Allo-CARs have thus far largely been used as donor lymphocyte infusions after allogeneic hematopoietic cell transplantation (allo-HCT) to either treat or mitigate relapse. Smith et al. previously reviewed clinical outcomes in patients who received either donor-derived allo-CARs or recipient-derived “pseudo-allo” CARs in the post-transplant setting (14). Among these trials, Kochenderfer et al. (15), Brudno et al. (16), Kebriaei et al. (17), and Lee et al. (18) enrolled patients with r/r DLBCL. Responses were mixed in these DLBCL cohorts: 3 patients achieved CR, 5 remained with stable disease, and 3 developed progressive disease. The time to cell production also differed between groups; Kochenderfer and Brudno et al. produced cells in 8 days (using retroviral transduction methods) (15, 16), Kebriaei et al. in 28 days (using a transposon/transposase system) (17), and Lee et al. in 11 days (using a simplified retroviral transduction system) (18, 19). Notably, no patients in these studies developed new GVHD, and 1 patient who had preexisting mild GVHD continued to have GVHD post-allo-CAR ( Table 1 ) (16). Preclinical data suggests that the presence of CD19+ targets is protective against GVHD as tonic signaling through both the CAR and T cell receptor (TCR) promotes an exhaustion phenotype and subsequent apoptosis of alloreactive, GVHD-inducing cells (22). Despite this loss of alloreactive T cells, bulk CAR T cell populations contain other cells capable of attacking CD19+ tumors while sparing host tissue, which may explain why early trials of allo-CARs have not reported significant rates of GVHD (16, 22).

More recently, donor-derived allo-CARs have been shown to induce responses in patients who previously received allo-HCT and auto-CAR therapy, raising the possibility of a graft-vs-lymphoma effect induced by the donor-derived allo-CARs (20). These early studies demonstrated the viability of allo-CARs as modified donor lymphocyte infusions for patients with otherwise refractory disease.

Allo-CARs have been successfully generated from various cell types ( Table 2 ), including those belonging to the innate immune system. Natural killer (NK) cells are a promising CAR candidate as they have intrinsic killing capabilities, do not cause GVHD, and are negatively regulated by major histocompatibility complex (MHC) class I molecules present on normal cells (39), which may limit off-target toxicities. They can also be derived from a variety of sources, including peripheral blood, NK cell lines, memory-like NK cells, human embryonic stem cells, CD34+ hematopoietic progenitor cells and induced pluripotent stem cells (iPSCs) (23). NK cells transduced with CARs (CAR-NKs) have demonstrated clinical efficacy; a Phase 1/2 trial of cord blood-derived CAR-NKs included 2 patients with r/r DLBCL, one of whom achieved a minimal residual disease (MRD)-negative CR for 15 months at data cutoff (21). No patients experienced cytokine release syndrome (CRS), immune effector cell-associated neurotoxicity (ICANS), or GVHD in this trial (21), providing proof-of-concept for decreased toxicity while maintaining anti-cancer effects. However, allogeneic NK cells are currently limited by lack of persistence in the absence of exogenous cytokine stimulation (40), decreased trafficking to tumor sites (41) and dysfunction induced by the immunosuppressive tumor microenvironment (39, 42, 43). Strategies to circumvent these obstacles include engineering CAR-NKs to express cytokine transgenes to improve persistence and chemokine receptors to improve homing capabilities (39). Either deleting or blocking checkpoint inhibitors on NK cells also may enhance their function.

Invariant NK/T (iNKT) cells have recently shown promise as another platform for CAR engineering. These relatively rare cells co-express T and NK cell markers with a semi-invariant TCR that recognizes antigens presented on CD1d, an MHC class I-like molecule (24). CAR-modified iNKT cells were first developed to target the GD2 ganglioside on neuroblastoma in murine models, where they demonstrated antitumor activity without causing GVHD (44). Further murine and human studies have shown that iNKT cells actually suppress GVHD, likely through expansion of T regulatory cells (24). When transduced with CD19 CARs, iNKT cells showed activity against B cell malignancies through both CAR-CD19 and CD1d-invariant TCR interactions, resulting in elimination of tumors in mice (including those with intracranial and relapsed lymphomas) (25). Subsequently, in a murine model of CD19+ lymphoma, Simonetta et al. showed that allogeneic CAR-iNKT cells exerted anti-tumor activity through cross-priming of CD8+ T cells, and were more effective at tumor control than conventional CAR T cells in the presence of host lymphocytes (26). This cross-talk between CAR-iNKT cells and CD8 T cells appears essential to CAR-iNKT function, which may be impaired by the typical lymphodepleting chemotherapy given prior to conventional auto-CAR therapy. Nonetheless, iNKT cells do not cause GVHD, can be easily expanded ex vivo, and utilize dual targeting mechanisms to eliminate tumors. These characteristics make iNKT cells excellent candidates for future CAR development strategies.

Like iNKT cells, γδ T cells participate in both innate and adaptive immunity and have cytotoxic mechanisms independent of the TCR-MHC interaction. γδ T cells recognize peptide antigens in an MHC-unrestricted fashion and can express ligands such as NK receptors and toll-like receptors to identify and destroy target cells damaged by infection, malignancy, or other stressors (27). Their multimodal approach to cell destruction makes γδ T cells good candidates for CAR engineering. An anti-CD20 allo-CAR made from γδ T cells has demonstrated manufacturing feasibility and improved cytotoxic activity over conventional CAR-T cells in a preclinical model of B-cell lymphoma (28), which has subsequently led to a Phase 1 trial in humans (NCT04735471).

The current dearth of easily available, antigen-specific T-cells could also be addressed by using iPSCs as CAR platforms as they have nearly unlimited replication potential. Themeli et al. demonstrated the feasibility of this approach by generating iPSC clones from peripheral blood lymphocytes using dedifferentiation methods, transducing a clone with a CD19-targeted 2^nd-generation CAR, and directing the differentiation of the CAR-iPSC clone towards T-lymphoid lineages (31). This method produced functional iPSC-derived CAR T-cells that expanded ex vivo and induced complete tumor regression in murine models, albeit more slowly than standard αβ or γδ CAR T cells (31). Despite retaining the αβ-TCR, the iPSC-derived CARs also exhibited phenotypic and functional characteristics typical of γδ T cells (31). The main limiting factor for iPSC-derived allo-CARs is the potential for GVHD due to HLA mismatches and endogenous TCR expression (32). Genome editing techniques could be employed to remove these proteins from allo-CARs, but this would potentially make them susceptible to NK cell-mediated destruction due to lack of MHC I expression (32).

Epstein-Barr virus accounts for 20% of DLBCL cases (33), making this virus an important therapeutic target. EBV-driven lymphoproliferative diseases (EBV-LPDs) are categorized as a DLBCL subtype that develops after allo-HCT and solid organ transplantation. EBV-LPDs have been shown to be amenable to T-cell therapies targeting EBV latent membrane proteins (LMPs). In one early study of autologous cytotoxic T cells, three out of 4 patients with r/r DLBCL achieved CR (34). Subsequently, donor-derived T-cells engineered to target LMPs were evaluated in patients after allo-HCT for r/r lymphoma and chronic lymphocytic leukemia (CLL). Three of the 19 patients in this study had r/r DLBCL and an additional 2 had EBV-LPDs. All 5 remained in CR for 8 weeks post-infusion, and 3 of these patients maintained a CR for over 3 years (33). Reactivation of GVHD was seen in 2 of the patients with DLBCL, one of whom died of allo-HCT-related complications 6 months post-infusion (33). Patients with B-cell diseases had an 80% overall survival at 2 years, demonstrating the safety and effectiveness of engineered allogeneic T-cells as adjuvant therapy for virally driven lymphomas. More recently, the EBV-specific allogeneic T-cell therapy tabelecleucel (which employs TCR-targeting and HLA restrictions to specifically attack EBV-transformed cells) (45) induced CRs in 32/76 patients (42%) with r/r EBV+ LPDs (46). The overall response rate (complete and partial remissions) in this trial was 63%, with 91% overall survival at 1 year and 86% at 2 years without any GVHD, CRS, neurotoxicity, or graft rejection (46). EBV-specific allo-CAR T-cells have also been developed and tested in the consolidation setting after AHCT for non-Hodgkin lymphomas. A Phase 1 trial demonstrated 100% CR rates in 4 such patients, with a median overall survival of 30.8 months and no CRS or neurotoxicity (35). This encouraging early data suggests that EBV+ lymphomas are attractive disease candidates for future allo-CAR trials.

In studies of DLBCL relapses after axi-cel, suboptimal T-cell fitness (measured by prolonged cell doubling time) was associated with treatment resistance (47). Prior chemotherapy exposure likely plays a role in decreasing T-cell fitness, as laboratory evidence suggests that cyclophosphamide, doxorubicin, and cytarabine impact mitochondrial function and induce lingering deficits that impair ex vivo stimulation (48). Studies of T-cell expansion in children undergoing chemotherapy for various malignancies demonstrated that T-cells from children with non-Hodgkin’s lymphoma had among the lowest post-chemotherapy expansion potentials (36). Another factor that impacts T-cell fitness is the phenotype of the cells comprising the CAR product. Preclinical evidence suggests that less-differentiated (CD62L+) T-cell populations are capable of enhanced engraftment, expansion, and persistence relative to more differentiated effector subsets (37). CCR7⁺CD45RA⁺ cells have a stem-like central memory (SCM) phenotype, and a decreased frequency of these cells relative to tumor burden has been associated with increased CAR-T doubling times and failure to achieve durable responses in patients with DLBCL (47). Enriching for SCM T-cells in CAR products may thus improve response rates. Further data supporting this hypothesis comes from transcriptomic analysis of CAR-T cells used for chronic lymphocytic leukemia, which identified a subpopulation of CD27⁺CD45RO^low CAR-T cells that were seen more consistently in patients with complete responses (38). These cells were then shown to proliferate extensively ex vivo while maintaining a less differentiated state (38). A Phase 1 trial using central memory-enriched CAR T-cells (albeit after AHCT) for r/r DLBCL demonstrated the safety and feasibility of this approach, and 5 of 11 patients maintained their best responses (either PR or CR) at 1 year post-therapy (49). CAR products engineered with defined 1:1 CD4:CD8 ratios have also shown improved antitumor activity relative to unselected T-cells in mouse models (50) and have demonstrated efficacy in patients with r/r DLBCL (51).

As alluded to in prior sections, the key obstacles that allo-CARs will need to overcome to become more broadly used include GVHD, allo-CAR rejection (i.e., alloimmunization), and potentially decreased long-term persistence. Though not specific to allo-CARs, resistance mechanisms such as antigen escape will also need to be addressed to maximize allo-CAR effectiveness.

GVHD and alloimmunization both reflect the challenge of addressing HLA-incompatibility between the allo-CAR donors and patients. Overall, early clinical data with allo-CARs revealed encouragingly low rates of GVHD (15, 16, 18), though innovations in allo-CAR engineering are aiming to lower that risk even further. GVHD is thought to be mediated through the interaction of donor αβ-TCR with host MHC complexes; therefore, strategies to abrogate GVHD thus far have involved elimination of the αβ-TCR or the use of non-alloreactive cells (such as NK, iPSCs, and virus-specific T-cells discussed above) (52). Several genome editing techniques are capable of creating insertions or deletions at the site of the TCR alpha constant (TRAC) locus, causing disruption of this gene and the inability to form the TCR complex (53, 54). Another technique is to use a recombinant adeno-associated virus or lentivirus to insert the CAR transgene directly into the TRAC locus, thereby replacing the TCR with a CAR construct and enabling homogenous and uniform levels of CAR expression (54). Generating allo-CARs from less-differentiated T-cell subsets (e.g., SCM T cells) may also reduce GVHD risk as these cells have limited TCR specificity (55, 56).

Alloimmunization is expected to be a significant issue with allo-CARs. Though lymphodepletion is necessary for expansion of infused cells, immune recovery with time is expected to result in rejection of HLA-mismatched cells. Preexisting donor-specific antibodies (e.g., from prior pregnancies or blood product transfusions) may also mediate rejection (57). Alemtuzumab has been used to further deplete host T-cells prior to allo-CAR therapy, though this increases the risk of subsequent cytopenias and infections (58). One potential way to minimize the risk of rejection is to bank T cells that match key HLA alleles (HLA-A, HLA-B, and HLA-DR) with most of the population, as has been done with solid organ and umbilical cord transplantation.

CAR T-cell persistence is a multifactorial issue that has yet to be fully explained. A prior review of auto-CAR resistance mechanisms (59) identified several contributors, including initial T cell quality (18, 60), phenotype (38, 61–63), and choice of costimulatory domain (2, 3). In a preclinical model, T cell quality was assessed by comparing the viability and phenotypes of CAR T cells created from young (6-12 weeks) and aged (72 weeks) mice. T cells derived from the older mice were more cytotoxic but were shorter-lived and had a less memory-like phenotype than cells from younger mice (64). Phenotype was also studied in a retrospective analysis of a Phase 1 trial evaluating CAR T cells for leukemia and lymphoma in children and young adults. In this study, CAR T cells that initially had greater expression of PD-1 and LAG-3 and decreased expression of TNF-α were associated with dysfunctional responses and decreased cytokine production in response to stimulation (65). Dysfunctional response was a clinically meaningful outcome as it was defined as either the inability to achieve remission or disease progression after remission but before CAR engraftment (65). Finally, costimulatory domains also play a significant role in CAR persistence; in the landmark study of tisa-cel in children and young adults with B-cell leukemias, 4-1BB-based auto-CARs persisted for a median of 168 days compared to 30 days for CD28-based auto-CARs (66). Whether CAR persistence equates to improved remission duration in r/r DLBCL is unclear and needs further investigation.

Antigen escape is a common CAR resistance mechanism observed with leukemias (59), though this phenomenon has also been reported in lymphomas. One case report demonstrated sequential loss of CD19 and CD22 over time in a pediatric patient with DLBCL treated with auto-CARs targeting those antigens (67). The patient also had a homozygous TP53 mutation, which the authors hypothesized may have driven clonal expansion of CD19- and CD22-negative cells (67). A similar case was previously reported demonstrating absent surface CD19 expression (with conserved cytoplasmic expression) in primary mediastinal B-cell lymphoma (PMBCL) treated with auto-CARs (68). In this case, missense mutations in exon 4 of the CD19 gene and deficiencies in DNA mismatch repair proteins were identified in PMBCL clones that were seen pre-treatment and ultimately became dominant post-treatment (68). More recently, CD19 exon 3 point mutations in a case of high grade B-cell lymphoma were found to confer resistance to certain subsets of auto-CAR populations while retaining sensitivity to others (69). These cases are illustrative of how mutations impact loss/alteration of CAR surface targets, presenting an important area for further research.

Numerous methods are under investigation to overcome antigen escape. Multi-antigen-targeted CARs (e.g., bispecific CD19/22) have been shown to induce durable remissions in CD19-low or negative disease previously treated with anti-CD19 auto-CARs (70). Studies of Bryostatin1, which can improve CAR functionality and durability by increasing expression of CD22 on DLBCL cell lines, have demonstrated the feasibility of modulating target antigen density in lymphomas (71, 72). Modifying the CAR construct can also increase affinity for lower antigen density tumors. Majzner et al. demonstrated that CD28 endodomain-containing CARs are better able to kill and proliferate in response to low antigen density cells in vitro compared to their counterparts with 4–1BB endodomains (73). Alterations to the hinge transmembrane domains or CAR zeta chains also can improve recognition of low density antigens (73). These findings will undoubtedly influence how allo-CARs will be engineered in the future.