Potential clinical relevance of the biomarker

Effective responses seen after cancer immunotherapy, while often durable in nature due to immunological memory, can take time to mount and remain difficult to foresee, which complicates selection of appropriate immune-based treatments for each patient and slows down the progress in clinical trials. Therefore predictive biomarkers are urgently needed for cancer immunotherapy that would optimally, not only indicate biological activity of the drug, but also discriminate between immunologically responding and non-responding patients for each immunotherapeutic application. Good examples of the latter are immune checkpoint modulating antibodies that directly act on the immune system: programmed death-1 (PD-1) pathway, among others, mediates tumor-induced immunosuppression and monoclonal antibodies against PD-1 receptor or its ligands, such as nivolumab, are showing promise in the treatment of e.g. melanoma, lung, and renal cancer. Importantly, clinical data suggests that only patients with PD-L1-positive tumors seem to respond to antibody treatment, although warranting for confirmation in prospective

99

biomarker trials (McDermott and Atkins 2013). This would be logical since these patients feature strong PD-1 mediated immunosuppression that could be combated by pathway blocking antibodies releasing the specific breaks of the immune system.

For immunogenic immunotherapy, aimed at inducing antitumor immune responses rather than releasing the existing suppressive breaks, however, no such predictive biomarkers exist yet. Tumor lytic adenoviruses belong to this class, by featuring a strong immunostimulatory anticancer component via release of danger signals and tumor-associated antigens, which can be further boosted by immunogenic transgenes and/or adjuvant therapies. Interestingly, a very recent phase II randomized multicenter immunotherapy trial using chemotherapy together with or without oncolytic reovirus for the treatment of relapsed or metastatic pancreatic cancer failed meet its main endpoint of progression-free survival (PFS) in the total patient population, but instead showed a trend for improved PFS (39% increase in median PFS) in a sub-analysis where only KRAS-mutated patients were included (ClinicalTrials.gov Identifier: NCT01280058). Importantly, this subgroup analysis was based on previous preclinical and clinical reports suggesting that the oncolytic reovirus was mostly active in cancer cells with an activated RAS pathway (Thirukkumaran and Morris 2009). This predefined biomarker assessment was thus implied to have a major clinical significance, suggesting therapeutic benefit only in a subtype of pancreatic cancer, although the overall survival data remains to be seen. Unlike the optimal biomarker arsenal, however, the KRAS-mutation status as studied here only seems to predict biological activity of the drug (virus replication), thus ignoring the immunological aspects of the patient and tumor.

Closer to this end, a smaller pilot study focusing on oncolytic vaccinia virus treated melanoma patients recently investigated potential immunological biomarkers by gene expression analyses on peripheral blood T-cells. The authors identified immunoglobulin-like transcript 2 (ILT2) expression on immunosuppressive subset of CD8+ T-cells (CD8+FoxP3+ILT2+) as a candidate predictive biomarker of vaccinia virus immunotherapy outcome. Although not as straightforward to assess as serum HMGB1 and pending for further validation in large-scale studies, these findings are compatible with our results: In both studies, the overall frequency of (activated) CD8+ T-cells did not differ between responders and non-responders, but the antitumor T-cell activity reflected to improved outcome only in patients with low HMGB1 baseline, and the same was observed in vaccinia virus treated patients with low immunosuppressive CD8+FoxP3+ILT2+ T-cell population. It would be interesting to assess if these observations are connected by studying suppressive ILT2-positive T-cell populations in our patient series, and vice versa.

In a previous report by our group we studied if germline differences in the immunological mechanisms of patients, in particular, single nucleotide polymorphisms in Fc gamma receptors (FcgRs), could predict survival or disease control after oncolytic immunotherapy (Hirvinen et al.

2013). In that study we focused on FcgRs because of their central role in the immune defense against infections and potential impact on cancer therapy (van Sorge et al. 2003, Mellor et al.

2013). Blood samples of 235 patients treated with oncolytic viruses in the context of the ATAP were studied, with the Ethics Committee approval because receptor polymorphisms might also impact safety or efficacy of the treatments, and genotyped for two different Fc gamma receptor polymorphisms, FcgRIIa-H131R and FcgRIIIa-V158F. Comparisons between allotypes revealed that one particular genotype combination, FcgRIIIa-VV + FcgRIIa-HR, differed from others and was predictive of poor overall survival after virotherapy, but since only 10 patients carried this genotype, multivariate and further correlative analyses could not be performed. Nevertheless, our results suggested that genotypic differences in immunological charateristics may also have some

100

impact on the responsiveness to oncolytic immunotherapy, although predictive significance of FcgR polymorphisms remained to be evaluated.

In study IV, we have extended our attempt in identifying patients that possibly benefit from oncolytic immunotherapy. We have identified serum HMGB1 protein as an independent prognostic and predictive biomarker for oncolytic immunotherapy that may potentially discriminate between immunologically responding and non-responding patients. It would be interesting to study some of the aforementioned putative predictive factors together with our findings. In theory, this could further increase the sensitivity and specificity, similar to what we saw with regards to predictive impact of antitumor T-cell activations. Potentially, our findings could also apply to other immunogenic immunotherapy approaches including other oncolytic viruses, adoptive T-cell therapy, and cancer vaccines. Further mechanistic and prospective clinical studies using predefined levels of HMGB1 as a putative biomarker are warranted. Identifying novel predictive factors for emerging cancer immunotherapies is crucial not only for the individual patient benefit, but also to improve cost-effectiveness of these often expensive therapies (Geynisman et al. 2014). Ultimately, our results may help in selecting the right patients for each therapy, thus reducing costs and human suffering.

101