Anti-tumor antibodies can drive therapeutic T cell responses

Abstract

The classical view of therapeutic monoclonal antibodies against tumor associated antigens is that their mechanism of action is dominated by signal blocking or the cytotoxicity of Fc-driven innate immune effector functions. We review here a mounting body of evidence that anti-TAA mAbs are capable of profoundly synergizing with T cell-directed immunotherapies such as checkpoint blockade and adoptive cell therapy. Two key components account for this synergy: 1) a self-vaccinal effect mediated by dendritic cells; and 2) an inflammatory repolarization of the tumor microenvironment. Efficient exploitation of these mechanisms has tremendous therapeutic potential.

Classical anti-TAA mAb pharmacology

The pharmacological objective of killing every single tumor cell is shared by the great majority of cancer drugs, from chemotherapy to targeted kinase inhibitors. However, extended courses of drug administration usually just result in outgrowth of resistant clones [1, 2], shifting the Kaplan-Meier curve midpoint rightward without improving long-term survival. To definitively counter such an evolving pathology, an adaptive therapy is required – and natural selection has provided us a means to accomplish this, in the form of our adaptive immune system.

Although originally extracted from the adaptive humoral immune system, monoclonal antibodies against tumor-associated antigens (anti-TAA mAbs; e.g. cetuximab, trastuzumab, and rituximab) were at first conceptualized as signal blockers rather than triggers of an endogenous therapeutic immune response. It was subsequently found in mouse studies [3] and retrospective analysis of clinical experiences [4] that antibody effector function contributes significantly to anti-TAA mAb efficacy, suggesting that antibody dependent cell-mediated cytotoxicity kills sufficient numbers of tumor cells to directly account for the observed clinical efficacy.

Roles for innate antibody effector function in priming a T cell response

However, in the natural course of an infection, the innate immune response serves primarily as an early restraining action to buy time while initiating the adaptive immune response necessary for long-term protection. Along these lines, it has been proposed that anti-TAA mAbs may foster therapeutic T cell responses [5, 6]. Multiple mouse model studies suggest that anti-TAA mAbs initiate a CD8+ T cell response - that is in turn required for anti-tumor efficacy. In the examples shown in Figure 1, anti-TAA mAb monotherapies cure syngeneic tumors in mice with wild-type immune systems, but lose all efficacy upon depletion of T cells. These results are not more widely known because most mAb preclinical studies in the past have been performed with human tumor cells in mice lacking T cells. A linkage from mAb therapy to T cell responses has also been observed clinically: treatment with trastuzumab [7], cetuximab [8], or a trastuzumab antibody drug conjugate [9] all stimulate T cell responses against tumors in patients.

Figure 1.

Efficacious anti-TAA mAb therapies in syngeneic murine tumor models require the presence of T cells for their mechanism of action. Four different published examples are shown. A) Anti-HER2 mAb fails to cure tumors after depletion of CD8+ T cells [28]. B) Anti-HER2 mAb loses efficacy after anti-CD8 depletion [29]. C) Enhanced survival from anti-TAA mAb therapy plus a TLR4 ligand fails upon depletion of CD4+ and CD8+ T cells [30]. D) Efficacy of anti-TRP1 mAb and extended-lifetime IL-2 fails when CD8+ T cells are depleted [20].

How do anti-tumor antibodies drive a T cell response? Two inter-related mechanisms are supported by significant available evidence: 1) a vaccinal effect following mAb effector-mediated tumor cell killing; and 2) inflammatory reprogramming of the tumor microenvironment. The vaccinal effect results from generating a bolus of antigenic material via tumor cell death, which antigen presenting cells such as dendritic cells or macrophages then cross-present to CD8+ T cells. Analogous vaccinal effects have been shown to occur with chemotherapies that drive immunogenic cell death [10], and with external beam radiation [11]. Particular advantages of antibody-directed immunogenic tumor cell death include: absence of the inadvertent cytotoxicity of chemotherapy and radiation against the very immune effector cells essential for an immune response; efficient concentration and packaging of antibody-bound immune complexes for focused uptake by professional antigen-presenting cells; and stimulation of activating Fc gamma receptors on DCs, which has been demonstrated to potentiate the process of cross-presentation of phagocytosed antigen [12, 13]. The ability of anti-TAA mAbs to drive vaccinal stimulation of T cell responses has been demonstrated in mouse models numerous times [5, 7, 8, 14–19]. Given this extensive convergent evidence, a vaccinal effect should reasonably be assumed to comprise at least a component of the mechanism of action of any anti-TAA mAb possessing an activating isotype (e.g. human IgG1 or murine IgG2a). A particular advantage that immune complexes containing tumor cell debris have over the limited number of synthetic neoantigens in a typical cancer vaccine is the inherent personalization of such self-vaccination, which carries the potential for antigen spreading (which has been observed experimentally [19]). Such antigen spreading could enable the adaptive immune response to track antigenic changes in a genetically labile heterogeneous tumor cell population despite loss of expression of the particular antigens selected for synthetic vaccination.

A second mechanism by which anti-TAA mAbs drive T cell responses is by reprogramming of the tumor microenvironment immune contexture, with significant local increases in inflammatory cytokines and chemokines in syngeneic tumor models (Figure 2). This intratumoral “cytokine storm” is likely a direct consequence of innate effector cell activation during anti-TAA mAb-driven ADCC. Across several of our published studies, consistent and robust increases in MIP-2, G-CSF, and IL-6 in particular have been observed upon addition of an anti-TAA mAb to an immunotherapy cocktail. In one instance [20], these increases were shown to result from a complex interaction amongst NK cells, neutrophils, eosinophils, and macrophages, with macrophages directly increasing MIP-2 production in a fashion reliant on the presence of the other innate immune effector cell types. Not surprisingly, these changes are then correlated with significant intratumoral increases in essentially all types of immune cells examined [19, 20]. The chemokines MIP1α (CCL3), MIP1β (CCL4), and RANTES (CCL5) that are increased intratumorally upon anti-TAA mAb treatment (Figure 2) are among a signature of six chemokines preferentially expressed in human metastatic melanoma biopsies infiltrated with T cells [21]. Given the observed clinical correlation between tumor-infiltrating lymphocytes and response to checkpoint blockade antibodies [22, 23], it is therefore an exciting possibility that existing anti-TAA mAbs may help convert nonresponding patients to obtain benefit from checkpoint blockade antibody treatment. Indeed, we have observed significant anti-TAA mAb potentiation of adoptive cell therapy, vaccines, and immunocytokines in mouse models of cancer (Figure 3;[19, 20, 24]).

Figure 2.

Intratumoral increases in cytokines and chemokines upon addition of an anti-TAA mAb to: Fc/IL-2 (black bars, Zhu et al.); an IgG-based IL-2 immunocytokine (dark grey bars, Tzeng et al.); and combination Fc/IL-2, anti-PD1, and vaccine immunotherapy (light grey bars, Moynihan etal.). In each of these studies, numerous additional analytes were statistically significantly increased; the particular inflammatory molecules most consistently upregulated upon anti-TAA mAb treatment are excerpted and highlighted here.

Figure 3.

Three published examples where addition of an anti-TAA mAb to a T-cell-directed immunotherapy protocol significantly improved its efficacy. A) CD8+ T cell adoptive therapy (ACT) plus IL-2/Fc fusion protein extended survival, but addition of an anti-TAA mAb leads to cures of most mice [20]. B) A vaccine that potently elicits CD8+ Tcell responses, together with an anti-PD-1 antibody and an IL-2/albumin fusion protein (IPV) extends survival of tumor-bearing mice but gives no cures. Addition of an anti-TAA mAb leads to cures for the majority of mice. C) Simultaneous treatment with an anti-TAA mAb, IL-2/Fc fusion, and interferon-alpha extends survival but leads to no cures. However, delaying the interferon-alpha dose by 48 hours leads to cures of all mice, with evidence supporting an effect of improving dendritic cell cross-priming of CD8+ T cell responses [27].

How might this immune response-kindling function be best exploited with anti-TAA mAbs? This mechanism of action is so distinct from simple signaling antagonism as to demand a fundamentally different dosing perspective. In order to kindle adaptive immunity, episodic treatments might be better matched to the natural temporal sequence in an immune response than the sustained high-level exposures typical with chemotherapy or kinase inhibitors. The acute inflammatory response initiates an autonomously feedback-limited time course, terminated by synthesis of anti-inflammatory molecules such as lipid mediators hours to days after onset [25]. There is the distinct possibility that continued saturation with mAbs during the resolution phase of the inflammatory cycle could actually paradoxically lead to an immunosuppressive state of chronic inflammation [26]. Antigen presentation under such tolerizing conditions would clearly be counterproductive to the intended therapeutic effect. Perhaps anti-TAA mAb administration should be reconceptualized as a pulsatile prime or boost component of a vaccination, for which one would never contemplate perpetual saturating dosing.

An anti-TAA mAb’s effector functions generate a significant amount of tumor cell debris in the first days following a bolus dose. One therefore might expect antigen presenting cells to possess an inventory of potential tumor antigens at this time, and not considerably before or after. If one were to administer a pharmacological pulse of dendritic cell stimulation, it would make the most sense to time it so as to coincide with maximal DC antigen loading. We indeed found this to be the case for several immunostimulatory molecules [27]. Coadministration of a type-I interferon with anti-TAA mAb was ineffective against large syngeneic tumors, but delay of the interferon to two days following the mAb administration led to a majority of cures after only two cycles of treatment (Figure 3C). Of particular note, premature administration of a type-I interferon was found to significantly inhibit the immune response to a protein vaccine antigen. This same general time dependence was observed for several other DC immunostimulatory agents [27].

Concluding remarks

Approaching the pharmacology of the extensive industrial portfolios of anti-TAA mAb clinical candidates (and approved drugs) with this vaccinal/immunotherapy mindset could potentially contribute to a considerable amplification of the already impressive inroads that checkpoint blockade antibodies have made against cancer. The pipelines of pharmaceutical companies were once full of anti-RTK antibodies that gave disappointing therapeutic responses despite early promise in preclinical studies in xenografted tumors in nude mice. For example, nine different antibodies against IGF1-R (cixutumumab, ganitumab, dalotuzumab, figitumumab, teprotumumab, robatumumab, AVE1642, BIIB022, and isiratumab) have been tested in clinical trials, but none provided sufficient efficacy for FDA approval. These same anti-TAA mAbs (except figitumumab and BIIB022, which are not human IgG1 isotype) might well be repurposed to serve as innate immune kindlers of a therapeutic CD8+ T cell response, if combined with checkpoint blockade antibodies (see outstanding questions). Dozens of anti-TAA mAbs and antibody drug conjugates are being evaluated in industrial clinical trials, and a case can be made that each of these should be clinically evaluated in a prime/boost dosing schedule together with an anti-PD-1 or anti-PD-L1 antibody.