Patient selection & biomarkers
Currently in clinical oncology drug development, targeted therapies rely upon the establishment of companion diagnostics that are reproducible and robust, such that they will have a high likelihood of predicting response to a given therapy. Notable examples in breast cancer therapy include the presence of ER/PR for endocrine therapies, as well as methods for assessing Her2 overexpression/amplification by immunohistochemistry and FISH, for predicting response to Her2 directed therapies such as trastuzumab and lapatinib. More recent examples of breast cancer biomarkers include those cancers with loss of BRCA1 or BRCA2 predicting for response to PARP inhibitors [94
]. Finally, the establishment of Oncotype DX as a predictive assay, apart from its utility as a prognostic test, affords the ability to estimate the likelihood of chemotherapy benefit for a given patient [118
]. Although chemotherapy is not considered a targeted therapy for breast cancer, this last example highlights the unmet need in establishing predictive assays to choose patients who will have the highest likelihood of benefit, thus sparing needless toxicity and costs to patients and society.
For PI3K pathway inhibitors, there has been intense interest in developing companion diagnostics for the reasons mentioned above. How reliably a given test performs will depend on a number of factors, including both biological and technical challenges. Using the above examples, PIK3CA
mutations, Pten loss and AKT1
(E17K) mutations might be logical choices as positive predictors of response to PI3K, Akt and mTOR inhibitors. To date, however, no single study has demonstrated that any of these genetic alterations have strong predictive abilities. In the case of PIK3CA
mutations and Pten loss, this may be due to the activation of other pathways as previously mentioned, suggesting that drugs targeting additional pathways may be needed. Interestingly, the E17K Akt1 mutation demonstrated no obvious phenotype when introduced as a somatic mutation into breast epithelial cells via gene targeting [119
]. The reasons for this are unknown but could include effects of single versus multiple copy numbers of the E17K allele, cell-specific context of the mutation, additional unknown but requisite mutations in cooperating genes, etc. However, suffice it to say that the mutation itself will not necessarily predict for pathway activation nor, consequently, for response to targeted therapies. Intriguingly, the Braf V600E mutation, which is both the target and a predictive marker of response for the small molecule inhibitor vemurafenib, was found to be highly active in metastatic melanomas harboring the mutation [120
] but not in colorectal cancers with mutant Braf V600E. This has recently been shown to be due to the presence of EGFR in experimental models, which modulates the activity of the mutant Braf protein [121
]. This example highlights that finding predictive markers of response to targeted therapeutics in breast cancer using somatic mutations such as PIK3CA
may not be applicable for other tumor types and vice versa.
Of concern for the use of PIK3CA
mutation status as positive predictors of response to PI3K-targeted therapies is the growing realization that tumor heterogeneity may make the use of somatic alterations problematic. Similar to recent data suggesting ER and Her2 status can change in metastatic lesions [122
], reports have also emerged suggesting that PIK3CA
mutational status in breast cancer patients is prone to discordance when comparing primary versus metastatic lesions [124
]. As most trials heretofore have used archival tissue specimens to assess mutation status, current thinking suggests that this may not truly reflect the status of disease at the time of consideration for directed targeted therapies. As such, biopsy of metastatic lesions is becoming more common place for breast and other cancers. However, the inability to obtain biopsies from metastatic sites presents a conundrum for many patients and potentially limits the utility of future PI3K targeted therapies. Therefore, attempts to evaluate disease and predictors of response in a non-invasive manner have gained much interest and attention.
Ideally, a non-invasive functional test would offer great utility as a predictive assay for PI3K-directed therapies in breast cancer. Because hyperactivation of the PI3K pathway predictably leads to increases in glucose metabolism, there has been great enthusiasm for researching the utility of [18F] fluorodeoxyglucose (FDG)-positron emission tomography (PET) to not only evaluate disease burden but to simultaneously assess whether FDG-PET scanning can offer predictive value for PI3K pathway inhibitors. In theory this idea seems plausible, as any perturbation of the PI3K pathway, whether genetic, epigenetic, or otherwise would lead to a functional state of increased glucose metabolism that should correlate with response to therapy. This approach, however, may have a “fatal flaw” in that not knowing where the “lesion” is could prevent efficacy and reduce the perceived predictive power. For example, hyperactivation of the PI3K pathway by Pten loss may lead to a positive result by FDG-PET scans, but an isoform specific p110α inhibitor would likely not be effective in such a tumor. Thus, FDG-PET may be useful as an adjunct test for functional analysis of the pathway in addition to evaluation of the genetic lesion that accounts for pathway activation.
Newer technologies in development offer the potential for mutation detection from blood. As example, recent studies have shown the utility of detecting PIK3CA
mutations using blood from breast cancer patients [126
]. However, for such a test to be clinical meaningful, sensitivity and specificity must be greatly increased beyond traditional sequencing methods and PCR assays. BEAMing [127
] is a novel technique based upon digital, emulsion-based PCR with mutation detection. This technology has been shown to reliably detect rare mutant alleles found in the blood of colorectal cancer patients [129
]. More recently, retrospective and prospective studies have also demonstrated the feasibility and utility of BEAMing for evaluating the PIK3CA
mutation status of breast cancer patients (manuscript in revision). Similar to other studies, however, discordance between primary and metastatic lesions was present again, suggesting that assessing mutation status at the time of targeted therapy would likely afford the best chance of predicting response. Future work with targeted therapies against the PI3K pathway in breast cancer will almost certainly incorporate additional novel assays such that targeted therapies truly are given to the proper patients.