Evasion of cell senescence in SHH medulloblastoma

ABSTRACT

The mechanisms leading to brain tumor formation are poorly understood. Using Ptch1^+/− mice as a medulloblastoma model, sequential mutations were found to shape tumor evolution. Initially, medulloblastoma preneoplastic lesions display loss of heterozygosity of the Ptch1 wild-type allele, an event associated with cell senescence in preneoplasia. Subsequently, p53 mutations lead to senescence evasion and progression from preneoplasia to medulloblastoma. These findings are consistent with a model where high levels of Hedgehog signaling caused by the loss of the tumor suppressor Ptch1 lead to oncogene-induced senescence and drive p53 mutations. Thus, cell senescence is an important characteristic of a subset of SHH medulloblastoma and might explain the acquisition of somatic TP53 mutations in human medulloblastoma. This mode of medulloblastoma formation contrasts with the one characterizing Li-Fraumeni patients with medulloblastoma, where TP53 germ-line mutations cause chromothriptic genomic instability and lead to mutations in Hedgehog signaling genes, which drive medulloblastoma growth. Here we discuss in detail these 2 alternative mechanisms leading to medulloblastoma tumorigenesis.

KEYWORDS:

• cell senescence
• Medulloblastoma; p53
• preneoplasia
• Ptch1
• sonic hedgehog
• TP53

Introduction

Medulloblastoma is the most common brain tumor in children. According to integrated genomics studies, medulloblastomas can be classified in 4 different molecular subtypes: WNT, SHH, Group 3 and Group 4.¹ Each of these molecular subtypes has specific underlying molecular features, gene expression, demographic characteristics and prognosis.² Sonic Hedgehog (SHH) medulloblastomas are characterized by activation of the Hedgehog (Hh) signaling pathway and are often driven by mutations in Hh pathway components.³ Here we focus on SHH medulloblastoma and discuss how cell senescence shapes the natural history and molecular evolution of this childhood cancer.

TP53 mutations and medulloblastoma

For many years, it was thought that TP53 mutations were infrequent in medulloblastomas^4,5 and that P53 signaling was dispensable for medulloblastoma tumor suppression. However, recent studies reported that TP53 mutations are frequent in human WNT (with a rate of 16%) and SHH primary medulloblastomas (with a rate of 13% to 21%, depending on the study)^3,6,7 and are indicators of poor prognosis exclusively in SHH medulloblastoma.⁷ Moreover, TP53 mutations have recently been identified as a key event in the pathogenesis of medulloblastoma recurrence.^8,9 Although the specific roles of P53 in medulloblastoma are still largely unknown,¹⁰ the presence of recurrent TP53 mutations suggests that P53 signaling plays an important role in driving medulloblastoma tumorigenesis.

Human genomic studies also established that TP53 mutations co-exist with mutations or amplifications of Hh signaling components such as SHH, SMO, SUFU, GLI2 and MYCN.³ These studies, however, do not establish the sequential order of genetic events that lead to medulloblastoma formation and how these mutations correlate with presumptive histopathological stages of medulloblastoma. Nevertheless, patients with Li-Fraumeni syndrome, caused by germ-line mutations in TP53, offer an opportunity to study how medulloblastomas arise at the genetic level. Li-Fraumeni patients are cancer-prone and sometimes develop medulloblastoma;¹¹ therefore, at least for this subset of patients, TP53 mutations are the first genetic event leading to medulloblastoma formation (Fig. 1A). Notably, most Li-Fraumeni medulloblastomas belong to the SHH subgroup, indicating that TP53 mutations specifically predispose to SHH medulloblastoma.¹² Because cerebellum granule cell precursors (GCPs) are the cells of origin of SHH medulloblastoma,^13-15 this indicates that GCPs are highly susceptible to transformation in absence of P53. Moreover, these SHH medulloblastomas seem to be the result of chromothripsis, a massive genome rearrangement caused by chromosomal shattering likely occurring in a single event.¹⁶ These chromothriptic events likely lead to mutations in components of Hh signaling, such as GLI2, BOC and MYCN amplifications.¹² Since Hh signaling is the most important mitogenic pathway for GCPs,¹⁷ it is therefore expected that acquisition of Hh pathway mutations efficiently causes SHH medulloblastoma in Li-Fraumeni patients.

Figure 1. Two alternative mechanisms of medulloblastoma formation. (A) In Li-Fraumeni patients with germ-line TP53 mutations, cerebellum GCPs experience chromosomal instability. Chromothriptic events lead to massive chromosomal rearrangements and high levels of amplification in Hh signaling genes such as GLI2 and MYCN. Acquisition of these Hh signaling mutations leads to medulloblastoma growth. (B) In Ptch1^+/− mice, loss of heterozygosity of the Ptch1 wild-type allele leads to the formation of preneoplasia. Preneoplastic lesions display high levels of cell senescence. Spontaneous p53 mutations or p16ink4a inactivation leads to senescence evasion and progression to advanced medulloblastoma. Proliferation and senescence levels during medulloblastoma formation are indicated.

Although p53 knockout mice or mouse models of Li-Fraumeni syndrome do not develop medulloblastoma,^18,19 elegant studies have demonstrated that the inactivation of p53 together with other DNA repair factors such as as Xrcc4, Ligase IV, Xrcc2 and Brca2 leads to Shh medulloblastoma.^20,21 Interestingly, those medulloblastomas also harbored spontaneous mutations in Hh signaling components such as Ptch1, Mycn and Gli2.²⁰ This result not only highlights how important it is to maintain genomic stability to prevent SHH medulloblastoma formation, but it also indicates that Hh signaling activation seems to be necessary for medulloblastoma formation even when p53 and DNA repair mechanisms are absent. In summary, germ-line P53 mutations in both mouse and human cause genomic instability and lead to mutations in Hh signaling components, conducing to medulloblastoma formation (Fig. 1A). This is an interesting paradigm showing that Hh signaling mutations happen subsequent to TP53 mutations in Li-Fraumeni syndrome.

Half of the SHH medulloblastomas with TP53 mutations have a germ-line (Li-Fraumeni) origin and are potentially explained by the mechanism described above.^3,12 However, it is not known how the other SHH medulloblastoma cases (including the ones with somatic TP53 mutations and the ones without TP53 mutations) develop and the temporal order in which they acquire their mutations. Another unresolved, yet related, question is whether advanced medulloblastomas arise in a step-wise manner from subclinical precancerous lesions. For many epithelial cancers, the availability of preneoplastic lesions allowed the establishment of tumor progression models with sequential histopathological stages and the molecular changes that characterize them.²² However, the problem for understanding brain tumor development lies in the inability to detect and obtain precancerous lesions; therefore, genome sequencing of advanced brain tumors only offers a snapshot of the mutations present in advanced tumors but does not show the order in which mutations are acquired during tumor progression.

To address this question, we used Ptch1^+/− heterozygous mice, a well-established model of Shh medulloblastoma (Fig. 2). Ptch1^+/− mice develop preneoplastic lesions with high frequency, but only a fraction of those animals develop advanced medulloblastoma.^13,23 Therefore, we hypothesized that an unidentified tumor suppressive mechanism might restrain the progression of medulloblastoma preneoplasia into advanced tumors. We found that apoptosis levels are the same between preneoplastic lesions and advanced medulloblastoma, eliminating apoptosis as an essential tumor suppressor in this model of Shh medulloblastoma. Surprisingly, when we looked at cell senescence, we found that while preneoplastic lesions display high numbers of p21^Cip1 and p16^Ink4a positive cells (which are effectors and markers of cell senescence and cell cycle arrest^24-26), advanced medulloblastomas have very low levels of senescence.²⁷ These high levels of senescence are paralleled by lower levels of proliferation and correlate with loss of heterozygosity (LOH) of the Ptch1 wild-type allele in preneoplastic lesions (Fig. 1B), suggesting that high levels of Hh signaling may contribute to cell senescence. Using laser capture microdissection, we found that one-third of all advanced medulloblastomas acquired p53 mutations that were not present in preneoplastic lesions, supporting the notion that p53 mutations allow senescence evasion and medulloblastoma progression (Fig. 1B). Moreover, we found that engineered p53 mutations prevent cellular senescence and accelerate medulloblastoma formation, showing that p53 mutations lead to senescence evasion.²⁷ Further supporting the idea that senescence evasion is necessary for medulloblastoma progression, advanced tumors without p53 mutations display low expression of p16^Ink4a due to promoter methylation. In summary, we found that, contrary to Li-Fraumeni syndrome (where TP53 germ-line mutations lead to Hh signaling mutations), Hh signaling hyperactivity leads to cell senescence in preneoplastic lesions and creates selection pressure for the inactivation of p53 or p16^Ink4a, which allows the progression from preneoplasia to advanced medulloblastoma (Fig. 1B).

Figure 2. Medulloblastoma formation in Ptch1^+/− mice. During postnatal development, granule cell precursors (GCPs) of the cerebellum, the cells of origin of Shh medulloblastoma, are located in the external granule-cell layer (EGL). After their proliferation, most GCPs differentiate, populate the internal granule-cell layer (IGL), and disappear from the EGL after the second postnatal week in the mouse. LOH of the Ptch1 wild-type allele causes a clonal expansion and leads to the formation of preneoplasia. While most preneoplastic lesions disappear, some of them progress to advanced medulloblastoma. See also Figure 1B.

Importantly, the finding of spontaneous p53 mutations is not limited to the Ptch1^+/− model, as we also found spontaneous p53 mutations in tumors from Olig1-Gnas mice, another model of Shh medulloblastoma.²⁸ This suggests that evasion of senescence could be a hallmark of Shh medulloblastoma.

Hh and p53 signaling in medulloblastoma

Since p53 is inactivated during the formation of Shh medulloblastoma (Fig. 1B), this may highlight some possible interactions between p53 and Hh signaling in the brain. Some studies have explored this relationship using a variety of approaches. Shh, a protein secreted by Purkinje cells of the cerebellum, is the most potent mitogen for GCPs.^17,29,30 Shh also promotes proliferation of neural progenitors of the adult hippocampus.³¹ In contrast to these proliferative effects of Shh, p53 activity supresses proliferation and self-renewal of adult neural stem cells of the subventricular zone,³² and this effect might be mediated by p21^Cip1.^32,33 This may explain why p53 activity is downregulated during neurogenesis³⁴ to allow cell proliferation and brain formation. Interestingly, Gli activity has been shown to downregulate p53 protein levels in cell lines.³⁵ Specifically, it has been proposed that high levels of Hh signaling in cell lines caused by expression of constitutively active Smo mutants or overexpression of Gli1 and Gli2 leads to an Mdm2-dependent degradation of p53.³⁶ Consistently, low levels of Mdm2 in vivo increase p53 levels, lead to cerebellar hypoplasia and reduce medulloblastoma development in Ptch1^+/− mice,³⁷ showing that p53 signaling negatively controls cerebellum growth and implying that p53 signaling is important for medulloblastoma tumor suppression. Similarly, the proto-oncogene PPM1D, a negative regulator of p53, is overexpressed in medulloblastomas³⁸ and increases medulloblastoma formation in mice when overexpressed together with Shh.³⁹ Together, these findings provide strong evidence that Hh signaling leads to a functional inactivation of p53 signaling that allows GCP proliferation and, in some instances, medulloblastoma formation. However, the presence of somatic p53 mutations in Ptch1^+/− medulloblastoma²⁷ demonstrates that the ability of Shh signaling to functionally suppress p53 signaling is not always sufficient to inactivate p53 activity in a tumorigenic context.

For many years, it has been known that p53 deletion accelerates medulloblastoma formation and increases medulloblastoma incidence in Ptch1^+/− mice;⁴⁰ however, the mechanism responsible for this was never investigated. The fact that p53 mutations, which are acquired spontaneously during medulloblastoma formation, lead to senescence evasion provides an explanation for the presence of somatic TP53 mutations in human SHH medulloblastoma.²⁷

Cell senescence as a driver of p53 inactivation in medulloblastoma

Gli1 or Smo overexpression in cell lines causes high levels of oncogenic stress.^35,36 Recently, we also found that the ligand Shh induces DNA damage in GCPs, an effect that requires the presence of the Hh receptor Boc and CyclinD1.²³ These results support the hypothesis that GCPs are sensitive to replicative stress⁴¹ and that hedgehog signaling likely causes replication stress. It has been demonstrated that oncogene activation leads to oncogene-induced cell senescence (OIS), a tumor-suppressive mechanism that prevents transformation of premalignant lesions into tumors.⁴² Moreover, oncogene-induced DNA damage seems to be required for OIS.^43,44 In light of this, the fact that we found cell senescence and Ptch1 LOH in medulloblastoma preneoplastic lesions²⁷ suggests that high levels of hedgehog signaling in preneoplasia likely cause high levels of oncogenic stress and this leads to cell senescence. Therefore, we propose that high levels of hedgehog signaling shape the molecular evolution of medulloblastoma by leading to OIS and creating selection pressure to inactivate p53 in order to evade OIS.

In addition to oncogenes, loss of tumor suppressor genes has been shown to cause OIS.⁴⁵ For example, in the context of prostate cancer, Pten inactivation causes senescence and leads to the acquisition of p53 mutations.⁴⁶ Neurofibromin (NF1) loss also leads to Ras-mediated induction of cell senescence.⁴⁷ Since we observed a strong association between Ptch1 loss and cell senescence in medulloblastoma preneoplasia,²⁷ Ptch1 seems to be a new tumor suppressor whose absence may be capable of causing OIS.

TP53 mutations are also present in WNT medulloblastoma.⁷ However, the molecular mechanism leading to TP53 mutations has never been investigated in WNT medulloblastoma. Interesting work has shown that, at least in cell lines, overexpression of the downstream Wnt effector ß-catenin leads to p53 stabilization and activation.⁴⁸ Wnt activation in colorectal cancer has been associated with decreased levels of proliferation⁴⁹ and accumulation of p53 protein in early tumor stages,⁵⁰ suggesting that Wnt signaling may lead to cell senescence in precancerous lesions and this could be the cause of the acquisition of TP53 mutations at late stages of colorectal cancer. Additionally, it was shown that ß-catenin overexpression causes OIS in thymocytes.⁵¹ We thus speculate that Wnt activation in the brainstem may also lead to changes compatible with cell senescence and this may be an explanation for the presence of TP53 mutations in advanced WNT medulloblastomas. This is supported by the fact that expression of an active ß-catenin mutant in the lower rhombic lip of mice leads to the formation of hyperplastic lesions that only progress to advanced medulloblastomas when p53 is also deleted.⁵²

High-grade astrocytomas frequently harbor TP53 mutations.⁵³ It has been shown that low-grade astrocytoma lesions display a DNA damage response^54,55 and that loss of components of the Atm-Chk2-p53 pathway accelerates astrocytoma development.⁵⁵ Although cell senescence was not interrogated in those reports, another study showed that pilocytic astrocytomas (PA), the most benign type of astrocytomas, display MAPK activation and OIS;⁵⁶ interestingly, PA are indolent tumors that display long periods of growth arrest, rarely become high-grade astrocytomas, and do not display TP53 mutations. This is in agreement with the idea that PA are long-term senescent astrocytic lesions that do not progress.⁵⁶

Abbreviations

EGL	=	External Granule-cell Layer
GCPs	=	Granule Cell Precursors
Hh	=	Hedgehog
IGL	=	Internal Granule-cell Layer
OIS	=	Oncogene-induced senescence
Shh	=	Sonic hedgehog

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Funding

Work done in the Charron lab was supported by grants from the Canadian Institutes of Health Research (CIHR), the Fonds de Recherche du Québec-Santé (FRQS), and the Canada Foundation for Innovation (CFI). LT-O is recipient of the Caldas fellowship (Colciencias). SMS is recipient of the McGill James O. and Maria Meadows fellowship and IRCM Challenge-Guépards et Gazelles gourmands scholarship. FC holds the Canada Research Chair in Developmental Neurobiology.