REVIEWS Engineering strategies to overcome the current roadblocks in CAR T cell therapy Sarwish Rafiq1,2, Christopher S. Hackett 3 and Renier J. Brentjens3,4,5* Abstract | T cells genetically engineered to express chimeric antigen receptors (CARs) have proven — and impressive — therapeutic activity in patients with certain subtypes of B cell leukaemia or lymphoma, with promising efficacy also demonstrated in patients with multiple myeloma. Nevertheless, various barriers restrict the efficacy and/or prevent the widespread use of CAR T cell therapies in these patients as well as in those with other cancers, particularly solid tumours. Key challenges relating to CAR T cells include severe toxicities, restricted trafficking to, infiltration into and activation within tumours, suboptimal persistence in vivo, antigen escape and heterogeneity, and manufacturing issues. The evolution of CAR designs beyond the conventional structures will be necessary to address these limitations and to expand the use of CAR T cells to a wider range of malignancies. Investigators are addressing the current obstacles with a wide range of engineering strategies in order to improve the safety, efficacy and applicability of this therapeutic modality. In this Review, we discuss the innovative designs of novel CAR T cell products that are being developed to increase and expand the clinical benefits of these treatments in patients with diverse cancers. Chimeric antigen receptors (CARs) are synthetic recep- CAR structure tors that redirect lymphocytes to recognize and elimi- CARs have a modular design with four major com- nate cells expressing a cognate target ligand1. CAR T cells ponents: an antigen- binding domain, a hinge, a trans- targeted at cancer cells are the most widely studied membrane domain and an intracellular signalling form of this technology. Following the demonstration domain (Fig. 1). Each of these elements has a distinct 1Department of Hematology and Medical Oncology, of remarkable response rates in patients with large function and, optimal molecular design of the CAR can Emory University School of B cell lymphoma or acute lymphoblastic leukaemia be achieved through many variations of the constituent Medicine, Atlanta, GA, USA. (ALL)2–5, adoptive transfer of autologous CD19-targeted protein domains. 2Winship Cancer Institute, CAR T cells became the first therapeutic approach with a Emory University, Atlanta, genetic engineering component to be approved by the Antigen recognition and binding domains GA, USA. FDA for use in the USA6,7. Investigators are attempting to The antigen- binding domain is the extracellular por- 3Department of Medicine, broaden the therapeutic benefits of CAR T cells by com- tion of the CAR that recognizes the target antigen and Memorial Sloan Kettering bining them with numerous other types of anticancer redirects the specificity of CAR-expressing lymphocytes Cancer Center, New York, 8 NY, USA. therapies (previously reviewed elsewhere ) or through accordingly (FIG. 1). The antigen- binding domains of 4Cellular Therapeutics Center, innovations in CAR design to address the safety and pro- CARs have traditionally been composed of the variable Memorial Sloan Kettering duction issues associated with the currently approved heavy (VH) and variable light (VL) chains of mono clonal Cancer Center, New York, agents as well as to enhance their efficacy and overcome antibodies, connected by a flexible linker to form a single- NY, USA. 9 treatment resistance. In parallel, novel CAR engineering chain variable fragment (scFv) . The (Gly4Ser)3 peptide 5Molecular Pharmacology strategies are being developed to expand the clinical suc- is the most commonly used linker, exploiting glycine and Chemistry Program, cesses achieved with CAR T cells to patients with other residues for flexibility and serine residues for solubility Memorial Sloan Kettering Cancer Center, New York, malignancies, including solid cancers. In this Review, we and resulting in a properly folded scFv capable of anti- 10 NY, USA. provide an overview of basic CAR design and discuss gen recognition and binding . CARs classically contain *e- mail: [email protected] emerging strategies to engineer safer and more effective scFvs targeting extracellular antigens of cell- surface https://doi.org/10.1038/ CAR T cells for the treatment of haematological and proteins expressed by cancer cells, thus enabling major s41571-019-0297-y solid cancers. histocompatibility complex (MHC)-independent T cell NATURE REVIEWS | CLINICAL ONCOLOGY VOLUME 17 | MARCH 2020 | 147 REVIEWS Key points Molecules other than scFv have been used as alter- native antigen- binding domains for CARs. For exam- • Chimeric antigen receptor (CAR) T cells have induced remarkable responses in ple, zetakine CARs comprising cytokines fused to patients with certain haematological malignancies, yet various barriers restrict intracellular signalling domains, such as those tar- the efficacy and/or prevent the widespread use of this treatment. geting IL-13 receptor α2 (IL-13Rα2) via membrane- • Investigators are addressing these challenges with engineering strategies designed tethered IL-13 linked to the intracellular 4-1BB and to improve the safety, efficacy and applicability of CAR T cell therapy. CD3ζ domains, have been tested in clinical studies27. • CARs have modular components, and therefore the optimal molecular design Other ligand- based CARs are being tested in preclini- of the CAR can be achieved through many variations of the constituent protein cal and clinical studies across a range of malignancies, domains. including those incorporating a proliferation- inducing • Toxicities currently associated with CAR T cell therapy can be mitigated using ligand (APRIL) to target B cell maturation antigen engineering strategies to make CAR T cells safer and that potentially broaden the range of tumour- associated antigens that can be targeted by overcoming on- target, (BCMA) and transmembrane activator and calcium- off- tumour toxicities. modulator and cyclophilin ligand (TACI), which are receptors implicated in the pathogenesis of multiple • CAR T cell efficacy can be enhanced by using engineering strategies to address the 28 + various challenges relating to the unique biology of diverse haematological and myeloma , FLT3 ligands to target FLT3 acute myeloid 29 solid malignancies. leukaemia (AML) , granulocyte–macrophage colony- • Strategies to address the manufacturing challenges can lead to an improved stimulating factor (GM- CSF) to target the GM- CSF CAR T cell product for all patients. receptor (CD116) involved in the pathogenesis of juve- nile myelomonocytic leukaemia30 or natural killer (NK) cell receptor D (NKG2D) to target NKG2D ligands on activation; however, MHC- dependent, T cell recep- the surface of cancer cells31. Finally, T cells expressing tor (TCR)-mimic CARs that enable the recognition of CARs with peptide domains developed de novo for intracellular tumour- associated antigens (TAAs) have binding to specific antigens, such as designed ankyrin also been described11–14. CARs have also been engineered repeat proteins (DARPins) targeting HER2 (REFS32,33) with scFvs that bind to soluble ligands present in the or adnectin peptides (derived from tenth type III tumour microenvironment (TME), such as transforming domain of human fibronectin) targeting EFGR34, have growth factor-β (TGFβ)15, in order to convert an immuno- demonstrated preclinical efficacy. suppressive signal often present in solid tumours into a potent T cell activator. Hinge and transmembrane domains scFv sequences are typically derived from murine or The hinge and transmembrane domains of CARs con- human monoclonal antibodies, although CARs have also nect the extracellular antigen- binding domain to the been engineered to contain smaller, naturally occurring intracellular signalling domains. The hinge provides single- domain antibodies (nanobodies) comprising the sufficient flexibility to overcome steric hindrance and VH domain of camelid heavy- chain antibodies (which adequate length to facilitate access to the target anti- 16 inherently lack light chains and thus VL domains) . gen. Of note, differences in the length and composition Human scFv phage display libraries are another source of the hinge can affect antigen binding and signalling of scFv sequences10,17,18. through the CAR35. The characteristics of the hinge Certain characteristics of the scFv can have effects and transmembrane domain also influence CAR T cell on CAR function beyond solely recognizing and bind- cytokine production and AICD36. Spacer sequences in ing the target antigen. For example, the mode of inter- the hinge domain enable the CAR to access membrane 37–40 action between the VH and VL chains, and thus the proximal antigen epitopes , albeit at the potential relative position of the complementarity- determining cost of decreased CAR T cell function41,42. Amino acid regions, can affect the specificity and affinity of the sequences from CD8, CD28, IgG1 or IgG4 have been CAR for its target antigen19. Importantly, scFv affinity utilized in CAR hinge domains (Fig. 1), although some for the target antigen is a fundamental determinant of of the IgG-derived peptides can interact with Fcγ recep- CAR function and should be high enough to effectively tors (FcγRs), leading to CAR T cell depletion and thus recognize tumour cells and induce CAR signalling decreased persistence in vivo43,44. and T cell activation; however, excessively high affin- The transmembrane domain anchors