Cohort Of Glioblastoma Patients Research Articles

P14.59 Rapid early progression of glioblastoma is not related to cortical/neural stem cells regions

Abstract BACKGROUND Rapid early progression (REP) of glioblastoma after surgery observed on pre-radiotherapy MRI scan is common. Subventricular zone (SVZ) and hippocampal regions are supposed to harbor astrocyte-like neural stem cells (NSC) with tumors arising from these transformed stem cells threatening of higher risk of REP. REP is defined as a new enhancing tumor or >25% increase in enhancement before radiotherapy. Lim′s classification of initial glioblastoma location related to these NSC regions predicts invasive and multifocal tumor phenotype. Glioblastomas are classified preoperatively into four groups by the spatial relationship of the contrast-enhancing lesion with the SVZ and cortex. The aim of this retrospective single-institutional study is to evaluate the relations of this Lim classification on REP in unselected cohort of glioblastoma patients. MATERIAL AND METHODS Patients receiving radiotherapy between 2014–2017 were analyzed, 95 were evaluable. 47 patients (30.5%) were treated with the Stupp regimen. Lim1 classification (contact with cortex as well as SVZ) was presented in 74(48%) patients, Lim2 (contact with SVZ only) in 22(14.3%), Lim3 (contact with cortex only) in 50(32.5%) and Lim4 in 8(5.2%) patients. A total of 52% of patients developed REP. RESULTS Significantly better overall survival was with Stupp regimen (23.3 vs. 8.6 months, p<0.001) and without REP (18.5 vs. 10.2 months, p=0.001). There was no significant impact of time to start of radiotherapy. No significant relation between REP and Lim classification was observed. CONCLUSION The initial location is not predictive for REP. Patients experiencing REP have significantly worse overall survival and modification of their management represents an urgent unmet clinical need. Molecular and clinical biomarkers indicating an increased risk of REP are needed.Presented will also be an already published analysis of clinical factors associated with REP in glioblastoma and the effect of REP and treatment on survival outcomes. Newly, we will introduce the investigator-initiated prospective academic clinical trial (GlioMET) focused on optimization of glioblastoma radiotherapy by 11C-Methionine PET scan in patients with REP. Supported by Ministry of Health of the Czech Republic AZV, No.18-03-00469 and AZV NU20-03-00148.

P14.01 Incidence and impact of venous thromboembolism and major bleeding in patients with glioblastoma

Abstract BACKGROUND Patients with glioblastoma are considered to be at high risk of venous thromboembolism (VTE) and major bleeding (MB), although reliable incidence estimates are lacking. The aim of this study was to assess the cumulative incidence, risk factors and prognostic impact of VTE and MB in a large cohort of glioblastoma patients. MATERIAL AND METHODS Patients diagnosed with glioblastoma between 2004–2020 from the Leiden University Medical Center and Haaglanden Medical Center were included and followed from 6 months before date of histopathological glioblastoma diagnosis up to two years after, or until an outcome of interest (VTE and MB) or death occurred. Cumulative incidences were estimated using Kaplan-Meier and cumulative incidence competing risk methods. Outcome predictors were determined with multivariable (time-dependent) Cox models. RESULTS Of the 967 included patients, 631 died, 101 were diagnosed with VTE and 130 with MB during a median follow-up of 15 months (IQR 9–22). The majority of MBs were intracranially (94%), of which half was within 4 weeks post-surgery (51%). Ambulant patients did not routinely receive pharmacological thromboprophylaxis. The adjusted 2.5-year-cumulative incidence of VTE was 12% (95%CI 10–14) and of MB 16% (95%CI 13–18). One year increase in age (HR 1.03, 95%CI 1.01–1.05), tumor resection (vs. biopsy: HR 0.56, 95%CI 0.35–0.89) and performance status (ECOG >1 vs ≤1: HR 1.8, 95%CI 1.2–2.7) were independent predictors of VTE (a biopsy rather than resection was performed in patients with the worst prognosis). MB and VTE predicted all-cause mortality (adjusted HR 1.7, 95%CI 1.3–2.1 and 1.3, 95%CI 0.99–1.6 respectively). CONCLUSION The incidences of VTE and MB in patients with glioblastoma are high, with both complications associated with a poor prognosis. Our observations emphasize the need for prospective studies to determine optimal thromboprophylaxis and VTE treatment strategy in these patients.

Prognostic role of Ki-67 in glioblastomas excluding contribution from non-neoplastic cells

Survival of glioblastoma patients varies and prognostic markers are important in the clinical setting. With digital pathology and improved immunohistochemical multiplexing becoming a part of daily diagnostics, we investigated the prognostic value of the Ki-67 labelling index (LI) in glioblastomas more precisely than previously by excluding proliferation in non-tumor cells from the analysis. We investigated the Ki-67 LI in a well-annotated population-based glioblastoma patient cohort (178 IDH-wildtype, 3 IDH-mutated). Ki-67 was identified in full tumor sections with automated digital image analysis and the contribution from non-tumor cells was excluded using quantitative double-immunohistochemistry. For comparison of the Ki-67 LI between WHO grades (II-IV), 9 IDH-mutated diffuse astrocytomas and 9 IDH-mutated anaplastic astrocytomas were stained. Median Ki-67 LI increased with increasing WHO grade (median 2.7%, 6.4% and 27.5%). There was no difference in median Ki-67 LI between IDH-mutated and IDH-wildtype glioblastomas (p = 0.9) and Ki-67 LI was not associated with survival in glioblastomas in neither univariate (p = 0.9) nor multivariate analysis including MGMT promoter methylation status and excluding IDH-mutated glioblastomas (p = 0.2). Ki-67 may be of value in the differential diagnostic setting, but it must not be over-interpreted in the clinico-pathological context.

Enzymatic Activity of Glycosyltransferase Glt8d1 Promotes Human Glioblastoma Cells Migration

Glioblastoma (GBM) is the most aggressive primary brain tumor characterized by infiltrative migration of malignant glioma cells into the surrounding brain parenchyma. In this study, our analysis of GBM patient cohorts displayed significant higher expression of Glycosyltransferase 8 domain containing 1 ( GLT8D1 ) compared to normal brain tissue and could be associated with impaired patient survival. In vitro an increased GLT8D1 expression significantly enhanced migration of two different sphere-forming GBM cell lines. By in silico analysis we predicted the 3D-structure as well as the active site residues of GLT8D1. The introduction of point mutations in predicted active site reduced its glycosyltransferase activity in vitro and consequently impaired GBM tumor cell migration. Examination of GLT8D1 interaction partners by LC-MS/MS implied proteins associated with cytoskeleton and intracellular transport as potential substrates. In conclusion, we demonstrated that the enzymatic activity of glycosyltransferase GLT8D1 promotes GBM cell migration.

BRCA1 Protein Expression Predicts Survival in Glioblastoma Patients from an NRG Oncology RTOG Cohort

Purpose: Glioblastoma, the most common malignant brain tumor, was associated with a median survival of <1 year in the pre-temozolomide (TMZ) era. Despite advances in molecular and genetic profiling studies identifying several predictive biomarkers, none has been translated into routine clinical use. Our aim was to investigate the prognostic significance of a panel of diverse cellular molecular markers of tumor formation and growth in an annotated glioblastoma tissue microarray (TMA). Methods and Materials: A TMA composed of archived glioblastoma tumors from patients treated with surgery, radiation, and non-TMZ chemother­apy, was provided by RTOG. RAD51, BRCA-1, phosphatase and tensin homolog tumor suppressor gene (PTEN), and miRNA-210 expression levels were assessed using quantitative in situ hybridization and automated quantitative protein analysis. The objectives of this analysis were to determine the association of each biomarker with overall survival (OS), using the Cox proportional hazard model. Event-time distributions were estimated using the Kaplan-Meier method and compared by the log-rank test. Results: A cohort of 66 patients was included in this study. Among the 4 biomarkers assessed, only BRCA1 expression had a statistically significant correlation with survival. From univariate analysis, patients with low BRCA1 protein expression showed a favorable outcome for OS (p = 0.04; hazard ratio = 0.56) in comparison with high expressors, with a median survival time of 18.9 versus 4.8 months. Conclusions: BRCA1 protein expression was an important survival predictor in our cohort of glioblastoma patients. This result may imply that low BRCA1 in the tumor and the consequent low level of DNA repair cause vulnerability of the cancer cells to treatment.

Clinical and Molecular Features of a Prospectively-Sequenced Cohort of Glioblastoma Patients with Inherited Disorders in Mismatch Repair (2210)

Clinical and Molecular Features of a Prospectively-Sequenced Cohort of Glioblastoma Patients with Inherited Disorders in Mismatch Repair (2210)

DDRE-20. TARGETING SPHINGOLIPID PATHWAY REVEALS VULNERABILITY IN IDH1MUT GLIOMA

BACKGROUNDWhile central carbon metabolism has been studied extensively in cancer, lipidomic research is sparse. Sphingolipids participate in cellular functions including secondary messengers, lymphocyte trafficking, inflammation, angiogenesis, migration, proliferation, necrosis and apoptosis, thus highlighting the importance of understanding their role to tumor phenotype. Our investigation into metabolic alterations involving sphingolipid pathway in patient-derived IDH1mut glioma cultures aimed to identify points of metabolic vulnerability.METHODSDysregulation of sphingolipid metabolism was interrogated for brain tumor cultures via LCMS. Expression of enzymes within the pathway was assessed for IDHmut 1/2 and IDHWT glioblastoma patient cohorts via The Cancer Genome Atlas (TCGA) analysis and Western blot for tumor cultures. Biostatic drug response was examined via viability and cytotoxicity assays.RESULTSWe probed the effect that decreasing D-2HG levels with IDH1mut inhibitor (AGI5198) treatments had on sphingolipid metabolism in tumor cultures. The probe revealed N,N-dimethylsphingosine (NDMS), and sphingosine were significantly elevated, while sphingosine-1-phosphate (S1P) was downregulated in IDH1mut cultures following treatment. Drug panel screening revealed that SPHK inhibitor (SPHKi), N,N-dimethylsphingosine in combination with sphingosine triggered lethal dose-dependent response in IDH1mut cultures; contrary to IDHWT. Westerns presented differential expression of SPHK1 and SPHK2 in IDHWT glioblastoma cells, while IDHmut exclusively expressed SPHK1.CONCLUSIONThis novel discovery showed how targeting sphingolipid metabolism in IDH1mut gliomas presents therapeutic implications. Elevated S1P was reported particularly for malignant glioblastomas in prior studies; whereas our research revealed relatively low S1P in the IDHmut compared with IDHWT cultures. In addition to reduced or silenced expression of SPHK2, we postulate that S1P levels in IDHmut gliomas might be closer to a critical threshold allowing treatment with SPHK1i to effectively suspend proliferation and anti-apoptotic defense mechanisms. Our findings revealed that the manipulation of pivotal, endogenous sphingolipids can ultimately trigger apoptosis in IDHmut gliomas. Future studies will probe these targets in preclinical models.

High Expression of FOXP2 Is Associated with Worse Prognosis in Glioblastoma

FOXP2 expression has been associated with the prognosis of some tumors, but the role of FOXP2 in glioblastoma remains unclear. The aim of the present work is to study the role of FOXP2 as a prognostic biomarker in glioblastoma. This is a retrospective observational case series study in which the expression of FOXP2 has been analyzed both at protein level (immunohistochemistry, n= 62) and at mRNA level (RNAseq, in a cohort of glioblastoma patients from The Cancer Genome Atlas [TCGA] database, n= 148). Other molecular and clinical data have also been included in the study, with special focus on miRNA expression data. Survival analysis using log-rank test and COX-regression have been used. Non-parametric statistical tests were also used to study differences between low and high FOXP2 expression groups. Patients with a high expression of FOXP2 protein showed a worse prognosis than those patients with low expression in progression-free survival (hazard ratio 1.711; P= 0.034) and overall survival (hazard ratio 1.809; P= 0.014). These associations were still statistically significant in multivariate analysis. No prognostic association was found with FOXP2 RNA expression. Interestingly, 2 miRNAs that target FOXP2 (hsa-miR-181a-2-3p and hsa-miR-20a-3p) showed an interaction effect on overall survival with FOXP2 expression. A low level of these miRNA expression was associated with a significantly worse prognosis in patients with high FOXP2 RNA expression (log-rank test; P < 0.05). Greater expression of FOXP2 at the protein level is associated with a worse prognosis. This protein expression may be regulated by the expression of specific miRNAs that target FOXP2 mRNA: hsa-miR-181a-2-3p and hsa-miR-20a-3p.

Identification and Validation of ERK5 as a DNA Damage Modulating Drug Target in Glioblastoma.

Simple SummaryGlioblastomas are high-grade brain tumours and are the most common form of malignancy arising in the brain. Patient survival has improved little over the last 40 years, highlighting an urgent unmet need for more effective treatments for these tumours. Current standard-of-care treatment involves surgical removal of as much of the tumour as possible followed by a course of chemo-/radiotherapy. The main chemotherapeutic drug used is called temozolomide, however even with this treatment regimen, the average patient survival following diagnosis is around 15 months. We have identified a protein called ERK5 which is present at higher levels in these high-grade brain tumours compared to normal brain tissue, and which is also associated with resistance to temozolomide and poor patient survival. Additionally, we show that targeting ERK5 in brain tumour cells can improve the effectiveness of temozolomide in killing these tumour cells and offers potential much-needed future clinical benefit to patients diagnosed with glioblastoma.Brain tumours kill more children and adults under 40 than any other cancer, with approximately half of primary brain tumours being diagnosed as high-grade malignancies known as glioblastomas. Despite de-bulking surgery combined with chemo-/radiotherapy regimens, the mean survival for these patients is only around 15 months, with less than 10% surviving over 5 years. This dismal prognosis highlights the urgent need to develop novel agents to improve the treatment of these tumours. To address this need, we carried out a human kinome siRNA screen to identify potential drug targets that augment the effectiveness of temozolomide (TMZ)—the standard-of-care chemotherapeutic agent used to treat glioblastoma. From this we identified ERK5/MAPK7, which we subsequently validated using a range of siRNA and small molecule inhibitors within a panel of glioma cells. Mechanistically, we find that ERK5 promotes efficient repair of TMZ-induced DNA lesions to confer cell survival and clonogenic capacity. Finally, using several glioblastoma patient cohorts we provide target validation data for ERK5 as a novel drug target, revealing that heightened ERK5 expression at both the mRNA and protein level is associated with increased tumour grade and poorer patient survival. Collectively, these findings provide a foundation to develop clinically effective ERK5 targeting strategies in glioblastomas and establish much-needed enhancement of the therapeutic repertoire used to treat this currently incurable disease.

What is the role of the subventricular zone in radiotherapy of glioblastoma patients?

Background and purposeCurrent glioblastoma (GBM) therapies prolong survival, but overall prognosis is still poor. Irradiation of the subventricular zone (SVZ) has recently been discussed as a promising concept as this tissue harbors stem cells which seem to play a role in the initiation and recurrence of GBM. In this study, we retrospectively examined the relationship of SVZ irradiation dose and survival in a large, homogeneous GBM patient cohort. Materials and methodsWe included 200 GBM patients who had been treated at our institution with trimodal therapy (surgery, radiotherapy and chemotherapy) between 2009 and 2020. The SVZ was delineated, and dose-volume histograms were calculated and extracted. Tumors were classified according to their contact with the SVZ. The Kaplan-Meier method was used for survival analysis, and univariable and multivariable Cox regression (MVA) were used to determine prognostic effects on progression-free survival (PFS) and overall survival (OS). ResultsMedian PFS of the study group was 7.2 months; median OS was 15.1 months. In MVA (with mean dose to the ipsilateral SVZ as a continuous covariable), PFS was significantly lower for patients with a Karnofsky performance status (KPS) < 70% and without MGMT promoter methylation. Factors prognostic for shorter OS were old age, lower KPS, unmethylated MGMT status, SVZ contact and biopsy instead of subtotal- or gross total resection. There was no significant correlation between survival and SVZ dose. ConclusionIn this cohort, an increased mean dose to the ipsilateral or contralateral SVZ did not correlate with improved survival in irradiated GBM patients in MVA. Patients whose tumor directly involved the SVZ showed worse OS in MVA.

MerTK activity is not necessary for the proliferation of glioblastoma stem cells

MerTK has been identified as a promising target for therapeutic intervention in glioblastoma. Genetic studies documented a range of oncogenic processes that MerTK targeting could influence, however robust pharmacological validation has been missing. The aim of this study was to assess therapeutic potential of MerTK inhibitors in glioblastoma therapy. Unlike previous studies, our work provides several lines of evidence that MerTK activity is dispensable for glioblastoma growth. We observed heterogeneous responses to MerTK inhibitors that could not be correlated to MerTK inhibition or MerTK expression in cells. The more selective MerTK inhibitors UNC2250 and UNC2580A lack the anti-proliferative potency of less-selective inhibitors exemplified by UNC2025. Functional assays in MerTK-high and MerTK-deficient cells further demonstrate that the anti-cancer efficacy of UNC2025 is MerTK-independent. However, despite its efficacy in vitro, UNC2025 failed to attenuate glioblastoma growth in vivo. Gene expression analysis from cohorts of glioblastoma patients identified that MerTK expression correlates negatively with proliferation and positively with quiescence genes, suggesting that MerTK regulates dormancy rather than proliferation in glioblastoma. In summary, this study demonstrates the importance of orthogonal inhibitors and disease-relevant models in target validation studies and raises a possibility that MerTK inhibitors could be used to target dormant glioblastoma cells.

Hypoxia-inducible miR-196a modulates glioblastoma cell proliferation and migration through complex regulation of NRAS.

Glioblastoma (GBM) is the most common and aggressive malignant brain tumor in humans. Hypoxia has been correlated with the aggressive form of glial tumors, poor prognosis, recurrence and resistance to various therapies. MicroRNAs (miRNAs) have emerged as critical mediators of hypoxic responses and have shown great potential for cancer diagnostics and therapeutics. Here, we focus on the regulatory and functional characterization of miR-196a, a hypoxia-inducible miRNA, in GBM. Hypoxia/HIF regulation of miR-196a was assessed by RT-qPCR, promoter-luciferase and ChIP assays in GBM cell lines. miR-196a levels were analyzed inThe Cancer Genome Atlas (TCGA)-GBM, Chinese Glioma Genome Atlas (CGGA) and Indian GBM patient cohorts. miR-target interactions were studied using RNA/protein quantification and 3'UTR luciferase assays. The effect of miR-196a overexpression/inhibition wasassessed on cellular viability, migration and apoptosis under hypoxia and normoxia. Microarray-based gene expression profiling studies wereperformrd to study the effect of miR-196a ontheGBM cellular transcriptome under hypoxia. Weidentified miR-196a as a hypoxia-inducible and hypoxia-inducible factor (HIF)-regulated miRNA that plays an oncogenic role in GBM. miR-196a was found to be significantly up-regulated in TCGA-GBM, CGGA glioma as well as Indian GBM patient cohorts. miR-196a overexpression was found to induce cellular proliferation, migration, spheroid formation and colony formation and to inhibit apoptosis, while miR-196a inhibition using anti-miR-196ayielded opposite results, suggesting an oncogenic role of miR-196a in GBM. We further unveiled NRAS, AJAP1, TAOK1 and COL24A1 as direct targets of miR-196a. We also report a complex competitive regulation of oncogenic NRAS by miR-196a, miR-146a and let-7 in GBM. Analysis of microarray-based gene expression data obtained by miR-196a inhibition under hypoxia revealed a role of miR-196a inHIF, calcium adhesion, Wnt and cell adhesion pathways. Interestingly, miR-196a was found to positively regulate the expression of various genes involved in the induction or stabilization of HIFs andin maintenance of hypoxic conditions, thereby suggesting the existence of an indirect miR-196a/HIF positive feedback loop under hypoxia. Overall, our work identifies a novel association between hypoxia/HIF signalling and miR-196a in GBM and suggests its therapeutic significance.

BIOM-59. TERT PROMOTER MUTATION AND MGMT PROMOTER METHYLATION-MEDIATED SENSITIVITY TO TEMOZOLOMIDE IN IDH-WILDTYPE GLIOBLASTOMA: IS THERE A LINK?

Abstract BACKGROUND Benefit from temozolomide (TMZ) chemotherapy in the treatment of isocitrate dehydrogenase (IDH)-wildtype glioblastoma is essentially limited to patients with O6-methylguanine DNA methyltransferase (MGMT) promoter-methylated tumors. Recent studies suggest that the impact of the MGMT status on chemosensitivity may be modulated by telomerase reverse transcriptase (TERT) promoter hotspot mutations. METHODS MGMT promoter methylation and TERT promoter mutation status were assessed in an exploratory prospective cohort of IDH-wildtype glioblastoma patients of the German Glioma Network (GGN) (n=298) and validated in a retrospective cohort from Düsseldorf, Germany, and Zurich, Switzerland (n=302). RESULTS In the prospective GGN discovery cohort of patients with MGMT promoter-unmethylated tumors, TERT promoter mutation showed inferior outcome (p=0.044). In contrast, TERT promoter mutations were not associated with improved outcome in patients with MGMT promoter-methylated tumors. Different TERT promoter hotspot mutations were not associated with distinct outcomes. The association of TERT promoter mutation in MGMT promoter-unmethylated tumors was not confirmed in the retrospective validation cohort. CONCLUSIONS Analysis of two independent cohorts of glioblastoma patients, including the prospective GGN cohort, did not confirm previous data suggesting that TERT promoter mutations confer an enhanced benefit from TMZ in patients with MGMT promoter-methylated glioblastoma. Thus, diagnostic testing for TERT promoter mutations may not be required for prediction of TMZ sensitivity in IDH-wildtype glioblastoma patients.

Investigating Glioblastoma Response to Hypoxia.

Glioblastoma (GB) is the most common and deadly type of primary malignant brain tumor with an average patient survival of only 15–17 months. GBs typically have hypoxic regions associated with aggressiveness and chemoresistance. Using patient derived GB cells, we characterized how GB responds to hypoxia. We noted a hypoxia-dependent glycolytic switch characterized by the up-regulation of HK2, PFKFB3, PFKFB4, LDHA, PDK1, SLC2A1/GLUT-1, CA9/CAIX, and SLC16A3/MCT-4. Moreover, many proangiogenic genes and proteins, including VEGFA, VEGFC, VEGFD, PGF/PlGF, ADM, ANGPTL4, and SERPINE1/PAI-1 were up-regulated during hypoxia. We detected the hypoxic induction of invasion proteins, including the plasminogen receptor, S100A10, and the urokinase plasminogen activator receptor, uPAR. Furthermore, we observed a hypoxia-dependent up-regulation of the autophagy genes, BNIP-3 and DDIT4 and of the multi-functional protein, NDRG1 associated with GB chemoresistance; and down-regulation of EGR1 and TFRC (Graphical abstract). Analysis of GB patient cohorts’ revealed differential expression of these genes in patient samples (except SLC16A3) compared to non-neoplastic brain tissue. High expression of SLC2A1, LDHA, PDK1, PFKFB4, HK2, VEGFA, SERPINE1, TFRC, and ADM was associated with significantly lower overall survival. Together these data provide important information regarding GB response to hypoxia which could support the development of more effective treatments for GB patients.

Fractionated Stereotactic Sequential Boost in a Selected Cohort of Glioblastoma Patients: A Mono-institutional Analysis

To retrospectively assess toxicity and survival in 15 selected Glioblastoma patients treated with a sequential fractionated stereotactic radiotherapy (FSRT) boost after chemo-radiotherapy (CHT-RT) and compare their survival outcomes with a control group. Toxicity was assessed with the CTCAE 3.0 scale. The Kaplan-Meier method was used to design survival curves, log-rank test for bivariate analysis and Cox proportional hazard regression model for multivariate analysis. The median follow-up was 16 months (range=5-60). One case of headache and one of radionecrosis (RN) occurred. Median overall survival (OS) was 25 months in the boost group vs. 14 in the no-boost group (p=0.004). Median progression-free survival (PFS) was 15 months in the boost group versus 8 in the no-boost group (p=0.046). At multivariate analysis FSRT boost resulted significantly associated with OS and PFS. In our series a sequential FSRT boost resulted in safe outcomes and significantly associated with survival.

Histopathology-validated machine learning radiographic biomarker for noninvasive discrimination between true progression and pseudo-progression in glioblastoma.

Imaging of glioblastoma patients after maximal safe resection and chemoradiation commonly demonstrates new enhancements that raise concerns about tumor progression. However, in 30% to 50% of patients, these enhancements primarily represent the effects of treatment, or pseudo-progression (PsP). We hypothesize that quantitative machine learning analysis of clinically acquired multiparametric magnetic resonance imaging (mpMRI) can identify subvisual imaging characteristics to provide robust, noninvasive imaging signatures that can distinguish true progression (TP) from PsP. We evaluated independent discovery (n=40) and replication (n=23) cohorts of glioblastoma patients who underwent second resection due to progressive radiographic changes suspicious for recurrence. Deep learning and conventional feature extraction methods were used to extract quantitative characteristics from the mpMRI scans. Multivariate analysis of these features revealed radiophenotypic signatures distinguishing among TP, PsP, and mixed response that compared with similar categories blindly defined by board-certified neuropathologists. Additionally, interinstitutional validation was performed on 20 new patients. Patients who demonstrate TP on neuropathology are significantly different (P<.0001) from those with PsP, showing imaging features reflecting higher angiogenesis, higher cellularity, and lower water concentration. The accuracy of the proposed signature in leave-one-out cross-validation was 87% for predicting PsP (area under the curve [AUC], 0.92) and 84% for predicting TP (AUC, 0.83), whereas in the discovery/replication cohort, the accuracy was 87% for predicting PsP (AUC, 0.84) and 78% for TP (AUC, 0.80). The accuracy in the interinstitutional cohort was 75% (AUC, 0.80). Quantitative mpMRI analysis via machine learning reveals distinctive noninvasive signatures of TP versus PsP after treatment of glioblastoma. Integration of the proposed method into clinical studies can be performed using the freely available Cancer Imaging Phenomics Toolkit.

QOLP-30. CLINICAL PREDICTIVE MODEL FOR THE DEVELOPMENT OF VENOUS THROMBOEMBOLISM IN GLIOBLASTOMA

Abstract INTRODUCTION The risk of venous thromboembolism (VTE) remains high for patients with glioblastoma (GBM) throughout the disease trajectory. Our previous work demonstrated the Khorana scale lacks specificity in this population. We therefore constructed, and attempted to validate a predictive model specific for the development of VTE during adjuvant chemotherapy in glioblastoma patients. METHODS A prior study of GBM patients treated at MD Anderson (MDACC) during the years 2005–2011 found from a multivariate analysis that male sex, BMI ≥ 35, KPS ≤ 80, and steroid therapy were significantly associated with the development of VTE. A predictive model from the MDACC cohort was created using these risk factors, and we attempted to validate the model in an independent cohort of GBM patients treated at Henry Ford from 2010–2015. RESULTS To develop the model 315 patients from the MDACC cohort were randomly divided into two parts: training (75% of data) used for model building, and validation (25% of data) used for model validation. Using the predictive model, the MDACC validation cohort found 80% sensitivity and 80% specificity. We then validated the model in the Henry Ford cohort of 190 GBM patients of which 50 developed a VTE. In the external validation set, the predictive model was found to have a sensitivity = 78% and specificity = 49.3% (Fisher test p-value = 0.0008). CONCLUSIONS Our predictive model for the development of VTE during adjuvant chemotherapy in GBM patients retained high sensitivity in an external data set, however high specificity was lost. While the specificity in our model was higher than in previous studies examining the Khorona scale in GBM patients, further refinement to improve the models reliability to correctly identify people who will not later develop a VTE may be helpful.

Over ten years overall survival in glioblastoma: A different disease?

PurposeThe reasons why a specific subset of glioblastoma (GBM) patients survive longer than others is still unclear. This study analyzed a cohort of long-term and very-long-term GBM survivors to determine which genetic alterations or patient's characteristics influence survival time. MethodsWe retrospectively reviewed a cohort of GBM patients treated at our institution over the last 20 years, stratifying patients in three groups: those with a survival time ≥ 36 months and < 120 months (LTS), ≥120 months (VLTS), and < 36 months, respectively. Clinical (age, sex, focality, resection degree, Karnofsky performance status), and immunohistochemical and molecular data (Ki-67 expression and multiple genes alterations) were collected. We then utilized principal component analysis, logistic regression, and Cox proportional hazard models to identify those variables associated with survival. ResultsYounger age at presentation (HR = 0.36, 95% CI 0.21–0.67, p = .001), and MGMT promoter [(MGMTp), methylated, HR = 0.57, CI 0.34–0.96, p = .034) were associated with higher odds of VLTS survival. The multivariate analysis showed how the combination of younger age (< 50 years), Ki-67 < 10%, and the coexistence of TERTp not mutated, MGMTp methylated, and IDH1/2 mutated in the same patient are also associated with higher odds of survival (HR = 0.10, CI 0.01–0.74, p = .025). ConclusionsWe confirmed younger age at presentation and MGMTp methylation as the only independent factors associated with VLTS. The exceptional survival of our VLTS patients is probably associated with different, still understudied, gene mutations, or with the coexistence of multiple factors.

P14.55 FOXP2 as a prognostic biomarker in glioblastoma

Abstract Background FOXP2 is an essential transcription factor for the development of language. It has also a high expression in some neosplasm, with a differential role as a prognostic biomarker. However, the role of FOXP2 in glioblastoma (GBM) has not been studied until now. AIM To study the role of FOXP2 as a prognostic biomarker in GBM. METHODS This is a retrospective observational case series study in which the expression of FOXP2 has been analyzed at the protein level (immunochemistry, in probes from patients treated in our Center) and at the level of mRNA (RNAseq, in a cohort of GBM patients from The Cancer Genome Atlas [TCGA] database). The expression of some target genes of FOXP2 (RNAseq) has also been studied in the TCGA cohort. Survival analysis using Log-Rank test and COX regression (uni- and multivariate) have been used. The analysis included also other molecular and clinical features of interest in GBM. RESULTS At protein level, FOXP2 was expressed in more than 90% of patients, with a mean positivity of 28.33% (SD=32.29). At mRNA level, FOXP2 was expressed in 79% of patients, with a mean expression of 29.72 RPKM (SD=63.77). Patients with a high expression of FOXP2 at protein level showed a worse prognosis than those patients with low expression in both, progression free survival (HR=1.711; p=.034) and overall survival (HR=1.809; p=.014). These associations maintained their significance in multivariate analysis. Nevertheless, no prognostic association was found with mRNA FOXP2 expression. Interestingly, FOXP2 target genes did not show the expected regulation described in previous works. Some of these genes has been involved in oncogenesis. CONCLUSION FOXP2 is expressed in many GBM patients. Higher expression of FOXP2 protein may be associated with a worse prognosis, but this has not been confirmed at mRNA. Furthermore, a dysregulation of FOXP2 function may be present in tumoral cells.

P14.101 Glioblastoma survival outcomes related to cortical/neural stem cells regions

Abstract BACKGROUND Subventricular brain zone (SVZ) and hippocampal regions are supposed to harbor astrocyte-like neural stem cells. While some tumors may arise from transformed SVZ stem cells, other may be initiated by neo-plastic transformation of non-SVZ progenitor cells or mature glial cells. Lim′s et al classification (Neuro-Oncology 2007) of initial glioblastoma location, related to these neural stem cells regions, was predictive for invasive and multifocal tumor phenotype. The aim of this retrospective single-institutional study is to evaluate the relations of this Lim classification on survival parameters in unselected cohort of glioblastoma patients. MATERIAL AND METHODS Patients treated between 2014–2017 were grouped according to initial location of their contrast enhancing lesion as follows: Lim1 (SVZ+⋯SVZ/hippocampal involvement and Cortex+⋯cortex involvement), Lim2 (SVZ+ and Cortex-), Lim3 (SVZ- and Cortex+) and Lim4 (SVZ- and Cortex-). All patients underwent radiotherapy, some patients were indicated to full treatment according to Stupp regimen (at least 3 cycles of adjuvant chemotherapy after postsurgery chemoradiotherapy). RESULTS In total, 144 patients were analyzed (94 men, mean age 59 years). 47 patients (30.5%) were treated according to Stupp regimen. Lim1 classification was presented in 74 (48%) patients, Lim2 in 22 (14.3%), Lim3 in 50 (32.5%) and Lim4 in 8 (5.2%) patients. Cortical structures (Lim1 and Lim3) were involved in 124 (80.5%) patients. Median overall survival was significantly better in patients treated according to Stupp regimen (23.3 vs. 8.6 months, p&lt;0.001). Median overall survival differs in respective Lim groups: 12.3, 5.6, 11.8 and 6.6 months (p=0.07). Better survival was in patients with cortical involvement (Lim1+Lim3): 12.3 vs. 6.4 months (p=0.02), especially in subgroup of patients who were not treated according to Stupp regimen (8.9 vs. 4.4 months, p=0.02) vs. those after Stupp regimen (23 vs. 23.4 months, p=0.7). CONCLUSION Initial location of enhancing glioblastoma was prognostic for overall survival, with better outcomes in patients presented by involvement of cortical structures comparing to subventricular/hippocampal zones. Molecular patterns may further clarify potential effects of neural stem cells in glioma genesis mirrored in different clinical behavior and location of initial tumor. Supported by Ministry of Health of the Czech Republic, grant No. 18-03-00469