(experimental ⁵¹Cr release − maximum ⁵¹Cr release) × 100%.

Briefly, 30 000 T cells were coincubated with 90 000 target cells for 16 hours, and the supernatant was harvested. The concentration of macrophage inflammatory protein 1-β (MIP1-β), TNFα, or IFNγ in supernatant was quantified using the respective detection kit (R&D Systems), according to the manufacturer’s instructions. When indicated, target cells were preincubated with 50 µM zoledronic acid (Sigma Aldrich) for 16 hours and washed extensively before coincubation with T cells.

All data were analyzed in GraphPad Prism software, unless specified otherwise.

Lentiviral transduction of primary T cells is greatly enhanced when the cells are actively dividing in response to TCR and costimulatory signals.³¹  To incorporate this enhancement and produce a simple, time-efficient methodology that could be applied with many existing TCR transduction systems, we activated T cells in the presence of 2 separate lentiviral populations, 1 encoding the TCR of choice as a transgene, the other CRISPR/Cas9 targeting the endogenous (but not codon-optimized) TCR-β constant region (trbc1 and trbc2) as described in “Materials and methods.” Four gRNAs targeting TCR-β were designed and showed >90% knockout efficiency in the Jurkat T-cell leukemia line (supplemental Figure 1A). gRNA 1 was selected for use in primary T cells due to the high degree of mismatch between endogenous and the standard, codon-optimized TCR-β sequences generally used during TCR transduction (supplemental Figure 1B). The TCR and CRISPR lentiviruses encoded 2 different selection markers (ectopically expressed rCD2 and puromycin resistance gene, respectively; Figure 1A), allowing selection of cells that had integrated the lentiviral cargo. In addition, the use of rCD2, which was stoichiometrically expressed with the TCR, allowed ready comparison between different donors and different transduction conditions (TCR only or TCR plus CRISPR). Following lentiviral transduction, transduced cells were selected by magnetic or fluorescence-based sorting and culturing with puromycin, where applicable, followed by a conventional T-cell expansion protocol (Figure 1B). Although the selection of transduced cells by rCD2-based purification and puromycin treatment resulted in a nearly 90% decrease in cell number (supplemental Figure 2A), the selected cells were then capable of expanding to the same extent as untransduced cells for at least 5 consecutive expansions with allogeneic feeders and PHA (supplemental Figure 2B). Notably, transduction efficiency with TCR-bearing lentivirus was decreased in the presence of CRISPR lentivirus, indicating that a fraction of cells were capable of accepting only 1 of the lentiviruses (supplemental Figure 2C).

Transduction of primary αβ T cells with a γδ TCR on its own resulted only in a minor downregulation of the endogenous αβ TCR expression. In contrast, αβ TCR expression was almost completely ablated when cells were cotransduced with TCR and CRISPR/Cas9 in all donors tested, showing high efficiency of the TCR replacement system (Figure 1C). We then compared the expression of pyrophosphate metabolite–specific γδ TCR (γδ20), or a melanoma antigen–specific αβ TCR (Mel13), in primary CD8⁺ αβ T cells that were either single (TCR) or double (TCR plus CRISPR) transduced. Although the expression of transgenic TCRs in single transduced T cells was relatively low (as detected by pan-γδ TCR antibody or a cognate Mel13 tetramer, HLA-A2:EAAGIGILTV), cotransduction with the CRISPR vector resulted in dramatically enhanced expression of the transduced TCR in all donors tested (up to 10-fold increase in mean fluorescence intensity, as well as a distinct shift of histogram peak; Figure 2). A high level of expression of the transgenic αβ TCR in double-transduced cells further confirmed that TCR-β targeting gRNA was unable to cleave the codon-optimized receptor, and that the presence of endogenous TCRα chains did not have a detrimental effect on the expression of the transgenic TCR.

It is generally acknowledged that the number of functional TCR molecules on the surface of a T cell is one of the factors governing T-cell sensitivity to an antigen.³²  Therefore, we activated the single- or double-transduced CD8⁺ T cells with their target cell lines (B-LCL line preincubated with zoledronate for γδ20 TCR, and a HLA-A2⁺ melanoma cell line for Mel13 TCR), and measured the percentage of cells expressing a marker of cytotoxicity CD107a, and 2 cytokines IFNγ and TNFα. The response of TCR-plus-CRISPR–transduced T cells was markedly stronger than that of cells transduced with only TCR (Figure 3A; supplemental Figure 3). Up to 90% of TCR-plus-CRISPR–transduced cells expressed at least 1 activation marker in response to target cells, and most of these cells expressed all 3 markers tested for, indicating a strong polyfunctional response to antigen. Conversely, <10% and 40% of cells transduced only with γδ20 and Mel13 TCR, respectively, were capable of mounting a response to the target cells, with only a small fraction of the cells that did respond exhibiting >1 function. In comparison, TCR-transduced cells with CRISPR knockout were capable of mounting a statistically significantly stronger response to their cognate antigen in all donors tested (Figure 3B), and the response of TCR-plus-CRISPR cells was comparable to that of parental clones. Both TCR only and TCR-plus-CRISPR cells were capable of downregulating the transgenic TCR upon stimulation with the cognate antigen (supplemental Figure 4). Importantly, the untransduced and single/double-transduced cells showed similar terminally differentiated effector memory phenotype,³³  plausibly resulting from CD3/CD28 bead expansion, but no signs of T-cell exhaustion, in terms of PD-1 expression³⁴  (supplemental Figure 5).

T cells require a given copy number of antigen to be present on target cells in order to mount a successful response, thus defining the antigen sensitivity. Although antigen sensitivity may be manipulated in the case of αβ T cells by affinity maturation of the TCR¹⁴  so that it can robustly respond to a very limited number of antigen copies,^35,36  no such technology has been developed for γδ TCRs. Therefore, we decided to investigate whether increasing the copy number of γδ TCR on transgenic T cells by CRISPR/Cas9 knockout of endogenous TCRs would increase the sensitivity to the cognate antigen (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate (HMBPP) by TCR-transduced cells. In parallel, we tested the model proxy system for studying the role of TCR in target cell recognition, namely the Jurkat T-cell leukemia line.⁹  We measured T-cell response to the antigen in terms of MIP-1β secretion (for T cells) or CD69 upregulation (Jurkat). In our experience, MIP-1β secretion has been the most sensitive method for detecting T-cell activation.^35-41  Indeed, when we used IFNγ as a readout of T-cell activation, we could not detect any meaningful response from γδ20 TCR-only transduced T cells, whereas the activation of TCR-plus-CRISPR cells closely replicated that of the parental clone (supplemental Figure 6A). Furthermore, the parental T-cell clone was more sensitive to the antigen by 4 orders of magnitude (Figure 4) than the Jurkat cell line (when using MIP-1β and CD69 as markers of activation, respectively). More importantly, TCR-only transduced CD8⁺ cells were only slightly more sensitive than the Jurkat cell line, and well over a thousandfold less sensitive than the T-cell clone. In contrast, TCR-plus-CRISPR–transduced cells showed a similar degree of antigen sensitivity as the parental clone, and were >50 000 or >5000 more sensitive than Jurkat or single-transduced T cells, respectively. The improvement in antigen sensitivity observed with Mel13 TCR-plus-CRISPR–transduced T cells compared with Mel 13 TCR-only transduced T cells was more modest than for γδ20 TCR (∼10-fold greater sensitivity in terms of MIP-1β production; supplemental Figure 6B-C). Importantly, the improvement observed with Mel13 TCR-plus-CRISPR T cells extended to superior cytotoxic activity against HLA-A2⁺ melanoma targets compared with cells transduced with Mel13 TCR only (supplemental Figure 7).

Vγ9Vδ2 TCRs are known to recognize metabolites of the mevalonate pathway in the context of the butyrophilin 3A1 molecule.^42-44  The mevalonate pathway is often dysregulated in cancer cells,⁴⁵  and can be further modulated by aminobisphosphonates such as clinically approved zoledronate.⁴⁶  Vγ9Vδ2 T cells and TCRs thus have the potential to target multiple different cancer types. Therefore, we first tested the cytotoxic activity of γδ20 TCR-transduced cells against an LCL line derived from the same donor as the parental T-cell clone. In line with the polyfunctionality profile described in Figure 3, TCR-plus-CRISPR–transduced CD8⁺ cells were able to exhibit stronger cytotoxic activity, especially at low E:T ratios (26% vs 4% at 0.8 E:T; Figure 5A) to LCL preincubated with zoledronate. No cytotoxicity was observed without zoledronate pretreatment, even in the case of the parental T-cell clone, thus indicating that the endogenous accumulation of mevalonate metabolites in that cell line was not sufficient to trigger T-cell activation.

Because Vγ9Vδ2 T cells do not require CD8/CD4 coreceptors for target recognition, we then investigated the potential of the TCR replacement system to redirect both CD8⁺ and CD4⁺ T-cell subsets to a panel of hematological malignancies. We tested the ability of single- (γδ20 only) and double-transduced (γδ20 plus CRISPR) T cells to undergo activation and cytokine secretion (TNFα and IFNγ) in response to established blood cancer cell lines (T-cell ALL [T-ALL], acute myeloid leukemia [AML], multiple myeloma) as well as primary, patient-derived B-ALL cells (Figure 5B; supplemental Figure 8). The primary B-ALL cells used here have been previously shown to closely replicate the characteristics of the cancer without cell culture–induced bias.^19,20  Single-transduced T cells showed only low reactivity, or no reactivity at all (especially in the case of the Molt3 T-ALL line and primary B-ALL cells), to hematological malignancies pretreated with zoledronate; conversely, TCR-plus-CRISPR–transduced cells responded to all cell lines tested, in a much stronger manner than the TCR-only transduced cells, even to cancer cells expressing an almost undetectable level of BTN3 on the surface (supplemental Figure 9A). No reactivity of γδ20 TCR-transduced cells (with or without TCR-β CRISPR) was observed against freshly isolated, zoledronate-treated healthy cells (Figure 6), despite strong BTN3 expression on the cell surface (supplemental Figure 9B).

TCR gene transfer has been proven as a clinically successful means of redirecting a patient’s immune system to combat different cancer types.⁴⁷  However, the preexistence of endogenous αβ TCRs in the recipient T cells has limited the clinical use to highly competitive/high-affinity αβ TCRs. Here, we demonstrate that cancer-specific αβ or γδ TCRs that do not compete well with recipient TCRs, and therefore exhibit weak functional activity, can be efficiently used to redirect recipient T cells to cancer if combined with simultaneous knockout of endogenous TCR-β. The resultant engineered T cells were as sensitive to antigen as the starting T-cell clone, suggesting that mispairing between endogenous TCR-α chains and transduced TCR-β must be minimal. This finding is in accordance with the results of Provasi et al using zinc finger nucleases where transgenic TCR activity was comparable in T cells deficient for only the endogenous TCR-β and both TCR-α and -β.⁴⁸  Furthermore, because TCR-α and -β chains cannot pair with TCR-γ and -δ chains, disruption of just the TCR-β chain is sufficient to achieve the optimal expression of transgenic γδ TCRs.

To date, there have been several reported attempts to combine endogenous TCR knockout, using zinc finger nucleases,^48,49  transcription activator-like effector nucleases,^50-53  or CRISPR/Cas9,⁵⁴  with redirecting the T cells to cancer, in most cases via CARs. This is the first report demonstrating successful redirection of primary T cells with a pan-cancer–reactive γδ TCR in combination with endogenous TCR-β knockout. We showed that removal of the endogenous TCR-β chain leads to a striking increase of surface expression of transgenic αβ and γδ TCRs that translates into a much stronger response of engineered T cells to cancer lines. Although it has recently been shown by Eyquem et al that CAR insertion into the TCR locus is beneficial due to limiting and controlling CAR expression by physiological means, thus preventing premature exhaustion, the antigen-binding kinetics and affinity of natural TCRs differ significantly from that of CARs, and therefore a high copy number of TCRs on the cell surface appears more desirable.⁵⁵  Indeed, CRISPR-plus-TCR–transduced T cells exhibited a significantly more polyfunctional response profile when presented with target cells than that observed with TCR transduction in the absence of TCR-β knockout, without any apparent changes in terms of T-cell differentiation and exhaustion. Importantly, Ding et al showed that polyfunctional T cells are crucial for achieving a successful clinical outcome in patients suffering from hematological malignancies.⁵⁶  Moreover, this is the first side-by-side comparison of the antigen sensitivity of the model Jurkat T-cell line, primary T cells transduced with a given TCR and the parental T-cell clone using the most sensitive readouts available. Our results indicate that the antigen sensitivity of model systems used in research such as Jurkat cells, or in the clinic (primary T cells) are up to several orders of magnitude lower than that of the parental T-cell clone, and that the sensitivity of the latter can be accurately replicated by combining TCR transfer with endogenous TCR knockout (TCR replacement) in primary T cells. Apart from having implications for designing more effective TCR-based immunotherapies, this result indicates that TCR replacement is preferable to TCR transfer for functional characterizations of TCRs of interest especially where these TCRs compete poorly with endogenous TCRs for surface expression or have a relatively low affinity for cognate antigen. This approach should also enable detailed analysis of TCR recognition in the absence of parental T-cell clones. Such TCRs may come from T cells that display poor growth characteristics (eg, as a result of cancer-mediated T-cell exhaustion⁵⁷ ), or directly from high-throughput sequencing of TCR repertoires.⁵⁸  One can also envisage that primary T cells transduced with a TCR of unknown specificity but not expressing the endogenous TCRs could be used for high-throughput, whole-genome screens²⁸  to identify new TCR ligands, and therefore new potential therapeutic targets.

γδ T cells offer an attractive tool for cancer immunotherapy, due to their ability to recognize ubiquitously expressed targets and no evidence of MHC restriction. This feature allows such γδ T cells to respond to cancer from any individual and also eliminates the risk of graft-versus-host disease.⁵⁹  To date, the majority of clinical trials utilizing γδ T cells have focused on the predominant subset in the periphery, namely Vγ9Vδ2 T cells, which respond to phosphoantigen metabolites. Several multicenter clinical trials^60-63  have demonstrated that in vivo activation of Vγ9Vδ2 T cells and cancer cell sensitization with aminobisphosphonates (zoledronate, pamidronate) was well tolerated and did not result in off-target toxicities (despite the ubiquitous expression of butyrophilin molecules and mevalonate pathway components). Encouragingly, aminobisphosphonate treatment resulted in objective clinical responses in a fraction of patients with non-Hodgkin lymphoma, multiple myeloma,⁶⁰  and AML,⁶²  demonstrating the potential of γδ T-cell–based immunotherapies for hematological malignancies. However, the therapeutic success of γδ T-cell immunotherapies remains underwhelming,¹¹  especially compared with CD19-CAR therapies.⁶⁴  One of the potential reasons for this poor success could be the use of a variable and largely undefined (especially in terms of TCR usage) cellular product. Furthermore, antigen-driven expansion of Vγ9Vδ2 T cells, as used so far, has been shown to lead to exhaustion and loss of functional activity in both animal models and in patients.^65,66  In contrast, TCR replacement by gene transfer, as used here, could be applied to a desirable T-cell subset (for instance, T memory stem cells) thereby potentially allowing improved host engraftment and/or function.⁶⁷  We propose that using a defined γδ TCR transferred to patient’s T cells in combination with the knockout of endogenous αβ TCRs could be a therapeutically beneficial strategy. Indeed, TCR-plus-CRISPR T cells showed a markedly stronger response (in terms of TNFα and IFNγ production) than TCR-only transduced T cells to established cancer cell lines, as well as all primary B-ALL blasts. It should be noted that TNFα production was shown to correlate with cancer-specific activity of cytotoxic T cells, and an elevated intratumoral TNFα concentration could serve as a favorable prognostic factor.⁶⁸  Similarly, IFNγ is a potent immunomodulatory cytokine that enhances T-cell–mediated recognition of cancer cells and plays a direct anticancer role.^69,70  No on-target off-tumor reactivity of γδ20 TCR-transduced cells was observed against primary, aminobisphosphonate-treated B cells, T cells, or whole PBMCs, in line with multiple clinical trials indicating that aminobisphosphonate treatment (with or without infusion of ex vivo–expanded γδ T cells) was generally well tolerated without severe adverse effects (reviewed in Legut et al¹¹ ). It should be noted, however, that even with using the TCR replacement technology for generation of Vγ9Vδ2 TCR-transgenic T cells, the patients will most likely require bisphosphonate treatment for efficient cancer sensitization, in addition to adoptive transfer of TCR-engineered T cells. The increasing clinical experience in coadministering aminobisphosphonates and Vγ9Vδ2 T cells^62,71-74  will undoubtedly facilitate designing of clinical trials testing the efficacy of Vγ9Vδ2 TCR-transgenic T cells. We therefore believe that the TCR replacement technology described here would be of use in fundamental and translational research where it could, for instance, be used to discover ligands of clinically relevant TCRs. In addition, this kind of approach has potential for developing a new generation of TCR-based immunotherapies, provided the method is optimized for the clinical scale, based on the wealth of experience in generating therapeutic CAR T cells. Widespread clinical application of gene-editing technology seems imminent, as demonstrated by the recent success of the off-the-shelf allogeneic CAR19 T cells in inducing remission of B-ALL in infants.⁵³  In summary, it seems likely that TCR replacement by CRISPR/Cas9, or other means, will generate clinically useful T cells that do not encompass the dangers of TCR mispairing and that can be orders of magnitude more sensitive than the products currently being trialed.

The authors thank Catherine Naseriyan from Central Biotechnology Services for the assistance with cell sorting.

M.L. was supported by the Cancer Research UK scholarship. A.K.S. is a Wellcome Trust Senior Investigator (WT100327MA).

Contribution: M.L., G.D., and A.K.S. conceived the study, designed the experiments, and wrote the manuscript; M.L. performed the experiments and analyzed the data; and A.A.M. and O.G.O. provided the reagents and guidance.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Andrew K. Sewell, Cardiff University, Henry Wellcome Building, Heath Park, CF14 4XN Cardiff, United Kingdom; e-mail: sewellak@cardiff.ac.uk.