|Home | About | Journals | Submit | Contact Us | Français|
Brainstem gliomas (BSGs) are a rare group of CNS tumors that arise mostly in children and usually portend a particularly poor prognosis. We report the development of a genetically engineered mouse model of BSG using the RCAS/tv-a system and its implementation in preclinical trials. Using immunohistochemistry we found that PDGFRα is overexpressed in 67% of pediatric BSGs. Based on this observation, we induced low-grade BSGs by overexpressing PDGF-B in the posterior fossa of neonatal Nestin tv-a mice. To generate high-grade BSGs, we overexpressed PDGF-B in combination with Ink4a-ARF loss, given that this locus is commonly lost in high-grade pediatric BSGs. We show that the likely cells-of-origin for these mouse BSGs exist on the floor of the 4th ventricle and cerebral aqueduct. Irradiation of these high-grade BSGs shows that while 2, 6, and 10 Gy single doses significantly increased the percent of TUNEL-positive nuclei, only 6, and 10 Gy significantly induces cell-cycle arrest. Perifosine, an inhibitor of AKT signaling significantly induced TUNEL-positive nuclei in this high-grade BSG model, but in combination with 10 Gy, it did not significantly increase the percent of TUNEL-positive nuclei relative to 10 Gy alone at 6, 24, and 72 hours. Survival analysis demonstrated that a single dose of 10 Gy significantly prolonged survival by 27% (p=0.0002) but perifosine did not (p=0.92). Perifosine + 10Gy did not result in a significantly increased survival relative to 10Gy alone (p=0.23). This PDGF-induced BSG model can serve as a preclinical tool for the testing of novel agents.
Gliomas may arise anywhere in the central nervous system (CNS). An uncommon location for gliomagenesis in the CNS is the brainstem. Brainstem gliomas (BSGs) develop predominantly in the pediatric population where they constitute up to 20% of childhood brain tumors and are the leading cause of death for children with brain tumors (1). According to histopathological characteristics, BSGs are divided into high-grade (WHO Grades III and IV) which comprise about 85% of BSGs (also called DIPGs for diffuse intrinsic pontine gliomas if the tumor is primarily localized to the pons), with the remaining being low-grade (WHO Grades I and II). By contrast, BSGs in adults are much less common, accounting for <2% of gliomas. Furthermore, brainstem gliomas in adults are less aggressive than those in children and are thought to be biologically different (2).
Focal radiation consisting of daily 2 Gy doses for approximately 30 doses over six weeks is the current standard of care for these tumors - a treatment modality that unfortunately provides only temporary relief of symptoms. Despite numerous clinical investigations to date there have been no chemotherapeutic or biological agents that have proven clearly beneficial for the treatment of high grade BSGs in children. The median survival for these children is less than 1 year after diagnosis, although it has been observed that children less than 3 years of age may fare better (3).
There is limited understanding of the biology of BSGs in children due to limited availability of clinical specimens. As these tumors are not routinely resected and autopsy rates are quite low, such clinical specimens are extremely precious (4). The limited studies published on this entity demonstrate that these tumors have similar genetic alterations as pediatric gliomas arising in other parts of the brain, but distinct from the molecular characteristics of adult gliomas (5-10). ERBB1 is amplified in a small subset of brainstem gliomas and overexpressed in a larger subset (11). Kras and AKT activation have been described in 60% of pediatric high-grade gliomas including several brainstem gliomas (12). P53 mutations have been described ranging from 29-71% of pediatric brainstem gliomas (8, 10, 13-16). The most common reported molecular alterations occur in PDGFR, which is overexpressed in about 30%-80% of high-grade gliomas including brainstem gliomas (17-19), and loss of expression of Ink4a-ARF, which has been described ranging from 17-100% of pediatric glioblastomas in the brainstem (8,16,20).
There are two published rat models of BSG, both are allograft models using cell-lines derived from rat cortical gliomas that are stereotactically implanted into the brainstem. The rats develop tumors in the brainstem but it is not clear to what extent these accurately model the histology and genetics of the human disease (21,22). The cell-lines are derived from a carcinogen-induced cortical glioma, and are maintained in serum-conditions prior to implantation, culture conditions that has been shown to be inferior to culturing in bFGF and EGF in maintaining the phenotype and genotype of primary gliomas (23). In addition, a human xenograft brainstem glioma model in rats has also been recently reported which uses adult glioma cell-lines that arose from adult cortical gliomas and are implanted into immunodeficient rats (24). To our knowledge there is no genetically engineered mouse model for this disease.
Here we initially observed that PDGFRα is overexpressed in 67% of high-grade BSGs overall, including 87% of biopsy specimens. We utilized the RCAS/tv-a system to overexpress PDGF-B in nestin expressing cells lining the 4th ventricle and aqueduct. This gene transfer resulted in the formation of low-grade BSGs, while PDGF overexpression in conjunction with Ink4a-ARF loss in the same periventricular cells resulted in the formation of high-grade BSGs. These tumors were histologically similar to diffuse pediatric BSGs in many respects. A short-term analysis of RT in this BSG model showed that higher doses of RT are more effective in inducing cell cycle arrest. As the AKT pathway is thought to mediate radioresistance (25), we tested perifosine, an inhibitor of AKT signaling, alone and in combination with RT in this BSG model. Perifosine treatment significantly increased TUNEL-positive nuclei, but in combination with RT, perifosine did not significantly increase TUNEL-positive nuclei relative to RT alone at 6, 24, or 72 hours post RT. Survival analysis demonstrated that single dose 10Gy increased survival of high-grade BSG-bearing mice by 27% (57days vs. 45days; p=0.0002) while perifosine monotherapy did not. Pretreatment of 10Gy RT with perifosine did not result in a significantly increased median survival relative to 10Gy alone (64 days vs. 57 days p=0.23). In summary, our PDGF-induced BSG model is the first genetically engineered BSG mouse model that may serve as useful preclinical model for testing of novel treatment regimens.
To generate BSGs, we injected 1μl (105cells) of RCAS-PDGFB expressing DF1 cells into Ntv-a or Ntv-a; Ink4a-ARF −/− mice. The genetic background of our Ntv-a mice is a mix of BALB/c, FVB, B6, 129 and SW strains. INK4a-ARF −/−; Ntv-a strain is a mix of BALB/c, FVB, B6, and 129 strains. Greater than 20 generations of backcrossing was done to generate the lines. Injections were intracranial, 2mm posterior to the bregma at the midline position within 72 hours of birth using a Hamilton syringe. Mice were monitored carefully for signs of tumor development (hydrocephalus, lethargy, head tilt). Upon the appearance of brain tumor symptoms, mice were euthanized with CO2, brain tissue extracted, fixed in formalin, and paraffin embedded.
To determine the cell of origin, we injected 1μl (105cells) of DF1 cells expressing RCAS-GFP or RCAS-RFP into Ntv-a mice as described above. Mice were euthanized 24 hours later (n=2) or at four weeks of age (n=6). Their brain tissue was then extracted, postfixed for 30 minutes in 4% PFA, transferred into 30% sucrose at 4°C for cryoprotection, embedded in OCT, frozen on dry ice, and stored at −80°C. Brains were sectioned in 10μm sections and inspected with fluorescent microscope for GFP or RFP labeled cells.
We used total-body irradiation (TBI) for all the irradiation experiments with the exception of the survival analysis. TBI was delivered with a Gamma Cell 40 Irradiator (MDS Nordion) that delivers 106cGy/min. For survival analysis, mice were sedated with ketamine and xylazine, and irradiation of the head was done using a X-RAD 320 from Precision X-Ray at 115 cGy/min (the rest of the mouse was shielded with a lead jig). Perifosine treatment was IP at a dose 30mg/kg/day (in normal saline) once daily and was for a minimum of three daily doses. Combination perifosine and RT was done as follows: perifosine given daily for 3 days as monotherapy, followed by single dose RT. At the end of the treatment, mice were euthanized, brains were dissected, and placed in formalin.
Immunohistochemistry on mouse and human FFPE sections was performed using Discovery XT (Ventana Medical Systems, Tucson, AZ). Human BSG tissue was obtained from MSKCC, CNMC, Montreal Children's Hospital, and Carolinas Heathcare System. Analysis of human BSG tissue was IRB approved at all institutions. The following antibodies were used: PDGFRα (Cell Signaling, #3164 1:100), Ki-67 (DAKO 1:100 Human; Vector for Mouse), PhosphoH3-Ser10 (Upstate 1:800), Olig2- (Millipore; 1:250), GFAP (DAKO 1:1000). For immunofluorescence, we used anti-RFP antibody from Rockland at 1:200 followed by Alexxa 488 anti-Rabbit secondary antibody both as per manufacturer's protocol. The TUNEL assay was performed per Roche's instructions (#3333566).
All immunohistochemistry slides were scanned. One observer (OJB) performed the quantification of TUNEL, pH3, and Ki-67 using the program Metamorph where a threshold was set for TUNEL, pH3 and Ki67 and applied to all sections analyzed. At least 5 high-magnification 20× representative areas were analyzed per tumor. For quantification of pH3, positive cells were at least 5 pixels of positive area. For quantification of TUNEL-positive nuclei, we excluded pseudopalisades.
MRI of murine gliomas has been previously described (26).
To specifically investigate the PDGFRα expression in pediatric high-grade BSGs biopsy and autopsy specimens, we immunostained paraffin sections of high-grade BSGs for PDGFRα and noted that twelve out of eighteen (67%) of high-grade BSGs contained tumor cells that express PDGFRα (Table 1). Interestingly, we noted that seven out of eight surgical biopsies of BSGs had PDGFRα overexpression (87.5%), while only five out of ten autopsies expressed PDGFRα (50%). The difference in expression of PDGFRα between surgical biopsies and autopsies was not statistically significant (Two tail fisher's exact test p=0.15).
PDGFRα overexpression is a common event in pediatric high-grade BSGs, and so we were interested to determine if overexpression of PDGF in the posterior fossa of neonatal Ntv-a mice can induce brainstem gliomas. We infected RCAS-PDGF into the posterior fossa of Ntv-a neonatal mice within 72 hours of birth (approximately 2mm posterior and midline to the bregma), and observed the mice for signs and symptoms of brain tumor formation. PDGF overexpression in Ntv-a mice resulted in low-grade BSGs (Grade II; Figure 1B). The incidence of tumor formation was 80% (12 out of 15) and most of the mice were asymptomatic until euthanasia at 12 weeks (Figure 1A). Of note, no embryonal tumors (such as medulloblastomas) were observed, and the cerebellar external granule layer was not altered.
The tumor cells of these low-grade gliomas were proliferating at a low rate with a Mib1 index of 0.83± 0.16%. There was minimal baseline apoptosis as observed with cleaved caspase 3 immunohistochemistry (data not shown). The low-grade gliomas were not purely astrocytic as most of the cells in these lesions immunostained for Olig-2. These lesions were not simply reactive gliosis as they were quite large at 12 weeks with several tumors occupying approximately 50% of the pons. These lesions were easily visible on MRI (Figure1B). Mice injected with vector alone did not show any lesions. The single mouse that died developed obstructive hydrocephalus due to tumor. Characteristic immunostains of these lesions are illustrated in Figure 2C (Ki-67, Olig-2, PDGFR), and can be contrasted with immunostains of normal brainstem (Figure2D). In addition there is a GFAP immunostain of a low-grade BSG in Supplemental Figure 1A.
High-grade BSGs in humans show Ink4a-ARF loss, and therefore, we added this genetic alteration to our model. Ntv-a mice with deletion of Ink4a-ARF infected with RCAS-PDGF developed symptoms of brain tumor formation such as macrocephaly, lethargy, and hemiparesis between four and eight weeks. We sacrificed these symptomatic animals and extracted their brains. All of the mice had grade IV BSGs with pathological elements of microvascular proliferation and pseudopalisading necrosis (Figure 1C). The incidence and latency for PDGF overexpression in Ink4a/Arf null mice was 96% (68 out of 71) with a median survival of 31.5 days (Figure 1A). The survival of mice with high-grade BSGs induced in Ink4a-ARF−/− background was significantly shorter than that of mice with low-grade BSGs (p<0.0001 using log-rank test). We obtained an MRI of a subset of symptomatic mice prior to sacrificing the animals and could easily visualize the brainstem tumors in all cases (Figure 1C). We examined the brainstems of at least fifty uninjected Ntv-a mice with deletion of Ink4a-ARF and did not observe any preneoplastic lesions.
The anatomical extent of the tumors at 4 weeks was that 80% of the tumors localized completely within the brainstem (based on T2 weighted MRI images). The natural course of these tumors based on serial MRIs was that the tumors would go on to invade into the cerebellum posteriorly and invade anteriorly and superiorly into the thalamus and top of the brainstem respectively. Eventually all the mice developed obstructive hydrocephalus that likely resulted in obstruction of the CSF flow at the level of the 4th ventricle and aqueduct and its associated symptoms. Mice were euthanized at that point. The tumor volume estimate at 4 weeks (measurements were done on T2 weighted MRI images of at least 10 high-grade BSGs) was 11.9mm3 +/−4.7mm3. Histological examination revealed that 70% of the tumors occupied at least 50% of the pons.
The histological appearance of our murine BSGs was similar to human pediatric BSGs, particularly with regard to the diffusely infiltrating arrangement of the tumor cells (Figure 2A and and2B).2B). We immunostained our murine BSGs for Ki-67, Olig 2, GFAP, and PDGFRα, and we noted further similarities between these tumors and pediatric BSGs. Pediatric BSGs stained positive for Olig-2 but with large variability in the percent of positive cells in different tumors ranging from 10% to nearly 100%, while greater than 90% of the tumor cells of our PDGF induced BSGs in Ink4a-ARF−/− mice stained strongly for PDGFRα and for Olig-2 (Figure 2A,B and Supplemental Figure 2). GFAP immunostaining was positive in both low and high-grade murine BSGs, with similarity to pediatric BSGs (Supplemental Figure 1,2). Of note, Mib-1 labeling was 10.2± 5.6% for PDGF-induced BSGs in Ink4a-ARF−/− background.
The RCAS/tv-a system allows for lineage tracing of tv-a expressing cells infected with RCAS. As we were using the Ntv-a mouse, we were infecting only nestin-expressing cells. Nestin expression in the ependymal layer of the 4th ventricle has been previously described before in humans (27). In addition, it has been hypothesized that cells lining the floor of the fourth ventricle may serve as cells-of-origin for brainstem gliomas (28). We therefore decided to investigate which cells we were targeting with our infections (29,30).
We injected DF-1 cells producing RCAS-RFP into the posterior fossa of neonatal Ntv-a mice at P2 (n=2). 24 hours later we euthanized both neonatal mice, extracted their brains, sectioned and looked for RFP labeled cells. Both mice had RFP labeled cells floating in the cerebrospinal fluid and were localized around the 4th ventricle and aqueduct (Supplemental Figure 3). In parallel, we injected DF-1 cells producing RCAS-GFP in a similar fashion at P2 (n=6). We monitored the mice for four weeks (DF-1 cells have been previously reported to survive up to three weeks after injection), euthanized the mice, extracted their brains, and looked for GFP labeled cells. The cells expressing GFP were located mostly in the outer layer of the 4th ventricle (Figure 3A) or aqueduct. Of note, the distribution was not random, (Figure 3B). We found that more than 90% of the GFP labeled cells were located symmetrically at specific locations on the ventricular surface of the floor of the 4th ventricle, while only 5% were from the roof of the fourth ventricle (266 GFP labeled cells counted). In addition, we immunostained sections of neonatal pups at postnatal days 1-3 and noted strong nestin expression in the cells lining the 4th ventricle during this early postnatal age (both roof and floor of fourth ventricle as seen in Figure 3C).
To confirm that the nestin-expressing cells of the 4th ventricle and aqueduct are the cells- of-origin for our murine BSG model, we infected RCAS-PDGF as described above, observed the mice for three weeks, euthanized them, and looked for neoplastic lesions. We noted that all of the mice had small precursor lesions and 80% of the precursor lesions (12 out 15) were arising from the floor of the 4th ventricle or aqueduct (Figure 3D).
In summary, we have developed a genetically engineered BSG model that has histological similarities to human BSGs and has comparable immunostaining characteristics. Our BSG model arises from the cells of the 4th ventricle or aqueduct. Next we were interested to use this BSG model in preclinical trials
RT is the only known therapeutic modality with antitumor activity for children with BSGs. Therefore, we investigated this treatment in our mouse model. We irradiated PDGF-induced high-grade BSG-bearing mice with single dose RT at 2 Gy (approximately the dose used in the clinic), 6 Gy, and 10 Gy, and sacrificed the mice at three time-points: 6 hours, 24 hours, and 72 hours (minimum 5 mice per group). We analyzed the response of these tumors to irradiation using TUNEL and evaluated the proliferation rate using phosphoH3 (pH3). 10 Gy and 6 Gy induced a near complete cell-cycle arrest six hours post treatment as evaluated by pH3, but 2 Gy did not significantly affect pH3 (Figure 4A and B). We also found significantly increased TUNEL-positive nuclei, relative to unirradiated BSGs at 24 hours post RT for all doses of RT analyzed (Figure 4 C and D). 10 Gy time-course demonstrated that the peak inhibition of proliferation was 6 hours post RT and that the highest percent of TUNEL-positive nuclei was 24 hours post RT (Supplemental Figure 4A,B). To determine if RT prolongs survival in this BSG model we treated BSG-bearing mice with single dose 10 Gy (by only irradiating the head). Treatment was begun after pretreatment MRIs at four weeks so that similarly sized tumors were assigned to each treatment group (including untreated group). 10 Gy single dose prolonged survival by 12 days (Figure 5D; 57days vs. 45 days; p=0.0002). A 2 Gy single dose did not prolong survival significantly (p=0.14). In summary, high-dose RT is effective in prolonging survival in this BSG model.
We recently noted that perifosine, an inhibitor of AKT signaling, cooperates with radiation in medulloblastomas (31). In addition, the AKT pathway has been demonstrated to be active in radioresistant glioma cells, and we have previously shown that perifosine cooperates with temozolomide in gliomas (25,32). Therefore, we determined whether perifosine has a therapeutic effect in brainstem gliomas as a single agent and in combination with 10 Gy RT, the most effective single radiation dose. We treated a group of PDGF-induced high-grade BSGs with perifosine alone for 3-5 days, as well as a second group with three days of perifosine followed by single dose 10 Gy. Perifosine treatment did not affect proliferation significantly as a single agent and there was no statistically significant difference in proliferation with the addition of perifosine to 10Gy RT at 6, 24 and 72 hours as measured with pH3 (Mann-Whitney p=0.7, p=0.28, and p=0.06 respectively, Figure 5A). The percent of TUNEL-positive nuclei increased significantly with treatment with perifosine (Mann-Whitney p=0.005), but the combination of perifosine+ 10 Gy did not significantly affect the percent of TUNEL-positive nuclei relative to 10 Gy alone at 6, 24, and 72 hrs post RT (Mann-Whitney p=0.46, p=0.15, and p=0.09 respectively; Figure 5B). To investigate the long-term effect of these treatments, survival analysis was carried out to determine if perifosine will prolong survival as monotherapy or in combination with RT (using head irradiation) in this high-grade BSG model. Single agent daily perifosine treatment for one week did not prolong survival (43 days vs. 45 days; p=0.92), and pre-treatment with perifosine prior to 10Gy RT did not result in a significantly increased survival relative to 10Gy RT alone (64 days vs. 57 days; p=0.23; Figure 5C, ,5D).5D). Our results suggest that perifosine, when used a single agent or with single dose 10Gy, does not have sufficient antitumor activity to prolong survival in this model.
In this study we applied RCAS/tv-a technology to generate brainstem gliomas that can be used for preclinical trials. There are numerous advantages to this preclinical model. First, it is consistent with the genetic alterations of a subset of the human disease according to the published literature and our analysis of human specimens. Second, It is histologically similar to the human disease. Third, the tumors arise in immunocompetent mice. Fourth, the model has high penetrance and short latency, and forms in the appropriate microenvironment- the brainstem. Fifth, this model responds to, but is not cured with RT, like the clinical experience of treating children with DIPGs with focal RT. Sixth, the pattern of invasion of the PDGF-induced high-grade BSGs is similar to the pattern of invasion of human high-grade BSGs as a subset of the latter have been described to contiguously involve the cerebellar peduncles, cerebellum, and/or thalamus (33). It thus may serve as a preclinical model for screening of novel agents. There are only a few hundred new cases of pediatric BSGs in the world each year, and as there are now an increasingly large number of novel agents and exponential combinations of these agents- a genetic mouse model may be useful to screen for the most active combinations and schedules.
The cell-of-origin for supratentorial NF-1 deleted gliomas have been proposed to be neural stem cells from the subventricular zone (34). However, other modeling systems indicate that nestin expressing cells in other areas of the brain are also able to serve as the cell of origin (26). Here we present evidence that the cells-of-origin for this murine brainstem glioma model are mostly localized to the surface of the 4th ventricle and aqueduct. Other cells in the brainstem may also serve as the cells-of-origin for the development of murine BSGs but as we used the Ntv-a mouse, we restricted our investigation to nestin expressing cells in the brainstem in the neonatal period – which are the cells lining the 4th ventricle and aqueduct.
We evaluated external beam irradiation in this BSG model and noted that similar to the clinical experience of treating children with DIPGs with focal RT, radiation therapy is active in this model. It is important to note that the results of our preclinical testing with single doses of irradiation do not suggest that higher doses of irradiation will be more beneficial to treat children with DIPGs. A single dose of 10 Gy is a potentially toxic dose to the normal cells in the brainstem and higher total doses of radiation have been tested in clinical trials for DIPGs using hyperfractionated radiation and did not demonstrate increased efficacy (35). As BSGs are heterogeneous tumors, observations with this mouse model may only apply to BSGs that overexpress PDGFR and have lost Ink4a-Arf. If this was the case, pretreatment tissue to determine genetic status would be needed, and obtaining biopsies for DIPGs has recently been reported to have minimal complications (36). However, it is important to note that to biopsy DIPGs is a controversial topic, its success is highly dependent on who performs it, and it is hard to obtain approval for it in clinical trials (37).
We extended the preclinical testing by evaluating perifosine, an inhibitor of AKT signaling, alone and in combination with RT using this model. Our analysis showed that while perifosine monotherapy has modest activity in inducing cell-death in short-term analysis in this model it does not have enough antitumor activity to increase survival. In addition, perifosine did not provide additional survival benefit to high dose RT. This is perhaps not surprising as there is no drug that has been shown to increase survival of children with high-grade BSGs in numerous clinical trials even in combination with RT, and may be due to the poor drug delivery into the brainstem due to the blood brain barrier. This observation reinforces our recommendation that novel combinations should be tested in a genetically and histologically accurate preclinical model first prior to their translation to the clinic.
We thank Katia Manova, Sho Fujiwama, Jim Finney, Quancho Zhang, Carl Le, Dov Winkelman, and Mihaela Lupu. OJB is a St. Baldrick's scholar and is supported by MSKCC BTC, MLF and Witmer foundations, K12 Award. ECH is supported by Kirby foundation, R01 CA100688, U01 CA141502, and U54 CA126518.