Identification of a Progenitor Cell of Origin Capable of Generating

Home , Arachnoid trabeculae, Pia mater

Oncogene (2011) 30, 2333–2344 & 2011 Macmillan Publishers Limited All rights reserved 0950-9232/11 www.nature.com/onc ORIGINAL ARTICLE Identiﬁcation of a progenitor cell of origin capable of generating diverse meningioma histological subtypes

M Kalamarides1,2,3, AO Stemmer-Rachamimov4, M Niwa-Kawakita1,2, F Chareyre1,2, E Taranchon1,2, Z-Y Han1,2, C Martinelli1,2, EA Lusis5, B Hegedus6,8, DH Gutmann6 and M Giovannini1,2,7

1Inserm, U674, Paris, France; 2Universite´ Paris 7—Denis Diderot, Institut Universitaire d’He´matologie, Paris, France; 3AP-HP, Hoˆpital Beaujon, Service de Neurochirurgie, Clichy, France; 4Molecular Neuro-Oncology and Pathology Department, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA; 5Department of Neurosurgery, Washington University School of Medicine, St Louis, MO, USA; 6Department of Neurology, Washington University School of Medicine, St Louis, MO, USA and 7Center for Neural Tumor Research, House Ear Institute, Los Angeles, CA, USA

Meningiomas are among the most common primary affect older adults, particularly women, and are often central nervous system tumours in adults. Studies focused associated with significant morbidity (Louis et al., 2000). on the molecular basis for meningioma development are Histologically, these tumours exhibit a wide range of hampered by a lack of information with regard to the cell of histological appearances, with meningothelial, fibroblas- origin for these brain tumours. Herein, we identify a tic and transitional meningiomas comprising the most prostaglandin D synthase-positive meningeal precursor as common World Health Organization subtypes. the cell of origin for murine meningioma, and show that Compared with adult gliomas, relatively less is known neurofibromatosis type 2 (Nf2) inactivation in prostaglandin about the cell of origin of these tumours or the critical D2 synthase (PGDS) ( þ ) primordial meningeal cells, genetic changes that drive tumourigenesis. Insights into before the formation of the three meningeal layers, accounts the genetic aetiology of meningioma have derived from for the heterogeneity of meningioma histological subtypes. the study of individuals with the inherited cancer Using a unique PGDSCre strain, we define a critical predisposition syndrome neurofibromatosis type 2 embryonic and early postnatal developmental window in (NF2), in which 50% of affected individuals develop which biallelic Nf2 inactivation in PGDS ( þ )progenitor meningioma (Evans et al., 1992). The central role of the cells results in meningioma formation. Moreover, we identify NF2 gene in regulating leptomeningeal cell proliferation differentially expressed markers that characterize the two is underscored by the finding of biallelic NF2 gene major histological meningioma subtypes both in human and inactivation in 50–80% of sporadic meningiomas mouse tumours. Collectively, these findings establish the cell (Ruttledge et al., 1994; Gutmann et al., 1997). In oforiginforthesecommonbraintumoursaswellasa particular, NF2 loss is most frequently observed in the susceptible developmental period in which signature genetic fibroblastic histological subtype (Hartmann et al., 2006). mutations culminate in meningioma formation. Previously, we showed that biallelic Nf2 gene in- Oncogene (2011) 30, 2333–2344; doi:10.1038/onc.2010.609; activation in leptomeningeal cells of genetically engi- published online 17 January 2011 neered mice (GEM) is sufficient for meningioma development. Nf2 inactivation in leptomeningeal cells Keywords: neurofibromatosis type 2; arachnoid; brain of newborn Nf2 conditional knockout mice (Nf2flox2/flox2) tumours was accomplished by non-selective cellular targeting with intrathecal adenoviral Cre recombinase delivery (Kalamarides et al., 2002). Many of these mice devel- Introduction oped meningiomas, firmly establishing the critical role of the Nf2 gene in meningioma pathogenesis. However, Meningiomas account for approximately one-third of all several important questions remained unaddressed. primary central nervous system tumours and constitute First, it is unclear which cell type(s) during development the most common brain tumour in adults over 35 years and differentiation of the arachnoidal cell lineage of age. These leptomeningeal neoplasms usually represents the cellular target for Nf2 gene inactivation. Second, it is not known how Nf2 loss in these Correspondence: Current address: Dr M Giovannini, Center for leptomeningeal cells results in the development of the Neural Tumor Research, House Ear Institute, 2100W Third Street, distinct meningioma histological subtypes (Kros et al., Los Angeles, CA 90057, USA. 2001; Lee et al., 2006). Third, is there a specific time E-mail: [email protected] window during mouse development when Nf2 inactiva- 8Current address: Department of Thoracic Surgery, Medical tion results in meningioma formation? University of Vienna, Vienna, Austria Received 6 May 2010; revised 9 November 2010; accepted 27 November The purpose of this study was to generate a novel 2010; published online 17 January 2011 GEM strain to define the cell of origin of meningioma Heterogeneity of meningioma histological subtypes M Kalamarides et al 2334 and to explain the spectrum of histological subtypes. These are the first recognizable cells of the developing Herein, we identified the prostaglandin D2 synthase mesoderm-derived meningeal layer (Kamiryo et al., (PGDS) gene as a specific marker of arachnoidal cells of 1990). At E15.5, the extracellular space became enlarged rats, mice and human beings (Urade et al., 1993; around the brain, showing a reticular structure that Yamashima et al., 1997; Kawashima et al., 2001). We resembled the subarachnoid space (Figures 1e–g) leveraged this marker to develop a new transgenic Cre (McLone and Bondareff, 1975). At this stage of strain to target Nf2 inactivation to PGDS-expressing development, neural crest-derived undifferentiated cra- cells and showed that they are the meningioma cell of nial mesenchyme rostral to the Rathke’s pouch (tele- origin capable of giving rise to both fibroblastic and ncephalic meninges) showed no PGDS immunostaining meningothelial histological subtypes. Moreover, we (Figure 1e). Starting at E18.5, PGDS immunostaining show that biallelic Nf2 inactivation in PGDS ( þ ) progressed in a caudal to rostral direction along the meningeal progenitor cells must occur during a defined telencephalic meninges and was complete by postnatal developmental window, such that Nf2 inactivation in day 5 (PN5) (Figures 1d, h), when all arachnoidal cells PGDS ( þ ) meningeal cells after this period does not were PGDS ( þ ) irrespective of their embryological result in meningioma development. Finally, we used origin or location. In contrast, dura mater cells were several human microarray data sets to identify tran- negative for PGDS expression (Figure 1h). scripts differentially expressed in fibroblastic and Next, we characterized the pattern of PGDSCre meningiothelial meningioma subtypes, and validated a expression and activity using a LacZ reporter strain subset of these markers at the protein level in both (ACZL), in which the chicken b-actin promoter and human and mouse meningioma tumours. Collectively, loxP-flanked CAT gene are upstream of the LacZ gene these findings establish a PGDS ( þ ) meningeal cell as (Akagi et al., 1997). In the resulting PGDSCre;ACZL the cell of origin for the two major meningioma mice, meningeal progenitor cells and all their progeny histological subtypes, and define a susceptible develop- are identified by LacZ expression and positive X-gal mental window during which Nf2 inactivation leads to staining. We isolated embryos from these crosses and meningioma formation. scored the brain and spinal cord after staining the sections with X-gal. Between E15.5 and E17.0, the primitive interface zone between the dura mater (outer Results zone) and the arachnoid (inner zone) is formed (Angelov and Vasilev, 1989). At E18.5, blue cells were observed in PGDSCre is expressed at E12.5 in the mouse primordial both outer (dural border cells, DBCs) and inner (ABCs) meningeal cells that give rise to the multiple meningeal zones, whereas only the internal layer was positive for layers PGDS expression (Figures 2a and b). These findings In the adult arachnoid, PGDS is expressed by arachnoid indicate that both zones derive from PGDS ( þ ) barrier cells (ABCs), whereas its expression is negligible primordial meningeal cells. In PN3 pups, X-gal ( þ ) in arachnoid trabeculae and pia mater (Beuckmann cells were also found in the arachnoid at the skull base. et al., 2000). Similar to previous studies on human By PN5, PGDS ( þ ) cells were observed in the meningiomas (Kalamarides et al., 2008), we found that meningeal convexity, a structure that originates from both human and mouse meningiomas exhibit intense the neural crest (Le Lievre and Le Douarin, 1975) PGDS immunoreactivity (Supplementary Figure S1), (Figure 1h). In adult mice, arachnoidal cells were both suggesting that PGDS is a marker of normal and Cre immunopositive and X-gal positive. In addition, neoplastic arachnoidal cells. some oligodendrocytes and choroid plexus cells were To determine whether early Nf2 inactivation, before positive for both Cre and LacZ (Figures 2c, e and f). the separation of the primordial meninges into the three To test whether these PGDS ( þ ) meningeal convexity terminally differentiated layers, might result in menin- arachnoidal cells derive from the neural crest, we gioma formation in vivo, we generated mice in which Cre crossed human tissue plasminogen activator promoter recombinase was expressed from the endogenous PDGS (HtPA)Cre mice with ACZL reporter mice. This promoter to reproduce the spatial and developmental transgenic line expresses Cre recombinase under the pattern of PGDS expression in the leptomeninges. We control of the HtPA (Pietri et al., 2003). Using this Cre elected to employ a knockin approach, as PGDS line, X-gal ( þ ) cells emanate from all the known heterozygous knockout mice are viable with no obvious derivatives of cranial, vagal and trunk migratory neural phenotype (Eguchi et al., 1999). A nuclear-targeted Cre crest cells, including craniofacial structures and cranial recombinase gene was integrated into the PGDS coding ganglia, cardiac and endocrine derivatives, melanocytes, region by replacing exon 1 of the murine Pdgs gene peripheral and enteric nervous system. At E15.5, Cre- (Supplementary Figures S2 and S3). mediated recombination, as revealed by LacZ expres- To define precisely the onset and spatial distribution sion, was evident only in the neural crest-derived of PGDSCre expression in the various cell populations meninges along the convexity of the telencephalon of the developing meninges, we first analysed embryos (Figure 2d). In PN3 PGDSCre;ACZL mice, X-gal- and pups for endogenous meningeal PGDS expression. positive cells were observed in both the DBC and ABC. At E12.5, a single layer of mesoderm-derived undiffer- However, only ABCs were PGDS ( þ ) (Figure 2g), entiated cranial mesenchymal cells caudal to Rathke’s indicating that both layers derive from a common neural pouch was positive for PGDS expression (Figures 1a–c). crest precursor. We also monitored PGDSCre-induced

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2335

Figure 1 Spatial and temporal PGDS expression pattern. Sagittal sections through E12.5 (a–c) and E15.5 (e–g) embryos, PN1 (d) and PN3 (h) pups were immunostained with anti-PGDS antibodies. PGDS immunoreactivity was limited to the embryonic meningeal layer (white arrowheads) covering the spinal cord, the rhombencephalon (myelencephalon and metencephalon) surrounding pontine flexure (a, e). No positivity was observed in the meninges covering the telencephalon. At higher magnification (b, f), PGDS immunoreactivity was present in the meninges overlying the basioccipital cartilage and the rhombencephalon in the interpeduncular fossa, caudal to the Rathke’s pouch. At E12.5, a single layer of PGDS ( þ ) cells (PrM) indicate the earliest stage in the development of the mesoderm- derived meninges (c). At E15.5, the subarachnoidal space has developed. DM, atc and PM are PGDS (À) in contrast to the PGDS ( þ ) AL (g). At PN1, PGDS immunostaining showed limited positivity in the meningeal convexity, excluding the DM (d). By PN5, positivity was complete through a caudorostral dynamic progression (h). AL, arachnoid layer; atc, arachnoid trabecular cell; boc, basioccipital cartilage; Br, brain; bv, blood vessel; chp, choroid plexus; Di, diencephalon; DM, dura mater; Me, metencephalon; My, myelencephalon; pfl, pontine flexure; Pg, pituitary gland; PM, pia mater; PrM, primordial meninges; Rp, Rathke’s pouch; SAS, subarachnoid space; SC, spinal cord; SV, skull vault; Te, telencephalon.

Nf2flox2 recombination by polymerase chain reaction shown). The low levels of recombination in the brain DNA analysis of adult mouse brain, leptomeninges, and spinal cord samples is consistent with previous spinal cord, peripheral nerves, liver, kidney, lung, heart, published data (Hoffmann et al., 1996), and may spleen, ovary and testis. Only the leptomeninges showed indicate the presence of contaminating leptomeninges robust levels of recombination, confirming the specifi- in these tissue samples. Recombination was also city of this promoter for the arachnoid (data not detected in testis and ovary, suggesting germline Cre

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2336

Figure 2 Origins and fates of meningeal progenitor cells. Temporal and spatial expression of lacZ in X-gal-stained sagittal sections of PGDSCre;ACZL mouse embryos at E18.5 (a, b) and adult mice (f). At E18.5, b-galactosidase activity was present in immunostained PGDS ( þ ) (AL) ABC as well as in PGDS (À) DBC (DM). A representative section of the arachnoid and cerebral cortex of a control (c) and PGDSCre (e) adult mice immunostained with anti-Cre antibodies. Cre-expressing cells are indicated by black arrowheads in the AL. One oligodendrocyte (white arrowhead) shows Cre immunopositivity. X-gal staining shows b-galactosidase activity restricted to arachnoidal cells covering the cerebral cortex in adult PGDSCre;ACZL mice (f). (d, g) Temporal and spatial expression of lacZ in X-gal-stained sagittal sections of HtPACre;ACZL in embryos (E15.5) and in PN3 pups.(d) Neural crest-derived meninges of the convexity were identiﬁed by X-gal staining (AL). At that time in development, there were no PGDS-immunopositive cells. (g) Only ABCs were PGDS ( þ ) at convexity (AL), whereas DBCs were PDGS (À). AL, arachnoid layer; bo, basioccipital; br, brain; DM, dura mater.

activity. Consequently, we found that all offspring from PGDSCre;Nf2flox2/À were viable and their survival male or female mice heterozygous for the PGDSCre was comparable to their control Nf2flox2/À littermates allele had demonstrable Cre activity irrespective of (Figure 3). In PGDSCre;Nf2flox2/À mice, meningiomas whether they inherited the transgene paternally or with histological features reminiscent of World Health maternally. Therefore, only PGDSCre;Nf2flox2/À mice Organization grade I (benign) human meningiomas were could be used for further analysis. frequently observed (Table 1). Meningioma development was restricted to the skull base ventral to the PGDS promoter-directed biallelic Nf2 inactivation leads brainstem, a site where PGDS expression was consis- to the development of different meningioma histological tently observed during embryonic development. No subtypes meningioma or meningeal abnormalities were found in To determine whether PGDSCre-mediated Nf2 inacti- neural crest-derived meninges along the convexity. vation results in meningioma formation, we intercrossed Two major histological meningioma subtypes were PGDSCre;Nf2flox2/ þ mice with Nf2flox/flox mice to generate observed in these mice: meningothelial meningiomas PGDSCre;Nf2flox2/À mice. PGDSCre;Nf2flox2/À mice were were restricted to the arachnoid layer, whereas fibro- born at the expected Mendelian frequencies and blastic meningiomas were exclusively found emanating were monitored closely over a period of 15 months. from the DBC layer of the dura mater. Meningothelial

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2337 meningiomas were found in six of 16 (38%) mice meningiomas had the particular feature, not described in analysed (Figures 4a–f), presenting most commonly as human meningiomas, of arising and being restricted to nascent tumours composed of meningothelial cells the dura mater, leaving the arachnoid layer intact. All (tumourlets) (Figure 4e). In nascent meningiomas, the analysed fibroblastic meningiomas were negative for tumour cells resembled normal arachnoidal cells with PGDS immunostaining (Figure 4i). Diagnostic ultra- oval nuclei that, on occasion, showed a central clearing. structural features of fibroblastic meningioma, including Nascent meningiomas, located at the arachnoid level spindle-shaped cells with long processes, interdigitating attached to the inner surface of the dura mater, showed cell processes and scattered desmosomes whorling consistent PGDS immunostaining (Figure 4f). Vimen- around collagen nodules were found in one representa- tin, a universal marker of human meningiomas tive meningioma analysed by electron microscopy (Schwechheimer et al., 1984), was also expressed in (Figures 4j–l). Proliferation of cells in the pia mater 100% of the murine tumours (Figures 4b and c). In and hydrocephalus caused by impaired cerebro-spinal some cases, strong PGDS immunostaining was observed fluid flow were observed in 19% and 13% of the mice, within some regions of the tumour (patchy distribution), respectively. whereas in other tumours, only few scattered positive or In four mice (25%), meningothelial and fibroblastic weakly positive cells were present (Figure 4d). meningiomas were concomitantly present at the medial Fibroblastic meningiomas were found in six of 16 skull base. Careful inspection of the meningioma (38%) PGDSCre;Nf2flox2/À mice analysed (Figures 4g–i). subtypes arising in this GEM model revealed that the These tumours were frequently large. Microscopic two different histological subtypes were each spatially examination showed lesions composed of spindle- restricted to one meningeal layer: meningothelial me- shaped cells embedded in a rich collagen matrix. These ningiomas emanated from the arachnoid layer, whereas fibroblastic meningiomas arose from the dura mater. Other tumours included osteomas (81%), as previously reported in Nf2 hemizygous mice (Giovannini et al., 2000), and pituitary adenomas (69%), based on microscopic examination and staining properties (Periodic Acid Schiff staining). The pituitary tumours were negative for PGDS immunostaining.

Additional nullizygosity for p16Ink4a or p53 does not increase the number and malignancy grade of meningiomas in PGDSCre;Nf2flox2/À mice We have previously shown using our first-generation GEM meningioma model that additional hemizygosity for p53 did not modify meningioma frequency or progression in AdCre;Nf2flox2/flox2 mice, whereas p16ink4a Figure 3 Kaplan–Meier cumulative survival curves. homozygosity increased the frequency of meningioma PGDSCre;Nf2flox2/À (n ¼ 24), PGDSCre;Nf2flox2/À;p16ink4aÀ/À (n ¼ 22) development without changing malignancy grade. and PGDSCre;Nf2flox2/À;p53flox/À (n ¼ 20) mice were examined The survival of PGDSCre;Nf2flox2/À;p53 þ /À mice was significantly To further examine this genetic cooperativity, two flox2/À flox/À reduced compared with PGDSCre;Nf2flox2/À (***log-rank test, cohorts of PGDSCre;Nf2 ;p53 and PGDSCre; P ¼ 0.0001). Nf2flox2/À;p16ink4aÀ/À mice were observed over a period of

Table 1 Summary of the phenotypic consequences of Nf2 gene mutation in PGDSCre-expressing cells in vivo

Phenotypic abnormality PGDSCre; PGDSCre;Nf2flox2/À; PGDSCre;Nf2flox2/À; Nf2flox2/À (n ¼ 16) p16ink4a-/- (n ¼ 16) p53flox/- (n ¼ 14) n (%) n (%) n (%)

Meningothelial meningioma 6 (38%)a 8 (50%)b 0 Fibroblastic meningioma in dura mater 6 (38%)a 8 (50%)b 6 (43%) Proliferation of pia mater 3 (19%) 3 (19%) 0 MPNST 0 0 4 (29%) Hydrocephalus 2 (12%) 5 (31%) 8 (57%) Osteogenic tumour 13 (81%)c 14 (87%)c 11 (79%)d Pituitary tumour 11 (69%) 1 (6%) 2 (14%) Choroid plexus abnormality 0 0 4 (29%)e

Abbreviations: MPNST, malignant peripheral nerve sheath tumours; Nf2, neurofibromatosis-2; PGDS, prostaglandin D2 synthase. aFour mice with concomitant fibroblastic and meningothelial meningiomas. bThree mice with concomitant fibroblastic and meningothelial meningiomas. cOsteoma. dOsteosarcoma. eOne choroid plexus carcinoma and three choroid plexus hypertrophy.

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2338

Figure 4 Histological analysis of phenotypic abnormalities in PGDSCre;Nf2flox2/À mice. (a) Haematoxylin- and eosin-stained sections showing a typical meningothelial meningioma (arrowheads) overlying the TG. (b) Some areas are positive for vimentin, a marker used to stain human meningiomas. (c) At higher magnification, black arrowheads denote vimentin positivity in this meningioma. (d) PGDS immunoreactivity (white arrowheads) was observed in rare cells. (e) Small MM (black arrowheads) composed of typical tumoral arachnoidal cells with oval nuclei and central clearing disposed in lobules (white arrowheads), adjacent to dm. (f) This meningioma (arrowheads) displays multifocal intracytoplasmic PGDS positivity. (g) FM (arrowheads) is located at the skull base. (h) Higher magnification of an FM surrounded by a small MM containing blood vessels. (i) Only the MM (arrowheads) showed immunoreactivity for PGDS. (j–l) Electron microscopy study of the FM in (g), showing ultrastructural features diagnostic of FMs: few cell surface specializations and intermediate filaments, spindled cells separated by collagen bundles (l). White arrowheads (j) and red asterisks (k) indicate desmosomal intercellular junctions and complex interdigitating cell processes, respectively. br, brain; c, collagen nodule; dm, dura mater; FM, fibroblastic meningioma; MM, meningothelial meningioma; tg, trigeminal nerve.

15 months. The survival of PGDSCre;Nf2flox2/À;p53flox/À heterozygous p53 mutant mice (Donehower et al., mice was significantly reduced compared with 1992), and the strong oncogenic synergy between Nf2 PGDSCre;Nf2flox2/À (Figure 3). This reduced viability and p53 inactivation in mice is well established was attributable to the early development (mean age 4.5 (McClatchey et al., 1998; Robanus-Maandag et al., months) of highly aggressive osteosarcomas, malignant 2004). In contrast, the choroid plexus hypertrophy and peripheral nerve sheath tumours (MPNST) and choroid carcinoma are likely related to p53 rather than Nf2 loss, plexus carcinoma. Three cases of choroid plexus as choroid plexus tumours are frequently observed in hypertrophy were also observed. The finding of osteo- the context of the Li–Fraumeni (TP53) inherited cancer sarcomas and malignant peripheral nerve sheath tu- syndrome (Krutilkova et al., 2005). mours is in agreement with our previous reports of In contrast, the survival of PGDSCre; osteosarcoma formation in P0Cre;Nf2flox2/flox2;p53 þ /À Nf2flox2/À;p16ink4aÀ/À mice was similar to that of mice, and in AdCre;Nf2flox2/À;p53 þ /À mice with loss of PGDSCre;Nf2flox2/À mice. The frequency of meningioma heterozygosity for both Nf2 and p53 (Kalamarides et al., development was only slightly increased relative to 2002; Robanus-Maandag et al., 2004). These tumour PGDSCre;Nf2flox2/À mice; however, this trend did not types belong to the typical tumour spectrum of reach statistical significance (w2, P ¼ 0.06). Surprisingly,

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2339 no pituitary tumours were observed in PGDSCre; Nf2ﬂox2/À;p16ink4aÀ/À mice.

Late postnatal and adult adenoviral Cre-mediated Nf2 inactivation in arachnoidal cells does not induce meningioma formation To more precisely define the developmental window during which PGDS ( þ ) cells are susceptible to malignant transformation, we injected Cre adenovirus (adCre) into the cerebro-spinal fluid at the convexity or at the skull base of Nf2flox2/flox2 mice at PN10 (n ¼ 20) or at 6 weeks of age (adult) (n ¼ 20). We found that Nf2 inactivation in arachnoidal cells at either late postnatal or adult ages did not result in meningioma formation (no tumours at 12 months of age). In contrast, early postnatal Nf2 inactivation in arachnoidal cells generated different meningioma histological subtypes in 23% of adCre-injected PN1–2 pups (Kalamarides et al., 2002) (w2, P ¼ 0.02). These results show that Nf2 loss in embryonic or very early postnatal PGDS ( þ ) arachnoidal cells induces meningioma formation, whereas arachnoidal cells at later times of development are not susceptible to tumour initiation following Nf2 gene inactivation.

Gene expression profiling supports the existence of a common meningioma progenitor cell for both meningothelial and fibroblastic meningioma development To support the hypothesis that early Nf2 loss in a common meningioma progenitor cell leads to the generation of meningiothelial and fibroblastic meningioma, we used existing human microarray data sets for the these two histological subtypes. First, we applied standard clustering methods to generate an initial list of 31 genes with greater than a fourfold change and a false discovery rate of Po0.1 differentially expressed between fibroblastic (n ¼ 4) and meningothelial (n ¼ 11) menin- Figure 5 Gene expression profiling of human meningothelial and giomas (Figure 5 and Table 2). We then leveraged fibroblastic meningiomas. Heat map representation of fibroblastic previous data from Fevre-Montange et al. (2009) on 17 (red) compared with meningothelial (blue) meningiomas using 31 meningiomas (two meningothelial and three fibrous genes that are differentially expressed in the two histological subgroups (fold change 44 and false discovery rate, P 0.1). meningiomas) to identify an additional 32 genes with o differential expression in the two histological subtypes. Six of the genes in our initial list were contained in the second data set (highlighted genes in Table 2). From this mouse meningothelial meningiomas (Figures 6e and e0) set of six differentially expressed genes, we selected two and FBLN1 staining only in murine fibroblastic genes for proof-of-concept validation studies: adenoma- meningiomas (Figure 6f). These results show that, in tosis polyposis coli downregulated (APCDD1) was human, as in mice, histologically distinct meningioma relatively overexpressed, whereas fibulin 1 (FBLN1) subsets arise from a shared PGDS ( þ ) progenitor cell was relatively underexpressed in meningothelial menin- following Nf2 inactivation, and argue against the giomas relative to fibroblastic meningiomas, respec- existence of separate cells of origin for these different tively. Using an independent set of human histological subtypes. meningiomas, we found that the differential pattern of APCDD1 and FBLN1 immunostaining distinguished between meningothelial and fibroblastic meningiomas, consistent with the microarray data. Specifically, Discussion APPCDD1 staining was exclusively detected in meningothelial meningiomas (Figures 6a and a0), whereas The arachnoid is one of the three membranous FBLN1 staining was only found in fibroblastic menin- envelopes (meninges), along with the pia mater and giomas (Figure 6d). As seen in the human tumours, we dura mater, which surround the brain and spinal cord. observed APCDD1 staining only in PGDSCre;Nf2flox2/À The pia mater is the meningeal envelope that firmly

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2340 Table 2 Genes with differential expression (>4-fold) in human meningothelial meningiomas compared with ﬁbroblastic meningiomas Gene symbol Gene name Chromosome Fold changea Overlapb

ADAMTS5 ADAM metallopeptidase with thrombospondin type 1 motif, 5 21q21.3 6.7 No APCDD1 Adenomatosis polyposis coli downregulated 1 18p11.22 9.3 Yes APOD Apolipoprotein D 3q26.2-qter 5.6 No ATP8B1 ATPase, class I, type 8B, member 1 18q21.31 6.1 No CAV1 Caveolin 1, caveolae protein, 22 kDa 7q31.1 4.1 No CFH Complement factor H 1q32 À4.5 No COL14A1 Collagen, type XIV, alpha 1 8q23 5.3 No EDIL3 EGF-like repeats and discoidin I-like domains 3 5q14 4.0 No ENPP2 Ectonucleotide pyrophosphatase/phosphodiesterase 2 8q24.1 5.8 No FBLN1 Fibulin 1 22q13.31 À4.5 Yes FIBIN Fin bud initiation factor homolog (zebraﬁsh) 11p14.2 À5.8 No GRB14 Growth factor receptor-bound protein 14 2q22–q24 5.7 No IGFBP3 Insulin-like growth factor binding protein 3 7p13–p12 4.1 No LRRN4CL LRRN4 C-terminal like 11q12.3 5.7 No LUM Lumican 12q21.3–q22 17.3 No NNMT Nicotinamide N-methyltransferase 11q23.1 À7.1 No PALMD Palmdelphin 1p22–p21 9.8 Yes PID1 Phosphotyrosine interaction domain containing 1 2q36.3 À4.3 No POSTN Periostin, osteoblast-speciﬁc factor 13q13.3 22.8 No RAB27B RAB27B, member RAS oncogene family 18q21.2 4.9 No RCAN2 Regulator of calcineurin 2 6p12.3 4.2 No RELN Reelin 7q22 4.9 No RHOJ Ras homolog gene family, member J 14q23.2 5.5 No RSPO3 R-spondin 3 homolog (Xenopus laevis) 6q22.33 À6.2 No SHC4 SHC (Src homology 2 domain containing) family, member 4 15q21.1–q21.2 À7.8 No THBS1 Thrombospondin 1 15q15 À4.1 Yes TNC Tenascin C 9q33 À14.0 Yes TNFRSF11B Tumour necrosis factor receptor superfamily, member 11b 8q24 7.5 No TPD52 Tumour protein D52 8q21 4.2 No TSPYL5 TSPY-like 5 8q22.1 À4.4 No VIT Vitrin 2p22.2 À5.2 Yes

aA positive fold change indicates increased expression in meningothelial relative to fibroblastic meningiomas, whereas a negative fold change indicates increased expression in fibroblastic relative to meningothelial meningiomas. bOverlap indicates genes identified in the Fevre-Montange paper (Yes or No).

adheres to the surface of the brain. External to the pia neural crest (Le Lievre and Le Douarin, 1975; Couly mater is the arachnoid, which is comprised of two and Le Douarin, 1987), whereas the meninges covering regions: a loosely organized inner part with arachnoid the brainstem and spinal cord arise from a different trabecular cells delimiting the subarachnoid space filled embryological lineage, likely the mesoderm (Catala, with cerebro-spinal fluid and an outer layer of closely 1998). Recent studies using transgenic mice with neural packed cells with numerous cell junctions (desmosomes crest (Wnt1Cre)- and mesoderm (Mesp1Cre)-specific and tight and gap junctions) and a basement membrane promoters have shown that the meninges covering the that forms the ABC to restrict the movement of fluids telencephalon are of neural crest origin, whereas the and large molecular weight substances (Haines et al., midbrain meninges derive from the cephalic mesoderm 1993). The outermost of the three meninges is the dura (McBratney-Owen et al., 2008; Yoshida et al., 2008). In mater (or pachymeninx), a dense membrane that this report, we conclusively establish that the inner part envelops and protects the brain. The dura mater is of the dura mater (DBC) and the outer part of the composed of fibroblasts embedded in an extracellular arachnoid (ABC) derive from a common PGDS ( þ ) matrix rich in collagen. One unique population of precursor cell. Moreover, we show that this common elongated fibroblasts, DBCs, are present where the precursor cell is of mesoderm origin at the skull base innermost part of the dura mater becomes contiguous (PGDS ( þ ) meningeal cell), and neural crest derived at with the arachnoid (Vandenabeele et al., 1996). the cerebral convexity (telencephalic region). In addi- Integrative data from electron microscopy and tion, we show that PGDS serves as both an early marker immunophenotypic studies in normal arachnoid and of mesoderm-derived progenitor cells during late em- meningiomas suggest that arachnoidal cap cells are bryogenesis as well as a marker of mature ABCs likely to be the precursor cells of meningioma (Couly specialized in cerebro-spinal fluid production. and Le Douarin, 1987; Tohma et al., 1992); however, Importantly, our results show that Nf2 inactivation this cellular origin has never been experimentally alone in a PGDS-expressing precursor cell is sufficient evaluated. In addition, the origin of the meningeal for the development of the two predominant meningio- layers remains controversial. Using quail-chick chi- ma subtypes: meningothelial (PGDS ( þ )) and fibro- meras, all layers of the forebrain meninges as well as blastic (PGDS (À)) meningiomas. The co-existence of bone and overlying dermis originate from the cephalic meningothelial and fibroblastic meningiomas restricted

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2341

Figure 6 Differential APCDD1 and FBLN1 staining pattern in meningothelial and fibroblastic meningiomas. Expression of APCDD1 (a, c, e) and FBLN1 (b, d, f) was analysed by immunohistochemistry in paraffin-embedded sections of human meningothelial (a and b), fibroblastic (c, d), and in PGDSCre;Nf2flox2/À mouse meningothelial (e) and fibroblastic (f) meningiomas. Human: (a) APCDD1 immunopositivity in meningothelial cells forming lobules surrounded by negative thin collagen septae. (a0) Representative immunopositive tumour cells at higher magnification. (b) The same meningothelial meningioma is negative for FBLN1 expression. (c) Conversely, APCDD1 is not expressed in a typical fibroblastic meningioma, whereas (d) FBLN1 is strongly expressed. Mouse: (e) APCDD1 is strongly expressed in a meningothelial meningioma. (E0) Representative positive immunostaining in the tumour cells, but not in the adjacent fibroblastic meningioma (black arrowheads), is shown from a PGDSCre;Nf2flox2/À mouse. (f) FBLN1 immunopositivity in a fibroblastic meningioma (black arrowheads) is shown. An adjacent small meningioma presenting as a single whorl (white arrowheads) is immunonegative for this marker. to specific and distinct meningeal layers in used these markers to show that both of these PGDSCre;Nf2flox2/À mice allowed us to identify the meningioma subtypes develop from a shared PGDS progenitor cell for each meningioma subtype (DBCs ( þ ) meningeal progenitor cells following Nf2 inactiva- for fibroblastic meningioma and ABCs for meningothe- tion in vivo. This finding is consistent with the normal lial meningioma). These observations explain why fate of this progenitor cell, which undergoes lineage fibroblastic meningiomas show histological features commitment to form the two different meningeal layers similar to DBCs, whereas meningothelial meningioma (dura mater and arachnoid layer). cells are reminiscent of normal arachnoid cells (ABCs). Lastly, our experimental data support the hypothesis We conclude that differences between the two menin- that the developmental stage of arachnoid cells may gioma subtypes relate to their respective normal cell of partly determine their susceptibility to tumour-initiating origin, DBCs versus ABCs, which share a common alterations, such as Nf2 loss. In this regard, we found PGDS ( þ ) precursor cell. To strengthen this conclu- skull base meningiomas in PGDSCre;Nf2flox2/À and early sion, we employed human meningioma subtype micro- postnatal adCre;Nf2flox2/flox2 mice, in which Nf2 loss array data to discover two differentially expressed genes occurs in embryonic precursor cells present in the foetal that distinguish these two major histological subtypes in or neonatal meninges. In contrast, meningiomas did not both human and mouse meningiomas. Furthermore, we form following Nf2 inactivation when it occurred later

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2342

Figure 7 Proposed model of meningioma initiation in the mesoderm- and neural crest-derived meningeal lineages. In the mesoderm- derived meninges of the skull base, PGDS ( þ ) primordial meningeal cells (E12.5) undergo lineage commitment to form both DBCs (PGDS (À)) in the dura mater and ABCs (PGDS ( þ )) at E15.5. In PGDSCre;Nf2flox2/À mice, Nf2 loss in embryonic PGDS ( þ ) primordial meningeal cells induces the development of fibroblastic meningiomas in the dura mater and meningothelial meningiomas in the arachnoid layer. These subtypes were also observed following early postnatal (oPN5) Nf2 inactivation in early postnatal PGDS ( þ ) and PGDS (À) meningeal cells following adenoviral Cre injection (AdCre;Nf2flox2/flox2 mice). More differentiated cells including later postnatal (4PN5) and adult meningeal cells do not form meningiomas following Nf2 inactivation. In the neural crest-derived meninges of the telencephalon, a PGDS (À) neural crest-derived meningeal precursor cell gives rise to both DBC (PGDS ( þ )) and ABC (PGDS (À)). PGDS immunopositivity in ABC appears at E18.5 and is completed by PN5. Susceptibility to Nf2 inactivation resulting in meningioma development is restricted to a temporal window between PN1 and PN5 as determined by the absence of meningioma formation in P0Cre;Nf2flox2/flox2 (Nf2 inactivation in the primordial meninges of neural crest origin), and in PGDSCre;Nf2flox2/À and AdCre;Nf2flox2/À mice (Nf2 inactivation in mature neural crest-derived ABCs and DBCs of the convexity).

in postnatal development or in adult mice. Meningiomas evaluate targeted therapies for this common human were also not found at the convexity of PGDSCre; brain tumour. Nf2flox2/À mice, in which Cre expression induces Nf2 loss in more differentiated meningeal cells, or in P0Cre; Nf2flox2/flox2 mice expressing Cre in the neural crest lineage Materials and methods and derivative meningeal layer (Giovannini et al., 2000) (Figure 7). In the mouse, meningioma initiation requires Generation and genotyping of PGDSCre mutant mice Nf2 loss during a narrow window of prenatal and Germline chimeras (PGDSCrefloxGFPHygro/ þ ) were generated by perinatal development. Remarkably, in NF2 patients, injection of 10 mutant embryonic stem cells into C57BL/6 the tumour burden (including meningiomas) is usually blastocysts, and crossed with FVB/N mice to produce outbred defined at a young age, with fewer tumours arising de heterozygous offspring. The genotypes of all offspring were novo later in adulthood (da Cruz et al., 2000). Moreover, analysed by polymerase chain reaction or Southern blot the typical growth rate of benign meningiomas is slow analysis on tail-tip DNA. To generate PGDSCre mice, (Olivero et al., 1995; Nakamura et al., 2003), and most PGDSCrefloxGFPHygro/ þ mice were crossed with EIIACre deletor meningiomas remain asymptomatic throughout life transgenic mice (Lakso et al., 1996). In the deriving double (Vernooij et al., 2007), such that nearly half of all transgenic offspring XbaI–NdeI digested tail DNA, the PGDSCre allele (13.0 kb) was detected by Southern blot with meningiomas are incidentally discovered at autopsy probe A. Mice carrying the PGDSCre allele were subsequently (Nakasu et al., 1987). These observations suggest that crossed with FVB/N mice to segregate the mutant allele. meningioma initiation from a common PGDS ( þ ) progenitor cell may be an early event, whereas tumour development may take decades more to form by the Mice flox2/flox2 flox2/flox2 addition of genetic and/or epigenetic events. To obtain PGDSCre;Nf2 , Nf2 mice (Giovannini Collectively, we have generated a new GEM model et al., 2000) on the FVB/N background were bred to PGDSCre mice on a FVB/N/129Ola mixed background, and the for meningioma, and used this unique strain to address resulting PGDSCre;Nf2flox2/ þ mice were bred to Nf2flox2/flox2, several unresolved issues germane to meningioma Nf2flox2/flox2;p53flox/flox or Nf2flox2/flox2;p16ink4aÀ/À mice. The p53flox/flox pathogenesis. With the identification of a putative cell (Marino et al., 2000) and p16ink4aÀ/À (Krimpenfort et al., 2001) of origin and a developmental window for tumour mice were kindly provided by A Berns. All mice used for this initiation, we are now positioned to develop and analysis were predominantly on a FVB/N mixed strain, with

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2343 contemporaneous littermates serving as controls. Genotyping LKB8801 ultramicrotome and diamond knife, stained with was performed as described previously (Giovannini et al., Sato’s lead and examined in a Phillips 301 transmission 2000). FVB/N mice were purchased from Charles River electron microscope. Laboratories (Wilmington, MA, USA). Pathogens were tested on a quarterly basis, and all serologies tested were negative Microarray expression data throughout the study. All animal care and experimentation Meningioma gene expression data (GEO4780) were obtained reported herein were conducted in compliance with the from the Gene Expression Omnibus, containing 62 gene guidelines and with the specific approval of Institutional expression arrays (five Affymetrix U133A and 57 Affymetrix Animal Care and Use Committee of the French Department of U133-Plus 2.0 microarrays) as published previously (http:// Agriculture. www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc ¼ GSE4780) (Fevre- Montange et al., 2009; Stuart et al., 2010). Histological Statistics subtypes were available for 11 meningothelial and four Significant differences in survival and tumour development were fibroblastic World Health Organization grade I meningiomas. identified using w2 test; Po0.05 was considered significant. These data were imported into the Partek Genomics Suite version 6.5 (Partek Inc., St Louis, MO, USA), and analysis of Molecular analysis of tissues variance was employed to detect differentially expressed genes Genomic DNA was extracted from various tissues of 2-month- in meningothelial versus fibroblastic meningiomas, using a fold old PGDSCre;Nf2flox2/ þ mice and the recombined Nf2D2 allele change of greater than 4 and a false discovery rate of Po0.01. detected by polymerase chain reaction as described previously (Giovannini et al., 2000). Immunohistochemical validation Immunohistochemical validation was performed in formalin- Histopathology and immunohistochemistry fixed, paraffin-embedded sections using an independent set of Mice were euthanized by CO2 inhalation when seriously ill 10 meningothelial and 10 fibroblastic World Health Organiza- (rarely) or at 15 months, and a detailed necropsy performed. tion grade I human meningiomas and five meningothelial and Histopathological analysis and detection of b-galactosidase five fibroblastic PGDSCre;Nf2flox2/À mouse meningiomas. activity were performed as described previously (Kalamarides FBLN1 and APCDD1 immunohistochemistry were per- et al., 2002). For embryo analysis, freshly taken en toto formed using affinity-purified rabbit polyclonal antibodies embryos were fixed for 2 h in 0.2% glutaraldehyde–2% (FBLN1: Atlas Antibodies, HPA001612 (1:200); APCDD1 formaldehyde–PBS, sagittally sectioned and stained with X- Atlas Antibodies, HPA014468 (1:100)). gal before embedding in paraffin as described previously (Kalamarides et al., 2002). PGDS immunohistochemistry was performed on adjacent paraffin sections using affinity-purified Conflict of interest rabbit polyclonal antibodies (1:500, sc-14825; Santa Cruz Biotechnologies, Santa Cruz, CA, USA) and standard techniques (Dako, Carpinteria, CA, USA). Cre immuno- The authors declare no conflict of interest. histochemistry was performed as described previously (Giovannini et al., 2000). Acknowledgements Electron microscopy Paraffin-embedded tissue was cut into nine pieces, placed in We thank A Berns for p16Ink4a and p53flox mutant mice; S xylene overnight, rehydrated in graded ethanols (100–25%) Dufour and JP Thiery for HtPACre mice; Y Urade for the and then placed in glutaraldehyde. Subsequently, the tissue PGDS genomic fragment; M Pla and staff of the Institut was post-fixed in osmium tetroxide, stained with uranyl Universitaire d’Hematologie (IUH), Universite Paris 7, for acetate, dehydrated in graded ethanol solutions, infiltrated mouse housing; N Karboul for technical assistance; and M with propylene oxide/Epon mixtures, flat embedded in pure Catala and W Van Furth for helpful discussions. This work Epon and polymerized over night at 60 1C. One-micron was supported by Grants from the US Army Medical sections were cut, stained with toluidine blue and examined Research and Materiel Command (DAMD17-02-1-0645 to by light microscopy. The best section containing meningioma MG), James S McDonnell Foundation (to DHG and MG), was chosen to proceed for electron microscopy study and Association Neurofibromatoses et Recklinghausen, the Vin- trimmed accordingly. Thin sections were cut with an cent Buono Research Fund (to DHG and MG) and Inserm.

References

Akagi K, Sandig V, Vooijs M, Van der Valk M, Giovannini M, Strauss Part I: the ventricular system, meninges and choroid plexuses. Arch M et al. (1997). Cre-mediated somatic site-specific recombination in Anat Cytol Pathol 46: 153–169. mice. Nucleic Acids Res 25: 1766–1773. Couly GF, Le Douarin NM. (1987). Mapping of the early neural Angelov DN, Vasilev VA. (1989). Morphogenesis of rat cranial primordium in quail-chick chimeras. II. The prosencephalic neural meninges. A light- and electron-microscopic study. Cell Tissue Res plate and neural folds: implications for the genesis of cephalic 257: 207–216. human congenital abnormalities. Dev Biol 120: 198–214. Beuckmann CT, Lazarus M, Gerashchenko D, Mizoguchi A, Nomura da Cruz MJ, Hardy DG, Moffat DA. (2000). Clinical presentation of a S, Mohri I et al. (2000). Cellular localization of lipocalin-type group of NF2 patients to a tertiary referral skull base unit. Br J prostaglandin D synthase (beta-trace) in the central nervous system Neurosurg 14: 101–104. of the adult rat. J Comp Neurol 428: 62–78. Donehower LA, Harvey M, Slagle BL, McArthur MJ, Montgomery Jr Catala M. (1998). Embryonic and fetal development of structures CA, Butel JS et al. (1992). Mice deficient for p53 are developmentally associated with the cerebro-spinal fluid in man and other species. normal but susceptible to spontaneous tumours. Nature 356: 215–221.

Oncogene Heterogeneity of meningioma histological subtypes M Kalamarides et al 2344 Eguchi N, Minami T, Shirafuji N, Kanaoka Y, Tanaka T, Nagata A Marino S, Vooijs M, van Der Gulden H, Jonkers J, Berns A. (2000). et al. (1999). Lack of tactile pain (allodynia) in lipocalin-type Induction of medulloblastomas in p53-null mutant mice by somatic prostaglandin D synthase-deficient mice. Proc Natl Acad Sci USA inactivation of Rb in the external granular layer cells of the 96: 726–730. cerebellum. Genes Dev 14: 994–1004. Evans DG, Huson SM, Donnai D, Neary W, Blair V, Teare D et al. McBratney-Owen B, Iseki S, Bamforth SD, Olsen BR, Morriss-Kay (1992). A genetic study of type 2 neurofibromatosis in the United GM. (2008). Development and tissue origins of the mammalian Kingdom. I. Prevalence, mutation rate, fitness, and confirmation of cranial base. Dev Biol 322: 121–132. maternal transmission effect on severity. J Med Genet 29: 841–846. McClatchey AI, Saotome I, Mercer K, Crowley D, Gusella JF, Fevre-Montange M, Champier J, Durand A, Wierinckx A, Honnorat Bronson RT et al. (1998). Mice heterozygous for a mutation at the J, Guyotat J et al. (2009). Microarray gene expression profiling in Nf2 tumor suppressor locus develop a range of highly metastatic meningiomas: differential expression according to grade or histo- tumors. Genes Dev 12: 1121–1133. pathological subtype. Int J Oncol 35: 1395–1407. McLone DG, Bondareff W. (1975). Developmental morphology of Giovannini M, Robanus-Maandag E, van der Valk M, Niwa-Kawakita the subarachnoid space and contiguous structures in the mouse. M, Abramowski V, Goutebroze L et al. (2000). Conditional biallelic Am J Anat 142: 273–293. Nf2 mutation in the mouse promotes manifestations of human Nakamura M, Roser F, Michel J, Jacobs C, Samii M. (2003). The neurofibromatosis type 2. Genes Dev 14: 1617–1630. natural history of incidental meningiomas. Neurosurgery 53: 62–70; Gutmann DH, Giordano MJ, Fishback AS, Guha A. (1997). Loss of discussion 70-1. merlin expression in sporadic meningiomas, ependymomas and Nakasu S, Hirano A, Shimura T, Llena JF. (1987). Incidental schwannomas. Neurology 49: 267–270. meningiomas in autopsy study. Surg Neurol 27: 319–322. Haines DE, Harkey HL, al-Mefty O. (1993). The ‘subdural’ space: a Olivero WC, Lister JR, Elwood PW. (1995). The natural history and new look at an outdated concept. Neurosurgery 32: 111–120. growth rate of asymptomatic meningiomas: a review of 60 patients. Hartmann C, Sieberns J, Gehlhaar C, Simon M, Paulus W, von J Neurosurg 83: 222–224. Deimling A. (2006). NF2 mutations in secretory and other rare Pietri T, Eder O, Blanche M, Thiery JP, Dufour S. (2003). The human variants of meningiomas. Brain Pathol 16: 15–19. tissue plasminogen activator-Cre mouse: a new tool for targeting Hoffmann A, Bachner D, Betat N, Lauber J, Gross G. (1996). specifically neural crest cells and their derivatives in vivo. Dev Biol Developmental expression of murine beta-trace in embryos and 259: 176–187. adult animals suggests a function in maturation and maintenance of Robanus-Maandag E, Giovannini M, van der Valk M, Niwa- blood–tissue barriers. Dev Dyn 207: 332–343. Kawakita M, Abramowski V, Antonescu C et al. (2004). Synergy Kalamarides M, Niwa-Kawakita M, Leblois H, Abramowski V, of Nf2 and p53 mutations in development of malignant tumours of Perricaudet M, Janin A et al. (2002). Nf2 gene inactivation in neural crest origin. Oncogene 23: 6541–6547. arachnoidal cells is rate-limiting for meningioma development in the Ruttledge MH, Xie YG, Han FY, Peyrard M, Collins VP, mouse. Genes Dev 16: 1060–1065. Nordenskjold M et al. (1994). Deletions on chromosome Kalamarides M, Stemmer-Rachamimov AO, Takahashi M, Han ZY, 22 in sporadic meningioma. Genes Chromosomes Cancer 10: Chareyre F, Niwa-Kawakita M et al. (2008). Natural history of 122–130. meningioma development in mice reveals: a synergy of Nf2 and Schwechheimer K, Kartenbeck J, Moll R, Franke WW. (1984). p16(Ink4a) mutations. Brain Pathol 18: 62–70. Vimentin filament–desmosome cytoskeleton of diverse types of Kamiryo T, Orita T, Nishizaki T, Aoki H. (1990). Development of the human meningiomas. A distinctive diagnostic feature. Lab Invest 51: rat meninx: experimental study using bromodeoxyuridine. Anat Rec 584–591. 227: 207–210. Stuart JE, Lusis EA, Scheck AC, Coons SW, Lal A, Perry A et al. Kawashima M, Suzuki SO, Yamashima T, Fukui M, Iwaki T. (2001). (2010). Identification of gene markers associated with aggressive Prostaglandin D synthase (beta-trace) in meningeal hemangioper- meningioma by filtering across multiple sets of gene expression icytoma. Mod Pathol 14: 197–201. arrays. J Neuropathol Exp Neurol (e-pub ahead of print 11 Krimpenfort P, Quon KC, Mooi WJ, Loonstra A, Berns A. (2001). December 2010). Loss of p16Ink4a confers susceptibility to metastatic melanoma in Tohma Y, Yamashima T, Yamashita J. (1992). Immunohistochemical mice. Nature 413: 83–86. localization of cell adhesion molecule epithelial cadherin in human Kros J, de Greve K, van Tilborg A, Hop W, Pieterman H, Avezaat C arachnoid villi and meningiomas. Cancer Res 52: 1981–1987. et al. (2001). NF2 status of meningiomas is associated with tumour Urade Y, Kitahama K, Ohishi H, Kaneko T, Mizuno N, localization and histology. J Pathol 194: 367–372. Hayaishi O. (1993). Dominant expression of mRNA for prosta- Krutilkova V, Trkova M, Fleitz J, Gregor V, Novotna K, Krepelova A glandin D synthase in leptomeninges, choroid plexus, and et al. (2005). Identification of five new families strengthens the link oligodendrocytes of the adult rat brain. Proc Natl Acad Sci USA between childhood choroid plexus carcinoma and germline TP53 90: 9070–9074. mutations. Eur J Cancer 41: 1597–1603. Vandenabeele F, Creemers J, Lambrichts I. (1996). Ultrastructure of Lakso M, Pichel JG, Gorman JR, Sauer B, Okamoto Y, Lee E et al. the human spinal arachnoid mater and dura mater. J Anat 189(Part (1996). Efficient in vivo manipulation of mouse genomic sequences 2): 417–430. at the zygote stage. Proc Natl Acad Sci USA 93: 5860–5865. Vernooij MW, Ikram MA, Tanghe HL, Vincent AJ, Hofman A, Le Lievre CS, Le Douarin NM. (1975). Mesenchymal derivatives of Krestin GP et al. (2007). Incidental findings on brain MRI in the the neural crest: analysis of chimaeric quail and chick embryos. general population. N Engl J Med 357: 1821–1828. J Embryol Exp Morphol 34: 125–154. Yamashima T, Sakuda K, Tohma Y, Yamashita J, Oda H, Irikura D Lee JH, Sade B, Choi E, Golubic M, Prayson R. (2006). Meningothe- et al. (1997). Prostaglandin D synthase (beta-trace) in human lioma as the predominant histological subtype of midline skull base arachnoid and meningioma cells: roles as a cell marker or in and spinal meningioma. J Neurosurg 105: 60–64. cerebrospinal fluid absorption, tumorigenesis, and calcification Louis DN, Scheitauer BW, Budka H, von Deimling A, Kepes JJ. process. J Neurosci 17: 2376–2382. (2000). Meningiomas. In: Kleihues P, Cavenee WK (eds). Pathology Yoshida T, Vivatbutsiri P, Morriss-Kay G, Saga Y, Iseki S. (2008). and Genetics of Tumours of the Nervous System. IARC Press: Lyon, Cell lineage in mammalian craniofacial mesenchyme. Mech Dev 125: pp 176–184. 797–808.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene