Coexistence of PIK3CA and Other Oncogene Mutations in Lung Adenocarcinoma–Rationale for Comprehensive Mutation Profiling Free

This article is featured in Highlights of This Issue, p. 255

The identification and targeting of specific oncogenic driver mutations have revolutionized the treatment of lung adenocarcinoma. Although mutations in Kirsten rate sarcoma viral oncogene homolog (KRAS) were identified decades ago and remain a target of investigation, the first therapeutic advance in lung cancer was the discovery of mutations in the epidermal growth factor receptor (EGFR) gene in patients who had experienced dramatic benefit from treatment with EGFR tyrosine kinase inhibitors (1–3). Since then, a number of other oncogenic driver mutations (missense mutations, insertions, and deletions) have been identified in v-Raf murine sarcoma viral oncogene homolog B1 (BRAF), phosphoinositide-3-kinase catalytic alpha polypeptide (PIK3CA), and human epidermal growth factor receptor 2 (HER2). In addition, the discovery of the anaplastic lymphoma kinase (ALK) fusion protein in lung cancer (4, 5) led to the rapid identification of the ALK inhibitor crizotinib revealing a 61% overall response rate in patients with ALK rearrangements (6, 7) and leading to expedited approval by the U.S. Food and Drug Administration. This early experience has bolstered the growing enthusiasm for our ability to target therapy for an individual based on the presence of a specific driver mutation in their tumor specimen.

The characteristics of EGFR-mutant lung cancer and ALK-rearranged lung cancer have been well described (4, 8). Recent reports have summarized the characteristics of BRAF- (9) and HER2-mutant populations (10). However, the patients with tumors harboring mutations in PIK3CA have not been characterized.

PIK3CA encodes the p110α subunit of phosphoinositide 3-kinase (PI3K), an integral signaling molecule in the pathways driven by growth factor receptors such as HER/ERBB2. Activated PI3K phosphorylates AKT and leads to downstream activation of mTOR that is essential for cell survival and proliferation. Because activating mutations and overexpression of PI3K are known to be oncogenic, the PI3K pathway has been intensively studied (11).

Multiple mutations have been identified in PIK3CA, the oncogene that encodes the p110α subunit of PI3K (12). The mutations that occur with regularity and in highly conserved regions of the gene lead to amino acid substitutions in the helical binding domain encoded by exon 9 (E542K and E545K) and in the catalytic subunit of p110α encoded by exon 20 (H1047R or L). Mutations in the helical binding domain interfere with p85 binding and allow activation of PI3K. The mutations in the catalytic subunit are thought to increase kinase activity (13). These specific mutations have been shown to be sufficient for tumorigenesis both in vivo and in vitro (14, 15).

The rate of PIK3CA mutations reported in non-small cell lung cancer (NSCLC) is estimated from 1% to 4% (16–19). In the United States alone, this represents 9,000 patients per year who may benefit from a therapy targeting the PI3K pathway. Interestingly, unlike other oncogenic driver mutations in lung adenocarcinoma that are rarely found in squamous cell carcinoma, PIK3CA has been reported to be amplified and mutated in squamous cell carcinoma as well as adenocarcinoma of the lung (20). Preliminary data in the genotyping of squamous cell carcinoma carried out at our institution confirms the occurrence of PIK3CA mutations in 2% of squamous lung cancer (21).

As part of the ongoing Memorial Sloan-Kettering Lung Cancer Mutation Analysis Project, we have routinely tested for PIK3CA mutations in patients with adenocarcinoma of the lung since 2009 (22). Given the prevalence of PIK3CA mutations in other diseases, multiple drugs targeting PI3K and AKT/m-TOR are in development, including trials targeting PIK3CA mutations in lung cancers, we evaluated their clinical and molecular characteristics to learn more about this patient population.

Between January 2009 and June 2010, all patients evaluated by the thoracic medical oncology and surgery services were offered participation in an institutional tissue analysis program entitled the Lung Cancer Molecular Analysis Project. In patients with sufficient tissue, assessment for driver mutations was carried out in 9 genes: EGFR, KRAS, BRAF, HER2, PIK3CA, v-akt murine thymoma vial oncogene homolog 1 (AKT1), v-ras neuroblastoma viral oncogene homolog (NRAS), dual specificity mitogen-activated protein kinase kinase 1 (MEK1), and ALK. EGFR exon 19 deletions were identified through a PCR-based assay (23). EGFR exon 20 and 21 mutations, as well as activating mutations in KRAS, BRAF, HER2, PIK3CA, AKT1, NRAS, and MEK1, were assessed by a mass spectrometry–based nucleic acid assay on the Sequenom platform. The platform was designed to include mutations in these genes that have been reported as activating. All detected mutations were confirmed by direct sequencing. These methods have been previously described (24). Rearrangements involving ALK were determined by the ALK breakpoint FISH assay (Vysis LSI ALK Dual Color).

Clinical characteristics were obtained from the medical record. Smoking definitions are as follows: never (<100 cigarettes lifetime), current (active smoker within the past year), former (>100 cigarettes lifetime and no tobacco use within the last year). Pathologic stage was determined at the time of surgery according to the American joint commission on cancer (AJCC), 7th edition tumor-node-metastasis staging system (25). Binary correlative variables were evaluated with the Fisher exact test; continuous variables were evaluated with the Wilcoxon signed-rank test. Overall survival was calculated among patients diagnosed with stage IIIB/IV lung adenocarcinoma using the Kaplan–Meier method. Disease free survival is calculated from the date of surgery. Patients were followed from the date of diagnosis of stage IIIB/IV disease until death or the last available follow-up. Group comparison was done with the log-rank test. All research was conducted under appropriate Institutional Review Board/Privacy Board protocols and waivers.

Twenty-three of 1,125 (2%, 95% CI: 1–3) patients had a mutation in PIK3CA (10 E545K, 2 E542K, 3 H1047L, and 8 H1047R; Fig. 1). There were no mutations identified in R88, N345, C420, and M1043. The clinical characteristics of the PIK3CA-mutant patients are presented in Table 1. Sixteen of 23 (70%, 95% CI: 49–86) had coexisting mutations in other oncogenes—10 KRAS, 1 MEK1, 1 BRAF, 1 ALK rearrangement, and 3 EGFR exon 19 deletions (Fig. 2). This is 1% (95% CI: <1–4) of the 260 EGFR-mutant cases, 3% (95% CI: <1–16) of the 34 ALK-rearranged cases, and 3% (95% CI: 1–5) of the 355 KRAS-mutant cases.

Patients had a median follow-up of 13 months (range 3–60 months). Of the 9 patients with early-stage disease, 5 received neoadjuvant chemotherapy and all underwent a complete resection. Four of the 5 patients treated with neoadjuvant chemotherapy received a cisplatin or carboplatin-based doublet. The other patient had the comutation BRAF V600E and had a marked treatment response to neoadjuvant gefitinib on a clinical trial (26). Five patients have had disease recurrence and 3 have died of disease. The mutation data, treatment course, pathologic stage, and survival of the patients are presented in Table 2.

The median survival of the 14 patients with stage IIIB and IV disease was 21 months. In this small group of patients, there was a shorter median survival in patients with a coexisting mutation (EGFR, KRAS, BRAF, and ALK) versus those with mutations in PIK3CA alone, median 13 versus 27 months (P = 0.03). Three patients had EGFR exon 19 deletions. One patient treated with first-line erlotinib had a prolonged radiographic partial response and then developed T790M-mediated resistance after 15 months on erlotinib. The other patient treated in the first-line had a partial radiographic response of 5 months duration and did not undergo a repeat biopsy at the time of progression. The third EGFR-mutant patient did not respond to second-line erlotinib; he had no evidence of T790M. The rebiopsy samples were not tested for persistence or loss of the PIK3CA mutation. The patient with an ALK rearrangement was initially treated with erlotinib with upfront progression and subsequently progressed again on docetaxel before being treated with crizotinib with stable disease as the best response. The remainder of patients were treated with standard first-line treatments (Table 2).

There was no difference in the stage or the frequency of coexisting mutations between patients with mutations in the PIK3CA kinase versus helical domain. Mutations in the kinase domain of PIK3CA occurred with higher frequency in patients who were never smokers (P = 0.009; Table 3).

Consistent with the published literature, we have confirmed that PIK3CA mutations occur in approximately 2% of patients with lung adenocarcinoma. Although single cases of adenocarcinoma harboring PIK3CA mutations and comutations have been previously reported (27–30), we found that the majority of tumors with PIK3CA mutations had another driver mutation as well (Fig. 2). This is in contrast to the mutual exclusivity of driver oncogene mutations seen in adenocarcinoma of the lung harboring EGFR, KRAS, and ALK, raising the possibility of tumor heterogeneity, though the high frequency of coexiting mutations, lack of 2 evident primary tumors microscopically, and the data regarding PIK3CA mutation in other diseases (31–33) make the explanation of tumor heterogeneity unlikely. Although 2 cases of PIK3CA mutation acquisition have been reported after the development of acquired resistance to erlotinib in EGFR-mutant lung cancer (34), in this series the mutations were present before treatment with targeted therapies. Additional data are required to characterize the effect of concurrent PIK3CA mutations on responses to erlotinib and crizotinib in patients harboring EGFR mutations and ALK rearrangements.

Mutation detection in this study makes use of a mass spectrometry–based system (Sequenom). This assay was designed to detect known “hotspot” mutations in specific oncogenes based on the published literature and available databases. Some mutations are confirmed by direct Sanger sequencing but full sequencing is not carried out on all specimens. Although this directed approach may miss unknown point mutations that would be detected by direct sequencing, it is optimal for detecting recurrent oncogene mutations that are known to be activating and minimizes the identification of new mutations of unclear significance. Our panel detects more than 90% of the PI3KCA mutations reported in all histologies of lung cancer in the COSMIC database.

Beyond lung cancer, PIK3CA mutations have been identified in breast, ovarian, endometrial, and colorectal carcinomas. Although the frequency of these mutations has been defined (Fig. 3), the influence of PIK3CA mutations on pathogenesis, prognosis, and response to therapy is not uniform across disease types or between studies. An example of this is the effect of PIK3CA mutation status on the efficacy of cetuximab in metastatic colorectal cancer. Although Prennen and colleagues found no effect of PIK3CA mutation on response to cetuximab with or without irinotecan (35), DeRoock reported an inferior response rate to cetuximab plus chemotherapy in patients harboring mutations in PIK3CA exon 20 but not the more commonly mutated exon 9 (32). In breast cancer, response to trastuzumab in PIK3CA-mutant HER2⁺ breast cancer cell lines is inferior to PIK3CA wild-type HER2⁺ cell lines only with coincident PTEN loss (36) but not with intact PTEN (37) The coexistence of PIK3CA mutations with mutations in KRAS, NRAS, and BRAF has been shown in colorectal adenocarcinoma (32) and with KRAS in pancreatic adenocarcinoma and ovarian carcinoma (31). Initial reports in endometrial carcinoma claimed that mutations in PIK3CA and KRAS were mutually exclusive (38), but more recent studies have found both KRAS mutations and PTEN loss in patients with PIK3CA mutant endometrial carcinoma (33). In the absence of comprehensive mutational profiling, this heterogeneity between diseases and mutational profiles, may explain the seemingly contradictory clinical outcomes described above.

Independent of treatment efficacy, mutations in distinct domains of PIK3CA may impart unique biologies. Studies in breast cancer have found clinical differences between patients with mutations in the helical and kinase domains with fewer lymph node metastases in individuals with mutations in the kinase domain (39) and inferior overall survival in those with mutations in the helical domain (40). These observations are further supported by findings in soft tissue sarcoma in which downstream activation of AKT is higher in tumors with helical domain mutations than those with kinase domain mutations (41). Our cohort of PIK3CA-mutant lung cancer patients has too many coexisting mutations to allow for comparison of outcomes between domain-specific mutation populations, although we can conclude that the never smokers were more likely to have mutations in the kinase domain. Interestingly, similar to the findings in TP53- and KRAS-mutant adenocarcinoma of the lung in which never smokers were more likely to harbor transition mutations (substitution purine for purine or pyrimidine for pyrimidine) and not transversion mutations (substitution pyrimidine for purine or purine for pyrimidine; refs. 42, 43), the PIK3CA kinase domain mutations, more commonly identified in never smokers, are transition and not transversion mutations.

Interestingly, in this small sample of patients with PIK3CA-mutant advanced disease, the presence of a coexisting oncogene mutation was correlated with an inferior outcome. Only one patient with a PIK3CA mutation received an experimental agent targeting the PI3K pathway, therefore, we cannot base the above average survival in this arm on effective targeted therapies. Although this sample size is too small to draw any conclusions, the survival findings are thought provoking.

Despite the uncertain effect of PIK3CA mutations on prognosis and response to standard therapies, it is an important target for drug development. We and others recommend the testing of agents specifically targeting PIK3CA only in patients with tumors that have evidence of dependency on PI3K pathway (PIK3CA mutation or PTEN loss). Cell lines with PIK3CA mutations are sensitive to downstream inhibitors such as everolimus, an inhibitor of mTOR, although this sensitivity can be abrogated by coincident mutation in KRAS (44). This is an expected yet important observation in light of the high frequency of coincident mutations found in this study. A recent report showed a 7-fold increase in response rate (35% vs. 5%) of PI3K pathway–targeted agents in patients with evidence of PIK3CA mutation in tumor specimens (31). These data call into question the use and appropriateness of testing PI3K pathway–targeted agents in patients whose tumors lack evidence of PI3K dependency and accentuates the importance of comprehensive genotyping of tumor specimens in all patients under consideration for molecularly targeted therapies.

Our data indicate that the majority of patients with lung adenocarcinoma harboring mutations in PIK3CA have coexisting mutations in other oncogenes. A shortcoming of this study is that the high throughput system used did not allow for testing of PTEN or TP53 loss; this is a step that we feel is essential moving forward and plan to incorporate into future studies to fully understand the effect of PI3K pathway alterations in lung adenocarcinoma. The timing of acquisition of PIK3CA mutation (and/or PTEN loss) in relation to that of other oncogenes and the contribution to tumor biology and response to therapy is unclear. As agents targeting various pathways including PI3K are in development, comprehensive (and perhaps sequential) mutation profiling should be carried out on tumor specimens from all patients to assess the impact of coincident mutations on the response to the targeted agents.

G.J. Riely is a consultant or is on the advisory boards of Boehringer-Ingelheim, Chugai, Tragara, and Ariad.

The authors thank L. Borsu for assistance with Sequenom assays, D. Ang for assistance with Sequenom data review, and E. Brzostowski and M. Pilloff for Sequenom data management.

This work was also supported by grants P01 CA129243 and RC2 CA148394 to M.G. Kris and Conquer Cancer Foundation of ASCO Young Investigator Award to J.E. Chaft. The MSKCC Sequenom facility was supported by the Anbinder Fund.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.