|Home | About | Journals | Submit | Contact Us | Français|
Although pilocytic astrocytoma (PA) is the most common brain tumor diagnosed in children, few prognostic variables have been delineated that stratify the risk of clinical progression in patients with this tumor. In this study, the MIB-1 labeling index was compared with 2 other immunohistochemical markers of cell proliferation, phospho-histone H3 (PHH3) and mini-chromosomal maintenance protein 2 (MCM2) in 80 incompletely resected PAs to see which was best able to identify patients at risk for tumor progression. 06-Methylguanine-DNA methyltransferase (MGMT) protein expression, which has been predictive of progression-free survival (PFS) in high-grade gliomas in children, was also evaluated in these cases. The mean follow-up period was 7.81 ± 3.9 years, and 42.8% of tumors have shown progression at the time of censoring. A MIB-1 labeling index ≥2.0 was associated with shortened PFS as a grouped variable by log-ranked analysis (P = .03) and by Cox regression analysis as a continuous variable (P = .007). None of the other potential biomarkers was significantly predictive of PFS. Although the amount of MCM2 staining correlated with the MIB-1 labeling index (P < .001), MCM2 reactivity was not independently associated with outcome. We conclude that MIB-1 labeling remains the best predictor of PFS in pediatric PAs.
Pilocytic astrocytomas (PAs) are the most common brain tumors diagnosed in children and can occur anywhere in the central nervous system. Although they are categorized as World Health Organization (WHO) grade I gliomas and generally have an excellent prognosis, many children have significant morbidity from these neoplasms and recurrence is not unusual.1 A 38.7% progression rate has been reported in a series of 141 cases of PAs from our institution.2 The only well-established prognostic variables that identify children at higher risk for recurrence are the closely interrelated factors of tumor location and the degree of surgical resection. An elevated MIB-1 labeling index (MIB-LI), an immunohistochemical estimation of the proportion of cells in the growth phase of the cell cycle, has also been shown to be associated with an increased likelihood of tumor progression in PAs.2
The use of immunohistochemistry-based biologic markers to predict cancer prognosis and response to therapy is becoming widespread in diagnostic pathology today. MIB-1 immunostaining has become an invaluable tool in accurately categorizing gliomas and other brain tumors.3 Recent studies have suggested that other biologic markers, such as phospho-histone H3 (PHH3) and mini-chromosomal maintenance protein 2 (MCM2), may more accurately identify cells within the proliferating phases of the cell cycle than MIB-1 antibody.4–6 Antibodies against PHH3 specifically target the phosphorylated version of this core histone protein. As phosphorylation of the H3 protein is seen maximally during mitotic chromosome condensation during early prophase and negligibly at any other times, immunohistochemistry for PHH3 provides an excellent marker of mitosis.7 Immunohistochemistry for PHH3 has been reported to be superior to MIB-1 in discriminating between grade II and grade III diffuse astrocytomas in adults and may offer advantages over MIB-1 staining in predicting the clinical outcome of patients with high-grade gliomas.6 MCM2 is part of a family of proteins that form prereplicative complexes, binding to DNA sites and allowing replication to occur. Upregulated expression of this protein is seen as cells transition into the G1/S phases of the cell cycle.8 MCM2 expression has recently been reported as a potential prognostic marker for a variety of human cancers, including lymphoma, meningioma, and oligodendroglioma.9–11 One objective of this study was to determine whether either of these 2 markers of cell proliferation was superior to MIB-1 in predicting tumor progression in incompletely resected pediatric PAs.
06-Methylguanine-DNA methyltransferase (MGMT) is another protein that holds potential as a predictor of prognosis and therapeutic response for CNS neoplasms. This enzyme, also known as alkylguanine methyltransferase, or AGT, is involved in repairing DNA damage and is variably expressed in high-grade glial neoplasms and other tumors.12,13 When present in tumor cells, MGMT appears to provide a mechanism for resistance to therapy with alkylating agents. High-grade gliomas with reduced or absent MGMT expression have increased sensitivity to temozolomide and other alkylating agents and improved overall survival (OS).13–15 Loss of MGMT expression in glial neoplasms often occurs from gene silencing due to methylation of the 5′ promoter; however, other mechanisms are likely also involved, as correlation studies between assays of tumor MGMT promoter hypermethylation and clinical outcome have shown variable results.13,16 Interestingly, hypermethylation of the MGMT promoter in adult gliomas is associated with a worse outcome whether or not the patients are treated with alkylating agents, suggesting that this may be an important independent biomarker of prognosis.17 Much less is known about the importance of MGMT expression in low-grade gliomas, especially in children. One study found that MGMT levels in pediatric low-grade gliomas was similar to those seen in high-grade gliomas, but there was no association with progression-free survival (PFS) or OS, possibly due to the small sample size.18 Only 1 prior study has examined MGMT expression by immunohistochemistry in pediatric PAs, and it found no association with clinical outcome.19 In our study, we examined whether MGMT protein expression, as estimated semiquantitatively by immunohistochemistry, correlated with PFS among a large group of children with subtotally resected pilocytic astrocytomas.
Cases of incompletely resected PAs were identified from the files of patients less than 19 years of age who underwent evaluation and treatment in the Neuro-Oncology Program from 1988 until 2005. The Neuro-Oncology Program includes patients seen at Children's Medical Center of Dallas, the Zale-Lipshy University Hospital at The University of Texas Southwestern Medical Center, and Medical City Hospital—Dallas. This study was reviewed for human subject protection and confidentiality and was approved under the expedited review process by the Institutional Review Board of the University of Texas Southwestern Medical School in accordance with standards of the National Institutes of Health. Recorded were demographic data, including the patient's date of clinical presentation, age at diagnosis, gender, tumor location, postoperative and follow-up imaging findings, use of adjuvant therapy, date of last contact, and patient outcome. Tumor progression was defined as an increase in the size of the tumor on imaging studies or worsening symptoms (eg, objective worsening of neurological exam, increase in seizure activity) significant enough to warrant a change in therapy. Fifty-five of these cases were included in the previously reported series of 141 consecutively diagnosed cases of PAs from our institution.2
The majority of patients had aggressive surgery as their primary and only therapy. The diagnosis of incomplete resection (residual tumor present) was based upon postoperative imaging studies performed from the first postoperative day until 3 months following surgery and correlated with the surgeon's intraoperative assessment of the degree of resection. Only patients with incompletely resected tumors were evaluated in this study.
As a result of the treating physician's judgment, adjuvant therapy was occasionally prescribed for tumors that were located in unfavorable surgical locations (such as the brainstem or optic pathway) or for rare metastatic tumors. Generally, chemotherapy, which consisted of carboplatin with or without vincristine, was prescribed for prepubertal children and for tumors located in the cerebral hemispheres, thalamus, and optic pathways. In a few cases, radiation therapy (5000 to 6000 cGy) was administered to older patients and patients with tumors located in the cerebellum and brainstem.
Pathology slides from all resected tumor specimens were reviewed by one of the authors (LRM) to confirm the diagnosis of a PA using the established criteria from the WHO classification of brain tumors1 and to confirm that there was sufficient suitable material for the immunohistochemical studies. Tumors that lacked the classic histopathologic features of PA were excluded from this study, including tumors with pilomyxoid features or mixed glioneuronal neoplasms.
Immunohistochemistry was performed using commercially available reagents according to manufacturer's recommendations. The following antibodies were used: MIB-1 mouse monoclonal anti-human Ki67 antibody, 1:150 dilution (Dako Corporation); anti-PHH3 (Ser 10) rabbit polyclonal antibody, 1:1000 dilution (Upstate Biomedical–Millipore Labs); MCM2 mouse monoclonal antibody, 1:40 dilution (Vision Biosystems–Novacastra); and MGMT mouse monoclonal antibody 1:20 dilution (Lab Vision). MIB-1 and MGMT staining was performed manually using the Envision Plus Staining System (Dako). PHH3 and MCM2 staining was performed on the Discovery XT autostainer (Ventana Medical Systems) using OmniMap reagents. All stains were performed on 4-µM-thick sections of formalin-fixed, paraffin-embedded sections of tumor with diaminobenzidine as the chromogen. Known positive control slides were processed with each batch of slides. For the quantification of MIB-1-, PHH3-, and MCM2-positive cells, microscopic images were viewed with an Olympus BH-2 microscope with the microscopic image projected onto a Sony DVM 1942Q color video monitor fitted with a 10 × 10 grid of equal squares via a Sony DXC-107 color video camera. Slides were scanned to identify the area with the highest density of positively staining nuclei, and these “hot-spot” areas were used for quantification. Counting was performed by a single observer, blinded to patient outcome, and intraobserver variability was not assessed. Nuclei showing obvious brown staining were considered positive on the slides stained for MIB-1, MCM2, and PHH3. For slides stained with PHH3, only the cells showing brown staining of a structure with the morphologic features of a mitosis (condensed nuclear chromatin forming chromosomes, visible mitotic spindles, etc.) were counted as positive. The number of positive nuclei in 15 to 20 contiguous 200× fields in the hot-spot area was quantified, and the total number of nuclei in each field was determined using the method of Going,20 examining at least 1000 cells per case. Nuclei of nontumor cells, including vascular endothelial cells and hematopoietic elements, were not included in the counts. The percentage of reactive nuclei was determined for each of these stains to determine the labeling index for MIB-1 (MIB-LI), PHH3 (PHH3-LI), and MCM2 (MCM2-LI). The immunoreactivity of MGMT protein was evaluated semiquantitatively by estimating the fraction of cells with positive nuclear staining (<5% was considered negative, 5%–25% low reactivity, 25%–50% moderate reactivity, and >50% high reactivity) in 15 to 20 ×400 magnification fields.13 Separate estimates of the degree of reactivity were made for the entire slide by examining 15 to 20 random fields and for 15 to 20 contiguous fields in the hot-spot area.
The clinical data were expressed as means ± SD where appropriate. Clinical and histologic parameters were compared using a t-test or χ2 test. PFS and OS were analyzed using the Kaplan–Meier method, and comparison of study groups was performed using the log-rank test. The Pearson correlation coefficient was used for MIB-1 L1, PHH3-LI, and MCM2-LI when analyzed as continuous variables; Kendall's tau correlation was used when values were grouped.
The statistical package SPSS, version 15.0, and Advanced Statistics, version 7.5 for Windows (SPSS Inc) were used to conduct statistical analyses. A P value of less than .05 was considered statistically significant, and all tests were 2-tailed.
Of 231 patients enrolled in the neuro-oncology program with PA during the study period, 80 with subtotally resected PAs were identified who had sufficient residual paraffin-embedded tissue to perform the immunohistochemical studies. The other 151 cases were excluded because of primary surgery elsewhere (n = 44), insufficient residual tissue (n = 47), gross total surgical resection (n = 44), no clinical follow-up (n = 5), change in pathologic diagnosis (n = 4), presurgical adjuvant therapy (n = 5), or age at diagnosis >18 years (n = 2). Of the study population of 80 patients, the mean age at diagnosis was 8.0 years (range, 0.1–18 years), with an equal number of males and females. Three patients (3.75%) had coexisting neurofibromatosis type I. Patients have been followed for a mean of 7.81 years (SD, 3.9 years) after the initial pathology-confirmed diagnosis. Additional clinical information is detailed in Table 1. At the time of censoring the data, 48 patients had experienced tumor progression (Kaplan–Meier PFS at 5 years was 42.8%) (Fig. (Fig.1).1). The mean time to tumor progression was 1.74 years after diagnosis (SD, 2.10 years). Two patients have died (OS rate, 97.5%).
Overall, the mean MIB-LI was 2.50 ± 2.33 (range, 0.08–13.9). The shortened progression-free interval was significantly associated with a MIB-LI greater than or equal to 2.0 (P = .03) and tumor location within the optic/hypothalamic areas (P = .038) by log-rank analysis. By Cox regression analysis, the MIB-LI as a continuous variable (P = .007) and as a grouped category of <2.0 MIB-LI vs. ≥2.0 MIB-LI (P = .03) was significantly associated with shortened PFS; no other variable was significant at P≤ .05 (Fig. (Fig.2).2). Using Cox regression analysis, MIB-LI was significant independent of location (optic/hypothalamic area; P = .037).
PHH3 immunostaining produced intense reactivity in the condensed chromosomes of mitotic cells, readily allowing their identification to determine the mitotic index (Fig. 3). Scattered immunoreactive interphase nuclei with no morphologic features of a mitotic figure were not included in the counted cells, as previously described.6 The mean number of PHH3 positive cells in all cases was 0.19 ± 0.24 (range 0.0–1.0). The PHH3-LI did not correlate with the MIB-LI and was not significantly associated with progression regardless of whether a cutoff of 0.1, 0.5, or 1.0 was used (Table 2).
Staining for MCM2 produced intense nuclear reactivity, which was readily distinguished from negative cells and was easy to quantify. Reactive tonsils, used as positive control tissue, showed staining in the majority of cells in the germinal centers. PAs, however, had only scarce positively stained tumor cells, and more than half the cases showed reactivity in less than 0.1% of cells (mean, 0.65 ± 1.0; range, 0.05–5.80). Although the MCM2-LI correlated with the MIB-LI (P < .001), the extent of MCM2 reactivity was not significantly associated with progression regardless of whether a cutoff of 0.1, 0.5, or 1.0 was used (Table 2).
Most tumors showed at least focal areas with intense nuclear reactivity for MGMT. At least 25% of tumor nuclei were immunoreactive in 34 cases (43%) in the hot-spot area and 18 (23%) had at least that amount of reactivity in randomly selected fields. There was no statistical association apparent in the amount of MGMT expression seen in tumor sections and in the patient's clinical progression (Table 2).
Although many children with PAs have indolent tumors that are essentially cured by gross total surgical excision, a significant number of patients have tumors that are not amenable to complete resection but recur or disseminate. To date, few definitive histopathologic, biologic, or molecular markers have been identified to stratify the patients who may need more aggressive therapy or follow-up. Recent publications have evaluated the potential predictive value of histologic features of pediatric PAs but with discrepant, inconclusive results. For example, necrosis, vascular hyalinization, and calcification have all been suggested to predict an adverse outcome in one study and not in another.19,21 Only the presence of oligodendroglial-like morphology was consistently predictive of poorer outcome in these 2 studies. Another publication found that increased microvessel density, as determined by quantified immunohistochemistry, was associated with a significantly higher rate of progression in 41 pediatric patients with optic pathway gliomas.22 This interesting observation warrants further evaluation in PAs in other locations.
Tumor location has been well established as an adverse predictor of PFS previously and our study again confirmed this with a significantly higher proportion of tumors in the optic/hypothalamic area showing progression.
This study reconfirmed the validity of using the MIB-LI as an important independent predictor of PFS in patients with incompletely resected PAs. None of the other immunohistochemical markers studied here was of use in predicting PFS in this large and unique series of patients with incompletely resected tumors and adequate clinical follow-up.
The use of MIB-1 immunohistochemistry has become well established in diagnostic neuro-oncologic pathology to idntify tumors with a large population of proliferating cells, which usually indicates more aggressive biologic behavior and a poorer patient outcome. The MIB-1 monoclonal antibody is one of the most frequently used and sensitive of the commercially available Ki67 antibodies, with over 4000 literature citations. Although the function of the Ki67 antigen is still unknown, this nonhistone nuclear protein is expressed throughout the active parts of the cell cycle (G1, S, G2, and mitosis).23 The most frequent criticism of the use of MIB-1 immunostaining is the variable intraobserver reproducibility depending on the method used for quantification.24 Our method of quantification is the one most widely used and is the same as was performed in our previous study.2 As the same counting methodology was used for the analysis of all 3 of the proliferative markers in this study, this provides a useful comparison of their benefits.
PHH3 is becoming well established as a prognostic marker for identifying and stratifying some types of adult brain tumors, as it allows easy identification of mitosis. While counting mitoses has long been a method of identifying proliferative activity in neoplastic cells, highlighting the mitotic structures with immunoreactive labels such as PHH3, makes the practice quicker and more accurate by eliminating technical artifacts and allowing the discrimination of mitoses from apoptotic cells. Our study indicates, however, that in PAs, where mitoses are very uncommon, quantification of PHH3 is not useful in identifying patients with shortened PFS. More than half of our cases had no PHH3-positive mitoses identified in the histology sections. In cases that had PHH3-immunoreactive mitoses, the number of reactive cells did not discriminate between patients who had stable disease or clinical progression. This suggests PHH3 immunohistochemstry may be of more benefit in high-grade malignancies, where mitoses are more numerous.
MCM2 is another proliferative marker that holds promise as a potential prognostic variable in cancer histopathology. MCM2 is one of a family of proteins that form a complex that appears to function as a DNA helicase, allowing DNA to unwind at the initiation of DNA replication.11 In normal nonneoplastic tissues, immunohistochemical expression of MCM2 is restricted to the nuclei of cells in proliferating areas, such as the basal regions of gastrointestinal and squamous epithelia.8 In carcinomas, a higher proportion of poorly-differentiated tumor cells express MCM2 than do well-differentiated tumor cells.8 Our study found tumors had a smaller proportion of positive cells with MCM2 than were seen with the MIB-1 staining; however, the 2 values were statistically related (P = .002). Due to the diminished degree of reactivity, however, MCM2 staining was not an independent predictor of PFS. It could be anticipated that MCM2 would stain a smaller proportion of tumor cells than MIB-1, as it is known to selectively identify cells in the early initiation phase of DNA replication and is downregulated in differentiated or quiescent cells.10,11 It is interesting that several studies have shown MCM2 staining to be more abundant than MIB-1 in a variety of normal and neoplastic tissues.8,10,11 In meningiomas in adults, MCM2-positive cells were often more than 4 times as prevalent as MIB-1 reactive cells, although it was reported that there was marked heterogeneity of MCM2 staining in different areas of the tumor.10 That study reported the majority of recurrent meningiomas had an MCM2-LI of >30%, a number far exceeding the degree of staining seen in any of our gliomas.10 A study examining MCM2 expression in a series of oligodendrogliomas also found a 4 times higher proportion of tumor cells staining for MCM2 than MIB-1, with a mean of 20% of the tumor cells showing reactivity for MCM2.11 It is not clear why in both those studies such a high percentage of tumor cells expressed this supposedly tightly regulated marker of DNA replication. It would be anticipated, as was seen in our study, that a smaller number of cells expressed MCM2 than MIB-1. It is possible that reactivity for MCM2 varies among different types of CNS tumors, and further studies of glial neoplasms with this marker may be of benefit.
Identifying specific biomarkers that are predictive of therapeutic response holds the potential to further improve therapy specifically tailored to the individual patient. A growing library of immunohistochemical reagents is now available to help guide therapeutic decisions. Immunohistochemical assays have the advantage over other types of tests in that they are readily available in most hospital laboratories and the studies can be done on routinely processed tissue sections in conjunction with the diagnostic pathology workup. One such potential biomarker, MGMT, has been studied fairly extensively in adult malignant gliomas, as it holds the potential to predict tumor sensitivity to alkylating agent chemotherapy.13,14 In children with high-grade gliomas, MGMT expression by immunohistochemistry strongly correlated with outcome in patients treated with alkylator-based adjuvant chemotherapy.15 As there have been suggestions that MGMT expression may be an independent predictor of outcome, regardless of the use of adjuvant therapy, we thought it was important to evaluate it in PAs. For our study, we selected a previously published, detailed scoring system to semiquantitatively estimate the amount of MGMT immunoreactivity, as this protein is known to be heterogeneous in its distribution and intensity in tumors.13 In addition to quantifying random fields, as had been previously described, we also examined a series of contiguous fields from the so-called hot-spot areas of greatest intensity to see if either method was predictive of PFS. Although there is great variability in the degree of MGMT expression among different PA cases, the amount of reactivity did not correlate with PFS. To the best of our knowledge, only one prior study has examined MGMT by immunohistochemistry in pediatric PA cases and, although it used a much simpler grading scheme than we did, it also found no association of MGMT expression with PFS.15 The lack of association of MGMT expression and PFS provides additional evidence that PAs are biologically distinct from high-grade infiltrating gliomas, and a search for other relevant biomarkers should continue.
Relatively little is understood about the molecular changes that prompt the development of PAs, and it is expected that better understanding of these alterations will identify prognostically relevant markers and potential therapeutic targets. High-sensitivity comparative genomic hybridization analysis has found a few common alterations, such as gains in chromosome 7q34 leading to increased copy number of the BRAF gene, in the majority of cases.25 The clinical and prognostic significance of these findings is not yet determined.
In conclusion, this study reconfirms the value of the MIB-1 labeling index as a predictor of outcome in patients with PAs and indicates that more selective markers of cell proliferation, such as PHH3 and MCM2, are not useful in determining progression in this neoplasm. The amount of MGMT expression does not show promise as a predictor of PFS. Until further details of the molecular pathways involving PA are delineated, the MIB-1 labeling index shows the highest promise in predicting which patients are at highest risk of disease progression.
The authors wish to thank Mary Holland and Estella Castillo for expert secretarial assistance and would like to acknowledge the Children's Medical Center Pediatric Biospecimen Repository for providing the archival paraffin slides used in this project.
Conflict of interest statement. None declared.