Conditional Cancer Immunotherapy as a Safer Way to Step on the Gas

Immunotherapy, particularly in the form of immune checkpoint inhibitors (CPI), has revolutionized cancer therapy in the last decade by “removing the brakes” on the endogenous antitumor immune response. Immunotherapies that agonize costimulatory molecules, such as CD40, CD28, and CD137, represent a complementary strategy to CPIs that use mAbs to enhance T-cell activation against tumors by triggering these receptors, thereby “stepping on the gas.” Although CPIs, systemic cytokine administration, and immune agonist antibodies have demonstrated clinical efficacy, there can be considerable toxicity that results from systemic activation of the immune system. The most well-described are the immune-related adverse events which occur with CPIs, typically characterized by delayed-onset inflammation of healthy organs and cytokine release syndrome or cytokine storm, a broader systemic inflammatory response, which has been seen with agonist antibodies, cytokines, and chimeric antigen receptor (CAR) T cells (1). Local injections of immune agonists are being studied, but this is logistically difficult and cannot be applied to widespread metastases.

CD137 (4-1BB and TNFRSF9), initially discovered as a molecule expressed by activated T cells, is an attractive target for anticancer therapy due to its role in enhancing T-cell proliferation and activation (2). As evidence of its crucial role in potentiating T-cell receptor signaling, it has been utilized in CAR T-cell constructs as a costimulatory domain. However, as has been the case with other agonist antibodies and multiple proinflammatory cytokines targeting immune effector cells, the toxicity induced with systemic administration has represented a significant challenge. Agonist antibodies targeting CD137 studied in early-phase clinical trials have ultimately failed to progress clinically because of a limited therapeutic window, for example, with urelumab and utomilumab (3).

In their article, Kamata-Sakurai and colleagues have aimed to uncouple toxicity from antitumor activity by designing a “conditional” antibody that only binds its target, CD137, in the presence of a high extracellular ATP concentration (or its metabolites). This agonist antibody, STA551, increases IFNγ production in T cells in vitro and has only agonist properties, with neither antibody-mediated phagocytosis or cytotoxicity nor complement-mediated cytotoxicity observed. To test their anti-human CD137 in vivo, they created a mouse surrogate antibody, STA-MB (the STA551 variable region coupled to a mouse constant region), and treated transgenic mice expressing human CD137, showing both local tumor control and inhibition of metastases, in a T-cell–dependent manner. STA-MB upregulated expression of granzyme B, PD-1, KLRG1, or ICOS on CD8⁺ T cells. Expression of CD8⁺ activation markers, as well as efficacy, could be enhanced with combinatory treatment with STA-MB and anti–PD-L1 or a T-cell engager bispecific antibody (targeting tumoral GPC3 and CD3 on T cells). Furthermore, there was minimal toxicity in mice and nonhuman primates at therapeutic doses compared with other CD137 agonist antibodies.

Tumors have higher extracellular concentrations of ATP, due to the breakdown and turnover of cancer cells. In the development of STA551, the authors have taken advantage of the differential in concentration of ATP between tumors and normal tissues to use ATP and its metabolites as a molecular switch to activate the binding of the agonist antibody to CD137. The adenosine pathway has recently been a target of cancer therapy in its own right, as adenosine is inherently immunosuppressive to effector T cells and can augment the function of immunosuppressive cells, such as myeloid-derived suppressor cells and T regulatory cells (Treg), via binding of the adenosine 2A receptor (A2AR). Targeting this pathway can lead to clinical responses in treatment- refractory tumors (4). Additional trials are ongoing to target A2AR as well as CD39 and CD73, enzymes that catalyze the breakdown of ATP to ADP/AMP and adenosine, respectively. STA155 has been designed to bind ATP, as well as its breakdown products (ADP and AMP), and was shown to retain efficacy even in tumor microenvironments (TME) with CD73 and CD39 expression. Nevertheless, degradation of ATP into adenosine through these and other mechanisms may ultimately limit the efficacy of STA551 due to depletion of ATP, ADP, and AMP, as well as due to adenosine's inherent suppressive properties on immune effectors such as T cells.

This conditional immunotherapy is not entirely novel. This approach has already been applied in the development of other second-generation CPIs, agonist antibodies, and cytokines. There are various ways in which specificity can be enhanced and toxicity can be decreased (Fig. 1). In one example, an anti-CTLA4 antibody was engineered with an outer peptide that can be cleaved by proteases found within the TME, shielding an inner active binding domain while in circulation. This anti-CTLA4 masked antibody had demonstrated antitumor activity in mouse tumor models as well as safety in nonhuman primates (5). In an early-phase clinical trial, the anti-CTLA4 masked antibody BMS-986249 was administered alone and in combination with nivolumab. Although higher doses of BMS-986249 could be given than what is deemed safe with the anti-CTLA4 antibody ipilimumab, there was still a 74% rate of treatment-related adverse events with the combination (6). Tumor-targeted bispecific antibodies combine a tumor antigen-binding site with an immune binding domain. One example of this is ABBV-428, a bispecific antibody targeting mesothelin (a tumor-associated protein) and CD40, which has minimal activity against CD40 when mesothelin is not present (7). In mesothelin-expressing tumor preclinical models, the bispecific antibody could control tumors without liver inflammation or systemic cytokine upregulation. In a phase I study of ABBV-428, a human bispecific antibody to CD40 and mesothelin, in patients with advanced solid tumors, neither hepatotoxicity nor cytokine release syndrome was seen, as has been observed with other CD40 agonists (8). Although there was minimal single-agent clinical activity, with no objective responses, there is potential for this drug to have greater efficacy when combined with other immunotherapies or chemotherapy. Tumor-targeted masked antibodies represent a similar approach, combining the features of tumor-antigen specificity and selective activity by using a dual variable domain. When the tumor antigen-binding domain binds to its target, tumor-associated proteases can cleave off this outer domain to expose the inner binding domain, thus keeping the antibody inactive until it reaches the TME. A tumor-targeted masked antibody with binding domains for CTLA4 and prostate stem-cell antigen retained efficacy against tumors in preclinical models via T-cell activation and Treg depletion with minimal toxicity compared with traditional anti-CTLA4, as tissue resident Tregs are spared (9).

With the ATP-activated CD137 agonist, Kamata-Sakurai and colleagues have used the concept of conditional immunotherapy similarly to some of these other examples. However, compared with antibodies that rely on activation within the TME, but then result in an irreversibly activated molecule, which may circulate systemically, the activity of STA551 is reversible if removed from areas of high ATP concentration. This represents an advancement on other described strategies, which should improve specificity and widen the therapeutic window. Cytokines represent an additional area of immunotherapy that would benefit from a conditional strategy, given the intense inflammatory responses that have previously been elicited and largely led to the abandonment of this modality for several decades (1). CAR T cells are also undergoing second- and third-generation engineering incorporating the use of molecular switches to ensure safety. A further strength of the approach taken in the article in this issue is that it does not rely on prior knowledge of tumor antigens, a challenge for the field of tumor immunotherapy due to the heterogeneity and mutability of human cancer cells. Rather, a strategy targeting ATP, found in high concentrations in the extracellular space in many tumor types, is more likely to be applicable to many, although not all, tumor types, depending on the presence of this molecule. However, the downside of targeting a ubiquitous molecule such as ATP and its products is that extracellular ATP also accumulates during inflammation or tissue damage and as a result of other physiologic processes, such as platelet activation and apoptosis, allowing for antibody binding and T-cell activation outside the TME (10). This approach also presumes that agonizing CD137 in the TME is the best strategy to induce antitumor immunity. Emerging studies are now highlighting that T-cell immunity with CPI is actually induced outside the TME (11, 12). Nevertheless, an “off the shelf,” highly specific, and active product with limited side effects makes for a highly desirable immunotherapy strategy, and we expect that the conditional immunotherapy approach will continue to evolve with the development of new treatments and technologies.

L. Fong reports grants from AbbVie, Bavarian Nordic, BMS, Dendreon, Janssen, Merck, and Roche/Genentech during the conduct of the study. No potential conflicts of interest were disclosed by the other author.

L. Fong was supported by a Prostate Cancer Foundation Challenge grant and NCI R01CA223484, U01CA233100, and U01CA244452. B.P. Keenan was supported by NIH T32AI007334 and an ASCO Young Investigator Award.