YKL-40 Is a Differential Diagnostic Marker for Histologic Subtypes of High-Grade Gliomas

2258 Vol. 11, 2258–2264, March 15, 2005 Clinical Cancer Research

Catherine L. Nutt,1 Rebecca A. Betensky,3 INTRODUCTION Melissa A. Brower,1 Tracy T. Batchelor,2 In modern neuro-oncology, tumor classification is the David N. Louis,1 and variable that most affects therapeutic decisions and prognostic 1 estimation. Among high-grade gliomas, anaplastic oligodendro- Anat O. Stemmer-Rachamimov gliomas and glioblastomas follow markedly different clinical 1 Molecular Neuro-Oncology Laboratory and Molecular Pathology Unit, courses. Anaplastic oligodendrogliomas remain the only subtype Departments of Pathology, Cancer Center, and Neurosurgical Service; 2Brain Tumor Center, Department of Neurology, Massachusetts General of high-grade glioma that commonly responds to chemotherapy Hospital and Harvard Medical School; and 3Department of Biostatistics, (1, 2) and patients with these lesions have a much more Harvard School of Public Health, Boston, Massachusetts favorable prognosis than patients with glioblastomas (3). The most widely used method of brain tumor classification is that of the WHO (3). The 2000 WHO system divides diffuse gliomas ABSTRACT into astrocytomas, oligodendrogliomas, and oligoastrocytomas. Purpose and Experimental Design: In modern neuro- Anaplastic oligodendroglioma and glioblastoma (an astrocytic oncology, no variable affects therapeutic decisions and tumor) are the most malignant forms of glial lesions in each of prognostic estimation more than tumor classification. We their two respective tumor lineages. These high-grade gliomas showed recently that class prediction models, based on gene are characterized by dense cellularity and high proliferation expression profiles, classify diagnostically challenging malig- indices as well as the presence of microvascular proliferation nant gliomas in a manner that better correlates with clinical and/or necrosis (although, unlike glioblastoma, microvascular outcome than standard pathology. In the present study, we proliferation and/or necrosis are not required for diagnosis of used immunohistochemistry to investigate YKL-40 protein anaplastic oligodendroglioma). Histologically, astrocytic tumors expression in independent sets of glioblastomas and ana- typically exhibit pleomorphic, spindle-shaped cells with copious plastic oligodendrogliomas to determine whether this single eosinophilic cytoplasm and fibrillary processes. In contrast, marker can aid classification of these high-grade gliomas. oligodendrocytic tumors are characterized by monomorphic cells Results and Conclusions: Glioblastomas show strikingly with rounded nuclei and perinuclear halos. Unfortunately, more YKL-40 expression than anaplastic oligodendroglio- malignant gliomas often display intratumoral histologic vari- mas. Only 2 of 37 glioblastomas showed completely negative ability or lack defining histologic features and, consequently, YKL-40 staining in both tumor cells and extracellular these tumors can be diagnostically challenging (4); as such, matrix, whereas 18 of 29 anaplastic oligodendrogliomas considerable interobserver variability can occur, resulting in were completely negative in non-microgemistocytic tumor limited diagnostic reproducibility (5, 6). cells and extracellular matrix. Tumor cell staining intensity To develop more objective approaches to tumor classifica- was also markedly different: 84% of glioblastomas showed tion in general, many recent investigations have focused on gene strong staining intensities of 2+ or 3+ whereas 76% of expression microarray analyses. In one such study of high-grade anaplastic oligodendrogliomas either did not stain or stained gliomas, we showed that class prediction models, based on gene at only 1+.YKL-40stainingprovidedabetterclass expression profiles, provided a powerful tool to classify distinction of glioblastoma versus anaplastic oligodendro- diagnostically challenging malignant gliomas (7). Although the glioma than glial fibrillary acidic protein, the current prediction model used in the study used expression information standard immunohistochemical marker used to distinguish from only 19 genes, many additional genes were differentially diagnostically challenging gliomas. Moreover, a combination expressed between glioblastomas and anaplastic oligodendro- of YKL-40 and glial fibrillary acidic protein immunohisto- gliomas. One gene, YKL-40, showed an average expression level chemistry afforded even greater diagnostic accuracy in in predicted glioblastomas 82-fold greater than the average anaplastic oligodendrogliomas. expression found in predicted anaplastic oligodendrogliomas. YKL-40, a secreted glycoprotein, is a member of the 18 glycosyl hydrolase family (8–10). Although the function of YKL-40 is not well characterized, it is expressed in numerous Received 8/11/04; revised 11/15/04; accepted 12/28/04. pathologic conditions of inflammation and tissue degradation Grant support: Oligo Brain Tumor Fund of the National Brain Tumor and remodeling, including rheumatoid arthritis (11), hepatic Foundation (C.L. Nutt) and NIH grant CA57683 (D.N. Louis). fibrosis (12), and inflammatory bowel disease (13). In addition, The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked high serum levels of YKL-40 have been detected in a number of advertisement in accordance with 18 U.S.C. Section 1734 solely to cancers including breast (14), colorectal (15), and ovarian (16). indicate this fact. Moreover, high levels of YKL-40 were associated with worse Requests for reprints: David N. Louis, Molecular Pathology Unit, prognosis in all of these studies (14–16). Taken together, these CNY7, Massachusetts General Hospital, 149 13th Street, Charlestown, MA 02129. Phone: 617-726-5690; Fax: 617-726-5079; E-mail: dlouis@ findings suggest potential roles for YKL-40 in tumor invasion partners.org. and angiogenesis. In fact, YKL-40 can facilitate migration of D2005 American Association for Cancer Research. endothelial cells and promote the formation of branching tubules

(17, 18). In addition, YKL-40 has been shown to act as a growth scale of 0 to 3+, indicating the number of positive cells relative factor for connective tissue (19, 20). to all reactive astrocytes in the microscopic field (0, none; High YKL-40 serum levels have also been shown in 1+, few; 2+, some; 3+, many). glioblastomas; glioblastomas were found to have 3- to 62-fold Statistical Analysis. Crude power calculations indicated elevation of YKL-40 levels over normal brain (21). Furthermore, that we had f80% power to detect a difference in median there is evidence that YKL-40 levels correlate with grade in percentage of positive cells, staining intensity, and extracellular astrocytic gliomas (21). One recent study hypothesized that staining intensity of 0.85 between glioblastomas and anaplastic elevated levels of YKL-40 in glioblastomas might be due to the oligodendrogliomas. This was based on two-sided 0.017 level presence of increased numbers of macrophages (22) as the Wilcoxon Rank Sum tests (corrected for the three comparisons) expression of YKL-40 in macrophages has been well docu- and assumed an SD of 1 unit (as seen in our data). The exact mented (10, 23–26). Because there have been no in situ studies Wilcoxon Rank Sum test was used to compare medians of to date on YKL-40 expression in gliomas, the cellular percentage of positive cells, staining intensity, and extracellular localization of YKL-40 expression remains unknown. In the staining intensity among histologic subtypes; estimates and present study, we have used immunohistochemistry to investi- nonparametric 95% confidence intervals (95% CI) are reported gate YKL-40 protein expression in high-grade gliomas and to for differences in these immunohistochemistry measures (27). determine whether this single marker can augment the histologic McNemar’s test and the Wilcoxon Signed Rank test were used classification of high-grade gliomas. We have specifically for comparisons of YKL-40 and GFAP within the same chosen to examine YKL-40 expression in anaplastic oligoden- individuals; exact binomial 95% CIs were computed for drogliomas and glioblastomas as they comprise the highest-grade probabilities of interest. A k-mean clustering algorithm (28) tumors of two distinct glial tumor lineages. was used to cluster the subjects based on YKL-40 and GFAP staining (k = 2). We elected to use unsupervised k-mean MATERIALS AND METHODS clustering to avoid making assumptions about how to model the ordinal immunohistochemistry measures in a logistic regression Tissue Samples. Tissue samples were collected from and to avoid the problems of sparse data that would result. Massachusetts General Hospital (Boston, MA) under the Histologic diagnoses were not used to train the clustering model approval of the Massachusetts General Hospital Institutional and, therefore, the k-mean algorithm was not biased by these Review Board. A total of 81 malignant gliomas were used in the labels; there is, however, additional variability induced due to the study: 37 glioblastomas, 29 anaplastic oligodendrogliomas, and estimation of the k-mean clusters. The histologic diagnosis was 15 anaplastic oligoastrocytomas. In initial experiments, a set of then assessed in the resultant two clusters. A Bonferroni 19 glioblastomas and 18 anaplastic oligodendrogliomas were correction for multiple comparisons was applied due to the randomly selected for YKL-40 and glial fibrillary acidic protein three measurements of YKL-40 (i.e., intensity, percentage of (GFAP) immunohistochemistry. Subsequently, an additional positive cells, and extracellular staining intensity). Adjustments 18 glioblastomas and 11 anaplastic oligodendrogliomas were were not made for analyses within histologic subtypes as they randomly selected and stained with YKL-40 when the were each of inherent scientific interest. All tests were two sided 15 anaplastic oligoastrocytomas were added to the study. and the significance level was taken to be 0.05. In addition, specifically for the evaluation of reactive astrocytes, serial sections of regions with reactive gliosis in 16 of the above- mentioned gliomas and five pathologic brain samples that did RESULTS AND DISCUSSION not contain malignant primary brain tumor (e.g., metastases and YKL-40 Immunohistochemistry Distinguishes Anaplas- vascular disease) were stained with YKL-40 and GFAP. tic Oligodendrogliomas from Glioblastomas. In a recent Immunohistochemistry. YKL-40 immunohistochemistry gene expression-based microarray analysis, we identified was done on formalin-fixed, paraffin-embedded sections with a large number of genes differentially expressed between a polyclonal anti-YKL-40 antibody (Quidel Corporation, San glioblastomas and anaplastic oligodendrogliomas (7). Class Diego, CA). Antigen retrieval was achieved with a 5-minute prediction models, based on gene expression profiles, were incubation in 0.05% saponin/1% bovine serum albumin/PBS used to classify diagnostically challenging malignant gliomas solution. Slides were incubated with a 1:400 dilution of primary and, utilizing this data, YKL-40 showed an average expression antibody overnight at 4jC and visualized with an avidin-biotin level in predicted glioblastomas 82-fold greater than the average complex (Vectastatin Elite ABC kit; Vector Laboratories, expression found in predicted anaplastic oligodendrogliomas. Burlingame, CA) in the presence of 3,3V-diaminobenzidine Among these microarray samples, 22 of 35 predicted glioblas- tetrahydrochloride. GFAP immunohistochemistry was done tomas had YKL-40 RNA expression levels above the back- in the absence of antigen retrieval with a monoclonal anti-GFAP ground of the assay, whereas only 1 of 15 predicted anaplastic antibody (1:1,000 dilution; DAKO, Carpinteria, CA). oligodendrogliomas had YKL-40 RNA expression above For evaluation of tumor samples, staining for YKL-40 and background. To determine if differential YKL-40 RNA expres- GFAP was scored semiquantitatively for both cell number and sion correlated with differential protein expression, we first did staining intensity. The number of positive cells was given as 0%, YKL-40 immunohistochemistry on six samples included in the <5%, 5% to 25%, 25% to 75%, and 75% to 100%. Signal original microarray study. Three glioblastomas that were intensity was graded on a scale consisting of 0 to 3+. When positive for YKL-40 RNA showed YKL-40 staining by specifically investigating reactive astrocytes, the same signal immunohistochemistry whereas no positive YKL-40 staining intensity scale of 0 to 3+ was used to score only the reactive was seen in the three anaplastic oligodendrogliomas that did not astrocytes and the cell number was also evaluated on a graded express YKL-40 RNA (Fig. 1A and C). Thus, these preliminary

Fig. 1 Examples of the range of YKL-40 immunopositivity within histologic subtypes of high-grade glioma. A, positive YKL-40 staining in a glioblastoma shown previously to express YKL-40 RNA (7). B, higher magnification of a second glioblastoma exhibiting high levels of YKL-40 immunopositivity. C, negative YKL- 40 staining in an anaplastic oligodendroglioma shown previously not to express YKL-40 RNA (7). D, region of a diagnosed glioblastoma with histology more similar to an anaplastic oligodendroglioma exhibiting almost no YKL-40 immunopositivity. E, anaplastic oligodendroglioma representative of positive YKL-40 staining. F, anaplastic oligoastrocytoma displaying stronger YKL-40 staining in astrocytic cells. G, normal brain was negative for YKL-40 protein by immunohistochemistry.

results suggested that YKL-40 immunohistochemical staining In order to determine if YKL-40 immunohistochemistry correlated with our YKL-40 RNA expression data. could be used to distinguish histologic subtypes of high-grade Next, independent sets of glioblastomas (n = 37) and gliomas, we did statistical analyses on the semiquantitatively anaplastic oligodendrogliomas (n = 29) were immunostained scored data. The percentages of positive cells were significantly for YKL-40. In addition to tumor cells, positive YKL-40 higher in glioblastomas as compared with anaplastic oligoden- staining was also detected in blood vessels, extracellular matrix drogliomas (P < 0.001; median location difference, 2.0; 95% CI, and, to a lesser degree, in a subset of reactive astrocytes. Fifty- 1.0-2.0). Furthermore, staining intensities were also significantly four percent of glioblastomas showed positive YKL-40 staining higher in glioblastomas (P < 0.001; median location difference, in at least 25% of tumor cells (Fig. 1A and B; Table 1). 1.0; 95% CI, 1.0-2.0). The extracellular staining intensities In contrast, 69% of anaplastic oligodendrogliomas were in glioblastomas were not significantly different from those positive for YKL-40 in less than 5% of tumor cells (Fig. 1C; of anaplastic oligodendrogliomas (P = 0.186). Table 1). An even greater difference was seen in tumor cell High-grade astrocytic gliomas are graded as either anaplastic staining intensity: 84% of glioblastomas showed strong staining astrocytoma (WHO grade III) or glioblastoma (WHO grade IV). intensities of 2+ or 3+ whereas 76% of anaplastic oligoden- In contrast, anaplastic oligodendroglioma (WHO grade III) is the drogliomas either did not stain at all (62%) or stained at an highest grade designation given to oligodendroglial tumors. It is intensity of 1+ (14%). Furthermore, when positive staining was therefore possible that high-grade oligodendroglioma could found in oligodendrogliomas, it was most often in micro- encompass a wider range of malignancy when compared with gemistocytes and/or endothelial cells (Fig. 1E and 2C). Only either anaplastic astrocytoma or glioblastoma alone. Because 2 of 37 glioblastomas showed completely negative YKL-40 YKL-40 expression has been shown to correlate with tumor grade staining in both tumor cells and extracellular matrix. In in astrocytic tumors (21, 22), it was possible that the lower levels contrast, 18 of 29 anaplastic oligodendrogliomas were of YKL-40 expression found in our anaplastic oligodendroglioma completely negative in both non-microgemistocytic tumor cells sample set were associated exclusively with a lower grade as and extracellular matrix. Normal brain was also negative for compared with the glioblastomas. We have two lines of evidence YKL-40 staining (Fig. 1G). that suggest this is not the case. First, in a number of tumors that

displayed regional histologic heterogeneity, YKL-40 staining oligoastrocytomas, 1.0; 95% CI, 0.0-2.0; between anaplastic seemed to correlate with histology. For example, one diagnosed oligoastrocytomas and anaplastic oligodendrogliomas, 0.51; 95% glioblastoma displayed 3+ intensity staining but only in 5% of the CI, 0.0-1.0). If decreased YKL-40 expression in anaplastic tumor cells. On histologic review, a large region of this particular oligodendrogliomas was due exclusively to differences in tumor tumor section exhibited histologic features more characteristic of grade, anaplastic oligoastrocytomas would not be expected an anaplastic oligodendroglioma in which YKL-40 staining was to display expression levels different from those seen in almost completely absent (Fig. 1D). Even within more diffuse anaplastic oligodendrogliomas. Assuming a uniform distribution histologic regions of oligoastrocytomas, astrocytic cells stained of malignancy within the ‘‘anaplastic’’ grade in both the more intensely than the rounded cells more typical of an oligodendroglioma and oligoastrocytoma sample sets, the finding oligodendroglioma (Fig. 1F). Second, to examine this issue from that YKL-40 expression levels in anaplastic oligoastrocytomas another perspective, we did YKL-40 immunohistochemical are intermediate to those of glioblastomas and anaplastic staining on 15 samples of anaplastic oligoastrocytoma (Fig.1F), oligodendrogliomas suggests that decreased levels of YKL-40 a high-grade malignant lesion exhibiting histologic features of staining in anaplastic oligodendrogliomas can be attributed to both astroglial and oligodendroglial differentiation; as for biological differences among these tumors independent of grade. oligodendrogliomas, the grade III ‘‘anaplastic’’ designation is At first glance, in light of the suggested roles for YKL-40 the highest given to oligoastrocytomas. The percent positive cells in cancer, the differential expression of YKL-40 between and staining intensity were significantly different between glioblastomas and anaplastic oligodendrogliomas seems counter- glioblastomas, anaplastic oligodendrogliomas, and anaplastic intuitive as invasion and angiogenesis can be seen in both oligoastrocytomas (P < 0.001). For both percent positive cells glioblastomas and anaplastic oligodendrogliomas. YKL-40 and staining intensity, anaplastic oligoastrocytomas gave results expression, however, has been correlated with specific signaling intermediate to glioblastomas and anaplastic oligodendrogliomas pathways. For example, YKL-40 is expressed selectively in (median location difference between glioblastomas and anaplastic murine mammary tumors initiated by neu/ras oncogenes (29) and

Fig. 2 Comparison of YKL-40 and GFAP immunohistochemistry in high-grade gliomas. A and B, serial sections of a glioblastoma demonstrating similar staining patterns with YKL-40 (A) and GFAP (B). C and D, serial sections of an anaplastic oligodendroglioma exhibiting less YKL-40 immunopositivity (C) as compared with staining with GFAP (D). E and F, serial sections of an invasive tumor edge demonstrating the detection of fewer reactive astrocytes by YKL-40 (E) as compared with GFAP (F).

Table 1 Summary of YKL-40 immunohistochemistry results for also significantly different (P = 0.012) but GFAP staining glioblastoma, anaplastic oligodendroglioma, intensity was not significantly different (P = 0.255). and anaplastic oligoastrocytoma Although both YKL-40 and GFAP were effective at (A) Number of cells staining positive for YKL-40 distinguishing glioblastomas from anaplastic oligodendrogliomas, YKL-40 staining seemed to be a stronger marker. When a GBM (n = 37) AO (n = 29) AOA (n = 15) Percentage of k-mean clustering algorithm was applied to the data set and positive cells n % n % n % asked to find two clusters, an algorithm using the GFAP data 0 4 11 18 62 4 27 correctly identified 17 of 19 glioblastomas and 9 of 18 anaplastic 0-5 5 14 2 7 3 20 oligodendrogliomas. In comparison, an algorithm utilizing the 5-25 8 22 5 17 3 20 YKL-40 data correctly identified 19 of 19 glioblastomas and 10 25-75 10 27 4 14 5 33 75-100 10 27 0 0 0 0 of 18 anaplastic oligodendrogliomas. Interestingly, when both the YKL-40 and the GFAP data were used, the algorithm (B) Staining intensity of cells positive for YKL-40 correctly identified 18 of 19 glioblastomas and 15 of GBM (n = 37) AO (n = 29) AOA (n = 15) 18 anaplastic oligodendrogliomas. Therefore, although YKL-40 Staining intensity n % n % n % provided a better class distinction than GFAP when distinguishing histologic subtypes of high-grade gliomas, using a 0 4 11 18 62 4 27 combination of YKL-40 and GFAP immunostaining afforded 1+ 2 5 4 14 7 47 2+ 17 46 7 24 3 20 even greater diagnostic accuracy with respect to identification of 3+ 14 38 0 0 1 7 anaplastic oligodendrogliomas. To determine which staining characteristics of YKL-40 (C) Staining intensity of extracellular matix for YKL-40 were most responsible for improved diagnostic accuracy over GBM (n = 37) AO (n = 29) AOA (n = 15) GFAP staining, we compared YKL-40 and GFAP staining Staining intensity n % n % n % within single histologic subtypes. Within glioblastomas, YKL- 025682586106740 staining was not significantly different from GFAP staining 1+ 8 22 4 14 4 27 when comparing the percent positive cells (P = 1.0) nor 2+ 4 11 0 0 1 7 cellular staining intensity (P = 1.0), suggesting that these two 3+ 0 0 0 0 0 0 markers provide equivalent information for glioblastomas; both Abbreviations: GBM, glioblastoma; AO, anaplastic oligodendro- markers stained similar percentages of glioblastoma cells at glioma; AOA, anaplastic oligoastrocytoma. comparable intensities. In contrast, the percent positive cells was significantly different between YKL-40 and GFAP staining within anaplastic oligodendrogliomas (P = 0.003); phosphorylation of both extracellular signal-regulated kinases 1/2 GFAP stained a greater percentage of anaplastic oligoden- and AKT occurs in a dose- and time-dependent fashion on droglioma cells at much higher intensities (73% of samples addition of YKL-40 in fibroblastic cell lines (20). The RAS and had a YKL-40 staining intensity of 0 or 1+ versus 78% AKT signaling pathways cooperate to induce glioblastoma formation in mouse models (30) and activation of the RAS- Table 2 Summary of YKL-40 and GFAP comparative immunohisto- mitogen-activated protein kinase and phosphatidylinositol-3- chemistry results for glioblastoma and kinase-AKT signaling pathways is very common in human anaplastic oligodendroglioma glioblastomas (31). Although an intriguing possibility, whether (A) Number of cells staining positive for YKL-40 and GFAP the differential expression of YKL-40 between glioblastomas and anaplastic oligodendrogliomas can be attributed to differences in YKL-40 GFAP these signaling pathways among high-grade glial tumors will GBM AO GBM AO require further investigation. (n = 19) (n = 18) (n = 19) (n = 18) Percentage of YKL-40 Is a Stronger Marker than Glial Fibrillary positive cells n % n % n % n % Acidic Protein for Distinguishing Histologic Subtypes of 0 0 0105615422 High-Grade Glioma. In current diagnostic pathology, immu- 0-5 2110 00 00 0 nohistochemistry detects GFAP, which is most often found in 5-25 4 21 5 28 4 21 9 50 astrocytic gliomas. However, GFAP is not restricted exclusively 25-75 7 37 3 17 8 42 4 22 to astrocytic lesions and anaplastic oligodendrogliomas may 75-100 6 32 0 0 6 32 1 6 show positive GFAP staining (32). In order to determine if (B) Staining intensity of cells positive for YKL-40 and GFAP YKL-40 might provide a better marker for distinguishing histologic subtypes of high-grade glioma, we compared YKL- YKL-40 GFAP 40 and GFAP staining in a subset of 19 glioblastomas and 18 GBM AO GBM AO anaplastic oligodendrogliomas (Fig. 2A-D; Table 2). As in our (n = 19) (n = 18) (n = 19) (n = 18) initial analysis above, YKL-40 staining was able to distinguish Staining intensity n % n % n % n % between glioblastomas and anaplastic oligodendrogliomas in 000105615422 this subset of tumors: the percent positive cells (P < 0.001) as 1+ 003171500 well as the staining intensity (P < 0.001) were significantly 2+ 9 47 5 28 0 0 0 0 different. In comparison, the percent GFAP-positive cells was 3+ 10 53 0 0 17 89 14 78

Table 3 Summary of YKL-40 and GFAP immunohistochemistry Moreover, the findings that regional histologic differences within results for reactive astrocytes (n = 21) tumors correlated with YKL-40 staining and that YKL-40 YKL-40 GFAP expression levels in anaplastic oligoastrocytomas were interme- Staining Cell Staining Cell diate to those of glioblastomas and anaplastic oligodendroglio- intensity number* intensity number* mas suggest that decreased levels of YKL-40 staining in Score n % n % n % n % anaplastic oligodendrogliomas can be attributed to biological differences among these tumors that are not simply reflective of 0 6 29 6 29 0 0 0 0 1+ 4 19 6 29 0 0 1 5 malignancy grade. When compared with GFAP, the current 2+ 9 43 7 33 1 5 5 24 standard clinical marker for distinguishing diagnostically 3+ 2 10 2 10 20 95 15 71 challenging high-grade gliomas, YKL-40 staining provided a *Cell number: 0, none; 1+, few; 2+, some; 3+, many—as compared better class distinction. Moreover, using a combination of YKL- with the total number of reactive astrocytes. 40 and GFAP staining afforded even greater diagnostic accuracy. In this study, we chose to focus on anaplastic oligodendrogliomas and glioblastomas, the highest-grade lesions of two distinct of samples displaying a GFAP staining intensity of 3+). Thus, tumor lineages. In the future, it will be interesting to examine the greater staining differential afforded by YKL-40 in YKL-40 with respect to tumor grade within a single tumor comparison with GFAP could most likely be ascribed directly lineage, such as comparisons specifically among glioblastomas to significantly lower levels of YKL-40 staining in anaplastic and anaplastic astrocytomas, as well as in other types of central oligodendrogliomas. nervous system tumors. One of the complicating characteristics of GFAP staining is its ability to stain reactive astrocytes strongly; immunohisto- chemically positive reactive astrocytes can falsely exaggerate the ACKNOWLEDGMENTS impression of a positively staining glial tumor. In order to We thank Jennifer Roy for help with tissue acquisition; Molly Dorfman, Whitney Nugent, and Loc Pham for clinical data; and Kevin determine if decreased staining of reactive astrocytes might be Park for photography. one of the explanations for YKL-40 being a better marker compared with GFAP, we stained regions of reactive gliosis in 16 high-grade gliomas and 5 additional pathologic brain samples REFERENCES without glioma and scored these samples for positive staining of 1. Cairncross JG, Macdonald DR. Successful chemotherapy for reactive astrocytes (Fig. 2E and F; Table 3). When scores were malignant oligodendroglioma. Ann Neurol 1988;23:360–4. clustered into two groups (0 and 1+ versus 2+ and 3+), the YKL- 2. Cairncross JG, Ueki K, Zlatescu MC, et al. Specific chromosomal 40 and GFAP distributions of positive cells were significantly losses predict chemotherapeutic response and survival in patients with anaplastic oligodendrogliomas. J Natl Cancer Inst 1998;90:1473–9. different (P = 0.003); GFAP immunohistochemistry detected 3. Kleihues P, Cavenee WK. WHO Classification of Tumours of the greater numbers of reactive astrocytes in 11 of 21 samples Nervous System. Lyon: WHO/IARC; 2000. (95% CI, 0.30-0.74) and stained the cells more intensely in 7 of 4. Louis DN, Holland EC, Cairncross JG. Glioma classification: a 21 samples (95% CI, 0.15-0.57) than did YKL-40. molecular reappraisal. Am J Pathol 2001;159:779–86. Anaplastic oligodendrogliomas are also associated with 5. Coons SW, Johnson PC, Scheithauer BW, Yates AJ, Pearl DK. 1p/19q loss of heterozygosity; these genetic markers also Improving diagnostic accuracy and interobserver concordance in the predict chemotherapeutic response and better prognosis in these classification and grading of primary gliomas. Cancer 1997;79:1381–93. lesions (33, 34). Although 1p/19q loss of heterozygosity status 6. Giannini C, Scheithauer BW, Weaver AL, et al. Oligodendrogliomas: Reproducibility and prognostic value of histologic diagnosis and grading. was not available for the samples included in this study, we J Neuropathol Exp Neurol 2001;60:248–62. investigated 1p status in the anaplastic oligodendrogliomas 7. Nutt CL, Mani DR, Betensky RA, et al. Gene expression-based used in our initial microarray analysis (7). Although the classification of malignant gliomas correlates better with survival than average YKL-40 mRNA expression was slightly lower in 1p histological classification. Cancer Res 2003;63:1602–7. loss of heterozygosity tumors as compared with 1p intact cases, 8. Hakala BE, White C, Recklies AD. Human cartilage gp-39, a major these groups were not significantly different.4 Thus, it seemed secretory product of articular chondrocytes and synovial cells, is a mammalian member of a chitinase protein family. J Biol Chem 1993; that YKL-40 provided diagnostic information in addition to 268:25803–10. that afforded by 1p/19q status. 9. Johansen JS, Jensen HS, Price PA. A new biochemical marker for joint injury. Analysis of YKL-40 in serum and synovial fluid. Br J SUMMARY Rheumatol 1993;32:949–55. In the present study, we used immunohistochemistry to 10. Renkema GH, Boot RG, Au FL, et al. Chitotriosidase, a chitinase, and the 39-kDa human cartilage glycoprotein, a chitin-binding lectin, are investigate YKL-40 protein expression in high-grade gliomas homologues of family 18 glycosyl hydrolases secreted by human and to determine whether this single marker can augment the macrophages. Eur J Biochem 1998;251:504–9. histologic classification of high-grade gliomas. YKL-40 immu- 11. Johansen JS, Stoltenberg M, Hansen M, et al. Serum YKL-40 nohistochemistry was able to distinguish anaplastic oligoden- concentrations in patients with rheumatoid arthritis: relation to disease drogliomas from glioblastomas in a highly significant manner. activity. Rheumatology (Oxford) 1999;38:618–26. 12. Johansen JS, Christoffersen P, Moller S, et al. Serum YKL-40 is increased in patients with hepatic fibrosis. J Hepatol 2000;32:911–20. 13. Vind I, Johansen JS, Price PA, Munkholm P. Serum YKL-40, a 4 C.L. Nutt and D.N. Louis. Uncorrected P value of 0.726, unpublished potential new marker of disease activity in patients with inflammatory data. bowel disease. Scand J Gastroenterol 2003;38:599–605.

14. Johansen JS, Cintin C, Jorgensen M, Kamby C, Price PA. Serum the gene for human cartilage gp-39 (CHI3L1), a member of the chitinase YKL-40: a new potential marker of prognosis and location of metastases protein family and marker for late stages of macrophage differentiation. of patients with recurrent breast cancer. Eur J Cancer 1995;31:1437–42. Genomics 1997;43:221–5. 15. Cintin C, Johansen JS, Christensen IJ, Price PA, Sorensen S, 25. Kirkpatrick RB, Emery JG, Connor JR, Dodds R, Lysko PG, Nielsen HJ. Serum YKL-40 and colorectal cancer. Br J Cancer 1999;79: Rosenberg M. Induction and expression of human cartilage glycoprotein 1494–9. 39 in rheumatoid inflammatory and peripheral blood monocyte-derived 16. Dehn H, Hogdall EV, Johansen JS, et al. Plasma YKL-40, as a macrophages. Exp Cell Res 1997;237:46–54. prognostic tumor marker in recurrent ovarian cancer. Acta Obstet 26. Boot RG, van Achterberg TA, van Aken BE, et al. Strong induction Gynecol Scand 2003;82:287–93. of members of the chitinase family of proteins in atherosclerosis: 17. Malinda KM, Ponce L, Kleinman HK, Shackelton LM, Millis AJ. chitotriosidase and human cartilage gp-39 expressed in lesion macro- Gp38k, a protein synthesized by vascular smooth muscle cells, stimulates phages. Arterioscler Thromb Vasc Biol 1999;19:687–94. directional migration of human umbilical vein endothelial cells. Exp Cell 27. Hollander M, Wolfe DA. Nonparametric statistical inference. New Res 1999;250:168–73. York: John Wiley & Sons; 1973. p. 68–75. 18. Nishikawa KC, Millis AJ. gp38k (CHI3L1) is a novel adhesion and 28. Hartigan JA, Wong MA. A K-means clustering algorithm. Appl Stat migration factor for vascular cells. Exp Cell Res 2003;287:79–87. 1979;28:100–8. 19. De Ceuninck F, Gaufillier S, Bonnaud A, Sabatini M, Lesur C, 29. Morrison BW, Leder P. neu and ras initiate murine mammary tumors Pastoureau P. YKL-40 (cartilage gp-39) induces proliferative events in that share genetic markers generally absent in c-myc and int-2-initiated cultured chondrocytes and synoviocytes and increases glycosaminogly- tumors. Oncogene 1994;9:3417–26. can synthesis in chondrocytes. Biochem Biophys Res Commun 2001; 30. Holland EC, Celestino J, Dai C, Schaefer L, Sawaya RE, Fuller GN. 285:926–31. Combined activation of Ras and Akt in neural progenitors induces 20. Recklies AD, White C, Ling H. The chitinase 3-like protein human glioblastoma formation in mice. Nat Genet 2000;25:55–7. cartilage glycoprotein 39 (HC-gp39) stimulates proliferation of human 31. Zhu Y, Parada LF. The molecular and genetic basis of neurological connective-tissue cells and activates both extracellular signal-regulated tumours. Nat Rev Cancer 2002;2:616–26. kinase- and protein kinase B-mediated signalling pathways. Biochem 32. Herpers MJHM, Budka H. Glial fibrillary acidic protein (GFAP) J 2002;365:119–26. in oligodendroglial tumors: gliofibrillary oligodendroglioma and 21. Tanwar MK, Gilbert MR, Holland EC. Gene expression microarray transitional oligo-astrocytoma as subtypes of oligodendroglioma. Acta analysis reveals YKL-40 to be a potential serum marker for malignant Neuropathol (Berl) 1984;64:265–72. character in human glioma. Cancer Res 2002;62:4364–8. 33. Cairncross JG, Ueki K, Zlatescu MC, et al. Specific chromo- 22. Shostak K, Labunskyy V, Dmitrenko V, et al. HC gp-39 gene is up- somal losses predict chemotherapeutic response and survival in regulated in glioblastomas. Cancer Lett 2003;198:203–10. patients with anaplastic oligodendrogliomas. J Natl Cancer Inst 1998; 23. Krause SW, Rehli M, Kreutz M, Schwarzfischer L, Paulauskis JD, 90:1473–9. Andreesen R. Differential screening identifies genetic markers of 34. Ino Y, Betensky RA, Zlatescu MC, et al. Molecular subtypes of monocyte to macrophage maturation. J Leukoc Biol 1996;60:540–5. anaplastic oligodendroglioma: implications for patient management at 24. Rehli M, Krause SW, Andreesen R. Molecular characterization of diagnosis. Clin Cancer Res 2001;7:839–45.

Downloaded from clincancerres.aacrjournals.org on September 25, 2021. © 2005 American Association for Cancer Research. YKL-40 Is a Differential Diagnostic Marker for Histologic Subtypes of High-Grade Gliomas

Catherine L. Nutt, Rebecca A. Betensky, Melissa A. Brower, et al.

Clin Cancer Res 2005;11:2258-2264.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/11/6/2258

Cited articles This article cites 29 articles, 5 of which you can access for free at: http://clincancerres.aacrjournals.org/content/11/6/2258.full#ref-list-1

Citing articles This article has been cited by 9 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/11/6/2258.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/11/6/2258. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.