Presence of Somatic Mutations within PIK3CA, AKT, RAS, and FGFR3 but not BRAF in Cisplatin-Resistant Germ Cell Tumors

Although most advanced germ cell tumors (GCT) can be cured with initial cisplatin-based chemotherapy, 20% to 30% of patients demonstrate cisplatin resistance and require more intensive salvage chemotherapy programs or desperation surgery with 10% to 20% ultimately dying of this disease (1). For such patients, novel therapies are needed.

Although other tumor types have benefited greatly from the introduction of molecularly targeted therapies into oncologic practice, few novel targets and no new agents have been identified as effective against refractory GCT. Furthermore, mutations that confer chemotherapy resistance in somatic tumor types such as TP53 alterations have been rarely observed in GCT and, when present, are not universally associated with cisplatin resistance (2–4).

In 2009, Honecker and colleagues challenged this prevailing notion when they reported that 26% of 35 patients with cisplatin-resistant GCT had BRAF mutations as compared with 1% of 100 unselected GCT cases (5). However, this finding has not been validated. One study of adult GCT (6) found an incidence of BRAF mutations of only 5% (9% among nonseminomas and 0% among seminomas) and a report of pediatric and adolescent GCT (7) identified no BRAF mutations among 66 patients, including 18 diagnosed at age ≥13 and 15 with progressive disease after primary therapy. One additional series of 65 testicular GCT and four GCT cell lines also failed to identify any BRAF mutations (8).

In this study, we attempted to validate the finding of frequent BRAF mutations among resistant GCT (in comparison with sensitive tumors), which would support routine BRAF testing in this population and potentially allow treatment with a BRAF inhibitor in mutation-positive patients. Mutations in KRAS, HRAS, NRAS, AKT1, PIK3CA, and FGFR3 were also interrogated to gain a better understanding of the spectrum of alterations in GCT, thereby providing a rationale to test novel agents targeting these signaling pathways in patients with resistant GCT.

Patients were eligible to have their tumors analyzed if they had a diagnosis of GCT (of any primary site) confirmed by pathologic review at Memorial Sloan Kettering Cancer Center (New York, NY), received cisplatin-based chemotherapy for advanced disease, and had adequate fresh-frozen or paraffin-embedded tumor tissue available for DNA extraction. In addition, information about patient outcome following cisplatin-based chemotherapy had to be available. Tumors could have been obtained from primary or metastatic sites and before or after chemotherapy. Written informed consent was obtained from all patients to allow use of their tumor for research purposes and the study was approved by the Institutional Review Board.

A tumor was classified as “cisplatin-sensitive” if the patient received one line of cisplatin-based chemotherapy [typically etoposide plus cisplatin (EP) or bleomycin plus etoposide plus cisplatin (BEP)] with or without surgery and achieved a complete response or partial response with negative tumor markers with no evidence of relapse as of their last follow-up before study entry (minimum of 2 years after completion of chemotherapy). For patients who underwent post-chemotherapy surgery, no residual GCT elements other than teratoma could be present within the pathologic specimen.

A tumor was classified as “cisplatin-resistant” when a patient developed progressive disease after first-line cisplatin-based chemotherapy (incomplete response or relapse) or if viable nonteratomatous GCT was identified at post-chemotherapy surgery. To select the most resistant cases, tumors from patients who died of progressive GCT were prioritized for inclusion.

Nine GCT cell lines with varying degrees of sensitivity and resistance to cisplatin were subjected to the Sequenom assay, including eight derived from nonseminomas (NT2101, 27X1, NCC15, 169A, 218A, 228A, 2102EP, and TERA-1) and one from a seminoma (TCAM2). Of note, 27X1, 169A, 218A, and 228A were generated in our laboratories whereas the remaining cell lines were obtained from commonly available sources. Maintenance and relative cisplatin resistance of the nonseminoma cell lines were previously described (2).

Thirty-six samples were from fresh-frozen tissue (FFT) and 39 from formalin-fixed paraffin-embedded blocks. All samples were reviewed by a board-certified GU pathologist (H. Al-Ahmadie or V.E. Reuter) to confirm the diagnosis and estimate tumor content. Whenever possible, samples with ≥70% tumor content were used. In addition, in mixed nonseminomas containing secondary somatic malignant differentiation, an effort was made to select blocks containing predominantly viable GCT elements rather than the secondary somatic malignant histology. For the FFT samples, DNA was readily available, having been extracted for use in our previous studies (9). DNA was extracted from paraffin samples using the Qiagen DNeasy Kit. All concentrations were measured using the Nanodrop 2000 spectrophotometer.

Tumors were screened for hotspot alterations using a custom iPLEX assay (Sequenom). Briefly, multiplexed PCR and extension primers were designed for a panel of known mutations (Supplementary Table S1). After PCR and single-nucleotide extension reactions were performed, the resulting extension products were analyzed using a matrix-assisted laser desorption/ionization–time-of-flight mass spectrometer, as previously reported (10).

Bidirectional Sanger sequencing of mutations detected by Sequenom was performed as previously reported (10, 11). Primer sequences are available upon request.

This analysis was designed as an exploratory study to quantify the frequency of BRAF mutations in patients with advanced GCT who required treatment with cisplatin. Because prior studies had demonstrated a higher rate of BRAF mutations in cisplatin-resistant tumors as compared with cisplatin-sensitive tumors, we sequenced samples from 50 resistant and 25 sensitive tumors. With an exact 95% confidence interval (CI), the proportion of patients whose tumor harbored a BRAF mutation was calculated overall and by cisplatin-sensitivity status (sensitive or resistant). Comparisons of mutation status by clinical factors were performed using the Fisher exact test or the Wilcoxon rank-sum test.

Of 75 patients, 73 had nonseminoma and 2 had seminoma. The median age at diagnosis was 30 (range, 14–60 years). Primary site was testis in 62 and mediastinum in 13. Twenty-five patients had cisplatin-sensitive tumors and 50 had cisplatin-resistant tumors. Additional patient characteristics divided by cisplatin sensitivity status are provided in Table 1.

With a median follow-up of 10.0 years (range, 4.0–16.7) from initiation of chemotherapy, all patients with cisplatin-sensitive tumors remained alive and disease-free at last follow-up. In contrast, 49 of 50 patients with resistant tumors have died, with 48 deaths due to progressive GCT. The cause of death in one patient was unknown. Median survival from initiation of chemotherapy for cisplatin-resistant patients was 1.6 years (range, 0.56–14.98).

In five cases, DNA quality was inadequate for evaluation, resulting in 70 samples suitable for mutation analysis, 43 of which were primary tumors (34 testicular and nine mediastinal) and 27 of which were metastatic samples (predominantly retroperitoneal lymph nodes). Of 27 metastatic samples, the primary tumor was testis in 23 and mediastinum in four.

Sixty-eight tumors were nonseminoma and two were seminoma. Of the 68 nonseminomas, histology was mixed in 32 cases, a solitary histology in 34 cases, and two cases were nonseminoma, not otherwise specified (NOS). Teratoma with secondary somatic malignant differentiation was present in five samples with the transformed malignancy consisting of sarcoma in three and primitive neuroepithelial tumor (PNET) in two cases.

Twenty-seven tumors were obtained before chemotherapy and 43 after ≥1 line of prior chemotherapy. Fourteen of 43 tumors obtained post-chemotherapy were pure teratoma (n = 12) or a combination of teratoma plus teratoma with malignant transformation (n = 2) without other viable GCT elements. Sample characteristics divided by cisplatin responsiveness are listed in Table 2.

Nineteen mutations were found in 16 (23%) of the 70 tumors evaluated, including one tumor with three mutations and a second with two mutations (Tables 3 and 4). The remaining 14 tumors harbored one mutation each. Mutations were found in FGFR3 (n = 9), KRAS (n = 3), AKT1 (n = 3), PIK3CA (n = 3), and HRAS (n = 1). Mass spectrometry traces with Sanger confirmation for FGFR3, AKT1, and PIK3CA mutations are illustrated in Supplementary Fig. S1. No mutations were identified in BRAF or NRAS. Of nine FGFR3 mutations, six were R248C, two A391E, and one S249C. The proportion of patients with FGFR3 mutations was equivalent among cisplatin-sensitive (13%) and -resistant tumors (13%; Table 3).

In contrast with FGFR3, all other mutations were identified only within cisplatin-resistant tumors (Table 3). All three AKT1 mutations were E17K, whereas there was heterogeneity among KRAS (G12D, G12C, and G13D) and PIK3CA mutations (one E542K and two E545K). The lone HRAS mutation was G12S.

Overall, mutations were found in 13% of cisplatin-sensitive tumors (all FGFR3) and 28% of cisplatin-resistant tumors. Coexisting mutations included FGFR3 and AKT1 in 1 patient and FGFR3, AKT1, and HRAS in another patient (Table 4).

There was no apparent association between mutation frequency and whether the sample was taken from the primary tumor or metastatic lesion (26% vs. 19%; Table 5). A higher proportion of mutations was observed among mediastinal as compared with testicular primary tumors (31% vs. 21%) but this did not reach statistical significance (Table 5). With the exception of HRAS [identified in one tumor from a patient with primary mediastinal nonseminomatous germ cell tumor (PM-NSGCT)], a subset analysis excluding patients with PM-NSGCT did not significantly affect the frequency of any mutation. Interestingly, both patients with multiple mutations in the same tumor had PM-NSGCT and their tumors contained FGFR3 and AKT1 mutations. All patients with solitary FGFR3 mutations had tumors of testicular origin (Table 4).

Mutations were observed across all histologies, including 14 of 68 patients with nonseminoma and both patients with pure seminoma (Table 5). Thirteen of the 16 samples with mutations had either a yolk sac tumor or teratoma component (yolk sac tumor in five, teratoma in three, and both in five). One FGFR3 and one KRAS mutation were found within tumors that contained teratoma with malignant secondary somatic differentiation (PNET in one, rhabdomyosarcoma in the other). In both cases, the tumor block selected for sequencing did not contain any secondary somatic malignant elements (embryonal carcinoma only in one and teratoma and embryonal carcinoma in the other; Table 4).

Similarly, there was no association between mutation frequency and sample timing relative to chemotherapy with six (22%) of 27 pretreatment samples and 10 (23%) of 43 posttreatment samples harboring mutations. Mutations were observed in tumors from patients in all International Germ Cell Cancer Collaborative Group (IGCCCG) risk groups, including 9 of 29 good-risk, 1 of 12 intermediate-risk, 4 of 27 poor-risk, and both unclassifiable patients (Table 5). Notably, mutations were found in 7 of 11 tumors from patients with cisplatin resistance despite having IGCCCG good-risk disease. Finally, mutations other than FGFR3 (PIK3CA, AKT1, KRAS, and HRAS) were associated with cisplatin resistance (P = 0.023) and the presence of yolk sac tumor without teratoma (P = 0.018). A trend toward more frequent non-FGFR3 mutations was also observed for mediastinal primary tumors (P = 0.055; Supplementary Table S2). None of the nine GCT cell lines contained mutations in BRAF or the other genes assessed.

In this study, mutations in several well-known oncogenes, including FGFR3, PIK3CA, AKT1, KRAS, and HRAS, but not BRAF or NRAS, were identified in tumors from patients with advanced GCT. Historically, mutations in well-characterized oncogenes and tumor suppressors have been thought to be extremely uncommon events in GCT. For example, two of the most commonly mutated genes across all malignancies, TP53 and RB1, are rarely altered in GCT (3, 4). In fact, the integrity of the p53 and Rb pathways in GCT, facilitating induction of apoptosis in response to DNA damage, has been postulated to explain the superior sensitivity of these tumors to cisplatin compared with other malignancies (2, 12).

However, prior evaluations for mutations within GCT have been limited by small sample sizes, lack of annotated clinical information to correlate with mutational results, and heterogeneous techniques for mutation detection. Overall, a low number of samples derived from patients with cisplatin resistance have been included in these series. In the only prior report to focus on cisplatin-resistant GCT, Honecker and colleagues identified BRAF mutations within 9 (26%) of 35 cisplatin-resistant GCT compared with one (1%) of 100 randomly selected GCT (cisplatin-resistance unknown; ref. 5). The authors hypothesized that BRAF mutations occur frequently in relapsed or refractory GCT and may play a role in the development of cisplatin resistance. Of note, cisplatin resistance in GCT is defined differently than in other solid tumor malignancies. Because its high cure rate, anything short of cure (incomplete response or relapse), no matter how significant the initial response, is considered indicative of cisplatin resistance.

We sought to confirm this finding in a larger cohort of cisplatin-resistant GCT. However, none of the 70 tumors (including 46 cisplatin-resistant) or nine GCT cell lines tested harbored a BRAF mutation. The discrepancy between our results and Honecker's findings could relate to differences in mutational analysis techniques, differences in definition of cisplatin resistance, or ethnic differences within the patient populations screened (German vs. American), as observed in other malignancies such as prostate cancer (13). However, it is notable that our findings are consistent with the other three series that evaluated GCT for BRAF mutations (Table 6; refs. 6–8). Two of these failed to identify any BRAF mutations and the remaining study reported a rate of only 5% among 62 samples. In addition, two of these studies were conducted in Germany (6, 7), and the other in an ethnically similar population in the United Kingdom (8). Although these reports either did not include cisplatin-resistant tumors or did not have clinical information available to complement the molecular findings, it is likely based upon our data that BRAF mutations are less common in GCT than suggested by Honecker and colleagues.

Of the seven genes evaluated, FGFR3 was the most frequently mutated despite no prior reports of FGFR3 mutations in GCT (although other FGFR3 alterations have been reported). As observed in other malignancies such as bladder cancer (14), the presence of an FGFR3 mutation in GCT was not associated with cisplatin sensitivity or resistance (identified in 13% of samples in both groups). In addition, to our knowledge, we report for the first time AKT1- and PIK3CA-activating mutations in patients with GCT. Importantly, all mutations in PIK3CA and AKT1 were present only within cisplatin-resistant tumors and mutations in these two genes were mutually exclusive as expected for genes in which protein products are found within the same oncogenic pathway.

The finding of PI3K/AKT mutations is notable, as prior studies in several tumor types have identified mutational activation of the PI3K–AKT pathway as a potential mechanism of resistance to cytotoxic chemotherapy, including cisplatin (15–18). Interestingly, phosphorylated AKT (pAKT) was recently demonstrated to play a role in cisplatin resistance in a preclinical model of GCT through preferential shuttling of p21 to the cytoplasm in which it binds to cyclin-dependent kinase 2, thereby inhibiting the apoptotic response to cisplatin-induced DNA damage (19). Inhibition of pAKT either directly or by blocking PI3K signaling resulted in reversal of cisplatin resistance with a marked increase in apoptosis (19). Thus, although the number of samples with any particular mutation was small, our results support the hypothesis that mutations in these genes could be linked to cisplatin resistance. As numerous selective inhibitors of PI3K and AKT are now in clinical testing, these results, if further validated, should prompt clinical trials of PI3K/AKT–selective inhibitors in combination with chemotherapy in patients with poor-risk previously untreated GCT and those with established cisplatin resistance. The data also suggest a need to develop novel cell lines more representative of the genetic profile of cisplatin-resistant GCT to serve as in vitro models for testing novel targeted therapies.

In addition to cisplatin resistance, mutations other than FGFR3 were associated with tumors that contained yolk sac elements but lacked teratoma (although observed across all histologies). Importantly, our analysis confirmed that these mutations represent events occurring within GCT components rather than being acquired during somatic malignant differentiation of teratoma. Only two of 16 tumors with mutations had a secondary somatic malignant component and in both cases, a block lacking the secondary somatic cancer was used for sequencing. There was also a strong trend toward a higher rate of non-FGFR3 mutations among patients with mediastinal as compared with testicular primary tumors. Even among poor-risk GCT, primary mediastinal nonseminomas tend to have higher rates of cisplatin resistance and more unfavorable outcomes. A higher predilection toward acquiring mutations could partially explain their more aggressive phenotype.

In contrast, there was no difference in mutational frequency among post-chemotherapy (vs. pre-chemotherapy) samples, suggesting mutation acquisition was not a consequence of chemotherapy exposure. However, such a conclusion is limited by the small number of samples studied and lack of paired primary/metastatic and pre-/post-chemotherapy tumors. Similarly, the lack of paired samples and evaluation of multiple areas within any particular sample prevents us from excluding intratumoral heterogeneity, which may have influenced mutational frequency, particularly within samples with mixed histologies. Finally, although the overall frequency of mutations identified among GCT samples in our study is lower than other tumor types such as bladder (mutation rate 36% for the 7 genes assessed here; ref. 20), colorectal (11), lung (21), and melanoma (22), such comparison is limited by our assessment for only hotspot mutations whereas more comprehensive assessment has been performed for these other tumor types. Use of an unbiased sequencing technique would be required to define the actual frequency of alterations on a global genomic level in GCT. On the basis of our results, we estimate that such a study will require 76 tumors (38 sensitive and 38 resistant) to have 80% power (α = 0.05) to detect a 20% difference (2% vs. 22%) in the frequency of PI3K–AKT–mTOR pathway mutations between sensitive and resistant samples.

We also performed an exploratory analysis (Supplementary Fig. S2) to evaluate the association between mutation status and survival. Although a trend toward worse survival was observed for patients whose tumors harbored mutations, given our cohort was enriched for cisplatin-resistant tumors and mutations were more common among cisplatin-resistant samples, this analysis was inherently biased. This was easily demonstrated by confining the analysis to only cisplatin-resistant tumors that resulted in no differences in survival between patients with and without mutations (Supplementary Fig. S3).

In summary, we were unable to validate the finding of frequent BRAF mutations in American patients with cisplatin-resistant GCT, despite evaluation of the largest cohort of cisplatin-resistant cases studied to date. However, we found mutations in other oncogenes in 28% of cisplatin-resistant GCT, including for the first time, FGFR3, AKT1, and PIK3CA. Although FGFR3 mutations were identified within both sensitive and resistant tumors, all other mutations were exclusive to platinum-resistant tumors. A more detailed genetic analysis of cisplatin-resistant GCT for these and other mutations using next-generation sequencing technologies will hopefully define their prevalence in this disease. Evaluation of paired primary/metastatic and pre-/post-chemotherapy samples will also aid in determining whether a certain proportion of GCT contain an intrinsically cisplatin-resistant clone before chemotherapy or whether resistance is more commonly the result of tumor acquisition of new mutations on treatment. Ultimately, such findings could lead to the identification of novel drug targets for the treatment of patients with recurrent, cisplatin-resistant disease, and the development of combinatorial therapies that prevent or delay the emergence of drug-resistant clones.

No potential conflicts of interest were disclosed.

Conception and design: D.R. Feldman, R.S. Chaganti, D.B. Solit

Development of methodology: D.R. Feldman, G. Iyer, D.B. Solit

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.R. Feldman, G. Iyer, H. Al-Ahmadie, G.J. Bosl, R.S. Chaganti, D.B. Solit

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.R. Feldman, G. Iyer, L. Van Alstine, S. Patil, H. Al-Ahmadie, G.J. Bosl, D.B. Solit

Writing, review, and/or revision of the manuscript: D.R. Feldman, G. Iyer, L. Van Alstine, S. Patil, H. Al-Ahmadie, V.E. Reuter, G.J. Bosl, R.S. Chaganti, D.B. Solit

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.R. Feldman, G. Iyer, D.B. Solit

Study supervision: D.R. Feldman, L. Van Alstine, D.B. Solit

This work was supported by the Sidney Kimmel Center for Prostate and Urologic Cancers (New York, NY)

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.