12p gain is predominantly observed in non-germinomatous germ cell tumors and identifies an unfavorable subgroup of central nervous system germ cell tumors

Abstract

Background

Central nervous system (CNS) germ cell tumors (GCTs) are neoplasms predominantly arising in pediatric and young adult populations. While germinomas generally respond to chemotherapy and radiation, non-germinomatous GCTs (NGGCTs) require more intensive treatment. This study aimed to determine whether 12p gain could predict the prognosis of CNS GCTs.

Methods

Eighty-two CNS GCTs were included in this study. The 12p gain was defined by an additional 12p in the background of potential polyploidy or polysomy. Cases were analyzed using an Illumina methylation 450K array for copy number investigations and validated by fluorescence in situ hybridization (FISH).

Results

A 12p gain was found in 25-out-of-82 cases (30%) and was more frequent in NGGCTs (12% of germinoma cases and 50% of NGGCT cases), particularly in cases with malignant components, such as immature teratoma, yolk sac tumor, choriocarcinoma, and embryonal carcinoma. 12p gain and KIT mutation were mutually exclusive events. The presence of 12p gain correlated with shorter progression-free (PFS) and overall survival (OS) (10-year OS: 59% vs. 94%, with and without 12p gain, respectively, P = 0.0002), even with histology and tumor markers incorporated in the multivariate analysis. Among NGGCTs, 12p gain still had prognostic significance for PFS and OS (10-year OS: 47% vs. 90%, respectively, P = 0.02). The 12p copy number status was shared among histological components in mixed GCTs.

Conclusions

12p gain may predict the presence of malignant components of NGGCTs, and poor prognosis of the patients. It may be associated with early tumorigenesis of CNS GCT.

Key Points.

• 12p gain is commonly found in NGGCTs, particularly in those with malignant components.

• 12p gain is a prognostic factor in NGGCT.

• 12p gain and polyploidy may be an early event in tumorigenesis of CNS GCT.

Importance of the Study.

This is the first to study the significance of 12p gain in central nervous system (CNS) germ cell tumors (GCTs) for its pathological, clinical, and pathogenesis perspectives. First, 12p gain is predominantly found in non-germinomatous GCTs (NGGCTs), and especially indicates the existence of malignant histological components. Second, 12p gain predicts unfavorable prognosis in CNS GCTs as a whole, as well as identifies NGGCTs with worse prognosis. Third, 12p copy number status is shared among different histological components in mixed GCTs, suggesting 12p gain as an early event in tumorigenesis. Our study thus demonstrated that 12p gain may serve as a strong prognostic biomarker in identifying CNS GCT patients with a poor prognosis who require more intensive treatment and provides insights in the determination of its pathogenesis.

Central nervous system (CNS) germ cell tumors (GCTs) are morphologically and immunophenotypically similar to gonadal and other extraneuraxial germ cell neoplasms.^1,2 CNS GCTs mainly affect the pediatric population and young adults. Incidence peaks in patients aged 10–14 years and are predominantly in males.³ There is a characteristic geographical distribution in CNS GCTs, which are more common in Eastern Asia than in Europe or North America. The incidence is 0.45 per 100,000 population aging <15 years in Japan and 0.49 per 100,000 population <19 years in Korea, accounting for 10.6%–15.6% of primary brain tumors under the age of 19,^4–7 compared with 4.1% in Europe and 3.8% in North America.^8,9 While germinomas (GEs) generally respond to chemotherapy and radiation, non-germinomatous GCTs (NGGCTs) are more likely to recur and require more intensive treatment.

Gonadal GCTs are considered to originate from primordial germ cells (PGCs), which arise from the yolk sac epithelium and migrate to testes or ovaries.¹⁰ Corroborated by their histological similarity,¹¹ mis-migrated PGCs have been considered the mutual cell of origin for CNS GCTs with testicular GCTs (TGCTs) by their shared genetic/epigenetic profiles, such as alterations of the MAPK and PI3K pathway¹² and global DNA methylation profile.^13–16 TGCTs are classified into two fundamentally different categories: germ cell neoplasia in situ (GCNIS)-associated (Type II) and non-GCNIS-associated (Types I and III) TGCTs.¹⁷ 12p gain, often in the form of isochromosome 12p [i(12p)], is the most common (77%–88%) genetic hallmark of type II TGCTs,^17–20 although rarely found in its precursor lesion, GCNIS. Therefore, 12p gain is not only supportive in the diagnosis of primary and metastatic TGCTs in clinical practice,¹⁷ but also is considered to be associated with the acquisition of invasiveness in TGCT.²¹ Recently, Shen et al. revealed that 12p gain was one of the early events in TGCT tumorigenesis occurring after whole-genome duplication.²⁰

By contrast, several studies demonstrated that 12p gain was less common in CNS GCT (20%–58%)^22–27 than in TGCT (77%–88%).^17,19,20 There are few studies regarding the clinicopathological significance of 12p copy number status in CNS GCT, such as its distribution in each histological subtype and its prognostic impact. This study performed genome-wide copy number analysis to identify the frequency of 12p gain in a large CNS GCT cohort, including all histological subtypes. This study aimed to determine whether 12p gain could serve as a prognostic or diagnostic marker in CNS GCTs.

Materials and Methods

Tissue Materials

The study was approved by the Institutional Review Board of the National Cancer Center, Tokyo, Japan (No. 2012-043) and the respective local institutional review boards. A total of 82 histologically verified primary CNS GCTs collected from the iGCT Genome Analysis Consortium were used in this study. In all cases, the histopathological diagnoses were reviewed by an expert neuropathologist (YN) and confirmed by another neuropathologist (KS) based on the microscopic review of the hematoxylin and eosin (H&E) staining slides, in addition to immunohistochemical evaluations as necessary (Supplementary Table 1). The cohort included 42 GEs, 22 mixed GCTs (MGCTs), 7 mature teratomas (MT), 4 immature teratomas (IT), 1 embryonal carcinoma (EC), 5 yolk sac tumors (YST), and 1 choriocarcinoma (CC). The median patient age was 15 years, ranging from 2 months to 45 years. There were 71 males (87.7 %) and 10 females (12.3 %) (one unknown; predicted as male by DNA methylation array-based analysis). Of these 82 CNS GCTs, the germinoma and NGGCT components of three MGCTs (GCT51, GCT58, and GCT86) were microdissected using corresponding formalin-fixed, paraffin-embedded (FFPE) sections. Four normal tissues (cerebrum, pineal gland, testis, and ovary) obtained from two autopsied bodies, and a neural stem cell line (NSC) supplied by Cambridge University (UK) were also included as control samples.

Tissue Microdissection

As shown in a previous study,¹³ the GE and NGGCT components in MGCTs were individually collected on FFPE sections mapped under light microscopy by a pathologist (SF). In this study, three cases (GCT51, GCT58, and GCT86) were included because they displayed a well-demarcated pattern between these two components on the H&E stained slides.

DNA Extraction

For all 82 GCTs and 5 normal samples, the genomic DNA was extracted from frozen tissues using a DNeasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA) according to the manufacturer's protocol. In 3 CNS GCTs in which microdissection was performed, the genomic DNA was extracted from the collected tissues using a DNeasy Blood & Tissue Kit (Qiagen). DNA concentration was measured using the Qubit dsDNA Assay Kit on a Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA). The quality of the FFPE DNA was assessed using the Illumina FFPE QC Kit (Illumina, San Diego, CA, USA) with the THUNDERBIRD SYBR qPCR Mix (Toyobo Life Science, Tokyo, Japan) according to the manufacturer's protocol.

Molecular Analysis

Somatic mutation data were obtained by whole-exome or targeted sequencing in 81-out-of-82 cases.¹² For genome-wide methylation analysis, the raw IDAT files of 93 samples were reviewed: 82 frozen GCT tissues, 5 normal frozen tissues, and 6 microdissected FFPE tissues from 3 MGCTs, which mostly overlapped with the cases from the previous study.¹³ Bisulfite modification was performed using 500 ng genomic DNA from frozen tissues or 100 ng genomic DNA from FFPE sections using an EZ DNA Methylation Kit (Zymo Research Corporation, Irvine, CA) according to the manufacturer's protocol. The bisulfite-modified FFPE DNA was retrieved using an Infinium HD Assay Kit FFPE Restore Kit (Illumina). These were subjected to a comprehensive methylation analysis using the Infinium Human-Methylation450 BeadChips (450K, Illumina), including 485,512 probes according to the manufacturer’s instructions.

Copy Number Analysis by 450K

Raw IDAT files from 450K were processed using the minfi package (version 1.34.0) in R statistical environment (version 4.0.2), and quality control was performed. Mset objects generated from the raw IDAT files were used as the input data for copy number variation analysis using the conumee package (version 1.22.0). Using the genome annotations, 470,870 probes were used for further analysis. Unprocessed IDAT files of 119 normal control samples were downloaded from the NCBI Gene Expression Omnibus (GEO) under the accession number GSE109381.²⁸ Copy number loci proceeded by conumee package were taken as the average of each chromosomal short arm (p) and long arm (q) using R. A widely used heuristic to identify gain or loss of each chromosome arm is determined to use a symmetrical absolute cutoff of ±0.1 for conumee processed data.²⁹ To examine the i(12p) in the background of potential polyploidy or polysomy, 12p gain was further determined when the average copy number of 12p exceeded 0.1 than that of 12q. Using the same principle, the gain or loss of each chromosome arm was determined as follows: the relative gain of the short arm of chromosome n (np) is determined when np > 0.1 and np − nq > 0.1; the relative loss of np is determined when np < −0.1 and np − nq < −0.1; and vice versa for long arms. Automatic scoring was verified by manual assessment of the respective loci for each profile on a conumee plot.^30–32 Probes assigned to chromosomes X and Y were excluded from the analysis. Probes on chromosome arms 13p, 14p, 15p, and 22p were inapplicable for interpretation of copy numbers due to the probe design of 450K.

IDAT files of eight CNS GCT cases that have been published in a report from the Children's Oncology Group³³ were also obtained from the GEO (GSE183798). Then, the same copy number analysis was performed in those eight cases, and they were assessed as a validation cohort.

Fluorescence In Situ Hybridization (FISH)

Dual-color 12p/chromosome enumeration probes (CEP) 12 and 19p/19q FISH studies were performed using 4-μm-thick tissue specimens. The probes used were Vysis LSI 12p Spectrum Green (05J03-012, Abbott Molecular, Abbott Park, IL, USA), CEP12 Spectrum Orange probes (07J22-012, Abbott Molecular), and 19p13/19q13 probe (KBI-00131, Leica Biosystems, Tokyo, Japan). Images were captured using the Metafer Slide Scanning Platform (MetaSystems, Altlussheim, Germany), and a minimum of 100 non-overlapping tumor cells were examined. For 12p gain, the signal ratio of the number of green signal (12p) to that of orange signal (CEP12) was calculated for each tumor cell nucleus. The average signal ratio of 1.5 or greater was considered evidence of 12p gain. In addition to the average signal ratio of the 12p/CEP12 probe set, 19p/19q was also calculated. For each of these probe sets, samples were considered to be polysomy for chromosome 12 or 19 when greater than 10% of cells demonstrated greater than 2 orange and 2 green signals per cell in the tumor cells analyzed.³⁴

Clinical Information

Clinical information, including age, sex, tumor location, treatment (surgery, radiation, and chemotherapy), tumor markers, and follow-up outcomes, were obtained. Tumor markers were measured in either or both the serum and cerebrospinal fluid (CSF) pre- or intra-operatively. CSF was obtained via preoperative lumbar puncture or intraoperative ventricular drainage. The cutoff values of 5 IU/L and 10 ng/mL were used for normal beta-human chorionic gonadotropin (hCG) and alpha-fetoprotein (AFP) levels, respectively, by referring to representative standards and the SIOP guidelines.³⁵

As a principle for treating CNS GCTs among the institutions belonging to the iGCT genome analysis consortium, tumors are treated based on the three prognosis groups as described previously.³ In brief, except for MT, platinum-based chemotherapy followed by the combination of radiotherapy; whole ventricle irradiation with or without additional tumor irradiation of the tumor, or craniospinal irradiation with boosted tumor irradiation were administered

Statistical Analysis

For statistical analysis, variables were compared using the Mann–Whitney U test or Chi-square test. Multivariate analyses were performed using the logistic regression method. Overall survival (OS) and progression-free survival (PFS) after surgery were analyzed using the Kaplan–Meier method and compared using a log-rank test. Statistical significance was set at P < 0.05. All statistical analyses were performed using the JMP version 15 (SAS Institute, Cary, NC).

Results

The Profile of Copy Number Alteration and 12p Gain in CNS GCTs

The copy number profile of each chromosome arm is shown in Figure 1A. A 12p gain, indicating i(12p), was found in 30% (25/82) of CNS GCTs, and was the most common copy number alteration. None of the samples exhibited a loss of 12q. The genome-wide copy number plot of a representative case (GCT23) solely accompanying the 12p gain is shown in Figure 1B. The patient had neither gain nor loss of 12q or other chromosome arms. Other chromosome arm gains were commonly found in 21q (27%) and 1q (21%). Chromosome arm losses were frequently observed in 17p (12%) and 4q (11%) (Figure 2A, Supplementary Table 2).

Fig. 1.

Profile of copy number alterations in central nervous system germ cell tumors (CNS GCTs). 12p gain is the most frequently altered chromosome arm, particularly in non-germinomatous germ cell tumors (NGGCTs) (A). In the conumee plot of a representative case with 12p gain, the copy number level of 12p is higher than what would be expected for a single copy gain, suggesting the presence of polyploidy. There was no sign of 12q loss (B).

Fig. 2.

The frequency of copy number alteration in each chromosomal arm in CNS GCT. 12p gain is the most frequently altered chromosome arms (A). 12p gain is significantly overrepresented in non-germinomatous germ cell tumors (NGGCT) (P = 0.0002). Among the NGGCTs, tumors with malignant histological components (immature teratoma, yolk sac tumor, choriocarcinoma, or embryonal carcinoma) harbored significantly more frequent 12p gain (B).

Correlation between 12p Gain and Clinicopathological Profiles

Among the 82 patients, age at diagnosis, sex, and tumor location were available for 77, 81, and 80 patients, respectively. None of these were related to the 12p status (Table 1). In contrast, the pathological diagnosis was significantly correlated with 12p gain (P < 0.001, Table 1). It was present in 12% (5/42) of the GEs and 50% of the NGGCTs (20/40) (P < 0.001, Table 1, sensitivity [Se] = 50%, specificity [Sp] = 88%, positive predictive value [PPV] = 80%, and negative predictive value [NPV] = 65%, Supplementary Table 3). Of the 40 NGGCTs, tumors with malignant histological components (IT, YST, CC, and EC) harbored a 12p gain (18/28, 64%) more frequently than those without (2/12, 17%) (P = 0.014, Table 1, Se = 64%, Sp = 83%, PPV = 90%, and NPV = 50%, Supplementary Table 3). Note that the 12p gain showed a high NPV in identifying malignant components among all CNS GCTs (Se = 64%, Sp = 87%, PPV = 72%, and NPV = 82%, Supplementary Table 3). The distribution of 12p gain in each histological subtype is graphically summarized in Figure 2B.

Table 1.

Clinicopathological Profile and 12p Gain in Central Nervous System Germ Cell Tumors

	N (%)
	All	12p Gain	12p Intact	P-value
Number	82	25 (30.5)	57 (69.5)
Age at diagnosis				0.634
Median (y)	15	15	15.5
Range (y)	0.17–45	2–38	0.17–45
<6	5	2 (40.0)	3 (60.0)
≥6	72	21 (29.2)	51 (70.8)
ND	5	2 (40.0)	3 (60.0)
Sex				1.000
Male	71	22 (31.0)	49 (69.0)
Female	10	3 (30.0)	7 (70.0)
ND	1	0 (0.0)	1 (100.0)
Tumor location				1.000
Typical†	54	17 (31.5)	37 (68.5)
Atypical‡	26	8 (30.8)	18 (69.2)
ND	2	0 (0.0)	2 (100.0)
Pathological diagnosis				<0.001*
GE	42	5 (11.9)	37 (88.1)
MT*	7	2 (28.6)	5 (71.4)
IT	4	1 (25.0)	3 (75.0)
EC	1	1 (100.0)	0 (0.0)
YST	5	3 (60.0)	2 (40.0)
CC	1	0 (0.0)	1 (100.0)
MGCT	22	13 (59.1)	9 (40.9)
Histological class				<0.001*
GE	42	5 (11.9)	37 (88.1)
NGGCT	40	20 (50.0)	20 (50.0)
NGGCT w/ mal.	28	18 (64.3)	10 (35.7)	<0.001*
NGGCT w/o mal.	12	2 (16.7)	10 (83.3)
Serum AFP (ng/mL)				<0.001*
>10 ng/mL	25	16 (64.0)	9 (36.0)
≤10 ng/mL	45	5 (11.1)	40 (88.9)
ND	12	4 (33.3)	8 (66.7)
CSF AFP (ng/mL)				0.006
>10 ng/mL	11	8 (72.7)	3 (27.3)
≤10 ng/mL	14	3 (21.4)	11 (78.6)
ND	57	14 (24.6)	43 (75.4)
Serum beta-hCG (IU/L)				0.217
>5 IU/L	14	7 (50.0)	7 (50.0)
≤5 IU/L	25	6 (24.0)	19 (76.0)
ND	43	12 (27.9)	31 (72.1)
CSF beta-hCG (IU/L)				0.252
>5 IU/L	9	5 (55.6)	4 (44.4)
≤5 IU/L	8	2 (25.0)	6 (75.0)
ND	65	18 (27.7)	47 (72.3)
MAPK pathway				0.636
Mutant	41	11 (26.8)	30 (73.2)
Wild type	40	14 (35.0)	26 (65.0)
ND	1	0 (0.0)	1 (100.0)
KIT				0.00476*
Mutant	24	2 (8.3)	22 (91.7)
Wild type	57	23 (40.4)	34 (59.6)
ND	1	0 (0.0)	1 (100.0)
PI3K pathway				0.803
Mutant	10	2 (20.0)	8 (80.0)
Wild type	71	23 (32.4)	48 (67.6)
ND	1	0 (0.0)	1 (100.0)
mTOR				0.602
Mutant	7	1 (14.3)	6 (85.7)
Wild type	74	24 (32.4)	50 (67.6)
ND	1	0 (0.0)	1 (100.0)

Abbreviations: ND, no data; GE, germinoma; MT, mature teratoma; IT, immature teratoma; EC, embryonal carcinoma; YST, yolk sac tumor; CC, choriocarcinoma; MGCT, mixed germ cell tumor; NGGCT, non-germinomatous germ cell tumor; mal., malignant component (IT, EC, YST, and CC AFP, alpha-fetoprotein; CSF, cerebrospinal fluid; HCG, human chorionic gonadotropin; ND, no data.

^†Typical; neurohypophysis and pineal region.

^‡Atypical; others.

*Mature teratoma group contains one case of mature teratoma with somatic malignancy, accompanying relative 12p gain.

Elevated AFP in either or both serum and CSF significantly correlated with 12p gain (P < 0.0001 and P = 0.006). Regarding somatic mutations, neither mutation in MAPK nor PI3K pathways was associated with 12p gain; however, KIT mutation, the most frequent mutation in the MAPK pathway,^12,23 was inversely correlated with 12p gain (P = 0.00476, Table 1). In GEs, all KIT mutations (n = 20) were found in cases without 12p gain, and 12p gain was mutually exclusive with the KIT mutation (Figure 1A).

Of the five GE patients with 12p gain, two (40%) showed a total hCG (6.5 IU/l in CSF) or an AFP level (33.1 ng/ml in serum) exceeding the cutoff value (Supplementary Table 1). Both two NGGCT cases with 12p gain unaccompanied by a malignant component showed elevated AFP (22,398 or 2996 ng/ml in serum and 5732 ng/ml in CSF) (Supplementary Table 1).

Copy Number Analysis in Microdissected Components of Mixed Germ Cell Tumors

First, FISH was performed to investigate the distribution of 12p gain among the histological components in MGCT. Three MGCT cases (GCT51, 58, and 86) were selected. Analysis at 450K revealed that 12p gain was shared in GE and NGGCT components in GCT51 (Figure 3). In addition, per FISH, 12p gain was observed in each histological component of GE, IT, and YST, as evidenced by the ratio of 12p/CEP12 of 1.79, 1.52, and 1.55 in GE, IT, and YST, respectively, which exceeded the cutoff value of 1.5, indicating a 12p gain. In the other two cases, intact 12p was shared by both the GE and NGGCT components based on the analysis at 450K (GCT58 and GCT86; Supplementary Figure 1). In line with this, 12p gain was absent in GE and NGGCT components by FISH in the corresponding FFPE sections of these two cases. The 12p/CEP12 ratio was 1.07 in GE and 1.04 in MT in GCT58, and 1.25 in GE, 1.14 in MT, 1.24 in YST, and 1.33 in EC in GCT86 (Supplementary Figure 1). These results demonstrated that FISH analysis was concordant with the results at 450K in detecting copy number alterations and that 12p status was shared in each histological component in the same tissue with mixed histological components.

Fig. 3.

Copy number analysis in a microdissected mixed germ cell tumor, GCT51. The case consists of approximately 50% of germinoma, 30% of YST, and 20% of immature teratoma on formalin-fixed, paraffin-embedded section. The histology is shown in the top row. 12p gain was shared in germinoma and NGGCT components by 450K (middle row, arrowhead). Fluorescence in situ hybridization (FISH) analysis also showed 12p gain in germinoma and NGGCT (yolk sac tumor; YST) components (bottom row). In FISH analysis, green and orange signals correspond to 12p and centromere of chromosome 12 (CEP12) signals, respectively. Scale bar = 5 mm, 100 μm, or 25 μm, as indicated in the figures.

To assess the ploidy of the tumor cells, polysomy of chromosomes 12 and 19 was evaluated. Chromosome 19 was selected because it is one of the most stable chromosomes in terms of copy number and the ratio of the short and long arms in CNS GCTs (Figures 1A and 2A, Supplementary Table 2). By FISH, all the above three microdissected cases were determined to have polysomy 12 in all the histological components, regardless of the presence or absence of 12p gain (Figure 3, Supplementary Figure 1). The fraction of tumor cells with polysomy 12 ranged from 32% to 81% in GCT51, 43% to 47% in GCT58, and 47% to 57% in GCT86 (Table 2). In contrast, all these three cases were judged to have balanced 19p/19q in GE and NGGCT components according to 450K analyses, and the 19p/19q ratio ranged from 0.85 to 0.97 in FISH analyses (Table 2). However, FISH detected polysomy 19 in all three cases, ranging from 33% to 44% in GCT51, 44% to 57% in GCT58, and 61% to 75% in GCT86 (Table 2). We inferred that these three MGCTs shared whole-genome duplication, rather than partial or focal chromosomal gain, across different histological components. Typical isochromosome 12p signal pattern, that is, gain of 12p accompanied by loss of CEP12 in the background of polysomy 12, was not detected in any components of samples due to the probe design.

Table 2.

Copy Number Analysis based on Genome-wide DNA Methylation Array (450K) and FISH in Three Mixed Germ Cell Tumors

	Chromosome 12			Chromosome 19
	450K	12p/CEP12 FISH	Polysomy (%)	450K	19p/19q FISH	Polysomy (%)
GCT51-GE	12p gain	1.79	80.8	19p/q intact	0.86	43.8
GCT51-IT	12p gain	1.52	49.1	19p/q intact	0.87	39.6
GCT51-YST		1.55	31.6		0.88	33.3
GCT58-GE	12p intact	1.07	47.1	19p/q intact	0.89	44.3
GCT58-MT	12p intact	1.04	42.9	19p/q intact	0.85	57.4
GCT86-GE	12p intact	1.25	47.3	19p/q intact	0.97	61.5
GCT86-MT	12p intact	1.14	57.4	19p/q intact	0.97	61.2
GCT86-YST		1.24	50.0		0.93	64.5
GCT86-EC		1.33	46.5		0.85	74.5

Abbreviations: GE, germinoma; MT, mature teratoma; IT, immature teratoma; EC, embryonal carcinoma; YST, yolk sac tumor; CEP, chromosome enumeration probes.

Survival Analysis

Survival data were available for 80 of 82 cases, including 40 patients with GE and 40 patients with NGGCT. During the median follow-up of 79 months (range, 0.46–297 months), 18 patients relapsed, and 11 died.

Among 80 CNS GCT cases, 12p gain was found in 25 cases, which showed a significantly worse prognosis compared to the remaining 55 cases without 12p gain for PFS (1-, 5-, and 10-year PFS: 78% vs. 96%, 78% vs. 88%, and 52% vs. 83%, respectively, P = 0.011, Figure 4A) and OS (1-, 5-, and 10-year OS: 91% vs. 100%, 77% vs. 94%, and 59% vs. 94%, respectively, P < 0.001, Figure 4B). In the germinoma group, 5 out of 40 (13%) patients had a 12p gain; however, there was no significant difference in either PFS or OS per 12p status (Figure 4C and D). Among the NGGCT cases, 20 out of 40 (50%) cases had a 12p gain, and these cases showed significantly shorter PFS (1-, 5-, and 10-year PFS: 71% vs. 90%, 71% vs. 90%, and 47% vs. 83%, respectively, P = 0.034, Figure 4E) and OS (1-, 5-, and 10-year OS: 88% vs. 100%, 71% vs. 90%, and 47% vs. 90%, respectively, P = 0.017, Figure 4F).

Fig. 4.

Survival analysis in 80 central nervous system germ cell tumors (CNS GCTs). Among all CNS GCTs, 12p-gained cases showed significantly shorter progression-free survival (PFS) (A) and overall survival (OS) (B) as compared to 12p-intact cases. There was no statistically significant difference in survival in germinomas with 12p-gained cases and without (C, D), although 12p gain cases showed a trend of shorter OS (D). Among non-germinomatous germ cell tumors, 12p-gained cases showed significantly shorter PFS (E) and OS (F) as compared to 12p-intact cases.

Finally, the impact of histology (GE vs. NGGCTs), positive tumor markers, and the 12p status on survival were examined by multivariate analysis. Tumor marker positivity was determined based on whether either or both tumor markers were positive using the above-mentioned cutoff values. The results showed that 12p gain was the sole factor significantly associated with survival (P = 0.033). Neither histology (P = 0.30) nor tumor markers (P = 0.62) was associated with survival.

Copy Number Analysis in a Validation Cohort

Of the eight CNS GCT cases from the COG study, a 12p gain was observed in 33% of GE cases (1/3) and in 60% of NGGCT cases (3/5). Three of the cases in the NGGCT group with 12p gain were MGCTs, while the rest were teratomas without any distinction of further histological classification. Survival analysis was not performed in the validation cohort because the patients' detailed clinical information was not available.

Discussion

This study investigated the landscape of copy number alterations in CNS GCTs and highlighted three potential clinical values of 12p gain in the setting of pathological diagnosis and management of the best therapeutic strategies. First, 12p gain was predominantly observed in NGGCTs, particularly in tumors accompanying malignant components (IT, YST, CC, EC), compared to germinomas, as in TGCTs. Few studies have contrasted the frequency between GE and NGGCT, possibly due to the less frequent 12p gain in CNS GCT (20%–58%)^22–27 than the testicular counterpart (77%–88%).^17,19,20 The overall frequency of 12p gain (30%) was concordant with that in the previous literature on CNS GCT identified in this study.

Second, 12p gain was correlated with shorter survival time in both PFS and OS among all CNS GCT cases. As patients with 12p gain showed significantly shorter PFS/OS in NGGCTs, it can be regarded as a marker to predict poor prognosis. Although 12p gain is frequently observed in non-seminomatous germ cell tumors, its role in prognosis has not yet been defined in TGCTs³⁶ and in pediatric and adolescent non-seminomatous GCTs, including ovary and extragonadal lesions.³⁷ In contrast to a previous study suggesting that either i(12p), which may appear as 12p gain, or polysomy 12, when examined by FISH, were of limited use in predicting the clinical course of CNS GCTs,³⁴ the current study demonstrated the prognostic significance of 12p gain using genome-wide DNA methylation array-based analysis.

Multivariate analysis showed that 12p gain was the sole factor significantly associated with survival, except histology or tumor markers. However, 12p gain is predominantly found in NGGCT cases with malignant histological components, as noted previously. When the association between 12p gain and survival was investigated among those NGGCTs with malignant histological components alone, there was a trend toward shorter PFS/OS in cases with 12p gain, although the difference was not statistically significant, possibly due to the small sample size and lack of statistical power (data not shown). Further studies in larger cohorts are required to validate the significance of 12p gain in predicting the prognosis of CNS GCTs, particularly for those with malignant NGGCT components.

Of note, other common copy number aberrations at chromosomal levels were also investigated, including 1q gain (21%) and 21q gain (27%). Neither of these alterations was associated with the prognosis, except that germinoma cases with 21q gain had shorter PFS than those without 21q gain (P = 0.03) (Supplementary Figures 3 and 4).

Third, we showed the utility of FISH analysis to identify the 12p gain and the shared 12p status among histological components in MGCTs. The FISH results were corroborated by the copy number data obtained using the 450K array. Homogeneous 12p copy number status across germinomatous and non-germinomatous components in MGCT cases suggested that 12p gain appeared to be an early event in CNS GCT tumorigenesis, reminiscent of the previous finding that MAPK/PI3K pathway mutations were shared between different histological components.¹³ Thus, FISH data, despite of the limited number of tumor samples, may represent the copy number status of 12p in the whole tumor tissue. As approximately 25 percent of NGGCTs are mixed tumors containing more than one histologic component,^38,39 histological evaluation with a small specimen is often challenging to estimate the overall status of the tumor due to the potential sampling error. Predicting patient outcomes based on information from a small biopsy sample may thus involve a potential risk of underestimation. Nonetheless, although the finding is based on a small number of samples, our results showed that the 12p status in a portion of the tumor may represent the whole tumor. Thus, FISH serves as a useful ancillary technique to provide an accurate assessment of the 12p status, which may help to identify NGGCT to plan the best therapeutic strategies in clinical management. Validation using a larger number of CNS GCT cases is warranted to establish the diagnostic utility of 12p testing. Currently, we have confirmed the relatively frequent 12p gain in NGGCT than in GE in a small set of CNS GCT samples included in the COG study, which corroborated our results. However, no existing independent large-scale international cohort study had investigated CNS GCT tissue resources with readily available molecular data to validate our findings. Hence, we are organizing a new clinical trial to evaluate CNS GCTs, in which tissue collection and molecular profiling will be mandated. This new clinical trial cohort will hopefully confirm the clinical significance of 12p gain in CNS GCTs.

In our study, we used the Illumina Infinium HumanMethylation450 BeadChip (450K) array for copy number assessment. The Illumina Infinium HumanMethylationEPIC BeadChip (EPIC) array, the successor of the 450K array containing over 850,000 probes, covers approximately 90% of the CpG sites represented on the 450K. Both the EPIC and 450K arrays provide equivalent average copy number results for each chromosomal arm using the conumee package. On the contrary, assessment of the absolute copy number using the 450K and EPIC arrays is challenging because they only provide the relative copy numbers compared with normal control samples. In this study, 12p/CEP12 or 19p/19q FISH was carried out as a complementary technique to assess the absolute copy number, and FISH analysis showed polyploidy on chromosomes 12 and 19 in CNS GCTs. This suggests the presence of whole-genome amplification in CNS GCTs. A similar phenomenon was reported for the testicular counterpart.

Moreover, KRAS, located at 12p12.1, has been suggested as a candidate driver gene in association with 12p gain; the ubiquitous gain of 12p-derived sequences include several genes, such as CCND2, KRAS, TNFRSF1A, GLUT3, REA, NANOG, DPPA3, and GDF3, involved in the development, pluripotency maintenance and/or progression of TGCTs.⁴⁰ A focal 12p amplification (12p11.2-p12.1) including the KRAS region (12p12.1) has been observed in TGCTs.⁴¹ We detected a single NGGCT case (YST) with amplification of KRAS (Supplementary Figure 2). Recurrent gain-of-function mutations of RAS family members including KRAS were identified, as were KIT and PI3K/ AKT/ mTOR pathway components (MTOR, PTEN, and PIK3 family members)^12,13,24,42 in CNS GCTs, and were mutually exclusive to KIT alterations, which are predominant in GEs.²³ Based on the results and previous reports, another possible consequence of 12p gain can be KRAS functional activation, which may play a role in tumorigenesis in KIT-negative GE or NGGCT. The significance of KRAS alterations, particularly as a target of 12p gain, requires further investigation.

In conclusion, this study showed that 12p gain was predominantly observed in NGGCT (50%), particularly in those with malignant components (64%), although the frequency of 12p gain in CNS GCTs (30%) was less than that in TGCT (77%–88%).^17,19,20 The shared 12p status among histological components in the same specimen suggests its role in the early phase of tumorigenesis. Our data strongly suggest that 12p gain is a molecular marker indicating the existence of malignant histological components and predicts unfavorable prognosis. FISH can serve as a clinically applicable technique for detecting 12p gain, even in small biopsy specimens. We hope that our results will help risk-stratifying CNS GCT patients and plan the best therapeutic strategies for each patient.

Funding

This work was supported in part by KAKENHI No. 17K15659 (K. Satomi), Grant-in-Aid for Young Scientists, KAKENHI No. 20K17918 (H. Takami) from the Japan Society for the Promotion of Science (JSPS) and a research program of the Project for Development of Innovative Research on Cancer Therapeutics (P-Direct), Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan (No. 15 cm0106066h0005).

Conflict of interest statement. No conflicts of interest directly related to this work.

Authorship statement. Conceptualization: K.S., H.T., K.I. Manuscript writing: K.S., H.T., K.I. Final editing and approval of the manuscript: K.S., H.T., S.F., S.Y., Y.M., Y.N., T.S., S.T., A.M., N.S., M.K., T.K., T.T., K.K., M.N., T.I., K.Y., K.T., T.M., Ke.S., Ka.S., K.Y., H.T., M.N., A.A., T.U., M.M., R.N., K.I.