Prediction of Response to Anticancer Immunotherapy Using Gene Signatures

Accompanying this article are three reports involving clinical outcomes and correlative parameters associated with the administration of adjuvant melanoma-associated antigen 3 (MAGE-A3) antigen-specific cancer vaccine for the treatment of resected, stage IB-II, non–small-cell lung cancer (NSCLC) and metastatic melanoma. The first study, directed at patients with stage III or IV M1a melanoma (N 75), compared the earlier immune stimulant AS02B to a newer one (AS15) in a randomized, multicenter, phase II trial. This trial showed improved clinical outcomes with AS15, administration of which resulted in a 44% reduction in the risk of death (overall survival [OS]: 33 months v 20 months in the AS15 and AS02B groups, respectively). This observation corroborates the concept that, beyond antigen specificity, the therapeutic activity of cancer vaccines can be significantly enhanced by the coadministration of an appropriate pro-inflammatory molecule. AS15, which contains a TLR9 agonist in addition to a TLR4 agonist and the QS21 saponin (already present in the AS02B) was selected for subsequent phase III trials. 1,2 A second randomized, multicenter, phase II trial performed in patients with NSCLC (N 182) did not observe any significant difference in terms of disease-free survival and OS between the experimental and placebo groups. MAGE-A3 administration was confirmed to be well tolerated. The major limitation of this (proof-of-concept) trial is represented by the small sample size. With an estimated power of 50% to detect a difference of 10% in absolute recurrence after 30 months (for 0.2), the study was unlikely to achieve conclusive results in term of efficacy. The trial, moreover, used as adjuvant in combination with the vaccine a relatively inert immune stimulant (AS02B). Despite these limitations, an encouraging reduction in the risk of relapse of 25% in the treatment goup compared with the placebo group (P .25) was observed. In line with the majority of the studies performed so far, no significant correlation between cellular and/or humoral immune response against the administered antigen and clinical outcome was detected in either the melanoma or the NSCLC trial. The third report, which is the focus of this editorial, summarizes the biologic insights obtained through the systematic application of global transcript analysis (microarray) for the phenotyping of tumor deposits obtained before initiation of treatment. By comparing patients with melanoma who achieved clinical benefit with those who did not, Ulloa-Montoya et al identified an 84-gene expression signature associated with favorable outcome. The signature was internally cross-validated and found to be predictive of prolonged OS. The expressions of 61 out of 84 genes was assessed by qualitative real-time polymerase chain reaction (qRT-PCR), with superimposable results. The predictive role of the qRT-PCR classifier (gene signature; GS) was further confirmed in the NSCLC adjuvant study. In addition, when survival analysis in the NSCLC trial was restricted to patients bearing the favorable gene signature (GS ), the investigators detected a reduction in the risk of relapse of 58% (P .06) associated with vaccine administration, whereas no such effect was found in GSpatients. Similar trends were observed in the OS analysis. These observations indicate that immune manipulations could modify the postoperative course of (some) patients with NSCLC, as suggested earlier by the positive results of a randomized phase trial assessing the impact of adoptive therapy and human recombinant interlekin-2 (hrIL-2) in this setting. The report of Ulloa-Montoya et al does not present isolated findings but substantiates previous observations from several groups that point to the existence of a cancer immune phenotype conducive to immune responsiveness. A decade ago, serial analysis of melanoma metastases using minimally invasive biopsies was applied to patients undergoing melanoma antigen-specific vaccination in conjunction with the systemic administration hrIL-2. Biopsies of the same lesion were obtained before and after at least one cycle of treatment. This strategy allowed prospective assessment within the same lesion of the weight that the pretreatment tumor phenotype had on responsiveness to treatment. The results suggested that melanoma metastases likely to undergo complete regression following immunotherapy display an immunologically active microenvironment before treatment. Those early efforts were, however, curtailed by the limited number of transcripts represented by prototype platforms and failed to provide a global view of the program associated with immune responsiveness. More recently, application of the same serial biopsy strategy to a small cohort of patients undergoing treatment with systemic administration of hrIL-2 using a new generation genome-wide array platform allowed a more comprehensive characterization. Serial sampling of tissues before and during treatment identified in lesions about to undergo clinical regression a signature predictive of immune responsiveness in pretreatmentbiopsies thatwasqualitativelysimilar, thoughquantitatively less, to the one observable during treatment. Moreover, this signature largely overlapped with the one described by Ulloa-Montoya et aland others. This observation suggests that the cancer phenotype associated JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 31 NUMBER 19 JULY 1 2013