RADT-27. FRACTIONATED RADIOTHERAPY AS FIRST-LINE TREATMENT FOR INTRACRANIAL MENINGIOMAS

BACKGROUND

Meningiomas are the most common primary brain tumour, and patients requiring intervention typically undergo surgical resection as first-line therapy. However, fractionated radiotherapy (fRT) as primary treatment remains an option for patients with larger tumours and surgical contraindications, although outcome prediction in this group remains uncertain. We aimed to assess the clinical factors that contribute to treatment failure in a cohort of meningiomas treated with fRT as first-line therapy.

METHODS

Patients treated with primary fRT for intracranial meningiomas at our institution from November 1, 1998 to December 30, 2017 were reviewed. Those who received upfront surgical resection, radiosurgery, or had less than 6 months of clinical follow up were excluded. We applied logistic regression and Cox regression analysis to ascertain key predictors of treatment failure, progression-free survival (PFS), and adverse events following radiotherapy.

RESULTS

Our cohort included 147 tumours, of which 25 progressed after fRT (median PFS of 3.45 years). Tumours that progressed following RT had a larger median gross tumour volume (GTV) than those that did not (32.3cm3 vs 20.2cm3,p = 0.03768). Pre-RT GTV greater than 11.27cm3 was independently predictive of treatment failure. Larger pre-RT GTV was also associated with significant (grade 3/4) adverse events following fRT. Finally, cerebellopontine angle meningiomas were more prevalent in the group that progressed (20% vs 3.3%, p = 0.0046), whereas cavernous sinus and optic nerve sheath meningiomas had longer PFS (p = 0.0067).

CONCLUSION

We present the largest known cohort of meningiomas treated with fRT as first line therapy and find pre-RT tumour GTV and location to be the critical predictors of outcome, adding to the discussion of treatment considerations in this heterogeneous disease.

NCOG-10. SURGERY FOR RECURRENT GBM: A RETROSPECTIVE CASE-CONTROL STUDY

OBJECTIVE

The role of surgery in recurrent GBM remains a controversial topic. The goal of this study was to perform a case-control analysis including time to tumor recurrence as an additional prognostic factor in order to determine which patients benefit most from repeat surgery.

METHODS

Our brain tumor database was reviewed over a ten-year period for all adult (≥ 18 years old) patients with primary IDH wildtype GBM that received surgery for recurrent disease. These patients were then age, sex, and treatment-matched to case-controls from our institution that received medical therapy for recurrent disease.

RESULTS

A total of 174 adult patients with GBM were included in the study, 87 patients that received surgery for recurrent GBM (surgery cohort) and 87 patients that did not receive surgery for recurrent GBM (non-surgery cohort). The surgery cohort had longer overall survival (p = 0.0003) and post recurrence survival (p = 0.001) than the non-surgery cohort. When the surgery cohort was split into two groups based on time to tumor recurrence, the long time to recurrence group ( > 6 months) demonstrated significantly increased survival compared to the short time to recurrence group (p < 0.0001). Multivariate analysis of both cohorts demonstrated surgery for recurrent GBM was independently significant after adjusting for age, KPS, and time to tumor recurrence (p < 0.0001).

CONCLUSIONS

Surgery for recurrent GBM leads to improved survival independent of age, KPS, and time to tumor recurrence. Patients with time to tumor recurrence greater than 6 months benefit most from additional surgery.

Peri-Operative Outcomes of Carotid Endarterectomy are Not Improved on Dual Antiplatelet Therapy vs. Aspirin Monotherapy: A Systematic Review and Meta-Analysis

OBJECTIVE

A systematic review and meta-analysis of the peri-operative outcomes of carotid endarterectomy (CEA) on dual antiplatelet therapy (DAPT) vs. aspirin monotherapy was carried out, to determine optimal peri-operative management with these antiplatelet agents.

DATA SOURCES

The Web of Science, Pubmed, and Embase databases were searched from inception to July 2021. The corresponding authors of excluded articles were contacted to obtain additional data for possible inclusion.

REVIEW METHODS

The main outcomes included ischaemic complications (stroke, transient ischaemic attack [TIA], and transcranial Doppler [TCD] measured micro-emboli), haemorrhagic complications (haemorrhagic stroke, neck haematoma, and re-operation for bleeding), and composite outcomes. Pooled estimates using odds ratios (ORs) were combined using a random or fixed effects model based on the results of the chi square test and calculation of I².

RESULTS

In total, 47 411 patients were included in 11 studies, with 14 345 (30.2%) receiving DAPT and 33 066 (69.7%) receiving aspirin only. There was no significant difference in the rates of peri-operative stroke (OR 0.87, 95% confidence interval [CI] 0.72 - 1.05) and TIA (OR 0.78, 95% CI 0.52 - 1.17) despite a significant reduction in TCD measured micro-emboli (OR 0.19, 95% CI 0.10 - 0.35) in the DAPT compared with the aspirin monotherapy group. Subgroup analysis did not reveal any significant difference in ischaemic stroke risk between patients with asymptomatic and symptomatic carotid artery stenosis. DAPT was associated with an increased risk of neck haematoma (OR 2.79, 95% CI 1.87 - 4.18) and re-operation for bleeding (OR 1.98, 95% CI 1.77 - 2.23) vs. aspirin. Haemorrhagic stroke was an under reported outcome in the literature.

CONCLUSION

This meta-analysis found that CEA while on DAPT increased the risk of haemorrhagic complications, with similar rates of ischaemic complications, vs. aspirin monotherapy. This suggests that the risks of performing CEA on DAPT outweigh the benefits, even in patients with symptomatic carotid stenosis. The overall quality of studies was low, and improved reporting of CEA outcomes in the literature is necessary.

Leveraging the CSF proteome toward minimally-invasive diagnostics surveillance of brain malignancies

Background

Diagnosis and prognostication of intra-axial brain tumors hinges on invasive brain sampling, which carries risk of morbidity. Minimally-invasive sampling of proximal fluids, also known as liquid biopsy, can mitigate this risk. Our objective was to identify diagnostic and prognostic cerebrospinal fluid (CSF) proteomic signatures in glioblastoma (GBM), brain metastases (BM), and primary central nervous system lymphoma (CNSL).

Methods

CSF samples were retrospectively retrieved from the Penn State Neuroscience Biorepository and profiled using shotgun proteomics. Proteomic signatures were identified using machine learning classifiers and survival analyses.

Results

Using 30 µL CSF volumes, we recovered 755 unique proteins across 73 samples. Proteomic-based classifiers identified malignancy with area under the receiver operating characteristic (AUROC) of 0.94 and distinguished between tumor entities with AUROC ≥0.95. More clinically relevant triplex classifiers, comprised of just three proteins, distinguished between tumor entities with AUROC of 0.75-0.89. Novel biomarkers were identified, including GAP43, TFF3 and CACNA2D2, and characterized using single cell RNA sequencing. Survival analyses validated previously implicated prognostic signatures, including blood-brain barrier disruption.

Conclusions

Reliable classification of intra-axial malignancies using low CSF volumes is feasible, allowing for longitudinal tumor surveillance.

EPCO-32. IDENTIFICATION OF PROGNOSTIC CHORDOMA SUBGROUPS USING DNA METHYLATION SIGNATURES IN TISSUE AND PLASMA

BACKGROUND

Chordomas are malignant bone cancers arising from the skull-base and spine that are rare but cause devastating central nervous system morbidities. Survival is highly variable despite surgery and radiotherapy as 10% live under 1 year and 30-35% survive over 20 years. There are currently no reliable prognostic factors and this limits our ability to tailor patient treatment to their risk. Accordingly, this work identifies epigenetic prognostic chordoma subgroups that are detectable non-invasively through plasma methylomes to guide treatment.

METHODS

A total of 68 chordoma surgical specimens resected between 1996-2018 across three international centres underwent DNA methylation profiling. Cell-free methylated tumor DNA immunoprecipitation and high-throughput sequencing was performed on available matched plasma samples.

RESULTS

Two stable tumor clusters were identified through consensus clustering of tissue methylation data. Clusters had statistically significantly different disease-specific survivals (log-rank p=0.0062) independent of clinical factors in a multivariable Cox analysis (HR=16.5, 95%CI: 2.8-96, p=0.0018). The poorer-performing “Immune-infiltrated” cluster had genes hypomethylated at promoters, typically resulting in transcription, within immune-related pathways and higher immune cell abundance within tumors. The better-performing “Cellular” cluster showed higher tumor cellularity plus cell-to-cell interaction and extracellular matrix pathway hypomethylation. Fifty chordoma-versus-other binomial generalized linear models built using plasma methylome data distinguished chordomas from meningiomas and spinal metastases, as representative clinical differential diagnoses, in random left-out 20% testing sets (mean AUROC=0.84, 95%CI: 0.52-1.00). Plasma-based methylation signatures were highly correlated with tissue-based signals within both poor-performing (median r=0.69, 95%CI: 0.66-0.72) and better-performing cluster tumors (median r=0.67, 95%CI: 0.62-0.72).

CONCLUSIONS

The first identification of two distinct prognostic epigenetic chordoma subgroups is shown here with “Immune-infiltrated” tumors having a poorer prognosis than “Cellular” tumors. Plasma methylomes can be utilized for non-invasive chordoma diagnosis and subtyping. This work may transform chordoma treatment decision-making by guiding surgical planning in advance to match resection aggressiveness with patient prognosis.

Central Nervous System Metastases

The proportion of patients developing central nervous system (CNS) metastasis is increasing. Most are identified once symptomatic. Surgical resection is indicated for solitary or symptomatic brain metastases, separation surgery for compressive radioresistant spinal metastases, and instrumentation for unstable spinal lesions. Surgical biopsies are performed when histological diagnoses are required. Stereotactic radiosurgery is an option for limited small brain metastases and radioresistant spinal metastases. Whole-brain radiotherapy is reserved for extensive brain metastases and leptomeningeal disease with approaches to reduce cognitive side effects. Radiosensitive and inoperable spinal metastases typically receive external beam radiotherapy. Systemic therapy is increasingly being utilized for CNS metastases.

ECOA-8. Lung adenocarcinoma brain metastasis prediction using tumor DNA methylation profiling

Introduction

The development of brain metastases from primary cancer profoundly impacts patient prognosis. Up to one quarter of lung cancers develop brain metastases and subsequent median overall survival is one year. Although clinical factors do not reliably predict brain metastasis development, DNA methylation signatures have been recently shown to predict outcomes in other cancers. It is hypothesized that DNA methylation signatures predicting brain metastasis development from lung cancer will be identified. This work may allow for treatment strategies that prevent brain metastasis development in high risk lung cancer patients.

Methods

DNA methylation profiling was undertaken on N=124 lung adenocarcinoma patients. In a randomly selected 70% training cohort, differentially methylated CpG sites between patients developing and not developing brain metastases were identified and used to build a generalized boosted regression model. Patients in the independent 30% testing cohort were assigned brain metastasis risk scores by the model.

Results

Brain metastases developed in 49/124 (39.5%) of patients and 2.3K CpG sites were significantly differentially methylated between patients developing and not developing metastases. Methylation-based brain metastasis risk scores predicted time to brain metastasis development in the testing cohort (Univariate cox: HR=3.2, 95% CI 1.1–9.4, p=0.03). A multivariate cox analysis assessing tumor size and nodal positivity together with methylation scores as covariates identified methylation scores as the only independent predictor of brain metastasis development in the testing cohort (HR=4.3, 95%CI 1.1–17, p=0.038).

Conclusions

DNA methylation signatures in lung adenocarcinomas predict brain metastasis development independently from the non-metastatic components of cancer stage. Future work developing a comprehensive nomogram utilizing methylation scores together with clinical factors to determine patient specific risk values may aid in treatment decisions and patient prognosis counselling.

BIOM-02. IDENTIFYING MGMT ALTERATIONS AS BIOMARKERS OF SURVIVAL IN LUNG ADENOCARCINOMA WITH BRAIN METASTASES

EGFR-mutant lung adenocarcinomas (EGFRm-LUAD) have a higher risk of developing brain metastases (BM) compared to non-EGFR-mutant tumors. BM development has significant prognostic impact and leads to poorer patient survival. MGMT promoter methylation is known to determine response to therapy in other cancer types including intracranial gliomas but has not been investigated in EGFRm-LUAD BM. This work aims to assess whether MGMT promoter methylation predicts patient survival or BM development in EGFRm-LUAD patients. A large cohort of 90 primary EGFRm-LUAD tumors, of which 33 (37%) developed BM, were profiled using the Illumina Infinium MethylationEPIC Bead chip. Using the previously reported MGMT-STP27 approach that uses two CpG sites to predict MGMT methylation status, Cox modeling was performed to assess whether MGMT methylation status correlates with overall survival independent of other clinical factors. MGMT methylation significantly predicted poorer survival in EGFRm-LUAD patients that developed BM (p=0.0003) and did not develop BM (p=0.003). A multivariate cox analysis, adjusting for cancer stage and smoking status as potential confounders, showed that MGMT methylation (HR=6.2, 95%CI:2.2–17.4, p=0.0005) and BM development (HR=2.6, 95%CI:1.3–5.3, p=0.007) were both independently predictive of worse overall survival in EGFRm-LUAD patients. This finding of poorer survival in MGMT methylated EGFRm-LUAD is validated in an independent LUAD patient cohort. Total mutation burden, calculated by the number of mutations per megabase of DNA, was substantially higher in MGMT methylated tumours with an interquartile range (IQR) of 58 (30–71) compared to MGMT unmethylated tumours with the IQR of 5.5 (4.3–6.1) resulting p-value of 0.01 for this comparison. Overall, this work shows that MGMT promoter methylation status is an important prognostic biomarker in LUAD patients. MGMT promoter methylation status in EGFRm–LUAD patients with BM may be used to guide patient treatment with potentially a greater extent of treatment for high-risk patients.

EPCO-01. LUNG ADENOCARCINOMA BRAIN METASTASIS PREDICTION, PREVENTION, AND NON-INVASIVE DIAGNOSIS USING METHYLATION SIGNATURES WITHIN TISSUE AND CIRCULATING TUMOUR DNA

BACKGROUND

One quarter of lung adenocarcinoma (LUAD) patients develop brain metastases (BM) and experience a poorer median survival of 12 months despite treatment. Clinical variables do not robustly predict who will develop BM and targeted preventative treatments are limited. Tumour DNA methylation signatures predict outcomes in other cancers and can be detected in circulating tumour DNA (ctDNA). This work predicts BM development from LUAD using methylation data, identifies novel potential treatment targets to prevent metastases, and detects LUAD-BM ctDNA non-invasively.

METHODS

DNA methylation profiling was undertaken on N=124 LUAD tumours. A gradient boosted regression model built on differentially methylated CpGs (DMCs) between tumours with and without BM in 70% of samples was validated in an independent 30% testing cohort. Nine paired BM samples were profiled and DMCs between their corresponding LUAD tissue were identified along with copy number (CN) alterations. A total of 47 LUAD-BM plasma samples underwent sequencing of immunoprecipitated methylated ctDNA and differentially methylated regions (DMRs) between LUAD-BM and intrinsic brain lesions were identified.

RESULTS

The methylation-based model significantly predicted time to brain metastasis development within the testing cohort independently from cancer stage in a multivariate analysis (HR=4.3, 95%CI 1.1–17, p=0.038). Genes/pathways involved in the process of brain metastasis were identified through assessment of 83K DMCs (FDR< 0.2, mean difference >|0.1|) between paired samples as well as the CN losses found in chromosome 12q/19 of BM samples. A total of 5.5K DMRs were identified that distinguish BM samples from gliomas or primary CNS lymphomas (FDR< 0.05, logFC >1).

CONCLUSIONS

DNA methylation signatures in lung adenocarcinomas predict brain metastasis development independently from prognostic clinical factors. Genes and pathways involved in metastasis were identified as novel potential therapeutic targets. Methylated circulating tumor DNA signatures differentiate lung brain metastases from other ring-enhancing brain lesions and may have potential for non-invasive diagnosis.

BIOM-62 SENSITIVE DETECTION AND DISCRIMINATION OF INTRACRANIAL TUMORS BY BLOOD

BACKGROUND

The diagnosis of intracranial tumors relies on tissue specimens obtained by invasive surgery. Non-invasive diagnostic approaches, particularly for patients with brain tumours, provide an opportunity to avoid surgery and mitigate unnecessary risk to patients. We reasoned that DNA methylation profiles of circulating tumor DNA in blood can be used as a clinically useful biomarker for patients with brain tumors, given the specificity of DNA methylation profiles for cell-of-origin.

METHODS

We generated methylation profiles on the plasma of 608 patients with cancer (219 intracranial, 388 extracranial) and 60 healthy controls using a cell-free methylated DNA immunoprecipitation combined with deep sequencing (cfMeDIP-seq) approach. Using machine-learning approaches we generated and evaluated models to distinguish brain tumors from extracranial cancers that may metastasize to the brain, as well as additional models to discriminate common brain tumors included in the differential diagnosis of solitary extra-axial and intra-axial tumors.

RESULTS

We observed high sensitivity and discriminative capacity for our models to distinguish gliomas from other cancerous and healthy patients (AUC=0.99, 95%CI 0.96–1), with similar performance in IDH mutant and wildtype gliomas as well as in lower- and high-grade gliomas. Excluding non-malignant contributors to plasma methylation did not change model performance (AUC=0.982, 95%CI 0.93–1). Models generated to discriminate intracranial tumors from each other also demonstrated high accuracy for common extra-axial tumors (AUCmeningioma=0.89, 95%CI 0.80–0.97; AUChemangiopericytoma=0.95, 95%CI 0.73–1) as well as intra-axial tumors ranging from low-grade indolent glial-neuronal tumors (AUC 0.93, 95%CI 0.80 – 1) to diffuse intra-axial gliomas with distinct molecular composition (AUCIDH-mutant glioma = 0.82, 95%CI 0.66 -0.98; AUCIDH-wildtype-glioma = 0.71, 95%CI 0.53 – 0.9). Plasma cfMeDIP-seq signals originated from corresponding tumor tissue DNA methylation signals (r=0.37, p< 2.2e-16).

CONCLUSIONS

These results demonstrate the potential for cfMeDIP-seq profiles to not only detect circulating tumor DNA, but to accurately discriminate common intracranial tumors that share cell-of-origin lineages.

Dysphagia rates in single- and multiple-level anterior cervical discectomy and fusion surgery: a meta-analysis

Background

To conduct a meta-analysis to assess dysphagia complicating single-level and multiple-level (≥2) anterior cervical discectomy and fusion (ACDF) surgery.

Methods

Electronic searches were performed using four electronic databases from their inception to December 2017. Relevant studies reporting the rate of dysphagia as an endpoint for patients undergoing ACDF for degenerative disease, myelopathy, cervical canal stenosis or ossification of the posterior longitudinal ligament were identified according to prior inclusion and exclusion criteria. Statistical analysis was performed using a fixed effect model. P-scores were used to rank the levels of ACDF based on the rate of dysphagia. I² was used to explore heterogeneity.

Results

Ten studies were identified and included in the systematic review and meta-analysis, with a total of 4,018 patients identified; 2,362 patients underwent single-level ACDF, while 1,656 underwent multiple level (≥2 ACDF). The mean age ranged from 49.45 to 57.77 years. Mean follow-up time ranged from 2 days to 27.3 months. Overall, meta-analysis demonstrated a statistically significant higher dysphagia rate for multiple-level ACDF (6.6%) than for single-level ACDF (4%) (P heterogeneity =0.151, OR =1.42, 95% CI: 1.05-1.91, I²=32%).

Conclusions

Dysphagia is a relatively common complication in the early postoperative period following ACDF and may cause patients significant discomfort and distress. This meta-analysis demonstrates a higher rate of dysphagia with multiple-level ACDF than with single-level ACDF at a period of 12-24 months.

71. MGMT PROMOTER METHYLATION IS A PROGNOSTIC BIOMARKER IN EGFR MUTANT LUNG ADENOCARCINOMA WITH BRAIN METASTASES

EGFR-mutant lung adenocarcinomas (EGFRm-LUAD) have a higher risk of brain metastasis (BM) development than non-mutant lesions regardless of cancer stage. BM development is a marker of tumor aggressiveness and has significant prognostic impact that leads to treatment failure. MGMT promoter methylation is known to determine response to therapy in other cancer types but it has not been investigated in EGFRm-LUAD brain metastases. This work aims to assess whether MGMT promoter methylation predicts patient survival or BM development in EGFRm-LUAD patients. A large cohort of 90 primary EGFRm-LUAD, 33(37%) of which developed BM, were profiled using Illumina Infinium MethylationEPIC Beadchip. We utilized genome-wide methylation signatures to determine MGMT methylation status using the previously reported MGMT-STP27 approach that uses two CpG sites to predict MGMT methylation status. Cox modeling was performed to assess whether MGMT methylation status correlates with overall survival independent of other clinical factors. MGMT methylation significantly predicted poorer survival in EGFRm-LUAD patients developed BM (p=0.0003) and those who did not (p=0.003). A multivariate Cox analysis, adjusting for stage and smoking status as potential confounders, showed that MGMT methylation (HR=6.2, 95%CI:2.2–17.4, p=0.0005) and BM (HR=2.6, 95%CI:1.3–5.3, p=0.007) were both independently predictive of worsened survival. Total Mutation Burden calculated by the number of mutations per megabase of DNA was higher in MGMT methylated tumours with an interquartile range (IQR) of 58(30–71) compared to MGMT unmethylated tumours with IQR of 5.5(4.3–6.1). This work shows that MGMT promoter methylation status is an important prognostic biomarker in EGFRm-LUAD patients. Further work will validate these findings obtained using whole-genome DNA methylation by comparing to results using methylation specific PCR assays. MGMT promoter methylation status in EGFRm–LUAD patients with BM may be used to guide patient treatment with potentially a greater extent of treatment for those higher risk patients.

PATH-21. CLINICAL UTILITY OF DNA METHYLATION PROFILING FOR DIAGNOSIS OF CHALLENGING CENTRAL NERVOUS SYSTEM TUMORS: THE TORONTO EXPERIENCE

The update on the WHO classification of central nervous system (CNS) tumors incorporated molecular signatures for a more accurate diagnosis. Recently, DKFZ has demonstrated the utility of DNA methylation profiling(MP) for molecular classification of CNS tumors. We performed a prospective clinical study over the last three years to evaluate the clinical utility ofDNA MP on FFPE samples of 66 challenging CNS tumor cases using online DKFZ classifier. Eleven samples were excluded due to low tumor DNA content or low calibration(predictive) scores(CS)< 0.3.DNA MP confirmed the original pathology diagnoses in 15(27%)cases. The integrated molecular diagnoses were changed in 38/55(70%) including establishment of a new diagnostic entity, change in molecular signature and subtyping. TheWHO grades were changed in 16(27%) of the tumors; about two-thirds resulted in upgrading. We detected non-canonical IDH mutations in 9 diffuse gliomas and the CNV plots revealed false positive FISH results for 1p/19q co-deletion in two diffuse gliomas. The CNV plots contributed to the final diagnosis in 40(72%) patients. The molecular subtypes of medulloblastoma, ependymoma and glioblastoma subclasses were determined in 36(65%) cases. Seventy-five percent of cases with confirmation of initial diagnosis or change in molecular diagnosis had CS > 0.5, among which 51% had a CS >0.9. The median and range CS of cases with new diagnostic entity and confirmed cases were 0.86(0.37–0.99) and 0.98(0.42–0.99), respectably. Furthermore, we detected higher CS in IDH-mutant gliomas in comparison to glioblastoma IDH-wild type(P=0.04). We also observed lower CS in mesenchymal glioblastoma in comparison to other subclasses. The MGMT promoter methylation was determined in 17/20(85%) glioblastoma cases. While the DKFZ group established CS of 0.9 as a cut-off for matching to methylation classes, our findings suggest lower threshold values in challenging CNS tumor cases. Our experience indicates clinical utility of MP of challenging CNS tumors as a reliable ancillary diagnostic tool in routine neuropathology practice.

CMET-30. COMPREHENSIVE METHYLOME ANALYSIS OF EGFR-MUTANT PRIMARY LUNG ADENOCARCINOMA AND MATCHED BRAIN METASTASIS

Brain metastasis (BM) in patients with EGFR-mutant lung adenocarcinoma is a major determinant of overall survival. Novel insight into the genetic and epigenetic underpinnings of BM development is lacking. The aim of this study is to compare the methylome of EGFR-mutant primary lung adenocarcinoma (EGFRM-PLA) and matched BM to identify important alterations for the mechanisms of BM. Matched EGFRM-PLA and BM tumors from seven patients were profiled using the Illumina Infinium MethylationEPIC BeadChip array. Unsupervised clustering analyses of the 14 samples showed a similar whole DNA methylation signatures between EGFRM-PLA and BM tumors. Furthermore, PCA plot highlighted that seven matched BM and lung tumor samples were clustered together closely based on matching pairs for the most variable probes (2.5K to 10K). These observations indicate high level of concordance and the same cell of origin. However, these fourteen samples clustered into two groups based on tumor site being lung or brain based on 83K differentially methylated CpG sites. Of the 83K probes, 2.4K were either hypermethylated or hypomethylated in all lung samples. A quarter of these 2.4K probes were located in promoter regions. Specifically, we identified differences in methylation status of EGFR/ALK promoter regions in lung tumors versus BM. CNV analyses showed higher deep deletions of chromosomes and genes in BM compared to EGFRM-PLA. Leukocytes unmethylation for purity (LUMP) scores which indicate immune cell infiltration were similar between lung and BM pairs (Mean LUMP_score=0.64) consistent with high immune cell infiltration. Our results indicated a similar whole DNA methylation signature of EGFRM-PLA and matched BM, while comprehensive analysis identified important differentially methylated probes. Distinct differences in CNV alterations were observed in lung versus brain samples. The BM and EGFRM-PLA showed similar tumor purity and immune cell components. Overall, tumor methylation profiling provides clinically important information regarding biology of BM in EGFRM-PLA.

CMET-32. DNA METHYLATION ALTERATIONS IN LUNG ADENOCARCINOMAS THAT DEVELOP BRAIN METASTASES

INTRODUCTION

The development of brain metastases from primary cancer profoundly impacts patient prognosis. Metastases are the most common adult brain tumor with up to one quarter of lung cancers developing metastases and median overall survival after metastasis being one year. Clinical factors do not reliably predict brain metastasis development and over 90% are identified after symptoms develop. DNA methylation signatures predict outcomes in other cancers and so identifying signatures that predict metastasis development may allow for treatment strategies that prevent development in high risk patients.

METHODS

Whole genome DNA-methylation profiling was undertaken on N=124 lung adenocarcinoma patients after bisulfite conversion of DNA from formalin-fixed paraffin-embedded tissue. In a randomly selected 70% training cohort, the most differentially methylated CpG sites between patients developing and not developing brain metastases were identified with p< 0.05. A generalized boosted regression model built on these selected features output brain metastasis risk scores for patients in the independent 30% testing cohort.

RESULTS

Brain metastases developed in 49/124 (39.5%) of patients and 2.3K CpG sites were significantly differentially methylated between patients developing and not developing metastases. Methylation-based brain metastasis risk scores predicted time to brain metastasis in a univariate cox regression model (HR=3.2, 95% CI 1.1–9.4, p=0.03). A corresponding area under the receiver operating characteristic curve at 52 months was 0.64. A multivariate cox analysis including tumor size and nodal status, representing the non-metastatic components of cancer stage, identified methylation score as the only independent predictor of brain metastasis (HR=4.3, 95%CI 1.1–17, p=0.038).

CONCLUSIONS

DNA methylation signatures in lung adenocarcinoma predict brain metastasis development independent of stage components, which classically predict patient outcome in cancer. Future work developing a comprehensive nomogram utilizing methylation scores together with other clinical factors to determine patient specific risk values may aid in treatment decisions and patient prognosis counselling.

BSCI-26. COMPARATIVE METHYLATION PROFILING OF EGFR MUTANT LUNG ADENOCARCINOMA AND PAIRED BRAIN METASTASIS

BACKGROUND: Adenocarcinomas (ADC) are the most common lung tumours and EGFR-mutant lesions have a higher risk of brain metastasis development, which confers poorer survival. Genetic and epigenetic signatures of brain metastasis have not been comprehensively identified. The aim of this study is to compare the methylome of EGFR-mutant primary lung tumor and matched brain metastasis to identify mechanisms of brain metastasis and new treatment targets. METHOD: Seven matched primary to brain metastasis tumours were profiled using the Illumina Infinium MethylationEPIC BeadChip array. Hierarchical clustering and principal component analyses (PCA) were performed using most variable CpG sites. Supervised analyses were performed between lung and brain tumour samples. Copy number variation (CNV) plots identified alterations between pairs along with Leukocytes unmethylation for purity (LUMP) score and prediction lymphocyte proportion analyses to measure immune infiltration. RESULTS: Unsupervised clustering showed that the fourteen tumours clustered according to patient with similar methylation profiles between each of seven matched pairs. On the supervised analysis using 83K significant CpG sites, the fourteen samples clustered into two groups based on tumour site being lung or brain. Of these 83K CpG sites, 2.4K were either hypermethylated or hypomethylated in all lung samples. One quarter of these 2.4K CpG sites were located in promoter regions. CNV analyses showed losses of FGFR1, C19MC, CDKN2A, PTCH1, and MYCN genes with higher deep deletions in brain versus lung primary samples. Immune infiltration measures were similar between lung and brain metastasis pairs (LUMP-score=0.64) consistent with high immune cell infiltration. CONCLUSION: In this EGFR-mutant lung adenocarcinomas and matched brain metastases, differentially methylated CpG sites and CNV alterations are identified that distinguish lung from brain samples. Further work with additional matched samples may further elucidate signatures specific to brain metastasis and aid in our understanding of the mechanisms of brain metastasis.

OTHR-09. IDENTIFYING EPIGENETIC SIGNATURES IN LUNG ADENOCARCINOMAS THAT PREDICT DEVELOPMENT OF BRAIN METASTASIS

INTRODUCTION: Metastases are the most common adult brain tumor with half spreading from lung cancers and they reduce median overall survival from 26 to 12 months. There are no robust patient-specific predictors of brain metastasis. Epigenetic signatures predict disease recurrence in other cancers and identifying brain metastasis methylation-based signatures may allow for treatment approaches to high-risk patients that prevent their development. METHODS: In 207 lung adenocarcinomas, multivariate cox time to brain metastasis analyses including clinically-relevant variables (lung tumor size and TNM nodal score) along with significant covariates on univariate analyses were performed. DNA was extracted from 142 of these tumors and profiled on the Illumina Infinium EPIC array. A generalized boosted regression classification model used differentially methylated CpG sites significantly predicting time to brain metastasis in a 70% training cohort cox analysis (p&lt; 0.05). Resulting methylation-based risk scores were compared to size and nodal status in a multivariate analysis of the independent 30% testing cohort. RESULTS: Of 207 patients with 72 brain metastatic events, tumor size (HR=1.5, 95%CI 1.1–2.0, p=0.01), N status (N3 vs. N0, HR=9.9, 95%CI 3.1–31, p=0.0001), EGFR status (HR=0.4, 95%CI 0.2–0.8, p=0.014), and age (HR=0.7, 95%CI 0.5–1.0, p=0.039) independently predicted their development. Methylation-based risk scores significantly predicted time to brain metastasis in a univariate analysis of the testing cohort (p=0.03). A multivariate analysis of testing cohort patients identified methylation score as the only independent predictor of brain metastasis (HR=4.3, 95%CI 1.1–17, p=0.038) accounting for tumor size and N score. CONCLUSIONS: Genome-wide DNA methylation signatures predict brain metastasis development in lung adenocarcinomas independent of tumor size and nodal disease. The design of a nomogram combining methylation profile other clinical factors may be used to determine patient specific brain metastasis risk values to guide patient counselling, extent of treatment, and screening.

Methylation profiling of EGFR mutant primary and metastatic lung cancer with brain metastasis

e20574

Background: EGFR-mutant lung cancer is a key molecular subtype of lung cancer. In recent years there is clear recognition in the value of using methylation signature of cancer for improving diagnosis and predicting outcome as well as understanding the biology of cancer progression. Methods: In this study we chose to characterize the methylome signature of early stage surgically resected EGFR-mutant lung adenocarcinomas in the primary lung tumor. 90 NSCLC cases and 7 matched metastatic brain samples were profiled using Illumina Infinium MethylationEPIC Beadchip. We compared methylation profiles of 1) smokers versus lifetime non-smokers and 2) matched primary lung versus brain metastasis to identify methylation biomarkers. We performed supervised analysis and unsupervised clustering of the methylation data. Results: Unsupervised clustering of all lung and brain samples based on 10K most variable probes showed a similar methylation signature between metastatic brain samples and lung samples. The 7-matched brain and lung samples formed close cluster groups based on matching pairs for the most variable probes from 2.5K to 10K, reflecting the same cell of origin. Supervised analysis of smokers versus lifetime non-smokers did not show any significant methylation differences between the two groups, while unsupervised analysis did not create clusters of smokers and non-smokers based on various number of probe sets we analyzed. Conclusions: Lung tumors that metastasized to the brain share similar methylation features with primary lung tumors. Comprehensive methylation profiling demonstrated no difference between EGFR mutant tumors in smokers versus non-smokers, suggesting that the EGFR mutation is a stronger determinant of outcome independent of smoking.