Medulloblastoma: Challenges and advances in treatment and research

Abstract

Background

Medulloblastoma (MB) is a pediatric brain tumor occurring in the posterior fossa. MB is a highly heterogeneous tumor, which can be grouped into four main subgroups: WNT, SHH, Group 3, and Group 4. Each subgroup is different both in its implicated pathways and pathology, as well as how they are treated in the clinic.

Recent Findings

Standard protocol for MB treatment consists of maximal safe resection, followed by craniospinal radiation (in patients 3 years and older) and adjuvant chemotherapy. Advances in clinical stratification of this tumor have allowed establishment of treatment de‐escalation trials aimed at reducing long‐term side effects. However, there have been few advances in identifying novel therapeutic strategies for MB patients due to difficulties in creating chemotherapeutics that can bypass the blood‐brain‐barrier—among other factors. On the other hand, with the help of whole genome sequencing technologies, molecular pathways involved in MB pathogenesis have become clearer and have helped drive MB research. Regardless, this advance in research has yet to translate to the clinic, which may be due to the inability of current in vivo and in vitro models to accurately recapitulate this heterogeneous tumor in humans.

Conclusions

There have been significant advances in knowledge and treatment of medulloblastoma over the last few decades. Whole genome sequencing has helped elucidate clear differences between the subgroups of MB, allowing physicians to better tailor treatments to each patient in an effort to reduce long‐term sequelae. However, there are still many more obstacles to overcome, including less cytotoxic therapies in the clinic and better modeling systems to accurately replicate this disease in the laboratory. Scientists and physicians must work in a more cohesive manner to create translatable results from the laboratory to the clinic—helping improve therapies for medulloblastoma patients.

1. INTRODUCTION

Despite decades of ongoing research, central nervous system (CNS) tumors remain one of the leading causes of cancer‐related childhood mortality.1, 2, 3 Among these, medulloblastoma (MB), a highly malignant cancer originating in the cerebellum, remains the most common pediatric CNS tumor—accounting for 25% of children with intracranial tumors.4 MB is mostly a pediatric tumor, with peak incidence occurring between ages 4 and 7, although adult cases have also been reported.5 Furthermore, MB occurs in more male than female patients.3, 6 While mortality rates for nodular lymphoid hyperplasia (NLH), acute lymphocytic leukemia (ALL), and acute myelogenous leukemia (AML) were reduced by 75% and 60% (NLH/ALL, AML), respectively, survival rates for solid tumors, especially those with disseminated disease (common in MB), have not significantly improved over the past 3 decades.2 MBs have an overall 5‐year survival rate of 75%; however, this relatively high survival rate does not accurately depict the devastating nature of MB. It is a highly heterogeneous tumor consisting of four clinically and pathologically different subgroups. Patients with specific subtypes have dramatically different prognoses, ranging from 50% overall survival in the most aggressive subtypes to 95% overall survival in the less threatening prognoses.1, 6 Although the tumor is known to be heterogeneous, standard treatment is typically consistent across all subgroups. Additionally, MB survivors, particularly former pediatric patients, suffer from developmental, psychosocial, and neurological deficits7, 8, 9, 10—effects of radiation and chemotherapy on a developing brain. The discrepancies in overall survival and the long‐term effects on quality of life highlights the importance of research and its necessity to discover subgroup specific treatment paradigms for these patients. In this review, we will highlight the advances and challenges in therapy and research that physicians and scientists have faced in studying this heterogeneous tumor.

2. MEDULLOBLASTOMA SYMPTOMS, DIAGNOSIS, AND CURRENT THERAPY

According to the 2016 World Health Organization (WHO) classification of tumors of the central nervous system, MB is classified as an embryonal tumor. Subclassifications include genetically defined MB (WNT, SHH + TP53 mutant, SHH + TP53 wild‐type, and non‐WNT/non‐SHH [Group 3 and 4]), histologically defined MB (classic desmoplastic/nodular, extensive nodularity, and large cell/anaplastic), and NOS (not otherwise specified).11, 12 Initial symptoms of MB pathogenesis can occur on an average of 2 months, but can range from a few days to years. Patients typically present with vomiting, headaches, ataxia, and loss of developmental achievements (infants)—symptoms that can be correlated with cerebellar dysfunction and hydrocephalus.13 Hydrocephalus can be caused by compression of the fourth ventricle, which occurs because it is adjacent to the posterior fossa—where MB develops. MB can arise either in the cerebellar hemisphere or along the vermis (midline) of the cerebellum.14 This tumor has a propensity to disseminate through the cerebrospinal fluid (CSF) and colonize the neural parenchyma, with leptomeningeal metastases found along the spinal column. Metastases are found in upwards of 30% of newly diagnosed cases; however, MB rarely spreads outside of the CNS.5, 15

MB diagnosis typically occurs first through a cranial magnetic resonance imaging (MRI), followed by an assessment of leptomeningeal metastasis through a spinal MRI—although computed tomography (CT) is sometimes the first‐line modality because of its availability in emergency settings.16 Metastatic spread in the neural environment is further verified by cytological assessment of CSF attained through a lumbar puncture. MRI shows a cerebellar tumor with compression of the fourth ventricle and dilation of the lateral and third ventricles due to aberrant CSF flow. MB is treated first through resection of tumor followed by adjuvant chemotherapy. The aim of surgery is maximal safe resection of the primary tumor with minimal damage to the surrounding neural architecture and brain function. Surgery reduces mass effect of the tumor, debulks the tumor tissue, and restores CSF flow. In most cases, CSF flow is adequately restored with tumor removal; however, to reduce risk of further CSF blockage, a ventriculoperitoneal shunt is implanted in a majority of patients. To assess residual tumor status, post‐operative MRI is performed in the first 72 hours after surgery, although some groups scan patients as early as 24 hours. Complications of surgery include bleeding, infection, and nonobstructive hydrocephalus. In addition, there is growing evidence that resection of the primary tumor may increase the likelihood of metastatic spread—further underscoring the importance of continued monitoring of the patient's tumor status.17, 18, 19

After tumor resection, patients are administered radiotherapy with adjuvant chemotherapy. Initially, in the latter half of the 20th century, progression‐free survival (PFS) and overall survival (OS) increased dramatically with the usage of post‐operative craniospinal therapy at 36 Gy with a posterior fossa boost of 54–56 Gy in patients greater than 3 years of age.5, 16 Craniospinal radiotherapy is not recommended for children under the age of 3 due to the destructive effects radiation has on a developing brain. In spite of increased overall survival, many patients receiving this therapy succumbed to significant sequelae, including cognitive decline, impairment of growth and endocrine function, hearing loss, and overall reduction in quality of life.7, 8, 9, 10, 20 Thus, efforts were made to reduce the amount of total radiation given to patients and instead to supplement the radiotherapy with maintenance chemotherapy.21, 22, 23 In low‐risk patients, those with no metastases at diagnosis, sparse residual tumor after resection, and classic histology, PFS and OS of over 80% have been achieved with a craniospinal radiation dose of 23.4 Gy and a boost to the posterior fossa of up to 54 to 55.8 Gy with adjuvant vincristine. Following the radiation therapy, maintenance chemotherapy is given, which can consist of vincristine, cisplatin, and either cyclophosphamide or lomustine.24, 25 Conversely, high‐risk patients, who include those with metastases at diagnosis, residual tumor greater than 1.5 cm, and large‐cell or anaplastic histology, continue to have unsatisfactory outcomes despite the advances in treatment. Although OS has increased to 60% to 70% over the last decade, this has been due to high‐dose chemotherapy and alternative fractioned radiotherapy21, 26—whose treatment‐induced late effects are still unknown. High‐risk patients are given 36 to 39.6 Gy radiation followed by four cycles of high‐dose chemotherapeutic agents including cyclophosphamide, cisplatin, carboplatin, and vincristine with autologous peripheral blood stem cell rescue.27 Although there is no defined standard chemotherapy regimen, consensus among physicians states that chemotherapy provides benefit to MB patients.5 However, adverse side effects are still rampant in MB patients, and thus, other methods for treatment de‐escalation are necessary. Factors such as molecular subtypes seem to be more powerful predictors of patient outcome than histology and help dictate responsiveness to therapy.3, 6 Therefore, due to the diversity seen among these tumors, using tools such as molecular subtyping may help physicians treat these tumors, because many MB experts view each subgroup as a distinct disease, which should be treated and studied as such.

Molecular stratification of MB tumors shows that specific subgroups have increased metastatic capabilities.28, 29, 30 Metastases are a major treatment challenge and a cause of death in many MB patients. Dissemination is thought to occur primarily through the CSF, although evidence for metastasis in the neural parenchyma through vasculature has also been shown.31 Anatomical patterning of metastatic nodules among the supratentorial, infratentorial, and spinal compartments shows no significant difference in distribution of metastatic lesions.15 In contrast to non‐metastatic MB patients, who can be treated with a variety of chemotherapeutic regimens, there is defined gold standard treatment for patients with evidence of metastasis. Patients with disseminated disease are treated with high‐dose chemotherapy and hyperfractionated accelerated radiotherapy or carboplatin during craniospinal irradiation (CSI) with a variety of pre‐/post‐CSI chemotherapy regimens. Currently, most studies are directed at identifying specific molecular markers that may help stratify patients with metastasis into low‐/medium‐/high‐risk subgroups to help dictate intensity of treatment.32

3. MEDULLOBLASTOMA STRATIFICATION AND ITS EFFECTS ON OVERALL SURVIVAL: THEN AND NOW

To reduce the drastic side effects of radio‐ and chemotherapy in MB patients, physicians initially stratified patients into risk groups dependent mostly on clinical variables. High‐ and low‐risk patients were grouped based on age, presence of metastasis, and post‐resection residual tumor status.33 Despite clinically defined risk groups allowing lower levels of radiation and chemotherapy to be given to low‐risk patients, MB treatment was still characterized by gradual treatment intensification at the price of deteriorated quality of life.34 This was due to the highly heterogeneous nature of this cancer, and therefore, efforts were made to histologically classify the tumor. Tumors were classified as classic, large cell, anaplastic, desmoplastic, and MB with extensive nodularity.33 However, histologically stratifying patients did not alleviate the heterogeneity in these tumors. Recently, advances in high throughput genetic sequencing technology has allowed researchers to genetically stratify these tumors. This was the first major step in significantly segregating MB patients. Whole genome sequencing of samples showed four clearly distinct tumor populations in MB—WNT (Group 1), SHH (Group 2), Group 3, and Group 4.1, 3, 6, 28, 29, 30 Additional molecular studies indicate additional substructures within these four groups3, 29 and state that MB may include up to 12 subgroups.35 The MB subgroups were so drastically different that it prompted some researchers to proclaim that each of these cancers was merely grouped under the umbrella term medulloblastoma, but should be diagnosed, treated, and studied as four separate cancers.6, 12, 36 Although it is beneficial to stratify this tumor as much as possible, increased group stratification does not necessarily indicate a better road for patient therapies, as this further convolutes treatment protocols for an already heterogeneous tumor.

Moreover, WNT and SHH subgroups are the names for the pathways that are affected. However, no implicated pathways have been found for Groups 3 and 4. Northcott et al1 and Liu et al6 published excellent reviews highlighting the genomics of the four subgroups and would be an excellent resource for further specifics, including genetic pathways implicated in tumorigenesis. Stratification of MB into molecularly segregated subgroups allowed for critical debates regarding de‐escalation in treatment of specific groups, namely, the WNT subgroup.37 The WNT subgroup is the least aggressive of all the subgroups with only 5% to 10% metastases at diagnosis and a mean OS of 95%. Efforts are being made to determine whether reduced chemotherapy will be equally effective and whether post‐operative care is even necessary in this subgroup. Furthermore, assigning patients to one of the four subgroups has also allowed physicians to predetermine how aggressive to be with the treatment paradigm in clinical trials based on survival outcomes and metastasis status.5, 12 In SHH and Group 4 tumors, patients now have over a 75% overall survival. Conversely, MB patients whose tumors are classified as Group 3 have a dismal prognosis with about 40% to 50% overall survival. While SHH patients have a moderate degree of metastasis at diagnosis (15%‐20%), Group 3 and 4 patients suffer metastases in 35% to 45% of cases. This disparity in degree of metastasis and the overall outcome in SHH and Group 3/4 speaks to the highly heterogeneous nature of this tumor. Interestingly, there are reports showing that MB patient samples can show presence of two subgroups in one single tumor, further complicating therapy protocols/identification of tumors.38 As such, further studies need to be performed to determine clear cut descriptions for MB.

4. CHALLENGES AND ADVANCES IN MEDULLOBLASTOMA THERAPY

MB is radically different from any of its cancerous counterparts, especially peripheral tumors.39, 40 Factors such as the blood‐brain‐barrier (BBB), the microenvironment, and responses of cancer stem cells versus the bulk of the tumor tissue to therapy must be considered for MB therapy. The BBB is the principal reason for why brain tumors have a dismal survival rate as many of our current therapeutics, sans radiation, cannot easily penetrate this barrier. The BBB is composed of astrocytes, pericytes, and endothelial cells, which act as a protective barrier to avoid penetration of toxins and other harmful compounds into the neural parenchyma. These include common chemotherapeutic medications as well as hydrophobic, protein bound drugs. Furthermore, the BBB is impervious to any large molecules, which negates the capability to use recombinant proteins as therapies. Moreover, small molecules, which can be in the form of inhibitors targeting specific pathways, also cannot readily cross the BBB, as the layer is 98% impermeable to these smaller molecules.41

Currently, efforts are being made to devise plans to circumvent the BBB. The main areas of focus are optimizing drug delivery and utilizing nanomedicines. While passive diffusion may allow some endogenous compounds and drugs to cross the BBB, more sophisticated methods being applied to enhance drug delivery include disrupting the BBB through use of hydrophobic substances, blocking efflux pumps, and by mechanically circumventing the BBB by local drug delivery. Interestingly, new studies have shown that certain factors released by tumor cells in WNT subgroups MB cancer can actually increase permeability of the BBB, which may increase penetration of chemotherapeutics and thus explain the relatively high cure rate of patients with this tumor.42 Therefore, it may be worthwhile to explore if BBB permeability can be manipulated from inside the brain itself to increase bioavailability of chemotherapeutic drugs in the tumor parenchyma. Furthermore, evidence has shown that utilization of transporters through absorption‐mediated transcytosis43, 44 and receptor‐mediated endocytosis43, 45 may also be beneficial for bypassing the BBB. As such, until scientists and physicians are able to learn how to bypass the BBB to effectively deliver medication, advances in MB therapy will be limited to dose de‐escalation studies and clinical trials experimenting with different chemotherapy cocktails—in order to reduce the devastating side effects of treatment. To date, a review of completed clinical trials with results in MB (Table 1) reveals no important discoveries in treatment with patients having no response or limited response to the medications. Most studies focus on novel drug therapies that are not adept at bypassing the BBB, de‐escalation of treatments, or different cocktails of chemotherapeutics; however, each study shows minimal progression free survival or observed overall survival.

Table 1.

Fourteen clinical trials for medulloblastoma that were completed and had results as of July 1, 2018^a

Study Title	Drug Interventions	Start/Completion Date	Final Enrollment	Age Range: Total # of Patients, Median (y), Range (y)	Primary Outcome Measure	Primary #2/Secondary Outcome Measure
A Phase II, Multi‐center, Open‐label, Single‐arm Study of the Efficacy and Safety of Oral LDE225 in Patients With Hh‐pathway Activated Relapsed Medulloblastoma	Chemotherapy: LDE225 (sonidegib, orally), temozolomide (orally)	May 6, 2013/October 5, 2016	22 patients	Children: 2, 8.5, 4‐13 Adults: 16, 37, 24‐51 Control: 4, 35.5, 31‐38	OOR with LDE225: 0% in children 18.8% in adults OOR with TMZ: 0%	PFS with LDE225: 1.6 mo in children 3.3 mo in adults PFS with TMZ: 2.9
Phase II Clinical Trial Evaluating the Efficacy and Safety of GDC‐0449 in Adults With Recurrent or Refractory Medulloblastoma	Chemotherapy: GDC‐0449 (vismodegib, orally)	June 2009/August 2015	31 patients	PTCH/SHH (−): 8, 23.8, 22.4‐40.6 PTCH/SHH (+): 20, 32.0, 23.5‐51.9 PTCH/SHH (?): 3, 32.9, 25.3‐38.6	OR: PTCH/SHH (−): 0% PTCH/SHH (+): 15% PTCH/SHH (?): 0%	PFS: PTCH/SHH (−): 1.64 mo PTCH/SHH (+): 2.76 mo PTCH/SHH (?): 1.48 mo
A Phase II Clinical Trial Evaluating the Efficacy and Safety of GDC‐0449 in Children With Recurrent or Refractory Medulloblastoma	Chemotherapy: GDC‐0449 (vismodegib, orally)	November 2010/August 2015	12 patients	12, 10.4, 3.9‐20	OR sustained for ≥ 8 wk: 1 patient	PFS: 1.41 mo
A Study Evaluating Limited Target Volume Boost Irradiation and Reduced Dose Craniospinal Radiotherapy (18.00 Gy) and Chemotherapy in Children With Newly Diagnosed Standard Risk Medulloblastoma: A Phase III Double Randomized Trial	Chemotherapy: Cisplatin (IV), cyclophosphamide (IV), lomustine (orally), vincristine (IV) Radiation: Craniospinal irradiation, involved‐field radiation therapy	April 2004/March 2016	493 patients	Arm I (LDCSI, IFRT): 63, 5.9, 3.2‐7.8 Arm II (LDCSI, PFRT): 64, 5.9, 3.3‐7.9 Arm III (SDCSI, IFRT): 60, 5.9, 3.0‐8.0 Arm IV (SDCSI, PFRT): 58, 5.4, 3.1‐7.7 Arm V (SDCSI, IFRT): 120, 12.4, 8.0‐19.8 Arm VI (SDCSI, PFRT): 128, 11.8, 8‐21.8	EFS (units: Probability of 3 y EFS with 95% CI) LDSCI EFS: 76.3 (67.9 to 84.7) SDCSI EFS: 84.9 (77.6 to 92.2) IFRT EFS: 85.8 (81.1 to 90.5) PFRT EFS: 85.8 (81.1 to 90.5)	OS (units: Probability of 3 y OS rate with 95% CI): LDSCI OS: 85.5 (78.4 to 92.6) SDCSI OS: 90.4 (84.3 to 96.5) IFRT OS: 90.3 (86.2 to 94.4) PFRT OS: 93.3 (90.0 to 96.6)
A Phase 1 Trial of TPI 287 as a Single Agent and in Combination With Temozolomide in Patients With Refractory or Recurrent Neuroblastoma or Medulloblastoma	Chemotherapy: TPI 287 (IV) Temozolomide (orally)	March 2009/February 2016	18 patients	18, 7.5, 4.0‐25.0	Patients with advese events as a measure of safety and tolerability: 14	OOR: 1 patient
Phase 2 Single‐Arm, Open Label Study Of Irinotecan In Combination With Temozolomide In Children With Recurrent Or Refractory Medulloblastoma And In Children With Newly Diagnosed High‐Grade Glioma.	Chemotherapy: Irinotecan (IV) Temozolomide (orally)	April 2007/December 2011	83 patients	Medulloblastoma: 66 2 y to 12 y: 49 12 y to 18 y: 17 Glioblastoma: 17 2 y to 12 y: 12 12 y to 18 y: 5	TMZ + irinotecan objective response of complete response or partial response in (% of participants): Medulloblastoma: 32.6 Glioblastoma: 0	Duration of response (week) in TMZ + irinotecan treatment: Medulloblastoma: 22.4 Glioblastoma: 36.3
Phase II Study of Bevacizumab Plus Irinotecan (Camptosar™) in Children With Recurrent, Progressive, or Refractory Malignant Gliomas, Diffuse/Intrinsic Brain Stem Gliomas, Medulloblastomas, Ependymomas and Low Grade Gliomas	Biological: Bevacizumab (IV) Radiation: Fludeoxyglucose F‐18 Chemotherapy: Irinotecan hydrochloride (IV)	August 2006/October 2015	97 patients	High grade gliomas: 18, 15.07 Brain stem tumors: 17, 8.45 Medulloblastoma: 10, 10.56 Ependymoma: 15, 10.16 Low grade glioma: 37, 8.58	Objective response rate sustained for greater than 8 wk: 0 patients across high grade gliomas, brain stem tumors, medulloblastoma, and ependymoma	Sustained disease stabilization rate associated with bevacizumab and irinotecan in patients with recurrent or progressive low‐grade glioma: 23/35
A Phase I/II Study of LDE225 in Pediatric Patients With Recurrent or Refractory Medulloblastoma or Other Tumors Potentially Dependent on the Hedgehog‐signaling Pathway and Adult Patients With Recurrent or Refractory Medulloblastoma	Chemotherapy: LDE225 (Sonidegib, orally)	February 2011/October 2014	Phase I: 59 patients started, 27 patients completed Phase II: 17 patients started, 5 patients completed	LDE225 233 mg/m2: 11, 4 patients ≤10 y, 7 patients >10‐17 y LDE225 372 mg/m2: 16, 3 patients ≤10 y, 13 patients >10‐17 y LDE225 425 mg/m2: 11, 5 patients ≤10 y, 6 patients >10‐17 y LDE225 680 mg/m2: 22, 11 patients ≤10 y, 11 patients >10–17 y LDE225 800 mg: 16, 16 patients 18‐65 y	Number of participants with DLT in phase I: 0 patients in 233, 425, and 680 mg/m2 groups, 1 patient in 372 mg/m2 group	% of participants with objective response rate by treatment: Pediatric/adult patients: 3.3%/12.5% complete response 0%/6.3% partial response 8.3%/37.5% stable disease 76.7%/37.5% progressive disease 3.3%/18.8% ORR
Molecular Biology and Phase II Study of Lapatinib (GW572016) in Pediatric Patients With Recurrent or Refractory Medulloblastoma, Malignant Glioma or Ependymoma	Chemotherapy: Lapatinib (orally)	October 2004/July 2010	52 patients	MB: Lapatinib + surgery: 2, 1 patient ≤18 y, 1 patient between 18 and 65 Surgery only: 2, 2 patients ≤18 y No surgery/lapatinib: 17, 12 patients ≤18 y, 5 patients between 18 and 65 High grade glioma: No surgery: 13, 12 patients ≤18 y, 1 patient between 18 and 65 Ependymoma: Lapatnib + surgery: 2, 2 patients ≤18 y Surgery only: 2, 2 patients ≤18 y No surgery/lapatinib: 14, 14 patients ≤18 y	Relative phosphorylation of ERBB2: Lapatinib prior to surgery: 4 patients analyzed, 0.21 ratio No lapatinib prior to surgery: 3 patients analyzed, 0.5 ratio	Participants with sustained objective response (complete or partial response): Recurrent medulloblastoma: 0/15 Recurrent high grade glioma: 0/13 Recurrent ependymoma: 0/13
A Randomized Phase III Study of Sodium Thiosulfate for the Prevention of Cisplatin‐Induced Ototoxicity in Children	Sodium thiosulfate (STS, IV)	June 2008/April 2015	131 patients	STS arm: 65, 65 patients ≤18 y Observation arm: 66, 66 patients ≤18 y	Incidence of hearing loss (4 wk after last dose of cisplatin): STS arm: 14/49 patients observation arm: 31/55 patients	Change in hearing thresholds (dB) in STS/observation arms: 500 Hz: −1.45, 5.8 SD/−1.11, 8.59 SD 1000 Hz: −0.676, 4.59 SD/−0.319 8.99 SD 2000 Hz: −1.18, 4.85 SD/0.638, 12.7 SD 4000 Hz: 1.05, 7.09 SD/9.58, 20.5 SD 8000 Hz: 9.73, 17.3 SD/17, 24.7 SD
A Phase II Study of Pemetrexed in Children With Recurrent Malignancies	Chemotherapy: Pemetrexed (IV)	September 2007/February 2010	72 patients	Osteosarcoma: 10, 14.94 Ewing's sarcoma/peripheral PNET: 11, 18.24 Rhabdomyosarcoma: 9, 8.74 Neuroblastoma (measurable disease): 6, 9.62 Neuroblastoma (mIGB+ evaluable): 6, 9.62 Ependymoma: 10, 8.42 Medulloblastoma/Supratentrial PNET: 11, 12 Non‐brainstem high grade glioma: 10, 12.75	Percentage of patients with overall tumor response: 0 patients across all tumor subtypes	Number of patients with AdEERs, discontinuations, deaths due to study drug: Osteosarcoma: 3 AdEERs/10 patients Ewing's sarcoma/peripheral PNET: 2 AdEERs, 1 discontinuation/ 11 patients Rhabdomyosarcoma: 0 Neuroblastoma (M.D.): 2 AdEERs/5 patients Neuroblastoma (mIGB+ evaluable): 3 AdEERs/6 patients Ependymoma: 2 AdEERs/ 10 patients Medulloblastoma/Supratentorial PNET: 2 AdEERs, 2 discontinued/11 patients Non‐brainstem high‐grade glioma: 2 AdEERs/10 patients
Phase II Study of Arsenic Trioxide in Neuroblastoma and Other Pediatric Solid Tumors	Chemotherapy: Arsenic trioxide (IV)	March 2001/May 2009	22 patients	22, 19 patients ≤18 y, 3 patients between 18 and 65 y	Response rate after every 3 courses during drug treatment and then every 2–3 mo for 1 y after completion of treatment: 21 patients analyzed, 16 had progression of disease, 5 had stable disease	…
Phase III Double Blind, Placebo Controlled Study of Donepezil in the Irradiated Brain	Chemotherapy: Donepezil hydrochloride (orally)	January 2008/June 2012	198 patients	Arm I (donepezil): 99, 72 patients between 18 and 65 y, 27 patients ≥65 y Arm II (control): 99, 83 patients between 18 and 65 y, 16 patients ≥65 y	Memory as quantified by HVLT‐immediate recall (units on a scale): Arm I (donepezil): 22.5, 0.45 SE Arm II (control): 22.2, 0.45 SE	Memory as quantified by HVLT‐discrimination (units): Arm I (donepezil): 10.1, 0.24 SE Arm II (control): 9.2, 0.24 SE
A Pilot Study of Donepezil Hydrochloride in Pediatric Brain Tumor Survivors After Cranial Irradiation	Chemotherapy: Donepezil hydrochloride (orally)	June 2006/February 2010	14 patients	13, 13 patients ≤18 y	Neurocognitive function as measured by the neurocognitive battery at 24 wk (units on a scale): Donepezil: 10, 3 SD	…

Abbreviations: AdEERs, adverse event expedited reporting system; DLT, dose‐limiting toxicity; EFS, event free survival; FRT, smaller volume boost (radiation to tumor bed only); HVLT, Hopkins Verbal Learning Test; IPFRT, standard volume boost (radiation to the entire posterior fossa); LDSCI, reduced‐dose craniospinal radiation; OOR, overall response rate; OR, objective response; OS, overall survival; PFS, progression free survival; SD, standard deviation; SDCSI, standard‐dose craniospinal radiation.

^a

5. CHALLENGES AND ADVANCES IN MEDULLOBLASTOMA RESEARCH

Studying medulloblastoma is a difficult task for many reasons. First, MB is not grouped as a carcinoma or a sarcoma, so prototypical pathways found affected in these cancers are not readily translatable to this tumor. Furthermore, because it typically occurs in children, the mutation rate is not high, with an average of only eight mutations per tumor.46 In comparison, lung cancer has upwards of 200 mutations, allowing for targetable proteins to be found. However, this is very difficult in MB, as evidenced by the lack of an implicated pathway in Group 3 and 4 MBs. Importantly, while genetic alterations are not common in MB, there are significant epigenetic and cytogenetic alterations that are implicated in its pathogenesis and will be discussed further in this section. Below, we list the challenges researchers face in studying this disease and advances that have helped overcome these challenges.

5.1. Origin of medulloblastoma

The cerebellum is arguably the most complex region of the brain. Not only does it house half of the mature neurons in the brain47 but also it houses several different types of cells—each requiring substantial migratory capabilities to reach their anatomical position. Furthermore, each cell type is responsible for differing functions. These cells include the Bergmann glia, inhibitory Purkinje cells, excitatory granule cells, and inhibitory interneuron subtypes.48, 49 Correct orientation of these vastly different cells is required for proper functioning of the cerebellum. During embryonic development, pathways in WNT/beta‐catenin, TGF‐B/BMP, SHH/PTCH1, and Hippo pathways are highly expressed. In MB development, these pathways are mutated, lost, or aberrantly regulated.50 Although these molecular pathways have been shown to be implicated in MB pathogenesis, the precursor cells involved in the development of the different subgroups of MB has been more difficult to ascertain. Currently, studies have shown: Dorsal brainstem neuronal progenitors lead to WNT tumors, EGL and rhombic lip granule neuron precursors lead to SHH, and Group 3 tumors and ventricular zone neural progenitors lead to Group 3 tumors as well. No precursor cell has been identified for Group 4 tumors currently.1, 6, 51

One of the theories on medulloblastoma formation focuses on aberrant migratory capabilities in cerebellar granule neural precursors as one of the factors for MB onset. In the developing brain, the external granule layer (EGL) is a collection of cells lining the outer wall of the immature cerebellum. As the brain develops, cells in the EGL migrate across the Purkinje cell layer and increase proliferation to eventually form the internal granule layer (IGL). However, scientists have shown evidence that SHH tumors can be formed when the migratory capacity of granule cells is disrupted, causing rapidly proliferating cells, induced by SHH secretion by Purkinje cells, to be stuck in the external granule zone that induce formation of SHH tumors.48, 52

5.2. Cancer stem cells

In the last decade, there has been increasing evidence that cancer stem cells (CSC) have self‐renewal abilities that can give rise to the heterogeneous progeny of tumors.53 Moreover, they are responsible for therapeutic resistance, tumor relapse, and invasion—which pose a great challenge for long‐term survival of patients.54 CSCs can be confused with cells of origin; however, there is a stark contrast between the two. While cells of origin are the normal cells that undergo transformation to induce tumorigenesis,55 CSCs are tumor cells that have multiple mutations allowing them to become “stem‐like” cells. As previously stated, MB cells of origin have been difficult to identify; however, multiple markers have been classified. Currently, four main markers have been identified as potential MB CSCs: CD15 (FUT4/SSEA),56 CD133 (PROM1),54, 57 CD271 (NGFR),58, 59 and SOX2.60 However, there is much discrepancy on whether these markers are indicative of CSCs or progenitor cells and if presence of other markers is also necessary to be deemed a CSC. CD15 and SOX2 have been shown to select for stem cells in the SHH subgroup of MB.56, 60 SOX2‐positive cells are enriched in chemotherapy‐treated SHH tumors, indicating that SOX2‐positive MB cells may be responsible for tumor relapse.60 Conversely, while there was evidence for CD15 as a CSC marker, conflicting reports suggest that CD15‐positive cells are also typically Math1 positive and are thus behave more like progenitor cells rather than CSCs.61

Moreover, CD133 is a common CSC marker among many subtypes of tumors.62 Initially, it was shown that CD133 cells, which were also Nestin positive, had an increase in proliferation, self‐renewal, and differentiation in vitro.54 Moreover, CD133‐positive cells were able to induce tumor in vivo, whereas CD133‐negative cells were not.57 However, this finding was called into question when it was shown that CD133‐negative cells also showed tumor‐initiating properties.61, 63 Discrepancies in these studies may be due to the fact that CD133 is not exclusive to tumor cells and is also expressed by normal stem cells64 and differentiated epithelial cells.65 This further suggests that MB CSCs most likely express multiple markers that indicate their ability to become self‐renewing and tumor propagating cells. Additionally, CD271‐positive cells have been shown to have a higher capacity for noninvasive self‐renewal in contrast to increased invasive capabilities in CD133‐positive cells.59 CD271 expression is higher in SHH subgroup tumors but lower in more aggressive Group 3 tumors.58, 59 This indicates an interesting link between cell surface marker expression and clinical presentation of these tumors—warranting further investigation as to the relation of the CSC marker expression, tumor subgroup, and how the tumor acts in vivo. Although these markers have been shown to be associated with CSC phenotypes, the lack of agreement between different investigators indicates that further experiments are needed to establish a gold standard for MB CSC markers.

5.3. In vitro and in vivo tools for medulloblastoma research

5.3.1. Cell lines

Utilization of cell lines for in vitro experiments is commonplace in the research field. It is preferred to utilize low passage cell lines derived from patient tissues because they are more likely to accurately represent the genotype of the original tumor. Outside of patient sample‐derived cell lines, established and verified MB cell lines are difficult to obtain. There are five cell lines readily available from ATCC: DAOY, D341 Med, D283 Med, CHLA‐01‐MED, and CHLA‐01R‐MED. For such a vastly heterogeneous tumor, only having five verified cell lines to utilize for experiments limits how well studies in the laboratory can be clinically translatable. This understates the importance of deriving novel cell lines from patient samples to increase the amount of cell lines that can be utilized by researchers. Efforts must be made between the patient, the operating physician, the researcher, and multiple hospitals/institutions to receive resected tumors and attempt to establish novel cell lines. This provides the best avenue for achieving improvements in the clinical realm because studies will be done on samples that have not been acclimated to in vitro settings. In addition to the ATCC lines, other cell lines, stratified into molecular subgroups, have also been used in numerous peer‐reviewed publications.66, 67, 68 Ivanov et al has published an excellent review highlighting the subgroup and other molecular features of these cell lines.69 More MB cell lines do exist; however, they have yet to be stratified into a molecular subgroup. Cell lines are an extremely important tool in research as they allow scientist to test hypotheses outside of in vivo conditions. However, to act as a bridge between the laboratory and the clinic, cell lines must accurately represent how MB behaves in vivo. Unfortunately, this is not the case in MB research. One of the most utilized cell lines, D283 Med, is a MB tumor that was derived from a tumor resected from the peritoneum of the patient.70 Although it may harbor similarities to neural‐derived MB cells, MB tumors are very rarely found outside the CNS, and thus, this cell line is not a good model to study MB. Furthermore, not only are there differing medical reports on the patients, including differences in sex of the patient and resection location, genetic aberrations and karyotypes often differ as well.70, 71 Although this can be attributed to mistakes in reporting, it is well known that some expression profiles are artifacts of cell culture. Gene expression and metabolic profiles can change drastically from original tumor to cell line. In MB, it is known that certain genetic aberrations may also change, as shown when a tumor mass that gave rise to MED6 cells had a CTNNB1 mutation, while the resultant cell line had a wild‐type CTNNB1.72 Moreover, gene expression profiles of subgroup‐specific cell lines differ from gene expression profiles from patient tissues obtained from bioinformatics databases. While this may be due to the fact that some tumors may contain stromal or normal brain tissues, the disagreement between expression profiles in patients versus cell lines is an obstacle in MB research advancement. These same challenges can also be found in MB mouse models.

5.3.2. Medulloblastoma mouse models

Mouse models are an excellent tool to recapitulate MB pathogenesis in vivo. Utilizing mouse models allows for highly controllable genetic alterations in specific cell subtypes, hosts with intact immune systems, and ability for tumors to grow in their natural tumor microenvironment. Mouse models for WNT and SHH can be established because the implicated pathways for these subgroups is well characterized. As of 2017, only one model of WNT medulloblastoma has been established, by Gibson and colleagues,73 which utilizes a CTNNB1 aberration in conjunction with TP53 mutation (mutations present in human WNT medulloblastoma) to form WNT like tumors in mice. This mouse model not only was important for studies leading to clinical trials but also helped identify the progenitor cells for WNT MB. This model was also used to show that blood vessels associated with WNT MB tumors are fenestrated, allowing chemotherapeutic drugs to readily reach tumor cells and exert their cytotoxic effects. This may explain why patients with WNT tumors have upwards of 95% overall survival rate. While WNT MB only has one mouse model, there are a number of experimental models depicting SHH MB. There are a wide range of mutations used that accurately depict histological as well as infant/adult SHH tumor phenotypes. SHH MB tumorigenic pathways are well defined, with mutations occurring in either/or PTCH1, SUFU, and SMO.74, 75, 76, 77, 78 Most existing models are based on mutations in PTCH179, 80 or SMO,81, 82 which models adult SHH. However, no good models exist with mutations and amplifications in downstream parts of the SHH pathway such as Gli1, Gli2, and SUFU. In addition, most of the models require mutations in TP53. This mutation is uncharacteristic of the general patient population,83 although germline TP53 are mutations that are found in a subset of SHH patients who have Li‐Fraumeni Syndrome.84, 85, 86, 87 However, a recent transgenic mouse model overcomes the need for a TP53 mutation for establishing SHH tumors in vivo. Shackleford et al utilized the Barhl1 homeobox gene promoter, which is specific to cerebellar granule neuron precursors, to express an avian retroviral receptor TVA. By targeting this specific cell subtype, which is known to give rise to MB,88 they were able to selectively overexpress SHH and MYCN in these cells by injecting an avian retroviral vector encoding these genes—causing formation of SHH‐like MB in these mice.89

In contrast to the well‐characterized WNT and SHH MB groups, the latter of which contains dozens of suitable models, the other two subtypes of MB, Group 3 and 4, have been more difficult to model in vivo. Group 3 MB has a few available models, and most models focus on MYC‐driven tumorigenesis. MYC overexpression has been targeted in Nestin‐positive brain cells in conjunction with overexpression of Bcl‐2.90 Other groups overexpressed MYC in TP53‐negative cerebellar stem cells expressing CD133 but not lineage defined and achieved tumors that resembled large cell anaplastic MB.91 It is important to mention that TP53 mutations do not exist in Group 3 MB,83 and MYC overexpression only occurs in 15% to 20% of Group 3 human tumors.92 On the contrary, OTX2 amplification is frequently seen in this subgroup93; however, no good models exist using this aberration. Group 4 MB tumors have the highest incidence in the clinic, yet no mouse models have been established, highlighting the lack of knowledge understanding the biology of this subgroup. Most models target overexpression of MYCN; however, most tumors derived from these models resemble SHH or Group 3 MB and cannot be reliably used as models for Group 4 MB.94, 95

Albeit the shortcomings, mouse models have led to great advances in how physicians treat MB patients. However, as optimistic as we can be that these models depict tumors in humans, many mouse models were created using additional mutations in DNA‐repair pathways and interferon signaling, whose specific roles in tumorigenesis are unclear in human MB. Moreover, certain mutations, such as in the TP53 gene, or amplification of MYC, either do not exist in the patient population or relates to a small percentage of patients with germline mutations in these genes. This limits the ability to accurately recapitulate this disease in vivo. Although mouse models represent a far greater tool to study MB, further advances in identifying and developing genetic aberrations that model mutations in humans are necessary.

5.4. Genetic studies in medulloblastoma

5.4.1. Whole genome sequencing of medulloblastoma samples

Throughout the past decade, advancements in whole genome sequencing technologies have allowed researchers to analyze hundreds of MB tissues for their genetic profiles. Sequencing has allowed researchers to identify and correlate certain alterations (amplification, deletion, single nucleotide variants, indels, overexpression, translocation, and inversion) with specific subgroups.96, 97, 98 Moreover, this has allowed for greater stratification of this heterogeneous tumor, by allowing aberrations to be correlated with age and gender of the patient as well. Although WGS has greatly advanced the knowledge of MB and its drastic heterogeneity, the implications of these alterations have not been studied extensively. Thus, in order to fully utilize the large amount of data generated by WGS, researchers must understand how these changes in the genome and subsequent alterations in gene expression affect tumorigenesis in patients.

Excitingly, this tall task is not limited to the laboratories that produced the datasets. Most of the data extracted from WGS is readily available to researchers through online databases. These databases include the R2 Genomics Analysis Visualization Platform (http://r2.amc.nl), cBioPortal for Cancer Genomics (www.cbioportal.org), and recently the St Jude Cloud (https://stjude.cloud). R2 continuously adds databases, most recently the database published by Cavalli et al,35 which contains 763 patients, the highest amount of MB patient data in one dataset. The St Jude Cloud is a new database that allows users to access data generated from the hundreds of patients that visit St Jude Children's Research Hospital. Having this data allows researchers to peek into the clinical world to understand which genetic aberrations are present in human patients. This gives scientists the ability to model their experimental systems using clinical data in hopes of achieving results more efficiently to help advance patient care.

5.4.2. Cytogenetic changes, SCNAs, and epigenetic alterations in medulloblastoma

MB tumors have on average eight mutations per tumor, which is on the lower spectrum of mutations typically found in cancers.46 However, MB patients (specifically those with WNT and SHH tumors) can be predisposed to form MB tumors by germline mutations. Patients diagnosed with WNT tumors have somatic mutations of CTNNB1,99 which promotes stabilization and nuclear localization of β‐catenin,100, 101 and patients are identifiable by WNT gene expression signature.102 SHH tumors harbor germline mutations that affect PTCH1 or SUFU, which are negative regulators of SHH signaling pathway.74, 76, 103, 104, 105, 106 Conversely, there are no known germline mutations that predispose patients to either Group 3 or 4 medulloblastoma.

Moreover, like most cancers, MB is susceptible to cytogenetic alterations. From a cohort of 827 medulloblastoma patients, Northcott et al derived cytogenetic profiles across all four subgroups.1 WNT tumors typically have balanced genomes, but are known to harbor chromosome 6 monosomy.98, 107 Patients with SHH tumors, typically older patients, exhibit chromothripsis, leading to extensive fracturing of one or more chromosomes.87 Most SHH patients have chromosome 3q and 9p gain and chromosome 9q, 10q, 14q, and 17p loss.107 Group 3 MB possesses chromosome 1q, 7, 17q, and 18 gain and chromosome 8, 10q, 11, 16q, and 17p loss. Finally, Group 4 MB harbors chromosome 4, 7, 17q, and 18 gain with chromosome 8, 10, 11, and 17p loss.1, 98

Changes in cytogenetics frequently lead to somatic copy number aberrations (SCNAs). Because WNT tumors typically have stable chromosomes, they are typically devoid of SCNAs.98, 107 In SHH tumors, mutations in PTCH1 and SUFU can amplify the expression of SHH target genes MYCN and GLI2.28, 29, 30, 100 Moreover, deletions in PTCH1 and TP53 are also observed.87, 108, 109 In Group 3 MB, patients frequently exhibit amplification of the MYC oncogene29, 30, 98 and OTX2.35, 110 OTX2 amplification is mutually exclusive with MYC amplification and is said to promote tumorigenesis by upregulating MYC expression.93 Among patients with MYC amplified tumors, a majority had fusion of PVT1‐MYC proteins,87, 97, 111, 112 which has been suggested to help stabilize the MYC protein.98, 113 Moreover, amplifications and deletions in transforming growth factor‐β (TGFβ) signaling have also been found in Group 3 MBs.98 Finally, Group 4 patients do not have common molecular aberrations6; however, a subset of tumors harbor MYCN,3 SNCAIP,35 and CDK698 amplifications.

Although germline mutations are rare in MB, whole genome sequencing has identified that these mutations can occur in proteins that can affect the global epigenetic landscape—specifically histone modifications. It has been elucidated that MB tumors have stark epigenetic changes, which can be reflective of their subgroup status and prognostic outcome.114 For example, mutations in MLL2 and MLL3, the catalytic component in trithorax proteins that is responsible for methylation of histone H3 at the fourth lysine position (H3K4me), are found across all groups in MB.97, 111, 112, 115, 116, 117 Furthermore, inactivation of the KDM family of proteins, which are lysine demethylases that erase H3K27me3 marks on histones, is also seen in MB tumors. H3K27me3 marks repress lineage specific genes in stem cells.118 When H3K27me3 is erased by KDM gene family proteins, chromatin remodelers are recruited to H3K4me marks to promote differentiation. The biological repercussions of such mutations is an inappropriate silencing of pro‐differentiation genes—keeping cells in a poorly differentiated state typically indicative of aggressive tumors.119, 120 Other epigenetic modifiers can also be mutated, and a list of these can be found in the following references.112, 121

6. CONCLUDING REMARKS

Although there are many challenges in studying and caring for patients with a heterogeneous tumor, research in MB is advancing at a fast rate. Physicians and scientists are making progress; however, there needs to be a stronger relationship between science and medicine. Currently, most clinical trials are aimed at reducing/altering chemotherapeutic and radiation treatments in hopes of reducing devastating side effects in patients. There are very few trials that are attempting to utilize novel drugs or to address the more aggressive subtypes of MB. Absence of novel therapeutics for patients can be directed at the lack of research to identify targets and their mechanism of action. Although researchers are continuing to find more data highlighting the intense heterogeneity in the tumor, there is still a lack of understanding of how these mutations are implicated in the development and pathogenesis of MB. Continued stratification of patients does not lead to better therapies, and thus, we must begin to utilize the genetic data to understand the biological mechanism of MB.

CONFLICT OF INTEREST

The authors have no conflicts of interest to report.

AUTHORS' CONTRIBUTION

All authors had full access to the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Conceptualization, V.M., J.N.; Methodology, V.M., J.N.; Investigation, V.M.; Formal Analysis, V.M.; Resources, V.M.; Writing ‐ Original Draft, V.M.; Writing ‐ Review & Editing, V.M., J.N.; Visualization, V.M., J.N.; Supervision, J.N.; Funding Acquisition, J.N.

Martirosian V, Neman J. Medulloblastoma: Challenges and advances in treatment and research. Cancer Reports. 2019;2:e1146. 10.1002/cnr2.1146