Next‐generation chimeric antigen receptors for T‐ and natural killer‐cell therapies against cancer

Adoptive cellular therapy using chimeric antigen receptor (CAR) T cells has led to a paradigm shift in the treatment of various hematologic malignancies. However, the broad application of this approach for myeloid malignancies and solid cancers has been limited by the paucity and heterogeneity of target antigen expression, and lack of bona fide tumor‐specific antigens that can be targeted without cross‐reactivity against normal tissues. This may lead to unwanted on‐target off‐tumor toxicities that could undermine the desired antitumor effect. Recent advances in synthetic biology and genetic engineering have enabled reprogramming of immune effector cells to enhance their selectivity toward tumors, thus mitigating on‐target off‐tumor adverse effects. In this review, we outline the current strategies being explored to improve CAR selectivity toward tumor cells with a focus on natural killer (NK) cells, and the progress made in translating these strategies to the clinic.


| INTRODUC TI ON
In recent decades, the field of adoptive cellular therapy has undergone substantial developments in various clinical settings. 13][4][5][6] The advent of CAR engineering resulted from extensive research endeavors primarily focused on developing "living drugs" capable of specifically targeting and eliminating tumors.The initial work in the field of adoptive cell therapy consisted of isolating tumor-infiltrating lymphocytes (TILs) with natural specificity against mutated proteins on tumor cells, expanding them ex vivo and reinfusing them to patients.While this strategy showed promising responses in certain types of cancers such as melanoma, 7 its broad applicability has been limited by the complexity of harvesting TILs and their poor expansion in vitro. 8The emergence of sophisticated genetic engineering techniques allowed the modification of T cells to express a T-cell receptor (TCR) against certain tumor antigens.As such, T cells have been engineered to recognize a variety of tumor antigens such as NY-ESO-1, 9,10 PRAME, 11,12 and selected MAGE-A family members, 10,13 to treat patients with various cancers such as sarcoma and multiple myeloma. 9,14,15 TCR T-cell therapy is restricted by specific human leukocyte antigens (HLA) alleles, and as cancer cells commonly evade TCR recognition by downregulation of their major histocompatibility complex (MHC) proteins, the emergence of CAR-based therapy constituted a major advancement in the field of adoptive cell therapy due to its MHC-independent mechanism of antigen recognition. 16 fact, CAR T-cell therapy has resulted in remarkable responses in some hematologic cancers and its application has quickly evolved to its current Food and Drug Administration (FDA) approval for Blymphoid malignancies and multiple myeloma.
However, challenges associated with the autologous nature of these products have limited their widespread implementation.The manufacturing process for CAR T-cell therapies is cumbersome and costly, resulting in lengthy collection to administration times, thus posing a challenge for patients who, due to rapidly progressing disease, are in urgent need of treatment. 17Moreover, patient-derived T cells may be limited in number or compromised in function, especially in patients who have been heavily treated prior to CAR T administration. 18,19These limitations sparked a growing interest in alternative allogeneic cell sources that are off-the-shelf and available for pointof-care use.One approach focuses on developing allogeneic CAR T cells through genome editing to abrogate the endogenous expression of αβTCR and/or MHC class I complexes, thus eliminating Tcell alloreactivity and reducing the risk of graft-versus-host disease (GvHD). 20,21These "universal" CAR T cells can be manufactured in large scale from healthy donor sources and administered to patients more safely.The first off-the-shelf CAR T-cell product to be investigated in clinical trials was UCART19 for the treatment of patients with B-cell acute lymphoblastic leukemia (ALL). 22,23While early results with allogeneic CAR T cells are promising, challenges remain, including allo-rejection of the infused product by the recipient immune system, technical difficulties with achieving 100% editing efficiency and thus the risk of GvHD, and potential risks related to gene editing such as off-target editing, genotoxicity, and acquisition of chromosomal abnormalities. 24Alternative immune effector cells, such as natural killer (NK) cells 25 and invariant NK T (iNKT)/NKT cells, 26 are actively being explored as vehicles for CAR engineering due to their high cytotoxic potential and low risk of GvHD in the allogeneic setting.Of these, NK cells have been one of the most extensively explored alternative immune cells for adoptive cell therapy.
As with any targeted approach, the broad application of CAR T-cell and CAR NK-cell therapies has been limited by the paucity of targetable tumor-specific antigens (TSAs).Indeed, many tumor antigens are either inherently expressed at low levels or eventually downregulated as a mechanism of tumor escape from the targeted CAR cell therapy.8][29][30] To overcome these challenges, extensive translational research has been conducted to design the next generation of CAR immune cells with high potency and tumor selectivity.
In this review, we highlight the current approaches to CAR engineering, with a focus on NK cells, including strategies to enhance their on-target antitumor selectivity, and discuss the advances achieved in translating these innovations to the clinic.

| THE MODUL AR DE S I G N OF A C AR
Antigen recognition by a CAR is achieved via its extracellular domain, which conventionally employs the binding domain of a single-chain variable fragment (scFv), derived from a monoclonal antibody (mAb), to specifically recognize a tumor antigen.The CAR then transmits an activation signal to the carrier cell, resulting in a potent and targeted cytotoxicity.The CAR molecule consists of three parts: an extracellular domain, a transmembrane domain, and an intracellular domain. 6e transmembrane domain is linked by a hinge region to an extracellular domain and is commonly derived from IgG, CD8, or CD28.than one costimulatory domain in addition to CD3ζ, and (iv) fourth generation incorporating a transgenic protein such as a cytokine with constitutive or inducible expression (referred to as TRUCKs for "T cells redirected for antigen unrestricted cytokine-initiated killing"). 31,32The CAR transgene is introduced into the effector cells through DNA plasmid transfection or viral-based transduction, leading to CAR expression on the plasma membrane.

| CURRENT LIMITATI ON S OF C AR T-CELL THER APIE S
4][35][36][37][38][39][40][41][42][43] Despite the remarkable clinical success of CAR T-cell therapies in hematologic malignancies, their expanded clinical use has been limited by several factors.In addition to the arduous, time-consuming, and costly manufacturing process of autologous products, CAR T-cell therapy has a unique toxicity profile, characterized by cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS). 19,44While more research is needed to determine the risk factors predisposing patients to CAR T-cell toxicity, certain associated factors have been established. 44,45Other major limitations of CAR T-cell therapy pertain to the target antigens themselves, including on-target off-tumor toxicity due to the expression of the target antigens on normal tissues.Hence, identifying an ideal target antigen with the right balance of sensitivity and selectivity for CARbased cell therapy has been a major challenge for cancers beyond lymphoid malignancies.
CAR NK cells offer several advantages, including a broad range of donor cell source for manufacturing, lower risk of toxicities, and multiple mechanisms of cytotoxicity, making them a promising approach for cell-based immunotherapy.Indeed, CAR NK therapy circumvents the requirement for autologous NK cells due to their reduced risk of alloreactivity and GvHD. 46Thus, existing NK cellsources such as NK-92 cell lines, umbilical cord blood (CB), peripheral blood (PB), and induced pluripotent stem cells (iPSCs) can be leveraged for the large-scale production of "off-the-shelf" CAR NK

| NK-CELL B I OLOGY
NK cells are part of the innate immune system and unlike T cells, exert their cytotoxicity in an HLA-independent manner and without the need for prior priming. 480][51] NK cells are characterized by the distinct expression of CD56 and the absence of TCR and CD3 expression. 52They are broadly categorized based on the relative expression of the cell surface receptors CD56 and CD16 into two classes: CD56 bright CD16 low/− NK cells, characterized by immunomodulatory and cytokine-producing properties, and CD56 dim CD16 + NK cells that are generally cytotoxic. 53Notably, high-parameter cytometry and single-cell proteo-genomics have revealed much greater phenotypic and functional heterogeneity of NK cells than previously appreciated and have enabled the identification of diverse NK-cell subpopulations extending far beyond the two known subsets.The CD16 receptor on NK cells binds to the Fc receptor on antibody-coated cells to exert potent antibody-dependent cell cytotoxicity (ADCC). 52Additionally, NK cells can be activated through an intricate interplay between activating and inhibitory germline-encoded receptors present on their cell surface. 54These receptors convey signals of either activation (via immunoreceptor tyrosine-based activation motifs or ITAMs) or inhibition (via immunoreceptor tyrosine-based inhibition motifs or ITIMs). 55Notably, cancer cells downregulate MHC class I expression or upregulate stress-induced molecules like MICA/MICB, thereby disengaging inhibitory killer Ig-like receptors (KIRs; the "missing-self" phenomenon) or engaging activating receptors such as NKG2D, respectively.Consequently, NK cells become activated, directing their cytotoxic efforts toward the target cells in a finely orchestrated manner. 56Furthermore, NK cells interact with other immune cells through the production of cytokines and chemokines.As such, they have been shown to impact the function of T and B lymphocytes, dendritic cells (DCs), macrophages, and neutrophils. 57These diverse roles highlight the complex biological functions of NK cells and underscore their promise for immunotherapy.
While our classical comprehension of NK cells as part of the innate immune system depicts them as fast-acting and short-lived without antigen specificity, exciting discoveries in recent years have defied this convention.In fact, multiple research groups have reported that under specific circumstances, NK cells undergo clonal expansion in response to antigen stimulation to give rise to long-lasting memory cells. 58Adaptive NK cells were first studied in murine models of cytomegalovirus (MCMV) infection, revealing that NK cells bearing virus-specific receptors such as Ly49H, Ly49L, and NKR-P1A exhibit rapid proliferation, cytokine production, and clonal expansion upon MCMV reactivation, reminiscent of the memory-like properties typically attributed to adaptive immune cells. 59Moreover, Wayne Yokoyama's group showed that NK cells in mice acquire memory-like features in response to stimulation with inflammatory cytokines. 60milarly, human adaptive NK cells have been identified 58 and studied most extensively in the context of human CMV infection. 61,62tokine-induced memory of human NK cells has also been investigated.In fact, human NK cells that were activated with IL-12, IL-15, and IL-18 and rested for 1-3 weeks, were shown to exhibit robust antitumor response characterized by augmented interferon (IFN)γ production and proliferation in response to cytokines or exposure to K562 leukemia cells. 63,64Numerous other research groups have reported analogous memory-like functionality of NK cells in various immunological contexts, challenging the classical delineation of innate and adaptive immunity. 64,65

| CLINI C AL PROG RE SS OF C AR NK-CELL THER APY
CAR NK-cell immunotherapy is emerging as an attractive therapeutic option for cancer, with encouraging clinical responses reported in multiple phase I/II trials (Table 1).Given the promising clinical activity and safety of NK-cell therapy, it is imperative to define the best NK-cell source and CAR design for adoptive cell therapy.

| NK-cell sources for clinical use
Current NK-cell therapy platforms rely on various sources of cells for therapeutic applications.These include PB, CB, cell lines, hematopoietic stem and progenitor cells (HSPCs), and iPSCs. 66Despite their potential for generating scalable and clinically significant NKcell doses for CAR NK-cell therapy, each of these sources has distinct characteristics, presenting both unique advantages and challenges.
PB-derived NK cells can be obtained through apheresis from healthy donors and have been extensively studied and widely used in clinical trials of CAR NK-cell therapy (e.g., NCT00995137, NCT01974479, NCT05020678, NCT04623944).CB is another valuable source of NK cells for clinical use.The ease of collecting CB units and the ability to cryopreserve them offer unique advantages of this source for NK immunotherapy.Our group has focused on the use of CB-NK cells for CAR engineering, demonstrating the ability to generate over a 100 dose of CAR NK cells from a single CB unit. 67,68-92 is an immortalized NK lymphoma cell line that has received Investigational New Drug approval by the FDA for clinical  testing.NK-92 cells offer a homogeneous and abundant cell source 69 for CAR engineering; however, their cancerous origin necessitates irradiation prior to administration that could limit their in vivo proliferation and persistence.Additionally, without extra engineering steps, cell lines like NK-92 may lack certain functional capabilities, such as the ability to mediate ADCC due to the absence of CD16 expression.70 HSPCs and iPSCs present exciting prospects for NK-cell therapy, as they are characterized by clonal growth and high expansion capacity.Differentiation of these stem cells into NK cells allows for the manufacturing of large numbers of homogeneous NK-cell products.71 However, challenges exist, such as concerns with epigenetic memory of their cellular origin with iPSC-derived NK cells.72 Ongoing research is focused on optimizing the use of these cell sources and determining their efficacy and safety in clinical applications.

| Clinical experience with CAR NK-cell therapy
In recent years, CAR NK-cell therapy has emerged as a promising approach for the immunotherapy of cancer.The clinical safety of administering CAR NK cells was demonstrated in a Phase I study in 2018 by a group based in China (NCT02944162). 73The trial used NK-92 cells engineered to express a third-generation CD33-directed CAR construct incorporating CD28 and 4-1BB co-stimulatory domains for the treatment of acute myeloid leukemia (AML). 73While the study reported an excellent safety profile in three patients with relapsed/refractory disease, no durable responses were achieved. 73is was mainly attributed to the limited in vivo persistence of the irradiated CAR NK-92 cells. 73r group investigated the use of cytokine armoring to enhance the in vivo persistence and proliferation of CAR NK cells. 67In a firstin-human phase I/II clinical trial, we reported the excellent safety and promising activity of IL-15 armored CB-derived CAR19 NK cells in patients with relapsed/refractory B-lymphoid malignancies (NCT03056339). 67Notably, CAR19 NK cells were detectable up to 1 year post-infusion, and patients who responded to treatment exhibited substantially higher blood peak copy numbers of CAR19 NK cells. 67These results support the incorporation of cytokine armoring to enhance the persistence and clinical activity of NK cells.
iPSC-derived NK cells offer another attractive platform for CAR engineering.FT596 is an engineered iPSC-derived NK-cell product that incorporates three genes encoding: (i) a high-affinity noncleavable Fc receptor (hnCD16) that has been modified to include the 158V variant in combination with an S197P amino acid substitution to prevent cleavage by ADAM17, 74 (ii) a membrane-bound IL-15/ IL-15 receptor (IL-15R) fusion protein, and (iii) an anti-CD19 CAR. 75terim clinical results reported as of June 2021 demonstrated the safety of FT596 with no dose-limiting toxicities. 76A total of 20 patients underwent dose escalation treatment, with 10 patients receiving FT596 alone (Regimen A) and 10 patients receiving FT596 cells combined with rituximab (Regimen B). 76 Of the evaluable patients, the overall response rate (ORR) following the first treatment cycle was five of eight patients (62%) in Regimen A and four of nine patients (44%) in Regimen B. 76 Longer follow-up data will help elucidate the durability of the response and the overall efficacy of this platform.
Similarly, preliminary results from a phase I study (NCT05020678) of off-the-shelf allogeneic CAR19-engineered PB-NK cells expressing a membrane-bound form of IL-15 (NKX019) were recently reported in a press release. 77This therapy achieved a complete response rate of 70% (seven of 10 patients) in patients with relapsed/ refractory non-Hodgkin lymphoma and durable responses of greater than 6 months in multiple patients. 77These encouraging results collectively support the promise of NK cells as alternative immune effectors for CAR cell therapy, with over 45 trials currently registered on clini catri als.gov (Table 1).

| THE APPLI C ATI ON OF C AR ENG INEERING FOR SOLID TUMOR S
A major challenge in the translation of CAR T-cell or CAR NK-cell therapies from hematologic malignancies to solid tumors is the identification of target antigens that are widely and homogeneously expressed at high levels on tumor cells while having virtually no expression on normal tissues.TSAs, such as EGFRvIII in glioblastoma, 78 are considered as ideal targets due to their exclusive expression on tumor cells.However, they also present unique challenges primarily due to their heterogeneous levels of expression on tumor cells.1][92] Moreover, ICANS toxicity in CAR19 T cell-treated patients may represent an on-target off-tumor side-effect due to CAR-mediated cytotoxicity targeting low levels of CD19 expressed on brain mural cells. 93milarly, cross-reactivity of CAR-modified immune cells against BCMA-expressing neurons and astrocytes is thought to contribute to the neurocognitive and hypokinetic movement disorders seen after infusion of BCMA CAR T cells. 94Treatment of patients with CAR T cells recognizing carbonic anhydrase 9 (CA-IX), 95,96 HER-2, 80 ERBB-2, 97 or MSLN 81 has been associated with severe side-effects including acute respiratory failure and organ damage, most likely from the expression of these antigens at differing levels on epithelial cells in the lungs and the bile ducts, respectively.
On-target off-tumor toxicity has also been associated with CARmediated antigen recognition of a mimotope, which mimics the structure of the targeted epitope but belongs to a distinct antigen expressed on normal cells. 98Notably, preclinical murine models may not adequately predict the off-tumor antigen binding and toxicity potential of a given CAR, and more research is needed to develop dedicated in vivo models for the study of on-target off-tumor toxicity. 99Some strategies have been devised to curb the deleterious on-target off-tumor toxicities in patients treated with CAR immune therapy, such as the administration of immunosuppressive regimens or activation of safety switches including the inducible caspase nine suicide (iC9) gene system as used in our iC9/CAR19/IL-15 NK cell clinical trial. 67[102][103] Lastly, two important CAR-mediated on-target off-tumor ef- to trogocytosis and reduced CD19 expression on tumor cells, associated with a greater likelihood of relapse. 109Taken together, these data indicate the need for innovative and rationally designed CARs to increase tumor specificity while preventing on-target off-tumor toxicities.

| Exploration of novel targeting approaches
To improve CAR-mediated on-target antitumor activity, research, and clinical investigations are now focusing on adapting CAR technology to other cancers, while also enhancing potency and/or safety (Tables 1 and 2).As scFvs have perceivable limitations such as promoting self-aggregation of the CAR molecule which can in turn lead to premature CAR activation and exhaustion of the transduced immune effector (Figure 1A), [112][113][114] alternative binding domains have been explored.These include nanobodies (NBs), recombinant antigen-specific scFvs derived from the heavy chain (V HH ) of a mAb.
An NB has a similar antigen-binding affinity to a conventional or fulllength mAb, but due to its smaller size, has greater solubility and more stable physiochemical properties (Figure 1B). 115Using NBs to generate a CAR could be advantageous due to their enhanced stability, making them more amenable to additional modifications such as multi-targeting. 116NB-CAR NK-92 cells targeting CD38 demonstrated targeted cytotoxicity against patient-derived CD38expressing multiple myeloma cells, but the in vivo efficacy of this approach remains to be validated. 117  to target the corresponding receptor (Figure 1D).Natural receptors and ligands used as targeting domains in CAR NK cells include NKG2D (recognizing stress ligands such as MICA, MICB, and ULBP), 119 DNAM-1 (targeting PVR/CD155 and Nectin-2/CD112 ligands), 120 PD-1 (targeting PD-L1), 121 and CD27 (targeting CD70; NCT05092451, NCT05703854), among others (Table 1).
Approaches to fine-tune the binding affinity the CAR extracellular domain in order to preserve or enhance the recognition of their cognate antigen while minimizing off-tumor recognition have also been explored.An example was the use of an optimized-affinity anti-CD38 CAR capable of targeting cancer cells while sparing CD38 low-expressing normal cell populations (Figure 2A). 122In this study, NK cells transduced with this optimized-affinity anti-CD38 CAR successfully killed primary AML blasts with minimal fratricide against CD38 low-expressing NK cells. 123Strategies to optimize the binding of CARs to low antigen expressing targets have also been explored.
Recently, innovative synapse-tuned CAR NK cells have been devised by incorporating a PDZ binding motif (PDZbm), important for cell polarization and synapse formation, within the CAR construct (Figure 2B). 124This modification enhanced the strength of the synapse and the polarization of CAR T and CAR NK cells, resulting in improved effector cell function both in vitro and in vivo. 124her emerging approaches to broaden the scope of antigen binding include the use of vaccines to promote antigen display on   DCs and thus increase the sensitivity of CAR cells to tumors expressing antigens at low levels (CARVac; Figure 2C).In one study, an antigen known as claudin 6 (CLDN6), which is not expressed in normal tissues yet is often present at low levels in tumors, 125 was targeted by a CARVac. 126The transient expression of this antigen on DCs was amplified by CLDN6 mRNA vaccine, leading to CAR T-cell activation and expansion despite low CLDN6 expression on tumor cells. 126This strategy has been investigated in a clinical trial for the treatment of relapsed/refractory testicular, ovarian, and endometrial cancer as well as soft-tissue sarcoma (NCT04503278), with an ORR of 43% and disease control rate of 86% (CT002 -BNT211; Table 2).It would be very interesting to apply the CARVac strategy with NK cells, given the strong cross talk between NK cells and DCs. 1279][130] In a recent study, NK-92 cells were engineered to express both a TCR specific to the E7 protein of HPV16 and a CAR targeting TROP2. 130This combination strategy led to a significant increase in NK-cell activation and antitumor activity against HPV-driven cancers. 130Lastly, in an innovative effort to target oncogenic drivers by CAR therapy, an intracellular unmutated oncogenic driver peptide (QYNPIRTTF) that is commonly expressed in neuroblastoma and presented by HLA was identified and targeted using a peptide-centric CAR that recognizes this TSA (Figure 2E). 131

| Dual-targeting strategies to improve tumor recognition and reduce toxicity
Dual antigen-targeting CARs have been investigated to overcome antigen escape, and to refocus CAR specificity toward tumor cells and away from normal cells.Targeting two or more antigens with CAR T or CAR NK cells can be achieved in different ways: (i) co-administration of CAR products with different antigen specificities, (ii) engineering a cell product with two or more CAR viral vectors, (iii) introduction of a bicistronic construct encoding for two CARs, or (iv) introduction of a tandem CAR with two different scFvs linked to the same signaling endodomain, whereby antigen recognition by either scFv leads to CAR signal transduction (Figure 2F). 132,133Dual-targeting CAR T cells have been used in clinical trials for patients with B-cell malignancies (targeting CD19/ CD20 134 and CD19/CD22) 135 and in multiple myeloma targeting BCMA and CD38 (ChiCTR1800018143 136 ; Table 2).However, the net clinical benefit of dual-targeting strategies requires further evaluation, with some suggestion that dual-targeting CAR T cells may not sustain their antigen specificity and potency against both antigens. 135al-targeting strategies have also been explored with NK cells (Table 1).Notably, in preclinical studies with the previously discussed iPSC NK cell product FT596, dual-targeting against lymphoma was achieved by engineering the cells to express a CAR against CD19 and by combining them with a CD20 mAb that mediated ADCC by binding to hnCD16. 137Similarly, FT596 cells engineered to express anti-BCMA CAR and co-administered with anti-CD38 antibodies were shown to mediate strong activity against multiple myeloma in multiple xenogeneic mouse models. 138Dual-targeting has also been successfully achieved in preclinical studies with NK-92 cells directed against both CD19 and BCMA, 139 or CD19 and CD138. 140al-targeting strategies may also be an attractive strategy to address the inherent challenge of tumor heterogeneity in solid tumors.As such, dual-targeting CAR NK cells against PD-L1 and ErbB2 were highly effective against solid tumor cell lines expressing both antigens and maintained their cytotoxicity even when one antigen was lost or became inaccessible. 141Dual-specificity CAR NK cells that simultaneously recognize two distinct antigens with a shared epitope have also been investigated.For instance, CAR NK cells targeting a shared epitope of EGFR and its mutant form EGFRvIII were shown to be superior to single targeting CAR NK cells in glioblastoma xenografts. 142Thus, dual-targeting and dual-specificity CAR NK cells could provide a promising approach to counteract the immune escape mechanisms employed by cancer cells.
Multi-antigen targeting can also be achieved through "targetswitchable" CARs, allowing for the recognition of multiple tumor antigens without re-engineering the cell product.One such platform is referred to as split, universal, and programmable (SUPRA) CAR (Figure 2G), which includes a leucine zipper extracellular domain linked to a conventional CAR signaling endodomain (zip-CAR). 143An antigen-recognition module that binds to the leucine zipper can then activate zipCAR expressing T cells upon target engagement. 143Other novel receptor design platforms that combine a CAR backbone with a versatile extracellular domain that binds a chemical or a genetic tag linked to a tumor-specific scFv include peptide neoepitope (PNE)-targeting CARs, [144][145][146] anti-Tag CARs (e.g., fluorescein isothiocyanate CAR), 147 SpyCatcher CAR, 148 fusion protein CAR, 149 and Fab-based adaptor CAR (AdCAR). 150specific antibody-binding adaptor CARs 151 and "split CARs" that require recognition of both targeted TAAs for full activation (Figure 2H) [152][153][154]  and activation after drug administration, etc.These designs have been evaluated in mouse models and some are also currently being assessed in clinical trials (Tables 1 and 2).

| Inducing CAR activation by cues from the solid tumor microenvironment
[157] To divert CAR T-or CAR NK-cell activity away from normal tissues, CARs that rely on sensing certain cues within the TME have been investigated.One such approach is the use of hypoxia-sensing CARs that only activate the CAR response under hypoxic conditions, which is a hallmark of the solid TME. 158For example, the promoter of hypoxiCARs contains a hypoxia-response element (HRE), such that the expression of hypoxia-inducible factor 1α (HIF-1α) that normally increases under low oxygen conditions is mandatory for CAR expression (Figure 3B). 159,160Another design fused the CAR molecule to the oxygen-dependent degradation domain (ODD) of HIF-1α, thus, promoting the degradation of CAR molecule under normoxic conditions via ubiquitination. 160This CAR T-cell system demonstrated high antitumor activity in mouse models of solid tumors, with the CAR T cells being exclusively present in the tumor and absent in non-tumor sites. 160other interesting approach leverages the abundance of proteases in the TME to limit the off-target toxicity of CARs (Figure 3C).
Here, the CAR is modified to express an inhibitory or masking peptide that hinders its ability to bind to its cognate antigen under normal conditions.The inhibitory peptide in a "masked CAR" is susceptible to protease cleavage, such that and in the protease-rich TME, it is cleaved to unmask the CAR antigen binding domain. 161The masking approach does not alter the CAR activity, as anti-EGFR masked CAR T cells showed similar in vivo activity to control CAR cells. 161e use of protease-sensitive CARs has also been tested with NK cells.In a preclinical study in glioblastoma, NK cells were engineered to express a dual-targeting GD2-NKG2D CAR that, in response to proteases in the TME, locally released an antibody fragment to block the immunosuppressive purinergic signaling mediated by CD73. 162 reducing adenosine levels in the glioblastoma TME, this combinatorial strategy addresses key drivers of glioblastoma resistance to CAR NK-cell therapy. 162Because normal tissues also express a wide variety of proteases, the safety of this approach requires further evaluation.
Finally, as the TME often contains chemokines and cytokines, modulation of chemokine signaling to improve the trafficking and localization of CAR T and CAR NK cells to tumor sites have also been investigated (Figure 3D). 163For example, in a preclinical study, EGFRvIII-directed CAR NK cells that also expressed CXCR4 had greater chemotaxis toward glioblastoma cells secreting CXCL12/ SDF-1α, resulting in superior tumor control and improved survival in xenograft models. 164Similarly, forced expression of the chemokine receptor CXCR1, which is activated by IL-8 (secreted by multiple solid tumors), enhanced the migration and activity of intravenously administered NKG2D CAR NK cells in peritoneal ovarian cancer xenografts. 165

| OFF-switch CARs
Alternative strategies to reduce CAR-mediated on-target offtumor activity consist of regulating the expression and activation of the CAR molecule through drug administration.The more classical models of regulatory CARs employ an inducible caspase "suicide switch" that can be pharmacologically activated leading to the elimination of the CAR T or CAR NK cells. 100Our group demonstrated efficient elimination of NK cells expressing a CAR and a suicide switch based on iC9 upon its pharmacologic activation using the small molecule dimerizer AP1903 or Rimiducid both preclinically 68 and clinically. 67Another suicide system that has been used in CAR NK cells is the herpes simplex virus (HSV) thymidine kinase (HSV TK), which converts ganciclovir into a toxic product. 166her strategies to control CAR expression incorporate a reversible OFF-switch.For example, CARs engineered with a C2H2 zinc finger degron motif can be induced to interact with an E3 ubiquitin ligase by lenalidomide, leading to CAR proteasomal degradation (Figure 3E). 167Another drug-controlled system termed signal neutralization by an inhibitable protease (SNIP) incorporates the hepatitis C virus (HCV) NS3 protease (NS3p) together with an NS3p cleavage site at the intracellular end of the transmembrane domain of a CAR to maintain the CAR in an inactive state. 168The CAR can in turn be activated upon exposure to an NS3p inhibitor to prevent its proteolytic cleavage (Figure 3F). 168

| ON-switch CARs
Similar tactics have been used to create ON-switches that control CAR induction and activation by drugs or other chemical or physical factors.For instance, a doxycycline inducible CAR was engineered where the tet response element 3G (TRE3G) was fused to the CAR vector, so that administration of doxycycline induced a conformational change in TRE3G to enable CAR expression. 169An inducible system shown to enhance CAR NK-cell activation and cytokine production combined the MyD88/CD40 signaling endodomain with anti-CD123 or anti-BCMA-CAR.Mimicking toll-like receptor (TLR) activation in DCs and as a potent costimulatory moiety in T cells, the inducible MyD88/ CD40 moiety could be activated with Rimiducid to enhance CAR NKcell function and synergize with IL-15 signaling. 170It will be important to validate the benefit of these novel approaches in reducing on-target off-tumor activity without comprising antitumor potency in the clinic.

| Synthetic circuits and logic-gating strategies
To further regulate CAR activity, other strategies were developed that required recognition and sequential signaling of multiple antigens for activation.One such strategy that has been widely investigated is the design of synthetic Notch (SynNotch) receptors (Figure 3G). 171A SynNotch receptor is designed to recognize a tumor antigen of interest, which, upon ligand binding, induces cleavage of an orthogonal transcription factor that in turn induces expression of a second CAR directed toward another tumor antigen. 172As such, CAR activity toward a second cognate antigen is only initiated in the presence of the first tumor antigen, thus requiring an "AND" logic of both antigens to be present in order to induce CAR activation. 173,174The potential benefits of this strategy were demonstrated in preclinical models of human mesothelioma, ovarian cancer, and glioblastoma through controlling tumor growth, preventing CAR-mediated tonic signaling and maintaining a long-lived memory phenotype. 173,175The SynNotch system was also employed to enable an inducible autocrine circuit to drive IL-2 expression in CAR T cells upon engagement with tumor antigens, resulting in more efficient CAR T-cell infiltration into solid tumors and enhanced antitumor activity. 176e SynNotch system has also been used to increase the secretion of granzyme B (GZMB) and perforin (PRF1) by NK cells and to regulate their intracellular pools by coupling them with pSHP inhibition. 177Similarly, NK cells engineered to express a logic-gated GPC3-SynNotch-inducible CD147-CAR were shown to have increased specificity against hepatocellular carcinoma and reduced toxicity in preclinical models, as both antigens were required for the full-targeted activity of these CAR NK cells. 178However, the immunogenicity potential of the SynNotch receptor, a nonhuman artificial protein, as well as the high level of background signaling due to ligand-independent receptor activity pose safety concerns. 171,179To overcome some of these limitations, a SynNotch system that uses humanized domains with tunable sensing and optimized transcriptional response with the ability to achieve the intended programmed gene regulation (referred to as SNIPR) was recently described. 179 alternative approach to prevent CAR activity against undesirable targets is the use of a "NOT" logic gating strategy.
This approach uses an inhibitory CAR (iCAR) directed against a self-antigen expressed on healthy cells linked to the signaling endodomain of a checkpoint molecule (e.g., PD-1 and CTLA-4). 180cognition of the self-antigen on a heathy cell by the iCAR results in inhibition of CAR T-or CAR NK-cell activity (Figure 3H). 180 overcome self-recognition of trogocytic antigen-expressing (TROG + ) NK cells by the activating CAR (aCAR), and the resultant fratricide and exhaustion, our research group combined the activity of two CARs-an iCAR directed against a self-antigen expressed on NK cells and an aCAR against a tumor antigen. 109is strategy resulted in CAR NK cells receiving a "don't kill me" signal when interacting with their TROG + NK siblings, while preserving the function of their aCAR against the tumor target. 109By combining the activity of these two CARs, we were able to reduce NK-cell exhaustion and fratricide and improve their antitumor activity in vivo.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing not applicable-no new data generated.

| 219 LI
et al. cells.The second major advantage of CAR NK cell therapy is their excellent safety profile with a lower incidence of CRS and neurotoxicity compared to CAR T-cell therapy.This difference could be attributed to the distinct profile of cytokines secreted by CAR NK cells, with lower production of cytokines classically associated with CRS such as interleukin (IL)-1β, IL-2, and IL-6. 47Thirdly, in addition to the targeted cytotoxicity mediated by the CAR, NK cells possess multiple intrinsic mechanisms for tumor recognition as described in the next section.Thus, CAR NK cells could theoretically overcome the challenge of tumor escape through downregulation of the target antigen reported with CAR T-cell therapy.

TA B L E 1
Current clinical trials of CAR NK cell-based antitumor therapy.

TA B L E 1
(Continued)  Thus, the generation of CARs targeting TSAs would require screening individual patients for target antigen expression and manufacturing a custom product, which is costly and time-consuming.79Due to the paucity of known TSAs, other antigens that have higher expression on tumor cells but are not exclusive to tumor cells [referred to as tumor-associated antigens (TAAs), have been explored].While a myriad of TAAs have been investigated for CAR T-cell therapy including HER-2, 80 mesothelin (MSLN), 81 CEA, 82 etc., their clinical success has been limited by concerns related to on-target off-tumor targeting of normal tissues, among others.The extent and severity of on-target off-tumor toxicity depend on a variety of factors: (i) CAR T-or CAR NK-cell accessibility to healthy tissues that express the target antigen; (ii) the type of tissue and its physiological activity; (iii) the expression level and cellular localization of the antigen; and (iv) the potency of the CAR-engineered cells.
are fratricide and trogocytosis.Fratricide occurs when CARexpressing immune cells kill their sibling cells that also endogenously express the cognate antigen.Examples include CAR T cells targeting CD7, 104 CD38,105,106 or CD70,107 antigens that are expressed on normal or activated T cells in addition to cancer cells.This may then result in manufacturing challenges, poor in vivo persistence of the infused product or prolonged immunodeficiency through targeting of normal recipient T cells.Trogocytosis is also an important mediator of fratricide.Trogocytosis corresponds to the receptor-mediated transfer of cognate antigen from tumor cells to the receptorexpressing immune cells,108 and contributes to tumor escape and poor responses after CAR T-and CAR NK-cell therapy by causing antigen loss, NK-cell exhaustion and fratricide.[109][110][111]Our studies on clinical samples from patients with lymphoid malignancies who received anti-CD19 CAR NK-cell treatment confirmed a direct correlation between elevated CD19 levels on CAR NK cells secondary Similarly, CD7-targeted NB-based CAR NK-92 cells mediated strong cytotoxicity in vitro against T-cell leukemia cell lines and primary tumor cells, and in vivo in T-ALL xenograft mouse models. 118These and other studies support the further exploration of NBs as the targeting domain of CAR constructs for NK-cell immunotherapy.An alternative targeting strategy beyond scFvs and NBs is the use of natural receptors and ligands for antigen targeting.To design a natural receptor-based (NRB)-CAR, the ectodomain of the receptor of interest, and often its transmembrane domain and/or signaling endodomain, are incorporated into the CAR construct (Figure 1C).A similar approach is also applied for natural ligand-based (NLB)-CARs TA B L E 2 Current clinical trials of CAR T cell-based antitumor therapy using "next-generation" CAR strategies.

7 |
CON CLUS I ON S AND FUTURE RE S E ARCHCAR-based cell therapy that combines targeted precision medicine with immunotherapy using living cells has proven therapeutic activity in patients with certain hematologic malignancies.Much progress has been made in designing the next generation of CARs to enhance the effector function, proliferation, persistence, and safety of immune cells.Indeed, we are currently witnessing a burst of innovative cell therapy approaches being explored both preclinically and in the clinic.However, considerable effort is still needed to replicate the success observed with CAR cell therapies in B-lymphoid malignancies in other malignancies.The future will likely focus on developing multipronged approaches that leverage our ever-growing scientific and clinical knowledge of tumor immunology and immunotherapy with novel engineering strategies to improve the safety, potency, and feasibility of this therapeutic platform.ACK N OWLED G M ENTSY.L. was supported by the CPRIT Research Training Award RP210028.H.R. was supported by an ASCO Young Investigator Award and the University Cancer Foundation via the Institutional Research Grant program at The University of Texas MD Anderson Cancer Center.This work was supported in part by the generous philanthropic contributions to The University of Texas MD Anderson Cancer Center Moon Shots Program, and The Sally Cooper Murray endowment; by Grants (1 R01 CA211044-01, 5 P01CA148600-03) from the National Institutes of Health (NIH), the Cancer Prevention and Research Institute of Texas (CPRIT) grant RP180466, the Leukemia Specialized Program of Research Excellence (SPORE) Grant (P50CA100632), the Specialized Program of Research Excellence (SPORE) in Brain Cancer grant (P50CA127001), the Stand Up To Cancer Dream Team Research Grant (SU2C-AACR-DT-29-19), and the Grant (P30 CA016672) from the NIH to the MD Anderson Cancer Center.Figures were prepared using biore nder.com.CO N FLI C T O F I NTE R E S T S TATE M E NT Y.L., H.R., K.R., and The University of Texas MD Anderson Cancer Center have an institutional financial conflict of interest with Takeda Pharmaceutical and Affimed GmbH.K.R. participates on the Scientific Advisory Board for GemoAb, AvengeBio, Virogin Biotech, GSK, Bayer, Navan Technologies, and Caribou Biosciences.K.R. is the scientific founder of Syena.The remaining authors declare that they have no competing interests.

CAR strategy NK-cell source Target antigen Cancer Phase Status NCT no.
79

CAR strategy NK-cell source Target antigen Cancer Phase Status NCT no.
a Combination therapy with anti-ROBO1 CAR T cells.
b B-cell lymphoma, multiple myeloma, solid tumors.c Relapsed and refractory malignancies.