Post-operative stereotactic radiosurgery versus observation for completely resected brain metastases: a single-centre, randomised, controlled, phase 3 trial

Background

After brain metastasis resection, whole-brain radiation therapy (WBRT) decreases local recurrence but may cause cognitive decline. We performed this study to determine if stereotactic radiosurgery (SRS) to the surgical cavity improved local tumor tumor-free recurrence rates compared to surgical resection alone as an alternative to the need for immediate WBRT.

Methods

The main entry criteria for the study included patients >3 years of age, with a Karnofsky Performance Score ≥ 70, who were able to undergo an MRI scan and who had a complete resection of 1–3 brain metastases (the maximum diameter of the resection cavity had to be ≤4cm). Patients were assigned randomly to either SRS treatment of the resection cavity (within 30 days of surgery) or observation (OBS). Patients were stratified by histology, tumor size, and number of metastases. Patients were recruited at a single tertiary cancer center. The primary endpoint was time to local recurrence in the resection cavity assessed by blinded central review of brain MRI scans in the intention-to-treat population. The trial was registered at clinicaltrials.gov (Trial NCT00950001, status: closed to new participants).

Findings

Between 8/13/2009 and 2/16/2016, 132 patients were randomized to OBS (N=68) or SRS (N=64), with 128 patients available for analysis. We stratified by metastasis size (maximum diameter of ≥3 cm vs. <3 cm), histology (melanoma vs. other), and number of metastases (one vs. two or three). The 12-month local tumor recurrence-free rate was 43% (OBS) (95% CI 31%–59%) and 72% (SRS) (95% CI 60%–87%) (hazard ratio [HR] 0.46, 95% confidence interval [CI] 0.24–0.88, p=0.015).

Interpretation

This prospective randomized trial of patients undergoing surgical resection for 1–3 brain metastases indicates that SRS administered to the resection cavity significantly lowers local recurrence compared to observation alone. Thus, the use of SRS after brain metastasis resection is an alternative to WBRT.

Abstract 3348: UniD: unified and integrated diagnostic pipeline for malignant gliomas based on DNA methylation data

Prognostic and predictive molecular diagnostics for patients with gliomas typically rely on multiple assays, requiring large amounts of tissue and high cost. DNA methylation has been utilized to identify prognostic subsets and high-throughput platforms exist that are suitable for archival tissue. Therefore, we developed a unified and integrated diagnostic pipeline that can assess multiple prognostic and predictive biomarkers using only the Illumina Infinium Methylation array. This pipeline includes two parts: data processing and diagnostic biomarkers. Data processing starts from the raw data and followed with quantitative sample, probe, and batch quality control. The diagnostic biomarkers include a glioma methylation assay that predicts radiation response (GaMA); tumor classification enriched for TCGA expression subclasses; copy number alterations including phosphatase and tensin homolog (PTEN) loss, epidermal growth factor receptor (EGFR) amplification, and chromosomes 1p/19q co-deletion; CpG island methylation phenotype (G-CIMP); isocitrate dehydrogenase (IDH) mutation, and O6-methylguanine-DNA methyltransferase (MGMT) promoter hypermethylation. WHO grade II-IV gliomas were analyzed in both publically available and institutional datasets. A signature was identified to effectively distinguished radiation resistant from radiation sensitive glioma stem-like cells (GSCs). Signature has been applied to 272 TCGA GBM samples from patients who received standard radiotherapy (RT). The survival analysis showed that the subgroup with RT-sensitive and RT-resistant have significant difference in survival time (log-rank test p-value = 0.0016). Gene expression subclasses prediction biomarker was build by using the revised TCGA gene expression subclasses as gold standard. The prediction accuracy in test data set was 83.5% in the homogeneous subgroup, 71.0% in the semi-heterogeneous subgroup, and 62.1% in the heterogeneous subgroup. The prediction accuracy decreased as tumor heterogeneity increased. Certain copy number alteration events were predicted by developing specific signatures. Revised methylation signatures were developed for IDH mutation and G-CIMP status respectively, which can identified 99% of those samples. 230 GBM samples with 450k data available were tested with MGMT methylation-specific real-time PCR for MGMT methylation status. The methylation-based MGMT prediction accuracy reached about 90%. In summary, we have developed a single, FFPE-based pipeline for unified and integrated determination of multiple biomarkers of malignant glioma.

Citation Format: Jie Yang, Qianghu Wang, Lihong Long, Ravesanker Ezhilarasan, Erik Sulman. UniD: unified and integrated diagnostic pipeline for malignant gliomas based on DNA methylation data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 3348. doi:10.1158/1538-7445.AM2017-3348

Abstract 1781: EGFR amplification-A candidate predictive biomarker of PARP inhibitor Talazoparib sensitivity in gliomas

Poly (ADP-ribose) polymerases (PARP) are enzymes involved in DNA-damage repair. Inhibition of PARP is a promising strategy for targeting cancers with defective DNA-damage repair. Several PARP inhibitors are currently in trials in the adjuvant, neoadjuvant, and metastatic settings for the treatment of ovarian, BRCA-mutated breast, and other cancers. Glioma Stem Cells (GSCs) exhibit higher oxidative base damage, single-strand DNA breaks and genomic instability and is reliant on single-strand break repair/base excision repair to tolerate additional stress. PARP critical for maintaining genomic stability by regulating a variety of DNA damage repair mechanisms is constitutively activated in GSCs. In this study, we show that talazoparib – a unique PARP inhibitor exhibits a strong single agent activity to inhibit proliferation of GSCs in vitro and suppress tumor progression in GSCs xenografts. Talazoparib cytotoxic activity was through inhibiting PARylation and by trapping PARP-DNA complexes resulting in accumulation of DNA damage during replication and ultimately leading to apoptosis. More importantly, we found EFGR amplification, which occurs in about 45% of GBM as a biomarker of response to PARPi. Gene expression and RPPA data showed that GSCs with EGFR amplification were associated with increased ROS, basal expression of DNA repair proteins and more DNA damage following talazoparib treatment, which may partially explain the sensitivity of this group of GSCs to talazoparib. EGFR enzyme activity was important for PARPi sensitivity since kinase dead mutant of EGFR or EGFR knock down cell lines were resistant to PARPi as was also shown by decreased PARP/DNA complex formation. Talazoparib showed strong inhibition of tumor growth in subcutaneous glioma model and 20% increase in median survival in an intracranial model with a 10% blood brain penetration of talazoparib. These data provide mechanistic insights into the anti-cancer activity of PARP inhibitors with mechanistic rationale of PARP inhibition and potential EGFR amplification as biomarker in the development for precision cancer.

Citation Format: Shaofang Wu, Feng Gao, Xiaolong Li, Jie Ding, Siyuan Zheng, Emmanuel Martinez-Ledesma, Ningping Feng, Erik Sulman, Roel Verhaak, John de Groot, Dimpy Koul, W.K.Alfred Yung. EGFR amplification-A candidate predictive biomarker of PARP inhibitor Talazoparib sensitivity in gliomas [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1781. doi:10.1158/1538-7445.AM2017-1781

Abstract 2916: The clonal evolution of glioblastoma

Background: Glioblastoma (GBM) is an aggressive cancer that often recurs despite multimodal therapy. Median survival is 12-15 months. Genomic profiling studies have shown marked tumor heterogeneity with distinct mutations among treated and untreated samples. The Cancer Genome Atlas (TCGA), revealed that mutations in TP53, CDKN2A, PTEN, EGFR and NF1 are predominant in untreated tumors. Of 19 patients who progressed on therapy, seven samples were hypermutated with mutations in mismatch repair proteins, mainly MSH6. We proposed to use the Cellecta Lentiviral-based tagging library, a novel model system to assess the role of intratumor heterogeneity (ITH) in the presumptive tumor-initiating fraction of primary GBM and in the development of resistance to temozolomide (TMZ). From a pool of 30,000,000 barcodes, the system enables the genetic integration of a unique barcode sequence into each cell. Each barcode can be quantitatively tracked via next-generation sequencing, allowing for dynamic monitoring of the subclonal architecture.

Methods: GSC 272 and 627 (sensitive and resistant to TMZ) were derived from core biopsies of GBM patients. Cells were transduced with a lentiviral vector containing luciferase. Cells were expanded and infected with the Cellecta Lentiviral library. Multiplicity of infection was determined. Transduced cells were injected into B6.Cg-Foxn1nu/J mice brains using a guide screw system. Half of the mice were treated with TMZ via gavage at two weeks. Brain tumors were removed and total DNA was extracted. Barcode inserts were amplified and the Illumina Sequencing platform was used for barcode quantification. Additionally, whole exome sequencing was performed to assess mutation status.

Results: Confirmation of luciferase signal has been performed using a Luciferase Reporter Assay. Multiplicity of infection was found to be 0.15 and 1 for GSC 6-27 and 272 respectively. Initial methylation studies showed GSC 272 to have methylation of MGMT while GSC 627 did not. There was no difference in OS in the TMZ Resistant (GSC 627) untreated vs. treated cohort (with a median OS of 55 days in both groups (N= 23 vs 25 respectively, P= 0.563). TMZ Sensitive (GSC 272) untreated mice did not survive as long as the treated cohort (N= 24 vs. 25, Median OS 43 vs. 206 days, P= 4.61e-09). Barcode and whole exome sequencing results is currently being analyzed.

Conclusions: The Cellecta lentiviral tagging system is an innovative way to track ITH and clonal evolution in glioblastoma orthotopic models. We hope to discover novel insights into TMZ treatment response and resistance.

Citation Format: Jennifer Brooke Goldstein, Ravesanker Ezhilarasan, Mona Jaffari, Alessandro Carugo, Giulio Draetta, Roeland Verhaak, Sahil Seth, Erik Sulman, Phillip Andrew Futreal. The clonal evolution of glioblastoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 2916. doi:10.1158/1538-7445.AM2017-2916

Identification of Histological Correlates of Overall Survival in Lower Grade Gliomas Using a Bag-of-words Paradigm: A Preliminary Analysis Based on Hematoxylin & Eosin Stained Slides from the Lower Grade Glioma Cohort of The Cancer Genome Atlas

Background:

Glioma, the most common primary brain neoplasm, describes a heterogeneous tumor of multiple histologic subtypes and cellular origins. At clinical presentation, gliomas are graded according to the World Health Organization guidelines (WHO), which reflect the malignant characteristics of the tumor based on histopathological and molecular features. Lower grade diffuse gliomas (LGGs) (WHO Grade II–III) have fewer malignant characteristics than high-grade gliomas (WHO Grade IV), and a better clinical prognosis, however, accurate discrimination of overall survival (OS) remains a challenge. In this study, we aimed to identify tissue-derived image features using a machine learning approach to predict OS in a mixed histology and grade cohort of lower grade glioma patients. To achieve this aim, we used H and E stained slides from the public LGG cohort of The Cancer Genome Atlas (TCGA) to create a machine learned dictionary of “image-derived visual words” associated with OS. We then evaluated the combined efficacy of using these visual words in predicting short versus long OS by training a generalized machine learning model. Finally, we mapped these predictive visual words back to molecular signaling cascades to infer potential drivers of the machine learned survival-associated phenotypes.

Methods:

We analyzed digitized histological sections downloaded from the LGG cohort of TCGA using a bag-of-words approach. This method identified a diverse set of histological patterns that were further correlated with OS, histology, and molecular signaling activity using Cox regression, analysis of variance, and Spearman correlation, respectively. A support vector machine (SVM) model was constructed to discriminate patients into short and long OS groups dichotomized at 24-month.

Results:

This method identified disease-relevant phenotypes associated with OS, some of which are correlated with disease-associated molecular pathways. From these image-derived phenotypes, a generalized SVM model which could discriminate 24-month OS (area under the curve, 0.76) was obtained.

Conclusion:

Here, we demonstrated one potential strategy to incorporate image features derived from H and E stained slides into predictive models of OS. In addition, we showed how these image-derived phenotypic characteristics correlate with molecular signaling activity underlying the etiology or behavior of LGG.