Computationally simple analysis of matched, outcome-based studies of ordinal disease states.

Outcome-based sampling is an efficient study design for rare conditions, such as glioblastoma. It is often used in conjunction with matching, for increased efficiency and to potentially avoid bias due to confounding. A study was conducted at the Massachusetts General Hospital that involved retrospective sampling of glioblastoma patients with respect to multiple-ordered disease states, as defined by three categories of overall survival time. To analyze such studies, we posit an adjacent categories logit model and exploit its allowance for prospective analysis of a retrospectively sampled study and its advantageous removal of set and level specific nuisance parameters through conditioning on sufficient statistics. This framework allows for any sampling design and is not limited to one level of disease within each set, such as in previous publications. We describe how this ordinal conditional model can be fit using standard conditional logistic regression procedures. We consider an alternative pseudo-likelihood approach that potentially offers robustness under partial model misspecification at the expense of slight loss of efficiency under correct model specification for small sample sizes. We apply our methods to the Massachusetts General Hospital glioblastoma study.

Supervised Bayesian latent class models for high-dimensional data.

High-grade gliomas are the most common primary brain tumors in adults and are typically diagnosed using histopathology. However, these diagnostic categories are highly heterogeneous and do not always correlate well with survival. In an attempt to refine these diagnoses, we make several immunohistochemical measurements of YKL-40, a gene previously shown to be differentially expressed between diagnostic groups. We propose two latent class models for classification and variable selection in the presence of high-dimensional binary data, fit by using Bayesian Markov chain Monte Carlo techniques. Penalization and model selection are incorporated in this setting via prior distributions on the unknown parameters. The methods provide valid parameter estimates under conditions in which standard supervised latent class models do not, and outperform two-stage approaches to variable selection and parameter estimation in a variety of settings. We study the properties of these methods in simulations, and apply these methodologies to the glioma study for which identifiable three-class parameter estimates cannot be obtained without penalization. With penalization, the resulting latent classes correlate well with clinical tumor grade and offer additional information on survival prognosis that is not captured by clinical diagnosis alone. The inclusion of YKL-40 features also increases the precision of survival estimates. Fitting models with and without YKL-40 highlights a subgroup of patients who have glioblastoma (GBM) diagnosis but appear to have better prognosis than the typical GBM patient.

Efficacy of systemically administered oncolytic vaccinia virotherapy for malignant gliomas is enhanced by combination therapy with rapamycin or cyclophosphamide.

PURPOSE: The oncolytic effects of a systemically delivered, replicating, double-deleted vaccinia virus has been previously shown for the treatment of many cancers, including colon, ovarian, and others. The purpose of this study was to investigate the oncolytic potential of double-deleted vaccinia virus alone or in combination with rapamycin or cyclophosphamide to treat malignant gliomas in vitro and in vivo. EXPERIMENTAL DESIGN: Rat (RG2, F98, C6) and human (A172, U87MG, U118) glioma cell lines were cultured in vitro and treated with live or UV-inactivated vaccinia virus. Viral gene [enhanced green fluorescent protein (EGFP)] expression by fluorescence-activated cell sorting, relative cell viability by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), and assays for cytopathic effects were examined. S.c. murine tumor xenografts (U87MG, U118, C6) and i.c. (RG2, F98) tumor models in immunocompetent rats were treated with systemic administration of EGFP-expressing vaccinia virus (vvDD-EGFP), alone or in combination with rapamycin or cyclophosphamide, or controls. Tumor size, viral biodistribution, and animal survival were assessed. Lastly, the oncolytic effects of vvDD-EGFP on human malignant glioma explants were evaluated. RESULTS: vvDD-EGFP was able to infect and kill glioma cells in vitro. A single systemic dose of vvDD-EGFP significantly inhibited the growth of xenografts in athymic mice. Systemic delivery of vvDD-EGFP alone was able to target solitary and multifocal i.c. tumors and prolong survival of immunocompetent rats, whereas combination therapy with rapamycin or cyclophosphamide enhanced viral replication and further prolonged survival. Finally, vvDD-EGFP was able to infect and kill ex vivo primary human malignant gliomas. CONCLUSIONS: These results suggest that vvDD-EGFP is a promising novel agent for human malignant glioma therapy, and in combination with immunosuppressive agents, may lead to prolonged survival from this disease.

Deregulation of a STAT3-interleukin 8 signaling pathway promotes human glioblastoma cell proliferation and invasiveness.

Inactivation of the tumor suppressor phosphatase and tensin homolog (mutated in multiple advanced cancers 1) (PTEN) is recognized as a major event in the pathogenesis of the brain tumor glioblastoma. However, the mechanisms by which PTEN loss specifically impacts the malignant behavior of glioblastoma cells, including their proliferation and propensity for invasiveness, remain poorly understood. Genetic studies suggest that the transcription factor signal transducers and activators of transcription 3 (STAT3) harbors a PTEN-regulated tumor suppressive function in mouse astrocytes. Here, we report that STAT3 plays a critical tumor suppressive role in PTEN-deficient human glioblastoma cells. Endogenous STAT3 signaling is specifically inhibited in PTEN-deficient glioblastoma cells. Strikingly, reactivation of STAT3 in PTEN-deficient glioblastoma cells inhibits their proliferation, invasiveness, and ability to spread on myelin. We also identify the chemokine interleukin 8 (IL8) as a novel target gene of STAT3 in human glioblastoma cells. Activated STAT3 occupies the endogenous IL8 promoter and directly represses IL8 transcription. Consistent with these results, IL8 is upregulated in PTEN-deficient human glioblastoma tumors. Importantly, IL8 repression mediates STAT3 inhibition of glioblastoma cell proliferation, invasiveness, and spreading on myelin. Collectively, our findings uncover a novel link between STAT3 and IL8, the deregulation of which plays a key role in the malignant behavior of PTEN-deficient glioblastoma cells. These studies suggest that STAT3 activation or IL8 inhibition may have potential in patient-tailored treatment of PTEN-deficient brain tumors.

Autocrine platelet-derived growth factor-dependent gene expression in glioblastoma cells is mediated largely by activation of the transcription factor sterol regulatory element binding protein and is associated with altered genotype and patient survival in human brain tumors.

A complex profile of gene expression elicited by autocrine platelet-derived growth factor (PDGF) signaling was identified in U87 MG glioblastoma cells by microarray analysis. The most striking pattern observed was a PDGF-dependent activation of at least 25 genes involved with biosynthesis and/or uptake of cholesterol and isoprenoids, including mevalonate pyrophosphate decarboxylase, 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) synthase, HMG-CoA reductase, and low-density lipoprotein receptor. Activity of the HMG-CoA synthase promoter was induced by autocrine PDGF activity as indicated by significant reductions following forced expression of dominant-negative PDGF-A (88%) or treatment with the PDGF receptor antagonist CT52923 (50%). Induction of the HMG-CoA synthase promoter required a binding site for sterol regulatory element binding proteins (SRE-BP), consistent with a key role for these transcription factors in the induction of this gene network. Neither proteolytic activation nor nuclear localization of SRE-BP was affected by disruption of the PDGF autocrine loop, indicating that PDGF signaling is required for other signaling events involved in activation of SRE-BP target genes. Analysis of an expression databank derived from human glial tumors (n = 77) identified a subgroup exhibiting a profile consistent with PDGF dependence, including increased expression of SRE-BP target genes. This subgroup displayed an absence of epidermal growth factor receptor gene amplification, decreased incidence of allelic loss of 10q, increased frequency of TP53 mutations and allelic losses of 1p and 19q, and longer patient survival. This study identifies genes associated with oncogenic activity of PDGF and provides important insights into biomarkers and therapeutic targets in malignant gliomas.

Activation of STAT3, MAPK, and AKT in malignant astrocytic gliomas: correlation with EGFR status, tumor grade, and survival.

Diffuse astrocytic gliomas are the most common human glial tumors with glioblastoma being the most malignant form. Epidermal growth factor receptor (EGFR) gene amplification is one of the most common genetic changes in glioblastoma and can lead to the activation of various downstream signaling molecules, including STAT3, MAPK, and AKT. In this study, we investigated the activation status of these 3 signaling molecules as well as wild-type (EGFRwt) and mutant (EGFRvIII) EGFR in 82 malignant astrocytic gliomas (55 glioblastomas and 27 anaplastic astrocytomas) using immunohistochemistry. The presence of EGFRwt, but not EGFRvIII, immunopositivity correlated significantly with prevalent EGFR gene amplification in glioblastomas. STAT3 and AKT activation correlated significantly with EGFR status, although the correlation for p-STAT3 was attributed exclusively to EGFRvIII. The distribution of these 3 activated molecules varied significantly with tumor grade; although activation of STAT3 was essentially identical between anaplastic astrocytomas and glioblastomas, an increase in the activation of MAPK and AKT appeared to correlate with the progression of anaplastic astrocytoma to glioblastoma. Finally, activated STAT3 and AKT were marginally predictive of improved and worse prognosis, respectively. Taken together, these findings begin to elucidate the interrelationship between these signaling pathways in astrocytic gliomas in vivo.

Expression of oligodendroglial and astrocytic lineage markers in diffuse gliomas: use of YKL-40, ApoE, ASCL1, and NKX2-2.

The phenotypic heterogeneity of astrocytic and oligodendroglial tumor cells complicates establishing accurate diagnostic criteria, and lineage-specific markers would facilitate diagnosis of glioma subtypes. Based on data from the literature and from expression microarrays, we selected molecules relevant to gliogenesis and glial lineage specificity and then used immunohistochemistry to assess expression of these molecules in 55 diffuse gliomas, including 8 biphasic oligoastrocytomas, 21 oligodendrogliomas (all with 1p/19qloss), 21 astrocytomas, and 5 glioblastomas. For the astrocytic lineage markers (GFAP, YKL-40, and ApoE), GFAP expression was significantly higher in the astrocytic component of oligoastrocytomas compared with the oligodendroglial part; similar patterns were detected for YKL-40 and ApoE, although the differences were not significant. GFAP, YKL-40, and ApoE reliably distinguished grade II-III oligodendrogliomas from grade II-IV astrocytomas (p < 0.0001, p = 0.002, and p < 0.0001, respectively). Among the oligodendroglial lineage markers (Olig2, Sox10, ASCL1, and NKX2-2), ASCL1 and NKX2-2 displayed significantly different immunostaining between oligodendrogliomas and astrocytomas (p = 0.017 and 0.004, respectively), but none clearly differentiated between the 2 glial populations of oligoastrocytomas. In addition to GFAP, therefore, YKL-40, ApoE, ASCL1, and NKX2-2 represent promising tumor cell markers to distinguish oligodendrogliomas from astrocytomas.

YKL-40 is a differential diagnostic marker for histologic subtypes of high-grade gliomas.

PURPOSE AND EXPERIMENTAL DESIGN: In modern neuro-oncology, no variable affects therapeutic decisions and prognostic estimation more than tumor classification. We showed recently that class prediction models, based on gene expression profiles, classify diagnostically challenging malignant gliomas in a manner that better correlates with clinical outcome than standard pathology. In the present study, we used immunohistochemistry to investigate YKL-40 protein expression in independent sets of glioblastomas and anaplastic oligodendrogliomas to determine whether this single marker can aid classification of these high-grade gliomas. RESULTS AND CONCLUSIONS: Glioblastomas show strikingly more YKL-40 expression than anaplastic oligodendrogliomas. Only 2 of 37 glioblastomas showed completely negative YKL-40 staining in both tumor cells and extracellular matrix, whereas 18 of 29 anaplastic oligodendrogliomas were completely negative in non-microgemistocytic tumor cells and extracellular matrix. Tumor cell staining intensity was also markedly different: 84% of glioblastomas showed strong staining intensities of 2+ or 3+ whereas 76% of anaplastic oligodendrogliomas either did not stain or stained at only 1+. YKL-40 staining provided a better class distinction of glioblastoma versus anaplastic oligodendroglioma than glial fibrillary acidic protein, the current standard immunohistochemical marker used to distinguish diagnostically challenging gliomas. Moreover, a combination of YKL-40 and glial fibrillary acidic protein immunohistochemistry afforded even greater diagnostic accuracy in anaplastic oligodendrogliomas.

Selection pressures of TP53 mutation and microenvironmental location influence epidermal growth factor receptor gene amplification in human glioblastomas.

Epidermal growth factor receptor (EGFR) gene amplification occurs in glioblastomas as so-called double minutes. Because double minutes are extrachromosomal fragments, selection pressures must operate to maintain high EGFR copy number over multiple cell divisions. In analyses of glioblastoma lysates, EGFR amplification has been observed almost exclusively in glioblastomas harboring wild-type TP53 genes, which raises the alternative hypotheses that TP53 mutation either prevents amplification or selects against maintenance of EGFR-amplified cells. To address these possibilities at the cellular level, we studied 14 glioblastomas for TP53 mutation and EGFR gene amplification status, using fluorescence in situ hybridization (FISH) for the latter. Remarkably, four of the six cases with TP53 mutation had isolated EGFR-amplified cells in different regions, demonstrating that EGFR amplification occurs frequently at the cellular level in TP53-mutant glioblastomas. Thus, TP53 mutation does not prevent EGFR amplification but does not facilitate selection of EGFR-amplified cells. Of the eight cases without TP53 mutation, five had widespread EGFR amplification. In four of these five cases, multiple regions of the tumor were available for examination; FISH demonstrated a gradation of EGFR amplification, with highly amplified cells, primarily at the invading edges rather than the relatively solid tumor centers, suggesting that EGFR overexpression, when selected for in vivo, may be related to tumor invasion.

Gene expression-based classification of malignant gliomas correlates better with survival than histological classification.

In modern clinical neuro-oncology, histopathological diagnosis affects therapeutic decisions and prognostic estimation more than any other variable. Among high-grade gliomas, histologically classic glioblastomas and anaplastic oligodendrogliomas follow markedly different clinical courses. Unfortunately, many malignant gliomas are diagnostically challenging; these nonclassic lesions are difficult to classify by histological features, generating considerable interobserver variability and limited diagnostic reproducibility. The resulting tentative pathological diagnoses create significant clinical confusion. We investigated whether gene expression profiling, coupled with class prediction methodology, could be used to classify high-grade gliomas in a manner more objective, explicit, and consistent than standard pathology. Microarray analysis was used to determine the expression of approximately 12000 genes in a set of 50 gliomas, 28 glioblastomas and 22 anaplastic oligodendrogliomas. Supervised learning approaches were used to build a two-class prediction model based on a subset of 14 glioblastomas and 7 anaplastic oligodendrogliomas with classic histology. A 20-feature k-nearest neighbor model correctly classified 18 of the 21 classic cases in leave-one-out cross-validation when compared with pathological diagnoses. This model was then used to predict the classification of clinically common, histologically nonclassic samples. When tumors were classified according to pathology, the survival of patients with nonclassic glioblastoma and nonclassic anaplastic oligodendroglioma was not significantly different (P = 0.19). However, class distinctions according to the model were significantly associated with survival outcome (P = 0.05). This class prediction model was capable of classifying high-grade, nonclassic glial tumors objectively and reproducibly. Moreover, the model provided a more accurate predictor of prognosis in these nonclassic lesions than did pathological classification. These data suggest that class prediction models, based on defined molecular profiles, classify diagnostically challenging malignant gliomas in a manner that better correlates with clinical outcome than does standard pathology.

Histology-based expression profiling yields novel prognostic markers in human glioblastoma.

Although the prognosis for patients with glioblastoma is poor, survival is variable, with some patients surviving longer than others. For this reason, there has been longstanding interest in the identification of prognostic markers for glioblastoma. We hypothesized that specific histologic features known to correlate with malignancy most likely express molecules that are directly related to the aggressive behavior of these tumors. We further hypothesized that such molecules could be used as biomarkers to predict behavior in a manner that might add prognostic power to sole histologic observation of the feature. We reasoned that perinecrotic tumor cell palisading, which denotes the most aggressive forms of malignant gliomas, would be a striking histologic feature on which to test this hypothesis. We therefore used laser capture microdissection and oligonucleotide arrays to detect molecules differentially expressed in perinecrotic palisades. A set of RNAs (including POFUT2, PTDSR, PLOD2, ATF5, and HK2) that were differentially expressed in 3 initially studied, microdissected glioblastomas also provided prognostic information in an independent set of 28 glioblastomas that did not all have perinecrotic palisades. On validation in a second, larger independent series, this approach could be applied to other human glioma types to derive tissue biomarkers that could offer ancillary prognostic and predictive information alongside standard histopathologic examination.

Statistical considerations for immunohistochemistry panel development after gene expression profiling of human cancers.

In recent years there have been a number of microarray expression studies in which different types of tumors were classified by identifying a panel of differentially expressed genes. Immunohistochemistry is a practical and robust method for extending gene expression data to common pathological specimens with the advantage of being applicable to paraffin-embedded tissues. However, the number of assays required for successful immunohistochemical classification remains unclear. We propose a simulation-based method for assessing sample size for an immunohistochemistry investigation after a promising gene expression study of human tumors. The goals of such an immunohistochemistry study would be to develop and validate a marker panel that yields improved prognostic classification of cancer patients. We demonstrate how the preliminary gene expression data, coupled with certain realistic assumptions, can be used to estimate the number of immunohistochemical assays required for development. These assumptions are more tenable than alternative assumptions that would be required for crude analytic sample size calculations and that may yield underpowered and inefficient studies. We applied our methods to the design of an immunohistochemistry study for glioma classification and estimated the number of assays required to ensure satisfactory technical and prognostic validation. Simulation approaches for computing power and sample size that are based on existing gene expression data provide a powerful tool for efficient design of follow-up genomic studies.

Molecular genetics of oligodendrogliomas: a model for improved clinical management in the field of neurooncology.

Over the last several years, oligodendroglial tumors have become a model for the positive role of molecular genetics in improved treatment of patients with brain tumors. Oligodendrogliomas, in contrast to astrocytic gliomas, frequently respond to chemotherapy and have a better overall prognosis. Combined loss of chromosomes 1p and 19q has proven to be a powerful predictor of chemotherapeutic response and survival in oligodendrogliomas. In contrast, other genetic alterations, such as TP53 and PTEN mutations, EGFR amplification, and homozygous deletion of CDKN2A have been correlated with worse outcome in these tumors. Furthermore, 1p/19q loss has been shown to correlate with unequivocal oligodendroglial tumor histology, location and growth pattern of tumors within the brain, and magnetic resonance imaging characteristics. Although much is also known about the molecular pathological characteristics of astrocytic gliomas, the significance of this information to clinical management in patients with these tumors has not been as striking as has been the case for oligodendrogliomas; possible reasons for this are discussed. In this paper the author will summarize these advances, thus attempting to highlight the molecular genetic study of oligodendrogliomas as a model for improved clinical management in the field of neurooncology.

The oligodendroglial lineage marker OLIG2 is universally expressed in diffuse gliomas.

Astrocytomas, oligodendrogliomas, and oligoastrocytomas, collectively referred to as diffuse gliomas, are the most common primary brain tumors. These tumors are classified by histologic similarity to differentiated astrocytes and oligodendrocytes, but this approach has major limitations in guiding modern treatment and research. Lineage markers represent a potentially useful adjunct to morphologic classification. The murine bHLH transcription factors Olig1 and Olig2 are expressed in neural progenitors and oligodendroglia and are essential for oligodendrocyte development. High OLIG expression alone has been proposed to distinguish oligodendrogliomas from astrocytomas, so we critically evaluated OLIG2 as a marker by immunohistochemical and oligonucleotide microarray analysis. OLIG2 protein is faithfully restricted to normal oligodendroglia and their progenitors in human brain. Immunohistochemical analysis of 180 primary, metastatic, and non-neural human tumors shows OLIG2 is highly expressed in all diffuse gliomas. Immunohistochemistry and microarray analyses demonstrate higher OLIG2 in anaplastic oligodendrogliomas versus glioblastomas, which are heterogeneous with respect to OLIG2 levels. OLIG2 protein expression is present but inconsistent and generally lower in most other brain tumors and is absent in non-neuroectodermal tumors. Overall, OLIG2 is a useful marker of diffuse gliomas as a class. However, expression heterogeneity of OLIG2 in astrocytomas precludes immunohistochemical classification of individual gliomas by OLIG2 alone.

Genetic alterations of phosphoinositide 3-kinase subunit genes in human glioblastomas.

Genetic alterations of PI3K (phosphoinositide 3-kinase) subunits have been documented in a number of tumor types, with increased PI3K activity linked to gene amplification and mutation of catalytic subunits, as well as mutations of regulatory subunits. Among high grade gliomas, activation of the PI3K-AKT signaling pathway through loss of PTEN function is common. We therefore investigated whether genetic alteration of class IA PI3Ks might provide a mechanism for deregulation of this pathway in glioblastomas. We studied a series of glioblastomas with FISH to assess copy number of catalytic subunits (PIK3CA and PIK3CD) and with PCR-SSCP to screen for somatic mutations of conserved regions of both catalytic and regulatory subunits. FISH revealed frequent balanced copy number increases of both PIK3CA and PIK3CD, and one case showed an extra copy limited to PIK3CA. One glioblastoma exhibited a 9-bp deletion that encompassed the exon-intron junction of exon 12 of PIK3R1, documenting for the first time a mutation within a PI3K regulatory subunit in human glioblastoma. This deletion would be predicted to yield a truncated protein that lacks the inhibitory domain, resulting in increased PI3K activity. Furthermore, the case with selected PIK3CA copy number gain and the case with a truncating PIK3R1 mutation both featured AKT activation without PTEN mutation. These results suggest that genetic alterations of class IA PI3K subunit genes can occasionally play a role in human glioblastoma by activating the PI3K-AKT signaling pathway independently of PTEN mutation.

A sensitive and specific diagnostic panel to distinguish diffuse astrocytoma from astrocytosis: chromosome 7 gain with mutant isocitrate dehydrogenase 1 and p53.

One of the major challenges of surgical neuropathology is the distinction of diffuse astrocytoma (World Health Organization grade II) from astrocytosis. The most commonly used ancillary tool to solve this problem is p53 immunohistochemistry (IHC), but this is neither sensitive nor specific. Isocitrate dehydrogenase 1 (IDH1) mutations arecommon in lower-grade gliomas, with most causing a specific amino acid change (R132H) that can be detected with a monoclonal antibody. IDH2 mutations are rare, but they also occur in gliomas. In addition, gains of chromosome 7 are common in gliomas. In this study, we assessed the status of p53, IDH1/2, and chromosome 7 to determine the most useful panel to distinguish astrocytoma from astrocytosis. We studied biopsy specimens from 21 World Health Organization grade II diffuse astrocytomas and 20 reactive conditions. The single most sensitive test to identify astrocytoma is fluorescence in situ hybridization for chromosome 7 gain (76.2%). The combination of p53 and mutant IDH1 IHC provides a higher sensitivity (71.4%) than either test alone (47.8%); this combination offers a practical initial approach for the surgical pathologist. The best overall sensitivity (95%) is achieved when fluorescence in situ hybridization for chromosome 7 gain is added to the p53-mutant IDH1 IHC panel.

Increased expression of angiogenic genes in the brains of mouse meg3-null embryos.

Maternally expressed gene 3 (MEG3) is a noncoding RNA highly expressed in the normal human brain and pituitary. Expression of MEG3 is lost in gonadotroph-derived clinically nonfunctioning pituitary adenomas. Meg3 knockout mice were generated to identify targets and potential functions of this gene in embryonic development and tumorigenesis. Gene expression profiles were compared in the brains of Meg3-null embryos and wild-type littermate controls using microarray analysis. Microarray data were analyzed with GeneSifter, which uses Kyoto Encyclopedia of Genes and Genomes pathways and Gene Ontology classifications to identify signaling cascades and functional categories of interest within the dataset. Differences were found in signaling pathways and ontologies related to angiogenesis between wild-type and knockout embryos. Quantitative RT-PCR and immunohistological staining showed increased expression of some Vascular Endothelial Growth Factor pathway genes and increased cortical microvessel density in the Meg3-null embryos. In conclusion, Meg3 may play an important role in control of vascularization in the brain and may function as a tumor suppressor in part by inhibiting angiogenesis.

A multigene predictor of outcome in glioblastoma.

Only a subset of patients with newly diagnosed glioblastoma (GBM) exhibit a response to standard therapy. To date, a biomarker panel with predictive power to distinguish treatment sensitive from treatment refractory GBM tumors does not exist. An analysis was performed using GBM microarray data from 4 independent data sets. An examination of the genes consistently associated with patient outcome, revealed a consensus 38-gene survival set. Worse outcome was associated with increased expression of genes associated with mesenchymal differentiation and angiogenesis. Application to formalin fixed-paraffin embedded (FFPE) samples using real-time reverse-transcriptase polymerase chain reaction assays resulted in a 9-gene subset which appeared robust in these samples. This 9-gene set was then validated in an additional independent sample set. Multivariate analysis confirmed that the 9-gene set was an independent predictor of outcome after adjusting for clinical factors and methylation of the methyl-guanine methyltransferase promoter. The 9-gene profile was also positively associated with markers of glioma stem-like cells, including CD133 and nestin. In sum, a multigene predictor of outcome in glioblastoma was identified which appears applicable to routinely processed FFPE samples. The profile has potential clinical application both for optimization of therapy in GBM and for the identification of novel therapies targeting tumors refractory to standard therapy.

MSH6 mutations arise in glioblastomas during temozolomide therapy and mediate temozolomide resistance.

PURPOSE: Over the past few years, the alkylating agent temozolomide has become the standard-of-care therapy for patients with glioblastoma, the most common brain tumor. Recently, large-scale cancer genome sequencing efforts have identified a hypermutation phenotype and inactivating MSH6 mismatch repair gene mutations in recurrent, post-temozolomide glioblastomas, particularly those growing more rapidly during temozolomide treatment. This study aimed to clarify the timing and role of MSH6 mutations in mediating glioblastoma temozolomide resistance. EXPERIMENTAL DESIGN: MSH6 sequence and microsatellite instability (MSI) status were determined in matched prechemotherapy and postchemotherapy glioblastomas identified by The Cancer Genome Atlas (TCGA) as having posttreatment MSH6 mutations. Temozolomide-resistant lines were derived in vitro through selective growth under temozolomide, and the MSH6 gene was sequenced in resistant clones. The role of MSH6 inactivation in mediating resistance was explored using lentiviral short hairpin RNA knockdown and MSH6 reconstitution. RESULTS: MSH6 mutations were confirmed in posttreatment TCGA glioblastomas but absent in matched pretreatment tumors. The posttreatment hypermutation phenotype displayed a signature bias toward CpC transitions and was not associated with MSI. In vitro modeling through exposure of an MSH6 wild-type glioblastoma line to temozolomide resulted in resistant clones; one clone showed an MSH6 mutation, Thr(1219)Ile, that had been independently noted in two treated TCGA glioblastomas. Knockdown of MSH6 in the glioblastoma line U251 increased resistance to temozolomide cytotoxicity and reconstitution restored cytotoxicity in MSH6-null glioma cells. CONCLUSIONS: MSH6 mutations are selected in glioblastomas during temozolomide therapy both in vitro and in vivo and are causally associated with temozolomide resistance.