Ivy Glioblastoma Atlas Project Research Articles

IMMU-34. PREDICTION OF TUMOR-ASSOCIATED ANTIGENS USING RADIOGENOMICS

Abstract INTRODUCTION Glioblastoma remains a disease with dismal prognosis. Immunotherapy demonstrated promising results against some refractory cancers. However, comparative results have not been accomplished in glioblastoma therapy due to lack of suitable tumor targets and considerable clonal heterogeneity. OBJECTIVE We propose leveraging radiogenomics using machine learning based on Pre-Op MRI to predict tumor antigen expression profiles displaying specific targets for immunotherapy. METHODS We collected RNAseq data and Pre-Op Axial T2-FLAIR images in 29 patients from IVY Glioblastoma Atlas Project. MRIs were segmented using LIFEx software as follows: 1) tumor + edema (T+E) and 2) whole brain (WB). The intended prediction was for tumor-associated antigens. H2O.ai software was used for prediction using an automated supervised machine learning algorithm. Antigens that achieved >55% area under the curve (i.e., prediction) were reanalyzed adjusting for reproducibility, time of analysis and maximum precision. RESULTS An average of 53.2 T2-FLAIR axial slices per patient were used for 3D reconstruction. We selected the following genes (ERBB2, TP53, IL10RA, CD163, EGFR and MGMT) for downstream analysis given their clinical, immunologic, and prognostic relevance. Regarding tumor-associated antigen prediction for ERBB2, TP53, IL10RA, CD163, EGFR and MGMT predicted 52%, 56%, 66%, 66%, 56% and 0%, on T+E segmentation respectively; The values were 49%, 93%, 87%, 47%, 37% and 0% for the respective WB group. Reanalysis revealed 100% prediction accuracy for the presence of mutant TP53 and IL10RA in both segmentation groups (T+E, WB). CD163 and EGFR prediction was 54% and 61% for the T+E group and 48% for both antigens in the WB group. CONCLUSION We successfully predicted the presence of two out of six antigens (i.e., mutant TP53 and IL10RA) in this patient cohort. Our approach can provide expeditious antigen identification to enable cellular therapy manufacturing for patients with unresectable newly diagnosed tumorsor those which relapsed geographically. Furthermore, this MRI-based methodology might provide a precision medicine approach to places that cannot afford costly testing (i.e., RNAseq).

BIOM-18. MAPPING TUMOR HABITATS IN IDH-WILD TYPE GLIOBLASTOMA: INTEGRATING MR IMAGING, PATHOLOGIC, AND RNA DATA FROM IVY GLIOBLASTOMA ATLAS PROJECT

Abstract BACKGROUND To identify biologic correlates and spatially validate intratumoral subregions (tumor habitat) using diffusion-weighted and dynamic susceptibility contrast-imaging on a comprehensive pathology map of the isocitrate dehydrogenase (IDH)-wild type whole glioblastoma sample using the Ivy Glioblastoma Atlas Project (Ivy GAP). METHODS Complete data of 20 patients with 168 slides of IDH-wild type glioblastoma were obtained from the IvyGAP (http://glioblastoma.alleninstitute.org/). On MRI, tumor habitats were defined using voxel-wise clustering of apparent diffusion coefficient (ADC) and cerebral blood volume (CBV) maps for contrast-enhancing lesion (CEL) and peritumoral non-enhancing lesion (NEL). On pathology slides, normalized areas of leading edge (LE), infiltrating tumor (IT), cellular tumor (CT), hypervascular lesion (CThypervascular), and perinecrotic lesion (CTperinecrotic) were obtained. Gross specimen pictures were co-registered on MRI images for co-localization and correlation between MRI tumor habitats. Pathologic areas were calculated using Pearson’s correlation coefficient. RNA sequencing of 67 RNA-seq samples was assessed using 4 Neftel subtypes and further correlated with pathology. RESULTS Six tumor habitats were correlated: hypervascular, hypovascular cellular, and hypovascular hypocellular habitats for CEL and NEL. CT was strongly correlated with hypovascular cellular habitat in CEL (C2, r = 0.238, p =.005). IT was strongly correlated with hypovascular cellular habitat in NEL (C5, r = 0.294, p =.017). CThypervascular was correlated with hypervascular habitat in NEL (r = 0.195, p =.023). CTperinecrotic was correlated with imaging necrosis (r = 0.199, p =.005). Astrocyte-like subtypes were positively correlated with IT (r = 0.256, p <.001), while mesenchymal-like subtypes were positively correlated with CTperinecrotic area (r = 0.246, p <.001). CONCLUSION Pathologically matched tumor subregions were cellular tumor with hypovascular cellular habitat in CEL and infiltrative tumor with hypovascular cellular habitat in NEL. Identification of the most aggressive tumor portion, and infiltrative tumor portion using non-invasive imaging can be achieved using MRI tumor habitats.

Mapping Tumor Habitats in IDH-Wild Type Glioblastoma: Integrating MR Imaging, Pathologic, and RNA Data from Ivy Glioblastoma Atlas Project.

To spatially validate intratumoral subregions (tumor habitat) using physiologic MRI on pathology of the isocitrate dehydrogenase (IDH)-wildtype whole-glioblastoma sample. Data of 20 patients (168 slides) were obtained from the Ivy Glioblastoma Atlas Project. On MRI, tumor habitats were defined using voxel-wise clustering of apparent diffusion coefficient (ADC) and cerebral blood volume (CBV) maps for contrast-enhancing lesion (CEL) and non-enhancing lesion (NEL). On pathology slides, normalized areas of leading edge (LE), infiltrating tumor (IT), cellular tumor (CT), hypervascular lesion (CThypervascular), and perinecrotic lesion (CTperinecrotic) were obtained. Gross specimen was co-registered on MRI and correlation between pathology-MRI habitats was calculated. RNA sequencing of 67 samples was assessed using 4 Neftel subtypes and further correlated with pathology. Six tumor habitats were identified: hypervascular, hypovascular cellular, and hypovascular hypocellular habitats for CEL and NEL. CT was correlated with hypovascular cellular habitat in CEL (r= 0.238, p =.005). IT was correlated with hypovascular cellular habitat in NEL (r= 0.294, p =.017). CThypervascular was correlated with hypervascular habitat in NEL (r= 0.195, p = .023). CTperinecrotic was correlated with imaging necrosis (r= 0.199, p =.005). Astrocyte-like subtypes were correlated with IT (r= 0.256, p <.001), while mesenchymal-like subtypes were correlated with CTperinecrotic area (r= 0.246, p <.001). Pathologically matched tumor subregions were cellular tumor with hypovascular cellular habitat in CEL and infiltrative tumor with hypovascular cellular habitat in NEL. Identification of the most aggressive as well as infiltrative tumor portion can be achieved using non-invasive MRI tumor habitats.

MRI-based classification of IDH mutation and 1p/19q codeletion status of gliomas using a 2.5D hybrid multi-task convolutional neural network.

IDH mutation and 1p/19q codeletion status are important prognostic markers for glioma that are currently determined using invasive procedures. Our goal was to develop artificial intelligence-based methods to noninvasively determine molecular alterations from MRI. Pre-operative MRI scans of 2648 glioma patients were collected from Washington University School of Medicine (WUSM; n = 835) and publicly available Brain Tumor Segmentation (BraTS; n = 378), LGG 1p/19q (n = 159), Ivy Glioblastoma Atlas Project (Ivy GAP; n = 41), The Cancer Genome Atlas (TCGA; n = 461), and the Erasmus Glioma Database (EGD; n = 774) datasets. A 2.5D hybrid convolutional neural network was proposed to simultaneously localize glioma and classify its molecular status by leveraging MRI imaging features and prior knowledge features from clinical records and tumor location. The models were trained on 223 and 348 cases for IDH and 1p/19q tasks, respectively, and tested on one internal (TCGA) and two external (WUSM and EGD) test sets. For IDH, the best-performing model achieved areas under the receiver operating characteristic (AUROC) of 0.925, 0.874, 0.933 and areas under the precision-recall curves (AUPRC) of 0.899, 0.702, 0.853 on the internal, WUSM, and EGD test sets, respectively. For 1p/19q, the best model achieved AUROCs of 0.782, 0.754, 0.842, and AUPRCs of 0.588, 0.713, 0.782, on those three data-splits, respectively. The high accuracy of the model on unseen data showcases its generalization capabilities and suggests its potential to perform "virtual biopsy" for tailoring treatment planning and overall clinical management of gliomas.

Methionine regulates self-renewal, pluripotency, and cell death of GIC through cholesterol—rRNA axis

BackgroundGlioma-initiating cells (GICs) are the source of glioma cells that can self-renew, have pluripotency, and are treatment-resistant, so are the starting point for relapse and eventual death despite multimodality therapy. L-[methyl-11C] methionine PET has observed high accumulation at the time of recurrence, it is important to understand the mechanism of tumor cell activation caused by the reorganization of methionine metabolism. MethodsWe cultured cells in methionine-deprived culture medium for comprehensive analysis. Based on the obtained results, the possible target molecules were chemically inhibited and the respective markers were analyzed.ResultsMethionine depletion markedly decreased proliferation and increased cell death of GICs. Decreased S-adenosyl-methionine, which is synthesized intracellularly by catalyzed by methionine adenosyltransferase using methionine, triggered the following: (i) global DNA demethylation, (ii) hyper-methylation of signaling pathways regulating pluripotency of stem cells, (iii) decreased expression of the core-genes and pluripotent markers of stem cells including FOXM1, SOX2, SOX4, PROM1, and OLIG2, (iv) decreased cholesterol synthesis and increased excretion mainly through decreased SREBF2, and (v) down-regulation of the large subunit of ribosomal protein configured 28S and ACA43, small nucleolar RNA guiding the pseudouridylation of 28S rRNA, which is essential for translation. In addition, inhibition of cholesterol synthesis with statin resulted in a phenotype similar to that of methionine depletion and decreases in stem cell markers and small nucleolar RNA ACA43. Moreover, suppression of FOXM1 decreased stem cell markers such as SOX4 and PROM1. The gene expression profile for cholesterol production was obtained from the Ivy Glioblastoma Atlas Project database and compared between tumor cells with relatively low methionine levels in areas of pseudopalisading arrangement around necrosis and tumor cells in the infiltrating region, showing that cells in the infiltrating region have higher capacity to produce cholesterol.ConclusionsMethionine metabolism is closely related with self-renewal, pluripotency, and cell death in GICs through modification of cholesterol biosynthesis, especially in the SREBF2-FOXM1 and ACA43 axis with modification of rRNA.

TMIC-52. RELATIONSHIP BETWEEN MACROPHAGE AND RADIOSENSITIVITY IN HUMAN PRIMARY AND RECURRENT GLIOBLASTOMA: IN SILICO ANALYSIS WITH PUBLICLY AVAILABLE DATASETS

Abstract The glioblastoma microenvironment predominantly contains tumor-associated macrophages that support tumor growth and invasion. We investigated the relationship between tumor radiosensitivity and infiltrating M1/M2 macrophage profiles in public datasets of primary and recurrent glioblastoma. We estimated the radiosensitivity index (RSI) score based on gene expression rankings. Macrophages were profiled using the deconvolution algorithm CIBERSORTx. Samples from The Cancer Genome Atlas (TCGA), Chinese Glioma Genome Atlas (CGGA), the Ivy Glioblastoma Atlas Project dataset, a single-cell RNA sequencing dataset (GSE84465), Glioma Longitudinal Analysis Consortium (GLASS), and an immunotherapy trial dataset (GSE121810) were included. RSI-high radioresistant tumors were associated with worse overall survival in TCGA and CGGA than RSI-low tumors. M1/M2 macrophage ratios and RSI scores were inversely associated, indicating that radioresistant glioblastoma tumor microenvironments contain more M2 than M1 macrophages. In the single-cell RNA sequencing dataset, the mean RSI of neoplastic cells was positively correlated with high M2 macrophages proportions. A favorable response to programmed cell death protein 1 (PD-1) therapy was observed in recurrent glioblastomas with high M1/M2 macrophage ratios and low RSI scores. In patients with recurrent glioblastoma, fewer M2 macrophages and low RSI scores were associated with improved overall survival. High M2 macrophage proportions may be involved in radioresistant glioblastoma.

Differential expression of checkpoint markers in the normoxic and hypoxic microenvironment of glioblastomas.

Glioblastoma is the most common primary malignant brain tumor in adults with an overall survival of only 14.6months. Hypoxia is known to play a role in tumor aggressiveness but the influence of hypoxia on the immune microenvironment is not fully understood. The aim of this study was to investigate the expression of immune-related proteins in normoxic and hypoxic tumor areas by digital spatial profiling. Tissue samples from 10 glioblastomas were stained with a panel of 40 antibodies conjugated to photo-cleavable oligonucleotides. The free oligo-tags from normoxic and hypoxic areas were hybridized to barcodes for digital counting. Differential expression patterns were validated by Ivy Glioblastoma Atlas Project (GAP) data and an independent patient cohort. We found that CD44, Beta-catenin and B7-H3 were upregulated in hypoxia, whereas VISTA, CD56, KI-67, CD68 and CD11c were downregulated. PD-L1 and PD-1 were not affected by hypoxia. Focusing on the checkpoint-related markers CD44, B7-H3 and VISTA, our findings for CD44 and VISTA could be confirmed with Ivy GAP RNA sequencing data. Immunohistochemical staining and digital quantification of CD44, B7-H3 and VISTA in an independent cohort confirmed our findings for all three markers. Additional stainings revealed fewer T cells and high but equal amounts of tumor-associated microglia and macrophages in both hypoxic and normoxic regions. In conclusion, we found that CD44 and B7-H3 were upregulated in areas with hypoxia whereas VISTA was downregulated together with the presence of fewer T cells. This heterogeneous expression should be taken into consideration when developing novel therapeutic strategies.

Hypoxia-induced polypoid giant cancer cells in glioma promote the transformation of tumor-associated macrophages to a tumor-supportive phenotype.

AimsPolypoid giant cancer cells (PGCCs) represent a unique subgroup of stem‐like cells, acting as a critical factor in promoting the recurrence of various solid tumors. The effect of PGCCs on the tumor malignancy of glioma and its immune microenvironment remains unclear.MethodsBioinformatic analysis was performed to investigate the relationship between M2 tumor‐associated macrophages (TAMs) infiltration and survival of glioblastoma (GBM) patients. The spatial location of M2 TAMs in GBM was also investigated using the Ivy Glioblastoma Atlas Project (Ivy GAP) database. PGCCs were quantified in glioma of different grades. CoCl2 was used to induce PGCCs in cultures of A172 cells. PGCCs, and their progeny cells in cultures were further evaluated for morphological features, tumorsphere formation, and TAMs activation.ResultsThe magnitude of M2 TAMs infiltration is significantly correlated with poor survival in GBM patients. M2 TAMs were enriched in the perinecrotic zone (PNZ) of GBM and positively correlated with hypoxic levels. Increased PGCCs were detected in glioma specimens of higher grades. CoCl2 induced hypoxia and the transformation of A172 cultures into PGCCs, producing the progeny cells, PGCCs‐Dau, through asymmetric division. PGCCs and PGCCs‐Dau possessed tumor stem cell‐like features, while PGCCs‐Dau enhanced the polarization of TAMs into an M2 phenotype with relevance to immunosuppression and malignancy in GBM.ConclusionsPGCCs promote malignancy and immune‐suppressive microenvironment in GBM. PGCCs or their progeny cells may be a potential therapeutic target for GBM.

CD44-Mediated Poor Prognosis in Glioma Is Associated With M2-Polarization of Tumor-Associated Macrophages and Immunosuppression.

BackgroundGlioma is the most common primary brain tumor with a poor prognosis. Key genes that are negatively related to prognosis may provide the therapy targets to cure glioma. To clarify the role of CD44 in glioma, we explored its function at bulk-transcriptome, spatial and single-cell transcriptome levels.MethodsIn total, expression profiles with survival data of whole-grade glioma from The Cancer Genome Atlas (TCGA) and the Chinese Glioma Genome Atlas (CGGA), RNA-seq data with anatomic information of glioblastoma (GBM) from the Ivy Glioblastoma Atlas Project, RNA-sequencing (RNA-seq) data from recurrent GBM receiving adjuvant anti-PD-1 immunotherapy accessed through GSE121810, and single-cell RNA-seq data of GBM under accession GSE103224 were enrolled in this study. CD44-specific findings were further analyzed by R language.ResultsCD44 is positively correlated with WHO grade of malignancy and is negatively related to prognosis in glioma. Meanwhile, CD44 predominantly expresses in GBM mesenchymal subtype, and gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses reveal that CD44 positively coexpressed genes are closely related to glioma immunity. Moreover, CD44+ cells mainly distribute in perinecrotic region with high expression of immune factors. At single-cell resolution, only malignant tumor cells, tumor-associated macrophages (TAMs), and T cells express CD44 in GBM. CD44+ malignant tumor cells are in mesenchymal-1-like (MES1-like) cellular state, and CD44+ TAMs are in M2 phenotype. CD44+ T cells have high expression of both PD-1 and PD-L1. CD44 and its directly interacted inhibitory immunomodulators are upregulated in patients with nonresponder recurrent GBM treated with PD-1 blockade therapy.ConclusionOur work demonstrates that CD44, a new M2 TAM biomarker, is involved in immune suppressor and promote glioma progression in glioma microenvironment. These results expand our understanding of CD44-specific clinical and immune features in glioma.

Relationship between Macrophage and Radiosensitivity in Human Primary and Recurrent Glioblastoma: In Silico Analysis with Publicly Available Datasets.

The glioblastoma microenvironment predominantly contains tumor-associated macrophages that support tumor growth and invasion. We investigated the relationship between tumor radiosensitivity and infiltrating M1/M2 macrophage profiles in public datasets of primary and recurrent glioblastoma. We estimated the radiosensitivity index (RSI) score based on gene expression rankings. Macrophages were profiled using the deconvolution algorithm CIBERSORTx. Samples from The Cancer Genome Atlas (TCGA), Chinese Glioma Genome Atlas (CGGA), the Ivy Glioblastoma Atlas Project dataset, a single-cell RNA sequencing dataset (GSE84465), Glioma Longitudinal Analysis Consortium (GLASS), and an immunotherapy trial dataset (GSE121810) were included. RSI-high radioresistant tumors were associated with worse overall survival in TCGA and CGGA than RSI-low tumors. M1/M2 macrophage ratios and RSI scores were inversely associated, indicating that radioresistant glioblastoma tumor microenvironments contain more M2 than M1 macrophages. In the single-cell RNA sequencing dataset, the mean RSI of neoplastic cells was positively correlated with high M2 macrophages proportions. A favorable response to programmed cell death protein 1 (PD-1) therapy was observed in recurrent glioblastomas with high M1/M2 macrophage ratios and low RSI scores. In patients with recurrent glioblastoma, fewer M2 macrophages and low RSI scores were associated with improved overall survival. High M2 macrophage proportions may be involved in radioresistant glioblastoma.

CBMS-10 Methionine metabolism closely related with self-renew, pluripotency and cell death in GICs through modification of cholesterol biosynthesis and ribosomal RNA

Glioma initiating cells (GICs) are the source of glioma cells that have the ability to self-renew and pluripotency, which are treatment-resistant, starting point for relapse and eventual death despite multimodality therapy. Since high accumulation is observed in 11cMet-PET at the time of recurrence, it is important to understand the mechanism of tumor cell activation caused by the reorganization of methionine metabolism. We cultured cells in methionine-deprived culture medium and performed a comprehensive analysis, and found that methionine depletion markedly decreased proliferation and increasing cell death of GICs. Decreased SAM, which is synthesized intracellularly catalyzed by methionine adenosyltransferase (MAT) using methionine, triggered the following: (i) global DNA demethylation, (ii) hyper-methylation of signaling pathways regulating pluripotentcy of stem cells, (iii) decreased expression of the core-genes and pluripotent marker of stem cells including FOXM1, SOX2, SOX4, PROM1 and OLIG2, (iv) decreased cholesterol synthesis and increased excretion mainly through decreased SREBF2 (v) down-regulation of the large subunit of ribosomal protein configured 28S and ACA43, snoRNA guiding the pseudouridylation of 28S ribosomal RNA, which has crucial role for translation. In addition, inhibition of cholesterol synthesis with statin resulted in a phenotype similar to that of methionine removal and a decrease in stem cell markers and snoRNA ACA43. Moreover, suppression of FOXM1 decreased stem cell markers such as SOX4 and PROM1. The gene expression profile for cholesterol production was obtained from the Ivy Glioblastoma Atlas Project (IVYGAP) database and compared between tumour cells with relatively low methionine levels in area of pseudopalisading arrangement around necrosis and tumour cells in the infiltrating region, showing that cells cells in the infiltrating region have a higher capacity to produce cholesterol. Taken together, methionine metabolism closely related with self-renew, pluripotency and cell death in GICs through modification of cholesterol biosynthesis: especially SREBF2-FOXM1 and ACA43 axis with modification of ribosomal RNA.

NIMG-60. PREDICTING OVERALL SURVIVAL IN GLIOBLASTOMA USING HISTOPATHOLOGY VIA AN END-TO-END DEEP LEARNING PIPELINE: A LARGE MULTI-COHORT STUDY

Abstract PURPOSE Glioblastoma is an aggressive and universally fatal tumor. Morphological information as captured from cellular regions on surgically resected histopathology slides has the ability to reveal the inherent heterogeneity in Glioblastoma and thus has prognostic implications. In this work, we hypothesized that capturing morphological attributes from high cellularity regions on Hematoxylin and Eosin (H&amp;E)-stained digitized tissue slides using an end-to-end deep-learning pipeline will enable risk-stratification of GBM tumors based on overall survival. METHODS A large multi-cohort study consisting of N=514 H&amp;E-stained digitized tissue slides along with overall-survival data (OS) was obtained from the Ivy Glioblastoma atlas project (Ivy-GAP (N=41)), TCGA (N=379), and CPTAC (N=94). Our deep-learning pipeline consisted of two stages. First stage involved segmenting cellular tumor (CT) from necrotic-regions and background using Resnet-18 model, while the second stage involved predicting OS, using only the segmented CT regions identified in the first stage. For the segmentation stage, we leveraged the Ivy-GAP cohort, where CT annotations confirmed by expert neuropathologists were available, to serve as the training set. Using this training model, the CT regions on the remaining cohort (TCGA, CPTAC) (i.e. test set) were identified. For the survival-prediction stage, the last layer of ResNet18 model was replaced with a cox layer (ResNet-Cox), and further fine-tuned using OS and censor information. Independent validation of ResNet-Cox was performed on two hold-out sites from TCGA and one from CPTAC. RESULTS Our segmentation model achieved an accuracy of 0.89 in reliably identifying CT regions on the validation data. The segmented CT regions on the test cohort were further confirmed by two experts. Our ResNet-Cox model achieved a concordance-index of 0.73 on MD Anderson Cancer Center (N=60), 0.71 on Henry Ford Hospital (N=96), and 0.68 on CPTAC data (N=41). CONCLUSION Deep-learning features captured from cellular tumor of H&amp;E-stained histopathology images may predict survival in Glioblastoma.

TAMI-32. CORRELATION BETWEEN RADIOSENSITIVITY INDEX AND M2 MACROPHAGE PROPORTION IN TUMOR MICROENVIRONMENT OF GLIOBLASTOMA

Abstract BACKGROUND Tumor-associated macrophages (TAMs) Macrophage are predominant in glioblastoma tumor microenvironment (TME), supporting for neoplastic cell expansion and invasion. We investigated the relationship between radiosensitivity of glioblastoma and M1/M2 macrophage profiles in bulk and single cell RNA sequencing datasets. METHODS We used radiosensitivity index (RSI) gene signature and estimated RSI score based on the ranking of genes by expression level. Two large glioma datasets – The Cancer Genome Atlas (TCGA) and Chinese Glioma Genome Atlas (CGGA) – were employed to identify whether RSI is clinically predictive of overall survival following radiation therapy. To analyze the association between M1/M2 macrophages and RSI within spatial context, the Ivy Glioblastoma Atlas Project dataset was investigated and single cell RNA sequencing dataset (GSE84465) was analyzed as well. Macrophages were profiled using a deconvolution algorithm, CIBERSORTx. RESULTS The RSI-high group having radioresistant tumors showed worse overall survival than the RSI-low group in both the TCGA (HR=1.87, 95% CI=1.06-3.29, P=0.031) and the CGGA (HR=1.61, 95% CI=1.04-2.50, P=0.031) glioblastoma population. In the Ivy Glioblastoma Atlas Project dataset, radiosensitive tumor having lower RSI was significantly more found in more vascular region including hyperplastic and microvascular region (coefficient=-0.07, P=0.001), meanwhile, radioresistant tumor was significantly clustered in necrotic region including perinecrotic and pseudopalisading regions (coefficient=0.07, P&amp;lt; 0.001). The proportion of M1/M2 macrophage and RSI score showed an inverse relationship (coefficient=-0.23, P=0.015), indicating that radioresistant glioblastomas are related with TME having more M2 than M1 macrophage. In single cell RNA sequencing dataset composed of immune and tumor cells collected from four patients, mean RSI of neoplastic cells was positively correlated with high proportion of M2 macrophages. CONCLUSION RSI can predict radiation response in terms of overall survival in glioblastoma patients. High proportion of M2 macrophage may play an important role in TME of radioresistant glioblastoma.

MRI-Based Radiomic Signature for PI3K/AKT-Activated Glioblastoma

Prior work has demonstrated that transcriptional signatures and immune cell infiltrates associated with phosphatidylinositol 3-kinase (PI3K) downregulation are able to distinguish glioblastoma patients who respond to immune checkpoint blockade. PI3K/Akt pathway is involved in tumorigenesis, immune microenvironment regulation, and chemo-radiation treatment resistance. We hypothesized that PI3K/Akt-activated glioblastoma may harbor imaging features that would allow for radiomic discrimination.Patients from The Cancer Genome Atlas and the Ivy Glioblastoma Atlas Project who underwent resection and received temozolomide-based chemoradiation from 1996-2003 were included. All tumors underwent RNA-sequencing and had pre-, post-operative and surveillance MRIs. Single sample Gene Set Enrichment Analysis (ssGSEA) was applied to gene expression data to identify samples with PI3K/Akt pathway activation. A programming environment-based machine learning software was used to extract unique quantitative image features from manually segmented pre-operative tumors from post-contrast T1 sequences. Prior to image analysis, all images underwent z-score normalization over all voxels. Extracted features included shape features, first-order statistical, intensity-histogram based, fractal, local intensity, and texture matrix-based features from both unfiltered and filtered images (wavelet decompositions). A threshold of P < 0.05 was used to identify significant radiomic features. Adaptive LASSO-penalized regression was used to develop a radiomic predictor for PI3K-Akt downregulation. Area under the ROC curve (AUC) assessed performance of the model with three-fold internal cross-validation.Twenty-two patients with glioblastoma were included in the analysis. Tumors from 8 of 22 patients (36.4%) had downregulation of PI3K/Akt signaling by ssGSEA. Among tumors with upregulation of PI3K/Akt signaling, 71% had upstream loss or deletion of PTEN. Tumors with downregulated PI3K/AKT signaling were also more likely to have intratumoral B cell lymphocytes identified by ssGSEA. Overall, 2800 radiomic features were extracted from pre-operative, post-contrast MRI images of 22 patients with glioblastoma. 49 features were significantly different between the two groups. Adaptive LASSO regression analysis was performed to develop a 6-feature radiomic signature with an AUC of 0.955 (95% Confidence Interval, 0.862-1.000).We identified a radiomic signature comprised of 6 features that classifies PI3K/Akt-activated glioblastoma, which is associated with increased immune cell infiltration and potential response to immunotherapy. This signature may help in identifying patients who may benefit from immunomodulation in glioblastoma.

Development and Validation of an Mesenchymal-Related Long Non-Coding RNA Prognostic Model in Glioma.

Glioma is well known as the most aggressive and prevalent primary malignant tumor in the central nervous system. Molecular subtypes and prognosis biomarkers remain a promising research area of gliomas. Notably, the aberrant expression of mesenchymal (MES) subtype related long non-coding RNAs (lncRNAs) is significantly associated with the prognosis of glioma patients. In this study, MES-related genes were obtained from The Cancer Genome Atlas (TCGA) and the Ivy Glioblastoma Atlas Project (Ivy GAP) data sets of glioma, and MES-related lncRNAs were acquired by performing co-expression analysis of these genes. Next, Cox regression analysis was used to establish a prognostic model, that integrated ten MES-related lncRNAs. Glioma patients in TCGA were divided into high-risk and low-risk groups based on the median risk score; compared with the low-risk groups, patients in the high-risk group had shorter survival times. Additionally, we measured the specificity and sensitivity of our model with the ROC curve. Univariate and multivariate Cox analyses showed that the prognostic model was an independent prognostic factor for glioma. To verify the predictive power of these candidate lncRNAs, the corresponding RNA-seq data were downloaded from the Chinese Glioma Genome Atlas (CGGA), and similar results were obtained. Next, we performed the immune cell infiltration profile of patients between two risk groups, and gene set enrichment analysis (GSEA) was performed to detect functional annotation. Finally, the protective factors DGCR10 and HAR1B, and risk factor SNHG18 were selected for functional verification. Knockdown of DGCR10 and HAR1B promoted, whereas knockdown of SNHG18 inhibited the migration and invasion of gliomas. Collectively, we successfully constructed a prognostic model based on a ten MES-related lncRNAs signature, which provides a novel target for predicting the prognosis for glioma patients.

Identification of Novel Transcriptome Signature as a Potential Prognostic Biomarker for Anti-Angiogenic Therapy in Glioblastoma Multiforme.

Glioblastoma multiforme (GBM) is the most common and devastating type of primary brain tumor, with a median survival time of only 15 months. Having a clinically applicable genetic biomarker would lead to a paradigm shift in precise diagnosis, personalized therapeutic decisions, and prognostic prediction for GBM. Radiogenomic profiling connecting radiological imaging features with molecular alterations will offer a noninvasive method for genomic studies of GBM. To this end, we analyzed over 3800 glioma and GBM cases across four independent datasets. The Chinese Glioma Genome Atlas (CGGA) and The Cancer Genome Atlas (TCGA) databases were employed for RNA-Seq analysis, whereas the Ivy Glioblastoma Atlas Project (Ivy-GAP) and The Cancer Imaging Archive (TCIA) provided clinicopathological data. The Clinical Proteomic Tumor Analysis Consortium Glioblastoma Multiforme (CPTAC-GBM) was used for proteomic analysis. We identified a simple three-gene transcriptome signature—SOCS3, VEGFA, and TEK—that can connect GBM’s overall prognosis with genes’ expression and simultaneously correlate radiographical features of perfusion imaging with SOCS3 expression levels. More importantly, the rampant development of neovascularization in GBM offers a promising target for therapeutic intervention. However, treatment with bevacizumab failed to improve overall survival. We identified SOCS3 expression levels as a potential selection marker for patients who may benefit from early initiation of angiogenesis inhibitors.

Differentiated glioblastoma cells accelerate tumor progression by shaping the tumor microenvironment via CCN1-mediated macrophage infiltration

Glioblastoma (GBM) is the most lethal primary brain tumor characterized by significant cellular heterogeneity, namely tumor cells, including GBM stem-like cells (GSCs) and differentiated GBM cells (DGCs), and non-tumor cells such as endothelial cells, vascular pericytes, macrophages, and other types of immune cells. GSCs are essential to drive tumor progression, whereas the biological roles of DGCs are largely unknown. In this study, we focused on the roles of DGCs in the tumor microenvironment. To this end, we extracted DGC-specific signature genes from transcriptomic profiles of matched pairs of in vitro GSC and DGC models. By evaluating the DGC signature using single cell data, we confirmed the presence of cell subpopulations emulated by in vitro culture models within a primary tumor. The DGC signature was correlated with the mesenchymal subtype and a poor prognosis in large GBM cohorts such as The Cancer Genome Atlas and Ivy Glioblastoma Atlas Project. In silico signaling pathway analysis suggested a role of DGCs in macrophage infiltration. Consistent with in silico findings, in vitro DGC models promoted macrophage migration. In vivo, coimplantation of DGCs and GSCs reduced the survival of tumor xenograft-bearing mice and increased macrophage infiltration into tumor tissue compared with transplantation of GSCs alone. DGCs exhibited a significant increase in YAP/TAZ/TEAD activity compared with GSCs. CCN1, a transcriptional target of YAP/TAZ, was selected from the DGC signature as a candidate secreted protein involved in macrophage recruitment. In fact, CCN1 was secreted abundantly from DGCs, but not GSCs. DGCs promoted macrophage migration in vitro and macrophage infiltration into tumor tissue in vivo through secretion of CCN1. Collectively, these results demonstrate that DGCs contribute to GSC-dependent tumor progression by shaping a mesenchymal microenvironment via CCN1-mediated macrophage infiltration. This study provides new insight into the complex GBM microenvironment consisting of heterogeneous cells.

Utilizing Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) to Analyze Interstitial Fluid Flow and Transport in Glioblastoma and the Surrounding Parenchyma in Human Patients.

Background: Glioblastoma (GBM) is the deadliest and most common brain tumor in adults, with poor survival and response to aggressive therapy. Limited access of drugs to tumor cells is one reason for such grim clinical outcomes. A driving force for therapeutic delivery is interstitial fluid flow (IFF), both within the tumor and in the surrounding brain parenchyma. However, convective and diffusive transport mechanisms are understudied. In this study, we examined the application of a novel image analysis method to measure fluid flow and diffusion in GBM patients. Methods: Here, we applied an imaging methodology that had been previously tested and validated in vitro, in silico, and in preclinical models of disease to archival patient data from the Ivy Glioblastoma Atlas Project (GAP) dataset. The analysis required the use of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), which is readily available in the database. The analysis results, which consisted of IFF flow velocity and diffusion coefficients, were then compared to patient outcomes such as survival. Results: We characterized IFF and diffusion patterns in patients. We found strong correlations between flow rates measured within tumors and in the surrounding parenchymal space, where we hypothesized that velocities would be higher. Analyzing overall magnitudes indicated a significant correlation with both age and survival in this patient cohort. Additionally, we found that neither tumor size nor resection significantly altered the velocity magnitude. Lastly, we mapped the flow pathways in patient tumors and found a variability in the degree of directionality that we hypothesize may lead to information concerning treatment, invasive spread, and progression in future studies. Conclusions: An analysis of standard DCE-MRI in patients with GBM offers more information regarding IFF and transport within and around the tumor, shows that IFF is still detected post-resection, and indicates that velocity magnitudes correlate with patient prognosis.

Reproducibility analysis of multi-institutional paired expert annotations and radiomic features of the Ivy Glioblastoma Atlas Project (Ivy GAP) dataset.

The availability of radiographic magnetic resonance imaging (MRI) scans for the Ivy Glioblastoma Atlas Project (Ivy GAP) has opened up opportunities for development of radiomic markers for prognostic/predictive applications in glioblastoma (GBM). In this work, we address two critical challenges with regard to developing robust radiomic approaches: (a) the lack of availability of reliable segmentation labels for glioblastoma tumor sub-compartments (i.e., enhancing tumor, non-enhancing tumor core, peritumoral edematous/infiltrated tissue) and (b) identifying "reproducible" radiomic features that are robust to segmentation variability across readers/sites. From TCIA's Ivy GAP cohort, we obtained a paired set (n=31) of expert annotations approved by two board-certified neuroradiologists at the Hospital of the University of Pennsylvania (UPenn) and at Case Western Reserve University (CWRU). For these studies, we performed a reproducibility study that assessed the variability in (a) segmentation labels and (b) radiomic features, between these paired annotations. The radiomic variability was assessed on a comprehensive panel of 11700 radiomic features including intensity, volumetric, morphologic, histogram-based, and textural parameters, extracted for each of the paired sets of annotations. Our results demonstrated (a) a high level of inter-rater agreement (median value of DICE≥0.8 for all sub-compartments), and (b) ≈24% of the extracted radiomic features being highly correlated (based on Spearman's rank correlation coefficient) to annotation variations. These robust features largely belonged to morphology (describing shape characteristics), intensity (capturing intensity profile statistics), and COLLAGE (capturing heterogeneity in gradient orientations) feature families. We make publicly available on TCIA's Analysis Results Directory (https://doi.org/10.7937/9j41-7d44), the complete set of (a) multi-institutional expert annotations for the tumor sub-compartments, (b) 11700 radiomic features, and (c) the associated reproducibility meta-analysis. The annotations and the associated meta-data for Ivy GAP are released with the purpose of enabling researchers toward developing image-based biomarkers for prognostic/predictive applications in GBM.

NIMG-09. PREDICTING OVERALL SURVIVAL OF GLIOBLASTOMA PATIENTS ON MULTI-INSTITUTIONAL HISTOPATHOLOGY STAINED SLIDES USING DEEP LEARNING AND POPULATION-BASED NORMALIZATION

Abstract MOTIVATION Glioblastoma is the most common and aggressive adult brain tumor. Clinical histopathologic evaluation is essential for tumor classification, which according to the World Health Organization is associated with prognostic information. Accurate prediction of patient overall survival (OS) from clinical routine baseline histopathology whole slide images (WSI) using advanced computational methods, while considering variations in the staining process, could contribute to clinical decision-making and patient management optimization. METHODS We utilize The Cancer Genome Atlas glioblastoma (TCGA-GBM) collection, comprising multi-institutional hematoxylin and eosin (H&amp;E) stained frozen top-section WSI, genomic, and clinical data from 121 subjects. Data are randomly split into training (80%), validation (10%), and testing (10%) sets, while proportionally keeping the ratio of censored patients. We propose a novel deep learning algorithm to identify survival-discriminative histopathological patterns in a WSI, through feature maps, and quantitatively integrate them with gene expression and clinical data to predict patient OS. The concordance index (C-index) is used to quantify the predictive OS performance. Variations in slide staining are assessed through a novel population-based stain normalization approach, informed of glioblastoma distinct histologic sub-regions and their appearance from 509 H&amp;E stained slides with corresponding anatomical annotations from the Ivy Glioblastoma Atlas Project (IvyGAP). RESULTS C-index was equal to 0.797, 0.713, and 0.703 for the training, validation, and testing data, respectively, prior to stain normalization. Following normalization, staining variations in H&amp;E and ‘E’ gained significant improvements in IvyGAP (pWilcoxon&amp;lt; 0.01) and TCGA-GBM (pWilcoxon&amp;lt; 0.0001) data, respectively. These improvements contributed to further optimizing the C-index to 0.871, 0.777, and 0.780 for the training, validation, and testing data, respectively. CONCLUSIONS Appropriate normalization and integrative deep learning yield accurate OS prediction of glioblastoma patients through H&amp;E slides, generalizable in multi-institutional data, potentially contributing to patient stratification in clinical trials. Our computationally-identified survival-discriminative histopathological patterns can contribute in further understanding glioblastoma.