Protein-based immunotherapies offer the possibility of generic or “off-the-shelf” immunotherapies, different from the highly personalized approach of engineered immune cell therapies. For example, T cell–engaging bispecific antibodies (CD3 BsAbs) trigger signaling of the CD3 surface receptor on T cells and also bind to a second target protein on tumor cells, thereby activating cytotoxic T cells to eliminate cancer cells with one antibody molecule. But this approach has challenges, including identifying shared cancer-selective targets and protein engineering to redirect T cells. A study by Hsiue et al. (1) on page 1009 of this issue and a study by Douglass et al. (2) describe the development of CD3 BsAbs that recognize mutation-associated neoantigens (MANAs). Additionally, Paul et al. (3) describe a CD3 BsAb that uses T cell receptor (TCR)–specific antibodies to selectively eliminate T cell malignancies. In the future, these immunotherapeutic agents could be used to treat diverse cancers with specific mutations.
Tumor antigens are classified as either tumor-associated antigens (TAAs) or tumor-specific antigens (TSAs). TAAs are aberrantly expressed by, but are not specific to, cancer cells. Because normal tissue also expresses TAAs, albeit at lower levels, on-target off-tumor toxicities of immunotherapies that target TAAs are a concern. Conversely, TSAs are found only on cancer cells. MANAs are a subclass of TSAs and are derived from the expression of common hotspot mutations in cancer genes. Next-generation sequencing combined with proteomics and bioinformatic tools have accelerated the discovery of MANAs, with particular focus on those derived from cancer-driver genes (which give a selective growth advantage to cells when mutated), such as the tumor suppressor gene TP53 (which encodes p53) and the RAS family of oncogenes (KRAS, HRAS, and NRAS) (4, 5).
MANAs, like other cellular proteins, are processed by the proteasome, and the peptides that are generated are transported into the lumen of the endoplasmic reticulum (ER) by the heterodimeric transporter associated with antigen processing (TAP1-TAP2) complex before being trimmed by ER aminopeptidases into 8 to 10 amino acid peptides. These mutated peptides (neopeptides) are loaded onto classical human leukocyte antigen (HLA) class I molecules that present MANAs on the cell surface as neoantigens (pHLA).
Finding neopeptides that bind to HLA molecules can be assisted by using in silico prediction software. Hsiue et al. and Douglass et al. provide a roadmap to discover and quantify neoantigens. Each group used computer modeling to predict binding of commonly mutated proteins in cancer—p53R175H, KRASG12V, HRASG12V, NRASG12V, NRASQ61L, and HRASQ61L—to different HLA class I molecules. For example, p53R175H binds to HLA-A2, KRASG12V binds to HLA-A3, and NRASQ61L binds to HLA-A1. To confirm in silico predictions, both teams used a highly sensitive mass spectrometry (MS)–based approach [MANA-selective reaction monitoring (SRM)] to analyze HLA-bound peptides. Hsiue et al. estimated on average that there are only 2.4 copies of p53R175H/HLA-A2 expressed per cell by a human multiple-myeloma cell line. Douglass et al. showed that both the 9- and 10-amino-acid–long peptides from KRASG12V are processed and presented by HLA-A3. However, only the 10-amino-acid–long KRASG12V peptide is presented by HLA-A3 on human lung papillary adenocarcinoma and pancreatic ductal adenocarcinoma cell lines at nine and three copies per cell, respectively. Although adoptive T cell therapies have been used to target mutant oncogenes and tumor suppressor genes (6), the findings from these studies directly confirm that MANA pHLAs are present but at very low amounts on human cancer cells.
TCR-mimic (TCRm) antibodies are a class of biologics that recognize pHLA derived from intracellular proteins (7). CD3 bispecifics that include a TCRm were previously described and shown to mediate T cell cytotoxicity of human colorectal adenocarcinoma cells in vitro and elimination of human transformed B cells in mice (8, 9). Hsiue et al. and Douglass et al. used phage display to screen naïve human antibody libraries to discover TCRms specific to their MANA pHLA of interest. Both groups then made CD3 BsAbs by using a single-chain diabody (scDb) format, comprising a single-chain TCRm antibody fragment (Fv) specific for MANA pHLA fused to an scFv CD3 antibody. The CD3 scFv moiety binds CD3 and activates a polyclonal T cell response when targeted to cancer cells expressing the MANA pHLA antigen.
As with any immunotherapy, safety and toxicity issues are a potential concern. Creation of scDbs with affinity for mutated residue(s) in the neopeptide could reduce the possibility for cross-reactivity and hence the risk for off-target toxicity. Characterization of the binding selectivity of TCRm from both studies revealed that scDbs could specifically recognize mutated peptides and not the corresponding wild-type or related mutated peptides. Hsiue et al. used x-ray crystallography to provide a structural basis for scDb binding only to mutated p53R175H/HLA-A2 and more broadly support the use of structural data to predict cross-reactivity of scDbs.
Hsiue et al. and Douglass et al. describe scDbs with high sensitivity that can activate T cells in vitro to lyse human myeloma cells and lung papillary adenocarcinoma cells expressing less than 10 copies of neoantigen pHLAs. Moreover, Hsiue et al. show that immunodeficient mice engrafted with human T cells suppressed growth of a human multiplemyeloma cell line expressing an average of 2.4 copies of p53R175H/HLA-A2 per cell after treatment with the H2-scDb. These studies provide the first evidence for CD3 BsAbs to eliminate cancer cells by activating polyfunctional T cell responses against previously nondruggable intracellular targets presented by tumor cells at very low density.
To date, more than 100 different CD3 BsAb formats have been developed for therapeutic use (10). Hsiue et al. and Douglass et al. created multiple formats of CD3 BsAbs, including using different CD3 antibodies, to identify scDb as the optimal format. It might be the small, compact, and rigid nature of the scDb structure that is responsible for the observed enhancement in antitumor activity over other bispecific formats. This is especially true when it comes to tumor targets expressed at low densities such as described by Hsiue et al. and Douglass et al. Perhaps the more potent antitumor activity reported for the scDb is related to establishing an optimized immunological synapse between tumor cells and T cells. Or possibly, the affinity of the CD3 antibody or its binding epitope are important. Regardless, these findings will likely foster rapid development of scDb CD3 BsAbs that target MANA pHLA and other TAAs and TSAs on cancer cells.
Paul et al. describe an approach to deplete T cell malignancies (T cell leukemias and lymphomas) while preserving healthy T cells by using scDb molecules. To establish proof of concept, the authors took advantage of previously identified antibodies to TCR β-chain variable regions, TRBV5-5 and TRBV12, which are two of 30 possible TRBV family members that form a functional αβ TCR expressed by T cells (see the figure). Because T cell malignancies are clonal and express only one TRBV family member on their surface, targeting a single TRBV type represents an ideal TAA for selectively treating with CD3 BsAbs. Paul et al. show that scDb composed of an antibody to TRBV5-5 or TRBV12 and tethered to a CD3 antibody can selectively eliminate malignant human T cells that express TRBV5-5 or TRBV12 in mouse models while preserving the majority of healthy human T cells not expressing the targeted TRBV.
Although the studies of Hsuie et al., Douglass et al., and Paul et al. are promising for advancing scDbs into the clinic, there are other factors to consider before therapeutic efficacy can be fully realized. scDbs are small molecules that are rapidly cleared from the blood in humans and mice, which will likely make it necessary to continuously infuse scDb drugs with an implanted pump. The addition of an immunoglobulin G (IgG) fragment crystallizable (Fc) domain to the scDb molecule would be expected to increase its plasma half-life. However, the addition of an Fc domain would also create a bulkier molecule that could sterically hinder epitope binding and result in reduced antitumor potency. Although further identification of commonly shared MANA pHLA will expand the targets available on solid tumors for CD3 BsAbs, the polymorphic nature of classical HLA class I limits the individuals who can be targeted with these agents. To broaden population coverage, an alternative may be to find MANA-derived peptides presented by nonclassical HLA (that is, HLA-E and HLA-G). Unlike classical HLA, HLA-E and HLA-G are essentially monomorphic, and their activity is up-regulated on cancer cells. Whether HLA-E and HLA-G present peptides from MANAs may be worth investigating.
There are potential issues related to the tumor microenvironment that might limit the effectiveness of CD3 BsAbs, including scDb against both hematological malignancies and solid tumors. For example, in cases in which limited T cell activation has been observed, combination with agonist antibodies to activate costimulatory receptors such as CD137 is being pursued (11). Both hematological and solid tumors actively evade immune cell responses through the expression of inhibitory immune checkpoints that would negatively affect T cell activation by CD3 BsAbs. Potential solutions include combination with immune checkpoint blockers, such as antibodies that block cytotoxic T lymphocyte–associated protein 4 (CTLA-4) and programmed cell death protein 1 (PD-1) and thereby allow T cell activation (12). Additionally, some solid tumors lack T cell infiltration, and the combination of BsAbs with oncolytic viruses and cancer vaccines is being evaluated to encourage T cell trafficking to tumors (10, 13). The studies by Hsiue et al., Douglass et al., and Paul et al. offer a potential avenue to achieve off-the-shelf, protein-based immunotherapeutics for treating cancers with specific TAAs or mutations that are expressed as MANAs. However, much more work will be required before this ambitious goal can be achieved.
Acknowledgments: J.W. is a cofounder of and chief scientist at Abexxa Biologics.