While some fusions serve only as predictive biomarkers for RTKi therapy, others may play a role both as predictive and prognostic biomarkers. Exemplifying this concept is the recurrent fusion between KIAA1549and BRAF found in >50% of pilocytic astrocytomas. Presence of this fusion was associated with an improved five-year progression-free survival in patients receiving chemotherapy for this childhood brain tumor.75,76 As survival was not correlated to a specific chemotherapy, the KIAA1549-BRAF fusion represents a prognostic marker in this setting. Moreover, this fusion has relevance for targeted therapy with RTKi, which could also classify it as a predictive biomarker. However, absolute associations between the presence of a fusion and either a beneficial or detrimental drug response may not be apparent without validation. Such is the case with KIAA1549-BRAF, where BRAF inhibitors have been evaluated for drug sensitivity and pathway inhibition in fusion-positive cell lines.77 Strikingly, paradoxical activation of the MAPK pathway, as measured by phospho-ERK, and in vivo tumor growth occurred with a first-generation BRAF inhibitor, but not with a second-generation inhibitor. This suggests that the presence of fusions can elicit differential response as compared to kinase-activating mutations and even unpredicted drug responses, which need to be further explored in order to appropriately stratify patients for targeted agents.77,78
In contrast to kinase fusions that most often result in kinase activation, transcription factor fusions can lead to gene activation or to altered gene expression affecting integral mediators of cell function. This can have a dominant negative effect on the cell. In follicular thyroid cancers with PAX8-PPARγ fusions, wild-type PPARγ inhibition is regarded as the mechanism of cellular transformation.79 The relevance of PPARγ agonists in treating fusion-positive, thyroid cancer patients with advanced disease stage is now being tested in phase II trials (clinicaltrials.gov). Intriguingly, a clinical study measuring PPARγ as a surrogate marker for the overexpressed fusion protein showed an association with favorable prognosis and improved outcome.80
Approximately 50% of prostate cancer cases are known to harbor fusions in the Ets transcription factor family with TMPRSS2, a serine protease.66,73,81 Almost 80% of these fusions are TMPRSS2-ERG, which in some studies correlate with poorer prognosis and increased disease recurrence and is thus implicated as a prognostic biomarker.82,83 A genomic deletion event that fuses the 5′ region of the androgen-responsive serine protease TMPRSS2 with the 3′ region of the transcription factor ERG results in an androgen-dependent overexpression of ERG leading to an altered, oncogenic transcriptome. Due to the high frequency and specificity of TMPRSS2-ERG fusions seen in prostate cancer, its clinical applicability as a biomarker for diagnosis84–86 and treatment87,88 is also being explored.
Due to their role, recurrence, and functional implications, fusions thus serve as powerful, predictive biomarkers for targeted therapy, have prognostic implications, and display significant translational relevance.
Current Knowledge of Fusions in Breast Cancer
Large-scale studies involving transcriptome and genomic sequencing have revealed the presence of several gene fusions in the more common types of breast cancer.3,34,36 These include fusions that have been noted in other cancers (Table 3 and Table 4). Though several in-frame fusions were observed, a recurrent gene fusion is yet to be identified.3,36 With a few exceptions, the majority of fusions reported in breast cancer are uncommon and present only in a limited number of samples. Whether these are truly single events or are not observed in more samples due to limited sample size, difference in specific breast cancer subtypes included in study cohorts, and/or the appropriate technologies for detection is yet to be determined.
(To view a larger version of Table 4, click here.)