TCR-like antibody and GITR signaling lead to effective CAR-T against solid tumor

Main text

Chimeric antigen receptor T cell (CAR-T) therapy is one of the great success stories of cellular immunotherapy for blood cancers (e.g., B cell leukemias, lymphomas, and multiple myeloma) but has remained ineffective for solid tumors. The failure of CAR-T in solid tumor models is partly due to the immunosuppressive nature of the tumor microenvironment (TME), and the choice of target is also a likely contributor. Most tumor-specific or tumor-associated antigens are intracellular proteins that cannot be simply targeted by antibodies, unlike B cell antigen CD19 or B cell maturation antigen that are targeted by CARs in current clinical use. In this issue of Molecular Therapy, Wang and co-workers¹ used a different approach: a novel intracellular signaling domain and an Fv targeting a major histocompatibility complex -peptide (pMHC) complex, the T cell receptor’s (TCR’s) target, to achieve promising pre-clinical control of a solid tumor in an immunodeficient mouse model.

The CARs in current clinical use consist of a single-chain engineered antibody-binding domain (scFv), a linker region, a transmembrane region, an intracellular domain consisting of a part of a co-stimulatory molecule (either CD28 or 4-1BB [CD137]), and the intracellular domain of the T cell signal transducer CD3ζ (Figure 1, left). In Wang and colleagues’ study, the intracellular domain is unusual in that the signaling domain from the glucocorticoid-induced tumor necrosis factor (TNFR)-related receptor (GITR) is placed downstream of the CD3ζ domain (Figure 1, right). Placing the GITR domain after the CD3ζ domain was a practical choice, as in the more standard orientation, they could not express the construct. GITR is related to 4-1BB and has a similar, but not identical, signal transduction pathway.² GITR signaling is reported to reduce the threshold for T cell activation³ and to enable CD8⁺ T cell resistance to regulatory T cells (T_regs).^4,5 These characteristics certainly imply that GITR signaling could be an advantage to a T cell in the TME.

Figure 1.

Classical and newly described CAR-T constructs and their target cells

An illustration of the standard anti-CD19 CAR-T cells approved for clinical use (left), compared to the anti-peptide-MHC class I (“TCR-like) CAR-T described in Wang et al.,¹ making use of a GITR co-stimulatory signaling domain. These CAR-Ts also have different targets on the cancer cells, as shown.

The other major difference from standard CAR design used in this study is the scFv against a classical T cell target: a peptide from the cancer testis antigen MAGE-A4 bound to the MHC class I molecule HLA-A02:01 (A2). Such TCR-like antibodies could be useful as CARs since they can recognize antigens that are intracellular and only appear on the cell surface due to antigen processing and presentation on HLA. Variations of these “TCR-like” antibodies have been tested in CARs, showing good in vitro activity.^6,7,8 However, none have been tested in vivo, and one did not recognize tumor-derived material because the relevant peptide was not presented in a sufficient amount.⁹

Wang et al. screened an scFv library for binding to the MAGE-A4 peptide (the 10-mer p230-239) bound to A2 but not to A2 with other peptides. They isolated one high-affinity scFv with very high specificity for the pMHC. TCR-T therapy with high-affinity TCRs led to cross-reactions on a peptide from a completely unrelated protein,¹⁰ or from related proteins,¹¹ leading to death. So, avoiding cross-reactions in any such therapy is clearly paramount. Wang and co-workers found that only 3/10 residues in the peptide were not required for the scFv to bind the pMHC. Screening of all possible amino acid substitutions at these 3 positions and screening of the human genome did not find any potential cross-reactions. The closest matches were peptides mostly from other MAGE proteins, and the scFv did not bind these. The CAR-T made using this scFv showed the expected antigen and MHC specificities, as well as good cytokine and killing responses to tumor cells in vitro. When compared to classical 2^nd generation CAR-Ts using 4-1BB or CD28 in vivo in an immunodeficient mouse model, the GITR CAR-T was successful in clearing the solid tumor, whereas the other classical CAR-Ts were not. Importantly, the CD8⁺ GITR CAR-Ts infiltrated the tumor, whereas neither of the classical CAR-T constructs were able to do this.

However, the mechanism by which the GITR CAR-Ts were more effective is unclear. For example, although they made more interferon-γ and responded to slightly lower concentrations of antigen than the 4-1BB or CD28 CAR-Ts, the differences were relatively minor—seemingly not enough to explain the profound differences in activity. This mechanism is well worth probing in future experiments. Another wrinkle in this story is that CD4⁺ T cells seemed to have an inhibitory function on the CD8⁺ T cells; together, CD4⁺ and CD8⁺ cells performed worse than CD8⁺ T cells alone, while CD4⁺ T cells alone were minimally effective. This finding was not explored in great depth, but it may be due to the proportion of T_regs in the CD4⁺ population. Additionally, CD8 was not used as a co-receptor in the CD4⁺ CAR-Ts. While this cannot be absolutely ruled out, CD8 was previously shown to have no positive effect on signaling by a TCR-like CAR, even though CD8 was recruited to the immunological synapse.⁷

The work by Wang and co-workers¹ highlights several important points in the development of CAR-T therapy that will be effective for solid tumors.

First, a cell surface protein might be ideal for a CAR-T approach, but such targets are very few and far between in the solid tumor field. The vast majority of tumor-specific or tumor-associated antigens are intracellular proteins. These are naturally targets of T cells, through recognition by the TCR of peptides from these proteins associated with the patient’s HLA molecules. A TCR-like, “MHC-restricted” antibody can provide a way to target such antigens, but it means that the therapy will only work on this HLA allele. In this case, the relevant allele, HLA-A02:01, is very common in the Caucasian population but much less so in other populations. If such an approach is taken, it is important to spread the wealth by targeting antigens presented by HLA alleles prevalent in other populations.

Second, this paper looked carefully for potential cross-reactions on other target antigens and demonstrated convincingly that the scFv is highly specific to the relevant pMHC. It is also possible for such antibodies to cross-react on related HLA alleles.¹² This potential issue could be addressed by testing for cross-reactivity on multiple different HLA alleles.

Third, GITR was identified as a useful domain with exceptional advantages in solid tumors, and it worked downstream of CD3ζ, as opposed to the standard method of putting CD3ζ at the C terminus of the CAR construct. This novel approach calls into question the accepted blueprint for making CAR constructs. The potential arrangement of myriad signaling domains is one issue. This has been approached by a combinatorial screening method for CAR-T activity.¹³ Results from this screen show the strength of this method, as it found known useful intracellular domains that are not in the canonical CAR constructs (including FcεR1γ, CD40, and DAP12). Additionally, the screen identified CD3ε and individual ITAMs of CD3ζ as good signaling domains, as opposed to CD3ζ. CD3ε (containing a single immunoreceptor tyrosine-associated motif: ITAM) was independently identified as a viable CAR signaling domain with superior characteristics to CD3ζ,¹⁴ and reducing the number of CD3ζ ITAMs has also been shown to improve CAR function.¹⁵ This approach is likely to lead to the identification of better CAR designs to target different types of cancer.