|Home | About | Journals | Submit | Contact Us | Français|
Gliomas consist of multiple histologic and molecular subtypes with different clinical phenotypes and responsiveness to treatment. However, enrollment criteria for clinical trials still largely do not take into account these underlying molecular differences. We have incorporated a high-throughput tumor genotyping program based on the ABI SNaPshot platform as well as other molecular diagnostic tests into the standard evaluation of glioma patients in order to assess whether prospective molecular profiling would allow rational patient selection onto clinical trials. From 218 gliomas we prospectively collected SNaPshot genotyping data on 68 mutated loci from 15 key cancer genes along with data from clinical assays for gene amplification (EGFR, PDGFRA, MET), 1p/19q co-deletion and MGMT promoter methylation. SNaPshot mutations and focal gene amplifications were detected in 38.5 and 47.1 % of glioblastomas, respectively. Genetic alterations in EGFR, IDH1 and PIK3CA closely matched frequencies reported in recent studies. In addition, we identified events that are rare in gliomas although are known driver mutations in other cancer types, such as mutations of AKT1, BRAF and KRAS. Patients with genetic alterations that activate signaling pathways were enrolled onto genetically selective clinical trials for malignant glioma as well as for other solid cancers. High-throughput molecular profiling incorporated into the routine clinical evaluation of glioma patients may enable the rational selection of patients for targeted therapy clinical trials and thereby improve the likelihood that such trials succeed.
As a result of the success of targeted therapy in several cancer types, recent drug development in oncology has shifted toward therapeutics that selectively target specific molecular alterations [1–6]. These successes were contingent upon the identification of molecular subtypes in histologically identical tumors that were susceptible to specific targeted agents. For example, subsets of advanced non-small cell lung cancers (NSCLC) with either activating mutations in the region encoding the kinase domain of the epidermal growth factor receptor (EGFR) gene or activating rearrangements of the anaplastic lymphoma kinase (ALK) gene render these tumors sensitive to specific tyrosine kinase inhibitors (TKI) targeting EGFR or ALK, respectively [2, 4, 7].
Current evidence suggests that histologically identical gliomas represent distinct molecular phenotypes with different prognoses and possibly differential responses to treatment [8–14]. For example, recurrent mutations of codon 132 of the isocitrate dehydrogenase 1 (IDH1) gene and the analogous codon (172) of the IDH2 gene are associated with younger age and longer survival for WHO grade II, grade III (anaplastic) and grade IV (glioblastoma, GBM) gliomas [10, 12, 13].
Although these and other molecular markers have strong prognostic value, genetic markers that have been prospectively validated to predict sensitivity to specific targeted therapies (predictive biomarkers) are currently lacking in gliomas. This contributes to the impression that targeted agents have failed in malignant glioma. However, it is possible that molecular subsets of responsive tumors may be masked in unselected trial populations.
We implemented a high-throughput genetic profiling platform (SNaPshot genotyping)  into a comprehensive molecular evaluation of malignant gliomas in order to prospectively detect genomic alterations and rationally direct patient enrollment onto clinical trials of molecularly targeted agents. The SNaPshot genotyping platform is a robust, highly sensitive and specific tumor genotyping assay that enables simultaneous detection of multiple somatic mutations in small quantities of tumor DNA extracted from formalin-fixed, paraffin-embedded (FFPE) tissue in a time frame of 2–3 weeks. This assay detects common mutations in 15 key cancer-related genes, several of which activate signaling pathways that are targeted by drugs currently in clinical development for malignant glioma.
Here, we describe our initial results of prospective, comprehensive molecular profiling of tumor tissues from glioma patients seen at the Dana-Farber/Harvard Cancer Center. We used SNaPshot genotyping in combination with fluorescence in situ hybridization (FISH) for gene amplification/loss and MGMT promoter DNA methylation analysis. We aimed to identify molecular markers that may predict sensitivity to targeted agents currently in investigation for malignant glioma.
From December 2009 to August 2011 we prospectively profiled 218 gliomas from 212 patients treated at Massachusetts General Hospital (MGH) and Dana-Farber Cancer Institute/Brigham and Women’s Hospital (DFCI/BWH) using SNaPshot genotyping. If tissue was sufficient, EGFR, MET, and platelet-derived growth factor receptor alpha (PDGFRA) gene amplification was assessed (details below). If clinically indicated, methylation-specific PCR (MS-PCR) to determine MGMT promoter methylation status and FISH for 1p and 19q deletion status was performed. The time from test requisition to report finalization of all molecular tests was approximately 2–3 weeks.
Testing was performed as part of routine medical care in CLIA-certified clinical laboratories. All results were obtained from the patient medical record and collected in an IRB-approved patient database. A counseling session with background information about SNaPshot genotyping was conveyed to patients undergoing testing along with the information that established treatment algorithms based on results of genetic profiling are lacking. Patients then signed an informed consent form. Patients in this cohort enrolled on recurrent GBM clinical trials were previously treated with radiation and temozolomide as described .
SNaPshot genotyping assays were performed in the MGH Translational Research Laboratory (TRL). This assay is based on the SNaPshot assay from Applied Biosystems and was developed in the TRL for clinical tumor genotyping . It can sensitively and specifically detect low-level somatic mutations in DNA extracted from FFPE tumor samples. The MGH SNaPshot assay consists of a multiplexed PCR step followed by a single-base extension reaction that generates allele-specific fluorescently labeled probes. Mutations in 68 commonly mutated loci from 15 genes (AKT1, APC, BRAF, CTNNB1, EGFR, HER2, IDH1, KIT, KRAS, MEK1, NOTCH1, NRAS, PIK3CA, PTEN, TP53) are detected in nine multiplexed reactions (Table 1).
To determine gene amplification and chromosomal deletion status, 5 μm sections of FFPE tumor material were prepared and regions containing a majority of tumor cells were selected. Three dual-color FISH assays were performed to assay the EGFR, MET and PDGFRA genes. BAC probes CTD-2113A18 (chromosome 7p EGFR locus) and CTB-13N12 (chromosome 7q MET locus) were used with a centromere 7 copy number control (CEP7; Abbott) and RP11-58C6 (chromosome 4q PDGFRA locus) was used with a centromere 4 copy number control (CEP4; Abbott). Signal quantitation was used to generate gene/ copy number control ratios. A ratio of >2.0 was considered amplified. Focal amplification was considered amplified and was specified in the diagnosis. EGFR amplification status at DFCI/BWH was determined by chromogenic in situ hybridization (CISH) with the SPOT-Light EGFR amplification probe (#84-1300) and CISH kit (#84-9246) (Invitrogen). Samples with greater than ten EGFR probe signals per nucleus in >50 % of the tumor cells were considered high level amplification. For data analysis, tumors positive by FISH or CISH were considered amplified. For clinical trial enrollment, FISH was used to confirm CISH results.
1p and 19q status was determined at MGH using two separate dual-color FISH assays using the Vysis 1p36/1q25 and 19q13/19p13 FISH Probe Kit (Abbott). Signal quantitation was used to generate 1p/1q and 19q/19p ratios. A ratio of ≥0.75 was considered loss and <0.75 as maintained. At DFCI/BWH, tumors were tested using disaggregated whole nuclei preparations from 50 μm FFPE sections and using a molecular probe assay for 1p/19q FISH (Abbott #04N60-020) using similar cutoffs for determination of loss.
To determine MGMT promoter methylation status, 5 μm sections of a FFPE tumor specimen were prepared and regions containing a majority of tumor cells were selected. Following manual microdissection, DNA was extracted and subjected to bisulfite treatment. MS-PCR was performed as described . MGMT promoter methylation status was not used as criteria for enrollment in the clinical trials reported herein.
Associations between the occurrences of genetic alterations were analyzed using Fisher’s exact test. All reported P values are two-sided.
A SNaPshot mutation was detected in of 84/218 tumor specimens (38.5 %) (Fig. 1; Table 2, Supplementary Table). Of the genes interrogated by SNaPshot, the most frequently mutated was IDH1, which was found in 47/218 specimens (21.6 %). IDH1 mutation was found significantly more frequently in secondary (5/8, 62.5 %) than primary GBM (12/156, 7.7 %) (p < 0.0004), and more frequently in WHO grade II and III gliomas (30/54, 55.6 %) than glioblastomas (17/164, 10.4 %) (p < 1 × 10−10). In the grade II and grade III glioma subset, ten of 11 oligodendrogliomas (90.9 %) had an IDH1 mutation, whereas nine of 15 (60 %) oligoastrocytomas and 11 of 28 (39.3 %) astrocytomas harbored an IDH1 mutation, indicating an association between IDH1 mutation and oligodendrogliomas versus tumors with an astrocytic component (p = 0.0147).
In gliomas of all types, IDH1 mutation was significantly associated with 1p/19q co-deletion and inversely associated with EGFR amplification. In tumors with 1p/19q co-deletion, IDH1 mutation was detected in 15 of 17 (88.2 %) cases whereas only 13 of 53 (24.5 %) tumors without 1p/19q co-deletion had IDH1 mutation (p < 1 × 10−5). Conversely, only 1 of 47 EGFR-amplified tumors (2.1 %) had IDH1 mutation whereas 11 of 66 (16.7 %) tumors without EGFR amplification were IDH1-mutated (p = 0.014). Interestingly, IDH1 mutation was concurrent with other oncogene alterations in several GBMs, including KRAS mutation, EGFR amplification, PDGFRA amplification, and one tumor with concurrent PIK3CA and PTEN mutations.
Point mutations in the PIK3CA gene were observed in 9/164 (5.5 %) of GBMs. Four tumors had mutation of R88, which is located in the P85-binding domain, and five had mutations in ‘‘hot-spot’’ regions (four H1047 kinase domain mutations and one helical domain Q546 mutation). One PIK3CA mutant anaplastic oligodendroglioma also harbored IDH1 and PTEN point mutations and two PIK3CA mutant GBMs had concurrent EGFR amplification.
Amplification of the EGFR gene was detected in 45 of 108 (41.7 %) malignant gliomas and 42 of 102 (41.2 %) of GBMs (Fig. 2). EGFR amplification was occasionally concurrent with other genetic alterations, including two tumors with PIK3CA mutation, three with both EGFR and MET amplification and one with IDH1 mutation. None of the EGFR-amplified tumors harbored 1p/19q co-deletion (0/9).
Our population included several oncogene mutations that are rare in malignant gliomas. Three GBMs had KRAS mutation (1.8 %), one CTNNB1 mutation (0.6 %), one BRAF V600E catalytic domain mutation (0.6 %), one AKT1 mutation, and one with an EGFR active site mutation (G719D) known to sensitize to EGFR inhibitors in NSCLC. Four GBMs displayed MET gene amplification (4/102, 3.9 %) (Fig. 2) and two had PDGFRA amplification.
In this dataset, there were seven matched sets of tumors that were profiled at diagnosis and at progression. Two tumors with IDH1 mutation and 1p/19q co-deletion retained these mutations at re-resection. One was a grade II oligodendroglioma that transformed into an anaplastic oligodendroglioma, and the other was an anaplastic oligodendroglioma that had similar histology at recurrence. Four matched GBM tumors had identical molecular profiles at diagnosis and at recurrence. Two had only EGFR amplification, one recurrent GBM retained the PIK3CA and TP53 point mutations detected at diagnosis, and one had no genetic alterations detected.
Two GBMs had a different molecular profile at recurrence. One EGFR-amplified tumor remained EGFR-amplified at recurrence however also developed a new PTEN point mutation in the recurrent tumor. Another GBM had both EGFR and MET amplification at diagnosis, however at recurrence both alterations were lost. Interestingly, a small area of this original tumor had oligodendroglial histology without either EGFR or MET amplification, and at recurrence the majority of the tumor contained this oligodendroglial population.
Prospective molecular profiling enabled rational selection of clinical trials for several patients. Recently, seven patients with EGFR amplification in their tumors at diagnosis were enrolled at recurrence onto a genetically-selective (EGFR-amplified) phase II trial of PF-00299804, an irreversible EGFR TKI (NCT01112527). Efficacy results are pending, however one patient who progressed within 2 months required urgent surgery for malignant cerebral edema. The tissue was molecularly profiled and intriguingly, the recurrent tumor harbored a PTEN point mutation that was not present in the original tumor at diagnosis. Another patient had a minor response and is currently stable after 14 months.
One GBM patient had MET gene amplification in her tumor at diagnosis, which occurs infrequently in GBM (5 % ). At recurrence she enrolled onto a phase I study of crizotinib, a dual MET and ALK tyrosine kinase inhibitor, for advanced solid tumor patients with MET gene amplification or ALK gene rearrangement/amplification in their tumors (A8081001). Within 2 weeks she had significant improvement in short term memory and orientation and she had a sustained 40 % reduction in the area of contrast enhancement on MRI . She ultimately progressed after 6 months.
One GBM patient who harbored both EGFR and MET amplification in his tumor at diagnosis was a potential candidate for one of two genetically-selective clinical trials; PF-00299804 for EGFR-amplified recurrent GBM and crizotinib for advanced solid cancers with MET or ALK genetic alterations. At recurrence, this patient required surgical resection due to rapid, symptomatic growth of the tumor. Surprisingly, both EGFR and MET amplification were lost in the recurrent tumor, and he therefore was not enrolled onto either trial.
We identified many newly diagnosed malignant glioma patients who may be candidates for genetically-selective targeted inhibitor clinical trials open at our institution at the time of recurrence, including EGFR-amplified patients (PF-00299804 trial), MET-amplified patients (crizotinib trial) and patients with PIK3CA or PTEN point mutations [BKM120 for recurrent GBM patients with activated PI3K pathway (NCT01339052)]. Patients in whom we detected alterations such as PDGFRA amplification are potential candidates for genetically un-selective clinical trials investigating agents such as cediranib, sunitinib and IMC-3G3.
Our institution has developed a practical yet robust platform to genotype tumors in a short time frame. This has allowed us to incorporate molecular profiling into the standard clinical evaluation of glioma patients and has enabled more rational enrollment into early-stage clinical trials.
In general, the results from our molecular profiling were consistent with previously reported prevalences. For example, the frequencies of EGFR amplification (41.2 %) and MGMT promoter methylation (53.5 %) in GBMs were similar to prior reports [11, 20–22]. IDH1 mutations were associated with secondary versus primary GBM, grade II and grade III gliomas versus GBMs, oligodendrogliomas versus astrocytic gliomas (in grade II and grade III gliomas) and 1p/19q co-deletion, whereas they were inversely associated with EGFR amplification [12, 13].
Our observed frequencies of uncommon genetic alterations were also consistent with prior reports, including MET amplification (3.9 vs. 5 % ), PIK3CA hot-spot mutation (5.5 vs. 6.6–15 % [18, 23]) and KRAS hot-spot mutation (1.8 vs. 0–1.1 % [18, 24, 25]). We detected slightly lower mutation frequencies in other genes, such as BRAF V600E mutation in GBM (0.6 vs. 3.2–6 % [24, 26]). These differences may be due to our limited sample size, however it is noteworthy that our data was obtained prospectively during routine patient care using clinical assays.
Prospective molecular profiling also revealed several potential clinical associations for further exploration, including potential predictive biomarkers, links between mutations and responses to therapy, and selection criteria for targeted therapies. We identified one genetic marker (MET amplification) that warrants further investigation in GBM. In the case of the EGFR-amplified GBM patient who required urgent surgery after rapid progression on an EGFR inhibitor, a PTEN point mutation had developed during the interval since diagnosis. Although speculative, the PTEN mutation could have had a role in her tumor’s intrinsic resistance to EGFR inhibition. Another patient’s recurrent tumor surprisingly lost amplification of two oncogenes just prior to enrollment onto genetically-selective clinical trials. Although this patient avoided treatment with therapies that he may have been resistant to and would have likely experienced toxicity from, this case raises a concern about the validity of using tumor tissue from initial diagnosis for enrollment onto targeted therapy trials at the time of recurrence.
Comprehensive molecular profiling may provide additional benefit if a resection is performed at tumor recurrence since the mutational status of tumors may change after therapy . Tissue can be analyzed to confirm the retention of previously observed alterations and to assess whether new alterations have developed. Although tumor genetic heterogeneity remains a challenge , determining the fraction or abundance of specific alterations in situ with methods such as FISH would assist the formulation of combination and sequential targeted strategies.
By applying a comprehensive assay that interrogates commonly mutated oncogenes across multiple cancer types, we identified oncogene mutations that are infrequent in GBM (e.g. EGFR G719D, BRAF, KRAS, and AKT1) although are currently being targeted in solid cancer phase I trials. Although CSF penetration of many solid cancer phase I drugs is either limited or unknown, these cases open opportunities to explore rare mutations as predictive biomarkersin GBM. TwoBRAFmutant glioma patients may be eligible for one of several BRAF-mutant selective phase I clinical trials. Three patients with KRAS mutation may be eligible for one of several trials testing MEK inhibitor and PI3K inhibitor combinations, a formula which has preclinical activity in KRAS mutant lung cancers .
A limitation of targeted sequencing platforms is that mutations at non-interrogated loci can be missed, resulting in false negative testing. For tumor suppressors, our SNaPshot panel interrogates loci that are mutated with at least 3 % frequency, however this design still covers less than 30 % of the TP53 and 10 % of the PTEN point mutations that have been reported in GBM . Our observed point mutation frequencies for TP53 (12.2 %) and PTEN (2.4 %) in GBM cases were much lower than previous reports (60–90 % for TP53 [31–33] and 14–32 % for PTEN [18, 34–36]), reflecting this limitation. However, our SNaPshot panel is intentionally weighted to interrogate oncogenes, which have activating mutations at only a few hot-spots. Oncogenes are better suited for targeted sequencing than tumor suppressors, where mutations tend to occur throughout the gene. Our panel also favors oncogenes that are being targeted either in clinical trials or with FDA-approved targeted therapies.
Sanger sequencing or next-generation sequencing methods would provide more comprehensive genetic information than our targeted sequencing panel. However, tissue quantity, cost, and time are significant limiting factors in the clinical setting. Although our method has limitations in scale and modest expense, newer approaches such as next-generation sequencing present additional challenges such as bioinformatics and technical expertise that currently limit their implementation in clinical practice. Our focused genotyping panel does not include some mutations commonly found in gliomas (e.g. non-kinase domain EGFR mutations) and includes others that are rare in glioma, however it is practical, moderately comprehensive, robust and ultimately feasible for application in a cancer center. We report examples where our profiling approach may have had clinical consequence and it provides targeted therapy clinical trials in GBM the opportunity for improved outcomes. Ultimately our approach may increase the likelihood of replicating the successes already observed with targeted therapy in other cancers.
Andrew S. Chi is supported by a Joan Ambriz American Brain Tumor Association Basic Research Fellowship and an Early Career research award from the Ben and Catherine Ivy Foundation. Ethical Standards: All results were obtained from the medical record of patients and collected in an IRB-approved patient database. Molecular testing was performed as part of routine medical care in CLIA-certified clinical laboratories. All molecular tests comply with the current laws of the USA.
Preliminary data of this manuscript was presented at the 2011 American Association of Neurology Annual Meeting in Honolulu, Hawaii.
Conflicts of interest The authors have no conflicts of interest to declare.
Andrew S. Chi, Department of Neurology, Stephen E. and Catherine Pappas Center for Neuro-Oncology, Massachusetts General Hospital Cancer Center, Yawkey 9E, 55 Fruit Street, Boston, MA 02114, USA.
Tracy T. Batchelor, Department of Neurology, Stephen E. and Catherine Pappas Center for Neuro-Oncology, Massachusetts General Hospital Cancer Center, Yawkey 9E, 55 Fruit Street, Boston, MA 02114, USA.
Dora Dias-Santagata, Translational Research Laboratory, Massachusetts General Hospital Cancer Center, Boston, MA, USA, Department of Pathology, Massachusetts General Hospital Cancer Center, Boston, MA, USA.
Darrell Borger, Translational Research Laboratory, Massachusetts General Hospital Cancer Center, Boston, MA, USA.
Charles D. Stiles, Department of Cancer Biology, Dana-Farber Cancer Institute, Boston, MA, USA.
Daphne L. Wang, Department of Neurology, Stephen E. and Catherine Pappas Center for Neuro-Oncology, Massachusetts General Hospital Cancer Center, Yawkey 9E, 55 Fruit Street, Boston, MA 02114, USA.
William T. Curry, Department of Neurosurgery, Massachusetts General Hospital, Boston, MA, USA.
Patrick Y. Wen, Center for Neuro-Oncology, Dana-Farber/Brigham and Women’s Cancer Center and Department of Neurology, Brigham and Women’s Hospital, Boston, MA, USA.
Keith L. Ligon, Center for Molecular Oncologic Pathology, Dana-Farber Cancer Institute and Department of Pathology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA.
Leif Ellisen, Translational Research Laboratory, Massachusetts General Hospital Cancer Center, Boston, MA, USA.
David N. Louis, Department of Pathology, Massachusetts General Hospital Cancer Center, Boston, MA, USA.
A. John Iafrate, Translational Research Laboratory, Massachusetts General Hospital Cancer Center, Boston, MA, USA, Department of Pathology, Massachusetts General Hospital Cancer Center, Boston, MA, USA.