Distribution of Copy Number Variants and Impact of Chromosome Arm Call Thresholds for Meningioma

doi:10.21203/rs.3.rs-3986481/v1

Download PDF

Article

Distribution of Copy Number Variants and Impact of Chromosome Arm Call Thresholds for Meningioma

https://doi.org/10.21203/rs.3.rs-3986481/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Chromosome-arm copy number variants (CNVs) are an important component of cancer molecular classifiers. CNVs are often translated into binary chromosome arm calls (arm gain/loss) using an arm call threshold before integration into classification schemes. However, substantial variability exists in thresholds used to define arm calls from CNV data. We analyzed 1042 meningiomas with whole-genome microarray data and 12 meningiomas with multifocal sampling to characterize how CNV thresholds influence molecular classification and prognostication. Changing arm call thresholds shifted the association of chromosomal arm calls with meningioma recurrence in an arm-dependent manner and upgraded 21.5% of cases from low-grade to high-grade in a molecularly Integrated Grade (IG) scheme. The impact of threshold differences in IG prediction of recurrence was most evident amongst intermediate grade (IG-2) tumors and CNV call thresholds approaching whole-chromosome arm length (> 95%). The designation of chromosome loss or gain remained stable across a majority of thresholds, although this varied in a chromosome-dependent manner. CNVs fluctuated among paired primary-recurrent tumors, mostly growing on recurrence, but clustered in discrete sizes within a tumor. Appreciation of the impact of chromosome arm call thresholds can help ensure robustness of molecular classification paradigms.

Biological sciences/Cancer/Cancer genomics

Biological sciences/Cancer/CNS cancer

molecular classification

genomics

meningioma

copy number variants

chromosome arm loss

chromosome arm gain

The introduction of molecular profiling has transformed our ability to classify and predict the outcome of patients with both solid and liquid cancers.¹ Molecular classification schemes incorporate information on genome copy number variants (CNVs), structural variation, gene mutations, and/or DNA methylation in a tumor-dependent manner.^2–7

CNVs reflect gains and losses of DNA across the genome,^8,9 which can occur on any chromosome arm and vary considerably in size.^10,11 Rather than using the precise chromosomal location and/or size of CNVs, molecular classifiers commonly convert this information into a binary chromosome arm call, also known as a broad gain or loss. CNVs are translated to chromosome arm calls by comparing the size of a CNV to an arm call threshold, defined by the research investigator or institution. When the size of the CNV exceeds this threshold, a chromosome arm call is generated. For example, with a threshold of 50%, if the size of a copy number gain/loss exceeds 50% of the arm, an arm gain/loss call is generated, while if the CNV affects < 50% of the chromosome arm, no arm call is generated.

Due to the lack of clarity regarding the minimum proportion of a chromosome arm that should be affected to consider it clinically meaningful, there is significant variability in the thresholds applied to generate chromosome arm calls across studies, ranging from 5–80%.^3,5,12 Additionally, clinical cytogenetics laboratories often use an absolute segment size, as opposed to a percentage of the chromosome arm, to report segmental chromosome arm alterations.¹³ The extent to which heterogeneity in thresholds affects interpretation of molecular classification schemes remains unknown.

Given this variability, we investigated the distribution and prognostic impact of CNVs in a large clinically annotated cohort of meningiomas, the most common primary central nervous system tumor in adults,¹⁴ for which robust CNVs have been shown to confer prognostic significance. Our analysis demonstrates the unique distribution of CNV gains and losses, as well as areas of CNV enrichment in the meningioma genome. We further explore CNV sizes in multifocal and primary-recurrent meningiomas, revealing the spatial and temporal dynamics of CNVs. Our findings reveal the value of defining arm call thresholds in meningioma molecular classification systems to allow reproducible genomic stratification across heterogenous groups of patients.

We profiled 1042 meningiomas using whole-genome microarrays and found that almost all cases (99.3%) had a detected copy number alteration. As such, we surmised that the determination of whether a chromosome arm is considered lost or gained would be significantly influenced by the chosen size threshold. To investigate this, we evaluated two CNV arm call thresholds commonly reported in the literature, 5% and 50%, and observed a significant shift in the prevalence of arm losses and gains (Fig. 1A-B). There was significant decrease in both arm loss calls (affecting 26 out of 40 arms tested, all p < 0.05) and arm gain calls (affecting 24 out of 40 arms tested, all p < 0.05).

CNV Thresholds Influence Integrated Grade Assignment

To explore the cumulative effects of varying copy number calls as produced by different CNV arm call thresholds, we interrogated a copy number-driven molecularly Integrated Grade (IG) previously shown to identify patients at risk for recurrence more accurately compared to the WHO grade.⁵ We assessed the assignment of IG classification while varying CNV arm call thresholds in meningiomas from 883 patients (68.2% WHO grade 1, 28.9% WHO grade 2, 2.9% WHO grade 3) with complete whole-genome microarray data, CDKN2A/B loss status, and recorded mitoses. As CNV arm call thresholds increased, meningiomas tended to shift to lower IG grades (Fig. 2A).

The distribution of low (IG-1) versus high (IG-2/3) grade meningiomas significantly differed at the upper and lower ends of CNV arm call thresholds. Applying a 5% CNV arm call threshold resulted in significantly greater number of patients classified with an IG-2/3 meningioma compared to thresholds of 75% or 95% (both p < 0.02). The converse was observed with a 95% CNV arm call threshold, with a significantly smaller proportion of patients with an IG-2/3 meningioma compared to those at the 25%, 50%, and 75% thresholds (all p < 0.01). In total, 21.5% of patients shifted to a higher IG grade when moving from a 95–5% thresholds, with 14.3% of patients shifting from a low (IG-1) to high (IG-2/3) grade. Although meningioma IG designation did shift at the 25%, 50%, and 75% CNV arm call thresholds, the shifts were not statistically significant between these thresholds.

The chromosome arms contributing to the shift in IG grades across CNV arm call thresholds varied (Fig. 2B). From a 5–25% threshold, a reduction in 19q arm loss calls predominantly contributed to shifting patients to lower IG grades. 19p and 19q losses in particular showed a stepwise decrease in calls starting from the lowest arm call threshold. Other chromosome arms showed an incremental reduction over thresholds, such as 3p, while arms such as 18p significantly contributed to shifting IG grades only at the 95% threshold.

CNV Thresholds and Integrated Grade Recurrence Prediction

The impact of CNV thresholds on predicting meningioma recurrence was most evident in aggressive meningiomas and at the extreme ends of the CNV threshold spectrum. Recurrence was examined in 487 patients with a primary, gross-total resected meningioma with no prior radiation or chemotherapy and at least six months of follow-up (75.4% WHO grade 1, 23.8% WHO grade 2, 0.8% WHO grade 3). Patients were a median of 57 years old at surgery (range: 23–89 years), 72.6% were female, and 8.5% subsequently received adjuvant radiation. During the follow-up interval, 10.5% of patients had tumor recurrence (median follow-up: 4.7 years, range: 0.5–15 years).

Across all studied CNV arm call thresholds, there was no difference in PFS for benign (IG-1) meningiomas, suggesting robustness of the IG-1 classification (Fig. 2C, Supplement 1, p = 0.67). In meningiomas designated as IG-2, median PFS varied significantly between the 5% and 95% CNV call thresholds (p = 0.044, Fig. 2D, Supplement 1). In the most molecularly aggressive meningiomas (IG-3), median PFS showed variation between the 5% and 95% thresholds but was not statistically significant (56.1 versus 32.2 months, respectively; Fig. 2E, Supplement 1, p = 0.06). Amongst the patients who experienced a recurrence, when arm call thresholds were shifted from 95–5%, 19.6% were upgraded from low (IG-1) to high (IG-2/3) grade designation, which could potentially influence upfront clinical management. There was no significant difference in PFS between tumors classified at the 25%, 50%, and 75% CNV thresholds across IG-1, IG-2, and IG-3 meningiomas (Supplement 1).

Testing the capability of the IG classifier to predict recurrence across every CNV arm call threshold showed the best aggregate performance on time-dependent receiver operating curves when arm call thresholds were in the 38–42% range (Fig. 2F-H, Supplement 2). The highest CNV arm call thresholds (> 95%) led to the worst prediction of meningioma recurrence, noted even 1.5 years post-surgery. Brier scores over time also reflected this, with increasing prediction error seen at thresholds restricting broad arm calls to whole arm CNV changes (thresholds > 95%).

In practice, clinical cytogenetics labs report clinically significant chromosome arm calls based on absolute size thresholds measured in megabases (Mb), rather than as a fraction of total chromosome length. For example, our clinical cytogenetics laboratory uses 3 Mb of DNA as a threshold to call chromosome arm gains and losses. Across all three IG grades, there was no significant difference in PFS between tumors classified with absolute CNV size thresholds between 3 and 15 Mb (Supplement 1).

Meningioma CNVs Demonstrate Heterogeneity in Size

We delved into the copy number landscape of the meningioma genome to understand how CNVs across chromosomes impact shifts in molecular classification. CNVs 1) involved one or both arms simultaneously, 2) were small or large in size, and 3) were either solitary (affecting one contiguous chromosome segment) or multiple (involving non-contiguous segments across a chromosome) (Fig. 3A). Specific chromosomes displayed preferential CNV patterns, with distinct involvement of the p arm (e.g., chr 1 losses), the q arm (e.g., chr 10 gains), or both the p and q arms (e.g., chr 18 gains). Additionally, copy number gains more frequently involved alterations present on both the p and q chromosome arms compared to CNV losses (p < 0.05).

Copy number gains and losses in meningioma genomes demonstrated a pattern of clustering into either small CNV segments (≤ 10% of chromosome length) or large CNV segments (≥ 80% of chromosome length), rather than a continuous distribution throughout the chromosome (Fig. 3B-C). Across gains and losses, 55.8% affected ≤ 10% of the chromosome arm and 30.1% affected ≥ 80%, while only 14.1% affected between 10%-90% of the arm (both p < 0.001, Fig. 4). Notably, the majority of CNVs manifested as a continuous solitary segment of the chromosome arm, with fewer than a quarter of events involving ≥ 2 fragments. A majority of copy number losses affected large segments of chromosome arms (median: 51.2% of the chromosome arm, average: 48.3%) (Fig. 4A-B).

On a megabase scale, copy number losses affected a median of 22.8 Mb (average: 34.1 Mb) and gains affected a median of 1.8 Mb (average: 17.2 Mb), with 64.4% of copy number losses and 38.0% of gains altering more than 3 Mb of a chromosome arm (Fig. 4C-D).

Meningioma Copy Number Changes Show Distinct Localization

After filtering CNVs which are frequently observed in healthy patients,¹⁵ meningioma copy number gains and losses showed differential frequency and location distributions across the genome (Fig. 5). Copy number losses distributed across broad regions of the chromosome arm, with few discrete high-frequency CNV hotspots. Losses on many chromosome arms followed a similar pattern, with a greater frequency of meningiomas sustaining copy number losses toward the telomeres (e.g., chromosome 1p, 6q, 19p). An exception was chromosome 9p, which had an enrichment of copy number losses in a region including CDKN2A/B and MTAP.

Regions of copy number gains which enriched proximal to the centromere included chromosome 1p (NOTCH2) and 22q (immunoglobulin lambda light chain locus, MAPK1). Gains toward the telomere included 5p (TERT) and 9q (NOTCH1). However, some regions of high-frequency gains, such as that in the proximal segment of 14q (TRAC), likely represent normal polymorphisms driven by frequent segmental duplications.^16,17

CNV Thresholds Affect Chromosome Arm Calls

We next sought to understand how CNV thresholds impact designation of chromosome arm calls. By incrementally adjusting CNV arm call thresholds, we identified threshold ranges where arm calls exhibited significant variations. Determination of arm-level loss remained stable between CNV thresholds of 5–95% for most chromosome arms, while changing precipitously between thresholds of 0–5% and 95–100% (Fig. 6A, Supplement 3). However, a few arms—namely 17p, 19p, and 19q—showed a steady decline in chromosome arm loss calls as thresholds were increased between 5–95%. Similarly, for gains, most arm calls remained stable between 5–95% thresholds (Fig. 6B, Supplement 3). Arms 21q and 22q showed steady declines of chromosome arm gain calls up to a 15% threshold.

Optimal Predictive CNV Thresholds are Arm Dependent

CNV arm call thresholds impacted the predictive value of chromosome arm calls for meningioma recurrence (Fig. 7). Recurrence was assessed in 487 patients who had a primary, gross-total resected (GTR) meningioma without any prior radiation or chemotherapy and at least six months of follow-up. Out of the 39 arms tested for copy number losses, 30 arms were associated with meningioma recurrence at one or more CNV arm call thresholds on univariate testing. Across arms, the strength of association with meningioma recurrence varied as thresholds were altered. For instance, loss of chromosome 1p, the most commonly altered arm in aggressive meningiomas, showed increasing hazard ratios for recurrence as the CNV arm call threshold was raised toward 50% and declining hazard ratios at thresholds above 60–70%. By contrast, chromosome arms such as 6p had a stronger association with meningioma recurrence when nearly the entire arm was lost. This resulted in different “optimal” CNV call thresholds per arm: chromosome 1p loss demonstrated the highest hazard ratio for meningioma recurrence at a 46% CNV arm call threshold, while chromosome 18q loss showed the highest hazard ratio when an 87% CNV arm call threshold was applied (Supplement 4). Compared to copy number losses, copy number gains were less often significantly associated with meningioma recurrence (Fig. 7B).

Chromosome 1p Loss is Clinically Significant across CNV Thresholds

We specifically focused on 1p loss, the most common CNV in aggressive meningiomas, to further explore how varying CNV thresholds applied to an individual arm impacts patient risk stratification. Chromosome 1p loss was significantly associated with meningioma recurrence at almost all CNV thresholds < 98%, underscoring the clinical significance of small segmental 1p losses as prognostic markers for recurrence. (Fig. 7A, Supplement 4). Across CNV arm call thresholds between 5% and 95%, patients with chromosome 1p loss had significantly shorter median time to recurrence compared to those designated with WHO grade 1 (all p < 0.001, Supplement 5). Meningiomas with 1p loss designated at the 50% and 75% thresholds were associated with significantly shorter median time to recurrence compared to those designated with WHO grade 2 meningiomas (both p < 0.05) while other 1p loss thresholds approached significance (5%: p = 0.085, 25%; p = 0.079, 95%; p = 0.056).

CNV Sizes Change Incrementally in Recurrent Meningioma

Given the spectrum of CNVs associated with recurrence risk in meningioma, we explored how CNV sizes change as tumors evolve and whether the accumulation of CNVs contributes to aggressive meningioma behavior. We examined CNVs in 13 patients with paired primary and recurrent meningiomas from serial resections. The median time between resections was 4.4 years: 30.8% of primary and 61.5% meningiomas were high-grade (grade 2 or 3) by the WHO system. Using the IG system, 61.5% primary and 69.2% recurrent meningiomas were of high-grade (IG-2, IG-3).

In 12 out of 13 patients, there was an increase in CNV size of ≥ 8% on at least one altered chromosome arm between the primary and recurrent tumor (Fig. 8A, Supplement 6). The size of the CNV change appeared to associate with meningioma grade. When examining the distribution of the largest CNV size change present for each patient, individuals with more aggressive primary/recurrent meningiomas (WHO grade 2/3) had a significantly greater CNV size change (mean: 86.8%) compared to those who had a primary WHO grade 1 meningioma which remained WHO grade 1 on recurrence (mean: 23.3%, p = 0.015).

Despite these large events, 40.6% of all CNVs remained stable in size upon recurrence while 29.4% of CNVs changed only slightly (1–5% of the chromosome arm length), predominantly showing growth in size (Supplement 7). 30.0% of CNVs changed by > 5% of the chromosome arm over time; within these, around half (52.3%) represented de-novo copy number changes in the recurrent tumor which did not exist in the primary meningioma. This phenomenon likely reflects the cellular heterogeneity of meningioma and dynamic subclonal expansion and consolidation during tumor progression.¹⁸

CNV Sizes Cluster in Multifocal Sampled Meningiomas

We investigated whether CNVs manifest as distinct sizes across different regions within the same tumor by evaluating samples from 12 patients who underwent multifocal sampling of their meningiomas with a median of 8 spatially distinct samples (range: 2–8).^4,18 While the pattern of chromosome CNVs varied across samples taken from the same meningioma, losses and gains within each affected chromosome in a given tumor were observed to cluster in two to three size groups rather than across a spectrum of different sizes (Fig. 8B-C, Supplement 8).

Copy number variations (CNVs) are frequently observed in meningioma genomes, carrying implications for tumor biology and clinical outcomes. As molecular classification paradigms develop that rely on chromosome arm calls generated from CNVs, it becomes crucial to understand how CNVs are distributed across the genome, how they evolve over space and time, and the consequences of the definitions used to convert CNVs to chromosome arm calls. With nearly a quarter of meningiomas shifting in IG classification across arm call thresholds, our analysis demonstrates critical considerations surrounding the utilization of these thresholds.

Meningioma CNVs primarily affected small (< 10%) or large (> 80%) amounts of the chromosome arm. Altering CNV thresholds resulted in a change in the frequency of arm calls in an arm-dependent manner. While we observed relative stability in the IG molecular classifier, which aggregates multiple chromosome arms, the differential effect of arm call thresholds is likely a factor influencing the stability of molecular classification schemes relying on chromosome arm calls. As such, sensitivity testing across arm call thresholds can be a helpful step to ensure the reproducible translation of molecular classification schemes across institutions and patient populations.

The arm-dependent differences in CNV sizes and arm call thresholds affected the ability of arm calls to predict meningioma recurrence, with hazard ratios varying across arm call thresholds. Despite this, the IG meningioma classification scheme remained relatively stable in predicting meningioma recurrence, especially amongst IG-1 and IG-3 meningiomas, highlighting the strengths of this grading system. These groups reflect tumors with a rarer, small CNV changes or larger, more frequent CNV events respectively, and therefore are less affected by varying CNV arm call thresholds. Absolute size-based thresholds (Mb) showed a similar effect, with no difference in time to recurrence for meningiomas classified as IG-1 or IG-3.

By contrast, performance across arm call thresholds for IG-2 meningiomas varied. These meningiomas represent a more heterogenous group, which proven challenging to categorize across histopathologic and molecular classification schemes.^5,20,21 Some demonstrate aggressive high-grade behavior while others remain more indolent, despite apparent similarities in pathological appearance or molecular composition. When a higher arm call threshold was applied (95%), this subset of meningiomas showed a shorter time to recurrence than IG-2 meningiomas identified at a lower arm call threshold (5%). This likely highlights the presence of a subset of IG-2 meningiomas with large CNV changes that behave more aggressively than IG-2 meningiomas with smaller CNV changes.

Assessing the predictive stability of the IG classifier across grades and timepoints showed a deterioration in performance when very high CNV arm call thresholds were applied. Even at shorter term follow-up timepoints of one or two years, time-dependent receiver-operating curves showed notable decreased performance in predicting meningioma recurrence at thresholds above 80%. This drop in performance sustained over longer time horizons: up to 8 years in our analysis. Brier curves also reflected this, with very high CNV arm call thresholds showing consistently greater prediction error over time compared to lower CNV arm call thresholds. Therefore, restricting CNV arm calls to near whole-arm chromosomal alterations is likely not the best strategy to accurately capture patients at risk of meningioma recurrence over extended follow-up windows. Rather, for the IG classifier, if a single arm call threshold is required across all included chromosome arms, thresholds around 40% appear to maximize the predictive power of this classification scheme.

Analysis of multifocal sampled and primary-recurrent meningiomas supported geographic and temporal heterogeneity but also revealed discrete CNV size clusters, suggesting a predilection for CNV breakpoint sites across the genome.¹⁹ The consistency in CNV sizes across spatially distinct samples from the same meningioma demonstrates that, while there is intratumoral heterogeneity, CNVs are not steadily changing in size across subpopulations of tumor cells. Rather, there was greater variability in the collection of arms with CNVs in each sampled location. In paired primary-recurrent meningiomas, all patients had at least one significant chromosome CNV change in size, with more than half of these large CNV changes showing no alteration in the primary tumor. However, almost three-quarters of CNVs changed ≤ 5% in size between primary and recurrent meningiomas. This suggests that amidst overall stability, there are occasional dynamic changes in the extent of CNVs in recurrent meningioma tumors, potentially linked to genomic instability that contributes to tumor recurrence.

Mapping the locations of meningioma CNVs revealed unique patterns in distribution of gains and losses. For several high-risk chromosome arms such as 1p, 6p, and 10q, copy number losses appeared to enrich near the telomeres. This result is consistent with findings from other cancer types, where broad CNV losses frequently start near the telomeres – among the most gene rich regions of the genome – but are more variable in their length of extension across the chromosome arm.¹⁹ Yet certain chromosome arms such as 9p and 19q did show a higher frequency of losses in segments closer to the center of the arm, potentially indicating discrete regions of the meningioma genome under selection. Copy number gains did show some more focal hotspots, notably around the peri-centromeres and near the telomeres, such as on 1p and 5p. Of note, we observed a high frequency of CNVs in regions such T-cell receptor clusters which were filtered from the meningioma genome after comparison to CNVs found in the normal population. This underscores the background population-level variability of CNVs and complexity of isolating and assessing the clinical significance of CNVs linked to meningioma evolution.

The framework presented here demonstrates how CNVs and their translation into molecular classification schemes can be systematically investigated. However, this study does have certain limitations, many of which are associated with the modalities used for genomic profiling. While we can identify CNVs through whole-genome microarray and DNA methylation data, drawing definitive conclusions on how these CNVs impact gene expression is challenging. Combining this data with RNA sequencing or proteomics might help determine the biological consequences of CNVs enriched particular locations. Additionally, while our primary cohort was limited to cases evaluated at a single institution, we aimed to ensure sufficient power and incorporated publicly available multi-institutional meningioma data to capture the heterogeneity in tumor molecular alterations. Collectively, our analysis demonstrates a pathway for integrating CNVs into molecular classification systems beyond meningioma, a step toward improving prognostic models and identifying therapeutic targets.

Patient Cohort

We assembled a cohort of patients with meningioma evaluated at Brigham and Women’s Hospital (BWH) with resection performed between 2007–2021. Patients were selected if whole-genome microarray data was available from our institutional clinical cytogenetics laboratory. Demographic information, clinical history, and follow-up information on overall survival (OS) was extracted from the electronic medical record. Meningioma histopathologic grade was extracted from pathology reports, with histopathologic grade assigned by neuropathologists using the 2007 and 2016 WHO classification guidelines.^22,23 Mitotic count per 10 high-powered fields was also extracted from the clinical record.

Progression-free survival (PFS) was defined as time to radiologically evident first recurrence of meningioma. Follow-up MRI images were reviewed independently by neurosurgeons and neuroradiologists to evaluate for recurrence. GTR was defined after volumetric post-operative image contouring which demonstrated no residual tumor.²⁴ Paired primary-recurrent samples were identified if molecular profiling was performed on a patient’s primary and recurrent meningioma. This study was approved by the Institutional Review Board at BWH.

Chromosome Copy Number Analysis

Chromosome copy number was assessed using the Affymetrix OncoScan assay, a whole genome single nucleotide polymorphism (SNP) based microarray with 217,611 probes and an average resolution of approximately 300 kilobases. Chromosome Analysis Suite Software (ThermoFisher Scientific) was used to perform an in-silico comparison between the hybridization pattern of a patient specimen against a pooled reference sample set.

Segmented microarray data for each meningioma samples was processed using probe start and end locations to determine the size of CNVs in DNA base pairs. GRCh37/hg19 was used for probe localization and subsequent chromosome arm calculations. CNVs present in healthy control populations were filtered out using a validated consensus stringent CNV map.¹⁵ The size of each CNV was divided by the total profiled chromosome size and chromosome arm size to determine the percent of total chromosome affected and percent of chromosome arm affected. If multiple CNVs of the same type (e.g., loss or gain) occurred on the same chromosome arm for a given sample, fragment sizes were combined. Differences in CNV sizes were statistically compared using the Wilcoxon rank-sum test at a significance level of p < 0.05.

The locations of CNVs across the entire sample set were compiled, generating location CNV frequency maps of the meningioma genome. For each arm, a baseline arm CNV value was established using the median number of meningiomas with a CNV affecting a given arm. Location peaks in CNV were defined as any discrete clusters across a chromosome arm exceeding one and half times the baseline arm CNV value. The NCBI Genome Data Viewer (assembly GRCh37/hg19) was then used to identify genes in each location peak.

To understand how chromosome arm CNV thresholds impact chromosome arm calls, thresholds were incremented between 0-100% at 1% increments. The change in chromosome arm call frequency was plotted separately for CNV gains and losses across each chromosome arm. The first derivative of each plot was taken to determine thresholds which result in the greatest change of chromosome arm calls.

CNV arm loss calls, CDKN2A loss status, and mitotic count were used to calculate a Molecularly Integrated Grade (IG) for meningioma.⁵ IG was calculated at various chromosomal arm loss thresholds. Differences in low (IG-1) versus high (IG-2/3) grade categorization across chromosome arm loss thresholds was evaluated using Chi-squared with Bonferroni adjustment at a significance level of p < 0.05.

Multifocal Sampled Meningioma

Using a publicly available dataset generated at the University of California, San Francisco (UCSF), we assessed chromosome copy number profiles in patients who had multiple stereotactic samples collected for each meningioma.⁴ DNA was extracted from each sample and processed on the Illumina Methylation EPIC Beadchip. GRCh37/hg19 was used for probe localization. CNV profiles were generated from this DNA methylation data using the conumee package in R, using whole blood methylation profile data as a copy number normal control.²⁵ CNV intensity value distributions were then processed according to previously published mean segment intensities: segments with a mean intensity value of less than − 0.1 were defined as copy number losses and segments with a mean intensity value of greater than 0.15 were defined as copy number gains.³

The size of each CNV segment, determined using the probe start and end location, was used to calculate the percent of the chromosome arm affected by each CNV. CNV profiles derived from the sample meningioma sample were plotted together to assess spatial variability in CNV size.

Integrated Grade and Progression Free Survival

The impact of chromosome arm call thresholds on predicting PFS was examined in patients with primary meningiomas who had no prior radiation or chemotherapy treatment and if gross-total resection was achieved. Patients were included if at least six months of follow-up data was available. CNV calls were computed at each threshold from 0-100% in increments of 1%. At each threshold, a logistic regression model was fit to chromosome arm calls and meningioma recurrence data, generating an odds ratio for recurrence at each chromosome arm call threshold. Odds ratios along with 95% confidence intervals were plotted across arm call thresholds. Odds ratios at the point of collinearity, when infinitely large or small due to complete positive or negative correlation, were excluded.

Kaplan-Meir curves were generated across selected chromosome arm call thresholds to determine how the IG classification system differences in predicting meningioma recurrence across thresholds. Five arm call thresholds were selected based on percentage of an arm lost: 5%, 25%, 50%, 75%, 95%. Five arm call thresholds were also assessed based on the absolute size of an arm lost in megabases (Mb): 3 Mb, 6 Mb, 9 Mb, 12 Mb, and 15 Mb. Kaplan-Meier curves for meningioma PFS across various thresholds were compared within each IG using the log-rank test, with curves significantly different at p < 0.05.

The capability of the IG classification system to predict meningioma recurrence was examined across all CNV arm call thresholds using time-dependent receiver operating curves, time-dependent average precision, and Brier scores over time. Cox models were fitted for Brier score computation and all analyses were performed in R (timeROC, APsurv, brier).

Funding: Courtney Meningioma Research Fund; Fleming Meningioma Research Fund

Conflict of Interest: The authors declare no conflicts of interest.

Authorship: Study conception and design: RVP, WLB. Development of methodology: RVP, HSG, DM, WLB. Data acquisition, analysis, interpretation, visualization: RVP, HSG, DM, SR, EBC, RB, AHL, SS, WLB. Manuscript production: RVP, HSG, DM, SR, EBC, RB, AHL, SS, WLB. All authors had full access to the data included in this manuscript and approved submission.

Golub, T. R. et al. Molecular Classification of Cancer: Class Discovery and Class Prediction by Gene Expression Monitoring. Science (1979) 286, 531–537 (1999).
Nassiri, F. et al. A clinically applicable integrative molecular classification of meningiomas. Nature 597, 119–125 (2021).
Choudhury, A. et al. Meningioma DNA methylation groups identify biological drivers and therapeutic vulnerabilities. Nat Genet 54, 649–659 (2022).
Magill, S. T. et al. Multiplatform genomic profiling and magnetic resonance imaging identify mechanisms underlying intratumor heterogeneity in meningioma. Nat Commun 11, 4803 (2020).
Driver, J. et al. A molecularly integrated grade for meningioma. Neuro Oncol 24, 796–808 (2022).
Sahm, F. et al. DNA methylation-based classification and grading system for meningioma: a multicentre, retrospective analysis. Lancet Oncol 18, 682–694 (2017).
Maas, S. L. N. et al. Integrated Molecular-Morphologic Meningioma Classification: A Multicenter Retrospective Analysis, Retrospectively and Prospectively Validated. Journal of Clinical Oncology 39, 3839–3852 (2021).
Steele, C. D. et al. Signatures of copy number alterations in human cancer. Nature 606, 984–991 (2022).
Beroukhim, R. et al. The landscape of somatic copy-number alteration across human cancers. Nature 463, 899–905 (2010).
Bi, W. L. et al. Genomic landscape of high-grade meningiomas. NPJ Genom Med 2, 15 (2017).
Zack, T. I. et al. Pan-cancer patterns of somatic copy number alteration. Nat Genet 45, 1134–1140 (2013).
Hielscher, T. et al. Clinical implementation of integrated molecular‐morphologic risk prediction for meningioma. Brain Pathology 33, (2023).
Neveling, K. et al. Next-generation cytogenetics: Comprehensive assessment of 52 hematological malignancy genomes by optical genome mapping. The American Journal of Human Genetics 108, 1423–1435 (2021).
Ostrom, Q. T., Cioffi, G., Waite, K., Kruchko, C. & Barnholtz-Sloan, J. S. CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2014-2018. Neuro Oncol 23, iii1–iii105 (2021).
Zarrei, M., MacDonald, J. R., Merico, D. & Scherer, S. W. A copy number variation map of the human genome. Nat Rev Genet 16, 172–183 (2015).
Young, J. M. et al. Extensive Copy-Number Variation of the Human Olfactory Receptor Gene Family. The American Journal of Human Genetics 83, 228–242 (2008).
Iben, J. R. & Maraia, R. J. tRNA gene copy number variation in humans. Gene 536, 376–384 (2014).
Vasudevan, H. N. et al. Intratumor and informatic heterogeneity influence meningioma molecular classification. Acta Neuropathol 144, 579–583 (2022).
Shih, J. et al. Cancer aneuploidies are shaped primarily by effects on tumour fitness. Nature 619, 793–800 (2023).
Aghi, M. K. et al. Long-term recurrence rates of atypical meningiomas after gross total resection with or without postoperative adjuvant radiation. Neurosurgery 64, 56–60 (2009).
Aizer, A. A. et al. Extent of resection and overall survival for patients with atypical and malignant meningioma. Cancer 121, 4376–4381 (2015).
Louis, D. N. et al. The 2007 WHO Classification of Tumours of the Central Nervous System. Acta Neuropathol 114, 97–109 (2007).
Louis, D. N. et al. The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary. Acta Neuropathol 131, 803–820 (2016).
Huang, R. Y. et al. Proposed response assessment and endpoints for meningioma clinical trials: report from the Response Assessment in Neuro-Oncology Working Group. Neuro Oncol 21, 26–36 (2019).
Owen, B. M., Phie, J., Huynh, J., Needham, S. & Fraser, C. Evaluation of quantitative biomarkers of aging in human PBMCs. Frontiers in Aging 4, (2023).

There is NO Competing Interest.

SupplementaryContentNatureComms.docx
Supplementary Content 1-8

Download PDF

Version 1

posted

You are reading this latest preprint version

Distribution of Copy Number Variants and Impact of Chromosome Arm Call Thresholds for Meningioma

Status:

Version 1

Abstract

Figures

Introduction

Results

CNV Thresholds Influence Integrated Grade Assignment

CNV Thresholds and Integrated Grade Recurrence Prediction

Meningioma CNVs Demonstrate Heterogeneity in Size

Meningioma Copy Number Changes Show Distinct Localization

CNV Thresholds Affect Chromosome Arm Calls

Optimal Predictive CNV Thresholds are Arm Dependent

Chromosome 1p Loss is Clinically Significant across CNV Thresholds

CNV Sizes Change Incrementally in Recurrent Meningioma

CNV Sizes Cluster in Multifocal Sampled Meningiomas

Discussion

Methods

Patient Cohort

Chromosome Copy Number Analysis

Multifocal Sampled Meningioma

Integrated Grade and Progression Free Survival

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1