Meningiomas: Loss of Heterozygosity on Chromosome 10 and Marker-Specific Correlations with Grade, Recurrence, and Survival

LOH³ of alleles on chromosome 10 has been reported in many cancers, leading to the identification of tumor suppressor genes on this chromosome. Several reports implicate LOH of chromosome 10 in meningioma progression, but the correlation of LOH with clinical indicators has not been well characterized. In a recent study of 208 meningiomas, we found a high incidence of LOH on chromosome 10 in benign (73.4%) meningiomas, with slightly increasing LOH occurring in atypical (80%) and malignant (86.7%) tumors (1). Because a large percentage of benign meningiomas had LOH for at least one locus, and a significant percentage of these had LOH at multiple loci (43.9%), the data suggest that one or more inferred tumor suppressor genes on chromosome 10 may contribute to meningioma progression. Mapping of the LOH patterns for these 208 meningiomas identified 4 regions of chromosomal deletion (1). Of the known or candidate tumor suppressor genes on chromosome 10, only the controversial tumor suppressor gene DMBT1 (2, 3) localized to one of the chromosomal deletions.

Further study is needed to identify those genes on chromosome 10 contributing to meningiomas; however, LOH of individual markers may serve as a surrogate marker to study tumor and clinical correlations that may be useful in the treatment or diagnosis of patients, as has been shown in other tumor types. For example, LOH at marker D10S209 was significantly associated with shorter survival in anaplastic astrocytoma patients, and LOH at D10S215 and D10S541 was associated with shorter survival in glioblastoma patients (4). Furthermore, LOH was age dependent, occurring in patients over 35 years of age (4). Such correlations have not been examined for meningiomas.

In previous reports examining LOH of alleles on chromosome 10 in meningiomas (5, 6), LOH was assessed relative to tumor histological subtype; however, either the sample size was too small, or the number of loci examined were too few to determine whether LOH at a particular locus occurred preferentially in a specific subtype. With respect to grade of meningiomas, we and others have demonstrated that LOH increases with increasing tumor grade (1, 5, 6, 7, 8, 9). However, LOH at individual markers has not been assessed for correlation with increased grade or for clinical correlations such as survival and recurrence.

Therefore, this study was performed to determine whether LOH at individual loci on chromosome 10 exhibits relevant biological or clinical predictive ability for meningiomas. We assessed LOH at 11 microsatellite dinucleotide repeat loci in 208 sporadic and recurrent meningiomas from 173 patients using LCM and fluorescence-based detection of PCR products. We investigated the correlations between the LOH results with respect to tumor location, histology, and grade, as well as patient race, age, gender, recurrence, and survival.

This study included a total of 173 patients with primary and/or recurrent meningiomas surgically treated at Henry Ford Hospital between 1968 and 2001. Race, gender, and age were documented. Dates of surgery, follow-up, recurrence, and death were recorded for the statistical analyses. This information was collected under an institutional review board-approved protocol.

For each patient, tumor grade, primary or recurrent status and the recurrence number, histological subtype, and tumor location were documented. A total of 208 paraffin-embedded tumors were collected for the 173 patients (Table 2). Normal DNA was either lymphocyte (113 samples) or microdissected adjacent normal brain tissue (60 samples). Two serial sections (5 μm) containing diagnostic tumor were obtained for each specimen. One section was stained with H&E, and the tumor was graded as either benign, atypical, or anaplastic/malignant, corresponding to the most recent WHO (10) grades I, II, and III, respectively (1). Histological subtype was also recorded from the H&E-stained sections (10). The neuropathologist circled the area of the tumor representative of the tumor grade. Institutional review board-approved informed consent was obtained from all patients or from the patient’s guardian for use of tumor tissue collected at the time of tumor resection. Since 1993, informed consent included the collection of blood for the extraction of lymphocyte DNA. Not all tumor specimens were available from surgeries performed before 1993, so patients with specimens collected before this date were excluded only from the calculation to assess the percentage of tumors by grade in the patient population.

All tumors were microdissected. LCM, DNA extraction, and PCR reactions have been reported previously (1). Briefly, the H&E-stained reference slides were used as reference for the selection of tumor and normal brain cells. DNA was extracted from the microdissected tissue in 50 μl of digestion buffer [10 mm Tris-HCl (pH 8.0), 1 mm EDTA, 1% Tween 20, and 0.1% proteinase K] at 52°C overnight. After digestion, the samples were centrifuged at 13,000 rpm for 5 min. Proteinase K was inactivated at 96°C for 8 min, and the extracts were used directly as templates for PCR. For all patients, tumor DNA was extracted from LCM specimens. For 60 patients, constitutional (normal) DNA was extracted from LCM brain samples. For 113 patients, normal DNA was extracted from lymphocytes using the Stratagene DNA extraction kit (La Jolla, CA). Matched tumor and normal DNAs were PCR amplified at 11 polymorphic microsatellite markers mapping to chromosome 10 (D10S179, D10S189, D10S89, D10S580, D10S109, D10S215, D10S574, D10S187, D10S209, D10S217, and D10S169), as reported (Ref. 1; Table 3). LOH was assessed using the CEQ 2000 XL fragment analyzer [Beckman Coulter, Fullerton, CA (1, 11)]. Specimens were scored as having maintenance of both alleles, having LOH, being noninformative, having microsatellite instability, or as not amplified. LOH was inferred when the peak area of one allele in tumor tissue was <40% of the other tumor allele compared with the ratio of peak areas of the two alleles in corresponding constitutional DNA (1, 12).

For all analyses, those specimens scored as not amplified, not informative, or as having microsatellite instability were excluded. Time was reported as the number of months or years. Significant association was concluded when P ≤ 0.05. The analyses were done using the SAS system (13). The lifetest procedure was used to calculate the median and 5-year survival, the median and 2-year recurrence, and the recurrence freedom by grade at 2 years. Significance of the lifetest was calculated using the log-rank test. Student’s t test was used to evaluate the relationship between LOH and the mean age. The Pearson χ² test was used to analyze the relationship between LOH and interval age, race, gender, tumor location, and histological subtype. Logistic regression analyses were used to examine the association between LOH and tumor grade, where benign tumors were compared with atypical and malignant tumors combined. Cox regression analyses were performed to examine the relationship between LOH and survival and recurrence. For the survival analysis, only the primary tumor was used, and follow-up was time to first recurrence, death, or date last seen. For recurrence analysis, the analysis was done using all specimens for a patient, given that we had the primary tumor and each of their subsequent tumors (if they had a recurrence), and follow-up was time to recurrence, death, or date last seen. Survival and recurrence curves were plotted according to Kaplan-Meier. To prioritize future investigations based on these results, Hochberg’s method was used to assess whether nominal Ps of ≤ 0.05 remain significant when adjusted for multiple comparisons (14). Separate Hochberg adjustment was made for each of the 11 loci.

The demographics of the patient population are presented in Table 1. The majority of patients were Caucasian or African American. We found a preference for meningioma occurrence in females versus males for the whole population, with tumors generally occurring in older patients. Of the 41 patients having 60 recurrent tumors (Table 2), 18 were male, and 23 were female. Recurrences occurred with an overall 2-year rate of 11.3% and a 5-year rate of 14.7%. The median recurrence freedom data are presented in Table 3A. The differences in 2-year recurrence freedom percentages for the grades were significant (P = 0.0001), with a greater percentage of recurrences occurring for atypical (50%) and malignant tumors (33.3%) in comparison with benign tumors (8.3%). Of the 19 patients having the 20 atypical tumors, nine were male. Of the ten female patients, one had a recurrent atypical tumor. Of the 11 patients having 15 malignant tumors, six were male. Two of the five female patients had two and one recurrent malignant tumor(s), whereas one of the six male patients had a malignant recurrence. The survival data are presented in Table 3B. The median overall survival was 20.3 years, with a 5-year survival rate of 79.4%. The median differences in 5-year survival rates for the grades were also significant (P = 0.0001), with no patients having malignant tumors surviving 5 years.

The specimen information is also presented in Table 2. Of the 208 tumors, 83.2% were benign, 9.6% were atypical, and 7.2% were malignant. For population comparisons, only those tumors collected after the tumor bank was started (1993–2001) were considered. Of these tumors, 88.6% were benign, 6.4% were atypical, and 5.0% were malignant. There were 71.2% primary tumors and 28.8% recurrent tumors. The major histological subtypes represented were meningothelial (41.3%), transitional (28.0%), and fibrous (19.2%). All other subtypes collectively represented 11.5% of the total. The majority of tumors occurred on the parietal and frontal dura.

No association was found between LOH and race or LOH and gender. A significant correlation between mean age and LOH status was found for D10S215. The mean age of patients without LOH was 52.6 ± 10.5 years, with a 95% confidence interval of 46.3–58.9 years. The mean age of patients with LOH was 65.3 ± 9.9 years, with a confidence interval of 59.0–71.6 years, indicating that LOH at D10S215 is more likely to occur in older patients.

There were significant differences between histological subtypes and LOH. LOH on D10S187 and D10S209 was associated with meningothelial and transitional histology, respectively. LOH on D10S217 correlated negatively with fibroblastic histology. A significant positive relationship was found between LOH at locus D10S89 and sphenoid tumors, suggesting that tumors in this location have increased likelihood of LOH. A significant negative relationship was found with LOH at locus D10S209 and parietal tumors, suggesting that tumors in this location have a decreased likelihood of LOH.

The percentage of tumors with LOH categorized by locus and grade is summarized in Table 4A. For all loci, LOH was observed in benign tumors. The percentage of benign tumors having LOH ranged from 19.5% to 46.2%. For the great majority of loci, the percentage of tumors with LOH increased with tumor grade. For D10S215, whereas LOH was observed for the single informative atypical tumor, there were no informative malignant tumors to assess the percentage of tumors with LOH at this locus. For D10S169, an increased percentage of tumors with LOH was observed for both atypical and malignant tumors compared with benign tumors; however, the atypical tumors demonstrated a greater percentage of tumors with LOH than the malignant tumors. Table 4 B summarizes the results analyzing the correlation between LOH and grade using all of the tumors. There were four significant relationships between grade and LOH on D10S179, D10S89, D10S580, or D10S169. A patient having a tumor with LOH on these loci had an increased likelihood of 10.94, 6.59, 4.99, or 6.61 times, respectively, of having a tumor that was either atypical or malignant.

Table 5A summarizes the results between LOH and long-term survival. LOH on locus D10S209 had a marginally significant result of P = 0.06. A patient having LOH for this allele would be 7.41 times more likely to die. A graph showing Kaplan-Meier curves for meningioma patients with or without LOH on D10S209 is presented (Fig. 1,A). Patients with tumors having LOH on D10S209 were more likely to have a shorter time of survival. Table 5B summarizes the results between LOH and recurrence. LOH on D10S169 was significant for predicting recurrence (P = 0.01). A patient having LOH on this allele would be 4.23 times more likely to have a recurrence. A graph showing Kaplan-Meier curves for meningioma patients with or without LOH on D10S169 is presented (Fig. 1 B). Patients with tumors having LOH on D10S169 had a shorter time to recurrence.

Table 6 summarizes the significant, marginally significant, and nonsignificant clinical and tumor correlations observed before and after adjustment for multiple comparisons. Of the loci having significant positive associations with tumor or clinical parameters, all are loci that define the smallest regions of allelic deletion reported previously (1) and illustrated in Fig. 2.

In this study, we investigated a series of 208 meningiomas for the correlation of LOH at 11 dinucleotide repeat loci on chromosome 10 with patient race, age, gender and tumor location, histomorphological subtype, grade, recurrence, and survival. We found significant correlations between specific loci and tumor location, grade, recurrence, and patient age and marginal correlations with survival using unadjusted Ps. The data retained significant correlations between location and grade and found marginally significant correlations between age, grade, and recurrence after using Hochberg’s multiple comparison method.

Analysis of the meningioma patient population used in this study indicated that it is representative, in part, of populations used in other studies. Consistent with the literature (15), tumors occurred more often in older adults, with a median patient age of 58 years. Benign tumors occurred twice as often in female patients as in male patients (15). Atypical and malignant tumors occurred similarly between male and female patients.

Different from previous reports (15), the percentage of malignant (5.0%) tumors collected since 1993 is slightly higher than what would have been expected (1.0–2.8%). In addition, the racial makeup of the patient population (63% Caucasian and 27% African American) differed from the incidence rate reported by the Central Tumor Registry of the United States, which indicates no difference in occurrence between African Americans and Caucasians (16). Furthermore, the 2-year recurrence freedom of 91.7% for benign, 50.0% for atypical, and 66.7% malignant tumor patients was higher for atypical tumors and slightly higher for malignant tumors than that previously reported [33% for atypical tumors and 56% for malignant tumors (17)]. However, no patients with malignant tumors survived past 5 years. These differences may reflect the local Detroit demographics of roughly 3 million Caucasians to 1 million African Americans, changes in grading criteria (10, 15), and/or a bias toward the complex or unusual cases referred to our center, respectively.

Using this patient population, we examined LOH at alleles reported in previous meningioma studies⁴ (1, 5, 6, 7, 8, 9) and additional loci mapping near known and candidate tumor suppressor genes on 10q (1). The high rate of LOH in benign tumors (73.4%) suggested that LOH of some of the markers might not be associated with tumor progression because only about 10% of the tumors progress. Thus, the LOH of some chromosome 10 alleles might serve as a genetic marker identifying specific subtypes of benign tumors. We therefore assessed the LOH on individual loci for correlations with both tumor and clinical parameters. No association was observed with respect to LOH and patient race. This outcome was expected because no difference in occurrence by race was reported by the tumor registry (16). We also observed no relationship between LOH and gender. These data suggest that the LOH of these alleles on chromosome 10 does not contribute to the 1.94:1 female:male ratio of occurrence that is consistent with gender differences in the tumor registry (16).

Correlations between LOH and several parameters were analyzed and assessed using both unadjusted and multiple comparison adjusted Ps. Multiple comparison adjustment was performed to identify those correlations whose nominally significant Ps might be due to random occurrence as a result of assessing many parameters at once. That is to say, when examining a large number of variables, the probability that one parameter will achieve statistical significance by chance increases. Such analyses help to prioritize outcomes when significance is altered from that obtained using the unadjusted analyses and point to those correlations that may benefit from an increase in the number of samples in follow-up studies.

Correlations between LOH and tumor location were assessed. A previous study (18) demonstrated an association between NF2 status (chromosome 22) and tumor location. Because tumors often reside in more than one location, LOH in each location was assessed for tumors occurring at least partly (unrestricted) or solely (restricted) in that region. Using either assessment, LOH on marker D10S209 was less likely to be found in tumors arising in the parietal region, and LOH on D10S89 was associated with tumors arising solely in the sphenoid region. These data suggest that tumors arising in different regions may result from genetic differences. This is supported by the observations that no restricted posterior fossa tumors had LOH at D10S580, D10S217, or D10S179, and no restricted temporal tumors had LOH at D10S179 (the small number of specimens prohibited statistical analysis). These data and the multiple comparison outcomes indicate that further study is warranted to determine whether these associations hold, and if so, the different LOH patterns suggest the possibility that tumors in different locations have different etiologies.

Although previous studies examining chromosome 10 found that meningiomas of all major histological subtypes had LOH, it was not established whether LOH occurring at a particular allele correlated with a histological subtype. However, one study (18) demonstrated an association between NF2 status (chromosome 22) and tumor histology, suggesting that genetic differences exist between the different histological subtypes. We found associations between LOH on chromosome 10 and this variable using statistical analyses without adjustment for multiple comparisons. Tumors with LOH for D10S187 or D10S209 were more likely to be meningothelial or transitional, respectively, suggesting that genes localized near these markers may contribute to these histologies. However, this analysis suggests other factors must also be important because, in our study, not every tumor having LOH at these loci was of these subtypes. Furthermore, the multiple comparisons adjusted analysis did not result in correlative significance. Additional study on a larger number of tumors may determine whether genetic allelotyping may be detecting subpopulations within these histological categories. It is interesting to note that the loci that correlated with specific histologies differ from the loci correlating with increased grade, supporting the concept that LOH of certain loci occurs early in tumorigenesis, whereas LOH of others may contribute to tumor progression.

An examination of the percentage of LOH occurring for each grade of tumor demonstrated that LOH slightly increased with increasing grade for all loci examined. Statistical analysis without multiple comparison adjustment determined that LOH at four of the loci correlated with increased grade. Two mapped to the p arm (D10S179 and D10S89), and two mapped to the q arm (D10S580 and D10S169), suggesting that tumors having LOH at these loci are likely to be atypical or malignant meningiomas. Two of these relationships were retained after multiple comparison analyses [one on each arm (D10S179 and D10S169)]. Because LOH at these loci also occurred in a percentage of the benign tumors, further study is needed to determine whether the LOH at these loci might serve as a marker of those benign tumors with the potential to progress to higher-grade tumors or identify those atypical tumors that have borderline histological parameters resulting in misclassification as benign tumors.

We reported previously three major patterns of LOH for the benign meningiomas (1), including tumors that had LOH restricted to the p arm, LOH restricted to the q arm, or LOH on both arms. Because the majority of atypical and all of the malignant tumors had loss on both arms, we speculated that benign tumors with loss on the q arm may be less aggressive than those benign tumors having loss on both arms. In this regard, it is interesting to note that LOH at D10S179 was associated with the greatest likelihood that the tumor was of increased grade, and it is located in the p arm deletion. Continued long-term follow-up of these patients is under way and will help to resolve these issues.

Several markers were found to be associated with prognosis. Survival analysis indicated that patients with tumors having LOH at D10S209 were more likely to have a shorter length of survival after diagnosis. Although this marginal significance was lost after multiple comparison adjustment, the Kaplan-Meier analysis suggests further study is warranted. Interestingly, this same locus was found to correlate with poor prognosis for anaplastic astrocytoma patients (4). This marker, therefore, has potential as a poor prognosis indicator, identifying patients who warrant closer follow-up and monitoring.

Recurrence is the primary cause of morbidity and mortality in meningioma patients. For this reason, the LOH for marker D10S169 may be the most useful clinical correlate from this study because of its significant unfavorable association with recurrence, despite dropping to marginal status when using the multiple comparisons analysis. Patients having tumors bearing LOH at D10S169 have increased risk of recurrence. Unfortunately, all of the recurrent tumors reported by Lamszuz et al. (9) were noninformative at this locus. Future analyses of these results with other clinical indicators that might correlate with tumor recurrence will be necessary to determine whether our LOH observation is an independent variable for tumor recurrence. We are analyzing such variables including location (skull base and parasagittal versus convexity), Simpson grade (to assess the amount of remaining postoperative tumor or dural attachment; Ref. 19), MIB-1 (to assess tumor growth), and invasion. If LOH at D10S169 is found to be predictive of future tumor recurrences, and a prospective study confirms that observation, clinicians might choose to administer a treatment such as postoperative irradiation to minimize such recurrence.

Interestingly, all of the loci having significant clinical associations are loci that define the smallest regions of chromosomal deletions (1). Because different loci correlated with each of these outcomes, the data suggest that specific genes at these loci may act independently to contribute to progression, recurrence, and overall survival. Such interpretations were supported by the Cox analyses that did not find further significance by combining alleles (data not shown).

In conclusion, we have observed frequent chromosome 10 allelic deletions early in the development of meningiomas, with increased incidence and complexity of allelic loss with progression to higher grade. Furthermore, by examining a large number of specimens for a large number of individual markers, we have identified statistically significant and marginal correlations between individual loci and age, histology, and location, suggesting that genetic differences may underlie the genesis of tumors with respect to age of onset, different histologies, or different locations. Of particular diagnostic and clinical relevance, we identified unfavorable prognostic significance for D10S179 (P = 0.001) or D10S169 (P = 0.004) and possibly D10S89 (P = 0.009) and D10S580 (P = 0.01), which predicted higher tumor grade; D10S209, which may predict shorter survival (P = 0.06); and D10S169, which may predict shorter time to recurrence (P = 0.01). These markers have the potential to determine the clinical approach to the treatment of patients.

We thank Lisa Randazzo for setting up and running the CEQ 2000 XL, Dr. James L. Fisher for database management, Debra Green for data entry, and Cheryl Spoutz, Lisa May, and Jennifer McCarthy for data abstraction. S. A. R. is indebted to the Levy family for their continued support of the Barbara Jane Levy Laboratory of Molecular Neuro-Oncology.