A Landscape of Pharmacogenomic Interactions in Cancer.

Systematic studies of cancer genomes have provided unprecedented insights into the molecular nature of cancer. Using this information to guide the development and application of therapies in the clinic is challenging. Here, we report how cancer-driven alterations identified in 11,289 tumors from 29 tissues (integrating somatic mutations, copy number alterations, DNA methylation, and gene expression) can be mapped onto 1,001 molecularly annotated human cancer cell lines and correlated with sensitivity to 265 drugs. We find that cell lines faithfully recapitulate oncogenic alterations identified in tumors, find that many of these associate with drug sensitivity/resistance, and highlight the importance of tissue lineage in mediating drug response. Logic-based modeling uncovers combinations of alterations that sensitize to drugs, while machine learning demonstrates the relative importance of different data types in predicting drug response. Our analysis and datasets are rich resources to link genotypes with cellular phenotypes and to identify therapeutic options for selected cancer sub-populations.

Effective drug combinations in breast, colon and pancreatic cancer cells.

Combinations of anti-cancer drugs can overcome resistance and provide new treatments1,2. The number of possible drug combinations vastly exceeds what could be tested clinically. Efforts to systematically identify active combinations and the tissues and molecular contexts in which they are most effective could accelerate the development of combination treatments. Here we evaluate the potency and efficacy of 2,025 clinically relevant two-drug combinations, generating a dataset encompassing 125 molecularly characterized breast, colorectal and pancreatic cancer cell lines. We show that synergy between drugs is rare and highly context-dependent, and that combinations of targeted agents are most likely to be synergistic. We incorporate multi-omic molecular features to identify combination biomarkers and specify synergistic drug combinations and their active contexts, including in basal-like breast cancer, and microsatellite-stable or KRAS-mutant colon cancer. Our results show that irinotecan and CHEK1 inhibition have synergistic effects in microsatellite-stable or KRAS-TP53 double-mutant colon cancer cells, leading to apoptosis and suppression of tumour xenograft growth. This study identifies clinically relevant effective drug combinations in distinct molecular subpopulations and is a resource to guide rational efforts to develop combinatorial drug treatments.

Predicting patient response with models trained on cell lines and patient-derived xenografts by nonlinear transfer learning.

Preclinical models have been the workhorse of cancer research, producing massive amounts of drug response data. Unfortunately, translating response biomarkers derived from these datasets to human tumors has proven to be particularly challenging. To address this challenge, we developed TRANSACT, a computational framework that builds a consensus space to capture biological processes common to preclinical models and human tumors and exploits this space to construct drug response predictors that robustly transfer from preclinical models to human tumors. TRANSACT performs favorably compared to four competing approaches, including two deep learning approaches, on a set of 23 drug prediction challenges on The Cancer Genome Atlas and 226 metastatic tumors from the Hartwig Medical Foundation. We demonstrate that response predictions deliver a robust performance for a number of therapies of high clinical importance: platinum-based chemotherapies, gemcitabine, and paclitaxel. In contrast to other approaches, we demonstrate the interpretability of the TRANSACT predictors by correctly identifying known biomarkers of targeted therapies, and we propose potential mechanisms that mediate the resistance to two chemotherapeutic agents.

Subunit composition of VRAC channels determines substrate specificity and cellular resistance to Pt-based anti-cancer drugs.

Although platinum-based drugs are widely used chemotherapeutics for cancer treatment, the determinants of tumor cell responsiveness remain poorly understood. We show that the loss of subunits LRRC8A and LRRC8D of the heteromeric LRRC8 volume-regulated anion channels (VRACs) increased resistance to clinically relevant cisplatin/carboplatin concentrations. Under isotonic conditions, about 50% of cisplatin uptake depended on LRRC8A and LRRC8D, but neither on LRRC8C nor on LRRC8E. Cell swelling strongly enhanced LRRC8-dependent cisplatin uptake, bolstering the notion that cisplatin enters cells through VRAC. LRRC8A disruption also suppressed drug-induced apoptosis independently from drug uptake, possibly by impairing VRAC-dependent apoptotic cell volume decrease. Hence, by mediating cisplatin uptake and facilitating apoptosis, VRAC plays a dual role in the cellular drug response. Incorporation of the LRRC8D subunit into VRAC substantially increased its permeability for cisplatin and the cellular osmolyte taurine, indicating that LRRC8 proteins form the channel pore. Our work suggests that LRRC8D-containing VRACs are crucial for cell volume regulation by an important organic osmolyte and may influence cisplatin/carboplatin responsiveness of tumors.

Pathogenic and targetable genetic alterations in 70 urachal adenocarcinomas.

Urachal cancer (UrC) is a rare but aggressive malignancy often diagnosed in advanced stages requiring systemic treatment. Although cytotoxic chemotherapy is of limited effectiveness, prospective clinical studies can hardly be conducted. Targeted therapeutic treatment approaches and potentially immunotherapy based on a biological rationale may provide an alternative strategy. We therefore subjected 70 urachal adenocarcinomas to targeted next-generation sequencing, conducted in situ and immunohistochemical analyses (including PD-L1 and DNA mismatch repair proteins [MMR]) and evaluated the microsatellite instability (MSI) status. The analytical findings were correlated with clinicopathological and outcome data and Kaplan-Meier and univariable/multivariable Cox regression analyses were performed. The patients had a mean age of 50 years, 66% were male and a 5-year overall survival (OS) of 58% and recurrence-free survival (RFS) of 45% was detected. Sequence variations were observed in TP53 (66%), KRAS (21%), BRAF (4%), PIK3CA (4%), FGFR1 (1%), MET (1%), NRAS (1%), and PDGFRA (1%). Gene amplifications were found in EGFR (5%), ERBB2 (2%), and MET (2%). We detected no evidence of MMR-deficiency (MMR-d)/MSI-high (MSI-h), whereas 10 of 63 cases (16%) expressed PD-L1. Therefore, anti-PD-1/PD-L1 immunotherapy approaches might be tested in UrC. Importantly, we found aberrations in intracellular signal transduction pathways (RAS/RAF/PI3K) in 31% of UrCs with potential implications for anti-EGFR therapy. Less frequent potentially actionable genetic alterations were additionally detected in ERBB2 (HER2), MET, FGFR1, and PDGFRA. The molecular profile strengthens the notion that UrC is a distinct entity on the genomic level with closer resemblance to colorectal than to bladder cancer.

TANDEM: a two-stage approach to maximize interpretability of drug response models based on multiple molecular data types.

MOTIVATION: Clinical response to anti-cancer drugs varies between patients. A large portion of this variation can be explained by differences in molecular features, such as mutation status, copy number alterations, methylation and gene expression profiles. We show that the classic approach for combining these molecular features (Elastic Net regression on all molecular features simultaneously) results in models that are almost exclusively based on gene expression. The gene expression features selected by the classic approach are difficult to interpret as they often represent poorly studied combinations of genes, activated by aberrations in upstream signaling pathways. RESULTS: To utilize all data types in a more balanced way, we developed TANDEM, a two-stage approach in which the first stage explains response using upstream features (mutations, copy number, methylation and cancer type) and the second stage explains the remainder using downstream features (gene expression). Applying TANDEM to 934 cell lines profiled across 265 drugs (GDSC1000), we show that the resulting models are more interpretable, while retaining the same predictive performance as the classic approach. Using the more balanced contributions per data type as determined with TANDEM, we find that response to MAPK pathway inhibitors is largely predicted by mutation data, while predicting response to DNA damaging agents requires gene expression data, in particular SLFN11 expression. AVAILABILITY AND IMPLEMENTATION: TANDEM is available as an R package on CRAN (for more information, see http://ccb.nki.nl/software/tandem). CONTACT: m.michaut@nki.nl or l.wessels@nki.nl SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Spatial relationships in the urothelial and head and neck tumor microenvironment predict response to combination immune checkpoint inhibitors.

Immune checkpoint inhibitors (ICI) can achieve remarkable responses in urothelial cancer (UC), which may depend on tumor microenvironment (TME) characteristics. However, the relationship between the TME, usually characterized by immune cell density, and response to ICI is unclear. Here, we quantify the TME immune cell densities and spatial relationships (SRs) of 24 baseline UC samples, obtained before pre-operative combination ICI treatment, using multiplex immunofluorescence. We describe SRs by approximating the first nearest-neighbor distance distribution with a Weibull distribution and evaluate the association between TME metrics and ipilimumab+nivolumab response. Immune cell density does not discriminate between response groups. However, the Weibull SR metrics of CD8+ T cells or macrophages to their closest cancer cell positively associate with response. CD8+ T cells close to B cells are characteristic of non-response. We validate our SR response associations in a combination ICI cohort of head and neck tumors. Our data confirm that SRs, in contrast to density metrics, are strong biomarkers of response to pre-operative combination ICIs.

Assessment of Predictive Genomic Biomarkers for Response to Cisplatin-based Neoadjuvant Chemotherapy in Bladder Cancer.

Cisplatin-based neoadjuvant chemotherapy (NAC) followed by radical cystectomy is recommended for patients with muscle-invasive bladder cancer (MIBC). It has been shown that somatic deleterious mutations in ERCC2, gain-of-function mutations in ERBB2, and alterations in ATM, RB1, and FANCC are correlated with pathological response to NAC in MIBC. The objective of this study was to validate these genomic biomarkers in pretreatment transurethral resection material from an independent retrospective cohort of 165 patients with MIBC who subsequently underwent NAC and radical surgery. Patients with ypT0/Tis/Ta/T1N0 disease after surgery were defined as responders. Somatic deleterious mutations in ERCC2 were found in nine of 68 (13%) evaluable responders and two of 95 (2%) evaluable nonresponders (p = 0.009; FDR = 0.03). No correlation was observed between response and alterations in ERBB2 or in ATM, RB1, or FANCC alone or in combination. In an exploratory analysis, no additional genomic alterations discriminated between responders and nonresponders to NAC. No further associations were identified between the aforementioned biomarkers and pathological complete response (ypT0N0) after surgery. In conclusion, we observed a positive association between deleterious mutations in ERCC2 and pathological response to NAC, but not overall survival or recurrence-free survival. Other previously reported genomic biomarkers were not validated. PATIENT SUMMARY: It is currently unknown which patients will respond to chemotherapy before definitive surgery for bladder cancer. Previous studies described several gene mutations in bladder cancer that correlated with chemotherapy response. This study confirmed that patients with bladder cancer with a mutation in the ERCC2 gene often respond to chemotherapy.

The Tumor Immune Landscape and Architecture of Tertiary Lymphoid Structures in Urothelial Cancer.

Candidate immune biomarkers have been proposed for predicting response to immunotherapy in urothelial cancer (UC). Yet, these biomarkers are imperfect and lack predictive power. A comprehensive overview of the tumor immune contexture, including Tertiary Lymphoid structures (TLS), is needed to better understand the immunotherapy response in UC. We analyzed tumor sections by quantitative multiplex immunofluorescence to characterize immune cell subsets in various tumor compartments in tumors without pretreatment and tumors exposed to preoperative anti-PD1/CTLA-4 checkpoint inhibitors (NABUCCO trial). Pronounced immune cell presence was found in UC invasive margins compared to tumor and stroma regions. CD8+PD1+ T-cells were present in UC, particularly following immunotherapy. The cellular composition of TLS was assessed by multiplex immunofluorescence (CD3, CD8, FoxP3, CD68, CD20, PanCK, DAPI) to explore specific TLS clusters based on varying immune subset densities. Using a k-means clustering algorithm, we found five distinct cellular composition clusters. Tumors unresponsive to anti-PD-1/CTLA-4 immunotherapy showed enrichment of a FoxP3+ T-cell-low TLS cluster after treatment. Additionally, cluster 5 (macrophage low) TLS were significantly higher after pre-operative immunotherapy, compared to untreated tumors. We also compared the immune cell composition and maturation stages between superficial (submucosal) and deeper TLS, revealing that superficial TLS had more pronounced T-helper cells and enrichment of early TLS than TLS located in deeper tissue. Furthermore, superficial TLS displayed a lower fraction of secondary follicle like TLS than deeper TLS. Taken together, our results provide a detailed quantitative overview of the tumor immune landscape in UC, which can provide a basis for further studies.

Ovarian Cancer-Specific BRCA-like Copy-Number Aberration Classifiers Detect Mutations Associated with Homologous Recombination Deficiency in the AGO-TR1 Trial.

PURPOSE: Previously, we developed breast cancer BRCA1-like and BRCA2-like copy-number profile shrunken centroid classifiers predictive for mutation status and response to therapy, targeting homologous recombination deficiency (HRD). Therefore, we investigated BRCA1- and BRCA2-like classification in ovarian cancer, aiming to acquire classifiers with similar properties as those in breast cancer.Experimental Design: We analyzed DNA copy-number profiles of germline BRCA1- and BRCA2-mutant ovarian cancers and control tumors and observed that existing breast cancer classifiers did not sufficiently predict mutation status. Hence, we trained new shrunken centroid classifiers on this set and validated them in the independent The Cancer Genome Atlas dataset. Subsequently, we assessed BRCA1/2-like classification and obtained germline and tumor mutation and methylation status of cancer predisposition genes, among them several involved in HR repair, of 300 ovarian cancer samples derived from the consecutive cohort trial AGO-TR1 (NCT02222883). RESULTS: The detection rate of the BRCA1-like classifier for BRCA1 mutations and promoter hypermethylation was 95.6%. The BRCA2-like classifier performed less accurately, likely due to a smaller training set. Furthermore, three quarters of the BRCA1/2-like tumors could be explained by (epi)genetic alterations in BRCA1/2, germline RAD51C mutations and alterations in other genes involved in HR. Around half of the non-BRCA-mutated ovarian cancer cases displayed a BRCA-like phenotype. CONCLUSIONS: The newly trained classifiers detected most BRCA-mutated and methylated cancers and all tumors harboring a RAD51C germline mutations. Beyond that, we found an additional substantial proportion of ovarian cancers to be BRCA-like.

Endocrine pulse identification using penalized methods and a minimum set of assumptions.

The detection of hormone secretion episodes is important for understanding normal and abnormal endocrine functioning, but pulse identification from hormones measured with short interval sampling is challenging. Furthermore, to obtain useable results, the model underlying hormone secretion and clearance must be augmented with restrictions based on biologically acceptable assumptions. Here, using the assumption that there are only a few time points at which a hormone is secreted, we used a modern penalized nonlinear least-squares setup to select the number of secretion events. We did not assume a particular shape or frequency distribution for the secretion pulses. Our pulse identfication method, VisPulse, worked well with luteinizing hormone (LH), cortisol, growth hormone, or testosterone. In particular, applying our modeling strategy to previous LH data revealed a good correlation between the modeled and measured LH hormone concentrations, the estimated secretion pattern was sparse, and the small and structureless residuals indicated a proper model with a good fit. We benchmarked our method to AutoDecon, a commonly used hormone secretion model, and performed releasing hormone infusion experiments. The results of these experiments confirmed that our method is accurate and outperforms AutoDecon, especially for detecting silent periods and small secretion events, suggesting a high-secretion event resolution. Method validation using (releasing hormone) infusion data revealed sensitivities and selectivities of 0.88 and 0.95 and of 0.69 and 0.91 for VisPulse and AutoDecon, respectively.

Preoperative ipilimumab plus nivolumab in locoregionally advanced urothelial cancer: the NABUCCO trial.

Preoperative immunotherapy with anti-PD1 plus anti-CTLA4 antibodies has shown remarkable pathological responses in melanoma1 and colorectal cancer2. In NABUCCO (ClinicalTrials.gov: NCT03387761 ), a single-arm feasibility trial, 24 patients with stage III urothelial cancer (UC) received two doses of ipilimumab and two doses of nivolumab, followed by resection. The primary endpoint was feasibility to resect within 12 weeks from treatment start. All patients were evaluable for the study endpoints and underwent resection, 23 (96%) within 12 weeks. Grade 3-4 immune-related adverse events occurred in 55% of patients and in 41% of patients when excluding clinically insignificant laboratory abnormalities. Eleven patients (46%) had a pathological complete response (pCR), meeting the secondary efficacy endpoint. Fourteen patients (58%) had no remaining invasive disease (pCR or pTisN0/pTaN0). In contrast to studies with anti-PD1/PD-L1 monotherapy, complete response to ipilimumab plus nivolumab was independent of baseline CD8+ presence or T-effector signatures. Induction of tertiary lymphoid structures upon treatment was observed in responding patients. Our data indicate that combined CTLA-4 plus PD-1 blockade might provide an effective preoperative treatment strategy in locoregionally advanced UC, irrespective of pre-existing CD8+ T cell activity.

Genomic data integration by WON-PARAFAC identifies interpretable factors for predicting drug-sensitivity in vivo.

Integrative analyses that summarize and link molecular data to treatment sensitivity are crucial to capture the biological complexity which is essential to further precision medicine. We introduce Weighted Orthogonal Nonnegative parallel factor analysis (WON-PARAFAC), a data integration method that identifies sparse and interpretable factors. WON-PARAFAC summarizes the GDSC1000 cell line compendium in 130 factors. We interpret the factors based on their association with recurrent molecular alterations, pathway enrichment, cancer type, and drug-response. Crucially, the cell line derived factors capture the majority of the relevant biological variation in Patient-Derived Xenograft (PDX) models, strongly suggesting our factors capture invariant and generalizable aspects of cancer biology. Furthermore, drug response in cell lines is better and more consistently translated to PDXs using factor-based predictors as compared to raw feature-based predictors. WON-PARAFAC efficiently summarizes and integrates multiway high-dimensional genomic data and enhances translatability of drug response prediction from cell lines to patient-derived xenografts.

Patient-derived organoids can predict response to chemotherapy in metastatic colorectal cancer patients.

There is a clear and unmet clinical need for biomarkers to predict responsiveness to chemotherapy for cancer. We developed an in vitro test based on patient-derived tumor organoids (PDOs) from metastatic lesions to identify nonresponders to standard-of-care chemotherapy in colorectal cancer (CRC). In a prospective clinical study, we show the feasibility of generating and testing PDOs for evaluation of sensitivity to chemotherapy. Our PDO test predicted response of the biopsied lesion in more than 80% of patients treated with irinotecan-based therapies without misclassifying patients who would have benefited from treatment. This correlation was specific to irinotecan-based chemotherapy, however, and the PDOs failed to predict outcome for treatment with 5-fluorouracil plus oxaliplatin. Our data suggest that PDOs could be used to prevent cancer patients from undergoing ineffective irinotecan-based chemotherapy.

MAP3K1 and MAP2K4 mutations are associated with sensitivity to MEK inhibitors in multiple cancer models.

Activation of the mitogen-activated protein kinase (MAPK) pathway is frequent in cancer. Drug development efforts have been focused on kinases in this pathway, most notably on RAF and MEK. We show here that MEK inhibition activates JNK-JUN signaling through suppression of DUSP4, leading to activation of HER Receptor Tyrosine Kinases. This stimulates the MAPK pathway in the presence of drug, thereby blunting the effect of MEK inhibition. Cancers that have lost MAP3K1 or MAP2K4 fail to activate JNK-JUN. Consequently, loss-of-function mutations in either MAP3K1 or MAP2K4 confer sensitivity to MEK inhibition by disabling JNK-JUN-mediated feedback loop upon MEK inhibition. In a panel of 168 Patient Derived Xenograft (PDX) tumors, MAP3K1 and MAP2K4 mutation status is a strong predictor of response to MEK inhibition. Our findings suggest that cancers having mutations in MAP3K1 or MAP2K4, which are frequent in tumors of breast, prostate and colon, may respond to MEK inhibitors. Our findings also suggest that MAP3K1 and MAP2K4 are potential drug targets in combination with MEK inhibitors, in spite of the fact that they are encoded by tumor suppressor genes.

Evaluation of the HER/PI3K/AKT Family Signaling Network as a Predictive Biomarker of Pathologic Complete Response for Patients With Breast Cancer Treated With Neratinib in the I-SPY 2 TRIAL.

PURPOSE: In the I-SPY 2 TRIAL (Investigation of Serial Studies to Predict Your Therapeutic Response With Imaging and Molecular Analysis 2), the pan-erythroblastic oncogene B inhibitor neratinib was available to all hormone receptor (HR)/human epidermal growth factor receptor 2 (HER2) subtypes and graduated in the HR-negative/HER2-positive signature. We hypothesized that neratinib response may be predicted by baseline HER2 epidermal growth factor receptor (EGFR) signaling activation/phosphorylation levels independent of total levels of HER2 or EGFR proteins. MATERIALS AND METHODS: Complete experimental and response data were available for between 130 and 193 patients. In qualifying analyses, which used logistic regression and treatment interaction analysis, 18 protein/phosphoprotein, 10 mRNA, and 12 DNA biomarkers that related to HER family signaling were evaluated. Exploratory analyses used Wilcoxon rank sum and t tests without multiple comparison correction. RESULTS: HER pathway DNA biomarkers were either low prevalence or nonpredictive. In expression biomarker analysis, only one gene (STMN1) was specifically associated with response to neratinib in the HER2-negative subset. In qualifying protein/phosphoprotein analyses that used reverse phase protein microarrays, six HER family markers were associated with neratinib response. After analysis was adjusted for HR/HER2 status, EGFR Y1173 (pEGFR) showed a significant biomarker-by-treatment interaction (P = .049). Exploratory analysis of HER family signaling in patients with triple-negative (TN) disease found that activation of EGFR Y1173 (P = .005) and HER2 Y1248 (pHER2) (P = .019) were positively associated with pathologic complete response. Exploratory analysis in this pEGFR/pHER2-activated TN subgroup identified elevated levels of estrogen receptor α (P < .006) in these patients. CONCLUSION: Activation of HER family phosphoproteins associates with response to neratinib, but only EGFR Y1173 and STMN1 appear to add value to the graduating signature. Activation of HER2 and EGFR in TN tumors may identify patients whose diseases respond to neratinib and implies that there is a subset of patients with TN disease who paradoxically exhibit HER family signaling activation and may achieve clinical benefit with neratinib; this concept must be validated in future studies.

Selective Loss of PARG Restores PARylation and Counteracts PARP Inhibitor-Mediated Synthetic Lethality.

Inhibitors of poly(ADP-ribose) (PAR) polymerase (PARPi) have recently entered the clinic for the treatment of homologous recombination (HR)-deficient cancers. Despite the success of this approach, drug resistance is a clinical hurdle, and we poorly understand how cancer cells escape the deadly effects of PARPi without restoring the HR pathway. By combining genetic screens with multi-omics analysis of matched PARPi-sensitive and -resistant Brca2-mutated mouse mammary tumors, we identified loss of PAR glycohydrolase (PARG) as a major resistance mechanism. We also found the presence of PARG-negative clones in a subset of human serous ovarian and triple-negative breast cancers. PARG depletion restores PAR formation and partially rescues PARP1 signaling. Importantly, PARG inactivation exposes vulnerabilities that can be exploited therapeutically.

Statistical validation of megavariate effects in ASCA.

BACKGROUND: Innovative extensions of (M) ANOVA gain common ground for the analysis of designed metabolomics experiments. ASCA is such a multivariate analysis method; it has successfully estimated effects in megavariate metabolomics data from biological experiments. However, rigorous statistical validation of megavariate effects is still problematic because megavariate extensions of the classical F-test do not exist. METHODS: A permutation approach is used to validate megavariate effects observed with ASCA. By permuting the class labels of the underlying experimental design, a distribution of no-effect is calculated. If the observed effect is clearly different from this distribution the effect is deemed significant RESULTS: The permutation approach is studied using simulated data which gave successful results. It was then used on real-life metabolomics data set dealing with bromobenzene-dosed rats. In this metabolomics experiment the dosage and time-interaction effect were validated, both effects are significant. Histological screening of the treated rats' liver agrees with this finding. CONCLUSION: The suggested procedure gives approximate p-values for testing effects underlying metabolomics data sets. Therefore, performing model validation is possible using the proposed procedure.

Identification of a Druggable Pathway Controlling Glioblastoma Invasiveness.

Diffuse and uncontrollable brain invasion is a hallmark of glioblastoma (GBM), but its mechanism is understood poorly. We developed a 3D ex vivo organotypic model to study GBM invasion. We demonstrate that invading GBM cells upregulate a network of extracellular matrix (ECM) components, including multiple collagens, whose expression correlates strongly with grade and clinical outcome. We identify interferon regulatory factor 3 (IRF3) as a transcriptional repressor of ECM factors and show that IRF3 acts as a suppressor of GBM invasion. Therapeutic activation of IRF3 by inhibiting casein kinase 2 (CK2)-a negative regulator of IRF3-downregulated the expression of ECM factors and suppressed GBM invasion in ex vivo and in vivo models across a panel of patient-derived GBM cell lines representative of the main molecular GBM subtypes. Our data provide mechanistic insight into the invasive capacity of GBM tumors and identify a potential therapy to inhibit GBM invasion.