Toward “off‐the‐shelf” allogeneic CAR T cells

Chimeric antigen receptor (CAR) T cell therapy represents a major breakthrough in the field of immuno‐oncology. Many potential issues are apparent for autologous CAR T cell therapy, such as time for manufacturing and need for interim therapies in progressing patients, wide variations in terms of quality and quantity of T cells, and difficulty to obtain enough cells for redosing. “Off‐the‐shelf” allogeneic CAR T cells premanufactured from third‐party donors may theoretically provide solutions to these different problems. However, allogeneic T cells possess foreign immunological identities that can lead to histocompatibility considerations such as graft‐versus‐host disease and rejection of allogeneic cells. This review outlines the major recent advances for off‐the‐shelf T cell therapies currently in clinical trials or in preclinical development and describes strategies for reengineering or selecting specific T cell immune identities to create safe and efficient immunotherapies for patients.

• T cell variability: Patient-specific variation in T cell phenotype and prevalence, like the presence of more or less 'early' or 'stem celllike' T cell subsets, may yield different expansion and persistence of cell products. 8 Many factors may be involved in this T cell variability: age, disease, previous lines of treatment, explaining the interpatient heterogeneity observed with the use of autologous CAR T cells. 8 • Risk of insufficient T cell expansion: Starting material is limited to T cells collected from leukapheresis and may not sufficiently expand to high numbers during manufacturing, especially in patients who are heavily pretreated and lymphopenic. 3,5 • T cell dysfunction: A cancer patient's T cells may be dysfunctional as a result of disease burden or prior lines of therapy, and may therefore be a poor drug delivery vehicle for CAR T cell therapies. [9][10][11][12] Collecting T cells during the first lines of cancer therapy may reduce the risk associated with the cancer treatments but not the effects of tumor-associated immunosuppression.
• Limited opportunity for redosing: CAR T cells are prepared in a single batch and may be of limited quantity, such that patients may not have the opportunity to quickly and easily receive a new infusion of CAR T cells.
Unlike autologous T cell therapies, "off-the-shelf" T cells from third-party donors would be expanded in high numbers prior to treatment and made quickly available to patients. 13,14 This strategy could overcome the different limitations discussed above. A manufacturing protocol performed on T cells from a single donor leukapheresis has the potential to supply enough cells to produce hundreds of doses, and a single donor may be able to support multiple rounds of manufacturing. 15,16 This high-volume manufacturing allows for an opportunity to quickly and easily re-dose patients with a new offthe-shelf cell therapy without delays due to manufacturing or scale of production. Of note, in the potential instance of an immunogenic response to an off-the-shelf therapy, the use of cells from another donor further facilitates re-dosing options. Furthermore, a pretreatment manufacturing protocol allows for featuring multiple edits, manipulations, or TCR selection strategies that otherwise would be difficult to accommodate in the autologous setting. As CAR T design gains in complexity, this may become even more important, with the weight and complexity of safety assays making testing on individual products prohibitive.
A T cell therapy sourced from healthy allogeneic donors, however, introduces potential risks for issues related to incomplete histocompatibility. Due to the highly polymorphic nature of human leukocyte antigen (HLA) genes, allogeneic donor lymphocyte infusions are difficult to achieve without including consideration for risk of graft-versus-host disease (GvHD) 17,18 and durability of donor cells. 19 Complete donor-recipient matching of HLA haplotype is limited by low compatible donor availability, and healthcare providers must assess risk factors based on the extent of HLA matches for unmodified, unselected, or unmanipulated allogeneic T cells. 20 In this Review, we will present the main approaches that are currently in development to deliver allogeneic T cell therapy and to address these potentials for GvHD and cell rejection. In this context, we will focus on TCR editing-based approaches and virus-specific memory T cells as examples of clinically advanced nonalloreactive T cells. We will discuss the challenge of persistence and the perspectives of optimization.

| S TR ATEG IE S TO AVOID GvHD
GvHD is believed to be largely driven by donor T cell recognition of host peptide-HLA complexes through the αβ T cell receptor complex (αβTCR). To eliminate αβTCR-mediated GvHD, two approaches can be employed: • Use of αβTCR negative cells: Editing cells to remove the TCR or using alternate cell types that lack αβTCR.
• Use of nonalloreactive T cells: Utilizing T cells with αβTCR specificity directed toward nonalloreactive targets.

| US E OF αβTCR NEG ATIVE CELL S
Cell therapy has broadly validated the use of αβ T cells as potent and efficient killers. [21][22][23] Given their relative abundance in peripheral blood and their ability to proliferate rapidly upon stimulation, they are attractive for large scale manufacturing of allogeneic cells to treat a substantial number of patients per batch. 24,25 Several strategies have been used to remove or decrease expression or function of the TCR complexes on these cells. In one approach, a truncated dominant-negative CD3ζ protein (termed TCR inhibitor molecule, or TIM) was expressed in T cells, resulting in reduced TCR signaling and protection from xenogeneic GvHD. 26 In another approach, an antibody-derived single-chain variable fragment specific for CD3ε was combined with amino acid sequences retaining it intracellularly. These protein expression blockers (PEBLs) colocalized intracellularly with CD3ε and inhibited surface expression of CD3 and αβTCR. 27 Small hairpin RNA (shRNA) has also been proposed to decrease expression of TCR chains. 28 Since the above approaches have the possibility to retain a low but functional amount of TCR, most entrants into the allogeneic space have opted for cellular gene editing approaches to ensure complete removal of the TCR. Several editing technologies are being used and all have in common their ability to cut DNA in a sequence-specific manner, creating a double strand break that then is patched using the cell's natural DNA repair machinery via nonhomologous end-joining (NHEJ). This restores the integrity of the chromosome but creates insertions or deletions (INDELS), which lead to disruption of the locus (most often TRAC, the TCRα common chain). With no functional TRAC gene, the TCR complex cannot form. 29 No approaches for αβTCR removal result in 100% of cells having αβTCR reduced or eliminated, necessitating a purification step to remove residual αβTCR positive cells. The efficiency of this process is critical to reduce the risk of any contaminating αβTCR positive cells that may induce GvHD.
Transcription activator-like effector nucleases (TALEN) were the first technology to be used in patients for αβTCR removal, 13 entering the clinic for compassionate use of a CD19 CAR for pediatric acute lymphoblastic leukemia (ALL) in 2015 30 and in two phase 1 trials for pediatric and adult ALL in 2016. 31 Grade 2 skin GvHD was observed in one of the two compassionate use cases, 30  Gene edited allogeneic CAR T cells are classed as gene therapies by the FDA, necessitating long-term follow-up of patients to monitor for safety. 33 Since gene editing may have a low frequency of off-target cutting, monitoring of predicted off-target cut sites and nonbiased approaches to identify off-target events are necessary.
Assays to look for any consequences of off-target cutting (such as eg inactivation of tumor suppressors) are also required, which includes demonstrating lack of growth factor independence. If more than one edit is made simultaneously, consequences of translocation must also be investigated and translocation frequencies must be monitored. 34 Although no instances of viral integration-related transformation have been reported with CAR T therapy to date, there is also a potential risk that this might occur given the random integration of both retroviral and lentiviral vectors. In addition to using gene editing to knock out αβTCR expression, it can also be used as a tool to add the CAR into the disrupted locus, thereby producing site specific integration. 35,36 Site specific integration may possess other advantages, for instance potentially lowering risk of translocations with two or more edits by favoring homology-directed repair rather than NHEJ. 37 Although αβ T cells represent the 'gold standard' of cell sources for therapy, they are not the only option available. Several effector cell types exist which lack an αβTCR or express an invariant version and are not expected to cause GvHD. 38 NK cells present in the peripheral blood for example can be expanded ex vivo to create an engineered product. CD19 CAR modified NK cells have been described, 39 showing activity in patients in a published phase I/II. 40 Eight 11 treated patients had a clinical response, including seven complete remissions (CRs) and one remission of a Richter's transformation with persistent chronic lymphocytic leukemia. Other cell types such as iNKT or γδT cells may be as potent as αβ T cells and may have advantages in terms of their innate receptor activity, their ability to be expanded readily to large numbers using their natural ligands 41 and ability to penetrate solid tumors. 42 . Their relative frequencies in peripheral blood are, however, very low and any contamination with alloreactive αβ T cells would require similar purification to avoid the risk of GvHD.
If a renewable cell source could be used to generate gene edited T cells or other cell types, clones with ideal characteristics and defined composition could be generated; for instance, absence of translocations, successful gene editing to result in a completely αβTCR negative cell source. Work has begun to derive T cells engineered from induced pluripotent stem cells (iPSCs). [43][44][45] These cells can be engineered, then genetically screened clones can be expanded and differentiated. Proof of concept for this approach has been generated for NK and T cells, 46 with at least one T cell product heading into the clinic (Table 1). Nonetheless, technical challenges still remain for alternatives to αβ T cells, including demonstrating functional equivalence to conventional T cells.
Most of the work described above has been carried out in the context of CD19 CARs since CD19 still represents the most clinically validated cell therapy target. Allogeneic cell products are now starting to be created against targets for other hematologic malignancies, such as BCMA, 47,48 and for some solid tumor targets. Although more challenges are expected for CAR therapies moving into solid tumor indications, these may be the areas where allogeneic approaches give the most benefit. Multiple gene edits to overcome solid tumor challenges may be possible, and the larger patient populations requiring treatment will be much more feasible with an allogeneic cell source.

| Selection of nonalloreactive T cells: A nongene edited approach
Alternative approaches to editing of αβTCR have also demonstrated significant clinical progress for off-the-shelf cell therapies. One example of this has been the use of virus-specific T (VST) cells derived from healthy donors. Exposure of lymphocyte collections to viral antigens ex vivo expand T cells expressing virus-specific αβTCR.
Virally-infected antigen presenting cells are commonly used for this stimulation and expansion of T cells expressing virus-specific αβTCR 49,50 and may be genetically engineered 51 or be directly loaded with viral peptides. 14 Similarly, virus-specific αβTCR can be directly isolated using MHC-bound antigen peptide. 52,53 The resulting virusspecific αβTCR repertoire functions via formation of a classical immune synapse and is thus activated via specific recognition of virus and not host peptide in complex with a specific HLA. As a primary therapeutic composition, VST cells can be manufactured from multiple donors, each associated with different HLA alleles, and banked in large numbers based on HLA restriction. 14,53 One immediate application for these off-the-shelf banks of VST cells is in the treatment of diseases originating from viral infection. For instance, Epstein-Barr virus (EBV) is a known driver of cell transformation and progression of several diseases including lymphoproliferative disorders, 50,54-56 nasopharyngeal carcinoma, 49 leiomyosarcoma, 57 and multiple sclerosis. 58 The first evidence of successful use of off-the-shelf VST therapy was performed with minimally HLA matched EBV-specific T cells to treat patients with EBV-positive lymphoma following solid organ transplant. 56

| E XPAN S I ON AND PER S IS TEN CE OF ALLOG ENEI C C AR T CELL S
The clinical efficacy of CAR T cells is associated with their initial expansion to achieve a sufficient effector-to-target ratio. Post-infusion CAR T cell expansion has been reported to be a major factor for clinical response in the case of autologous CD19 CAR T cells in B-ALL. 77 This expansion is enhanced by homeostatic cytokines such as IL-7 and IL-15, whose levels are elevated after lymphodepletion. 78,79 CAR T cell persistence can result in long-term immunosurveillance.
The persistence required to achieve a prolonged remission may vary according to the disease. One school of thought proposes that prolonged persistence may be required for a long-term control of diseases such as B-ALL. 5 In contrast, most of DLBCL patients achieving a CR at 6 months remain disease-free. 4 This difference probably reflects the biology of the disease and the persistence or not of a tumor reservoir after a first step of tumor killing.   82 However, autologous T cells expressing a CAR with a CD28 costimulatory domain rarely persist more than 1 or 2 months, whereas those with 4-1BB can persist for months or years. 5,77 Because the half-life of allogeneic CAR T cells will be reduced in most of cases due to rejection by the host immune system, it may be suggested in this context to use preferentially CD28 as a costimulatory domain, in order to obtain a rapid expansion of the CAR T cells and thus optimize the antitumor effect on a limited period of time.
The use of CD28 may be, however, associated with a higher risk of cytokine release syndrome compared to 4-1BB. 83 Of note, there is no data available to date in the clinic to compare allogeneic CAR T cells using CD28 versus 4-1BB as a costimulatory domain.
Once the risk of GvHD has been controlled, a potential issue re- CAR T cells are in this case protected by deletion of CD52. 18 This approach has been shown to lead to expansion of the allogeneic CAR T cells associated with efficacy. Two compassionate use patients first treated with UCART19 showed durable CR 30 and 82% of patients in the two phase 1 trials who were treated with the anti-CD52-containing lymphodepletion regimen achieved CR. 31 Furthermore, in two of three patients who failed to show benefit from the initial infusion, reinfusion gave minimal residual disease negative-CR. In a similar way, deletion of the dCK protein may provide resistance against purine nucleotide analogue chemotherapies such as fludarabine used for lymphodepletion. 85 These approaches have the advantage of suppressing all immune cell classes which may mediate rejection such as T, B, NK and potentially monocytes. However, the flip side of this benefit is that a prolonged immunosuppression is associated with a higher risk of opportunistic infections and viral reactivation.
Another approach consists in modifying allogeneic cells in order to obtain "universal cells" that will evade host immune detection. In the case of CD8 T cell rejection, the priority is to eliminate the expression of HLA class I molecules on CAR T cells. Deletion of the conserved gene beta 2-microglobulin completely removes surface expression of HLA class I. 86 However, while immunogenic recognition by CD8 T cells will be reduced by this approach, the complete loss of HLA class I molecules will increase the risk of recognition of the allogeneic CAR T cells by NK cells, the so called 'missing self' mechanism. NK cell mediated destruction of HLA-edited T cells has been proposed to be prevented by expression on CAR T cells of nonpolymorphic HLA molecules such as HLA-E that will bind inhibitory receptors on NK cells. 87 88,89 Currently it is unclear if some of the above will be dominant, or if all modifications will be needed to avoid rejection.
The different possibilities of T cell engineering open new perspectives in the field of immune cell therapy. However, it remains to be elucidated whether the production of universal cells that completely evade the host immune system may also induce safety issues, since the immune system will not be able to recognize transformed cells in case of an oncogenic event. The initial screening of the administered allogeneic T cells for oncogenic mutations would therefore be of crucial importance in the context of extensive editing of the T cell genome. Given that late oncogenic events can still occur, and because rare oncogenic cells may have not been detected during the process, an optimized suicide system may be advantageous in modified T cells to allow their efficient elimination. 90 Linking the suicide gene to a cell-division gene, as described by Liang et al, 91 may be particularly useful in this context.

| CON CLUS IONS
Engineered adoptive cell immunotherapies are quickly showing promise for patients. As logistical and scientific insights into the manufacturing of these therapies evolve, patients could greatly benefit from readily available off-the-shelf cell therapies tailored to the patient and their disease. By engineering T cell immune identities, off-the-shelf allogeneic T cell therapies currently in clinical development (Table 1)

E TH I C A L S TATEM ENT
The authors confirm that the ethical policies of the journal, as noted on the journal's author guidelines page, have been adhered to. No ethical approval was required as this is a review article with no original research data.