An scFv-based ectodomain is the most common antigen-binding motif used in CARs, due to its ease of appropriation from existing antigen-specific monoclonal antibodies [12
]. The incorporation of scFv binding domains has enabled antigen recognition with a range of affinities, but typically in orders of magnitude higher than TCR–peptide MHC (pMHC) interactions. In recent years, significant knowledge has been generated in the field regarding the effect of antibody modifications to the scFv ectodomain on CAR T cell efficacy and function, such as by tuning specificity or re-engineering CAR structures to other binding domains such as dual chains, native receptors, or ligand-based designs.
2.1. Affinity Tuning Antibody Domains
Single-chain antibody domains are amenable to affinity tuning, whereby the binding affinity between antigen and antigen-binding ectodomain can be selectively mutated, influencing affinity and CAR T cell function. Whilst CAR T research has traditionally utilised high affinity scFvs, there is some evidence to suggest that targeting the antigen using the highest affinity scFv possible does not necessarily result in the optimal outcome. Early studies using human epidermal growth factor receptor 2 (HER2)–CAR T cells with varying affinities (3.2 × 10−7
to 1.5 × 10−11
) has demonstrated that the activation threshold of CAR T cells is inversely correlated with scFv affinity [13
]. More recently, evidence has suggested that high affinity TCR interactions (above 10 µM) increase T cell signalling strength, as determined by T cell calcium flux and extracellular signal-regulated kinase (ERK) phosphorylation, without necessarily improving the efficacy of T cell therapy in B16/A2-Kb
melanoma mouse models [14
]. These results suggest that high affinity binding domains may not necessarily improve therapeutic outcome. Affinity of binding has been shown to be critical in a clinical trial using trastuzumab (Herceptin), a widely used high affinity monoclonal antibody to HER2. In this trial of metastatic colon cancer, the high affinity HER2-specific CAR T cells also recognised low-level HER2 antigen expression on healthy lung epithelial cells, causing a cytokine storm and resulting in the death of a patient [15
]. Conversely, in other phase I clinical trials, targeting HER2 with a lower affinity scFv has been shown to be safe [16
]. This discrepancy highlights that the affinity of the scFv can selectively detect varying antigen loads and drive on-target off-tumour T cell responses. Researchers demonstrated that the selectivity threshold to antigen could be controlled, as tuning trastuzumab-based CARs to a lower affinity could spare “healthy” tissue levels of HER2 on PC3 cells whilst maintaining efficacy against “high” tumour levels of antigen expression on HER2-overexpressing SK-OV3 cells [17
]. Therefore, controlling the affinity and avidity of CAR T cell interactions is of critical importance when designing a CAR targeting tumour-associated antigens (TAAs) which may also be expressed on healthy tissues. These findings have been replicated in other studies involving the affinity tuning of scFvs against TAAs such as epidermal growth factor receptor (EGFR), CD123, and CD38 [17
2.4. Receptor and Ligand-Based CAR
Relatively less is known about the clinical efficacy of the natural ligand-based CAR, which involves the fusion of a membrane-tethered ligand (such as the interleukin (IL)-13 ligand recognising a tumour-associated or specific receptor such as IL13Rα2) to the CAR endodomains [28
]. This approach is new, with few, such as the IL-13Rα2 and T1E CARs, having been translated to the clinic [28
]. In one such promising trial, a single patient receiving multiple doses of CAR T cells targeting IL-13Rα2 demonstrated tumour regression of all intracranial and spinal recurrent multifocal glioblastoma in a phase I clinical trial; however, the patient relapsed 7.5 months following therapy [29
Another ligand CAR is the T1E CAR (named after the promiscuous T1E peptide) targeting HER2/ErbB2 and ErbB dimers associated with many solid tumours, which entered a phase I clinical trial for squamous cell head and neck cancers (NCT01818323). This clinical trial was initiated following promising results using in vivo xenograft models where no accompanying toxicity was observed [30
]. The T1E CAR was co-expressed in T cells alongside a chimeric cytokine receptor termed 4αβ, consisting of an IL-4 receptor a subunit ectodomain fused to the IL-2/IL-15 receptor β chain transmembrane and endodomain to facilitate gene-modified cell enrichment following IL-4 administration [31
]. The use of the ErbB ligand allowed native binding to ErbB1- and ErbB4-based homo- and hetero-dimers, and ErbB2/3 heterodimers, which would not have been possible using any other CAR format. Other ligand-based CAR designs currently in pre-clinical testing include CARs targeting adnectin [32
], follicle-stimulating hormone (FSH) [33
], and CD27 (targeting CD70) [34
]. The expansion of the CAR ectodomain to include designs beyond antibody CDR binding moieties has allowed a broader range of potential antigens to be targeted, particularly those with more complex binding interactions reliant on heterodimer/multimer receptors.
2.5. Universal Immune Receptors (UIRs)
Current approved CAR T cell immunotherapies are manufactured on a patient-specific basis, which limits agile responses to antigen escape and limits the ability for dose titration or repeat administration of CAR T cells. One approach to overcome these obstacles is to design an “off-the-shelf” flexible product. It is becoming clear that targeting a single cancer antigen may be insufficient for lasting remission. Even after early CAR T cell clinical successes, antigen escape, which is the persistence and outgrowth of antigen-negative tumour cells, remains a significant problem in the field, both for haematological cancers [35
] and also solid cancers [36
]. To mitigate antigen escape, recent CAR designs have adopted a “lock-key” system, relying on an intermediate molecule in antigen targeting to allow for broader antigen specificity. This approach results in the creation of a “universal” CAR which can then be used with various adaptor molecules.
There are several different kinds of UIRs with adaptable specificity, including (1) antibody dependent cytotoxicity receptors, (2) bispecific protein mediated linkage, (3) anti-Tag CARs, and (4) tag-specific interactions (reviewed extensively in [37
One form of UIR utilises the antibody dependent cytotoxicity receptors on effector cells, such as natural killer (NK) cells, to facilitate antibody dependent cell cytotoxicity (ADCC) of target tumour cells by incorporating receptors such as NKG2D [38
], CD16 [39
], NKp30 (targeting B7116, BAG6, Gal3) [42
], or DNAM-1 (targeting PVR and adnectin) [43
] as the antigen binding domains of the CAR. In particular, T cells transduced with CD16 CARs, which consists of an extracellular CD16 (or FcyRIIIa) domain attached to an FceRIy endodomain, cause ADCC or the antibody-mediated apoptosis of target tumour cells coated with therapeutic monoclonal antibodies, such as rituximab (anti-CD20) or trastuzumab (anti-HER2) by recognising their IgG component [39
]. Researchers achieved cancer regression in vivo, first in CD20-positive B-lymphoblastoid cell lines using rituximab [39
], and later in human HER2-positive BT474 breast cancer with trastuzumab engrafted in immunodeficient mice [40
]. A later study, using anti-CD20, anti-HER2, and anti-GD2 antibodies to facilitate ADCC, along with a second generation CD16 CAR, demonstrated superior T cell activation, proliferation, and cytotoxicity compared to the first generation CD16 CAR [41
]. Due to these promising results, the CD16 CAR is currently in clinical trials (NCT02776813, NCT03189836, NCT03266692, NCT03680560) for non-Hodgkin’s lymphoma, HER2-positive cancers, or multiple myeloma. However, while ADCC is a powerful mechanism in the control of some cancers, it has yet to be determined whether the downregulation of the NK cell receptor ligands from tumour cells will prevent complete clearance and long-term remission.
In another “lock-key” interaction, a biotin-binding immune receptor CAR (BBIR CAR), utilises biotin–avidin to act as an intermediary mechanism bridging the CAR T cell to an antigen [45
] (Figure 2
). By replacing a CAR scFv with an avidin molecule, it allows recognition of any biotinylated antigen-specific molecule from a number of sources and allows for multiple antigens to be targeted simultaneously based on diverse biotinylation. This approach has shown to result in both in vitro and in vivo tumour suppression [45
]. This was replicated in a second study [46
] with similar success, but concerns were raised regarding the immunogenicity of avidin, which caused off-target anti-tag effects. Therefore, whilst promising, this approach must be extensively tested in preclinical in vivo models to determine efficacy and safety.
Another approach is to use anti-tag CAR designs which utilise scFvs targeting molecular tags or chemically conjugated peptides which in turn bind to tumour antigens. Some tags that have been explored include the fluorescent label fluorescein isothiocyanate (FITC) [47
] and a peptide neoepitope (PNE) [50
]. This approach is quickly being adopted by a number of groups and are appealing as they introduce a measure of external control over the CAR [50
]. This is best shown using a Nalm-6 xenograft model in which the switch anti-PNE CAR system was able to be dose-titrated by adjusting the concentration of antibody administered, with low antibody doses (0.5 mg/kg) maintaining a larger central memory CAR T cell population, improving long term CAR T cell persistence compared to high doses (2.5 mg/kg), which produced more effector CAR T cells [50
Another example of a universal CAR using tag-specific interactions is the split, universal, programmable (SUPRA) CAR system. This design incorporates an extracellular leucine zipper as the CAR ectodomain, bound to the intracellular T cell signalling domains (zipCAR). The leucine zipper of the zipCAR binds a second leucine zipper on an scFv or other antigen-specific molecule (zipFv), allowing for T cell activation following zip–zip interaction toward any leucine-zipper bound molecule [53
]. Using an immunodeficient NOD Scid gamma (NSG) xenograft mouse model, zipCAR T cells were shown to successfully eliminate HER2-positive SK-BR-3 human breast cancer cells grown subcutaneously following regular doses of anti-HER2 zipFv, with higher doses of the zipFv correlating with better killing but not affecting cytokine production [53
]. Moreover, affinity tuning the interaction resulted in a correlation shown between increasing binding affinity and cytokine release [53
Whilst these UIR platforms have yet to be tested in the clinic, their broad applicability and flexibility in various cancer types and antigen targets suggest that the UIRs are a promising strategy that allow for increased utility to be derived from the once off transduction of a patient’s T cells.