Introduction

Intracranial germ cell tumors (GCTs) comprise 3–15 % of pediatric brain tumors and can be sub-classified into pure germinomas (2/3 of cases) and non-germinomatous germ cell tumors based on histopathology and serum and cerebral spinal fluid (CSF) biomarkers [1–5]. Although survival rates for exceed 75 %, curative therapies come with significant morbidities, and salvage rates after relapse are as low as 50 % [1, 2].

Recent studies have identified the proto-oncogene KIT as one of the most frequent genetic abnormalities in CNS germinoma [3, 6–8]. When constitutively activated by mutations or amplification, KIT induces uncontrolled ligand-independent cellular proliferation and resistance to apoptosis [9, 10]. KIT mutation has been reported in up to 55 % of CNS germinomas, with the majority occurring in chromosome 4 codon 816. In addition to mutation, KIT over-expression is a pathologic characteristic of CNS germinoma [3, 6–8].

For CNS germinoma, there are neither cell lines nor animal models that replicate the mutated KIT and/or over-expressed KIT seen in patient tumors. To meet the critical need for in vivo model systems in pediatric neuro-oncology, our laboratory has successfully developed over 70 orthotopic xenograft models of seven tumor pathologies from patient surgical specimens. Many of these patient-derived orthotopic xenograft (PDOX) models have been extensively characterized and maintain molecular and genetic fidelity with the model-initiating human tumors [11, 12].

We consequently set out to explore whether clinically-relevant PDOX models could be developed from fresh surgical specimens of pediatric germ cell tumors. In this report, we describe the establishment and characterization of two PDOX models of metastatic pediatric CNS germinoma. We hypothesized that these models would replicate the original human tumors’ pathologic and immunohistochemical signatures and serve as biologically-accurate models for future pre-clinical investigations.

Materials and methods

Tumor specimens

Tumor specimens were obtained from five patients with intracranial germ cell tumors treated at Texas Children’s Hospital. Signed informed consent was obtained from each legal guardian prior to sample acquisition in accordance with local Institutional Review Board policy. The patients’ clinical characteristics and outcomes are described in Table 1.

Table 1 Clinical characteristics and outcomes of patients used to establish PDOX models
Full size table

Freshly resected metastatic pure germinoma from C6 spinal cord was obtained from the first patient at the time of tumor recurrence and labelled 6999GCT [3]. This metastatic tumor sample utilized for model establishment was obtained 9 months after initial biopsy of the primary tumor and following treatment with chemotherapy and radiation. Tumor tissue was washed and minced with fine scissors. Single tumor cells were collected with a 40 µm strainer and re-suspended in Dulbecco’s modified Eagle’s medium (DMEM) to achieve a concentration of 1 × 108 live cells/mL on trypan blue staining.

Metastatic CSF tumor cells were obtained from the second patient and labelled 9320GCT [3]. The diagnosis of pure germinoma was made from the resected primary supratentorial tumor mass; unfortunately, no primary site tissue was available for research. The patient had received no treatment prior to tissue acquisition.

Primary site tumor samples were obtained from the remaining three patients and processed as described above. All samples were subsequently transferred on ice to the animal facility for intracranial implantation.

Orthotopic transplantation of human tumor cells into mouse cerebra

Rag2/severe combined immune deficiency (SCID) mice were bred and housed in a pathogen-free animal facility. All animal experiments were conducted according to an Institutional Animal Care and Use Committee-approved protocol. Surgical transplantation of tumor cells into mouse cerebrum was completed via our previously described protocol [11, 12]. Briefly, mice aged 6–8 weeks were anesthetized with sodium pentobarbital (50 mg/kg, intraperitoneal). Tumor cells (1 × 105) suspended in 2 µL of culture medium were injected into the right cerebral hemisphere (1 mm to right of midline, 1.5 mm anterior to lambdoidal suture, 3 mm deep) via a 10 µL 26-gauge Gastight 1701 syringe needle (Hamilton Company).

Serial tumor sub-transplantation in vivo in mouse brains

Whole murine brains containing visible intra-cerebral (IC) tumors were aseptically removed and transferred to the tissue culture laboratory. Tumors were microscopically dissected from surrounding brain tissue, mechanically dissociated into cell suspensions, and filtered. Single tumor cells were subsequently injected into the brains of SCID mice as described above. For models IC-6999GCT and IC-9320GCT, the sub-transplantation process was repeated to complete a total of five tumor passages. In addition, injections of only 1 × 103 and 1 × 104 as well as the standard 1 × 105 tumor cells were performed for passage 5 of IC-9320GCT to determine the impact of implanted tumor cell number on xenograft survival.

Histologic and immunohistochemical characterization of xenograft tumors

After the development of neurologic deficits or once 12 months had passed following tumor implantation, mice were euthanized. Whole brains were aseptically removed and embedded in paraffin. Brains from models IC-6999GCT and IC-9320GCT at various passages were serially sectioned (5 mm thickness per slice) and stained with hematoxylin and eosin (H&E) following standard protocol.

Immunohistochemical (IHC) staining was performed using a Vectastain Elite kit (Vector Laboratories) [11, 12]. Primary antibodies included human-specific mitochondria (1:50) (Chemicon International), β-human chorionic gonadotropin (β-HCG) (Biomedia), placental alkaline phosphatase (PLAP) (Dako), and CD117 (KIT) (AB-1, Oncogen Science) [13]. After slides were incubated with primary antibodies for 90 min, the appropriate biotinylated secondary antibodies (1:200) were applied and incubated for 30 min. The final signal was developed using the 3,3′-diaminobenzidine substrate kit for peroxidase.

Pyrosequencing

Pyrosequencing was performed to detect KIT mutations using the following primers to amplifying a 134-bp genomic fragment spanning the codon 816 sequence of KIT: Forward 5′-TTGGCAGCCAGAAATATCCTCC-3′, reverse 5-Bio-CTGTCAAGCAGAGAATGGGT-3′. PCR was performed using AmpliTaq Gold (Roch) under standard conditions with 50 ng of DNA template in 50 µL. After the amplified PCR product was checked by gel electrophoresis, pyrosequencing was performed on a PSQ 96 (Qiagen). 40 µL of PCR product, 3 µL Streptavidin Sepharose High Performance beads, and 40 µL of binding buffer were well mixed in a 96 well plate. PCR products attached to the beads were washed and subsequently denatured in 0.2 Normal NaOH. Purified DNA samples were annealed to the sequencing primer (0.3 µM) and denatured for 5 min at 85 °C; samples were then processed in the PSQ 96 instrument. The sequence was analyzed using following dispensation order: GAGTAGXACA. Incorporation of a non-wild type nucleotide at position 7 with allelic frequency >10 was considered positive for mutation.

Flow cytometry analysis of putative germinoma stem cell markers

Flow cytometry analysis using LSR II (BD Biosciences) was performed to quantify the presence of candidate stem cell surface markers in the PDOX models. Single cell suspensions were stained with propidium iodine to gate out dead cells and with a four antibody anti-mouse cocktail (CD90.1, CD133, CD140a, and CD24 conjugated to allophycocyanin (Biolegend)) to gate out murine cells. Staining was then done with CD133 and CD44 antibodies conjugated to phycoerythrin and CD15, CD24, and CD49f antibodies each conjugated to FITC (MACS Miltenyi Biotec). Cells and antibodies were incubated at 4 °C prior to re-suspension in a mixture of Dulbecco’s Phosphate-Buffered Saline (Corning), bovine serum albumin (Sigma), and ethylenediaminetetraacetic acid (GIBCO-Invitrogen). Analysis of flow data was performed using Kaluza Analysis Software (Beckman Coulter).

Statistical analysis

Animal survival times were evaluated by log-rank analysis followed by pairwise multiple comparison procedures using the Holm-Sidak method with SigmaPlot 13 (Systat Software).

Results

Two metastatic germinomas formed transplantable xenograft tumors in mouse brains

Implanted germinoma cells obtained from metastatic sites (patient tumors 6999GCT and 9320GCT) successfully formed intra-cerebral xenografts (designated IC-6999GCT and IC-9320GCT) to constitute tumor passage 1. At passage 1, xenograft tumors formed in 4 of 6 implanted mice for IC-6999GCT (67 % tumor take rate) and in all nine implanted mice for IC-9320GCT (100 %). Cells from both models maintained their in vivo tumorigenicity over four serial sub-transplantations to complete five total tumor passages as documented either by grossly visible tumors upon brain harvesting or on histopathologic analysis of paraffin-embedded, serially-sectioned brains. In both models, 100 % of implanted mice formed tumors in passages 3 and 5.

In contrast, however, no xenograft tumors formed from implanted patient tumors 2597GCT, 2655GCT, or 9806GCT. Although sample size is small, these data suggest that tumor cells obtained from metastatic sites are more biologically “aggressive” than those from the primary site and thus better able to form transplantable tumors.

For both PDOX models, animal survival times were monitored over all 5 tumor passages in mice implanted with identical numbers of tumor cells (1 × 105) to examine the reproducibility of tumor growth rates (Fig. 1). In IC-6999GCT, average xenograft survival decreased significantly from passage 1–2 (p < 0.05) but subsequently stabilized, with an average median survival in passage 1 of 263 days compared to 155 days in passages 2–5 (Fig. 1a). In IC-9320GCT, the overall animal survival times were similar, particularly in passages 2 through 5 (p > 0.05) (Fig. 1b). This stabilization in animal survival times in passages 2 through 5 is in agreement with what we have observed in our other established pediatric brain tumor PDOX models and suggests that tumor growth velocity is not significantly altered over serial tumor sub-transplantations beyond passage 1 [11, 12, 14].

Fig. 1
figure 1

Germinoma PDOX model survival data. Median survival time of model IC-6999GCT passage 1 (263 days) was significantly different (p < 0.05) than passages 2–5 (155 days) (a). Median survival time of model IC-9320GCT passages 1 through 5 (53 days) was not significantly different (p = 0.097) (b). Median survival time of IC-9320GCT passage 5 (86 days) was not significant different with changes in implanted cell number (p = 0.357) (c)

Full size image

Mice bearing IC-9320GCT xenografts had significantly shorter median survival (53 days) than those implanted with IC-6999GCT (263 days in passage 1 and 155 days from passage 2–5). One explanation is that model IC-9320GCT originated from relatively pure CSF metastatic tumor cells, while the C6 spinal cord solid tumor metastasis used for IC-6999GCT establishment likely contained not only germinoma cells but also normal human stromal cells that did not grow following implantation; consequently, the selective in vivo growth and subsequent enrichment of the germinoma tumor cell likely accounts for the decreased animal survival times in passages 2–5. Of note, the significantly shorter survival of IC-9320GCT xenografts compared to IC-6999GCT corresponds to the respective patients’ outcomes, as patient 9320 died within 10 weeks of diagnosis and patient 6999 lived at least 3 years following initial therapeutic intervention.

In passages 2–5, IC-6999GCT xenograft tumors yielded an average of 1.02 × 107 live cells per tumor while IC-9320GCT tumors produced an average of 3.4 × 107 live cells per tumor, consistent with the larger size of these tumors compared to those of model IC-6999GCT (Fig. 2a). Live cells from each passage were both sub-transplanted as well as cryopreserved in liquid nitrogen for future use.

Fig. 2
figure 2

PDOX model histopathology. H&E whole brain imaging of normal mouse and xenograft models 9320 and 6999 (arrows show tumor) with corresponding patient MRI imaging (circles show tumor) (a). H&E imaging of patient and xenograft tumors from both PDOX models showing tumor cell microscopic similarities (b) and malignant features (c). IHC imaging of patient and xenograft tumors showing strong (+++) expression of KIT, no expression (−) of β-HCG, and faint expression (+) of PLAP as expected for pure germinoma (d). Scale bars white = 200 µm, black = 100 µm, green = 50 µm

Full size image

As few as 1000 PDOX tumor cells was sufficient to repopulate new xenografts

In order to determine the impact of implanted tumor cell number on xenograft survival times, xenografts of IC-9320GCT passage 5 were injected with the standard 1 × 105 tumor cells as well as both 1 × 103 and 1 × 104 cells. Animal survival times were relatively unchanged (median 86 days) despite the differences in implanted tumor cell number (p = 0.357) (Fig. 1c). This suggests that by passage 5, the xenograft model had become well established, with tumor formation occurring even after implantation of only 1 × 103 cells. The results also demonstrate that animal survival times are not solely dictated by implanted tumor cell number but by the tumor’s biological characteristics and intrinsic tumorigenicity.

The PDOX tumors replicated the histopathologic and immunohistochemical characteristics of the original patient tumors

Histopathologic and immunohistochemical examination of paraffin-embedded xenograft brains from both tumor models at various passages was performed by our institutional pediatric neuropathologist (A.A.) and confirmed that the xenograft tumors replicated the pathologic phenotypes of the original patient tumors even after 5 in vivo sub-transplantations (Fig. 2). IC-9320GCT tumors were grossly much larger than IC-6999GCT tumors (Fig. 2a). Key histopathologic malignant features including tumor cell invasion, necrosis, and intra-tumoral vascularity were noted in both models and correlated well with findings in the patient tumors (Fig. 2b, c). IHC staining of patient tumors and germinoma models revealed a profile classic for pure germinoma, including strong (+++) expression of KIT, faint expression (+) of PLAP, and no expression (−) of β-HCG (Fig. 2d).

KIT mutation and KIT over-expression were well maintained in the PDOX models

As we recently described, whole-exome sequencing of primary tumor from patient 9320 detected a D816H mutation in KIT, with a mutant allele frequency of 14 % [3]. We then performed KIT pyrosequencing on the metastatic tumor sample, which detected a mutant allele frequency of 39 % (Fig. 3). To determine the persistence of this mutation throughout serial in vivo sub-transplantations, pyrosequencing was performed on IC-9320GCT xenografts at passages 1–5. Compared to the patient tumor sample, the KIT mutation allele frequency was enriched in all xenograft passages (59–70 %) and maintained at a stable fraction over all five sub-transplantations. In contrast, as previously described, primary site tumor from patient 6999 was negative for KIT mutation as assessed using targeted deep sequencing with a custom-designed AmpliSeq array [3]. We completed pyrosequencing on the metastatic patient tumor sample, which also did not detect a codon 816 KIT mutation; in addition, no mutation was detected in any of the five xenograft passages of model IC-6999GCT (Fig. 3). However, patient and xenograft tumor cells from both models did have overexpression of KIT on IHC staining (Fig. 2d). Since KIT is the most frequently mutated gene in CNS germinoma, these models—with and without KIT D816H mutation—provide a novel resource for pre-clinical drug testing of targeted therapeutics.

Fig. 3
figure 3

KIT pyrosequencing data. Allele frequency of wild type (black) and mutant (red) KIT in patient and xenograft tumors from both PDOX models

Full size image

Putative cancer stem cell markers were detected in the PDOX tumors

Xenograft cells from both models at passages 1, 3 and 5 were stained with a panel of established cancer stem cell markers, including CD133, CD15, CD24, CD44 and CD49f, to identify candidate GCT stem cell marker(s) for future functional validation (Fig. 4) [12, 14–24]. Double staining of putative brain tumor stem cell markers CD133 and CD15 detected CD15 mono-positive (CD133−/CD15+) cells both models, ranging from 2 to 4 % in IC-9320GCT and 1–8 % in IC-6999GCT; CD133 mono-positive (CD133+/CD15−) cells, however, were only detected in model IC-6999GCT and with high variability, ranging from 1 % at passage 3–15 % in passage 5 (Fig. 4a). There were no CD133/CD15 double positive cells.

Fig. 4
figure 4

FACS analysis of candidate germinoma stem cell markers. Xenograft cells of both models at various passages had low but consistent staining for CD15 but not CD133 (other than model 6999 passage 5) across various in vivo passages (a). Neither model had significant staining for CD24 or CD44 (other than model 9320 passage 1) (b). Both models had low but consistent staining for CD49f (c)

Full size image

Similarly, CD24 and CD44, markers of breast, colon, and prostate cancer stem cells and found in medulloblastoma and glioma, were analyzed through double staining [14–17, 20]. Although high level (21 %) CD44 mono-positive (CD24−/CD44+) cells were detected in IC-9320GCT at passage 1, these cells decreased to 5 % in passage 3 and 0.5 % in passage 5 (Fig. 4b). Xenograft cells expressing CD24 mono-positive (CD24+/CD44−) or dual positive (CD24+/CD44+) were scarce (<1 % except at passage 3 in IC-9320GCT).

For CD49f, a marker of prostate and breast cancer stem cells23,24, positive cells were detected in both xenograft models at all analyzed passages (Fig. 4c). There was a steady increase of CD49f+ cells in IC-9320GCT (from 0.4 % in passage 1–4 % in passage 3 and further to 9 % in passage 5) and in IC-6999GCT (from 7 % in passage 1–22 % in passage 5), indicating an active enrichment of CD49f+ cells during serial in vivo passages in mouse brains.

Taken together, this initial screening of putative cancer stem cell markers demonstrated that CD15, CD133, and CD49f were expressed in a small population of GCT xenograft tumor cells. As they were detected in both models, CD15+ and CD49f+ GCT cells should be prioritized for future functional analysis of their self-renewal and multi-lineage differentiation capacities to determine if they are truly GCT stem cells.

Discussion

This study described the establishment of the first two known PDOX models of pediatric intracranial germinoma. These unique germinoma animal models, established from metastatic tumor tissue, replicated the key histopathologic and clinical features of the original patients. We confirmed that KIT over-expression does not require the presence of a mutation (i.e. IC-6999GCT); however, when the mutation is present (i.e. IC-9320GCT), it is not only maintained over serial xenograft passages, but the allele frequency is in fact enriched when compared to the patient tumor.

Surgical resection of intracranial germinoma is rare, with diagnosis typically occurring from tumor biopsy or assessment of tumor markers. Consequently, it is difficult to gain a clear overall pathologic picture of the invasive patterns of this tumor type from patient samples alone. Histopathologic analysis of our xenograft tumors allows a more complete understanding of germinoma invasiveness, vascularity, and mitotic rates throughout various anatomical regions of the tumor.

Furthermore, our data demonstrated that metastatic germinoma cells can form xenograft tumors from the intracranial implantation of as few as 1000 cells. In addition, tumor-induced animal death occurs in the same time frame for mice injected with 1 × 103, 1 × 104, and 1 × 105 germinoma cells. These foreboding findings underscore the fact that the goal of anti-neoplastic therapy must be to kill all malignant cells, not simply the majority, in order to eliminate the tumor and provide meaningful extension of patient survival. In addition, we showed that KIT overexpression can occur independently of KIT D816H mutation (i.e. IC-6999GCT). As KIT is the most commonly mutated gene in intracranial germinoma and KIT over-expression a pathologic characteristic of these tumors, our PDOX models—one with KIT mutation and both with KIT over-expression—are excellent representations of the genomic and pathologic spectrum of CNS germinomas.

These models represent both untreated (IC-9320GCT) and previously-treated (IC-6999GCT) tumors. This novel pair of GCT PDOX models thus provides an opportunity to evaluate the role of new therapeutics both in untreated and pre-treated/relapsed GCT.

Our data identified CD15 and CD49f as candidate cell surface markers of intracranial germinoma stem cells given their presence at low but consistent levels in both PDOX models. Although only two patient tumors were analyzed, the stability of CD49f and CD15 expression throughout serial in vivo tumor sub-transplantations supports our conclusion that these cell surface molecules should be highly ranked as candidate CNS germinoma stem cell markers for future functional validation studies.

Our PDOX models do have limitations. As the tumors are implanted into SCID mice, the models do not allow assessments of tumor immunology or investigations of immuno-therapies. In addition, we implanted germinoma tumor cells in the mouse cerebra due to the significant technical difficulty of implantation into the midbrain, the most common site of these tumors. Finally, our PDOX models have lengthy survival times and that there is a significant difference in average survival time between the two models. However, we feel that this in fact accurately reflects the widely variable clinical phenotype of germinomas.

Overall, these newly established orthotopic xenograft models of metastatic pediatric intracranial germinoma meet the critical research and clinical need for well characterized in vivo systems in which to pre-clinically test new therapeutic agents.