Preclinical Modeling of Image-Guided Craniospinal Irradiation for Very-High-Risk Medulloblastoma

Abstract

Background:

Craniospinal irradiation (CSI) is a crucial component of treatment for medulloblastoma (MB), a brain tumor clinically stratified into prognostically distinct molecular subgroups. Preclinical models of clinically-relevant CSI offer the potential to study radiation dose and volume effects in these subgroups and identify subgroup-specific combination adjuvant therapies, particularly for very-high-risk MB in which treatments are often unsuccessful.

Methods:

The commercially available Small Animal Radiation Research Platform equipped with a motorized variable collimator was used for image-guided CSI. Mice were implanted in brain cortices with patient-derived orthotopic xenografts (PDOXs) of very-high-risk Group 3 (G3) or Sonic Hedgehog (SHH) MB and were treated with fully-fractionated CSI at 2 Gy/fraction to a cumulative 36 Gy. Radiation therapy dose response effects on tumor burden and overall survival were assessed. The pattern of treatment failure was determined by bioluminescence and confirmed histologically. Acute toxicity was appraised by body weight measurements and blood work.

Results:

We established an accurate, efficient preclinical protocol to reproducibly administer CSI to mice harboring MB. CSI improved the survival of mice bearing very-high-risk G3- or SHH-MB PDOXs. However, radiation therapy dose responses across models suggested significant radioresponsiveness to conventionally-fractionated CSI ≥20 Gy. CSI was well tolerated; mice had no significant changes in body weight and acute leukopenia developed but resolved soon after therapy completion.

Conclusion:

Our protocol for preclinical CSI delivery was effective, well tolerated, and can be readily integrated into preclinical pipelines for MB and other central nervous system–seeding tumors.

Introduction

Medulloblastoma (MB), the most common malignant pediatric brain tumor, is molecularly classified into 4 major subgroups: WNT, Sonic Hedgehog (SHH), Group 3 (G3), and Group 4 (G4) (1). Patients with MB are stratified into favorable-, standard-, high-, and very-high-risk groups based on their molecular subgroup and clinicopathologic factors (2). Patients who have SHH- or G3-MBs with metastasis at presentation and/or amplification of proto-oncogenes MYCN (SHH) or MYC (G3), are classified as being at very-high-risk, with a 5-year survival probability of 28% (2).

The standard-of-care for children and adolescents with MB is surgical resection, adjuvant craniospinal irradiation (CSI), followed by focal radiation and combination chemotherapy. Patients receiving regimens that include CSI have the highest rates of tumor control, whereas those in whom CSI is delayed, omitted, or the targeted volume/delivered dose are compromised have inferior outcomes (3,4). However, CSI may cause subsequent malignancies or result in neurologic, endocrine, and cognitive deficits, especially in the youngest patients (5). For children with potentially favorable outcomes, efforts are underway to reduce radiation dose and volume or omit radiation therapy (RT). In patients with aggressive MB, efforts are aimed at intensifying treatment (e.g., incorporating potential radiosensitizers) (2,6). However, the response differences across MB subtypes remain ill defined. Given that the timing and patterns of recurrence are distinct across MB subtypes treated with adjuvant RT (6), the response to radiation may differ across MB subtypes. Thus, it is important to define RT responses across MB subgroups to refine RT guidelines and inform design of molecularly stratified trials. Experimental progress in this arena has been hampered by an inability to deliver clinically-relevant radiation protocols within the preclinical setting.

We developed a clinically-relevant, preclinical CSI protocol using the Small Animal Radiation Research Platform (SARRP), a state-of-the-art microirradiation system which allows for target localization, dose planning, image verification, and precise radiation delivery (7). We utilized patient-derived orthotopic xenograft (PDOX) models for the very-high-risk MB subtypes and established CSI dose responses across these groups. This work allows for CSI to be readily incorporated into preclinical pipelines and for the accurate preclinical evaluation of modifications to the CSI dose and field and the addition of surgical and systemic therapies.

Methods

Animal Studies

PDOXs were established in NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ mice (The Jackson Laboratory). Amplification of PDOXs and preclinical RT studies were performed in Crl:CD1-Foxn1nu (CD-1 nude) mice (Charles River Laboratories). Mice were housed in an AAALAC-accredited facility. Studies were approved by the Animal Care and Use Committee and performed in accordance with best practices outlined by the NIH Office of Laboratory Animal Welfare.

PDOX Models

The PDOXs included 2 MYC-amplified G3-MBs, the HD-MB03 cell line (8) and a primary tumor from a patient at our institution (TB-12–5950) (9), and a SHH-MB with MYCN amplification and a TP53-T125R germline mutation (TB-13–5634). All MBs were labeled with firefly luciferase for in vivo bioluminescence imaging (BLI) (9). To establish 100% luciferase-positive PDOXs, a 2-step process was used (Fig. 1): PDOX cells were infected with high-titer lentiviruses encoding luciferase and the yellow fluorescent protein (YFP) and implanted into cortices of naïve mice. Resulting tumors were sorted by flow cytometry for YFP⁺ cells and amplified by serial orthotopic implantation. Amplified PDOXs were molecularly characterized by Infinium Human Methylation 450K BeadChip (Illumina), immunohistochemistry, and fluorescence in situ hybridization and routinely tested by DNA fingerprinting to ensure tumor identity.

Fig. 1.

Schematic of establishing patient-derived orthotopic xenografts (PDOXs) marked with luciferase (A) for fluorescence-based cell sorting by flow cytometry (B). Representative bioluminescent imaging (BLI) (C) and MRI (D) of Group3 (G3)- and sonic-hedgehog (SHH)-PDOXs. Yellow circles indicate tumors.

Bioluminescent Imaging

Bioluminescence imaging (BLI) was performed weekly to monitor tumor progression in live mice. Mice were anesthetized with isoflurane and injected intraperitoneally with 125 mg/kg luciferin (ACROS Organics, Belgium, #337050500). Five minutes thereafter, the first whole-body BLI was acquired (Xenogen IVIS-200, PerkinElmer) to determine brain signals.

Orthovoltage Whole-Brain Irradiation

CD-1 nude mice were intra-cranially implanted with 1.0×10⁵ G3 HD-MB03 or 1.0×10⁶ SHH-MB cells. Mice bearing G3 HD-MB03 tumors were randomized into treatment groups 4–7 days after implantation, and those with SHH-MB 28 days after implantation. A published protocol for fully-fractionated whole-brain irradiation was used (10). In brief, a planar posterior–anterior (PA) beam was delivered at 2 Gy/day, 5 days on and 2 days off, for a total of 40 Gy for G3 HD-MB03 and 36 Gy for SHH-MB PDOXs (10).

Image-Guided CSI

Image-guided CSI was delivered with the SARRP microirradiator (XStrahl, Inc.). The treatment plan included 3 isocenters with 1 brain field using an arc treatment (−90°, 90°) and 2 spine isocenters placed in the anterior spinal cord with a PA beam arrangement (Supplementary Fig. S1). For detailed experimental procedures, see Supplementary Methods.

Toxicity Measurement

Toxicity was monitored twice weekly by measuring body weight and weekly by blood work. Complete blood counts from whole blood with EDTA were measured using a Forcyte Hematology Analyzer with impedance and laser technology (Oxford Scientific). Diagnostic chemistry panels were tested on sera by using the Trilogy chemistry analyzer (Drew Scientific).

Pattern-of-Failure Analysis

Treatment failure was monitored by real-time BLI for tumor progression along the entire neuroaxis. To determine the presence of spinal metastasis, a shield was custom-built to prevent high brain bioluminescence signals from interfering with the detection of smaller spinal tumors. Several animals expired long before we could harvest their tissue making it impossible to analyzed spinal metastasis.

The incidence and distribution of recurrent tumors were assessed for each mouse at the end stage and confirmed by MRI and/or histopathology.

Statistical Analyses

Overall survival (OS) was defined from study enrollment to euthanasia at predefined humane endpoints, as previously described (9). Statistical significance between treatment groups was determined by the log-rank test. For toxicity assessments, mixed-effects models were used to explore differences across groups over time. Results were considered significant if P ≤0.05.

Results

Whole-Brain Irradiation

Mice harboring G3 HD-MB03 or SHH-MB were enrolled in a protocol for fully-fractionated cranial RT (Fig. 2A). Median OS of mice with G3 HD-MB03 (Fig. 2B) or SHH-MB (Fig. 2C) improved from days 22.5–42 (P <0.001) and 95–122 (P = 0.0093) days, respectively. All nonirradiated mice with G3 HD-MB03 developed tumors. However, in mice with SHH-MB, 1 of 6 control and 3 of 8 treated mice were alive 30 weeks post-enrollment. Histological examination of surviving mice revealed no visible tumor in the control mouse or 1 of the 3 irradiated mice; those mice were excluded from the study. Small tumors arose in the other 2 surviving irradiated mice, and they were censored at 30 weeks. Because these experiments were done with tumors labelled with YFP and luciferase at only 50%, the bioluminescence signal did not reflect tumor burden after treatment with radiation. For all subsequent studies, tumor cells were sorted for 100% YFP positivity allowing correlation between bioluminescence (BLI), tumor growth, and enrollment based on BLI.

Fig. 2.

Protocol for fully-fractionated whole-brain RT for mice implanted with medulloblastoma (MB) patient-derived orthotopic xenografts (PDOXs). (A) Mice implanted with the Group3 (G3) HD-MB03 or sonic-hedgehog (SHH) TB-13–5634 PDOXs were treated with 40 Gy or 36 Gy, respectively, of fully-fractionated, brain-directed, fixed-beam RT. Kaplan-Meier survival curves of control (black) and irradiated (red) mice implanted with G3 HD-MB03 or (B) SHH-MB (C). Significance was determined by the log-rank test. Bioluminescent imaging of mice harboring G3 HD-MB03 (D) or SHH-MB (E) PDOXs with and without RT. Representative Haematoxylin and eosin staining of nonirradiated (control) and irradiated whole-brain radiation therapy (WBRT) G3 HD-MB03 (F) or SHH-MB (G) tumors. Yellow dashed lines outline the tumors.

In radiation-treated mice bearing G3-MB (75%) or SHH-MB (67%), tumors recurred predominantly in the olfactory bulb, a region anterior to the injection site and shielded from radiation (Fig. 2D and E, Supplementary S2A–C). In contrast, all G3 HD-MB03 and SHH-MBs tumors in non-irradiated mice were at or near the injection site. Among those with G3 HD-MB03, 80% non-irradiated and 83% irradiated mice developed spinal metastasis (Fig. 2F). In those with SHH-MB, 67% non--irradiated and 83% irradiated mice developed spinal metastasis (Fig. 2G).

Optimization of Preclinical Image-Guided CSI

We developed an image-guided protocol that consistently treated the entire neuroaxis. Proper mouse positioning was essential to deliver uniform doses to the target tissue. The natural arch of the mouse spine caused nonuniform dose distributions; thus, we used elastic athletic tape to immobilize anesthetized mice in the prone position (Supplementary Fig. S2A). We evaluated the dosimetric effects caused by variations in beam arrangement, field size, and isocenter placement; delineated targeted and non-targeted tissues with cone-beam CT (CBCT) (Supplementary Fig. S2B and C); and generated dose-volume histograms (DVHs) with MuriPlan® software.

We examined 3 simulated treatment plans (trials A-C) on a single CBCT image (Supplementary Table S1). Preliminary studies demonstrated superior sparing of the upper aerodigestive tract (data not shown) with a 180° arc treatment for cranial irradiation, and this plan was subsequently used for all simulations. Excellent target coverage was achieved: 99.8% of the brain volume received at least 95% of the prescription (V_95%).

In Trial A (Fig. 3A–D), 1 PA spinal field with an isocenter anterior to the vertebral bodies within the abdominal cavity, at a depth equal to the deepest portion of the spinal cord yielded V_95% spinal cord of 95%. However, the spinal cord and proximal tissue within the thorax and abdomen were overdosed (Fig. 3D, Supplementary Table S1). In Trial B, the spinal isocenter was positioned in the anterior portion of the spinal cord (Fig. 3E–3H), which reduced overdosing of thoracic and abdominal tissues. However, the V_95% spinal cord was inadequate (77.6%) (Fig. 3H, Supplementary Table S1). In Trial C, we used 2 spinal fields with isocenters positioned in the anterior portion of the spinal cord (Fig. 3I–L), yielding V_95% spinal cord=90.8% (Fig. 3L, Supplementary Table S1). Thus, Trial C was selected as the optimal treatment plan requiring approximately 15 minutes per mouse, giving satisfactory coverage of the spinal cord and sparing normal tissue, providing a reasonable throughput.

Fig. 3.

Illustrations of 3 simulated Craniospinal irradiation (CSI) protocols using the Small Animal Radiation Research Platform (SARRP). All mice received brain arc treatment (−90°, 90°), but spinal fields varied to determine the optimal treatment plan. Trial A used a posterior-anterior (PA) beam with a single isocenter placed anterior to the spinal column in the abdomen (A). CT images with respective isodose lines (B), isodose color wash (C), and dose-volume histogram (DVH) (D). Trial B used a PA beam with a single isocenter placed in the anterior spinal cord (E). CT images with respective isodose lines (F), isodose color wash (G), and DVH (H). Trial C used a PA beam with 2 isocenters placed in the anterior spinal cord (I). CT images with respective isodose lines (J), isodose color wash (K), and DVH (L).

To evaluate the variability in daily target and normal-tissue dosimetry, we gave 5 non–tumor-bearing mice a cumulative dose of 20 Gy over 10 fractions and generated DVHs for each animal over each fraction. Measurement of inter-animal, day-to-day variations in tissue exposure and inter-fraction variability for the brain and spinal cord indicated that the spinal cord had the most inter-fraction variability (Supplementary Fig. S3A–N). The average intra-animal variability (n=5) was calculated, with the mean V_95% brain and spinal cord being 98.8% ± 0.3% and 86.85% ± 1.08%, respectively (Supplementary Table S2).

CSI of Mice with Very-High-Risk Medulloblastoma PDOXs

To evaluate the effect of CSI dose on OS of mice with very-high-risk MB PDOXs, we treated tumor-bearing mice with 10 Gy (5 fractions), 20 Gy (10 fractions), or 36 Gy (18 fractions) and compared their OS with that of non-irradiated, tumor-bearing mice (Fig. 4A). Mice with G3 HD-MB03 (Fig. 4B–D), G3 TB-12–5950 (Fig. 4E–G), or SHH-MB (Fig. 4H–J) were enrolled.

Fig. 4.

Fractionated Craniospinal irradiation (CSI) treatment of mice implanted with medulloblastoma (MB) patient-derived orthotopic xenografts (PDOXs). (A) Schematic of the protocol. Mice implanted with Group3 (G3) HD-MB03 (B-D), G3 TB-12–5950 (E-G), or sonic-hedgehog (SHH)-MB (H-J) were treated with 0 Gy (black), 10 Gy (red), 20 Gy (green), or 36 Gy (blue) CSI. Kaplan-Meier survival curves for CSI-treated MB PDOXs (B, E, H). In vivo tumor imaging via bioluminescent imaging (C, F, I). Haematoxylin and eosin staining of PDOX tumors from control and CSI-treated (10 Gy) mice (D, G, J). Significance was determined by the log-rank test.

G3 HD-MB03 tumors were sensitive to RT; all but 3 treated mice showed decreased BLI signal (Fig. 4C). MB recurred in only 2 of 6 mice receiving 10 Gy, 1 of 6 receiving 20 Gy, and 0 of 9 receiving 36 Gy. All 7 control mice succumbed to tumor burden (Fig. 4B), with a median OS of 22 days. In contrast to G3 HD-MB03, TB-12–5950 G3-MB tumors recurred in 5 of 5 mice receiving 10 Gy, resulting in significantly higher median OS than that of untreated animals (126 days vs. 38 days, P=0.001) (Fig. 4E). Two of 5 mice receiving 20 Gy developed tumors that reached a BLI signal at enrollment greater than 1×10⁶ photons/s but died with smaller tumors (BLI signal=1.91×10⁷ and 1.88×10⁶ photons/s) compared to control mice (Fig. 4F, Supplementary Fig. S4). One of 5 mice receiving 36 Gy developed a recurrent tumor with a large tumor burden (BLI signal=1.77×10⁹ photons/s) (Fig. 4F). BLI signals for G3 TB-12–5950 also decreased immediately after RT initiation (Fig. 4F). All 7 mice harboring SHH-MB receiving 10 Gy died of tumor burden but had significantly higher median OS than did untreated mice (80 vs. 39.5 days, P=0.0002) (Fig. 4H). All mice with SHH-MB treated with 20- or 36-Gy CSI survived without recurrence and were followed for 30 weeks post-enrollment (Fig 4H). The bioluminescent signal for SHH-MB remained near enrollment threshold throughout CSI (1×10⁶ photons/s) but fell to background levels (< 5×10⁵ photons/s) post-treatment (Fig. 4I), suggesting a delayed radiation response compared with that in the G3-MBs.

Pattern of Treatment Failure Post-CSI

BLI was used to evaluate the pattern of treatment failure in mice with G3 TB-12–5950 or SHH-MB, before and after 10-Gy CSI. Whole-body BLI was used, and signals in the brain were often stronger than those in the spinal tumors. Thus, to evaluate spinal metastases, we custom-made a brain shield (Fig. 5A). Spinal metastases developed in all 7 G3- and 2 of 6 SHH-MB non-irradiated control mice compared to 3 of 5 G3- and 4 of 7 SHH-MB irradiated mice. In all cases, we detected recurrent brain tumors before spinal metastases. In 1 mouse with G3-MB, we detected spinal metastases at enrollment. After treatment, the mouse showed no evidence of spinal metastases, but ultimately the tumor recurred in the brain. When tissue was available, BLI of spinal tumors was confirmed by hematoxylin-and-eosin staining (Fig. 5B and C). In cases when the animals expired and were found after their death, the integrity of tissues was compromised preventing their analysis.

Fig. 5.

(A) Bioluminescent imaging of whole mouse (left) and head-shielded mice (right) to facilitate measurement of spinal tumors. Representative haematoxylin and eosin staining of spinal tumors of Group3 (G3)-medulloblastoma (MB) (B) and sonic-hedgehog (SHH)-MB (C) patient-derived orthotopic xenografts. Yellow arrows indicate spinal tumor metastases.

Toxicity of Preclinical CSI

We assessed the toxicity of CSI by assessing body weight and blood work (Fig. 6, Supplementary Fig. S4). CSI was well tolerated and mice maintained their body weight throughout treatment (Fig. 6A); additional food and hydration were provided as mush food when we detected minor weight loss. As expected, mice lost weight as the tumor burden increased in the control and 10-Gy groups. Mice also developed acute leukopenia (Fig. 6B) and mild neutropenia (Fig. 6C) within 1-week post-treatment. Blood cell counts recovered within 2 weeks post-treatment, indicating that CSI did not cause severe, long-term bone marrow hyperplasia. Treated mice had no anemia (Fig. 6D and E), but hemoglobin levels decreased in controls as tumor burden increased. After 1 week of CSI, mice developed mild thrombocytopenia (Fig. 6F), which resolved within 2 weeks post-treatment. Liver, kidney, and pancreatic function and general nutritional status were unremarkable, indicating that CSI did not acutely affect those organs (Figure 6G–O).

Fig. 6.

Toxicity evaluation of Craniospinal irradiation (CSI) in mice. Average body weight by treatment group (A). Complete blood count by treatment group: white blood cell (B), absolute neutrophil (C), red blood cell (D), hemoglobin (E), and platelet (F) counts. Blood chemistry panels by treatment group include aspartate aminotransferase (G), alanine aminotransferase (H), total bilirubin (I), alkaline phosphatase (J), creatinine (K), blood urea nitrogen (L), amylase (M), glucose (N), and albumin (O). Treatment groups: control (black), 10-Gy (red), 20-Gy (blue), and 36-Gy (green) CSI.

Discussion

Mice bearing PDOXs derived from very-high-risk G3-MBs with MYC amplification or SHH-MB with MYCN amplification and a TP53-T125R mutation showed increased OS on treatment with our novel preclinical protocol for fully-fractionated, image-guided CSI. The treatment plan included 1 brain field and 2 spinal PA fields, similar to clinical CSI. Although the brain treatment plan was within clinical guidelines, spinal coverage was below the common clinical threshold (V_95% ≥95%) due to the arched mouse spinal column. The number of spinal fields can be increased to obtain greater coverage if required, but this will increase the total treatment time per mouse and decrease throughput.

The first report of image-guided preclinical delivery of CSI used the X-Rad 225CX (Precision Xray) system equipped with a fixed collimator to study the effects of CSI with partial tumor resection in a genetically engineered mouse model of SHH-MB (11). Cranial RT was delivered similar to that in our study; however, spinal RT was delivered at 4.76 Gy over 6 fractions twice weekly, resulting in an estimated equivalent total dose of 34.25 Gy if delivered in 2-Gy fractions, assuming an α/β ratio of 10 (11). This regimen significantly improved OS; approximately 40% of mice were cured (11). In our study, all high-risk MBs treated with fully-fractionated CSI comprising 2 Gy for a cumulative dose of 36 Gy, closely mimicking the clinical standard of 1.8 Gy/fraction and the cumulative dose recommendation for very-high risk MB. Although our PDOXs retained characteristics of very-high-risk human MBs, current therapy incorporating RT often cannot cure these patients. When the cumulative dose was reduced to 10 Gy, G3 TB-12–5950 and SHH-MB tumors recurred in all mice, whereas G3 HD-MB03 did not.

Mice were enrolled in our protocol when their bioluminescence signal was between 1×10⁶ and 1×10⁷ photons/s indicating a low tumor burden relative to the BLI signal of a mouse at end-stage (~1×10⁹ photons/s). This BLI threshold represented the optimal range at which tumors reliably progress to endpoint. Because patients with MB generally undergo maximal-safe tumor resection before RT, we chose to enroll mice at the minimum detectable tumor size which could not be detected by magnetic resonance imaging (MRI) but was identified by immunohistochemistry (Supplementary Fig. S4). It is possible that the relative tumor burden in our PDOX models may be lower than that seen clinically, despite apparent clinical gross total resection of tumor. Additionally, low-energy photons used in small-animal irradiators may exert a larger relative biological effectiveness in mice than do high-energy sources used in the clinic (Hooshang et al, 2010). Thus 2 Gy per-day fractions at orthovoltage energies may result in an increased iso-effective dose. Finally, the radiosensitivity of murine and human tumor and normal tissue can vary (Gordon 1975; Purschke 2004; Parrinello, 2003), and the number of tumor clonogens might be smaller (Meyn RE, 1977)”.

Perhaps this contributes to differences in dose -responses between the murine SHH-MB model (11) and our human PDOX models.

Dose per fraction and total treatment time can influence clinical responses to RT (17,18). Although mice given 10-Gy fractionated CSI had a survival advantage over untreated mice in all models tested, total treatment duration was only 5 days. In contrast, clinical CSI is given for 13–20 days, depending on risk. To balance apparent cure with high-dose CSI in several MB PDOX models tested, to recapitulate standard treatment duration, and to facilitate clinically-relevant dosing schedules for combination systemic therapy, future studies should investigate the effect of reduced daily dosing over protracted treatment times. This may better model clinical treatment and provide a preclinical window to test putative radiosensitizing agents in a clinically-relevant dosing schedule.

In summary, our preclinical CSI protocol and initial RT-dose response across models of very-high-risk MB enable a more thorough evaluation of RT responses of this common pediatric brain tumor. These findings set the stage to evaluate modifications to the CSI dose and field and the efficacy and toxicity of combination therapy in the context of accurate, molecularly defined MB models, and facilitates future preclinical studies integrating surgery or systemic therapies with radiotherapy.