Using engineered multifunctional T cells for precise and effective immune therapies

Meganucleases, or homing endonucleases, are early gene-editing tools that recognize long DNA sequences (14–40 bp) with high specificity and minimal off-target effects (12); however, they are difficult to reprogram for new target sequences (27). Engineered variants, such as megaTALs (TALE fused to meganucleases), have been applied to T-cell editing, such as T-cell receptor alpha constant (TRAC) region disruption. Zinc-finger nucleases (ZFNs), introduced in 1996 (28), combine zinc finger DNA-binding domains with FokI nucleases to induce targeted DNA breaks (29). They showed genome-editing potential in eukaryotic cells (30) and were used for CCR5 disruption in human T cells (31). However, ZFNs are complex and costly to design (32 –34), leading to the development of TALENs in 2010 (14, 35). TALENs recognize individual nucleotides with high precision, target longer sequences than ZFNs, and offer a simpler design; however, their large size (~3 kb) limits their multiplex gene-editing applications (36).

Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)-Cas9 has emerged as a flexible and scalable gene-editing system that enables efficient multiplex editing via a simple guide-RNA (sgRNA) design. Discovered as bacterial repetitive sequences (37) and later identified as part of bacterial adaptive immunity (38), CRISPR-Cas9 enables the RNA-guided targeting of almost any DNA sequence with high efficiency (15). Unlike ZFNs and TALENs, CRISPR-Cas9 does not require target-specific protein engineering and induces double-strand breaks (DSBs) that are repaired by cellular pathways such as non-homologous end joining (NHEJ) or homology-directed repair (HDR) (39). Thus, it remains a promising tool for precise multiplex genome engineering (39, 40). However, CRISPR-Cas9-induced DSBs can cause off-target mutations and chromosomal translocations, compromising genome integrity (41) (see quality control considerations below). NHEJ is error-prone and reduces precision (42), whereas HDR is inefficient, particularly in non-dividing cells. Strategies to enhance HDR include tumor suppressor p53-binding protein 1(53BP1) inhibition and RAD18 (Radiation-sensitive 18) (43, 44), modified CRISPR systems, small molecules modulating DNA repair, and co-localization of repair templates with Cas9 (45).

In 2016, base editing was introduced by Liu et al. to enable precise single-nucleotide changes without DSBs or donor DNA templates. Cytosine and adenine base editors allow direct base conversions and reduce genomic rearrangements (46), although they are limited to transitions and are prone to off-target deamination and bystander editing. Prime editing (PE) was developed to address these limitations, enabling diverse substitutions, small insertions, and deletions without DSBs using Cas9 nickase fused to reverse transcriptase and pegRNA to direct edits (47). Despite their high precision, challenges include low efficiency in some cells, large construct sizes, and pegRNA mispriming (47). Prime editing evolved from PE1 with natural reverse transcriptase to PE2 using engineered enzymes, and PE3 added a second nick to increase efficiency but with increased indels, which PE3b mitigated by timing the second nick (48). PE4 and PE5 further enhance precision with DNA repair inhibitors (49), and enhanced Prime Editors (ePE) improve pegRNA stability and editing yield (50). These developments have progressively improved the efficiency, specificity, safety, and suitability of prime editing for multiplex gene-editing applications.

In addition to DNA-targeting tools, RNA-targeting technologies, such as CRISPR-Cas13, have emerged as powerful alternatives for modulating T-cell function without permanent genomic changes (51). Cas13 cleaves single-stranded RNA transcripts to transiently and reversibly regulate gene expression (52). In a striking demonstration of Cas13-based knockdown possibilities in T cells, Tieu et al. revealed that the co-transduction of Cas13d and multiple sgRNAs could reduce the expression of multiple target genes simultaneously, enabling the suppression of multiple immune checkpoints or entire metabolic pathways (53). This system can also be used for combinatorial screens and can be modified to permit drug-controlled Cas13d expression and graded target gene suppression, which may be advantageous over complete ablation in certain settings. Catalytically inactive Cas9 (dCas9) can be fused to transcriptional activators, leading to specific gene expression (CRISPR activation) in CAR T cells (54) without altering the DNA sequence at the targeted loci. The fusion of dCas9 E to epigenetic modulators enables targeted chromatin remodeling and gene expression changes without DNA breaks (55, 56). Early studies in primary T cells revealed that this approach could stabilize Foxp3 expression in mouse regulatory T cells and delay replicative senescence in stimulated human T cells through the expression of telomerase reverse transcriptase (TERT) (57, 58). Although all are at the pre-clinical stage, these RNA- and epigenetic-targeting approaches offer great promise for the precise control of T cell phenotypes while reducing the risks associated with permanent genomic alterations.

The insertion of new genetic material or genome editing requires the delivery of different cargoes, depending on the method used. Gene delivery methods can be classified into viral and non-viral approaches, each with specific advantages and disadvantages. Gammaretroviruses allow stable integration but require dividing cells to do so. While insertional mutagenesis is a theoretical concern for any retroviral vectors, experimental evidence in mature T lymphocytes suggests that these cells are relatively resistant to transformation (59). Long-term follow-up studies of patients treated with gammaretroviral-modified T cells have not reported malignant transformation (60, 61). Lentiviruses, a retroviral subclass, can transduce both dividing and non-dividing cells (62) and support stable gene expression, in addition to enabling multiplex gene editing (63) and the insertion of large polycistronic constructs. Lentiviral vectors offer key advantages for CAR T-cell therapy, including efficient T-cell transduction, durable expression, and a safer integration profile than gammaretroviruses (64). Third-generation lentiviral systems further enhance safety by separating viral components and using self-inactivating elements (65, 66). Importantly, no significant genotoxicity or malignant transformation has been reported in clinical CAR T-cell applications using lentiviral vectors (67), although concerns regarding insertional mutagenesis remain (68).

Adenoviruses are non-integrating viruses in the host genome and provide transient gene expression. However, they can elicit strong immune responses directed against the viral vector, which may limit their therapeutic efficacy (69). AAVs are less immunogenic and can support longer transgene expression durations; however, they are limited by their small cargo capacity and high production costs (70). Although less of a concern for other cellular engineering approaches, the risk of immunogenicity is a preoccupation whenever a foreign (natural or synthetic) molecule is introduced into a therapeutic cellular product. Cas9 nucleases (bacterial proteins) can elicit T-cell and humoral responses, and pre-existing immunity is prevalent in the population (71, 72). Despite the possibility of reducing immunogenicity through protein engineering (73), current clinical protocols using Cas9 modified T cells insist on transient exposure to Cas9 and the absence of the protein in the final product. Synthetic proteins (artificial receptors for example) may also be recognized as non-self and impact the persistence of the transferred T cells (74, 75). Even if heavily treated cancer patients may not be able to mount immune responses against foreign proteins as well as normal individuals, multiplex editing and/or the introduction of multiple artificial transgenes could increase the risk of early rejection.

Transposon-based systems, such as Sleeping Beauty, offer non-viral, nuclease-free integration with low cost and large cargo capacity (18, 76 –78). However, they face limitations in transfection efficiency, delivery synchronization, and the risk of semi-random integration (79, 80).

Non-viral methods such as electroporation, nucleofection, CPPs, lipofection, nanoparticles, and transposons offer low immunogenicity, simplified production, and reduced biosafety risks (10, 23, 81 –83), but often result in lower efficiency and transient expression, unless paired with integration systems. Among non-viral methods, electroporation and nucleofection are efficient for delivering genetic material into primary T cells, supporting the simultaneous delivery of multiple components (84, 85). This makes them ideal for multiplex editing strategies in T-cell engineering. Electroporation is a widely used method for delivering ribonucleoproteins (RNPs) or mRNA for CRISPR-based gene editing, allowing the effective delivery of Cas proteins and gRNAs without viral vectors (86). Despite being scalable to suit clinical purposes and yielding a high number of genetically modified T cells, electroporation and nucleofection can lead to significant cytotoxicity, especially when applied to minimally cultured or naïve T cells, potentially compromising the quality of the final product (10, 23, 83). Editing is typically performed prior to, or early after, T-cell activation (within 24–48 h) to maximize repair efficiency and viability (87, 88). This is important when both CRISPR-Cas9 and gammaretro-lentiviral methods are used on the same T cells for multiplex engineering purposes. Early editing with CRISPR-Cas9 avoids the cleavage of integrated vectors if CRISPR targets overlap with viral vector sequences (88). Recent clinical studies have highlighted that editing resting or minimally activated T-cells reduces chromosomal abnormalities linked to DSBs, supporting carefully timed editing workflows for safety and efficacy in clinical manufacturing (see Section 2.3). Another non-viral method for gene delivery into T cells is the use of CPPs which are short peptides that can traverse cell membranes and facilitate the intracellular delivery of various cargos, including nucleic acids and proteins. This approach has been investigated for the delivery of CRISPR/Cas9 components into T cells. For example, CPPs such as PepFect14, LAH5, TAT peptide, Transportan-10, and MPG have been successfully applied to deliver CRISPR/Cas9 plasmids or RNP complexes into primary human T cells (89 –92). Although promising, CPP-mediated delivery still faces challenges, such as potential cytotoxicity, limited efficiency compared to viral or electroporation-based methods, and the need for optimization to achieve robust genome editing in clinical-grade T cell products (93 –95). Hence, combining gene-editing tools (e.g., CRISPR-Cas9, base editors, transposons) with gene delivery platforms, such as lentiviral vectors, allows the creation of customized multi-edited or multi-functional T-cell products.

Multiplex gene editing in T cells offers substantial therapeutic potential but also raises significant concerns about genotoxicity due to the induction of multiple DNA double-strand breaks (DSBs). These breaks can lead to chromosomal translocations, large deletions, and complex rearrangements, such as chromothripsis, which compromise genome stability, reduce cell viability, and may even induce transformation or cancer (88, 96). Several strategies have been developed to reduce the risk of genotoxicity. These include the use of nucleases with different cutting patterns (such as Cas9 and Cas12a/b), temporal separation of sgRNA delivery, and favoring ribonucleoprotein (RNP) delivery over plasmids to limit the active time of nucleases. Promoting HDR over NHEJ, carefully designing sgRNAs, and transiently inhibiting NHEJ using small molecules, such as NU7441, can further improve genomic safety (96). Additionally, studies have shown that performing gene editing within 24 to 48h after T cell activation reduces p53-dependent DNA damage responses, thereby enhancing editing efficiency and cell recovery (88). It is also important to consider the risk of damaging the integrated vector sequences when combining multiplex editing and lentiviral transduction. Cas9 activity near or within a vector can disrupt transgene expression or cause loss of function (88). Therefore, delaying transduction until 48–72 hours post-editing is often beneficial, although this must be balanced by the activation status and susceptibility of T cells to infection. To evaluate the genotoxic impact of the editing process, assessment of cellular stress and DNA damage response (DDR) markers, such as p53, γH2AX, and apoptosis indicators, is essential, as DDR plays a central role in detecting and repairing DNA damage, maintaining genome stability, and preventing mutagenesis and tumorigenesis (97). Chromosomal rearrangements caused by CRISPR-Cas9-induced double-strand breaks (DSBs), especially during multiplex editing with multiple sgRNAs, are concerning (96). To mitigate these risks, high-fidelity Cas9 variants and novel editing platforms, such as base and prime editors, are being actively explored. Base editing enables precise nucleotide substitutions without double-strand breaks (DSBs), reduces genotoxic risks such as deletions, translocations, and p53 activation (46, 98), and is especially suited for multiplex or subtle edits. Off-target effects are also a major consideration in both CRISPR and base editing technologies, and require comprehensive assessment using techniques, such as GUIDE-seq, CIRCLE-seq, or deep whole-genome sequencing to evaluate specificity (99, 100).

This review focuses on multiplex gene editing, as it is increasingly moving into the clinical stage. Hence, safety concerns must be stratified and addressed according to the number and type of edits. CRISPR-Cas9-mediated double-strand breaks (DSBs) can activate the p53 pathway, induce chromosomal translocations, and drive immune responses, with the frequency of deleterious events increasing when multiple loci are targeted (101). Base and prime editors reduce DSB-related risks but will require long-term surveillance to monitor low-frequency off-target effects (102). Long-term surveillance is required to monitor the potential clonal expansion of edited T cells. The maximum number of gene edits or off-target lesions that a T-cell can sustain before functional impairment or death is unknown and likely depends on the genes being targeted. However, "over-engineering" remains a theoretical issue. Strategies that do not induce double-stranded DNA breaks, such as base and prime editing, appear to be ideal for the genetic editing of multiple genes through a single engineering step. Multiple gene knockdown with limited genetic risk may also be achieved using traditional viral vectors encoding multiple miRNAs (103) and Cas13-based approaches (as described above).

However, viral vectors require vigilance and well-designed quality control (QC) strategies. One key parameter is the vector copy number (VCN), as high VCNs (>5–10 copies per genome) in lentiviral and gammaretroviral systems are linked to an increased risk of insertional mutagenesis and oncogene activation (104). Therefore, clinical protocols aim for a VCN of 1–5 copies per cell to balance transgene expression with genomic safety (105). Additionally, verifying full-length transgene integration and expression is essential when using viral vectors, which is typically assessed using digital droplet PCR (ddPCR), long-read sequencing, or functional assays (106). These assays are usually required to complete other QC assessments necessary to evaluate the identity, purity, potency, and sterility of the final cellular product. The regulatory landscape for cell and gene therapies is rapidly evolving and can vary according to jurisdiction, mandating careful planning of quality control strategies by advanced T-cell product developers.

Modulating intrinsic T-cell physiology, conferring drug resistance, and mitigating alloreactivity are relevant to the design of T-cell immunotherapies. However, optimized cellular immunotherapies must also consider the biology of cancer cells and their environments. This section reviews how multi-functional and multiplex T-cell editing can address the crucial issues of cancer cell immune escape, trafficking into tumor beds, and resistance to hostile tumor environments. Most cellular engineering designs used thus far rely on a single engineering step consisting of the viral transduction of vectors containing multiple genes, conferring multiple functions. However, multimodal (viruses and nucleases) and non-viral methods are increasingly being used. Early phase clinical trials are being conducted to test several of these strategies, with clinical results increasingly available (see Section 4 below).

CAR T-cell therapy has changed the treatment of hematologic malignancies; however, its clinical application is often hindered by various toxicities. Among the most common and severe adverse effects are cytokine release syndrome (CRS), marked by systemic inflammation, fever, hypotension, and the potential for multiorgan failure due to excessive cytokine secretion by CAR T cells and other immune cells (214). Neurotoxicity, also known as immune effector cell-associated neurotoxicity syndrome (ICANS), presents with neurological symptoms such as confusion, seizures, and encephalopathy, likely caused by endothelial activation and blood–brain barrier disruption (215). On-target/off-tumor toxicity occurs when CAR T cells attack healthy tissues expressing the target antigen, leading to collateral damage to healthy tissues. B-cell aplasia is a specific on-target toxicity observed with CD19-directed CAR-T therapies, causing the depletion of normal B cells and increased infection risk (216). Tumor lysis syndrome (TLS) may follow rapid tumor cell destruction, causing metabolic imbalances and renal impairment (217). Additionally, macrophage activation syndrome (MAS) or hemophagocytic lymphohistiocytosis (HLH) represents a severe hyperinflammatory state linked to CAR T-cell therapy (218). In solid tumors, direct organ damage caused by the expression of the target antigen by epithelial or stromal cells may lead to significant toxicity (219). Although less frequent, off-target toxicity arises from unintended gene editing or cross-reactivity and may harm non-target tissues (220 –222).

Various safety strategies have been developed to reduce these risks (Figure 3). First, suicide switches, such as inducible caspase-9 (iCasp9), allow for rapid T-cell elimination upon administration of a small molecule and have been tested mainly in hematologic cancers (223). Clinical studies have validated iCasp9 in early phase trials, demonstrating reproducible elimination of infused cells and control of adverse events, with multiple dosing cycles feasible without cumulative toxicity (223, 224). Second, elimination markers, including truncated EGFR and CD20, provide targets for antibody-mediated depletion mechanisms (225, 226). Clinical evaluation of elimination markers has shown efficient CAR T cell depletion in vivo, confirming their potential to mitigate severe toxicities when necessary (227, 228). Next, logic-gated CARs that require dual antigen recognition enhance tumor specificity and reduce off-tumor effects (229). Preclinical validation of logic-gated CARs has demonstrated improved tumor selectivity and reduced off-tumor cytotoxicity, providing evidence of their translational potential (229, 230). Synthetic Notch (SynNotch) receptors enable the spatial restriction of CAR expression (231), whereas inhibitory CARs (iCARs) attenuate activation upon engagement with antigens on healthy tissues (232). Preclinical studies of SynNotch and iCAR systems have shown reduced systemic toxicity while maintaining antitumor efficacy (231, 232). Moreover, tunable systems, including tetracycline-responsive promoters and small-molecule "ON-switch" CARs, offer external control of CAR activity (233, 234). ON-switch CARs have been functionally validated in preclinical models, showing controlled T cell activation and mitigation of cytokine-mediated toxicity (235, 236). Additionally, tumor-selective protease-activated CARs and activation-inducible TRUCKs (for IL-12 production for instance as described above) confine potentially toxic cytokine secretion to the tumor microenvironments, and similarly, promoters such as NR4A2 and RGS16 have been designed to drive CAR or cytokine expression specifically in the tumor microenvironment, ensuring that the transgene is predominantly active only in tumor tissue, thereby enhancing safety and minimizing off-tumor effects (194, 237 –240). Clinical or preclinical validation of TRUCKs and promoter-restricted CARs has demonstrated reduced off-tumor activity and lower systemic cytokine release (240 –243). Finally, dual CAR systems, multi-step activation designs, and so-called masked CARs that rely on intra-tumoral proteolytic removal of a peptide blocking the CAR's antigen-binding site for localized activation further restrict activation to malignant contexts (229, 239).

The tumor environment can be further leveraged through metabolic switches. The best example of this is hypoxia (244). For example, the incorporation of hypoxia-responsive elements (HREs) within transgene promoters, combined with oxygen-dependent degradation domains (ODDs) derived from HIF1α, enables selective transgene expression and stabilization only under low-oxygen conditions, which are prevalent in tumors. Hypoxia-sensing CAR T cells (HypoxiCAR) have demonstrated significant mitigation of systemic toxicities while preserving robust antitumor efficacy in preclinical models (244, 245). While clinical data are still emerging, preclinical studies have confirmed that HypoxiCAR significantly reduces systemic cytokine release and off-tumor toxicity while retaining antitumor activity (246, 247). Similarly, CAR-T cells engineered with multiple HREs in their promoters (e.g., 5H1P-CEA CARs) showed enhanced tumor specificity and reduced activity in normoxic environments, leading to improved safety profiles (248). Furthermore, the fusion of cytokines, such as IL-12, to ODD domains (CAR19/hIL12ODD) ensure controlled cytokine release restricted to hypoxic tumor microenvironments, minimizing systemic inflammatory side effects (249). These innovations complement existing molecular safety switches and underscore the potential of microenvironment-responsive CAR designs to optimize therapeutic windows and reduce adverse events.

Preclinical development should incorporate sensitive biomarkers to detect early signs of toxicity and evaluate next-generation platforms in the context of tumor heterogeneity, antigen escape, and the tumor microenvironment. Collectively, these strategies build a safety framework that is necessary to optimize both the efficacy and tolerability of engineered CAR T-cell therapies.

Several T cell therapies have integrated many of the concepts described above into single T cell products. Representative examples that led to the clinical trials summarized in the next section illustrate the vast potential of contemporary cellular engineering methods.

Using a combination of viral and non-viral gene engineering approaches, several groups have pursued the objective of creating safe, rejection and fratricide-resistant, allogeneic "off-the-shelf" CAR T cells to treat T-cell leukemia. Targeting the universal T-cell marker CD7, this strategy involves the expression of anti-CD7 CAR through lentiviral transduction, CRISPR-Cas9 for CD7, and TCR alpha chain (TRAC) ablation to avoid fratricide and GVHD, respectively (250). This approach has been refined with the use of base editing and the addition of CD52 deletion to evade alemtuzumab-based lymphodepletion and has been tested in early phase clinical trials (251, 252). Base editing technologies have high potential for multiplex editing of therapeutic T cells. It was recently demonstrated that the simultaneous knock-in of a CAR transgene and knockout of four genes, B2M, to suppress MHC class I expression, CD52, TRAC, and PD-1, could be achieved without genotoxicity (253). Pushing even further, it is reported that a combination of adenine base editing and Cas12b nuclease could generate "stealth" knock-in CAR T cells resistant to allorejection, GVHD, adenosine, PD-L1, and TGF-β through the editing of B2M, CIITA (to suppress MHC-class II expression), CD3E (to suppress TCR expression), Adenosine A2a receptor (ADORA2A), PD-1, and TGFBR2 genes (254). Along the same lines, other groups used either non-viral or viral (transduction of multiple sgRNAs) to perform multi-editing in CAR T cells, similarly focusing on the ablation of the TCR/MHC axis and immune checkpoints (138, 255, 256). Multiplex editing has also been applied in transgenic TCR therapy. Deletion of TIM-3, LAG-3, and 2B4 genes by CRISPR-Cas9 led to superior functionality and resistance to exhaustion of transgenic TCR NY-ESO-1-specific T cells in preclinical myeloma models. The modular nature of electroporation-delivered CRISPR-Cas9-sgRNA complexes allowed for the comparison of single versus multiple edits, outlining the role of each immune checkpoint molecule (257). Taken together, multiple studies support multiplex editing as a reliable and clinically applicable approach for adoptive T-cell immunotherapy.

To date, the published clinical data on multiplex-edited therapeutic T cells principally relate to two key concepts we previously described: resistance to exhaustion through the editing of immune checkpoint genes and the avoidance of alloreactivity, especially to facilitate the persistence of allogeneic products.

Although all early phase, highly heterogeneous (engineering approaches, diseases, allogeneic vs. autologous, number and impact of previous or concomitant therapies, multiple dose levels, etc.) and generally including small numbers of patients, these clinical studies offer crucial insights. Many pioneering studies have included detailed assessments of genotoxicity, which are overall reassuring for both CRISPR-Cas9 and base editing approaches. Although off-target editing and chromosomal anomalies have been reported in some studies, no functional impact has been noted. While firm conclusions about safety will require a longer follow-up (few studies report outcomes beyond a few months), these early results support the further development of advanced engineering methods. Compared to standard CAR T cells, multiplex-edited CD19 CAR T cells appear to confer a similar risk of adverse events, such as CRS and cytopenia, the latter being largely attributed to the lymphodepleting regimen. Similarly, early evidence of efficacy is difficult to interpret in the absence of a control group, which is expected in phase I-II or proof-of-concept studies. As previously reported, hematological cancers respond better to T-cell therapy than solid cancers. Other currently investigated designs, including TRUCK or migration-enhanced T cells (as described above), may improve the response in solid cancers and feed a new wave of multiplex engineered T cells in solid tumors (as discussed in Section 4.2). A note of caution regarding allogeneic products is the relatively limited persistence of engineered cells, as reported in several studies. Suppression of all MHC I molecule expression through B2-microglobulin editing may be conducive to NK cell-mediated rejection to a greater extent than selective HLA-A/B editing (as (258)). Preclinical studies suggest that other strategies to mitigate allogeneic T cell rejection by NK cells include the ablation of the adhesion ligands CD54 and CD58 and may be considered in multiplex engineering designs (195). Multi-engineered T cells may have a survival disadvantage in certain settings. Loss of TCR expression has been shown to affect persistence in one of the studies (133). Successful TRAC-edited mesothelin CAR T cells did not persist as long as unedited T cells, suggesting a plausible homeostatic role for TCR signaling (259). In addition, PD-1 editing may precipitate T-cell dysfunction and loss, as suggested in the first study reporting on multiplex and CRISPR-Cas9 engineered T cells in human (260), and in line with a previous study in PD-1 knock-out mice (261). Hence, multiplex engineering allows for the counteraction of certain constraints of T-cell immunology but may unveil both predictable and unsuspected vulnerabilities.

Long-term remissions, complete responses, or bridging to potential curative therapies following engineered T-cell administration, have mostly been described for hematopoietic neoplasms. This is also true for multiplex-edited T cells and approaches targeting multiple antigens. Although not thoroughly discussed here because the therapeutic T cells tested had not undergone multiplex editing, recently published clinical studies investigating CRISPR-Cas9 edited cancer patient T cells or TILs from cancer patients are reassuring about the feasibility and safety of gene-engineered T cells. In the first study, lung cancer patients received autologous peripheral blood PD-1 edited T cells manufactured from peripheral blood (262). Detailed analysis revealed no major genotoxicity, and the treatment was well tolerated. In a landmark study that included clinical results, autologous NY-ESO-1 transgenic TCR T cells were generated following lentivirus delivery to patient T cells previously edited at the TRAC and TCR beta chain (TRBC) gene loci to avoid TCR chain mispairing and at the PDCD1 (PD-1) locus using the CRISPR-Cas9 approach (260). In another study using gammaretrovirus-mediated neo-antigen-specific TCR transgenic expression in patients with metastatic colon cancer, three out of seven patients had objective responses (156). Multiantigen targeting with personalized neoantigen-specific TCR is thus feasible but poses significant financial and logistical challenges that may be partly alleviated with non-viral methods. Such strategy was used in another trial, where personalized neo-antigen-specific TCR were inserted in situ at the TRAC locus and reinjected in patients with metastatic cancer. Approximately one-third of patients had stable disease following treatment, and no significant toxicity was observed (155). The discovery of "public" cancer-specific antigens, such as KRAS^G12V will further facilitate the design of multi-antigen targeting in solid tumors (263 –265). In lymphoblastic leukemia, patients who received CD19/CD22 (tandem or sequential) had a higher complete response and minimal residual disease-negative rates than those who received CD19 CAR T cells alone (266). Although additional evidence will be required to firmly conclude on the efficacy of the various CD19/CD22 approaches (267, 268), the concept of multi-antigen targeting using CAR T cells may be translated to solid tumors with recent pre-clinical data suggesting that a tandem mesothelin-MUC16 CAR is superior to monospecific CARs (269).

Other improvements in TILs have also been tested in the clinic. The production and injection of CISH-deleted gastrointestinal tumor-derived TILs has recently been reported. Gene-edited TILs were successfully manufactured in 86% (19/22) of the recruited patients, and 12 patients were treated. Side effects were as expected and unrelated to the TIL product; six patients had stable disease, and one patient with microsatellite instability achieved a complete response (144). Other approaches are currently under clinical investigation, including PD-1 deletion and cytokine signaling modulation (dominant-negative TGF-β receptor, etc.) to improve T-cell function, as well as several armored T-cell products aimed at altering the hostile tumor microenvironment (as described above), but very limited published clinical data are available.

Although disappointing compared to the results in hematopoietic neoplasms, engineered T-cell therapies for solid tumors are feasible. It is to be expected that clinical trial designs will have to account for the multi-layered complexity of solid tumors by enabling the simultaneous and coherent targeting of multiple antigens, intrinsic T-cell dysfunction and extrinsic constraints of the tumor microenvironment. While multiplex T-cell engineering can address some of this complexity, multi-product treatment schemes and optimization of the integration of cellular products relative to other treatments (timing, repeated dosing, etc.) will be required.