Morphological Analysis of CDC2 and Glycogen Synthase Kinase 3β Phosphorylation as Markers of G2 → M Transition in Glioma

G2 → M transition is a strategic target for glioma chemotherapy. Key players in G2 → M transition include CDC2 and glycogen synthase kinase 3β (GSK3β), which are highly regulated by posttranslational phosphorylation. This report is a morphological analysis of CDC2 and GSK3β phosphorylation using immunohistochemistry in gliomas with different biological properties. GBM showed a 2.8-fold and 5.6-fold increase in number of cells positive for pThr161CDC2 and a 4.2- and 6.9-fold increase in number of cells positive for pTyr15CDC2 relative to oligodendroglioma and ependymoma, respectively. Elevated labeling for inhibited phospho-CDC2 (pTyr15CDC) correlates with elevated levels of phosphorylated glycogen synthase kinase 3β (GSK3β). 71% of the GBM cases showed intermediate to high intensity staining for pSer9SGK3β 53% of oligodendroglioma, and 73% of ependymoma showed low intensity staining. CDC2 gene amplification correlates with increased survival in glioblastoma multiforme (GBM) and astrocytoma WHO grades II-III, but not in oligodendroglioma WHO grades II-III.


Introduction
Cell cycle progression is partly regulated by a group of proteins whose expression is cyclical during the cell cycle. These proteins, known as cyclins, exert their function on the cell cycle partly through-regulating the activity of their binding partners, the cyclin dependent kinases (CDKs)/cell division control (CDC) proteins [1]. Cyclin/CDC complexes, in concert with other proteins, control the cell cycle by regulation of multiple cell cycle checkpoints. Although many of the molecular pathways activated in gliomas have been implicated in the G1 → S phase transition of the cell cycle [2], the role of other cell cycle checkpoints is less clear. Furthermore, temozolomide-(TMZ-) induced cell cycle arrest occurs at the G2 → M transition in glioma cell lines [3]. Repair of TMZ-induced DNA damage is critical for TMZ toxicity, and thus the G2 → M transition is a target for chemotherapy. A central player in the G2 → M phase transition is CDC2 (also known as CDK1) [4]. CDC2 is overexpressed in gliomas, and inhibition of CDC2 expression by transfection of small interfering RNA targeted to CDC2 inhibits glioma growth in vivo and in vitro [5].
CDC2 associates with cyclin-B and cyclin-A. This complex can be either positively or negatively regulated by the state of CDC2 phosphorylation. A model of CDC2 activity is shown in Figure 1. Phosphorylation of a conserved threonine (Thr161) in the T-loop of CDC2 by the CDK Activating Kinase (CAK, also known as CDK7) is required for activation of the cyclin-B/CDC2 complex [4]. Conversely, phosphorylation of CDC2 at threonine 14 (Thr14) and tyrosine 15 (Tyr15) by the Wee1/Mik1 family of protein kinases inhibits the cyclin-B/CDC2 complex [6,7]. Adding to this complexity, Kang and colleagues demonstrated that CDC25, a promitotic phosphatase that dephosphorylates CDC2 at Tyr15 [8], is targeted for ubiquitin-mediated proteolysis by GSK3β-mediated phosphorylation, and that accumulation of CDC25 is highly associated with GSK3β inactivation in human tumors [8]. GSK3β, a target of the EGF signaling pathway [9], is an ubiquitous kinase with various functions, including but not limited to, initiation of β-catenin [10] and cyclin D1 proteolysis [11] and regulation of cell cycle arrest [12]. These and other findings suggest that GSK3β may act as a tumor suppressor protein in these setting. Prior studies have shown both CDC2 and GSK3β to regulate growth and invasion of cell lines derived from GBM [5,13]. Evaluation of CDC2 and GSK3β activation states in infiltrative primary glial tumors of other lineages has not been thoroughly evaluated. As these proteins' activities are highly regulated through post-translational phosphorylation, a morphological evaluation of their activation states using immunohistochemistry to phospho-specific forms of CDC2 and GSK3β was performed.  negative controls. Immunohistochemical optimization was performed by serial titration of antibodies on formalinfixed, paraffin-embedded tissue sections obtained from the UCSF Pathology Archives. Anti-phospho-Thr161CDC2 (pThr161CDC2) shows optimal staining at 1 : 50 dilution in invasive ductal carcinoma of the breast. Antiphospho-Tyr15-CDC2 (pTyr15CDC2) shows optimal staining at 1 : 25 dilution in colorectal adenocarcinoma. Antiphospho-Ser9GSK3β ()pSer9GSK3β) shows optimal staining at 1 : 50 in colorectal adenocarcinoma. EGFR, pTyr15CDC2, pThr161CDC2, showed no labeling in normal white matter and cortex. EGFR immunolabelling was detected in the syncytiotrophoblastic cells of placenta, which served as our positive control. Focal positive nuclear labeling of pThr161CDC2 and pTyr15CDC2 was also noted in carcinoma of breast metastatic to brain and in normal lymph nodes.

Quantification of CDC2
Labeling. All immunohistochemical labeling was scored by counting all the cells of interest in three high power magnification fields. Only the cells with similar or higher staining intensity as the control tissue were counted as positive. Occasional positive staining was noted sporadically in endothelial cells. Endothelial cells were not included in the CDC2 quantifications. The total area counted constituted approximately 10% of each tissue core. The CDC2 cell labeling was performed according to previously published counting methods with some modifications [14]. Briefly, cores with no CDC2 labeling are score = 0. Score = 1 shows 1-10 CDC2 positive cells per three high power fields, score = 2 shows 11-20 CDC2 cells per three high power fields, and score = 3 shows 21 or more CDC2 positive cells per three high power fields. To determine the mean CDC2 labeling in the patient, the total number of cells counted were averaged over all of the cores from the same patient. The final patient score was designated as low (scores 0 and 1), intermediate (score 2), or high (score 3). As the tissue cores represent a fraction of the original tissue, a separate category of "negative" was considered inaccurate. Hence, Score = 0 and Score = 1 were grouped together in the low labeling category. Percent positive cells were determined as follows: (#CDC2+ cells/all tumor cells quantified by hematoxylin counterstain)×100.

Quantification of PSerGSK3β and EGFR.
The scoring method used in evaluation of the pSerGSK3β and EGFR immunohistochemical stains was adopted from Atkins et al. [15]. Briefly, tumors with less than 10% labeling in tumor cells were considered negative (score 0). In all others, a minimum of 10% staining of tumor cells were required. Tumors with minimal labeling intensity were scored as LOW (score 1). Intermediate labeling was designated to cores if moderate labeling was detectable in more than 10% of the tumor cells (score 2). A high labeling designation was awarded to cores where more than 10% of the tumor cells displayed strong immunolabeling (score 3). Both authors evaluated the samples with similar results. The labeling score for EGFR and pSer9GSK3β is based on immunohistochemical intensity. Different cores from the same patient occasionally showed a variable labeling index, ranging from a low score to a high score. In these patients, the highest labeling score from the various cores is reported. The patient data is illustrated in Table 1. To determine if the labeling indexes of CDC2 results were dependent on cell densities, the percentage labeling in select cases is quantified. In these calculations, the denominator is the total number of tumor cells, and a minimum of 1000 tumor cells were counted per sample.

Statistical Analysis and Survival Data.
Analysis of variance (ANOVA)/Tukey-HSD, Chi-squared test, and heat maps of the Bredel et al. dataset [16] were performed with R v2. 12, an open source programming language and statistical framework (http://cran.r-project.org/mirrors.html). Data from the Rembrandt Database [17] were queried for survival data related to expression and copy numbers of CDC2, CDK7/CAK, WEE1, GSK3β, Cyclin A2, and Cyclin B1 in patient populations diagnosed with GBM, Astrocytoma WHO grades II-III, and Oligodendroglioma WHO grades II-III. The data from these queries are presented in

Infiltrative Glial Tumors Show Diverse Levels of CDC2
Phosphorylation by Immunohistochemistry. GBMs showed elevated pThr161CDC2 (activated) labeling relative to oligodendrogliomas and ependymomas ( Figure 2 and Table 1). The mean pThr161CDC2 positive cells in GBM was 2.8-fold and 5.6-fold higher than oligodendrogliomas and ependymomas, respectively (Figure 4(c)). Similarly, anaplastic/WHO Grade III oligodendroglioma showed more labeling relative to WHO grade II oligodendroglioma ( Figure 4(e)). Mean percent pThr161CDC2 labeling in GBM was 18% (SEM = 5.2), whereas for oligodendroglioma grade II percent labeling was 1.7% (SEM = 0.8) and for ependymoma percent labeling was 2.3% (SEM = 1.5). Of note, all patients with ependymoma showing elevated pThr161CDC2 labeling were found in the pediatric cohort ( Table 1). The pThr161CDC2 cell counts from the GBM group were statistically significant from both the oligodendroglioma and ependymoma group (P < .05 by Tukey HSD Test). pThr161CDC2 cell counts were not statistically significant between the oligodendroglioma and ependymoma group. Three distinct labeling patterns are noted in the pThr161CDC2 positive cells. In non-mitotic cells, the predominant labeling pattern was nuclear ( Figure 2(f), arrow heads). In mitotic cells, the pThr161CDC2 stain showed cytoplasmic localization (Figures 2(b), 2(c), and 2(f), single arrows). All mitotic cells identified showed pThr161CDC2 labeling in the GBM, oligodendroglioma, and ependymoma samples. However, in 13 of the GBM cores (17.3% of total cores), the pThr161CDC2 labeling showed cytoplasmic localization in non-mitotic cells (Figure 2(c), double arrow). Cytoplasmic labeling in non-mitotic cells was not identified in the oligodendroglioma and ependymoma cores.
Quantification of the number of positive cells per three high power fields demonstrated elevated pTyr15CDC2 (inhi-bited) labeling in GBM compared to oligodendrogliomas and ependymomas ( Figure 4). pTyr15CDC2 in GBM was 4.2-fold and 6.9-fold-increased relative to oilgodendrogliomas and ependymomas, respectively (Figure 4(c)). These data are also confirmed by quantification of the percent positive cells. In GBM, the percent positive pTyr15CDC2 is 17% (SEM = 4.1), whereas in oligodendroglioma grade II percent positive pTyr15CDC2 is 5.4% (SEM = 2.2) and Ependymoma is 1.7% (SEM = 0.6). For the pTyr15CDC2 cell counts, the cell counts from the GBM group were statistically significant from the oligodendroglioma and ependymoma group (P < .05 by Tukey-HSD Test). pTyr15CDC2 counts were not statistically significant between the oligodendroglioma and ependymoma groups.  The relationship of pTyr15CDC2 to pThr161CDC2 in each individual core was evaluated and the results plotted in Figure 4(d). The majority of oligodendroglioma and the ependymoma cores showed low pTyr15CDC2 labeling with low pThr161CDC2 labeling. In the GBM arrays, half of the cores showed high pTyr15CDC2 labeling with high pThr161CDC2 labeling. The three oligodendroglioma cores showing high pTyr15CDC2 labeling with high pThr161CDC2 labeling were anaplastic/WHO grade III oligodendrogliomas, but this group constituted less than 10% of total anaplastic oligodendrogliomas (Figure 4(e)). Although rare, pTyr15CDC2 positive mitotic was noted in GBM and in one patient with anaplastic/WHO grade III oligodendroglioma (Figures 4(a) and 4(b)). There were no mitotic pTyr15CDC2 positive cells identified in the WHO grade II oligodendroglioma or the ependymoma arrays. Nearly half of the WHO grade III oligodendroglioma cores show elevated pThr161CDC2, whereas in the WHO grade II oligodendroglioma, only 4% show elevated pThr161CDC2 (Figure 4(e)). This is in contrast to pTyr15CDC2 where the percentage of low labeling cores in grade II and grade III oligodendroglioma are 88% and 75%, respectively. There is no statistically significant association between pThr161CDC2 and pTyr15CDC2 labeling with the MIB-1 labeling indexes obtained from the tissue cores (data not shown).          The presence of pSer9GSK3β labeling highly correlates to CDC2 phosphorylation in all three glial tumors studied (Figure 6(g)). The majority of tumors with low pSer9GSK3β labeling showed low levels of both pTyr15-and pThr161-CDC2 levels, whereas the majority of oligodendroglioma and the GBM cores with high pSer9GSK3β also had high levels of both pTyr15-and pThr161-CDC2 levels ( Figure 6(g)). This trend was particularly evident in the GBM and oligodendroglioma samples.

Infiltrative Glial Tumors Show
Correlation between pSer9GSK3β levels and EGF receptor overexpression was tested by labeling with anti-EGF receptor antibody (Figure 6). Nearly thirty percent of the GBM cores had either moderately or highly elevated EGF receptor membrane labeling (Figures 6(a)-6(c) and, Table 1), whereas only three oligodendroglioma cores (3.2%) and no ependymoma cores showed EGF receptor expression by immunohistochemistry (Table 1). Although half of the GBM cores with high EGF receptor expression showed elevated pSer9GSK3β labeling, roughly 40% of the cores with elevated pSer9GSK3β staining were associated with low EGF receptor expression (Figure 6(h)).

G2 → M Checkpoint Protein Transcript Levels Are Changed in GBM Cell Lines after Creating Resistance to the O-6 Methylating Drugs Temozolomide and BCNU.
To evaluate the effect of resistance to the O-6 methylating drugs BCNU and Temozolomide on CDC2 interacting proteins, the data set published by Bredel and coworkers was interrogated [16].
In these experiments, glioblastoma cell lines from three patients (initialed DI, LX, and ME) were derived before and after treatment. Treatment included surgical debulking, radiotherapy, and chemotherapy. The cell lines were then selected for resistance to the O-6 methylating drugs Temozolomide (TMZ) and BCNU, and gene expression analysis was performed (for further experimental details, see Bredel et al. [16]). Selected CDC2 interacting and G2 → M transitioning proteins were evaluated, and the log base 2 relative expression value of these genes was used to draw the heat map illustrated in Figure 7. The Y -axis of the heat map shows the genes, labeled on the right, and the hierarchical    Figure 8 demonstrates the Kaplan-Meyer survival curves in GBM, astrocytoma WHO grades II-III, and oligodendroglioma WHO grades II-III for CDC2 gene expression and CDC2 copy number. Data on ependymomas was not available. Numerical data associated with these graphs are presented in Table 2. In the GBM and the astrocytoma group, CDC2 gene amplification was associated with increased survival (log ranked P-value = .029 for each group), whereas CDC2 amplification did not significantly lead to increased survival in the oligodendroglioma group (log ranked P-value = .86).
To evaluate if CDC2-interacting proteins showed similar effects on patient survival, similar queries for GSK3β, WEE1, CDK7/CAK, Cyclin A2, and Cyclin B1 were performed. The results of these queries are illustrated in Figure 9 (GBM) and Figure 10 (oligodendroglioma WHO grades II-III), while the numerical data are presented in Table 3 (GBM)  and Table 4 (oligodendroglioma WHO grades II-III). No significant difference, as determined by log ranked P-value, was seen on patient survival with expression or copy number changes in any of the involved genes. Overall, the trends of expression (i.e., upregulation versus downregulation versus intermediate expression) were very similar between the GBM and oligodendroglioma groups. However, Cyclin A2 was more commonly amplified in the oligodendroglioma group relative the GBM group (Pearson's Chi-squared test with Yate's continuity, X 2 = 12.5, df = 1, P-value = .00004).

Discussion
Many investigators have reported studies in which either pharmacological or radiological induced cell cycle arrest at the G2 → M checkpoint slows GBM growth [3,5,[18][19][20][21][22]. These data are in overall concurrence with the notion that increased CDC2 copy number is associated with increased survival in astrocytic neoplasms and underscore the importance of the G2 → M cell cycle checkpoint in GBM. However, similar results in oligodendroglioma and ependymoma have not been reported. We found no studies in the literature that examined the CDC2 or GSK3β activation states in these tumors. GBM's showed higher labeling of both the promitotic pThr161CDC2 and antimitotic pTyr15CDC2 relative to both oligodendroglioma and ependymoma. In addition, promitotic CDC2 levels (pThr161) positively correlated with the histologic grade in oligodendroglioma. This is of no surprise, as similar findings with other mitotic markers have been reported in glioma [14,23,24]. Similarly, all ependymomas with elevated pThr161CDC2 levels occurred in pediatric patients, possibly reflecting the reportedly more proliferative behavior of these neoplasms in this age group [25,26]. The positive correlation between oligodendroglioma grade and phospho-CDC2 raises the possibility that phospho-CDC2 labeling may correlate with tumor behavior.
Tyrosine-15 phosphorylated CDC2 is expected to prevent cells to enter mitosis, whereas threonine-161 phosphorylation promotes mitosis. 15 of 20 WHO grade II oligodendrogliomas, 16 of 20 ependymomas, and 5 of 17 of the WHO grade III oligodendrogliomas showed low pTyr15CDC2 with low pThr161CDC2 labeling. However, dual elevation of pTyr15-and pThr161CDC2 was seen nearly in half of GBM patients. Elevations of both pTyr15-and pThr161-CDC2 are unexpected and is of uncertain significance. Although we have found no reports that suggest phosphorylation at both residues, Tyr15 and Thr161 are mutually exclusive, it is nevertheless unclear what function a dual phosphorylated CDC2 molecule would have. Alternatively, two intracellular pools of CDC2 may be present, with each pool diverted towards one phosphorylation state. In this scenario, the intracellular ratio of pTyr15CDC2 to pThr161CDC2 would determine the final fate of the cell. As the CDC2 phosphospecific antibodies used are both derived from immunized rabbits, we were unable to perform dual immunolabeling to determine if the same cells with elevated pTyr15CDC2 also were pThr161CDC2 positive, and hence we cannot exclude that the elevated labeling of these two epitopes are present in different cells of the same tumor.
Hirose et al. showed that temozolomide (TMZ) treatment of glioma cell lines in vitro leads to CDC2 phosphorylation at Tyr15, and that inhibition of cell cycle progression by pTyr15CDC2 may be an important step in TMZinduced cytotoxicity [20]. Suppression of CDC2's promitotic activity has been reported in many other chemotherapeutic agents [18,[27][28][29][30]. At least a subset of cells in GBM and anaplastic oligodendroglioma is capable of proceeding with mitosis despite signals that ought to prevent such an act (i.e., phosphorylation at Tyr15CDC2, see Figure 4). In addition, GBM cell lines selected in vitro for TMZ resistance showed downregulations in CDC2. This raises the question of whether or not a glioma cell, which is capable of proceeding to mitosis despite the presence of an arrest signal (i.e., pTyr15CDC2), would be responsive to drugs that require a G2 → M arrest. Additional prospective studies of basal pTyr15CDC2 levels in GBM and their response to TMZ treatment are thus necessary to evaluate the use of pTyr15CDC2 labeling as a prognostic indicator of survival in a manner similar to CDC2 copy number.
The distinct subcellular localizations noted for CDC2 and GSK3β in the different tumor types is interesting and may suggest distinct biological functions. Studies performed using antibodies that detect total CDC2 show that its subcellular localization changes in relation to cell cycle in yeast, plants, and animals [31][32][33]. CDC2 is located in the nucleus where it is tightly bound to chromatin [31]. During mitosis, CDC2 localizes to the centrosomes, where it closely interacts with spindle-associated microtubules, regulating their stability through phosphorylation [34][35][36][37]. The presence of activated CDC2 (pThr151CDC2) in the cytoplasm in GBM cells may thus suggest a role in microtubular reorganization in those cells. Of interest to the nuclear GSK3β noted in GBM and one case of anaplastic oligodendroglima is a report by Caspi et al. showing that nuclear GSK3β actively inhibits the canonical Wnt/βcatenin signaling pathway through a nonphosphorylationdependent mechanism [38]. In summary, the differences in subcellular localization of CDC2 and GSK3β noted in GBM and anaplastic oligodendroglioma suggests that in those cells these with distinct subcellular localization, these molecules may have unique functions.

Conclusions
GBM showed elevated levels of activated (pThr161CDC2) and inhibited (Tyr15CDC2) relative to oligodendroglioma and ependymoma. This suggests that G2 → M transition regulation is highly important in GBM. In support of this, glioma cell lines show changes in gene expression in CDC2 interacting genes after developing resistance to TMZ, a known target of the G2 → M phase transition. In addition, increased CDC2 copy number leads to increased patient survival in astrocytomas WHO II-IV, but not oligodendroglioma.