The odds of immunotherapy success

See allHide authors and affiliations

Science  09 Oct 2015:
Vol. 350, Issue 6257, pp. 158-159
DOI: 10.1126/science.aad4140

Cancer immunotherapy has advanced to the forefront of molecular medicine as a consequence of the success of monoclonal antibodies (mAbs) that block immune checkpoints. Such antibodies, like ipilimumab, reverse cancer-induced immunosuppression and induce durable therapeutic responses in certain cancer patients (1). However, because only some patients respond to checkpoint blockade therapy, there is a need for reliable biomarkers that identify individuals most likely to respond to such treatment. On page 207 of this issue, Van Allen et al. (2) report the genomic analyses of tumors from 110 melanoma patients prior to ipilimumab therapy. The study not only validates features of responsive melanomas suggested in smaller-scale analyses, but also refutes claims that associate responsiveness to ipilimumab with tumor antigens that show putative similarities to microbial proteins (3).

During “cancer immunoediting,” tumor cells may evolve antigens with a reduced ability to provoke an immune response and/or establish an immunosuppressive tumor microenvironment (4). Both characteristics are interrelated because the latter can arise as a consequence of chronic but ineffective T cell stimulation. These events trigger an inhibitory signaling pathway in T cells, involving the expression of immunosuppressive immune checkpoint proteins [such as cytotoxic T lymphocyte antigen–4 (CTLA-4) or programmed cell death–1 (PD-1)] that impair T cell effector mechanisms (1). Indeed, mAb blockade of CTLA-4 on T cells promotes their antitumor activity. Despite the success of checkpoint blockade therapy, many patients do not benefit from treatment for reasons that are unclear but may be related to the reduced antigenicity of tumor cells. Tumor antigens can be either aberrantly expressed normal proteins or abnormal proteins displaying tumor-specific expression (5). Although somatic cancer mutations could give rise to “nonself” tumor-specific mutant antigens (neoantigens), this concept had to await experimental demonstration of its generalization. Next-generation sequencing and epitope prediction algorithms identified neoantigens in mouse tumors that functioned as tumor-specific targets for T cells (6, 7). Clinical studies showed that neoantigens were recognized by preexisting T cells in melanoma patients and were likely the major contributor to positive clinical effects seen with adoptive immune cell transfer therapy. Mouse models established that neoantigens were the targets of T cells activated by checkpoint blockade therapy, and that synthetic long peptides comprising these neoantigens were effective when administered as vaccines with CTLA-4 and/or PD-1 mAbs (8). Subsequent studies implied that cancers with higher mutation burdens, and therefore a greater likelihood of expressing neoantigens, were most likely to respond to checkpoint blockade therapy (5). Indeed, in melanoma, both the numbers of mutations and neoantigens correlated with patient response to ipilimumab (3). Further studies suggested a similar relationship among mutational burden, neoantigens, and response to PD-1 mAb therapy in non–small cell lung cancers (9) and colorectal cancers (10).

Neoantigen roulette.

Melanomas with fewer mutations are less likely to contain “winning” neoantigen(s) and are thus more likely to be unresponsive to immunotherapy (ipilimumab). Those melanomas with greater numbers of mutations have an increased chance of responding to immunotherapy because they have an increased chance of having neoantigens that activate T cells. Functional activity of these T cells can be sustained by blocking CTLA-4 (with ipilimumab), thereby rendering the tumor susceptible to elimination by the immune system.


Van Allen et al. found that the number of nonsynonymous mutations per tumor correlated with clinical responses in ipilimumab-treated melanoma patients. They then pipelined their mutational data into algorithms to determine whether the number of potential neoantigens recognized by T cells correlated with clinical response more strongly [the algorithms pertained to binding of peptides to major histocompatibility complex class I (MHC-I), which are then presented to T cells]. Although there was a correlation, no increase in significance was observed; in fact, the mutational load itself was a better indicator of response (P = 0.0076 for mutational load versus P = 0.027 for neoantigens). An intriguing explanation for the latter observation surrounds the focus of Van Allen et al. on neoantigens formed by missense point mutations while not considering other potential alterations that could also lead to neoantigen formation [e.g., insertions/deletions, generation of fusion proteins, or posttranslational modifications]. Along similar lines, another study has suggested that any relationship between neoantigens and patient responsiveness to checkpoint blockade therapy take into account both MHC class I and class II neoepitopes (5).

Van Allen et al. further analyzed gene expression in a subset of tumors and demonstrated that an increased transcript expression of PD ligand 2 (PD-L2) and an immune “cytolytic” gene signature correlated with neoantigen load and tumor response to ipilimumab. Interestingly, the expression of CTLA-4 was an indicator of response, with both CTLA-4 and PD-L2 likely expressed in the tumor-infiltrating immune cells. These findings may reflect ongoing preexisting T cell responses, as an inflamed tumor microenvironment prior to ipilimumab treatment has been associated with response (11). Thus, even though the numbers of mutations and neoantigens correlate with response, it may be important to consider biomarkers that reflect a tumor's immune context (12). Indeed, the expression of PD-L1 on tumor-infiltrating immune cells has been put forward as a possible biomarker in PD-L1 mAb treatment (13). Other factors such as increased density and decreased diversity in antigen specificity of T cells within the tumor may also provide predictive value, as observed in melanoma patients treated with PD-1 mAb (14).

Van Allen et al. also investigated whether any recurring neoantigens or neoantigens with shared features might better predict ipilimumab response. Nearly all neoantigens were patient-specific and most likely represent “passenger” mutations that do not directly contribute to tumorigenesis. Examination of neoantigens did not reveal any features or motifs exclusive to responders.

Notably, the results of Van Allen et al. do not support the notion of a four–amino acid (tetrapeptide) motif within the predicted neoepitopes of melanoma patients who received durable clinical benefit from ipilimumab (3). Van Allen et al. combed for this motif within their data set and found no enrichment of this signature in the clinical benefit cohort. Questions have been raised (15) about the validity of this tetrapeptide signature, as it violates widely accepted rules governing antigen presentation and T cell recognition. Further examination should clarify whether the concept is indeed correct or dangerously misdirecting the field.


View Abstract

Stay Connected to Science

Navigate This Article