Expression of MATH1, a marker of cerebellar granule cell progenitors, identifies different medulloblastoma sub-types

Neuroscience Letters 370 (2004) 180–185

Expression of MATH1, a marker of cerebellar granule cell progenitors, identifies different medulloblastoma sub-types Ettore Salsanoa , Bianca Pollob , Marica Eolib , Maria Teresa Giordanac , Gaetano Finocchiaroa,∗ a

Department of Experimental Neurosciences and Diagnostics, Unit of Experimental Neuro-Oncology and Gene Therapy, via Celoria 11, 20133 Milano, Italy b Department of Clinical Neurosciences, Istituto Nazionale Neurologico Besta, Milan, Italy c Department of Neurosciences, University of Turin; Italy Received 28 May 2004; received in revised form 30 July 2004; accepted 9 August 2004

Abstract In order to look for genetic markers helpful for the biological risk stratification of medulloblastomas (MBs) we assayed by real-time PCR expression levels of the following genes: MATH1, encoding a critical transcription factor for the differentiation of cerebellar granular cells (CGCs); PEDF, that encodes a trophic factor for CGCs and is located in a region of frequent allelic imbalance in MBs; and BIRC5, encoding the antiapoptotic protein survivin, usually overexpressed in malignancies. Expression levels of TRKC, higher in MBs with a more favorable prognosis, were also studied. Twenty-three patients were considered: MATH1 expression was strong in 14/23 and undetectable in the others. PEDF was up-regulated in 8/23, TRKC in 9/23, and BIRC5 in 23/23. MATH1 expression was significantly correlated with adult age (p < 0.0001), tumor location in hemispheres rather than the vermis (p < 0.0004), and PEDF and TRKC up-regulation (p < 0.008 and p < 0.04, respectively). During development MATH1 is selectively expressed in the external germinal layer (EGL) of the cerebellum. Thus, MATH1 expression identifies a subgroup of MBs that derive from the EGL and arise during adult age into cerebellar hemispheres. MATH1 mRNA-positive MBs express high levels of PEDF and show a trend towards longer survival, in agreement with increased expression of TRKC. BIRC5 expression, which is strong in all MBs and absent in normal cerebellum, lacks any prognostic value but could be explored for selective targeting of therapeutic factors to MBs. © 2004 Elsevier Ireland Ltd. All rights reserved. Keywords: Medulloblastomas; MATH1; PEDF; TRKC; BIRC5; Real-time PCR

Medulloblastomas (MBs) originate in the cerebellum and are the most common malignant central nervous system (CNS) tumor of the childhood; a significant fraction, however, is diagnosed in adulthood [36]. The quest for biological markers helpful to investigate their embryological origin is active as well as that for markers relevant to the prognosis and the treatment of medulloblastomas [9,18]. Clinical parameters like age, extent of surgery and presence of metastases have been used for risk classification [53]. Histopathological analysis may provide a grading for medulloblastomas: large/cell anaplastic medulloblastomas recur and metasta∗

Corresponding author. Tel.: +39 02 2394 453; fax: +39 02 26681 688. E-mail address: [email protected] (G. Finocchiaro).

0304-3940/$ – see front matter © 2004 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.neulet.2004.08.053

size more frequently than others [22] and a review of a large series of patients indicated that anaplasia, identified in 24% of the cases, is significantly associated with shorter survival [15]. Medulloblastomas with extensive nodularity, on the other hand, have a more favorable prognosis [21]. The impact of molecular prognostic markers increased in recent years. High levels of TRKC expression have been associated with a good clinical outcome [27,43], while loss of heterozygosity (LOH) on the short arm of chromosome 17, increased expression of the ErbB2 receptor and c-Myc amplification are associated with diminished survival [23,26,30]. More recently the technology of DNA microarray has been used for gene expression profiling and prediction of clinical outcome of medulloblastomas [13,40]. New markers

E. Salsano et al. / Neuroscience Letters 370 (2004) 180–185

originated from these studies may become more powerful factors in predicting the clinical evolution of medulloblastomas [17,19]. Furthermore, several studies suggest that signaling pathways implicated into the cerebellar development such as the SHH and WNT act as important tumor suppressor pathways in medulloblastoma [32]. In this context, we have investigated the expression of MATH1, PEDF and BIRC5, three genes that may have a role in cerebellar development and are not significantly expressed in the adult normal cerebellum, to assess their role for the biological risk stratification of medulloblastomas. Using real-time PCR we also evaluated the expression of TRKC, a molecular marker previously investigated by in situ hybridization or immunohistochemistry. MATH1 encodes a basic helix-loop-helix transcription factor, specifically expressed in the granule cell precursors (GCPs), a committed neuronal cell population located in the external granular layer (EGL) of the developing cerebellum [6,25]. Math1 is essential for the genesis of cerebellar granule cells (CGCs) and controls their differentiation by regulating multiple components of the Notch signaling pathway [20]. In the absence of Math1, the mouse cerebellum develops but there is a selective loss of both the GCPs and their descendants, the post-mitotic CGCs [6]. Math1 expression is shut-off in post-mitotic CGCs and undetectable in the adult cerebellum [6]. However, Math1 is expressed in MBs arising in PARP-1/p53 double null mice where it has been suggested that it may act as an oncogene [50]. PEDF is a neuroprotective factor for cerebellar granular cells [45,46] and its gene is located on chromosome 17p13.3 [49], where loss of heterozygosity (LOH) is frequently detected in MBs [7]. Moreover, PEDF encodes a powerful anti-angiogenic factor [14] and its expression is lost in glioma progression [28]. Survivin, encoded by BIRC5, is a member of the family of inhibitors of apoptosis and plays an important role in the control of cell division [2]. Its expression is up-regulated in most cancers [2], including those of the nervous system [42]. Expression levels of survivin, in particular, seem of prognostic value in gliomas [12]. Twenty-three MB specimens were obtained from patients who underwent surgery. Ten MBs occurred in pediatric patients (<16 years, 18 months to 14 years), whereas other tumors arose in adulthood (>16 years, 19–56 years). All pediatric MBs occurred in a midline or vermal location within the cerebellum, while most of the adult MBs were located in a cerebellar hemisphere. Only 2 of the 13 adult MBs had a midline location. Thirteen of the 23 MBs were nodular/desmoplastic, while 10 were diagnosed as classical MBs. Almost all the hemispheric tumors were of nodular/desmoplastic histology (9/11), while the midline tumors were commonly of classic histology (8/12) (Fisher’s Exact Test, p value = 0.0361). Twelve tumor specimens were frozen in liquid nitrogen after resection and stored at −80 ◦ C until RNA preparation. The others were archival, Carnoy-fixed, paraffin-embedded (CFPE) tumor specimens. The paraffin was removed from the

181

tissue by incubations in xylene at room temperature (20 min twice) and washing in ethanol (5 min × 3). Total RNA was isolated by homogenizing the tissue in the presence of at least 1 ml of EUROzol, following the protocols of the manufacturer (EuroClone). RNA from CFPE tumors usually required a clean-up with the Easy RNA kit (Qiagen). RNA yield was measured by spectrophotometer and analyzed by electrophoresis on 2% agarose gel. Two micrograms of total RNA were reverse-transcribed in a 40 ␮l reaction volume (including reverse transcriptase and DTT, added after DNase I treatment) containing 40 U of RNAse inhibitor (Ambion Diagnostic), 8 ␮l of 5× Expand Reverse Transcriptase Buffer (Roche Applied Science), 20 U of RNase-free DNase I (Roche Applied Science), 100 pmol of random hexanucleotides (Roche Applied Science) and 40 nmol of dNTPs. The reaction was incubated at 37 ◦ C for 45 min, and heated at 75 ◦ C for 5 min followed by 5 min at 65 ◦ C. The reaction was immediately cooled on wet ice. Two microliters of M-MuLV Reverse Trancriptase (50 U/␮l) (Applied Biosystem) and 4 ␮l of 100 mM DTT (Roche Applied Science) were then added, and reverse transcription was performed in an Applied Biosystem 9600 PCR System (10 ◦ C for 10 min, 42 ◦ C for 30 min, 95 ◦ C for 5 min, followed by 4 ◦ C soaking). To check for the presence of residual genomic DNA after DNase treatment, 3 ␮l of each RT mixture were added to 22 ␮l of PCR mixture, using the following primers for beta-actin (ACTB) amplification: 5 -CCCTGGACTTCGAGCAAGAG3 and 5 -CGGATGTCCACGTCACACTT-3 . The PCR mixture (25 ␮l) contained: 1× PCR Buffer II, 1.5 mM MgCl2 , 0.2 U of AmpliTaq DNA Polymerase (Perkin-Elmer), 1 mM of each dNTP, and 0.5 ␮M upstream and downstream primers. Cycling condition for ACTB included an initial denaturation step of 5 min at 95 ◦ C, followed by 95 ◦ C for 30 s, 55 ◦ C for 30 s, and 72 ◦ C for 30 s for 35 cycles. After 35 cycles, extension was performed at 72 ◦ C for 10 min followed by 4 ◦ C soaking. PCR reactions (10 ␮l) were analyzed by electrophoresis on 2% agarose gel. The remaining 37 ␮l of RT mixture were diluted at least 40 times and used for real-time PCR using the Assays-on-Demand reagents (Applied-Biosystems) for MATH1 (Hs00245453 s1), PEDF (Hs00171467 m1), BIRC5 (Hs00153353 m1), TRKC (Hs00176797 m1) and 18S (Hs99999901 s1). Gene-specific products were measured continuously in an ABI PRISM 5700 Sequence Detection System during 55 cycles. 18S rRNA was used for normalization. Expression values for tumor samples were compared with the expression in a normal cerebellum pool from 24 adult individuals (age 16–70; Clontech). The expression of all these genes was also evaluated in two normal brains (Clontech and Ambion). Before using the CT method for relative quantification, we performed a validation experiment to demonstrate that amplification efficiencies of the target gene and the reference marker (18S) were similar. The absolute value of the four slopes of log input amount versus CT was < 0.1.

182

E. Salsano et al. / Neuroscience Letters 370 (2004) 180–185

We investigated 23 MBs for the expression of MATH1, BIRC5, PEDF, and TRKC using real-time PCR. Archival, CFPE material from tumor specimen was a precious resource for RNA extraction and gene-expression studies since, using an easy and high-throughput method, we obtained small RNA fragments (under 500 bp, as previously described [16]), that could be used to quantify gene expression by real-time PCR. The quality of amplified products from CFPE material was similar to that obtained from frozen tissues as a starting material, as assessed by beta-actin amplification (not shown). Results are summarized in Table 1. MATH1 expression divided sharply MBs in two groups: one with high expression of MATH1 (n = 14) and one in which MATH1 expression was absent (n = 9). An example of sigmoid profiles obtained by real-time PCR of MATH1 is shown in Fig. 1. All lateral MBs (11/11) displayed very strong expression of MATH1, whereas only 3/12 midline located MBs showed such a strong expression (Fisher’s Exact Test, p value < 0.0004). All MBs (23/23) displayed uniformly strong expression of BIRC5, independently from their location or from patient age. PEDF mRNA was detected in all MBs, but the expression varied from weak to very strong. MBs were dichotomized into groups expressing high (7/23) or low (16/23) levels of PEDF by using the mean index value as a reference (overall trends were similar using the median index value). All MBs with high levels of PEDF expression also displayed very strong expression of MATH1 (Fisher’s Exact Test, p value = 0.019), Table 1 Clinical and molecular characteristics of MBs Age Pediatric (<16 years)

Adult (≥16 years)

10

13

10 0

2 11

7 3

3 10

MATH1 Yes (n = 14) No (n = 9)

1a 9

13 0

PEDFb Up (n = 8) Down (n = 15)

0 10

8 5

TRKCc Up (n = 9) Down (n = 14)

2 8

7 6

10 0

13 0

Medulloblastomas (n = 23) Location Vermis (n = 12) Hemisphere (n = 11) Subtypes MB (n = 10) DMB (n = 13)

BIRC5 Yes (n = 23) No (n = 0) a b c

Less than 2 years. Maximum PEDF/minimum PEDF = 1009; mean value = 27. Maximum TRKC/minimum TRKC = 178; mean value = 9.1.

Fig. 1. Real-time PCR analysis of MATH1 in human medulloblastomas (MBs) and normal adult cerebellum. The figure shows sigmoid curves of one MB not expressing MATH1 (#1, blue) and of another MB expressing high levels of MATH1 (#2, yellow). Amplification of MATH1 in control brain is also shown (purple). In all three experiments control amplification is performed with 18S.

and were more commonly of nodular/desmoplastic histology (7/8) (Fisher’s Exact Test, p value = 0.027). Using real-time PCR we have found highly variable levels of TRKC expression in MB specimens: as for PEDF we used the mean index value as a reference to identify tumors with high or low levels of TRKC expression. Interestingly, MBs with high TRKC expression usually express MATH1 (9/10) (Fisher’s Exact Test, p value = 0.04) and high levels of PEDF (Fisher’s Exact Test, p value = 0.0228). Our data confirm that human adult brain and cerebellum do not express or show very weak expression of MATH1 and BIRC5. Analogously, PEDF is weakly expressed in the adult cerebellum as previously found [47] while TRKC, encoding a marker of more mature and differentiated cells within the developing cerebellum, is expressed in adult normal cerebellum. Our study presents the evidence that MBs can be dichotomized into two subgroups on the basis of MATH1 expression (Table 2). All the MBs arising in the cerebellar hemispheres express high levels of MATH1, while the majority of vermal MBs do not. Since MATH1 is selectively expressed by GCPs of the EGL, this result implies that hemispheric and midline MBs have different histological origins, suggesting that laterally located MBs derive from the EGL, whereas vermal MBs originate from pluripotent progenitor cells of the ventricular matrix (VM) [32] or, in few cases, from the EGL: in our study 25% of midline MBs (3/12) express high levels of MATH1. Our results point to MATH1 as a robust molecular genetic marker that clearly dichotomizes MBs, delineating the cellular origin of one subset of MBs from GCPs of the EGL. The histogenesis of MBs has been controversial. Although many investigators believe that MBs originate from the EGL, other studies have proposed that MBs have more than one cell of origin [10,32]. Some authors believe that adult MBs derive from EGL, like cerebellar granule cells (CGCs), and that childhood MBs originate from the ventricular matrix,

E. Salsano et al. / Neuroscience Letters 370 (2004) 180–185

like Purkinje cells (PCs). However, no specific biomarker defines unequivocally the histogenesis of MBs. ZIC1, a potential CGC marker [52], is also expressed by more immature and undifferentiated neuroepithelial cells [5]. The P75 neurotrophin receptor (P75NTR), binding all neurotrophins with similar affinity, is present in CGCs as well as in the PCs [10,11]. The calcium-binding protein Calbindin-D28K (CalD) is expressed at variable levels in primitive neuroepithelial cells [33] and is considered as a marker of MBs originating from the VM [34], but one report also found CalD staining in CGCs during cerebellar development [51]. Differently from ZIC1 or p75NTR that are not exclusively expressed in the EGL, and from calbindin-D28K, that is accepted as a VM marker, MATH1 appears as a highly specific marker of the GCPs. Comparative studies have raised the possibility that pediatric and adult MBs are biologically different [24,41]. We have demonstrated that most of childhood MBs do not express MATH1 (9/10), while all the adult MBs express MATH1 (13/13). Thus, our data confirm the proposition that childhood MBs derive from VM neuroepithelial cells, while adult MBs originate from EGL cells. These data are also supported by the propensity of childhood MBs to arise in the midline, where VM neuroepithelial cells are located. A minority of pediatric MBs, however, could also originate from EGL cells, explaining why we have detected MATH1 expression in one of them. On the other hand, since EGL persists in the human cerebellum for 12–15 months after birth [1] and the patient was 18-month-old, it is possible that MATH1 expression was

Table 2 Clinical and molecular characteristics of MBs on the basis of MATH1 expression MATH1 Yes (n = 14)

No (n = 9)

Age Children (<16 years) (n = 10) Adult (≥16 years) (n = 13)

1 13

9 0

Location Vermis (n = 12) Hemisphere (n = 11)

3 11

9 0

Subtypes MB (n = 10) DMB (n =13)

3 11

7 2

PEDF Up (n = 8) Down (n = 15)

8 6

0 9

TRKC Up (n = 9) Down (n = 14)

8 6

1 8

MATH1 vs. age: Fisher’s Exact Test, p value for two-tailed test is <0.00002; MATH1 vs. location: Fisher’s Exact Test, p value for two-tailed test is <0.0004; MATH1 vs. subtypes: Fisher’s Exact Test, p value for two-tailed test is <0.02; MATH1 vs. PEDF: Fisher’s Exact Test, p value for two-tailed test is <0.008; MATH1 vs. TRKC: Fisher’s Exact Test, p value for two-tailed test is <0.04.

183

due to the presence of normal EGL cells intermingled to tumor cells. A recent study also suggested that MATH1 is up-regulated in a subgroup of MBs [37]. The location of the tumors was not described, however, and all the patients were of pediatric age, making correlations with our data difficult. In our study we also found that 30% of MBs (7/23) express high levels of PEDF. This result was surprising since PEDF is located on 17p13.3, a chromosomal region of frequent LOH in MBs [44]; moreover, low PEDF levels should favor tumor progression because of the anti-angiogenic properties of PEDF [14,28,48]. PEDF, however, is also a survival factor for immature CGCs [4,45,47]. Interestingly, we have found that all MBs expressing high levels of PEDF also express MATH1. This may suggest that PEDF plays a role in the pathogenesis of MBs originating from the EGL. During cerebellar development high levels of PEDF could inhibit vascularization of the EGL and prevent neuronal loss induced by hypoxia, like in the retina [38]. Subsequently, reduced PEDF expression could favor the loss of CGCs in excess and the capillary branching within the subpial surface of the cerebellar cortex. Thus, high PEDF expression could help the persistence of immature CGCs and their transformation or favor the survival of tumor cells, albeit inhibiting vessel growth. MBs with 17p LOH, on the other hand, may have reduced PEDF expression. These MBs might lose the advantage of high PEDF expression in preventing neuronal apoptosis, but could augment the vascularization and reduce the extent of hypoxia [31]. TRKC, the high affinity receptor for neurotrophin-3 (NT3), is widely distributed in the developing cerebellum [39]. TRKC is considered as a marker of neuronal differentiation and there is evidence that high levels of TRKC favor the apoptosis of neoplastic cells in MBs [35]. Indeed, high TRKC expression is an independent predictor of favorable clinical outcome [27]. We have found that high levels of TRKC mRNA are more common in MATH1 mRNA-positive MBs than in others (p < 0.04), suggesting that MATH1 mRNA-positive MBs are associated with a higher degree of differentiation. BIRC5, the gene encoding survivin, is a unique member of the inhibitor of apoptosis (IAP) family and maps on 17q25, a chromosomal region frequently amplified in MBs [8]. Survivin expression is undetectable in most adult tissues, including cerebellum, but BIRC5 is strongly expressed in embryonic and fetal organs and in virtually every tumor studied [42]. We have found that BIRC5 is overexpressed in all MBs and undetectable in adult cerebellum or brain. This may suggest that survivin, because of its requirement for maintaining cancercell viability, is a rational target for MB therapy. Furthermore, because of its differential expression in tumor versus normal tissues, a survivin-based therapy would be expected to carry limited toxicity. Indeed, pharmacological approaches, as well immune- and gene-therapy based strategies have been proposed for survivin-expressing tumors [3]. The pattern of MATH1 expression in MBs could also make MATH1 an attractive target for therapeutic strategies.

184

E. Salsano et al. / Neuroscience Letters 370 (2004) 180–185

Since the levels of BIRC5 mRNA in MBs are homogeneous, survivin expression, differently from human gliomas [12], appears as a poor prognostic indicators in MBs. Conversely, survival analysis on patients with tumor recurrence (n = 11; the number of patients that could be considered for overall survival was too low) suggests that the expression of MATH1 is associated with lower risk of tumor recurrence. Patients with MATH1 mRNA-positive MBs (n = 8) had a longer progression-free survival (PFS) than those with MATH1 mRNA negative MBs (n = 3): 63.0 ± 22.6 S.E. months vs. 11.3 ± 6.6 S.E. months. This better prognosis can be related to tumor location or patients’ age, since these parameters are strongly associated to MATH1 expression. TRKC up-regulation, that is significantly associated with MATH1 expression (p < 0.04), is also associated with increased PFS: 81.0 ± 32.4 S.E. months versus 22.2 ± 11.8 S.E. months in patients with TRKC mRNA up (n = 5) versus patients with TRKC mRNA down (n = 6). Larger, more powered studies will be necessary to evaluate the effect of MATH1 stratification as a prognostic indicator. Interestingly, in the cerebellum of transgenic mice overexpressing Math1, Helms et al. observed increased levels of Ptch, an antagonist of sonic hedgehog signaling. This expression profile could result in decreased proliferation of MATH1 mRNA-positive MBs and in a better prognosis [29]. In summary, we identified two subgroups of biologically different MBs that can be recognized by levels of MATH1 expression using real-time PCR. MBs expressing MATH1 originate from EGL, correlate more frequently with high TRKC and PEDF mRNA levels, and occur more commonly in cerebellar hemispheres. MBs that do not express MATH1 originate from the VM and are located in the midline. We suggest that this difference in MATH1 expression may help in clarifying the biology of MBs and should be taken into account when new biomarkers of prognostic value are assessed.

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15] [16]

[17]

Acknowledgements [18]

We thank Lucia Limido and Emanuela Maderna for help in analysis of PEDF expression. This work was partially supported by grants from the Italian Minister of Health to GF. References [1] H. Abraham, T. Tornoczky, G. Kosztolanyi, L. Seress, Cell formation in the cortical layers of the developing human cerebellum, Int. J. Dev. Neurosci. 19 (2001) 53–62. [2] D.C. Altieri, Survivin, versatile modulation of cell division and apoptosis in cancer, Oncogene 22 (2003) 8581–8589. [3] D.C. Altieri, Validating survivin as a cancer therapeutic target, Nat. Rev. Cancer 3 (2003) 46–54. [4] T. Araki, T. Taniwaki, S.P. Becerra, G.J. Chader, J.P. Schwartz, Pigment epithelium-derived factor (PEDF) differentially protects immature but not mature cerebellar granule cells against apoptotic cell death, J. Neurosci. Res. 53 (1998) 7–15. [5] J. Aruga, T. Tohmonda, S. Homma, K. Mikoshiba, Zic1 promotes the expansion of dorsal neural progenitors in spinal cord

[19]

[20]

[21]

[22]

[23]

by inhibiting neuronal differentiation, Dev. Biol. 244 (2002) 329– 341. N. Ben-Arie, H.J. Bellen, D.L. Armstrong, A.E. McCall, P.R. Gordadze, Q. Guo, M.M. Matzuk, H.Y. Zoghbi, Math1 is essential for genesis of cerebellar granule neurons, Nature 390 (1997) 169–172. J.A. Biegel, C.D. Burk, F.G. Barr, B.S. Emanuel, Evidence for a 17p tumor related locus distinct from p53 in pediatric primitive neuroectodermal tumors, Cancer Res. 52 (1992) 3391–3395. S.H. Bigner, R.E. McLendon, H. Fuchs, P.E. McKeever, H.S. Friedman, Chromosomal characteristics of childhood brain tumors, Cancer Genet. Cytogenet. 97 (1997) 125–134. A.A. Brandes, M.K. Paris, U. Basso, Medulloblastomas: do molecular and biologic markers indicate different prognoses and treatments? Expert Rev. Anticancer Ther. 3 (2003) 615–620. J. Buhren, A.H. Christoph, R. Buslei, S. Albrecht, O.D. Wiestler, T. Pietsch, Expression of the neurotrophin receptor p75NTR in medulloblastomas is correlated with distinct histological and clinical features: evidence for a medulloblastoma subtype derived from the external granule cell layer, J. Neuropathol. Exp. Neurol. 59 (2000) 229–240. A.R. Carter, E.M. Berry, R.A. Segal, Regional expression of p75NTR contributes to neurotrophin regulation of cerebellar patterning, Mol. Cell Neurosci. 22 (2003) 1–13. A. Chakravarti, E. Noll, P.M. Black, D.F. Finkelstein, D.M. Finkelstein, N.J. Dyson, J.S. Loeffler, Quantitatively determined survivin expression levels are of prognostic value in human gliomas, J. Clin. Oncol. 20 (2002) 1063–1068. A. Chopra, K.M. Brown, B.R. Rood, R.J. Packer, T.J. MacDonald, The use of gene expression analysis to gain insights into signaling mechanisms of metastatic medulloblastoma, Pediatr. Neurosurg. 39 (2003) 68–74. D.W. Dawson, O.V. Volpert, P. Gillis, S.E. Crawford, H. Xu, W. Benedict, N.P. Bouck, Pigment epithelium-derived factor: a potent inhibitor of angiogenesis, Science 285 (1999) 245–248. C.G. Eberhart, P.C. Burger, Anaplasia and grading in medulloblastomas, Brain Pathol. 13 (2003) 376–385. R.A. Edelstein, R.J. Krane, R.K. Babayan, A. de las Morenas, R.B. Moreland, A rapid and simple method for the detection of prostatespecific antigen mRNA in archival tissue specimens using a reverse transcription-polymerase chain reaction assay, Urology 45 (1995) 597–603. A. Fernandez-Teijeiro, R.A. Betensky, L.M. Sturla, J.Y. Kim, P. Tamayo, S.L. Pomeroy, Combining gene expression profiles and clinical parameters for risk stratification in medulloblastomas, J. Clin. Oncol. (2004). P.G. Fisher, P.C. Burger, C.G. Eberhart, Biologic risk stratification of medulloblastoma: The real time is now, J. Clin. Oncol. (2004). A. Gajjar, R. Hernan, M. Kocak, C. Fuller, Y. Lee, P.J. McKinnon, D. Wallace, C. Lau, M. Chintagumpala, D.M. Ashley, S.J. Kellie, L. Kun, R.J. Gilbertson, Clinical, histopathologic, and molecular markers of prognosis: toward a new disease risk stratification system for medulloblastoma, J. Clin. Oncol. (2004). R. Gazit, V. Krizhanovsky, N. Ben-Arie, Math1 controls cerebellar granule cell differentiation by regulating multiple components of the Notch signaling pathway, Development 131 (2004) 903–913. F. Giangaspero, G. Perilongo, M.P. Fondelli, M. Brisigotti, C. Carollo, R. Burnelli, P.C. Burger, M.L. Garre, Medulloblastoma with extensive nodularity: a variant with favorable prognosis, J. Neurosurg. 91 (1999) 971–977. F. Giangaspero, L. Rigobello, M. Badiali, M. Loda, L. Andreini, G. Basso, F. Zorzi, A. Montaldi, Large-cell medulloblastomas. A distinct variant with highly aggressive behavior, Am. J. Surg. Pathol. 16 (1992) 687–693. R. Gilbertson, C. Wickramasinghe, R. Hernan, V. Balaji, D. Hunt, D. Jones-Wallace, J. Crolla, R. Perry, J. Lunec, A. Pearson, D. Ellison, Clinical and molecular stratification of disease risk in medulloblastoma, Br. J. Cancer. 85 (2001) 705–712.

E. Salsano et al. / Neuroscience Letters 370 (2004) 180–185 [24] M.T. Giordana, P. Cavalla, A. Dutto, L. Borsotti, A. Chio, D. Schiffer, Is medulloblastoma the same tumor in children and adults? J. Neurooncol. 35 (1997) 169–176. [25] D. Goldowitz, K. Hamre, The cells and molecules that make a cerebellum, Trends Neurosci. 21 (1998) 375–382. [26] M.A. Grotzer, M.D. Hogarty, A.J. Janss, X. Liu, H. Zhao, A. Eggert, L.N. Sutton, L.B. Rorke, G.M. Brodeur, P.C. Phillips, MYC messenger RNA expression predicts survival outcome in childhood primitive neuroectodermal tumor/medulloblastoma, Clin. Cancer Res. (2001) 7. [27] M.A. Grotzer, A.J. Janss, K. Fung, J.A. Biegel, L.N. Sutton, L.B. Rorke, H. Zhao, A. Cnaan, P.C. Phillips, V.M. Lee, J.Q. Trojanowski, TrkC expression predicts good clinical outcome in primitive neuroectodermal brain tumors, J. Clin. Oncol. 18 (2000) 1027–1035. [28] M. Guan, H.F. Yam, B. Su, K.P. Chan, C.P. Pang, W.W. Liu, W.Z. Zhang, Y. Lu, Loss of pigment epithelium derived factor expression in glioma progression, J. Clin. Pathol. 56 (2003) 277–282. [29] A.W. Helms, K. Gowan, A. Abney, T. Savage, J.E. Johnson, Overexpression of MATH1 disrupts the coordination of neural differentiation in cerebellum development, Mol. Cell Neurosci. 17 (2001) 671–682. [30] J. Herms, I. Neidt, B. Luscher, A. Sommer, P. Schurmann, T. Schroder, M. Bergmann, B. Wilken, S. Probst-Cousin, P. HernaizDriever, J. Behnke, F. Hanefeld, T. Pietsch, H.A. Kretzschmar, CMYC expression in medulloblastoma and its prognostic value, Int. J. Cancer. 89 (2000) 395–402. [31] H. Huber, A. Eggert, A.J. Janss, R. Wiewrodt, H. Zhao, L.N. Sutton, L.B. Rorke, P.C. Phillips, M.A. Grotzer, Angiogenic profile of childhood primitive neuroectodermal brain tumours/medulloblastomas, Eur. J. Cancer. 37 (2001) 2064–2072. [32] C.D. Katsetos, L. Del Valle, A. Legido, J.P. de Chadarevian, E. Perentes, S.J. Mork, On the neuronal/neuroblastic nature of medulloblastomas: a tribute to Pio del Rio Hortega and Moises Polak, Acta Neuropathol. (Berl.) 105 (2003) 1–13. [33] C.D. Katsetos, A. Frankfurter, S. Christakos, E.L. Mancall, I.N. Vlachos, H. Urich, Differential localization of class III, beta-tubulin isotype and calbindin-D28k defines distinct neuronal types in the developing human cerebellar cortex, J. Neuropathol. Exp. Neurol. 52 (1993) 655–666. [34] C.D. Katsetos, M.M. Herman, L. Krishna, J.R. Vender, S.A. Vinores, D.P. Agamanolis, D. Schiffer, P.C. Burger, H. Urich, Calbindin-D28k in subsets of medulloblastomas and in the human medulloblastoma cell line D283 Med, Arch. Pathol. Lab. Med. 119 (1995) 734–743. [35] J.Y. Kim, M.E. Sutton, D.J. Lu, T.A. Cho, L.C. Goumnerova, L. Goritchenko, J.R. Kaufman, K.K. Lam, A.L. Billet, N.J. Tarbell, J. Wu, J.C. Allen, C.D. Stiles, R.A. Segal, S.L. Pomeroy, Activation of neurotrophin-3 receptor TrkC induces apoptosis in medulloblastomas, Cancer Res. 59 (1999) 711–719. [36] P. Kleihues, D.N. Louis, B.W. Scheithauer, L.B. Rorke, G. Reifenberger, P.C. Burger, W.K. Cavenee, The WHO classification of tumors of the nervous system, J. Neuropathol. Exp. Neurol. 61 (2002) 215–225 (discussion 226–229). [37] Y. Lee, H.L. Miller, P. Jensen, R. Hernan, M. Connelly, C. Wetmore, F. Zindy, M.F. Roussel, T. Curran, R.J. Gilbertson, P.J. McKinnon, A molecular fingerprint for medulloblastoma, Cancer Res. 63 (2003) 5428–5437. [38] N. Ogata, L. Wang, N. Jo, J. Tombran-Tink, K. Takahashi, D. Mrazek, M. Matsumura, Pigment epithelium derived factor as a neuroprotective agent against ischemic retinal injury, Curr. Eye Res. 22 (2001) 245–252.

185

[39] S.L. Pomeroy, M.E. Sutton, L.C. Goumnerova, R.A. Segal, Neurotrophins in cerebellar granule cell development and medulloblastoma, J. Neurooncol. 35 (1997) 347–352. [40] S.L. Pomeroy, P. Tamayo, M. Gaasenbeek, L.M. Sturla, M. Angelo, M.E. McLaughlin, J.Y. Kim, L.C. Goumnerova, P.M. Black, C. Lau, J.C. Allen, D. Zagzag, J.M. Olson, T. Curran, C. Wetmore, J.A. Biegel, T. Poggio, S. Mukherjee, R. Rifkin, A. Califano, G. Stolovitzky, D.N. Louis, J.P. Mesirov, E.S. Lander, T.R. Golub, Prediction of central nervous system embryonal tumour outcome based on gene expression, Nature 415 (2002) 436–442. [41] C. Sarkar, P. Pramanik, A.K. Karak, P. Mukhopadhyay, M.C. Sharma, V.P. Singh, V.S. Mehta, Are childhood and adult medulloblastomas different? A comparative study of clinicopathological features, proliferation index and apoptotic index, J. Neurooncol. 59 (2002) 49–61. [42] T. Sasaki, M.B. Lopes, G.R. Hankins, G.A. Helm, Expression of survivin, an inhibitor of apoptosis protein, in tumors of the nervous system, Acta Neuropathol. (Berl.) 104 (2002) 105–109. [43] R.A. Segal, L.C. Goumnerova, Y.K. Kwon, C.D. Stiles, S.L. Pomeroy, Expression of the neurotrophin receptor TrkC is linked to a favorable outcome in medulloblastoma, Proc. Natl. Acad. Sci. U.S.A. 91 (1994) 12867–12871. [44] I. Slavc, I.R. Rodriguez, K. Mazuruk, G.J. Chader, J.A. Biegel, Mutation analysis and loss of heterozygosity of PEDF in central nervous system primitive neuroectodermal tumors, Int. J. Cancer. 72 (1997) 277–282. [45] T. Taniwaki, S.P. Becerra, G.J. Chader, J.P. Schwartz, Pigment epithelium-derived factor is a survival factor for cerebellar granule cells in culture, J. Neurochem. 64 (1995) 2509–2517. [46] T. Taniwaki, N. Hirashima, S.P. Becerra, G.J. Chader, R. Etcheberrigaray, J.P. Schwartz, Pigment epithelium-derived factor protects cultured cerebellar granule cells against glutamate-induced neurotoxicity, J. Neurochem. 68 (1997) 26–32. [47] J. Tombran-Tink, C.J. Barnstable, PEDF: a multifaceted neurotrophic factor, Nat. Rev. Neurosci. 4 (2003) 628–636. [48] J. Tombran-Tink, C.J. Barnstable, Therapeutic prospects for PEDF: more than a promising angiogenesis inhibitor, Trends Mol. Med. 9 (2003) 244–250. [49] J. Tombran-Tink, H. Pawar, A. Swaroop, I. Rodriguez, G.J. Chader, Localization of the gene for pigment epithelium-derived factor (PEDF) to chromosome 17p13.1 and expression in cultured human retinoblastoma cells, Genomics 19 (1994) 266–272. [50] W.M. Tong, H. Ohgaki, H. Huang, C. Granier, P. Kleihues, Z.Q. Wang, Null mutation of DNA strand break-binding molecule poly(ADP-ribose) polymerase causes medulloblastomas in p53(−/−) mice, Am. J. Pathol. 162 (2003) 343–352. [51] D.T. Yew, C.B. Luo, C.W. Heizmann, W.Y. Chan, Differential expression of calretinin, calbindin D28K and parvalbumin in the developing human cerebellum, Brain Res. Dev. Brain Res. 103 (1997) 37–45. [52] N. Yokota, J. Aruga, S. Takai, K. Yamada, M. Hamazaki, T. Iwase, H. Sugimura, K. Mikoshiba, Predominant expression of human zic in cerebellar granule cell lineage and medulloblastoma, Cancer Res. 56 (1996) 377–383. [53] P.M. Zeltzer, J.M. Boyett, J.L. Finlay, A.L. Albright, L.B. Rorke, J.M. Milstein, J.C. Allen, K.R. Stevens, P. Stanley, H. Li, J.H. Wisoff, J.R. Geyer, P. McGuire-Cullen, J.A. Stehbens, S.B. Shurin, R.J. Packer, Metastasis stage, adjuvant treatment, and residual tumor are prognostic factors for medulloblastoma in children: conclusions from the Children’s Cancer Group 921 randomized phase III study, J. Clin. Oncol. 17 (1999) 832–845.