Molecular Classification Defines 4 Prognostically Distinct Glioma Groups Irrespective of Diagnosis and Grade

J Neuropathol Exp Neurol Copyright Ó 2015 by the American Association of Neuropathologists, Inc. Vol. 74, No. 3 March 2015 pp. 241Y249 ORIGINAL ARTICLE Molecular Classification Defines 4 Prognostically Distinct Glioma Groups Irrespective of Diagnosis and Grade Pilar Mur, PhD, Manuela Mollejo, MD, PhD, Teresa Hernández-Iglesias, Ángel Rodríguez de Lope, MD, PhD, Javier S. Castresana, PhD, Juan F. García, MD, PhD, Concepción Fiaño, MD, Teresa Ribalta, MD, PhD, Juan A. Rey, PhD, and Barbara Meléndez, PhD Abstract According to World Health Organization criteria, diffuse gliomas are divided into several histological subtypes, including astrocytomas, oligodendrogliomas, and oligoastrocytomas, and 4 malignancy grades (IYIV). Molecular alterations, such as the isocitrate dehydrogenase gene (IDH) mutation or 1p/19q loss, are found in these tumors but are not included in the current classification system. Recently, mutation of > thalassemia/mental retardation syndrome X-linked (ATRX) gene and its loss of expression have been reported in infiltrating gliomas. We evaluated ATRX protein expression in 272 gliomas and its association with molecular and clinical features. Loss of ATRX expression was more common in tumors with an astrocytic component (astrocytomas II/III, 46.4%; oligoastrocytomas, 47.5%) but was uncommon in oligodendrogliomas (7.3%) and glioblastomas (0.9%). In astrocytic tumors, loss of ATRX expression was significantly associated with longer overall survival. Remarkably, on the basis of IDH mutation, 1p/19q codeletion, and ATRX expression, our study defined 4 molecularly and prognostically different groups of gliomas, showing the relevance of ATRX expression as a new marker for refining the molecular classification of gliomas and for distinguishing clinically distinct prognostic subgroups of patients. Key Words: 1p/19q codeletion, ATRX alteration, Glioma, IDH mutation. From the Molecular Pathology Research Unit (PM, MM, THI, BM) and Departments of Pathology (MM) and Neurosurgery (ARDL), Virgen de la Salud Hospital, Toledo; Department of Biochemistry and Genetics, University of Navarra School of Sciences, Pamplona (JSC); Department of Pathology, MD Anderson International, Madrid (JFG); Department of Pathology, University Hospital Complex of Vigo, Vigo (CF); Department of Pathology, University Clinic Hospital, Barcelona (TR); and IdiPaz Research Unit, La Paz University Hospital, Madrid (JAR), Spain. Send correspondence and reprint requests to: Barbara Meléndez, PhD, Molecular Pathology Research Unit, Department of Pathology, Virgen de la Salud Hospital, Avda Barber 30, Toledo 45004, Spain; E-mail: bmelendez@sescam.jccm.es This work was supported by grants PI10/01974, PI10/01972, CA12/00318, PI13/00800, INT13/046, PT13/0010/0007, and PI13/00055 from the Fondo de Investigaciones Sanitarias of the Instituto de Salud Carlos III, Ministerio de Ciencia e Innovación, cofunded by the Fondo Europeo de Desarrollo Regional (FEDER 2007-2013) Unión Europea Una manera de hacer Europa. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal s Web site (www.jneuropath.com). J Neuropathol Exp Neurol Volume 74, Number 3, March 2015 INTRODUCTION Gliomas are the most common primary tumors of the CNS and are classified according to morphologic criteria established by the World Health Organization (WHO) into various histologic subtypes and grades, including astrocytomas (AAs), oligoastrocytomas (OAs), and oligodendrogliomas (ODs) (1). Histologic classification of gliomas essentially relies on morphologic similarities of the tumor cells with nonneoplastic glial cells and the presence of particular architectural features. However, in recent years, increasing numbers of genetic and epigenetic alterations have been described in gliomas as a result of our increased knowledge of the molecular biology of brain tumors (2Y4). As a consequence, clinically relevant tissue-based biomarkers have gained significance in the molecular diagnostics of gliomas, namely, mutation of the isocitrate dehydrogenase (IDH) 1 and 2 genes, and the combined deletion of chromosomal arms 1p and 19q. These alterations are currently being used in routine practice because of their clinical significance for the diagnosis, prognosis, and, in the case of the 1p/19q codeletion, for treatment of patients with gliomas (5Y7). The most common genetic alteration first observed in oligodendroglial tumors is the 1p/19q codeletion, which is present in 60% to 80% of ODs and 30% to 50% of OAs (1). This alteration is closely associated with coding mutations of the CIC and FUBP1 genes, as well as with mutations within the promoter of the TERT gene, suggesting their involvement in the pathogenesis of oligodendroglial tumors (8Y11). Mutations of IDH constitute an early alteration, as this mutation is found in most diffuse gliomas of WHO grades II and III, as well as in secondary glioblastomas (GBMs), suggesting a common glial precursor cell, whereas these mutations are not present in primary GBMs, suggesting a different cell of origin. Patients with mutated IDH have a better prognosis (3, 4, 12, 13), although this mutation has not so far proved able to predict response to any particular type of therapy. Several studies have reported that mutation of IDH produces an accumulation of 2-hydroxyglutarate, inhibiting TET2 and histone demethylation and giving rise to the glioma CpG island methylator phenotype (G-CIMP+) (14, 15). Our group recently reported that the G-CIMP+ phenotype is composed of 2 distinct CIMP profiles, both of which are associated with the IDH mutation but are characterized by distinct and specific molecular, histopathologic, and prognostic features (16). The 241

Mur et al J Neuropathol Exp Neurol Volume 74, Number 3, March 2015 first profile, CD-CIMP+, encompassed mostly pure ODs with 1p/19q loss, whereas the other, CIMP+, occurred in 1p/19qintact tumors with frequent TP53 mutation and involved mixed tumors. These observations suggest that the G-CIMP profile may be modulated by specific genetic alterations such as the 1p/19q codeletion or mutation of TP53 (16). Recently, mutation of the > thalassemia/mental retardation syndrome X-linked (ATRX) gene and loss of ATRX protein expression detected by immunohistochemistry have been observed in gliomas of various subtypes and grades (8, 10, 17Y19). This protein enables the incorporation of the variant histone 3.3 into heterochromatin, giving rise to changes in telomere length and genomic instability (9, 20, 21). Mutations in the ATRX gene were detected for the first time in pediatric GBMs (30%Y40%) linked to H3F3A mutations (21). A significant correlation with alternative lengthening of telomeres (ALT) was also identified (22, 23). In adult gliomas, this alteration was found to be more prevalent in astrocytic than in mixed glial tumors and rare in pure ODs. Also, mutations in the ATRX gene were strongly associated with IDH and TP53 mutations (9, 17Y19). The histologic classification of gliomas, of mixed OA in particular, has so far been difficult because the morphologic interpretation is subject to considerable interobserver variation. In fact, recent studies have provided strong evidence against the existence of an independent OA entity (24). Here, we sought to determine the prognostic value of ATRX protein expression and to examine its value in refining the classification and prognosis of glioma subtypes in the context of other relevant molecular alterations in gliomas, such as IDH mutation and 1p/19q codeletion. MATERIALS AND METHODS Tissue Samples and Patient Characteristics Two hundred seventy-two glioma samples were collected at 4 Spanish Hospitals included in the National Tumour Bank Network. Most tumor samples in the present study were included in tissue microarrays (TMAs), which were constructed using formalin-fixed, paraffin-embedded archival tissue blocks, as reported elsewhere (25). Appropriated nontumoral brain tissue samples obtained from autopsies were included as controls for each TMA. Informed consent was obtained, and the investigation was approved by the institutional review board of the Virgen de la Salud Hospital, Toledo, Spain. Tissue sections from all hematoxylin and eosinystained samples were reviewed to verify tumor viability and the histopathologic diagnosis according to the WHO classification of tumors of the CNS (1). This series includes different histologies and grades of gliomas as follows: 68 ODs (grade II, 39; grade III, 29); 40 OAs (grade II, 23; grade III, 17); 56 AAs (grade II, 34; grade III, 22); and 108 GBMs. Clinical information was obtained from the 252 cases; 153 of whom had died (60.7%). Patient characteristics are summarized in Table, Supplemental Digital Content 1 (http://links.lww.com/nen/a701). Immunohistochemistry Immunohistochemical studies were performed on selected glioma samples (TMA or sections) using different 242 primary antibodies. To detect the IDH1 R132H mutation, a mouse monoclonal antibody (dilution 1:100, DIA H09, CKPIIIGHHAYGD; Dianova GmbH, Hamburg, Germany) was used. When enough material was available, IDH-negative cases were sequenced for IDH1 and IDH2 using previously described primers and conditions to detect other potential mutations or to confirm the absence of the mutation (12, 13). P53 immunohistochemical assays were using a primary antibody against p53 (Novocastra Laboratories, Newcastle upon Tyne, UK) and the semiautomatic AutostainerLink 48 (DAKO Cytomation, Glostrup, Denmark) system as reported (25). P53 protein reactivity was scored as positive with more than 50% positive tumor cells. In addition, ATRX immunohistochemistry of polyclonal rabbit antibody (dilution 1:100, HPA001906; Sigma-Aldrich, St. Louis, MO) was performed using an automated immunostainer (Benchmark Ultra, Ventana, Tucson, AZ) with standard protocols. Non-neoplastic tonsil and brain tissues were included in TMAs as internal positive controls for ATRX. For individual tissue sections, endothelial cells, cortical neurons, and infiltrating inflammatory cells were evaluated as internal positive controls. Cases with more than 10% nuclear positive-immunostained tumor cells were scored as positive, as reported elsewhere (18). Cases were evaluated simultaneously by 2 observers on a multiheaded microscope, and a consensus score was agreed on. Fluorescence In Situ Hybridization Analysis Chromosomal loss of 1p and 19q was evaluated in 101 gliomas in which the presence of an oligodendroglial component was determined after pathologic examination. Fluorescence in situ hybridization (FISH) analysis was performed on paraffinembedded tissue sections using the LSI 1p36/1q25 and 19p13/ 19q13 probes (Vysis), as previously described (2). Telomerespecific FISH was performed in 47 cases to detect ALT using an Alexa Fluor 488 PNA (peptide nucleic acid) telomere C probe (N-CCCTAACCCTAACCCTAA-C) (Panagene, Yuseong-gu, Daejeon, Korea), as described (26). Telomere FISH in U2OS cells showing the bright telomere DNA aggregates was used as a positive control for ALT. Tissue sections with more than 1% ALT-positive tumor cells were scored as positive, as established by other authors (23, 27). Methylation Analysis We evaluated the methylation profile of 161 glioma samples: 46 oligodendroglial tumors previously reported by us (16) and 115 GBMs from The Cancer Genome Atlas database for which methylation A values of the 450 K Beadchip platform (Illumina, San Diego, CA) and somatic mutation data were available. Statistical Analysis The t-tests or analysis of variance was used to compare differences between 2 or more groups, respectively. The association between parameters was assessed using Spearman correlation coefficient. Univariate survival analysis was performed using Kaplan-Meier curves and the log-rank test. Multivariate analyses were done, involving a Cox proportional hazards model for which values of p G 0.05 were considered significant. Analyses were carried out using Predictive Analytics Software (SPSS, Chicago, IL). Ó 2015 American Association of Neuropathologists, Inc.

J Neuropathol Exp Neurol Volume 74, Number 3, March 2015 RESULTS Association Between Loss of ATRX Expression and Molecular and Clinical Features The ATRX protein expression and its associations with other molecular alterations were evaluated in 272 glioma samples (Table 1; Fig. 1; Table, Supplemental Digital Content 1, http://links.lww.com/nen/a701). The ATRX protein expression was found to be absent in 51 (18.7%) of the samples. Most of the tumors lacking ATRX expression were IDHmutated (41 of 48 nonyatrx-expressing samples, 85.5%), p53-positive (35 of 48, 72.9%), and intact 1p/19q (25 of 25, 100%) tumors. Of note, 7 nonyatrx-expressing and IDH1 wild-type samples could not be analyzed by sequencing; thus, additional mutations other than the R132H one detected by immunohistochemistry could be present in those samples. Alternative lengthening of telomeres was identified in 21 (95.5%) of 22 cases evaluated as not expressing the ATRX protein. Therefore, ATRX loss of expression seems to be strongly correlated with IDH mutation, the absence of the 1p/19q codeletion, p53 expression, and ALT (p G 0.0001). The association of loss of ATRX expression with tumor subtype was such that the alteration was more frequent in AA grades II (19 of 34, 56%) and III (7 of 22, 32%) and OA grades II (12 of 23, 52%) and III (7 of 17, 41%) than pure grade II and III ODs (4 of 39, 10.3% and 1 of 29, 3.4%, respectively) and GBMs (1 of 108, 0.9%) (p G 0.0001). Furthermore, patients TABLE 1. Molecular and Histopathologic Features With Respect to ATRX Status ATRX Lost, n = 51 (%) ATRX Expressed, n = 221 (%) 1p/19q Codeletion Yes 0 (0) 47 (100) No 25 (46) 29 (54) Total 25 76 IDH mutation Yes 41 (36.6) 71 (63.4) No 7 (5) 140 (95) Total 48 211 P53 expression Yes 35 (41) 50 (59) No 12 (7.2) 154 (92.8) Total 47 204 ALT positive Yes 21 (95.5) 1 (0.5) No 1 (4) 24 (96) Total 22 25 Subtype and grade GBM 1 (1) 107 (99) AA III 7 (32) 15 (68) AA II 19 (56) 15 (44) OA III 7 (41) 10 (59) OA II 12 (52) 11 (48) OD III 1 (3.4) 28 (96.6) OD II 4 (10.3) 35 (89.7) Total 51 221 Molecular Classification of Gliomas who did not express the ATRX protein were younger (mean age, 38 years; range, 20Y55 years) than ATRX-positive patients (mean age, 55 years; range, 23Y82 years) (Table 1; Fig. 1; Table, Supplemental Digital Content 1, http://links.lww.com/ NEN/A701). Favorable Prognostic Markers in Gliomas We examined whether the survival of these patients was related to their molecular and clinical features. In this series, the histology, WHO grade, and age were prognostic factors (p G 0.0001); patients with the OD subtype (mean overall survival [OS], 181.4 months) who were low grade (178.0 months) or relatively young (G65 years) (111.9 months) had a better clinical outcome compared with patients with a different diagnosis (OA, 109.9 months; AA, 28.3 months), with highgrade tumors (44.6 months), or who were aged 65 years or older (13.8 months) (Figure, Supplemental Digital Content 2, http://links.lww.com/nen/a702). Regarding molecular factors, IDH mutation and 1p/19q loss were prognostic of OS in the whole series or within each subtype, grade, or age group (p G 0.01). Also, although ALT was analyzed in a few cases (n = 47), the favorable prognostic value of ALT-positive cases was observed (p G 0.0001) (Fig. 2). Conversely, p53 expression was not significantly correlated with OS (p = 0.81). Given that the frequency of the ATRX alteration in pure ODs and GBMs was low and that it was mutually exclusive with the 1p/19q codeletion, we evaluated the prognostic value of ATRX expression in cases with an astrocytic component (OAs and AAs grades II and III) and intact 1p/19q. In this set of tumors, the patients lacking ATRX expression had a better prognosis (mean OS, 111.8 months) than patients with positive ATRX expression (47.5 months) (p G 0.0001). P53 expression also had a significant prognostic value in these patients (p = 0.039), whereby survival was longer in p53-expressing cases than in those that did not express it (mean OS, 93.5 vs 47.2 months, respectively) (Fig. 2). A multivariate Cox analysis of the clinical and molecular factors showed that the presence of the IDH mutation (p = 0.0001; hazard ratio, 7.712; 95% confidence interval, 3.28Y18.11) and the WHO grade (p = 0.031; hazard ratio, 2.306; 95 % confidence interval, 1.08Y4.92) were significant independent prognostic factors of OS in our model. Prognostic Value of the Molecular Classification Based on IDH Mutation, ATRX Expression, and 1p/19q Codeletion To determine whether molecular classification of the gliomas based on the molecular prognostic factors analyzed could establish different prognostic subgroups of tumors, we classified the samples according to their IDH mutational status, ATRX expression, and 1p/19q status. On the basis of these molecular markers, 235 tumors could be distributed among the 4 groups, whereas 6 could not be classified into any of them (Table 2; Fig. 3A). An additional 31 samples could not be classified because of insufficient data. First,weidentifiedagroupof46tumorswiththeIDH mutation, 1p/19q codeletion, and ATRX-positive expression, which we named molecular group 1 (I-CD). This group was mostly composed of pure ODs (89%, 41 of 46) and grade II Ó 2015 American Association of Neuropathologists, Inc. 243

Mur et al J Neuropathol Exp Neurol Volume 74, Number 3, March 2015 FIGURE 1. Association between loss of ATRX expression and molecular and clinical features in 272 glioma samples. (A) Loss of ATRX expression was strongly correlated with IDH-mutated, 1p/19q-intact, p53-positive, and ALT-positive status (p G 0.0001). (B) Loss of ATRX expression was more frequent in astrocytomas II/III and oligoastrocytomas II/III. (C) Age distribution (years) with respect to ATRX protein expression. Patients with loss of ATRX protein expression were younger than patients with ATRX-positive tumors. (D) Immunohistochemical staining for ATRX of 1 tumor sample with loss of ATRX nuclear expression. Red arrow indicates endothelial cells as non-neoplastic positive internal control. (E) The ALT phenotype determined by ultrabright signals (green) using telomere-specific fluorescence in situ hybridization in a tissue section of a representative case with loss of ATRX expression. AA II, III, grade II, III astrocytoma; ALT, alternative lengthening of telomeres; ATRX, > thalassemia/mental retardation syndrome X-linked; GBM, glioblastoma; OA II, III, grade II, III oligoastrocytoma; OD II, III, grade II, III oligodendroglioma. gliomas (61%, 28 of 46) with absence of p53 protein expression (87%, 40 of 46). The molecular group 2 (I-A) (38 patients) was characterized by IDH mutation, absence of ATRX protein expression, and intact 1p/19q. These alterations were significantly enriched in astrocytic(89%,22 AAs grades IIYIII and 12 OAs grades IIYIII) and grade II gliomas (71%, 27 of 38). P53-positive expression was frequently detected in this group (71.4%,25of35cases).Athirdsetof16patients, molecular group 3 (I), showed IDH mutation but intact 1p/19q and ATRX-positive expression. These patients included different diagnoses (2 GBMs, 8 AAs II/III, 3 OAs, and 3 ODs), more frequently had grade II tumors (75%, 12 of 16), and showed p53-positive expression (67%, 8 of 12). Finally, a large group of 135 patients were assigned to molecular group 4 (nonaltered [NA]), which was characterized by wild-type IDH, intact 1p/19q, and ATRX-positive expression. These tumors were mostly p53-negative (68%, 87 of 128) and had a histologic diagnosis of GBM (78.8%; 104 of 135), although grade II tumors (8 cases) and gliomas with an oligodendroglial component (14 cases) were also included. Age at diagnosis and clinical behavior were also evaluated in these groups. Patients of molecular groups 1 (I-CD), 2 (I-A), and 3 (I) were younger than those of molecular group 4 (NA) (mean age, 43.9, 39.2, and 40.8 years old vs 60.7 years old, respectively). Moreover, Kaplan-Meier curves revealed significant differences in OS between the molecular groups (p G 0.0001) (Table 2; Fig. 3B). Patients with the I-CD and I-A 244 signatures survived longer than patients of the I group (mean OS, 179.5, 118.0, and 82.7 months, respectively; p = 0.029), and patients without IDH, 1p/19q, and ATRX alterations ( NA ) had the poorest prognosis (mean OS, 10.6 months). Therefore, based on these findings, we propose a molecular classification of gliomas according to the status of IDH mutation, 1p/19q codeletion, and ATRX protein expression (Fig. 4). Association of the CIMP Methylation Profile With ATRX Expression We examined whether the expression of the ATRX protein was associated with the 3 previously identified methylation profiles, CIMP+, CD-CIMP+, and CIMPY (16). Samples classified as CIMP+ (n = 7) were associated with loss of ATRX expression, whereas all CIMPY (n = 14) and CD- CIMP+ (n = 25) tumors showed ATRX-positive expression. Evaluation of The Cancer Genome Atlas series showed ATRX mutation (which was reported to correlate with the lack of protein expression) detected in CIMP+ (n = 4) and CIMPY (n = 4) samples, whereas all tumors with wild-type ATRX were characterized by a CIMPY profile (n = 107). DISCUSSION Across recent decades, molecular genetic studies have greatly increased our knowledge and understanding of the development of gliomas. More recently still, next-generation Ó 2015 American Association of Neuropathologists, Inc.

J Neuropathol Exp Neurol Volume 74, Number 3, March 2015 Molecular Classification of Gliomas FIGURE 2. (AYD) Kaplan-Meier survival curves of all glioma cases based on IDH mutational status (A) and 1p/19q codeletion (B). In World Health Organization grade II and III astrocytomas and oligoastrocytoma patients with intact 1p/19q, loss of > thalassemia/ mental retardation syndrome X-linked (ATRX) protein expression (C) and p53-positive expression (D) were associated with significantly better survival. P values are those from log-rank tests. ms, median survival (months). sequencing technologies have enabled novel mutated genes to be identified, thereby providing biomarkers that may be of use in the histologic classification of diffuse gliomas, the prediction of prognosis, and the choice of the most appropriate available therapy. Mutation of the IDH 1 and 2 genes is frequent in secondary glioblastomas and diffuse gliomas of WHO grades II and III and, therefore, may be considered an early event in glioma development (3, 28). This mutation leads to an increase of 2-hydroxyglutarate intracellular concentrations, which modulates the epigenome, that is, genomic DNA methylation and histone methylation, inhibiting TET2 and histone demethylation. As a consequence, the glioma CpG island methylator phenotype (G-CIMP+) is found in IDHmutated gliomas (14, 15, 29, 30). However, our group recently observed that the G-CIMP+ profile actually is composed of 2 hypermethylator phenotypes, both of which are associated with the IDH mutation, although each characterized by specific additional molecular alterations. The CD-CIMP+ form encompasses tumors with 1p/19q loss, whereas the CIMP+ occurs in 1p/19q-intact tumors that frequently carry the TP53 mutation (16). In addition, other somatic mutations in genes coding for histone- or chromatin-modifying proteins that cause disruption of multiple epigenetic regulatory processes (by affecting histone modification, DNA methylation, and chromatin remodeling) Ó 2015 American Association of Neuropathologists, Inc. 245

Mur et al J Neuropathol Exp Neurol Volume 74, Number 3, March 2015 TABLE 2. Molecular Classification of 235 Glioma Samples IDH Mutated, 1p/19q Codeleted, ATRX Positive IDH Mutated, 1p/19q Intact, ATRX Negative IDH Mutated, 1p/19q Intact, ATRX Positive IDH wild type, 1p/19q Intact, ATRX Positive Subtype and grade GROUP I-CD GROUP I-A GROUP I GROUP NA n = 46 (%) n = 38 (%) n = 16 (%) n = 135 (%) GBM 0 (0) 0 (0) 2 (12.5) 104 (77.04) AA III 0 (0) 6 (15.79) 2 (12.50) 13 (9.63) AA II 0 (0) 16 (42.11) 6 (37.50) 4 (2.96) OA III 2 (4.35) 4 (10.53) 0 (0) 5 (3.70) OA II 3 (6.52) 8 (21.05) 3 (18.75) 2 (1.48) OD III 16 (34.78) 1 (2.63) 0 (0) 5 (3.70) OD II 25 (54.35) 3 (7.89) 3 (18.75) 2 (1.48) Mean age (range), years 43.9 39.2 40.8 60.7 (23Y78) (21Y55) (28.2Y71.5) (24Y82) Mean overall survival (range), months 179.5 117 82.7 10.63 (152Y207) (87.4Y149) (49Y116.3) (6.6Y14.6) AA III, anaplastic astrocytoma; OA III, anaplastic oligoastrocytoma; OD III, anaplastic oligodendroglioma; ATRX, > thalassemia/mental retardation syndrome X-linked; AA II, diffuse astrocytoma; GBM, glioblastoma; IDH, isocitrate dehydrogenase; OA II, oligoastrocytoma; OD II, oligodendroglioma. have been reported in adult as well as childhood gliomas (17, 19, 21, 31, 32). In line with this, it has been suggested that mutation of the ATRX gene, which codes for a core component of a chromatin remodeling complex acting in subtelomeric regions, contributes to deregulated telomere length and chromosomal instability. Presumably, this alteration is related to the maintenance or increase of telomere length through the phenomenon of ALT that is observed in some cancer cell types. In this study of the Spanish cohort, we observed a frequent lack of ATRX protein expression in astrocytic tumors (~45% in AAs and OAs of WHO grades II and III), in contrast to the infrequent loss of this protein in ODs (7.3%) and primary GBMs (0.9%). This absence of ATRX expression was strongly associated with IDH mutation (85.5%), p53-positive expression (72.9%), and ALT positivity (95.5%), whereas it was mutually exclusive with 1p/19q codeletion (p G 0.0001). Furthermore, the methylation analyses revealed that IDHmutated samples classified as having a CIMP+ phenotype do not express the ATRX protein. Our results are partially supported by other studies (9, 17Y19, 24, 33, 34) and indicate that ATRX alteration defines a subgroup of astrocytic gliomas with specific genetic (concurrent alterations of IDH, ATRX, and TP53) and epigenetic (CIMP+) profiles, in which the ALT mechanism for the maintenance of telomere length takes place. In addition, these results may suggest that the loss of ATRX function in this specific group of gliomas is involved FIGURE 3. Molecular classification based on IDH status, 1p/19q codeletion, and > thalassemia/mental retardation syndrome X- linked (ATRX) protein expression. (A) Representation of the different molecular groups showing diagnosis, grade, IDH status, 1p/ 19q codeletion, and ATRX and p53 protein expression. A set of 30 cases without IDH or 1p/19q codeletion data was not included in this classification. (B) Kaplan-Meier curves according to the molecular classification showed differences between groups (p G 0.0001). Molecular groups correspond to Group I-CD, IDH mutated, 1p/19q codeleted, and ATRX expressed; Group I-A, IDH mutated, 1p/19q intact, and ATRX loss; Group I, IDH mutated, 1p/19q intact, and ATRX expressed; Group NA, IDH wild type, 1p/19q intact, and ATRX expressed; Non-classifiable: 6 tumors that do not belong to any of these groups. ms, mean survival (months). 246 Ó 2015 American Association of Neuropathologists, Inc.

J Neuropathol Exp Neurol Volume 74, Number 3, March 2015 Molecular Classification of Gliomas FIGURE 4. Prognosis algorithms proposed for gliomas based on IDH mutational status, 1p/19q codeletion, and ATRX protein expression. ALT, alternative lengthening of telomeres; ATRX, > thalassemia/mental retardation syndrome X-linked; GBM, glioblastoma; H&E, hematoxylin and eosin; A, OA, OD, II, III, astrocytoma, oligoastrocytoma, oligodendroglioma, World Health Organization grades II, III; OS, overall survival. in the gliomagenesis initiated by IDH mutation. However, the contributions of mutant TP53 to telomere length or the modulation of the methylation profile require further investigation. On the other hand, gliomas of another subgroup, mostly with a histopathologic diagnosis of OD, frequently have the IDH mutation (as well as CIC and FUBP1 mutations) and the 1p/19q codeletion (8, 11, 35). In addition, these tumors have a CD-CIMP+ profile and show ATRX-positive expression. It is worth noting that a different mechanism for alteration of telomere length caused by upregulated telomerase (htert) expression as a consequence of TERT promoter hotspot mutations in the presence of functional p53 has been reported in these ODs (10). Therefore, based on the evaluation of the IDH mutation, 1p/19q codeletion, and ATRX expression, we define here for the first time 4 molecular subtypes of gliomas, of which 3 have the IDH mutation ( I-CD, I-A, and I tumors). In addition, we found associations between these molecular signatures and the clinical outcome of glioma patients. I-CD tumors (IDH mutated, 1p/19q codeleted, and ATRX expressing) had a better prognosis (mean OS, 179.5 months) than the intermediate I-A tumors (IDH mutated, altered ATRX and p53, 1p/19q intact) and the prognostically poorer I tumors (mean OS, 117.0 months). Here we provide evidence that the lack of ATRX protein expression in the context of intact 1p/19q codeletion could play an important role in the clinical course of astrocytic tumors. These results demonstrate the importance of assessing the 1p/19q codeletion and ATRX alteration in IDH-mutated tumors to identify the distinct prognostic group to which a case belongs. Interestingly, we detected a new and clinically different group of tumors ( I ); the patients in this group, despite being IDH mutated, low grade (75%), and young (mean age, 40.8 years), have a significantly worse prognosis (mean OS, 82.7 months) than the previously mentioned molecular groups. A recent study of somatic genomic alterations in GBMs showed that more than 40% of tumors harbored at Ó 2015 American Association of Neuropathologists, Inc. 247

Mur et al J Neuropathol Exp Neurol Volume 74, Number 3, March 2015 least 1 nonsynonymous mutation among chromatin-modifier genes (CMGs), suggesting potential biologic relevance of chromatin modification in GBM (36). In that study, although most of the tumors with CMG mutations had concurrent IDH and ATRX mutations, IDH-mutated GBMs without ATRX mutation were also detected. In the present report, although additional molecular alterationsinthesamplesofthis I group remain to be studied, the possibility of concurrent mutations of IDH and CMGs should not be excluded. Finally, the poorest-prognosis molecular group NA (mean OS, 10.6 months) lacks all of the alterations used in this classification. Low-grade tumors or tumors with an oligodendroglial component were also represented in the latter subgroup, even though it is mostly composed of primary GBMs, illustrating the need to consider these molecular markers when classifying and evaluating the prognosis of glioma patients. Previous reports have also demonstrated the prognostic value of ATRX but have proposed different classification schemes (9, 18). The first of these studies (by Jiao et al) classified samples into 3 molecular groups of gliomas according to IDH, ATRX, CIC, and FUBP1 mutational status: group I-CF (similar to I-CD in this study), group I-A, and group I-X, the latter including mutated IDH, ATRX-positive tumors, as well as wild-type IDH gliomas (9). Nevertheless, we have shown here that the genetically heterogeneous samples classified by these authors within the same group (I-X) have different outcomes. On the other hand, molecular classification of anaplastic gliomas as suggested by Wiestler et al (18) was ambiguous because their classification grouped anaplastic AAs (with or without ATRX loss) with anaplastic OAs with ATRX loss. We demonstrate here that anaplastic AAs with ATRX loss (group I-A in our study) show better OS than those without this alteration (group I ). Other WHO grade II tumors (histopathologically ODs, AAs, or OAs) with a bad prognosis are also included in the latter group. Of note, differences with these studies could also be attributed to the population under study. Remarkably, this is the first study on a large Spanish cohort that analyzes alterations of IDH, 1p/19q, and ATRX. In conclusion, our findings reveal ATRX expression to be a probable new biomarker in adult gliomas that, in combination with IDH and 1p/19q determinations, could be useful for the diagnosis of cases with ambiguous histologic features and for determining clinical and prognostic groups. ACKNOWLEDGMENTS In loving memory of Alberto Mur. We acknowledge the Spanish Tumour Bank Network of the Centro Nacional de Investigaciones Oncológicas, Madrid, Spain, and the Tumor Bank of the Hospital Virgen de la Salud (BioB-HVS), Toledo, Spain, for providing tumor samples. REFERENCES 1. Louis DN, Ohgaki H, Wiestler OD, et al. The 2007 WHO classification of tumours of the central nervous system. Acta Neuropathol 2007;114: 97Y109 2. Blesa D, Mollejo M, Ruano Y, et al. Novel genomic alterations and mechanisms associated with tumor progression in oligodendroglioma and mixed oligoastrocytoma. J Neuropathol Exp Neurol 2009;68:274Y85 248 3. Parsons DW, Jones S, Zhang X, et al. An integrated genomic analysis of human glioblastoma multiforme. Science 2008;321:1807Y12 4. von Deimling A, Korshunov A, Hartmann C. The next generation of glioma biomarkers: MGMT methylation, BRAF fusions and IDH1 mutations. Brain Pathol 2011;21:74Y87 5. Cairncross JG, Ueki K, Zlatescu MC, et al. Specific genetic predictors of chemotherapeutic response and survival in patients with anaplastic oligodendrogliomas. J Natl Cancer Inst 1998;90:1473Y79 6. Smith JS, Perry A, Borell TJ, et al. Alterations of chromosome arms 1p and 19q as predictors of survival in oligodendrogliomas, astrocytomas, and mixed oligoastrocytomas. J Clin Oncol 2000;18:636Y45 7. Weller M, Berger H, Hartmann C, et al. Combined 1p/19q loss in oligodendroglial tumors: Predictive or prognostic biomarker? Clin Cancer Res 2007;13:6933Y37 8. Bettegowda C, Agrawal N, Jiao Y, et al. Mutations in CIC and FUBP1 contribute to human oligodendroglioma. Science 2011;333:1453Y55 9. Jiao Y, Killela PJ, Reitman ZJ, et al. Frequent ATRX, CIC, FUBP1 and IDH1 mutations refine the classification of malignant gliomas. Oncotarget 2012;3:709Y22 10. Killela PJ, Reitman ZJ, Jiao Y, et al. TERT promoter mutations occur frequently in gliomas and a subset of tumors derived from cells with low rates of self-renewal. Proc Natl Acad Sci USA 2013;110:6021Y26 11. Yip S, Butterfield YS, Morozova O, et al. Concurrent CIC mutations, IDH mutations, and 1p/19q loss distinguish oligodendrogliomas from other cancers. J Pathol 2012;226:7Y16 12. Balss J, Meyer J, Mueller W, et al. Analysis of the IDH1 codon 132 mutation in brain tumors. Acta Neuropathol 2008;116:597Y602 13. Hartmann C, Meyer J, Balss J, et al. Type and frequency of IDH1 and IDH2 mutations are related to astrocytic and oligodendroglial differentiation and age: A study of 1,010 diffuse gliomas. Acta Neuropathol 2009; 118:469Y74 14. Lu C, Ward PS, Kapoor GS, et al. IDH mutation impairs histone demethylation and results in a block to cell differentiation. Nature 2012; 483:474Y78 15. Turcan S, Rohle D, Goenka A, et al. IDH1 mutation is sufficient to establish the glioma hypermethylator phenotype. Nature 2012;483:479Y83 16. Mur P, Mollejo M, Ruano Y, et al. Codeletion of 1p and 19q determines distinct gene methylation and expression profiles in IDH-mutated oligodendroglial tumors. Acta Neuropathol 2013;126:277Y89 17. Kannan K, Inagaki A, Silber J, et al. Whole-exome sequencing identifies ATRX mutation as a key molecular determinant in lower-grade glioma. Oncotarget 2012;3:1194Y203 18. Wiestler B, Capper D, Holland-Letz T, et al. ATRX loss refines the classification of anaplastic gliomas and identifies a subgroup of IDH mutant astrocytic tumors with better prognosis. Acta Neuropathol 2013; 126:443Y51 19. Liu XY, Gerges N, Korshunov A, et al. Frequent ATRX mutations and loss of expression in adult diffuse astrocytic tumors carrying IDH1/IDH2 and TP53 mutations. Acta Neuropathol 2012;124:615Y25 20. Clynes D, Higgs DR, Gibbons RJ. The chromatin remodeller ATRX: A repeat offender in human disease. Trends Biochem Sci 2013;38:461Y66 21. Schwartzentruber J, Korshunov A, Liu XY, et al. Driver mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma. Nature 2012;482:226Y31 22. Heaphy CM, de Wilde RF, Jiao Y, et al. Altered telomeres in tumors with ATRX and DAXX mutations. Science 2011;333:425 23. Nguyen DN, Heaphy CM, de Wilde RF, et al. Molecular and morphologic correlates of the alternative lengthening of telomeres phenotype in high-grade astrocytomas. Brain Pathol 2013;23:237Y43 24. Sahm F, Reuss D, Koelsche C, et al. Farewell to oligoastrocytoma: In situ molecular genetics favor classification as either oligodendroglioma or astrocytoma. Acta Neuropathol 2014;128:551Y59 25. Ruano Y, Ribalta T, de Lope AR, et al. Worse outcome in primary glioblastoma multiforme with concurrent epidermal growth factor receptor and p53 alteration. Am J Clin Pathol 2009;131:257Y63 26. Meeker AK, Gage WR, Hicks JL, et al. Telomere length assessment in human archival tissues: Combined telomere fluorescence in situ hybridization and immunostaining. Am J Pathol 2002;160:1259Y68 27. Abedalthagafi M, Phillips JJ, Kim GE, et al. The alternative lengthening of telomere phenotype is significantly associated with loss of ATRX expression in high-grade pediatric and adult astrocytomas: A multi-institutional study of 214 astrocytomas. Mod Pathol 2013;26:1425Y32 Ó 2015 American Association of Neuropathologists, Inc.

J Neuropathol Exp Neurol Volume 74, Number 3, March 2015 28. Gupta R, Flanagan S, Li CC, et al. Expanding the spectrum of IDH1 mutations in gliomas. Mod Pathol 2013;26:619Y25 29. Noushmehr H, Weisenberger DJ, Diefes K, et al. Identification of a CpG island methylator phenotype that defines a distinct subgroup of glioma. Cancer Cell 2010;17:510Y22 30. Duncan CG, Barwick BG, Jin G, et al. A heterozygous IDH1R132H/WT mutation induces genome-wide alterations in DNA methylation. Genome Res 2012;22:2339Y55 31. Sturm D, Witt H, Hovestadt V, et al. Hotspot mutations in H3F3A and IDH1 define distinct epigenetic and biological subgroups of glioblastoma. Cancer Cell 2012;22:425Y37 32. Sturm D, Bender S, Jones DT, et al. Paediatric and adult glioblastoma: Multiform (epi)genomic culprits emerge. Nat Rev Cancer 2014;14:92Y107 Molecular Classification of Gliomas 33. Cai J, Yang P, Zhang C, et al. ATRX mrna expression combined with IDH1/2 mutational status and Ki-67 expression refines the molecular classification of astrocytic tumors: Evidence from the whole transcriptome sequencing of 169 samples samples. Oncotarget 2014;5:2551Y61 34. Gorovets D, Kannan K, Shen R, et al. IDH mutation and neuroglial developmental features define clinically distinct subclasses of lower grade diffuse astrocytic glioma. Clin Cancer Res 2012;18:2490Y501 35. Sahm F, Koelsche C, Meyer J, et al. CIC and FUBP1 mutations in oligodendrogliomas, oligoastrocytomas and astrocytomas. Acta Neuropathol 2012;123:853Y60 36. Brennan CW, Verhaak RG, McKenna A, et al. The somatic genomic landscape of glioblastoma. Cell 2013;155:462Y77 Ó 2015 American Association of Neuropathologists, Inc. 249