Intracranial germ cell tumors (ICGCT) are rare in North America and Europe, occurring in 2-4% of children with brain tumors between birth and 19 years of age^6,4465). By contrast, ICGCT is the most common supratentorial brain tumor in Japan, Taiwan, and Korea, accounting for 11-15% of brain tumors in children and adolescents^14,3049,67). ICGCT occurs as midline lesions around the third ventricle, most commonly involving the pineal gland and the neurohypophyseal region of the brain, which includes the infundibulum and/or pituitary stalk. Less commonly, tumors occur in the basal ganglia or thalamic nuclei and, rarely, in the floor of the fourth ventricle at the medulla (Fig. 1A). ICGCT arising in the basal ganglia (Fig. 1B-F) is worthy of mention because of its clinical characteristics, especially male predominance, and the difficulties in diagnosing and evaluating the response to treatment (Fig. 1C-F). Hemiparesis is the most common symptom, and it rarely resolves even with long-term disease control. It affects 4-15% of patients with ICGCT and predominantly occurs in boys^47,5961,66). Up to 20% of ICGCT present as multiple midline tumors. It is unclear whether multicentric tumors are present from tumor initiation or whether microscopic tumors infiltrate into the subependymal periventricular area from a single tumor. Bifocal tumors are more common in germinomas than in mixed ICGCT²⁹⁾. However, it has been reported that a small number of patients with normal or slightly elevated levels of α-fetoprotein (AFP) and β-human chorionic gonadotropin (β-HCG) have tumors with a malignant germ cell component on pathologic examination¹⁾. The natural course of bifocal germinomas is not significantly different from that of localized intracranial germinoma, which can be controlled with an extended focal radiotherapy field⁶⁴⁾. However, 23-55% of bifocal tumors were shown to harbor metastatic tumors when closely evaluated by imaging or cytology imaging studies or cytologic examinations were performed^48,64). Therefore, careful evaluation of brain and spinal cerebrospinal space metastases is critical at initial presentation.

The survival rate of patients with pure germinoma appears to be very high, reaching a 10-year overall survival rate of 90%^{13,2432,3756),} whereas that of patients with non-germinomatous GCT (NGGCT) tends to be 30-80%^27,3435). A Japanese study group examined 153 patients with ICGCT, of which 147 patients whose tumors were surgically resected were assessed to examine the correlation between histologic examination and survival outcome. According to their study, three groups of patients showed significantly different survival rates of 94.1%, 70%, and 9.3%. The three therapeutic groups were classified as showing good, intermediate, and poor prognosis, respectively (Table 1)^20,29).

NGGCT is diagnosed as ICGCT with any malignant germ cell component and/or any tumor that secretes AFP or high levels of serum human chorionic gonadotropin. High β-HCG refers to a level greater than 50-100 IU; however, the cut-off value for the diagnosis of NGGCT varies between institutes or study groups. Tumors with high β-HCG concentrations were considered to be associated with worse prognosis by some researchers^4,5160). However, many other studies have shown that an elevated β-HCG level is not particularly associated with poor prognosis^{23,2628,3442,43)} when there is no evidence of dissemination through the cerebrospinal fluid (CSF). It is uncertain what level of β-HCG should be regarded as high, because the cut-off level of elevated β-HCG differs in each study. In a study of 58 patients with a newly diagnosed intracranial germinoma who underwent histologic confirmation by biopsy and complete staging work-up, 60% had normal β-HCG levels in both serum and CSF, 34.5% had a normal serum level and an elevated CSF level, 3.5% had elevated levels in both serum and CSF, and 2% had an elevated serum level and a normal CSF level³⁾. Although a β-HCG cut-off value of ≥50 mIU/mL in either serum or CSF was arbitrarily selected by several cooperative groups, the absolute β-HCG ceiling consistent with a pure central nervous system (CNS) germinoma has not been established. Previously, the Children's Oncology Group study ACNS0232 used serum or CSF β-HCG levels of 50 mIU/mL as the cut-off value for NGGCT diagnosis, and the recently started ACNS1123 study includes patients with β-HCG levels ranging from 50 to 100 mIU/mL. By contrast, the Japanese CNS GCT study group uses a much higher β-HCG cut-off level of 2000 mIU/mL for the diagnosis of NGGCT in patients with a poor prognosis, and includes patients with a histological diagnosis of choriocarcinoma, embryonal carcinoma, and yolk sac tumors. The difference between the current treatment strategy of the Japanese CNS GCT study group and other groups is that the former strongly advocates surgical biopsy for all ICGCTs. Therefore, the diagnosis is less dependent on serum and CSF tumor markers (personal communication, Masao Matsutani, Saitama International Medical Center, Hidaka, Japan). Currently, there is a debate about the role of surgery regarding its contribution to a superior treatment outcome when given before^35,51) or after radiotherapy combined with chemotherapy⁵⁰⁾.

The origin of ICGCT is considered to be primordial germ cells of the embryo²⁰⁾. Human primordial germ cells from the yolk sac endoderm migrate along the dorsal mesentery of the hindgut to the genital ridges for incorporation into developing gonads during week 4 of embryogenesis. Failure to initiate migration, abnormal migration, abnormal termination of migration, and entrapment during the migration, observed in midline extragonadal areas in chick, human, and mouse embryos, appears to coincide with the occurrence of extragonadal GCT, including CNS GCT²⁰⁾. Studies of the molecular and cytogenetic characteristics of ICGCT are rare because of the infrequency of surgical biopsy and the small size of specimen even if biopsies are performed. As shown in methylation and comparative genomic hybridization studies, both germinoma and NGGCT originate from primordial germ cells^52,57). DNA microarray analysis and hierarchal cluster analysis for ovarian dysgerminomas versus ovarian endodermal sinus tumors showed a differential expression of the genes in the Wnt/β-catenin signaling pathway. β-catenin expression was observed in all of endodermal sinus tumors and immature teratomas whereas there was no expression in other types of ICGCT. Especially, nuclear expression of β-catenin were observed in 50-70% endodermal sinus tumor and immature teratoma compared to those in dysgerminoma, embryonal carcinoma, and choriocarcinoma^16), suggesting activation of Wnt/β-catenin signaling pathway in these tumors. In one DNA methylation study, yolk sac tumors had a greater number of methylated tumor suppressor genes than seminomas and teratomas¹⁸⁾. Oct-4, a transcription factor encoded by the POU5F1 gene is involved in the initiation, maintenance, and differentiation of pluripotent and germline cells during normal development. It is expressed in mouse and human embryonic stem and germ cells but is absent in all differentiated somatic cell types. Oct-4 is considered to be a highly sensitive and specific and immunohistochemical marker for primary ICGCT^12,22).

Recently, genome-wide analyses of DNA copy number alterations were published^17,5962). Analysis of 62 ICGCT samples from three countries revealed that the overall pattern of genomic aberration was similar in germinoma and mixed GCT⁵⁹⁾. There were frequent aberrations of CCND2 (12p13) and RB1 (13q14), suggestive of Cyclin/CDK-RB-E2F pathway abnormalities in the pathogenesis of ICGCT. Additionally, the frequent gain of PRDM14 (8q13) suggested an abnormality in the transcriptional regulation of primordial germ cells. Somatic mutations in KIT and RAS have been reported in ICGCT and extracranial GCT. A recent targeted sequencing study confirmed that activating mutations in KIT and RAS were frequent and mutually exclusive in pure germinoma, suggesting that changes in the KIT signaling pathway play an important role in the development of germinomas¹⁷⁾. A whole-exome sequencing study also detected mutations in components of the KIT/RAS signaling pathway in >50% of ICGCTs that included recurrent somatic mutations in KIT and KRAS as well as in NRAS and its negative regulator, CBL⁶³⁾. In addition, somatic alterations in the AKT/mTOR pathway were discovered, and these alterations increase the copy numbers of the AKT1 locus at 14q32.33. Global microRNA profiling studies of both ICGCT and other GCT have shown distinct expression profiles in germinoma and NGGCT^19,3945). These studies also revealed a high expression of miR-371-373 and miR-302 clusters in all ICGCT histological subtypes. The microRNA from these two clusters was present at high levels in the patients' sera at the time of extracranial malignant GCT diagnosis. A decrease in the level of miR-372 following treatment suggested a diagnostic value of these markers in patients with marker-negative disease^19,45).

The biological factors responsible for the different ethnic incidence of ICGCT are poorly understood. Nevertheless, the whole-exome sequencing study of Japanese ICGCT patients detected significant enrichment of novel and rare germline variants in JMJD1C, which encodes a histone demethylase, a coactivator of the androgen receptor⁶³⁾. This finding could indicate a strong association between JMJD1C variants and the risk of developing ICGCT, but further international genome-wide association studies are needed to confirm the genetic and ethnic risk factors for the development of ICGCT.

Based on retrospective studies, craniospinal low-dose radiotherapy plus boost results in a 96-100% relapse-free survival rate in patients with ICGCT. The cure rate of germinoma is more strongly influenced by a radiotherapy volume that generously covers the cerebrospinal space than the radiation dose itself. The accumulating long-term treatment results show excellent control rates with reduced radiation doses with or without chemotherapy^{11,1340,5556)}. Neoadjuvant chemotherapy has been given to reduce the volume and dose of radiation reaching the gross tumor; however, many retrospective and prospective studies have shown that additional chemotherapy does not prevent periventricular tumor recurrence when only a small radiation field is used^2,49,1555). Currently, a radiation volume that covers at least the whole ventricular space is the standard for radiation therapy of ICGCT. Whole-ventricular radiation plus a boost, with or without chemotherapy, achieved reasonably good relapse-free survival rates of 88-96%. The relapse-free survival rate of patients with ICGCT reaches 90-100% with low-dose radiotherapy alone. Recent prospective studies showed that radiotherapy alone or chemotherapy followed by radiotherapy had similar treatment outcomes^9,56). The superiority of chemotherapy combined with radiotherapy would be difficult to prove unless the combined therapy substantially reduced the side-effects of radiotherapy. Accordingly, the role of neoadjuvant chemotherapy should be examined in a well-designed clinical trial in which the end-points include late sequelae, quality of life, and survival outcomes⁵³⁾. Preexisting endocrinopathies, growth problems, and the patient's neurocognitive status due to both the tumor itself and surgical complications observed in the majority of ICGCT patients should be considered as baseline data in the assessment of any treatment-related morbidity.

Historically, a radiation dose up to 45-60 Gy has been used for various radiotherapy volumes in patients with ICGCT. However, published results from various institutes worldwide show that much lower radiation doses in the range of 30-39 Gy for gross tumors and 19.5-25 Gy for microscopic tumors are sufficient for good disease-control rates^{11,1321,4168)}. It has been suggested that the radiation dose needed to control germinomas ranges from 36-45 Gy, depending on the volume of the tumor⁵⁴⁾. Although ICGCT mostly affects patients in late childhood and adolescence and the effect of radiation therapy on neurocognitive function is relatively mild^31,37), recent Japanese data have shown a decrease in the social function of survivors of ICGCT²⁴⁾. The optimal radiation dose and volume should be selected to minimize late sequelae while maintaining long-term survival. The median time to recurrence ranges from 30 to 50 months^25,46). Late recurrence after 8-15 years of age is not uncommon, suggesting the need for an observation period of ≥10 years^2,1323,2546).

NGGCT is diagnosed by the presence of any of the following germ cell types, such as embryonal carcinoma, endodermal sinus tumor, choriocarcinoma, or teratoma. Despite intensive chemotherapy and high-dose radiotherapy, 10-year overall survival rates range from 30% to 70%, with worst survival rate for patients in the poor prognosis group defined by Japanese prognostic classification. Currently, there is a debate on the contribution of surgery to a superior treatment outcome when performed before^35,51) or after radiotherapy combined with chemotherapy⁵⁰⁾. Radiotherapy combined with cisplatin-based chemotherapy provides the best treatment outcome, increasing the survival rate of patients with NGGCT to 60-70%. In a retrospective analysis of 32 patients with NGGCT, the relapse-free and overall survival rates were 77.6% and 74.6%, respectively. In this study, the most important prognostic factors were the implementation of craniospinal radiotherapy and the Japanese risk-group classification²⁷⁾. However, it has not yet been determined whether the whole craniospinal axis must be included in the radiation field for all patients with NGGCT. The International Society of Pediatric Oncology GCT phase II study defined the high-risk group as patients with a serum AFP level >1000 ng/mL and age <6 years. This study utilized focal radiotherapy with radiation doses of 54 Gy for the standard-risk group. By contrast, craniospinal radiotherapy with a radiation dose of 30 Gy plus an additional 24 Gy to the gross tumor was used to treat the high-risk group. The Japanese CNS GCT study group set an AFP level >2000 ng/mL and/or a serum β-HCG level >2000 mIU/mL as cut-off values for their phase II single-arm prospective study of concomitant radiochemotherapy for NGGCT³³⁾. Craniospinal radiotherapy was applied at a dose of 30.6 Gy for patients with poor prognosis and 23.4-27 Gy of whole-ventricular radiation for patients with intermediate prognosis, yielding a final dose to the focal tumors of 50.4 Gy and 61.2 Gy, respectively. The North American protocol COG ACNS1123 is underway in which the dose and volume of radiotherapy is adapted according to the tumor response to six cycles of chemotherapy, regardless of prognosis or tumor markers. This study is also testing whether the volume of irradiation can be safely restricted to the whole ventricle. The study is using a radiation dose of 30.6 Gy for whole-ventricular irradiation with a 23.4 Gy primary site boost for patients with localized NGGCT. This treatment regimen will only be used in patients confirmed by magnetic resonance imaging (MRI) to have a complete response to induction chemotherapy. In addition, serum and CSF tumor markers must be in the normal range for this therapy. The same treatment scheme can also be applied to patients who show a partial response to chemotherapy and to those who have undergone second-look surgery revealing mature teratomas.

Proton beam therapy (PBT) has dosimetric benefit in reducing radiation dose to the normal tissues while concentrating radiation dose to the tumor because of its inherent physical characteristics of absence of exit radiation dose, i.e., Bragg peak. This characteristic have been considered to be particularly useful when treating children with brain tumors because it reduces radiation dose to the various substructures of brain which are critical in intellectual functioning and hence preserving neural functions in brain tumor survivors. Both germinomas and mixed GCTs require a generous radiotherapy volume that covers the whole ventricle, if not the whole craniospinal space. Comparison of the dose-volume histograms of intensity-modulated radiotherapy (IMRT) and three dimensional-conformal radiotherapy (3D-CRT) for whole-ventricular irradiation showed that IMRT reduced the irradiated volume receiving radiation doses of 20 Gy, 30 Gy, and 40 Gy by 7.5%, 12.2%, and 9.0%, respectively, compared with 3D-CRT. IMRT was associated with a statistically significant reduction of the median irradiated volumes at all dose levels (p<0.002) compared with 3D-CRT; however, with IMRT the radiation dose was 1.9 times higher than 3D-CRT and had greater effects on peripheral areas of the body compared with 3D-CRT¹⁰⁾. PBT significantly reduces the incidence of early and late toxicities of radiotherapy, and is expected to help improve the therapeutic ratio because of its physical characteristics involves absence of radiation dose beyond target tissue. The use of PBT for whole craniospinal irradiation also significantly reduces hematological toxicity^7,858) as well as decreases the risk of secondary malignancy (Fig. 2)^38,69). In addition, whole-ventricular PBT brings the volume of the cerebral cortex receiving ≤10 Gy and ≤15 Gy down to two-third to a half of those of 3D-CRT and IMRT (Fig. 3). In tumors arising in the pineal gland, suprasellar area, or basal ganglia, PBT also delivers lower mean radiation doses to the cochlea, pituitary gland, and temporal lobes (Fig. 4). When combined with a focal boost radiotherapy field, IMRT and PBT significantly differed in the volume of brain parenchyma that received doses of 5-20 Gy, and the volume of temporal lobes that received doses of 5-30 Gy when a total of 30.6 Gy was delivered to the primary tumor. The benefit of PBT increased when the pencil beam scanning (or spot scanning) technology of PBT was used for the same target volume, because numerous small proton beams from thousands directions paint the target area in this technique, resulting in greater conformity to the irregularly shaped tumors. When radiation doses of ≥25 Gy were delivered to the whole brain, dose-cognitive effect modeling revealed that the radiation dose to the supratentorial and infratentorial brain parenchyma, temporal lobes, and the hippocampus was linearly associated with a decline in late neurocognitive function^5,36). It is expected that this dosimetric benefit of PBT can be translated into clinical benefit.

ICGCT affects includes heterogeneous disease entities which have variable prognosis. Radiotherapy and chemotherapy can be used in various combinations, resulting in 80-100% and 60-70% survival rates, respectively for germinoma and NGGCT. However, most of the patients who survive the disease are at risk for cognitive impairments and poor quality of life secondary to disease itself and treatment related factors. To achieve high long-term disease control rate and good quality of life of the survivors, carefully selected multimodality treatment needs to be combined with following systemic assessment of late tumor- and treatment-related functional outcome.