Calculating a cure for cancer: managing medulloblastoma MATH1-ematically

Keywords::

• BMI1
• brain tumors
• cerebellum
• external graule layer
• genomics
• granule cell precursors
• MATH1
• medulloblastoma
• neuronal differentiation
• Sonic Hedgehog

Medulloblastoma (MB), an aggressive embryonal tumor of the cerebellum, is the most common malignant pediatric brain tumor and a major cause of pediatric morbidity and mortality. Recent advances in genomic technologies have permitted in-depth analysis of gene expression, copy number and epigenetic abnormalities in these tumors, and have revealed genetic heterogeneity far greater than originally suspected [1–4]. The Sonic Hedgehog (Shh) subgroup of MBs closely resembles, and is thought to arise from, normal granule cell precursors (GCPs) of the external granule cell layer, the predominant cell type of the developing cerebellum [5]. Similarities between Shh subgroup MBs and GCPs include a high level of expression of the transcription factor MATH1 [5,6].

Cerebellar development & the role of Math1

Cerebellar development is a highly regulated and well-studied process with important contributions made by the Shh signaling pathway [7–10]. Undifferentiated GCPs migrate tangentially from the rhombic lip, forming an outer, proliferative layer of the cerebellum known as the external granule layer (EGL). Mitotic activation of these cells occurs through paracrine stimulation by Shh protein released by the underlying Purkinje cells, resulting in extensive proliferation of EGL cells [11]. Shh acts on the transmembrane receptor patched (Ptc1), thereby releasing inhibition of smoothened (Smo) and allowing downstream activation of Shh signaling, including the transcription of Shh effector genes including Gli1, Gli2and Mycn[7,11]. Following Shh-induced proliferation, cells of the EGL eventually become tachyphylactic to Shh, exit the cell cycle and migrate to the interior of the cerebellum, forming a terminally differentiated internal granule layer [7].

Math1encodes a basic helix–loop–helix (bHLH) transcription factor necessary for cerebellar development [12]. Expression of Math1 is an early marker of GCPs, guiding these largely undifferentiated cells towards a neuronal identity [13] and facilitating the formation of the proliferative EGL. However, Math1 expression is temporally restricted, and is extinguished in post-mitotic GCPs of the internal granule layer [11]. Transcriptional control is governed by numerous developmental pathways forming a complex regulatory mechanism that involves bone morphogenetic proteins (BMPs; BMP2, BMP4, BMP7) [14,15], Notch signaling (Hes5) [16], and Shh effectors (Shh) [15,17]. These pathways converge to facilitate the interplay between proliferation and differentiation guiding the formation and architecture of the developing cerebellum.

Gain and loss of function experiments in vivo have shown Math1to be critical for survival and normal cerebellar development [12,18]. Math1-deficient mice die shortly after birth [12], displaying no discernable EGL and abnormal GCP differentiation [16]. While loss of Math1 does not preclude the formation of the rhombic lip or the GCP population, the proliferative nature of the cells is grossly impaired [12]. Similarly, overexpression leads to postnatal lethality and premature lineage commitment, demonstrating the need for highly regulated and directed expression [18].

Math1 in the pathogenesis of MB

Medulloblastoma is a pediatric cerebellar tumor that can be attributed (at least in part) to perturbations of normal developmental processes. Indeed, extensive gene-expression profiling has confirmed the presence of commonly deregulated developmental pathways such as Wnt signaling and Shh signaling [1]. Integrative genomics has demonstrated MBs are in fact comprised of four distinct subgroups – SHH, WNT, Group C and Group D; each of which is defined by specific genetic perturbations and gene-expression signatures [1]. Upregulation of Math1 occurs specifically in most Shh subgroup MBs, but not other MB subgroups [19]. Elegant experiments by Yang et al. have demonstrated that MB can initiate from the EGL using a Ptch conditional knockout and Math1-GFP reporter [20].

Overexpression of Math1 may be governed by a number of tumorigenic events including promoter hypermethylation and subsequent silencing of the tumor suppressor gene HIC1[21]. HIC1is found on chromosome 17p13.3, a region deleted in 30–50% of MBs [22], and which acts as a direct negative regulator of Math1[21]. Promoter CpG methylation of Hic1 in heterozygote (Hic1^+/-) mice causes complete allelic silencing, uncoupling one modality of Math1 regulation [21]. Although Hic1 loss is not sufficient to initiate MB in mice, it is a progression factor, as when Hic mutant mice are crossed to Ptc1^+/- animals, the incidence of MB increases fourfold [21].

In order to bypass the perinatal lethality seen in Math1^-/- mice, Flora et al. deleted Math1 postnatally using a floxed allele and an inducible cre recombinase line, thereby demonstrating that Math1 participates in Shh signaling, and is necessary for MB formation [17]. In the absence of Math1, a sharp decrease in the proliferative capacity of GCPs with a hypoplastic EGL was observed. Math1acts as an enhancer of Gli2 transcription, directly regulating its expression [17]. Deletion of Math1completely inhibited tumor formation in a MB mouse model driven by expression of a constitutively active mutant of smoothened[17].

Proliferation or differentiation?

Although our current understanding of Math1 has suggested a pivotal role in proliferation of both normal GCPs and in Shh subgroup MB, as early as 2001, Helms et al. demonstrated Math1 overexpression is also linked to defects in GCP differentiation [18]. Specifically, elevated expression of Math1impedes the transition of neural precursors to fully differentiated neurons, allowing cells to retain their proliferative phenotype [18]. These conclusions were recently validated in an elegant series of experiments by Ayrault et al., which demonstrated a specific role of Math1 in regulating differentiation [23]. Viral transduction of Math1and Gli1 into Ptc^+/-; Cdkn2c^-/- GCPs results in a dramatically increased incidence of MB as compared with GCPs transduced with Gli1alone [23]. Furthermore, Math1-induced MBs demonstrate markedly reduced levels of neuronal differentiation markers as demonstrated by both immunohistochemistry and gene-expression profiling [23]. Interestingly, genes involved in cell proliferation remain unchanged or downregulated in spite of Math1 overexpression, suggesting the proliferative nature of Math1 stems from cooperation with the Shh pathway. However, Math1 overexpression alone was unable to induce MB formation, suggesting that hedgehog activation is a requisite element of Math1-induced tumorigenesis, and that Math1 is important in tumor progression rather than tumor initiation [23].

Math1 as a therapeutic candidate target

The aberrant expression of Math1 in MB led researchers to suggest that it might represent an interesting therapeutic target [17]. Zhao et al. used BMP-2 and BMP-4 to antagonize Math1 expression in both normal GCPs and MB [14,15]. BMP signaling resulted in both reduced proliferation and induced differentiation through negative regulation of Math1 by its proteasomal-mediated degradation [14]. However, there are numerous limitations associated with BMP-mediated treatment. The size of these proteins makes them particularly difficult to deliver, and the effect of BMP on Math1-positive cells appears temporary [15]. Specifically, BMP arrests the proliferative state of Math1-expressing cells, rather than forcing apoptosis. Furthermore, overexpression of Math1 decreases the therapeutic effect of BMPs [15].

Alternative hedgehog therapies

Currently a number of inhibitors have been proposed as treatments for Shh-driven neoplasms [24]. Cyclopamine, a plant steroid alkaloid, and HhAntag, a benzimidazole derivative, are agents that block activation of Smo, reducing the downstream activity of effectors [15,25]. These treatments, which differ in their affinity for SMO binding, not only decrease the proliferative index of tumor cells, but, unlike BMP treatment, also cause increased apoptosis of human tumor cells [26]. Further in vivo testing has demonstrated that HhAntag is able to prevent MB formation, or greatly reduce tumor mass in hedgehog-activated (Ptc1^+/-, p53^+/-) MB mouse models [25]. Ultimately, these treatments are limited by the genetic event that hyperactivates Shh signaling. Gene amplification of GLI1, GLI2 and mutation of the downstream Shh pathway inhibitor HSUFU have been observed in human MBs. Drugs targeting Shh signaling at the level of the membrane are unlikely to be effective in these latter tumors and may suggest that targeting Hh-effector proteins may be more efficacious in a larger percentage of MBs [27,28]. Both synthetic and naturally occurring drugs targeting Gli-family proteins have been described [24]. Although only preliminary results have been obtained, these molecules have added to the growing list of Shh modulators that might be used alone or in concert to inhibit Shh signaling in MB [24].

Bmi1: a reason for healthy skepticism

There are striking parallels in both the developmental and tumorigenic roles described for Math1and the polycomb factor Bmi1. Originally identified as a proto-oncogene in the Eµ-myc transgenic mouse [29], Bmi1 is a pro-proliferative gene that has been shown to be necessary for cerebellar development [30]. Specifically, Bmi1^-/- mice are ataxic due to improper growth of the cerebellar EGL [30]. Overexpression of BMI1 is observed in 30–40% of MBs, resulting in inhibition of apoptotic pathways through suppression of p16ink4a and p19arf [31]. Similar to what is described by Flora et al. for Math1[17], deletion of Bmi1 in a medulloblastoma mouse model driven by activated Smo (SmoA1) completely abrogates tumorigenesis [31].

Bmi1transcription is driven in part by SHH-pathway activation; however, Bmi1 overexpression alone appears insufficient to drive MB initiation [32]. When Bmi1 was overexpressed in vivo under the Nestin promoter (Nestin-BMI1-HA^+/-), a marker expressed but not restricted to GCPs, there was no detectable neoplastic phenotype [32]. While this largely rules out an initiating role for Bmi1 in MB, it does not address its role in tumor maintenance or progression. Although MBs show a large number of varied regions of genomic gain and/or amplification, amplification of Bmi1 or Math1have not been seen [27].

Adding it all up: Math1 inhibition as a treatment for MB?

While Math1 overexpression appears to be a hallmark of SHH subgroup MBs, it is unclear whether this event represents a driver or passenger event in the pathogenesis of MB. Amplifications of Bmi1 or Math1are not found in large cohorts of human MB [27]. SHH subgroup MBs represent only 25–35% of all MBs [1], underlining the necessity of stratifying MBs into appropriate subgroups. Overexpression of Math1is linked to a proliferative phenotype, as demonstrated in normal development, and in vitro in MB cell lines. Direct targeting of Math1 using BMPs has demonstrated the possible therapeutic utility of this approach. Perhaps the gold standard to test the role of Math1 in MB maintenance, and hence its suitability as a therapeutic target, would be best tested through inducible withdrawal of Math1 expression in an established MB in an inducible mouse model of MB, as was recently shown for Mycn[33].

Optimal targets for cancer therapy are driver mutations with a critical role in tumor maintenance. While amplification of Math1 is not commonly seen in human MBs, its overexpression may lie downstream of other currently unknown genetic or epigenetic events. Currently, proof for a driver role for Math1in the pathogenesis of MB is lacking. In vivo overexpression experiments have largely discounted a role for Math1 in MB initiation, and its role in tumor maintenance and/or progression remains unclear. Induced deletion of Math1 prior to tumorigenesis may attenuate the malignant phenotype by decreasing the number of EGL cells available for transformation. However, the role of Math1 in prevention of EGL differentiation, and the positive therapeutic result after Math1 inhibition are certainly suggestive of a role in tumor maintenance. The radical, albeit temporary, effect of anti-smoothened therapies in MB, which is followed by drug resistance, cries out for combination therapies. As Math1 expression appears to show great synergy with Shh signaling in the pathogenesis of murine Shh subgroup MBs, it may be an excellent candidate to move forward in combination with anti-hedgehog therapies in this devastating pediatric cancer.

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.