As for CAR-T cell based therapy, viral vectors were widely exploited for ex vivo TCR gene transfer into recipient T cells ( Figure 2 ). In clinical trials, the most common viral systems used are gamma retrovirus- (RV) and lentivirus-based vectors (LV), as reviewed in (31). Both RVs and LVs allow for stable integration and efficient long-term expression of exogenous TCRs. However, safety concerns to RVs and LVs remained, that are mainly insertional mutagenesis and neoplastic transformation (32), as well as generation of replication-competent viral particles. The latter limitations boosted the development of novel vector designs, such as addition of insulator sequences (33), disruption of the long terminal repeats for self-inactivating viral vectors (34, 35) or pseudotyping (36). Of note, RV transduction requires mitotic cells for the transgene to penetrate the nucleus and integrate in the genome and thus recipient cells must be activated beforehand. Conversely, LVs allow the effective transduction of a variety of not actively dividing and terminally differentiated cells. Yet, human resting T cells are scarcely susceptible to transduction by LVs and need to be minimally stimulated to enter the G1b phase (37, 38). Importantly. several studies highlighted a positive correlation between the proliferative potential of adoptively transferred cells and their in vivo persistency (1), sustaining the use of LVs in the clinic to limit recipient cell stimulation in vitro.

As alternative to RVs and LVs, adeno-associated virus-derived vectors (AAVs) can be used for TCR-engineering. AAVs are replication-deficient systems and are not known to induce any side effects, making them ideal candidates from a safety standpoint. AAV-based cancer treatments have not yet been used in the clinic, however multiple strategies have been developed and hold great promise for the future (39–41). It is to be noted that vector manufacturing remains time consuming (over six months) and expensive (42).

Non-viral methods for stable gene transfer into T cells were investigated in CAR-T clinical trials and represent a future opportunity for TCR gene transfer clinical studies. Non-viral engineering strategies include transposon-based vector systems, such as “Sleeping Beauty”, by which a transgene of interest flanked by inverted terminal repeats is provided to target cells as discrete DNA molecule and randomly integrates in its genome when a transposase is co-supplemented in trans (i.e. mRNA, plasmid or protein) (43). Over the past decades, transposons and transposases have been extensively optimized to increase their activity and reduce toxicity (e.g. insertional mutagenesis) (44). Alternatively, exogenous TCRs can be transiently expressed in trans if transfected into T cells in the form of mRNA molecules or non-integrating vectors (45, 46). All these non-viral systems may be preferable to viral ones for the clinic because of their easier handling and cheaper production costs. Of important note, gene editing of T cells with TCRs, as opposed to CARs, has to face an extra challenge. Mispairing of exogenous and endogenous TCR chains can indeed occur and lead to off-target toxicity. Researchers have therefore developed several platforms to specifically silence the expression of the endogenous TCR, based on Sleeping Beauty, Zinc finger nucleases (47), transcription activator-like effector nucleases (TALEN) (48, 49), mega-nucleases or Clustered Regularly-Interspaced Short Palindromic Repeats (CRISPR)–Cas9 (50, 51) ( Figure 2 ).

The CRISPR-Cas9 technology relies on short RNA sequences which are used to target the site of insertion instead of proteins and which are easily synthesized in vitro. As such, CRISPR-Cas9 enables the simultaneous targeting of multiple genome sites and site-specific mutagenesis, allowing the knockout (KO) of endogenous TCRs or of immune checkpoint such as PD-1 (52) and the induction of cytokine expression (e.g. IL-7, IL-12, IL-15, IL-18) (17) ( Figure 2 ). The safety and feasibility of multiplex CRISPR-Cas9 gene editing of T cells was demonstrated in a study with advanced refractory cancer patients infused with autologous T cells KO for the three genes TRAC, TRBC and PDC1 (PD-1 loci) and transduced with a NY-ESO1 TCR (53). Of interest, the KO by CRISPR-Cas9 of cytokine-induced SH2 protein (CISH), an immune checkpoint, was found to increase the in vitro proliferation and functionality of TCR-engineered T cells and such strategy will be further explored in an upcoming clinical trial (NCT04426669) (54). Despite technical challenges which still need to be overcome, in particular regarding the delivery systems currently based on AAV and transfection (55–59), CRISPR-Cas9 will likely become the method of choice for therapeutic gene engineering in the upcoming years.

T lymphocytes are the most common source of cells used as carrier for gene transfer therapy. Following TCR engineering into recipient cells, TCR-T cells need to be expanded to reach sufficient numbers for ACT. The primary starting material is most often autologous peripheral blood ( Figure 2 ).

Thus far, TCR gene therapy mainly focused on CD8 T cells, which represent key players of ACT. In particular, a retrospective analysis of TIL ACT infusion products and clinical efficacy from 92 patients highlighted an association between the fraction of CD8 T cells and clinical benefit (60). There is, however, an emerging clinical relevance of CD4 T cells. Early evidence was provided by Tran and colleagues who showed the antitumor potential of neoantigen-specific CD4 TILs by ACT (61). Of interest, the direct cytolytic capacity of CD4 T cells was demonstrated (62–64). In addition, CD4 T cell’s help is essential to generate efficient tumor-reactive effector CD8 T cells (65), notably during the process of epitope spreading (29, 66). Yet TCR gene transfer therapy with CD4 tumor-specific TCRs is also limited by the challenging prediction and detection of MHC class II-restricted neoantigen-specific CD4 T cells, despite major advances (67, 68).

T cell differentiation state is critical for ACT. The profile of adoptively-transferred cells is indeed likely to affect their in vivo persistency and thus treatment efficacy. T cell differentiation states range from naïve to central memory, effector memory and finally terminally differentiated (EMRA) (69). Accumulating evidence shows that the effector phenotype acquired in vitro negatively impacts the antitumor potential of T cells in vivo (70), while ACT efficacy requires a long-term persistence of transferred cells. Thus, less-differentiated T cell populations that maintain self-renewal capabilities are preferred and were associated with improved clinical benefit (71, 72). Naïve and memory T cells maintain the highest proliferative potential combined to the most potent fitness and stemness (73–76) ( Figure 2 ). Consistently, naïve and memory subsets were found more effective than effector T cells for ACT (70, 74–77). Furthermore, Hinrich and colleagues have demonstrated that naïve CD8 T cells had a higher anti-tumor potential for ACT as compared to central memory cells (77).

The T cell state is also modulated during the expansion phase in vitro. TCR-transduced T cells are commonly expanded with anti-CD3/CD28 beads in the presence of IL-2 (12, 38, 78). It has been demonstrated that the addition to the culture medium of alternative cytokines, such as IL-7, IL-15 and/or IL-21, enabled to the generation of less-differentiated TCR-engineered T cells thus leading to increased persistency and ultimately improved efficacy (71, 79–81).

Autologous T cells are available in limited quantities and their state is likely to be affected by the multiple rounds of cancer treatments which patients undergo before ACT, complicating the manufacturing process and the feasibility of TCR gene transfer therapy. Therefore allogenic universal T cells have also been exploited for gene transfer therapy (82) ( Figure 2 ). A few issues are to be noted, including competition with endogenous TCRs, mispairing of TCR subunits, risk of off-target toxicity due to allogenic TCR-T cell infusion and ACT product rejection by the host (83). It is thus required to KO endogenous TCRs to improve the safety and efficacy of TCR-ACT with allogenic cells (49). For all these reasons, several alternative non-T cell types were evaluated.

Mensali and colleagues provided the first proof of concept that cells other than αβ T cells could be used as recipient cells for TCR gene therapy. They transferred a TCR into NK cell line, NK-92, and demonstrated efficacy in vivo (84). More recently, Parlar and coworkers have engineered NK cells with a tyrosinase-specific TCR and highlighted their cytolytic potential in vitro (85). A potential benefit of using NK cells as a carrier is their ability to remain cytotoxic in an MHC-independent manner. This could be of interest in case of MHC loss, which is a common immune suppressive mechanism exerted by tumors (86). Alternatively, γδ T cells can also be TCR-engineered ( Figure 2 ), thus avoiding the issue of TCR mispairing. The latter strategy was proven efficient in leukemia (87). Furthermore, as T cell exhaustion is a critical component of ACT efficacy, there is an increasing interest in using induced-pluripotent stem cells (iPSCs) for gene transfer therapy ( Figure 2 ). Nishimura and colleagues were able to generate iPSCs from antigen-specific CD8 T cells and to re-differentiate them. In this way, they obtained rejuvenated cells with longer telomeres and a high proliferative potential, making them fitter for therapy (88). The efficacy of TCR-engineered iPSCs was shown in vivo (89).

Neoantigen-specific T cells, and hence their cognate TCR, can be identified in different samples including tumor or blood from patient or naïve T cells from healthy donors. Briefly, neoantigens can be identified from expanded TILs (7, 95, 101). However, due to low frequencies, sensitive detection sometimes requires antigen-specific in vitro stimulation (IVS). Of interest, we developed a novel strategy to improve the detection of neoantigens in TILs, based on the addition of pools of predicted neo-epitopes at the initiation of the TIL culture (7). Alternatively, TILs can be enriched prior to culture by sorting of dissociated tumor material ex vivo, based on various activation markers like PD-1 (CD279), OX40, CD137 (4-1BB), CD39 and CD103 (102–106).

Peripheral blood lymphocytes (PBLs) can also be exploited for neoantigen identification. IVS with antigen presenting cells (APC) loaded with neoantigen candidates have been extensively used to detect neoantigen reactivity (7, 30, 96, 107–110). Prior enrichment strategies have also been used by different groups including: the isolation of memory (107) or naïve T cells (108) or the sorting with PD-1 (109). Most identification processes use PBLs from autologous origin but it has been shown to be possible from allogenic sources as well (108).

Upon their identification, neoantigen-specific T cells can be purified based on pMHC multimer staining or based on the up-regulation of activation markers following specific-activation, such as 4-1BB and OX40 (7, 107, 108, 111, 112). Isolated cells then undergo bulk α and β TCR sequencing in order to select the dominant α and β TCR clonotypes. Of note, TCR repertoire analysis strategies are challenging due to the high diversity of TCR repertoires (113). Advances in next-generation sequencing has improved the interpretation of TCR repertoires. Using RNA as a source of material allows allelic exclusion thereby avoiding to overestimate repertoires diversity. RNA is also more sensitive than DNA despite being less quantitative due to variation in expression levels (114). Among the different sequencing methods, multiplex polymerase chain reaction (PCR) (115) remains the most commonly used strategy, despite misrepresentation of clonotypes proportion introduced by heterogeneity in primers efficiency (116). Other approaches rely on the addition of adaptors prior to PCR amplification (117, 118), such as the 5’ RACE PCR or TCR amplification following gene capture (119). For each method, the bias in quantification reduces the ability for easy pairing of α and β chains, which is required for therapeutic applications. A concept based on multiple sequencing and combinatorial analysis was developed to pair αβ TCR chains, yet this strategy is limited to high-frequency clonotypes and requires large cell numbers (120). To avoid the above-mentioned limitations, both T cell cloning or single-cell sequencing can be used.

Interrogation of neoantigen-specificity and antitumor-reactivity of candidate TCRs can be assessed following the expression of TCR candidates into recipient cells, a strategy hereafter referred as TCR cloning. Antigen-specificity can be challenged by transducing TCRs into activated PBLs or Jurkat T cells which are then co-cultured with APCs loaded with neoantigens (29, 86, 94, 95, 99, 101, 102, 107, 121–123). Next, tumor-reactivity can be measured by co-culture of TCR-engineered T cells with autologous tumor cells or APCs pulsed with tumor lysate (86, 94, 101, 102, 121, 124, 125). To enable a rapid identification of neoantigen-reactive TCRs, Paria and colleagues have developed a TCR cloning methodology using Jurkat T cells electroporated with RNAs encoding TCR α and β chain, respectively (126). The benefit of this approach lies in the use of RNA electroporation, which is faster and more efficient than TCR transduction by genetic engineering. TCR expression is transient but sufficient for TCR interrogation. Interestingly, they used the Jurkat luciferase system (under NFAT promotor) which is a rapid and easy read-out.

Of important note, the reactivity of validated neoantigen-specific TCRs to the wild-type peptide should be evaluated to avoid autoimmunity and thus ensure patient safety. This can done by performing a peptide dose response with APCs (86, 94, 95, 99, 102, 107). Another method consists of using a high-throughput genetic platform (127).

As an alternative to the aforementioned time-consuming strategies (based on TCR isolation and downstream TCR validation), new developments enable the direct identification of tumor-specific TCRs by pre-selecting the most frequent clonotypes from fresh tumor samples (121) ( Figure 2 ). Selected αβ TCR pairs are then challenged against tumor cells by TCR cloning. Others have however found that the tumor reactivity of intra-tumoral TCR repertoire was low (125), highlighting a potential limitation of using this strategy for gene transfer therapy. Identified tumor-reactive TCRs can then be examined retrospectively for antigen specificity using ligand discovery approaches, based for instance on trogocytosis, a process in which T cells exchange membrane proteins with APC presenting candidate antigens (128). Other possibilities for TCR ligand discovery include the use of chimeric signaling and antigen presenting bifunctional receptors (SABRs) (129). It is to be noted that the development of microfluidic-based platforms will accelerate and automate the isolation of relevant TCRs for TCR-ACT in the near future (130–132).

Instead of direct TCR identification, combined single-cell transcriptomics and TCR sequencing (133) could be exploited to define intra-tumoral signatures of neoantigen-specific or tumor-reactive TCRs (134). Old-generation single-cell methodologies are tedious and allow the sequencing of a limited number of cells (135, 136). Single-cell TCR sequencing and profiling of T cells was recently eased by the development of a commercially available strategy by 10x Genomics (137–139). Briefly, the latter microfluidic-based technology is able to generate an emulsion containing one trapped cell with a uniquely barcoded bead and, upon cell lysis polyadenylated mRNA is captured. The process results in barcoded libraries, which undergo both downstream single-cell TCR sequencing and transcriptomic profiling. Overall, 15’000-20’000 cells can be covered. The 10x Genomics strategy allows the analysis of clones in a timely manner and was applied for the identification of a signature of clone persistency in the circulation after ACT (140). In the near future, intra-tumoral signatures of neoantigen and/or tumor-reactive TCRs may be defined based on a comprehensive database of TCRs combined to their ex vivo transcriptomic profiling. We foresee that these signatures will enable the direct identification of relevant TCR pairs from ex vivo tumor, which will considerably facilitate and accelerate the selection of TCR candidates for gene transfer therapy.