Molecular Subtypes of Anaplastic Oligodendroglioma: Implications for Patient Management at Diagnosis1

Vol. 7, 839–845, April 2001 Clinical Cancer Research 839

Yasushi Ino, Rebecca A. Betensky, without TP53 mutations, which are poorly responsive, ag- Magdalena C. Zlatescu, Hikaru Sasaki, gressive tumors that are clinically and genotypically similar David R. Macdonald, to glioblastomas. Conclusions: These data raise the possibility, for the Anat O. Stemmer-Rachamimov, first time, that therapeutic decisions at the time of diagnosis 2 David A. Ramsay, J. Gregory Cairncross, and might be tailored to particular genetic subtypes of anaplastic David N. Louis oligodendroglioma. Molecular Neuro-Oncology Laboratory, Department of Pathology and Neurosurgical Service, Massachusetts General Hospital and Harvard Medical School [Y. I., H. S., A. O. S-R., D. N. L.] and Department of INTRODUCTION Biostatistics, Harvard School of Public Health, Boston, Massachusetts Malignant gliomas are the most common type of primary 02114 [R. A. B.], and Departments of Clinical Neurological Sciences, ϳ Oncology, and Pathology, University of Western Ontario and London brain tumor, with 12,000 new cases diagnosed each year in the Regional Cancer Centre, London, Ontario N6A 4L6, Canada United States (1). For nearly a century, malignant gliomas have [M. C. Z., D. R. M., D. A. R., J. G. C.] been classified on the basis of their histological appearance as astrocytomas (including glioblastomas), oligodendrogliomas, ependymomas, or mixed gliomas. For each type, surgical resec- ABSTRACT tion and radiation therapy have been the mainstays of treatment. Purpose: In a prior study of anaplastic oligodendrogli- Cytotoxic drugs have had a relatively minor therapeutic role omas treated with chemotherapy at diagnosis or at recur- because responses to chemotherapy generally have been infre- rence after radiotherapy, allelic loss of chromosome 1p cor- quent, brief, and unpredictable. The only notable exception has related with better chemotherapeutic response and overall been tumors with oligodendroglial histology, which have a survival. However, in this group of patients in whom ther- greater likelihood of radiographic response to chemotherapy. apeutic management was not uniform, loss of 1p did not Unfortunately, the microscopic distinction between high-grade identify all chemosensitive tumors, nor did all patients oligodendrogliomas, which often are chemosensitive, and glio- whose tumors harbor a 1p loss have long survival. blastomas, which are notoriously recalcitrant to the chemother- Experimental Design: To clarify the clinical relevance of apies used at present, is problematic because these two types of molecular genetic testing at the time of diagnosis for patients malignant glioma may share histological features such as small with anaplastic oligodendrogliomas, we studied a larger, cells, vascular proliferation, and necrosis. Such problems raise more homogeneous group of 50 patients with histologically the question of whether histological diagnosis can be refined in defined anaplastic oligodendrogliomas treated with a che- a clinically useful way. Clearly, more precise identification of motherapeutic regimen as the principal initial therapy. chemosensitive tumors and potential long-term survivors at di- Results: We demonstrate that these tumors can be diagnosis would afford greater flexibility in initial treatment de- vided genetically into four therapeutically and prognosti- cisions for such patients, and improved identification of resistant cally relevant subgroups. Patients whose tumors have com- tumors might justify prescribing novel treatments at diagnosis bined but isolated losses of 1p and 19q have marked and for patients with a poor prognosis (2). With a rational basis for durable responses to chemotherapy associated with long delivering or withholding chemotherapy, effective treatments survival, with or without postoperative radiation therapy. will be appropriately prescribed, and ineffective, toxic, and Other tumors with chromosome 1p alterations also respond costly treatments specifically avoided. to chemotherapy, but with shorter duration of response and Allelic loss of chromosomal arm 1p is emerging as a patient survival. Tumors lacking 1p loss can also be divided marker of chemotherapeutic response (3) and long survival (3, into two subgroups: those with TP53 mutations, which may 4) in patients with histologically defined anaplastic oligoden- also respond to chemotherapy but recur quickly, and those drogliomas. Furthermore, 1p loss may also identify other treatment-sensitive malignant gliomas, including rare glioblastomas (5). However, 1p loss does not identify all chemosensitive anaplastic oligodendrogliomas, nor do all patients whose tumors Received 11/16/00; revised 1/8/01; accepted 1/11/01. harbor 1p loss have long survival, although those with chromo- The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked some 1p and 19q loss may have particularly favorable outcomes advertisement in accordance with 18 U.S.C. Section 1734 solely to (4). In addition, prior studies included patients treated with a indicate this fact. variety of chemotherapeutic and radiotherapeutic regimens (3, 1 Supported by NIH Grants CA57683 and MRC-MOP-37849. 4). For example, in our previous study, some patients were 2 To whom requests for reprints should be addressed, at London Re- gional Cancer Centre, 790 Commissioners Road, London, Ontario, N6A treated with chemotherapy at diagnosis, whereas others were 4L6 Canada. Phone: (519) 685-8615; Fax: (519) 685-8611; E-mail: treated with radiation therapy at diagnosis and only received [email protected]. chemotherapy at recurrence (3). The practical utility of assess-

Table 1 Clinical and genetic features oligodendrogliomas, WHO grade III (Fig. 1; Ref. 6). Cases with Number of patients 50 definite astrocytic tumor components were considered oligo- Mean age at diagnosis (range), years 45.3 (17.4–82.1) astrocytomas and excluded. Thirty-four tumors showed radio- M/F 26/24 graphic contrast enhancement, with 9 displaying ring enhance- KPS (median) 80 ment. These investigations have been approved by the Enhancement at diagnosis 34/50 (68%) Massachusetts General Hospital Subcommittee on Human Stud- Ring enhancement at diagnosis 9/50 (18%) 1p LOHa 29/50 (58%) ies and the Review Board for Health Science Research Involv- 19q LOH 33/50 (66%) ing Human Subjects at the University of Western Ontario. None 10q LOH 12/47 (26%) of the patients in this analysis are included in ongoing chemo- PTEN alteration 6/49 (12%) therapy trials for oligodendroglioma. CDKN2A deletion 7/50 (14%) EGFR amplification 7/50 (14%) Molecular Genetic Analysis. Tumor DNA was ex- TP53 mutation 9/50 (18%) tracted from microdissected formalin-fixed, paraffin-embedded a LOH, loss of heterozygosity. sections; constitutional DNA was extracted from blood lympho- cytes or from formalin-fixed, paraffin-embedded sections of adjacent, uninvolved brain or other tissues (7). Allelic chromosomal loss was assessed by loss of heterozygosity assays in ing chromosome 1p status in anaplastic oligodendrogliomas at constitutional DNA/tumor DNA pairs using microsatellite the time of initial diagnosis, therefore, required clarification. To markers on 1p36.3 (D1S2734, D1S199, and D1S508), 19q13.3 explore further the potential of molecular genetic analysis to (D19S219, D19S112, D19S412, and D19S596), 10q23-24 enhance diagnosis and guide treatment for patients with malig- (D10S185 and D10S2491, near PTEN), and 10q25-26 nant gliomas, we undertook a detailed clinical-molecular genetic (D10S587; Refs. 3, 7). Exons 5–8 of the TP53 gene were correlative study of a larger and more homogeneous set of screened for mutation by single-strand conformation polymor- patients with histologically defined anaplastic oligodendroglio- phism analysis and direct sequencing (8). Exons 1–9 of the mas, all of whom had been treated with a chemotherapeutic PTEN gene were screened for mutations in tumors with 10q regimen as the principal initial therapy. We hypothesized that loss, using similar single-strand conformation polymorphism molecular subtyping could be carried beyond chromosome 1p analysis and sequencing approaches (9). Homozygous deletion analysis and that there were multiple distinct biological types of of the CDKN2A gene and PTEN gene was evaluated by com- anaplastic oligodendroglioma. parative multiplex PCR (10, 11) and EGFR gene amplification by differential PCR (12). MATERIALS AND METHODS Statistical Analysis. The primary goal of the analysis Clinical Parameters. The study included 50 patients was to test the a priori hypothesis that a particular grouping of treated at the London Regional Cancer Center with histologi- the patients according to molecular genetic events is associated cally confirmed anaplastic oligodendrogliomas in whom chem- with chemotherapeutic response, duration of response to chemotherapy was used as an integral part of an overall patient otherapy, and overall survival. Secondary analyses were under- management strategy from diagnosis. The cases are detailed in taken to investigate the effects of all of the observed genetic Table 1. Forty-eight patients received combination chemother- alterations, without regard for the a priori hypothesis. Associ- apy with PCV,3 1 received carmustine (BCNU), and 1 received ations among all of the observed genetic alterations were exam- temozolomide. Thirty-four of these patients received radiation ined, and the effects of the genetic alterations on therapeutic therapy after chemotherapy (as consolidation or at recurrence), response, duration of response, and survival were investigated. 5 received radiation therapy concurrent with chemotherapy, and The added effects of nongenetic features were also examined. 11 were not irradiated. At the start of chemotherapy, 38 tumors Odds ratios and tetrachoric correlations were used to meas- were assessable for radiographic response to chemotherapy: 25 ure the pair-wise associations among genetic alterations (13). (66%) responded to chemotherapy, with 10 complete responses. Fisher’s exact test was used to compare response rates, logistic All nonresponding tumors had been treated with PCV. Median regression was used to model response, and Cox proportional follow-up time from diagnosis was 107 months (minimum, 7 hazard regression was used to model duration of response and months). Although half of these patients were included in a survival. Duration of response and survival were both censored previous study (3), the present series reports analysis of addi- at the time of last follow-up. Classification trees were con- tional clinical and genetic parameters on these cases, together structed using CART (14) to best group the genetic alterations with longer follow-up times, and excludes patients treated with with respect to response. All reported Ps are two-sided. chemotherapy only at recurrence after radiation therapy. This Response to chemotherapy was defined radiographically as therefore represents a substantially larger and more homoge- a decrease in tumor size of Ն50% (15). Care was taken to neously managed group than studied previously. control for steroid effects and other false-positive responses (16, All 50 tumors were classified and graded by at least two 17). For comparison, the absence of disease progression 6 neuropathologists, blinded to the genetic results, as anaplastic months after the start of chemotherapy was considered as an alternative definition of response (18). Duration of response was measured from the start of chemotherapy to the first sign of radiological or clinical progression. 3 The abbreviations used are: PCV, procarbazine, lomustine (CCNU), Duration of response was also analyzed using the alternative and vincristine; KPS, Karnofsky performance status. definition of response, but with measurement from 6 months

Fig. 1 Histological features from different genetic groups are classical for oligodendroglioma and strikingly similar. Case 1720 (left) had combined, isolated 1p and 19q loss (group 1); this patient had a complete neuroradiological response to chemotherapy, a durable response, and long survival (Ͼ9 years). Case 1920 (right) had CDKN2A deletion and 10q and 19q loss (group 4), as well as radiological ring enhancement. There was no response to chemotherapy, and the patient died 8 months after diagnosis.

after the start of chemotherapy. Survival was measured from diagnosis. The analyses were refined further by stratifying for the starting time of radiation therapy. This was done through the use of time-dependent strata (19), defined by whether the patient had started radiation therapy.

RESULTS Molecular Genetic Findings and Groups. The molecular genetic results are listed in Table 1 and demonstrate that this series of anaplastic oligodendrogliomas is genotypically similar to those of other published series (3, 6, 9, 20–23). Allelic losses of 1p and 19q were positively associated with one another Fig. 2 Genetic analysis suggests the presence of four clinically rele- (P Ͻ 0.001), whereas loss of 10q, PTEN alteration, TP53 vant groups of histologically defined anaplastic oligodendrogliomas. Median values are given for age, duration of response, and survival from group 2 patients have 1p loss but either do not have ,ء .(mutation, EGFR amplification, and CDKN2A deletion were diagnosis (dx negatively associated with 1p and 19q loss and positively asso- 19q loss or have other genetics alterations such as TP53 mutation, PTEN ,ءء .ciated with each other. These associations suggested the pres- mutation, 10q loss, EGFR amplification, or CDKN2A deletion ence of at least two groups of tumors: those with 1p and 19q group 4 patients often had PTEN mutation, 10q loss, EGFR amplification, CDKN2A deletion, and ring enhancement. loss, and those with the other genetic alterations. However, because not all patients whose tumors display 1p loss have long survival and not all chemosensitive oligodendrogliomas harbor 1p loss (2–5), we postulated that these cases could be further Analyses of Chemotherapeutic Response by Molecular subdivided. For example, it has been suggested that patients Genetic Parameters. Rates of neuroradiological response to whose tumors have combined 1p and 19q loss have longer chemotherapy were significantly different among the four survivals (4); we therefore sought to determine whether com- groups (P Ͻ 0.0001), with those of groups 1 and 2 significantly bined 1p and 19q loss also predicted response to chemotherapy higher than those of groups 3 and 4 (P Ͻ 0.0001; Fig. 2). All 21 and duration of chemotherapeutic response. In addition, a num- neuroradiologically evaluable tumors with 1p loss in groups 1 ber of studies have suggested that TP53 status may be related to and 2 had visible responses to chemotherapy, whereas radio- clinical course in patients with gliomas (24, 25), directing in- graphic responses were observed in 33% of tumors in group 3 terest toward TP53 status as a predictor in our cases. On the and only 18% in group 4. Similar differences in rates of re- basis of such a priori possibilities, we postulated that cases with sponse to chemotherapy were also noted when the alternative 1p loss could be separated into those with combined 1p and 19q definition of response was used; all responders in groups 1 and loss only (combined, isolated 1p/19q loss; genetic group 1) and 2, except for one, remained progression free 6 months after those with 1p loss that either lacked 19q loss or had one of the starting chemotherapy, compared with 75% in group 3 and only other genetic alterations (“other” 1p loss; group 2) and that those 15% in group 4. Although group 3 had a higher response rate cases lacking 1p loss could also be separated into those with than group 4, the difference was not statistically significant, TP53 mutations (group 3) and those lacking TP53 mutations, most likely because of the small number of patients; the approx- which were tumors that often had alterations of the other genes imate power for this comparison using Fisher’s exact test (group 4; Fig. 2). When we used these definitions, only 1 of the was Ͻ2%. 50 patients could not be classified because chromosome 10q Classification trees were constructed to find the best data- loss was not evaluable. derived groupings of the genetic alterations with respect to

Table 2 Predictors of neuroradiological response, duration of response, and overall survival Response results from logistic regression models; duration of response and survival results from Cox regression models. Response Duration of response Survival Feature Odds ratio P Hazard ratio P Hazard ratio P Age at diagnosis Յ45 years 1.264 0.733 0.869 0.825 0.519 0.132 KPS Ն80 0.917 0.899 1.419 0.581 0.584 0.218 Enhancement at diagnosis 0.0667 0.577 8.662 0.041 3.867 0.015 Ring enhancement at diagnosis 0.036 0.004 7.576 0.080 18.024 Ͻ0.0001 Ͻ0.0001 0.115 0.017 0.180 ء bء 1p LOHa 19q LOH 3.694 0.072 0.101 0.001 0.177 0.0002 Ͻ0.0001 9.064ءء 10q LOH 0.057 0.002 PTEN alteration 0.067 0.021 23.495 0.026 18.782 Ͻ0.0001 CDKN2A deletion 0.139 0.034 3.489 0.026 7.969 Ͻ0.0001 Ͻ0.0001 11.965ءء EGFR amplification 0.139 0.034 TP53 mutation 0.307 0.170 2.518 0.253 1.717 0.300 a LOH, loss of heterozygosity. .(not estimable in this model (e.g., all patients with 1p loss of heterozygosity responded ,ء b

response. When we used radiographic response, patients were duced to adjust for variable start of radiation therapy. Likewise, not subdivided beyond their 1p loss status (misclassification similar results were obtained for the alternative definition of error rate, 3 of 37). However, when we used the alternative response. Multivariate models are not reported because of the definition of response, patients were divided into three genetic small numbers in this analysis. groups: those with 1p loss, those without 1p loss and with TP53 Analyses of Overall Survival by Molecular Genetic Pa- mutation, and those without 1p loss and without TP53 mutation rameters. For patients in group 1, the median survival time (misclassification error rate, 4 of 47). These groups correspond from diagnosis was Ͼ123 months. In contrast, patients in groups to the hypothesized groups 1 and 2 combined, group 3, and 2 and 3 had median survivals of ϳ70 months, and patients in group 4, respectively. group 4 had a median survival of only 16 months (Figs. 2 and The predictive power of each of the genetic alterations for 3). Each of the pairwise comparisons of survival among the four response was further assessed via logistic regression (Table 2). hypothesized groups was statistically significant, with the ex- In addition to 1p loss, which was a perfect predictor of response, ception of the comparison between groups 2 and 3. Identical response was negatively associated with 10q loss, PTEN muta- results were obtained when time-dependent strata were used to tion, CDKN2A deletion, and EGFR amplification, and positively adjust for the variable starting time of radiation therapy. associated with 19q loss. These relationships were also demon- In Cox models of survival, the hazard of death was signif- strated using the alternative definition of response. Multivariate icantly decreased for patients with 1p and 19q loss, and it was models were not fit because of the dominance of 1p loss in significantly increased for patients with 10q loss, PTEN alter- predicting response and the small number of patients without 1p ation, CDKN2A deletion, and EGFR amplification (Table 2). In loss. a multivariate model, selected using backward elimination, only Analyses of Duration of Response by Molecular Genetic 1p loss and CDKN2A deletion remained significant predictors at Parameters. Patients whose tumors had combined, isolated the 0.05 level. Loss of 1p predicted increased survival (hazard losses of 1p and 19q (group 1) had a significantly longer ratio, 0.13; P ϭ 0.0002), and CDKN2A deletion predicted de- duration of radiographic response to chemotherapy than patients creased survival (hazard ratio, 5.71; P ϭ 0.001). in genetic group 2 (P ϭ 0.004). They also had a significantly Addition of Nongenetic Parameters in Predicting Re- longer duration of response than patients in genetic groups 3 and sponse and Survival. Nongenetic parameters (patient age, 4(P ϭ 0.004). A significantly different duration of response KPS, neuroradiological enhancement, and ring enhancement) was not detected between patients in group 2 and patients in were included in additional analyses of response, duration of groups 3 and 4 (P ϭ 0.757). These results were also seen when response, and overall survival. For both definitions of response, we stratified additionally for the starting time of radiation ther- ring enhancement was a strong univariate predictor of lack of apy and when we used the alternative definition of response. response [odds ratio, 0.036 (P ϭ 0.004) for neuroradiological The median duration of response for group 1 was Ͼ31 months, response; odds ratio, 0.059 (P ϭ 0.002) for 6-month progres- compared with 10.6 months for group 2, 6.7 months for the 2 (of sion-free response]. In this model, therefore, ring enhancement 6) responders in group 3, and 5.0 months for the 2 (of 11) segregated with 10q loss, PTEN alteration, CDKN2A deletion, responders in group 4 (Fig. 2). and EGFR amplification as predictors of poor response to chem- In Cox models of the influence of individual genetic alter- otherapy. With regard to duration of response, radiological ations on duration of response to chemotherapy, the hazard of enhancement was highly associated with an increased hazard of progression was significantly decreased for patients with 1p and progression after response to chemotherapy (odds ratio, 4.37; 19q loss and was significantly increased for patients with 10q P ϭ 0.007). Because of the small numbers of patients, the effect loss, PTEN alteration, and CDKN2A deletion (Table 2). Similar of ring enhancement was not estimable in these models. results were obtained when time-dependent strata were intro- In univariate models of overall survival, enhancement and

Fig. 3 Kaplan-Meier curve. Group 1 patients (tumors with combined, isolated 1p and 19q loss) have good long-term overall survival, whereas group 4 patients (neither 1p loss nor TP53 mutation, often with other genetic alterations and ring enhancement) have a poor prognosis. Patients in groups 2 and 3 (see text) have intermediate prog- noses.

ring enhancement joined 10q loss, PTEN alteration, CDKN2A sponded to chemotherapy, but these responses were brief. Those deletion, EGFR amplification, and TP53 mutation as predictors in group 3 may have responded to chemotherapy, but briefly, of poorer overall survival. None of the clinical variables, in- and those in group 4 seldom responded to chemotherapies in use cluding age Ͻ45 years, was associated with differences in at present. overall survival. In a multivariate model, KPS, ring enhance- These findings suggest that molecular diagnosis may soon ment, and diffuse (nonring) enhancement joined 1p as signifi- influence the clinical management of patients with anaplastic cant predictors of survival. Ring enhancement and diffuse (non- oligodendrogliomas, particularly those in groups 1 and 4. For ring) enhancement predicted decreased survival (hazard ratios, example, group 1 patients might be treated initially with chem- 64.1 and 11.7; P Ͻ 0.0001 and P ϭ 0.001, respectively), otherapy because durable responses can be anticipated, at the whereas high KPS and 1p loss predicted increased survival same time deferring radiation therapy, which is the customary (hazard ratios, 0.15 and 0.04; P ϭ 0.0004 and P Ͻ 0.0001, postoperative treatment for malignant glioma. Indeed, in this respectively). series, 14 patients whose tumors had the genetic signature of isolated 1p and 19q loss were alive Ͼ7.4 years after diagnosis, DISCUSSION including 6 patients who never received radiation therapy. De- We previously reported that specific chromosomal losses spite a Kaplan-Meier curve that is remarkably flat for a cohort predicted chemotherapeutic response and survival in patients of malignant gliomas, 9 of the 23 group 1 patients have not yet with anaplastic oligodendrogliomas (3). The clinical utility of received radiation therapy. Given the neurotoxicities of radia- this earlier, smaller study was compromised, however, by the tion therapy, which become evident only in long-term survivors, inclusion of many patients who had received chemotherapy only an argument could thus be made to delay irradiation in patients at the time of recurrence, limiting its ability to address the issue whose anaplastic oligodendrogliomas have combined, isolated of therapeutic management at the time of initial diagnosis. In 1p/19q loss, especially if the location and size of the tumor are addition, the previous work analyzed a smaller number of clin- such that large volumes of normal brain tissue will be included ical and genetic parameters; for example, the important variable in the irradiated field. At the least, the results provide a strong “duration of response to chemotherapy” (as distinct from “time rationale for incorporating molecular analyses into prospective to tumor progression”) was not adequately addressed. The pres- trials that would evaluate whether radiation therapy can be ent results extend those prior observations by demonstrating delayed in patients with particular biological subtypes of ana- that, at the time of diagnosis, histologically defined anaplastic plastic oligodendroglioma. oligodendrogliomas can be divided into four prognostically and Patients in genetic groups 2 and 3 remain good candidates therapeutically relevant genetic groups (Fig. 2). Patients whose for chemotherapy at diagnosis and someday may have durable anaplastic oligodendrogliomas had isolated, combined allelic responses to new cytotoxic agents. At present, however, these losses of 1p and 19q (group 1) had a uniformly excellent patients must be monitored closely for recurrence and should be prognosis; patients with anaplastic oligodendrogliomas harbor- treated with radiation therapy postoperatively. Patients in ge- ing other 1p losses (group 2) or TP53 mutations (group 3) had netic group 4, on the other hand, lack meaningful responses to an intermediate prognosis, and patients whose anaplastic oligo- chemotherapy and have a particularly ominous prognosis. dendrogliomas had neither 1p loss nor TP53 mutation (group 4) Group 4 patients are highly likely to have ring enhancement on had a poor prognosis (Fig. 3). Moreover, patients in group 1 had neuroradiological examination, which was the only clinical or durable responses to chemotherapy; those in group 2 also re- radiographic feature highly predictive of aggressive behavior

and which was never observed in cases with isolated, combined ACKNOWLEDGMENTS 1p/19q allelic loss. The present results argue against the use of We thank Emi Moore at the Canadian Brain Tumor Tissue Bank initial PCV chemotherapy in this group of patients; because for expert assistance. there is little expected therapeutic gain, avoidance of PCV could both spare bone marrow reserves and provide additional time for REFERENCES other therapeutic approaches. Although patients in group 4 1. CBTRUS. 1996 Annual Report. Chicago: Central Brain Tumor Reg- should be offered radiation therapy at diagnosis, one could argue istry of the United States, 1997. that these patients are the very group who are logical candidates 2. Mason, W., Louis, D. N., and Cairncross, J. G. Chemosensitive for experimental therapies at diagnosis, as standard treatment gliomas in adults: which ones and why? J. Clin. Oncol., 15: 3423–3426, yields such poor outcomes. 1997. The genetic features noted in groups 3 and 4 are also 3. Cairncross, J. G., Ueki, K., Zlatescu, M. C., Lisle, D. K., Finkelstein, D. M., Hammond, R. R., Silver, J. S., Stark, P. C., Macdonald, D. R., characteristic of high-grade astrocytomas, such as glioblastoma Ino, Y., Ramsay, D. A., and Louis, D. N. Specific chromosomal losses (26, 27). Approximately one-third of glioblastomas have TP53 predict chemotherapeutic response and survival in patients with ana- mutations, most often in young adults, and a distinct one-third plastic oligodendrogliomas. J. Natl. Cancer Inst. (Bethesda), 90: 1473– have EGFR gene amplification, often with chromosome 10, 1479, 1998. PTEN, and CDKN2A/p16 alterations, and typically occur in 4. Smith, J. S., Perry, A., Borell, T. J., Lee, H. K., O’Fallon, J., Hosek, S. M., Kimmel, D., Yates, A., Burger, P. C., Scheithauer, B. W., and older adults (26, 28). Notably, in the present series, EGFR, Jenkins, R. B. Alterations of chromosome arms 1p and 19q as predictors PTEN, chromosome 10, and CDKN2A abnormalities clustered of survival in oligodendrogliomas, astrocytomas, and mixed oligoastro- in anaplastic oligodendrogliomas in generally older adults, and cytomas. J. Clin. Oncol., 18: 636–645, 2000. TP53 mutations were more common in tumors of younger adults 5. Ino, Y., Zlatescu, M. C., Sasaki, H., Macdonald, D. R., Stemmer- (Fig. 2). These observations suggest parallels between the two Rachamimov, A. O., Jhung, S., Ramsay, D. A., von Deimling, A., Louis, groups of non-1p-losing anaplastic oligodendrogliomas and the D. N., and Cairncross, J. G. Long patient survival and therapeutic responses in histologically disparate high grade gliomas with chromo- two well-defined molecular subtypes of glioblastoma. To ex- some 1p loss. J. Neurosurg., 92: 983–990, 2000. clude the possibility that group 3 and 4 tumors were misdiag- 6. Reifenberger, G., Kros, J. M., Burger, P. C., Louis, D. N., and nosed glioblastomas, these cases were reviewed by a third Collins, V. P. Oligodendrogliomas and oligoastrocytomas. In: P. neuropathologist, also blinded to the genetic results, and found Kleihues and W. K. Cavenee (eds.), World Health Organization Clas- to be in accord with current criteria for anaplastic oligodendro- sification of Tumours of the Central Nervous System. Lyon: IARC/ WHO, 2000. glioma (6). Indeed, some cases were histologically classic for 7. Louis, D. N., von Deimling, A., and Seizinger, B. R. A (CA)n oligodendroglioma, showing no features even suggestive of dinucleotide repeat assay for evaluating loss of allelic heterozygosity in glioblastoma (Fig. 1, right-hand panel). Thus, histologically small and archival human brain tumor specimens. Am. J. Pathol., 141: defined anaplastic oligodendrogliomas that are genotypically 777–782, 1992. similar to glioblastomas (group 4) appear clinically more akin to 8. Rubio, M-P., von Deimling, A., Yandell, D. W., Wiestler, O. D., glioblastomas. For purposes of treatment planning and prognos- Gusella, J. F., and Louis, D. N. Accumulation of wild-type p53 protein tication, therefore, genotyping may someday override the gold in human astrocytomas. Cancer Res., 53: 3465–3467, 1993. standard of histopathology, a prospect alluded to in the most 9. Du¨rr, E-M., Rollbrocker, B., Hayashi, Y., Peters, N., Meyer-Puttlitz, B., Louis, D. N., Schramm, J., Wiestler, O. D., Parsons, R., Eng, C., and recent iteration of the WHO classification for brain tumors (29). von Deimling, A. PTEN mutations in gliomas and glioneuronal gliomas. The above genetic markers clearly and objectively distin- Oncogene, 16: 2259–2264, 1998. guish between different biological types of malignant glioma. In 10. Ueki, K., Ono, Y., Henson, J. W., von Deimling, A., and Louis, the setting of a lesion histologically consistent with an oligo- D. N. CDKN2/p16 or RB alterations occur in the majority of glioblas- dendroglioma, therefore, combined but isolated 1p and 19q loss tomas and are inversely correlated. Cancer Res., 56: 150–153, 1996. constitutes an effective marker for a malignant glioma that is 11. Ono, Y., Tamiya, T., Ichikawa, T., Kunishio, K., Mtsumoto, K., Furuta, T., Ohmoto, T., Ueki, K., and Louis, D. N. Malignant astrocy- especially sensitive to therapies available at present. This sen- tomas with homozygous CDKN2/p16 gene deletions have higher Ki-67 sitivity may also pertain to chemotherapies other than PCV, as proliferation indices. J. Neuropathol. Exp. Neurol., 55: 1026–1031, well as to radiation therapy (30). From a diagnostic point of 1996. view, the ability to identify a malignant glioma that is poten- 12. Louis, D. N., Rubio, M-P., Correa, K., Gusella, J. F., and von tially curable with available therapies allows emphasis to be Deimling, A. Molecular genetics of pediatric brain stem gliomas. Ap- placed on a management strategy that maximizes tumor control plication of PCR techniques to small and archival brain tumor specimens. J. Neuropathol. Exp. Neurol., 52: 507–515, 1993. while minimizing neurotoxicity. The molecular genetic tests 13. Agresti, A. (ed.). Categorical Data Analysis. New York: Wiley, used in this analysis, although not yet widely available, should 1990. be relatively easy to establish in most pathology laboratories; 14. Breiman, L., Friedman, J. H., Olshen, R. A., and Stone, C. J. (eds.). many of the alterations can be detected by either PCR or Classification and Regression Trees. Monterey: Wadsworth, 1984. fluorescent in situ hybridization assays, with good concordance 15. Macdonald, D. R., Cascino, T. L., Schold, S. C., and Cairncross, between these techniques (22). In a recent editorial, we sug- J. G. Response criteria for Phase II studies of malignant glioma. J. Clin. gested that molecular typing of malignant gliomas could in the Oncol., 8: 1277–1280, 1990. future prove a viable means for identifying therapeutically re- 16. Cairncross, J. G., Macdonald, D. R., Pexman, J. H. W., and Ives, F. J. Steroid-induced CT scan changes in patients with recurrent malig- sponsive tumors, as well as providing a basis for placing a priori nant glioma. Neurology, 38: 724–726, 1988. nonresponsive tumors into alternative therapeutic arms (2). The 17. Watling, C. J., Lee, D. H., Macdonald, D. R., and Cairncross, J. G. rapid progress in this field over the past few years suggests that Corticosteroid-induced magnetic resonance imaging changes in patients this time has come. with recurrent malignant glioma. J. Clin. Oncol., 12: 1886–1889, 1994.

18. Yung, W. K., Prados, M. D., Yaya-Tur, R., Rosenfeld, S. S., Brada, 24. Kyritsis, A. P., Bondy, M. L., Hess, K. R., Cunninghan, J. E., Zhu, M., Friedman, H. S., Albright, R., Olson, J., Chang, S. M., O’Neill, D., Amos, C. J., Yung, W. K. A., Levin, V. A., and Bruner, J. M. A. M., Friedman, A. H., Bruner, J., Yue, N., Dugan, M., Zaknoen, S., Prognostic significance of p53 immunoreactivity in patients with gli- and Levin, V. A. Multicenter Phase II trial of temozolomide in patients oma. Clin. Cancer Res., 1: 1617–1622, 1995. with anaplastic astrocytoma or anaplastic oligoastrocytoma at first re- 25. Pollack, I. F., Hamilton, R. L., Finkelstein, S. D., Campbell, J. W., lapse. Temodal Brain Tumor Group [Published erratum in J. Clin. Martinez, A. J., Sherwin, R. N., Bozik, M. E., and Gollin, S. M. The Oncol., 17: 3693, 1999]. J. Clin. Oncol., 17: 2762–2771, 1999. relationship between TP53 mutations and overexpression of p53 and 19. Kalbfleisch, J. D., and Prentice, R. L. (eds.). The Statistical Anal- prognosis in malignant gliomas of childhood. Cancer Res., 57: 304– ysis of Failure Time Data, p. 186. New York: Wiley, 1980. 309, 1997. 20. Kraus, J. A., Koopman, J., Kaskel, P., Maintz, D., Brandner, S., 26. Louis, D. N., and Cavenee, W. K. Molecular biology of central Louis, D. N., Wiestler, O. D., and von Deimling, A. Shared allelic losses nervous system tumors. In: V. T. DeVita, S. Hellman, and S. A. on chromosomes 1p and 19q suggest a common origin of oligodendro- Rosenberg (eds.), Cancer: Principles and Practice of Oncology. Phila- glioma and oligoastrocytoma. J. Neuropathol. Exp. Neurol., 54: 91–95, delphia: Lippincott-Raven, pp. 2091–2160, 2001. 1995. 27. Louis, D. N. A molecular genetic model of astrocytoma histopa- 21. Reifenberger, J., Reifenberger, G., Liu, L., James, C. D., Wechsler, thology. Brain Pathol., 7: 755–764, 1997. W., and Collins, V. P. Molecular genetic analysis of oligodendroglial 28. Louis, D. N., and Gusella, J. F. A tiger behind many doors: multiple tumors shows preferential allelic deletions on 19q and 1p. Am. J. genetic pathways to malignant glioma. Trends Genet., 11: 412–415, Pathol., 145: 1175–1190, 1994. 1995. 22. Smith, J. S., Alderete, B., Minn, Y., Borell, T. J., Perry, A., 29. Kleihues, P., and Cavenee, W. K. (eds.). World Health Organization Mohapatra, G., Hosek, S. M., Kimmel, D., O’Fallon, J., Yates, A., Classification of Tumours of the Nervous System. Lyon: WHO/IARC, Feuerstein, B. G., Burger, P. C., Scheithauer, B. W., and Jenkins, R. B. 2000. Localization of common deletion regions on 1p and 19q in human 30. Bauman, G. S, Ino, Y., Ueki, K., Zlatescu, M. C., Fisher, B. J., gliomas and their association with histological subtype. Oncogene, 18: Macdonald, D. R., Stitt, L., Louis, D. N., and Cairncross, J. G. Allelic 4144–4152, 1999. loss of chromosome 1p and radiotherapy plus chemotherapy in patients 23. Louis, D. N. The p53 gene and protein in human brain tumors. with oligodendrogliomas. Int. J. Radiat. Oncol. Biol. Phys., 48: 825– J. Neuropathol. Exp. Neurol., 53: 11–21, 1994. 830, 2000.

Downloaded from clincancerres.aacrjournals.org on October 1, 2021. © 2001 American Association for Cancer Research. Molecular Subtypes of Anaplastic Oligodendroglioma: Implications for Patient Management at Diagnosis

Yasushi Ino, Rebecca A. Betensky, Magdalena C. Zlatescu, et al.

Clin Cancer Res 2001;7:839-845.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/7/4/839

Cited articles This article cites 23 articles, 10 of which you can access for free at: http://clincancerres.aacrjournals.org/content/7/4/839.full#ref-list-1

Citing articles This article has been cited by 38 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/7/4/839.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/7/4/839. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.