Genomic hallmarks of depot medroxyprogesterone acetate-associated meningiomas

Abstract

Background:

Population-based studies have linked progestin exposure to increased meningioma risk. However, the molecular basis of meningiomas associated with depot medroxyprogesterone acetate (DMPA) – a common injectable contraceptive – remains undefined.

Methods:

We performed an integrated clinicopathologic and genomic analysis of meningiomas from 10 women with long-term DMPA exposure. Tumors underwent histopathological analysis, targeted sequencing, and DNA methylation profiling. Data were integrated with reference cohorts (Baylor and Heidelberg) and analyzed through classifier assignment, consensus clustering, copy number analysis, differential methylation testing, and dimensionality reduction.

Results:

DMPA-associated meningiomas were all newly diagnosed, WHO grade 1 tumors with a predilection for the anterior and central skull base (n=6). Nine patients harbored multiple meningiomas. Four experienced regression of untreated meningiomas following DMPA cessation, while five demonstrated stabilization. Histopathology demonstrated relative overrepresentation of metaplastic morphology, an uncommon meningioma subtype. All DMPA-associated meningiomas mapped to benign molecular groups, and most exhibited low copy number alteration burden. Targeted sequencing revealed enrichment for TRAF7 mutations (n=5), with no NF2 mutations detected. Eight tumors shared consensus cluster identity, with cohesive grouping on principal component analysis and t-distributed stochastic neighbor embedding. No differential methylation was identified at the progesterone receptor locus.

Conclusions:

DMPA-associated meningiomas represent a recognizable phenotype within the broader NF2-wildtype/TRAF7-enriched spectrum of benign meningiomas, characterized by chromosomal stability, a shared methylation profile, tumor multiplicity, and regression or stabilization following DMPA cessation. While derived from a small single-institution cohort, these findings provide a molecular framework for understanding progestin-associated meningioma biology, re-interpreting epidemiologic literature, and informing population-level risk stratification.

Lay Summary

Meningiomas are common brain tumors that occur more often in women than men. Long-term use of the injectable hormone-based contraceptive depot medroxyprogesterone acetate (Depo-Provera) has been linked to higher meningioma risk, but the underlying biology of these tumors has not been described. We performed a clinical and molecular analysis of meningiomas from ten women with long-term exposure. They shared a recognizable set of features: patients often had several tumors, tumors shrank or remained stable after stopping the medication, and the tumors had similar underlying molecular characteristics. These findings provide a foundation for understanding hormone-associated meningiomas and informing patient care.

Introduction

Meningiomas are the most common primary brain tumor in adults and are rising in incidence due to an aging population and greater use of neuroimaging.^1–3 They occur two to four times more commonly in women compared to men and frequently express progesterone receptor.^1,4,5 These epidemiologic and histologic findings have supported a decades-long interest in the role of hormonal signaling on meningioma growth and development.⁶ Early observations noted a relationship between meningioma growth and pregnancy, a state of increased endogenous progesterone.^7–10 More recently, several studies have implicated exogenous progestins (i.e. from contraceptives or hormonal therapy) with meningioma behavior, with multiple groups reporting regression of meningiomas after cessation of exogenous progestins.^11–16 In the last five years alone, several European studies using nationwide registries demonstrated a link between exogenous progestin use and meningioma risk.^17–21 These studies examined commonly-used progestins in Europe including cyproterone acetate, desogestrel, levonorgestrel, nomegestrol acetate, and chlormadinone acetate.²² In the United States, one of the most common forms of progestin therapy is depot medroxyprogesterone acetate (DMPA), an injectable, high dose, long-acting progestin.²³ Similar to European counterparts, two case control studies in large United States cohorts recently demonstrated an association between DMPA use and meningioma risk, with one study demonstrating a dose-dependent relationship.^24,25 These findings have significant implications for both brain tumor management and for broader public health, given the widespread use of progestin therapy on a global scale.

In parallel, targeted sequencing and DNA methylation have recently become integral parts of CNS tumor classification, including in meningiomas.²⁶ Early work identified NF2 as a driver mutation in 40-60% of all sporadic meningiomas, with an additional 20-25% harboring mutations in TRAF7.²⁶ Several recent studies developed prognostic classification schemes in meningiomas defined by combinations of methylation state, mutational profile, and chromosomal instability. Sahm et al. proposed six methylation groups (Heidelberg groups): benign-1 (Ben-1), benign-2 (Ben-2), benign-3 (Ben-3), intermediate-A, intermediate-B, and malignant.²⁷ Nassiri et al. identified four molecular groups (Toronto groups): MG1/Immunogenic, MG2/NF2-wildtype, MG3/Hypermetabolic, and MG4/Proliferative.²⁸ Choudhury and Chen et al. proposed three methylation groups (UCSF groups): merlin-intact, immune-enriched, and hypermitotic.^29,30 Patel et al. also identified three molecular groups (Baylor groups) using integrated exome, RNAseq, and copy number profiling: MenG A (NF2 wildtype, no cytogenetic changes), MenG B (NF2-deficient, low chromosomal instability), and MenG C (NF2-deficient, high copy number alterations).³¹ Bayley et al. subsequently demonstrated that the MenG framework is robust across platforms, showing that the same three biological subtypes can be identified using RNAseq, DNA methylation, and copy number profiling, and that integrating these modalities improves classification accuracy.³² Notably, all of these schemes include methylation and NF2 status as key differentiators for meningioma classification.

Despite these advances and mounting epidemiologic evidence, there are no studies characterizing the underlying genetic and epigenetic landscape of DMPA-associated meningiomas. Moreover, the optimal clinical management of these unique patients remains undefined, though withdrawal of progestins is known to result in some degree of tumor regression or stabilization. Better understanding of the biology of these tumors would have immediate implications for classification and clinical management. It could also provide another lens through which to re-analyze existing meningioma literature. In this study, we address this important gap and present the first integrated molecular analysis of a curated cohort of DMPA-associated meningiomas using histopathologic analysis, targeted next-generation sequencing, and genome-wide DNA methylation profiling. We hypothesized that DMPA-associated meningiomas would represent a subtype of benign meningiomas with predominantly NF2-wildtype biology and a shared methylation signature.

Materials and Methods

Patient cohort and clinical annotation

Inclusion criteria were females with long-term DMPA use, defined as greater than 10 years. All patients underwent surgery at a single tertiary referral center between 2019 and 2024 for histologically confirmed WHO grade 1 meningiomas. Electronic charts of included patients were reviewed for acquisition of relevant clinical data. DMPA cumulative exposure was determined by patient self-report at neurosurgery clinical encounters, which consistently documented DMPA history. This exposure was confirmed against electronic medication administration and pharmacy records when available. MRIs of all patients were reviewed to assess for presence of multiple meningiomas. Representative MRI snapshots were obtained for each patient using an institutional clinical picture archiving and communication system platform. For the four patients with documented regression of untreated meningiomas after DMPA cessation, volumetric analysis of measurable untreated meningiomas was performed using 3D Slicer (slicer.org). Full segmentation and classification details are provided in the Supplemental Methods. Pathology reports were reviewed to collect WHO grade, progesterone receptor expression, and Ki-67 percentage. Histopathologic subtyping was performed by a board-certified neuropathologist during clinical workflows. Progesterone receptor expression was independently quantified by two neuropathologists using a modified H-score.³³ Raw methylation files from reference cohorts were downloaded from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus using accession numbers GSE189521 (Baylor) and GSE109381 (Heidelberg).³⁴ Corresponding clinical annotations were downloaded from the original published reports.³²

Sex as a biological variable

All patients in the study cohort were female, consistent with the clinical indication for DMPA. Reference cohorts from Baylor and Heidelberg included males and females. Probes on sex chromosomes were filtered out during methylation data processing to minimize sex-associated bias.

DNA extraction, targeted next-generation sequencing, and methylation array

FFPE tumor specimens were reviewed and macrodissected by a board-certified neuropathologist. Specimens then underwent targeted sequencing and EPIC v2 methylation profiling using clinically validated institutional workflows. Additional specimen processing details are provided in the Supplemental Methods.

Processing of methylation data

DMPA, Baylor, and Heidelberg methylation datasets were independently normalized, harmonized across shared probes, quality filtered, batch-corrected, and analyzed in a unified beta value matrix using the ChAMP pipeline. Full preprocessing details are provided in the Supplemental Methods.

Methylation bioinformatic pipeline

Study cohort patients were assigned to Heidelberg, Baylor, and UCSF methylation groups using established classifier workflows. Copy number profiling was performed from methylation array data using SeSAMe and cross-validated against independent Heidelberg Epignostix classifier copy number profiles and targeted next-generation sequencing panels. Consensus clustering was performed using ConsensusClusterPlus. Differential methylation analysis was performed at the PGR locus and across the 11q22.1 cytoband. Pathway-level differential methylation was performed for two curated Gene Ontology gene sets (GO:0050847, progesterone receptor signaling pathway; GO:0032570, response to progesterone) using a composite median-centered methylation score per sample. Dimensionality reduction was performed using principal component analysis (PCA) and t-distributed stochastic neighbor embedding (t-SNE). Full details are provided in the Supplemental Methods.

RNA sequencing and gene expression analysis

Bulk RNA sequencing was performed on FFPE tumor specimens. Differential gene expression and pathway enrichment analyses were performed to compare TRAF7-wildtype and TRAF7-mutant tumors within the DMPA cohort. Full details are provided in the Supplemental Methods.

Computational tools, generative artificial intelligence, and figure generation

Portions of the R code and icons in the study schematic (Figure 1B, Supplementary Figure 1A) were assisted by Claude (Opus 4.7; Anthropic, San Francisco, CA) and ChatGPT (GPT-5.5; OpenAI, San Francisco, CA). All outputs were verified and finalized by the authors. Visualizations in Figure 1B and Supplementary Figure 1 were generated using BioRender (Figure 1B: https://BioRender.com/flj3pd6, Supplementary Figure 1: https://BioRender.com/o2gaadn). Figures were assembled in Adobe Illustrator (version 29.8.2). Copy number plots using the Heidelberg classifier were downloaded from the Epignostix website.³⁵ The authors take responsibility for all final text and figures.

Figure 1. Patient cohort and study design.

A) MRI and histologic data for patients included in depot medroxyprogesterone acetate (DMPA)-associated meningioma study cohort. Snapshots are representative preoperative MRI images of resected tumors. In many instances, these snapshots do not capture all meningiomas for each patient. Imaging was not available for patient DMPA-4. Corresponding representative hematoxylin & eosin-stained histology slides are included for each patient (10x magnification). B) Schematic demonstrating study design and integrative molecular analysis of DMPA-associated meningiomas. Females with long-term exposure to DMPA and surgically resected WHO grade 1 meningiomas were identified. Histopathologic analysis was performed. FFPE tumor tissue was obtained and analyzed for molecular profiling for each patient, including targeted next-generation sequencing (NGS) and DNA methylation array analysis. Methylation data from the study cohort (n=10) were processed using the ChAMP bioinformatic pipeline and analyzed alongside publicly available reference meningioma cohorts from Baylor (Bayley et al., 2022) and Heidelberg (Capper et al., 2018). Integrated analyses included mutation profiling, copy number analysis, consensus clustering, differential methylation analysis, and dimensionality reduction.

Statistics

All statistics were performed in R (version 4.5.1) using the packages described above. For differential methylation analysis, comparison of beta values between DMPA and reference cohorts were performed using Welch’s t-test and Wilcoxon rank-sum test with adjustment for multiple hypothesis testing using the Benjamini-Hochberg method. P values less than 0.05 were considered statistically significant.

Study Approval

This study was approved by our center’s Institutional Review Board (STUDY23050143).

Results

Patients with DMPA-associated meningiomas exhibit a unique clinical and histopathologic phenotype

The study cohort included 10 females who underwent surgery for histologically confirmed WHO grade 1 meningiomas at a single tertiary referral center between 2019 and 2024 (Figure 1A, Supplementary Table 1). Mean age was 44.2 years (range 32-54), and mean duration of DMPA use was 21.4 years (range 14-32 years) (Table 1). Nine out of ten patients had multiple meningiomas. Seven patients had actively received DMPA within the standard 13-week dosing interval at the time of neurosurgical consultation (Supplementary Table 1). Resected tumors most frequently arose along the anterior and central skull base (n=6, including n=1 tuberculum/planum, n=3 sphenoorbital, n=1 sphenoid wing, n=1 sphenocavernous/planum), though they were also found along the convexity and parasagittal/parafalcine regions (n=4). All patients harbored at least one skull base meningioma, including those that were not targeted for treatment. All tumors had positive progesterone receptor expression. Mean H-score was 237 (range 135-295), with eight of ten having strong expression (H-score > 200) (Supplementary Table 2). All tumors had low proliferative indices (Ki-67 range 1-6%).

Table 1.

Summary of demographic information and clinical characteristics for 10 patients with depot medroxyprogesterone acetate (DMPA)-associated meningiomas in the study cohort. For patients who underwent multiple surgeries, histological and molecular analyses were performed on specimens from the first surgery. Tumor location refers to the dominant portion of the resected tumor(s).

Characteristic	Value
Demographics
Patients (n)	10
Age (mean (range))	44.2 (32-54)
Ethnicity (n)
Caucasian	8
Black	1
Hispanic	1
Depot medroxyprogesterone acetate use
Any use > 10 years (n)	10
Duration of use (years, mean (range))	21.4 (14-32)
Tumor regression of untreated tumors after stopping progestin (n)	4
Multiple meningiomas present (n)	9
Multiple surgeries performed (n)	4
Tumor location (predominant location of resected tumor)
Skull base	6
Convexity	3
Parasagittal/parafalcine	1
WHO grade 1 (n)	10
Progesterone receptor expression (n)	10
Maximum Ki-67 % (n)
1	3
2	2
3	3
4	1
6	1
Histology
Metaplastic	3
Meningothelial with myxoid features	2
Meningothelial	2
Transitional with myxoid features	1
Transitional	1
Secretory	1
Follow up (months, mean (range))	30.2 (13-50)

Histopathology demonstrated relative enrichment for metaplastic morphology, an uncommon feature among benign meningiomas.³⁶ Three out of ten patients met criteria for metaplastic subtype (Table 1). An additional three tumors without frank metaplastic morphology also demonstrated a prominent myxoid background (n=2 meningothelial with myxoid features, n=1 transitional with myxoid features). The remaining four patients harbored meningothelial (n=2), transitional (n=1), and secretory (n=1) subtypes. Representative clinical, radiographic, and histologic findings are shown in Figure 1A and Supplementary Figure 1.

Four patients with multiple meningiomas had radiographic regression of untreated meningiomas after DMPA cessation. Mean volumetric reduction across eight measurable lesions at last follow-up was 44.5% (range 1.8% to 67% reduction). The interval time from baseline MRI to first documented regression ranged from 11.4 to 14.0 months across the cohort. Volumetric trajectories for measurable lesions are shown in Supplementary Figure 2; corresponding volumetric measurements are shown in Supplementary Table 3. The remaining five patients with multiple meningiomas showed tumor stabilization with no further growth.

Benign methylation groups in existing classifiers encapsulate DMPA meningiomas

To frame this analysis within the broader meningioma methylation literature, the 10 DMPA cases were first assigned to three well-established classifiers from the Baylor, Heidelberg, and UCSF groups.^29,32,37 These were selected based on public availability of raw data with clinical annotations as well as general acceptance in the literature. Using a random forest model trained on Baylor samples and MenG groups, all 10 DMPA tumors were found to be assigned to MenG A, suggesting benign, NF2 wildtype biology (Table 2).³² By uploading raw methylation data to the Heidelberg classifier, all DMPA cases were similarly assigned to benign Heidelberg methylation groups.³⁵ Seven samples were assigned to Heidelberg Ben-2, two were assigned to Ben-3, and one remained unclassified. Similarly, all DMPA cases were assigned to UCSF methylation classes using the publicly available classifier, and all were determined to be Merlin-intact, the molecular group with the best clinical outcomes (Table 2).

Table 2.

Molecular classification of depot medroxyprogesterone acetate (DMPA)-associated meningiomas across previously described Baylor, Heidelberg, and UCSF classification systems. Baylor MenG groups were assigned using a random forest classifier trained on designations from their paper (Bayley et al., 2022). Heidelberg (DKFZ) assignments were made using the web-based Heidelberg Epignostix CNS Tumor Classifier (v12.8) (https://app.epignostix.com/). Confidence for group assignments was considered high if the probability output from the respective random forest classifier was ≥0.90, moderate if it was ≥0.75, and low if it was ≥0.60. Probabilities <0.60 were considered unclassified. UCSF assignments were made using a publicly available online classifier (https://william-c-chen.shinyapps.io/MeninMethylClass_V2_450K_added/). All 10 DMPA-associated meningiomas were assigned to Baylor MenG A, mostly with high confidence for the assignment. The 10 samples were assigned to a combination of Heidelberg Benign-2 (Ben-2) and Benign-3 (Ben-3), though confidence for these assignments was frequently lower. Several samples had features of both Ben-2 and Ben-3, and one sample did not cluster to any Heidelberg group. All 10 DMPA-associated meningiomas were assigned to the Merlin-intact UCSF methylation class.

Sample	MenG Group	MenG A Probability	MenG A Confidence	DKFZ Group	DKFZ Ben-1 Probability	DKFZ Ben-2 Probability	DKFZ Ben-3 Probability	DKFZ Group Confidence	UCSF Methylation Class
DMPA-1	A	0.950	High	3		0.202	0.795	Moderate	Merlin-intact
DMPA-2	A	0.901	High	Unknown		0.506	0.493	N/A	Merlin-intact
DMPA-3	A	0.819	Moderate	2		0.654	0.325	Low	Merlin-intact
DMPA-4	A	0.965	High	2		0.993		High	Merlin-intact
DMPA-5	A	0.782	Moderate	2		0.844	0.155	Moderate	Merlin-intact
DMPA-6	A	0.962	High	3		0.301	0.698	Low	Merlin-intact
DMPA-7	A	0.957	High	2		0.999		High	Merlin-intact
DMPA-8	A	0.621	Low	2	0.173	0.679	0.130	Low	Merlin-intact
DMPA-9	A	0.923	High	2		0.936		High	Merlin-intact
DMPA-10	A	0.950	High	2		0.675	0.324	Low	Merlin-intact

Targeted sequencing reveals a TRAF7-enriched, NF2-wildtype mutational landscape

To assess the mutational landscape of DMPA-associated meningiomas, all tumors underwent targeted next generation sequencing with well-established GlioSeq (n=9) and Oncomine Plus (n=1) assays (Figure 2A). Both of these assays encompass key mutations known to be implicated in meningioma biology, including NF2, TRAF7, AKT1, PIK3CA, SMO, POLR2A, TERT, CDKN2A, SMARCB1, FGFR1, KLF4, and SUFU.^38,39 Five patients had TRAF7 mutations, all of which localized to WD repeat domains (Figure 2A–B). Two of the TRAF7 mutations were from skull base tumors while the other three were from convexity and parafalcine locations (in patients who also harbored untreated skull base meningiomas). One patient with a sphenocavernous and planum meningioma had mutations in PIK3CA and FGFR1, both of which localized to kinase domains. All identified mutations were missense. Four patients had no detectable mutations. Notably, no patients had mutations in NF2.

Figure 2. Targeted sequencing and copy number analysis of depot medroxyprogesterone acetate (DMPA)-associated meningiomas reveals NF2-wildtype biology, enrichment for TRAF7 mutations, and low copy number burden.

A) Oncoprint summarizing pathogenic mutations detected by GlioSeq and Oncomine panels across the study cohort. Half of the patients had mutations in TRAF7, while one patient had mutations in both PIK3CA and FGFR1. Notably, no patients had mutations in NF2. B) Protein domain-level mapping of detected mutations identified in targeted sequencing panels. TRAF7 mutations localized to WD repeat domains. Mutations in PIK3CA and FGFR1 occurred in kinase domains. All identified mutations were missense mutations. C) Analysis of arm-level copy number alterations across the study cohort using SeSAMe pipeline shows low copy number variation. Each row represents an individual patient in the study cohort and each column represents an autosomal chromosomal arm. Blue boxes indicate copy number loss, defined as log2 copy number ratio ≤ −0.25. Arm-level calls were cross-validated against both the Heidelberg Epignostix classifier and targeted sequencing panels (GlioSeq or Oncomine) and included only if present on at least two independent platforms. No copy number gains (defined as log2 copy number ratio ≥ 0.30) were identified. D) Representative genome-wide copy number profile for a patient with a bland copy number profile (DMPA-1). E) Representative genome-wide copy number plot for a patient with multiple copy number losses (DMPA-3).

Copy number analysis reveals general chromosomal stability

Given the known importance of copy-number burden and chromosomal instability on meningioma behavior, copy number variation (CNV) analysis was next performed on the DMPA cohort by loading raw methylation intensity data (IDAT) files into SeSAMe. In parallel, IDAT files were also uploaded to the online Heidelberg classifier, which provides CNV data produced in Conumee 2.0.³⁵ Interestingly, most samples demonstrated bland copy number profiles (Figure 2C). Seven out of ten cases had no copy number gains or losses (Figure 2C–D). Of the three patients with any CNV, two had copy number losses on 1p and 2p, and one had multiple copy number losses (1p/q, 2p/q, 7p/q, 13q, 18p/q, 22q) (Figure 2C, Figure 2E). The latter tumor with the most copy alterations was a predominantly non-skull base (right frontal convexity with anterior clinoidal component) tumor without any driver mutations. All CNV calls reported in the main figures were independently supported by at least two platforms; discordant SeSAMe-only DMPA-10 calls were excluded (Supplementary Figures 3–4, Supplementary Table 4). There were no arm-level gains noted.

Clustering analysis identifies a shared methylation signature

To ascertain methylation subgroup information, the DMPA cohort was combined with additional WHO grade 1 meningiomas from Baylor (n=90) and Heidelberg (n=66), yielding a harmonized dataset. Unsupervised consensus clustering of the 2,000 most variable CpG probes resolved two stable methylation clusters (k=2) (Figure 3A), supported by a Delta area curve demonstrating a sharp inflection point at k=2 (Figure 3B). Sensitivity analyses that repeated clustering with the top 5,000 and 10,000 probes returned k=2 as the optimum (Supplementary Figure 5A–B). Alluvial concordance plots demonstrated general cluster stability, with the same patients tending to cluster together whether using the 2,000, 5,000, or 10,000 MVPs (Figure 3C). Cluster heatmaps with k=2 showed clear partitioning. Eight out of 10 DMPA-associated meningiomas clustered together (Cluster 2), while two clustered separately (Cluster 1) (Figure 3D).

Figure 3. DNA methylation profiling of depot medroxyprogesterone acetate (DMPA)-associated meningiomas reveals a cohesive epigenomic group without differential methylation at the progesterone receptor locus.

A) Consensus clustering of combined WHO grade 1 meningiomas from the DMPA, Baylor, and Heidelberg cohorts identifies two stable methylation-defined clusters (k=2). The consensus matrix heatmap demonstrates high agreement within clusters (blue) and low agreement between clusters (white). B) Delta area plot depicting relative change in area under the cumulative distribution function curve across values of k shows a clear inflection point at k=2, supporting two as the optimal number of clusters. C) Alluvial plot illustrating cluster membership across cohorts when clustering is performed using the 2,000 (2k), 5,000 (5k), or 10,000 (10k) most variable probes (MVPs). Numbers within bars indicate numbers of patients. DMPA-associated meningiomas largely retain cluster identity across analyses, with limited cluster crossover observed primarily in the Baylor and Heidelberg cohorts. D) Heatmap of DNA methylation beta values from the combined cohort highlights distinct methylation patterns between the two clusters. Eight of ten DMPA-associated meningiomas clustered together, while reference cohort samples were evenly distributed across clusters. E) Principal component analysis (PCA) demonstrates separation of the two methylation clusters, with DMPA-associated meningiomas forming a relatively cohesive subgroup that preferentially aligns with Cluster 2. F) t-distributed stochastic neighbor embedding (t-SNE) similarly demonstrates separation of methylation clusters and relatively cohesive grouping of DMPA-associated meningiomas, concordant with PCA results. G) Butterfly plot comparing mean DNA methylation beta values for 15 shared CpG probes mapping to the progesterone receptor (PGR) locus across the DMPA, Baylor, and Heidelberg cohorts shows no significant differences in CpG-level methylation between the study and reference cohorts. Dashed vertical line at β=0.3 indicates positive methylation. H) Gene-level comparison of mean beta values for all genes along the 11q22.1 cytoband, including PGR and adjacent genes, demonstrates no significant regional methylation differences between DMPA-associated meningiomas and reference cohorts.

Dimensionality reduction plots illustrate cohesive grouping

To further compare methylation signatures between DMPA and reference WHO grade 1 meningiomas, dimensionality reduction analysis was performed on the combined cohort using parallel techniques of PCA (Figure 3E) and t-SNE (Figure 3F). PCA captures major axes of variance in a linear fashion with an emphasis on global structure, while t-SNE can reveal nonlinear structure in high-dimensional genomic data with an emphasis on local neighborhoods.⁴⁰ Notably, in both methods, DMPA meningiomas formed a relatively cohesive group (Figure 3E–F). This was more pronounced on the PCA plot, in which the DMPA meningiomas occupied a compact, intermediate region between the two clusters that preferentially aligned with Cluster 2. On the t-SNE plot, most samples also occupied a compact space situated between the two larger clusters. Together, these orthogonal methods suggest reproducible, shared biology among these tumors.

Progesterone receptor and 11q22.1 cytoband methylation are unchanged

Given the central role of progesterone in DMPA-associated meningiomas, DNA methylation of the progesterone receptor locus (PGR on chromosome 11q22.1) was compared between DMPA-associated meningiomas and reference WHO grade 1 meningiomas from Baylor and Heidelberg. A total of 15 shared CpG probes were identified across the study and reference cohorts that mapped to the PGR gene and passed quality control. No differences in methylation were identified at any of these CpG sites (Figure 3G). To assess for potential regional effects, this analysis was repeated with all other genes along the 11q22.1 cytoband based on Illumina EPIC array probe annotations (YAP1, C11orf70, KIAA1377, ANGPTL5, TRPC6, TMEM133, FLJ32810, and CNTN5). Similarly, no differential methylation was found at any of these genes, suggesting that hormone responsiveness in DMPA-associated meningiomas is unlikely to be mediated by differential methylation of PGR or nearby loci (Figure 3H). This negative finding persisted when reference cohorts were restricted to NF2-wildtype and TRAF7-mutant subgroups (Supplementary Figure 6). Pathway-level analyses of methylation data using two curated progesterone-related Gene Ontology gene sets (progesterone receptor signaling pathway and response to progesterone) similarly demonstrated no significant methylation differences between the DMPA and reference cohorts (Supplementary Figures 7–8).

Exploratory transcriptomic analysis of TRAF7-mutant versus TRAF7-wildtype tumors

Given the enrichment of TRAF7 mutations in our cohort, exploratory bulk RNA sequencing was performed comparing TRAF7-mutant vs. TRAF7-wildtype tumors in our cohort. There was no differential expression between the progesterone receptor or composite scores for the progesterone response pathway or PGR signaling pathway (Supplementary Figure 9A–C). Differential expression analysis identified a total of 148 upregulated and 126 downregulated genes at nominal p<0.05 (Supplementary Figure 9D–E).

Integrated molecular profiling highlights a coherent DMPA-associated meningioma phenotype

To synthesize key clinical and molecular features of DMPA-associated meningiomas, an integrated oncoprint was generated (Figure 4). This visualization highlights the clinicopathologic phenotype of multiple meningiomas, skull base predilection, tumor regression or stabilization after DMPA cessation, and relatively high frequency of metaplastic and myxoid morphology; it also illustrates the molecular features of TRAF7 enrichment, absence of NF2 mutations, low copy number burden, and a shared methylation signature.

Figure 4. Integrated clinical, histologic, molecular, and genomic features define a cohesive phenotype of depot medroxyprogesterone acetate (DMPA)-associated meningiomas.

Oncoprint-style summary integrating clinical features, histologic subtype, DNA methylation classification, targeted sequencing results, and copy number alterations across the DMPA cohort. Columns represent individual patients (n=10). Relevant clinical features include patient age, sex, WHO grade, skull base (SB) location of the sequenced tumor, presence of skull base tumors (including untreated lesions), multiplicity of meningiomas, and radiographic regression following cessation of DMPA. Histologic subtype was determined by a neuropathologist. Molecular features include methylation cluster assignment, Baylor MenG group, Heidelberg (DKFZ) subclass assignment probability (from random forest model), UCSF group, mutation status, and arm-level copy number alterations. DMPA meningiomas were uniformly WHO grade 1 and NF2-wildtype, with enrichment for metaplastic and myxoid histology, a predominance of TRAF7 mutations, and generally low burden of copy number alterations. All patients harbored skull base meningiomas, even when not clinically targeted for treatment. Four patients showed radiographic regression of untreated meningiomas after stopping DMPA. Collectively, these integrated clinical and molecular features highlight a recognizable, hormone-responsive phenotype within the broader NF2-wildtype/TRAF7-enriched benign meningioma spectrum.

Discussion

Meningiomas are common, female-predominant tumors that frequently express the progesterone receptor, suggesting a hormonal influence on tumor behavior. Several large case-control studies across multiple European countries and the United States have demonstrated a reproducible association between exogenous progestin exposure and meningioma risk.^17–21 These epidemiologic findings are supported by multiple institutional reports demonstrating regression of meningiomas after cessation of exogenous progestins.^11–16 Separately, there has been a recent surge of studies characterizing the molecular landscape of meningiomas, with a particular emphasis on mutation and DNA methylation analysis. Despite this mounting clinical evidence and the rising use of molecular diagnostics, the molecular hallmarks of DMPA-associated meningiomas remain undefined. To address this unmet need, we conducted the first integrated histopathologic, genetic, and epigenetic analysis of progestin-associated meningiomas in a curated cohort of 10 patients with long-term DMPA use. We hypothesized that these tumors would represent a subtype of benign meningiomas with a shared epigenetic signature and NF2-wildtype biology.

Key findings and interpretation

The clinical features of our DMPA cohort are consistent with prior descriptions of hormone-associated meningiomas. Resected tumors were uniformly WHO grade 1, exhibited low proliferative indices, and demonstrated a predilection for the anterior and central skull base.^{6,11,12,15,16,41} Even patients undergoing resection for convexity or parafalcine meningiomas also harbored skull base meningiomas. Multiplicity of meningiomas was common, and several patients experienced radiographic regression of untreated tumors following DMPA cessation. This regression aligns with prior literature and reinforces the concept that exogenous progestin exposure contributes meaningfully to tumor growth; it also furthers the clinical relevance of defining this phenotype at a molecular level.^{6,11,12,15,16,41}

Across three established methylation classification frameworks, all DMPA-associated meningiomas mapped to benign subgroups, with uniform assignment to Baylor MenG A and UCSF Merlin-intact and predominant assignment to Heidelberg Ben-2 or Ben-3 groups. These findings align with prior studies demonstrating a general overlap between Baylor MenG A, Heidelberg Ben-2/Ben-3, and UCSF Merlin-intact subgroups, all of which are generally characterized by favorable clinical behavior and limited chromosomal instability.¹ Indeed, our CNV plots demonstrated striking chromosomal stability across most of the cohort, with most tumors lacking detectable arm-level gains or losses. All three patients with any arm-level CNV had 1p loss, though two of these did not have co-occurring 22q loss, a pattern consistent with less aggressive behavior.⁴² Our CNV plots generated using the SeSAMe pipeline generally aligned with similar plots from the Heidelberg classifier and targeted sequencing panels.³⁷ Cross-validation led to exclusion of false-positive calls for one patient (DMPA-10) called by SeSAMe but neither by Heidelberg nor GlioSeq (Supplementary Table 4). This cross-validation increases confidence that our findings represent true biologic signal. This relative chromosomal stability distinguishes DMPA-associated meningiomas from copy number-driven meningiomas, including NF2-mutant subtypes, suggesting an alternative molecular mechanism underlying progestin-associated growth.

Elucidation of the mutational landscape of our cohort substantially strengthens the biological coherency of this DMPA-associated meningioma phenotype. A key consideration in meningioma classification is mutation of NF2, an important tumor suppressor gene on chromosome 22q that is mutated in approximately half of sporadic meningiomas.⁴³ Another mutation of interest is TRAF7, which is found in 20-25% of sporadic meningiomas, classically associated with skull base location, mutually exclusive with NF2, and frequently co-occurring with KLF4 and AKT1 mutations.⁴³ Interestingly, our sequencing panel showed that no patients had NF2 mutations, while half had TRAF7 mutations. Two of the five TRAF7-mutant tumors were skull base meningiomas, while the other three were from convexity and parafalcine locations (in patients who also harbored untreated skull base meningiomas). None of the TRAF7-mutant tumors had other co-occurring mutations. Meanwhile, one patient had both PIK3CA and FGFR1 mutations, and four had no detectable mutations. These findings generally align with prior work by Peyre et al. on a French cohort of progestin-associated meningiomas (with exposure to cyproterone acetate, megestrol acetate, and chlormadinone acetate), in which only 7.5% of the cohort had NF2 mutations while 40% had TRAF7 mutations and 35% had PIK3CA mutations.⁴¹ In a study of four patients with cyproterone acetate-associated meningiomas and documented regression, Passeri et al. similarly demonstrated presence of a PIK3CA mutation, though of the four patients in their series, two had NF2 mutations.⁴⁴ Differences between the present study with DMPA exposure and these prior studies on French populations predominantly using cyproterone acetate do raise the interesting question of whether there may be different molecular drivers depending on the specific progestin used. This will be an interesting area for future research. Ultimately, our study extends these observations by integrating mutational data with histopathology, genome-wide methylation analysis, and copy number profiling, providing orthogonal evidence that DMPA-associated meningiomas arise within an NF2-independent, TRAF7-enriched, benign molecular framework.

Unsupervised consensus clustering supported our hypothesis of a shared epigenetic signature among DMPA-associated meningiomas. Our identification of two clusters of WHO grade 1 meningiomas generally aligns with prior classification systems from Baylor, Toronto, and UCSF.^28,29,32 The shared cluster identification among eight out of ten patients in our DMPA cohort suggests a shared methylation signature. Our PCA and t-SNE plots corroborated our consensus cluster findings. On both plots, the DMPA-associated tumors occupied a cohesive intermediate space relative to other meningiomas in Cluster 1 and Cluster 2. Taken together, the concordance of our clustering and dimensionality reduction approaches argues against stochastic grouping and supports the notion of reproducible biology in DMPA-associated meningiomas. Of the two patients that clustered separately (DMPA-1 and DMPA-8), one (DMPA-8) had PIK3CA and FGFR1 mutations, which was unique in the cohort. This differential clustering may reflect a combination of these mutations, biological heterogeneity, or perhaps additional modifying factors not captured in the present analysis.

Despite uniform progesterone receptor expression, no differential methylation was identified at the progesterone receptor locus, nearby genes along its cytoband, or in curated progesterone-related signaling pathways. We postulate that hormone responsiveness could potentially arise downstream of receptor methylation, perhaps through changes in gene expression, chromatin remodeling, tumor microenvironment factors, or post-transcriptional processing. The lack of methylation is consistent with sustained responsiveness to withdrawal of progestins and also consistent with the lack of the hormone resistance that is seen in hypermethylated cancers. This is demonstrated by the reliable shrinkage or stabilization of DMPA-associated meningiomas once DMPA is withdrawn.

The overrepresentation of metaplastic features in this cohort represents an intriguing finding of uncertain significance, as this subtype is generally uncommon in the literature.³⁶ Given their rarity, there is insufficient evidence to establish a direct relationship between histologic subtype and hormone sensitivity; however, we hypothesize that these morphologic features could reflect downstream differentiation states arising in part from an underlying hormone-responsive molecular phenotype.

Taken together, we propose that the convergence of several features of this cohort may be specific to DMPA exposure rather than reflecting generic NF2-wildtype, TRAF7-enriched meningioma biology. These include tumor multiplicity, regression of untreated tumors after DMPA cessation, and the relative overrepresentation of metaplastic morphology. Other features of our cohort including skull base predominance, copy number stability, and benign methylation classification are consistent with the broader NF2-wildtype, TRAF7-enriched meningioma spectrum. At this time we therefore suggest that DMPA-associated meningiomas represent a clinically actionable phenotype within the broader NF2-wildtype/TRAF7-enriched spectrum.

Clinical and public health implications

Our data, coupled with robust related literature, warrant attention from clinicians, researchers, and policymakers to ensure optimal care for at-risk patients. While the proportion of all meningiomas attributable to progestins is not known, the widespread global use of progestins suggests that even a small attributable risk could translate to meaningful population-level impact. DMPA use is particularly common among underserved and high-risk populations.⁴⁵ We emphasize careful interpretation of our data and caution against overtly dissuading women with legitimate indications for progestin therapy from using it. In particular, our report focuses on DMPA and is not necessarily generalizable to all progestin-based therapies (which are generally much lower dose formulations). However, we do support the need for interdisciplinary guidelines to guide care of patients with meningiomas who also have indications for progestin therapy.

Clinicians may face a number of challenging clinical scenarios. Surgeons may consider delaying resection of otherwise operative meningiomas by first trialing cessation of exogenous progestins. For recurrent or residual meningiomas associated with progestins, clinicians may debate whether to re-treat (i.e. with surgery or radiation) or observe with progestin cessation. Though there are retrospective case series reporting response to synthetic progestin withdrawal, in the absence of prospective data, such decisions should be individualized and made in multidisciplinary contexts.⁴⁶ We do not advocate routine MRI screening of all DMPA users, but rather, heightened vigilance and tailored multidisciplinary surveillance for patients with known meningiomas who remain on progestin therapy. Ultimately, risk-stratified guidelines are needed, with the recognition that each patient must be treated on an individualized basis. A recent collaboration between European neurosurgical, endocrine, and gynecological societies offers an excellent framework for how to approach future efforts in an interdisciplinary, patient-oriented, and holistic manner.⁴⁷

Biologic drivers and need for further research

The present study represents a comprehensive molecular characterization across multiple orthogonal platforms, rather than a mechanistic dissection of how DMPA drives tumorigenesis. It remains uncertain if these tumors are specifically driven by TRAF7 mutations, methylation changes, or other downstream pathways. Another unanswered question is why there are differences in the degree of regression of untreated meningiomas after stopping DMPA. A relevant next step would be comparison of gene expression between DMPA meningiomas and controls. We were unable to feasibly perform this because our study cohort RNA was derived from FFPE tissue while references were generated from fresh-frozen specimens, leading to significant batch effects. This limits mechanistic interpretation. Further work incorporating transcriptomic, chromatin accessibility, single-cell, and phosphoproteomic approaches will be critical to defining involved pathways and clarifying whether progestin exposure initiates tumorigenesis or preferentially drives expansion of a pre-existing molecular phenotype that is exquisitely progestin-responsive.⁶ Insights from breast oncology suggest a plausible role for sex steroid reprogramming of enhancer architecture.⁴⁸ This work also raises the theoretical possibility of hormone receptor antagonism as a therapeutic avenue. Finally, the preponderance of NF2 wildtype status and TRAF7 mutations in our cohort, along with general widespread use of progestins in the global population, raises the hypothesis-generating possibility that some previously-reported NF2-wildtype, TRAF7-mutant meningiomas in the literature may have in fact been cases of progestin-associated meningiomas. The present study, coupled with multiple population-level studies on this recently-recognized clinical phenotype, offer a potentially exciting new lens through which to re-analyze previously characterized meningioma cohorts.

Limitations

This was a single-institution study with a small cohort size, given the relatively recent recognition of this clinical entity and emphasis on longitudinal characterization. Banked tissue was only available in FFPE form, which limited additional analytic approaches that would have required prospective collection of frozen tissue. There was no institutional control group with which to compare the clinical phenotype, though a prior, small case control study was used as the initial basis for this phenotype.⁴⁶ While integration with multiple large reference cohorts mitigates some analytic limitations, additional multi-institutional studies are needed to validate these findings. Analysis of the DMPA cohort is limited by the granularity of clinical data available for the reference cohorts; in particular, some Baylor and Heidelberg patients may have been on progestin therapy (though if true, this would bias data toward the null). The cross-platform probe intersection may have eliminated relevant CpG probes in the newer arrays. Mutations and methylation analysis were the focus of this study, given their central role in the meningioma literature; however, these genomic and epigenomic changes arise upstream of gene expression and protein translation and may not be a reliable surrogate for these changes in all cases. They also do not provide insight into the potential role of the tumor microenvironment on progestin responsiveness. Further multiomic and single cell studies on prospectively collected fresh frozen tissue are warranted to better understand these dynamics.

Conclusion

The present study represents the first integrated histopathologic, genetic, and epigenetic analysis of meningiomas associated with progestin exposure. The convergence of skull base predilection, multiple meningiomas, regression after progestin cessation, NF2 independence, TRAF7 enrichment, copy number stability, and a shared methylation signature support the characterization of DMPA-associated meningiomas as a clinically actionable phenotype within the broader NF2-wildtype/TRAF7-enriched spectrum. These findings establish a molecular foundation for understanding progestin-associated meningioma biology and offer a framework for future translational studies to improve patient care.

Key Points.

• DMPA-associated meningiomas are NF2-wildtype and enriched for TRAF7 mutations

• They are often multiple and chromosomally stable with a shared methylation profile

• DMPA cessation is associated with tumor regression or stabilization

Importance of the Study.

Population-based studies have demonstrated a robust association between progestin exposure and increased meningioma risk, generating substantial clinical and public health interest. Despite this expanding epidemiologic literature, the biological basis for progestin-associated meningiomas remains undefined. This gap is particularly notable for depot medroxyprogesterone acetate (DMPA), the most commonly prescribed injectable contraceptive worldwide. In this study, we present the first integrated clinicopathologic, genomic, and epigenomic analysis of meningiomas arising in the setting of long-term DMPA exposure. We demonstrate that DMPA-associated meningiomas represent a recognizable phenotype within the broader NF2-wildtype/TRAF7-enriched spectrum of benign meningiomas, characterized by chromosomal stability and a shared DNA methylation signature. These tumors also exhibit a distinctive clinical phenotype marked by tumor multiplicity, skull base predilection, and regression or stabilization following DMPA cessation. Taken together, these findings position DMPA-associated meningiomas as a cohesive phenotype within an established molecular framework and offer a new lens for interpreting emerging epidemiologic associations, with direct implications for patient counseling and population-level risk stratification.

Data Availability:

DNA methylation array data and RNA-seq data have been deposited in GEO under accession GSE335177. R code used for data processing and analysis has been uploaded to https://github.com/sakibhuq/DMPA-Meningiomas.