Establishing a Preclinical Multidisciplinary Board for Brain Tumors

Despite decades of research, the treatment of brain tumors has remained largely unchanged. These cancers are treated with an aggressive combination of neurosurgery, radiotherapy, and chemotherapy that frequently fails to cure but inflicts significant side effects (1–4). This limited progress has occurred despite an active clinical trials effort: more than 2,580 brain tumor trials are currently registered with clinicaltrials.gov, but only six drugs are approved for treatment of brain tumors, of which only two—Everolimus, an inhibitor of the mTOR (5), and Bevacizumab, an inhibitor of VEGFA (1)—are molecular-targeted treatments.

So why have we failed to identify effective new brain tumor therapies? One possibility is that the preclinical systems used to select drugs for clinical trial do not predict therapeutic activity in patients (6). This explanation is plausible when one considers that most preclinical studies are conducted in mice harboring subcutaneous brain tumor xenografts that cannot recapitulate accurately the pharmacology or biology of brain tumor treatment. Furthermore, although brain tumor patients receive complex, multimodality therapy, mice in preclinical studies usually receive drugs as monotherapies. Such studies are unlikely to predict the survival benefit of a new treatment above that afforded by standard of care. Prioritizing treatments with the greatest potential for clinical efficacy is especially important for rare tumors that have limited patient populations available for clinical trial.

Modifying long-established treatment regimens that have evolved empirically over many years is also challenging. For example, the treatment of supratentorial ependymoma (SEP) and choroid plexus carcinoma (CPC)—two rare pediatric brain tumors—has evolved over decades to include maximum surgical resection and postoperative cranial irradiation (7–14). These treatments are effective, but evidence suggests that this efficacy varies with tumor subtype. For example, although most SEPs containing the C11ORF95-RELA translocation (hereon, SEP-CR[+]) resist combination surgery and irradiation, the majority of SEP-CR(–) tumors are cured with this therapy (11, 15, 16). Despite these differences in treatment sensitivity, ongoing clinical trials are testing whether classic histology SEPs, regardless of molecular tumor subtype, can be cured with total tumor resection alone (NCT01096368). Thus, there is a pressing need to determine if SEP-CR(+) resists surgery, radiotherapy, or both. Such knowledge is also important if we are to combine conventional and novel therapies to better treat these tumors; however, this knowledge is unlikely to be acquired solely in the clinic, especially given the rarity of disease variants. Therefore, to better understand the response of SEP and CPC subtypes to surgery and radiotherapy, and to design clinical trials that integrate conventional and new treatments, we established a preclinical multidisciplinary team (pMDT) with the capacity to conduct randomized, multimodality trials in mice harboring accurate models of SEP or CPC.

The isolation, culture, and orthotopic implantation of all mouse and human tumor cells was described previously (15–18). The nomenclature, species, tumor type, driver oncogene, and implanted cell number of each xenograft and allograft are provided in Supplementary Table S1. All cells were maintained as in vivo grafts and confirmed by ELISA as mycoplasma negative prior to and following in vitro studies. All animal studies were approved by the Animal Care and Usage Committees at St Jude Children's Research Hospital and the University of Cambridge. As discussed in detail in the Supplementary Methods, host mice for all allografts and xenografts were CD-1 nude mice (strain code: 086; Charles River). All preclinical surgery, radiotherapy, and chemotherapy studies were performed among randomized cohorts of mice harboring tumors with ≥1e10⁷ photons/sec bioluminescence (16). Tumor progression and treatment response were assessed clinically and by weekly bioluminescence (16). Mice displaying signs of excessive clinical morbidity (≥20% weight loss and/or neurological impairment) were euthanized.

Following baseline bioluminescence imaging, mice were appropriately anaesthetized, a craniotomy fashioned over the site of maximum bioluminescence, and tumors resected using a small suction tip. Postoperative hemostasis was achieved with thrombin-soaked gel foam prior to skin closure. Mice were reimaged in the immediate postoperative period, monitored on heating pads, and treated for 3 days with ibuprofen-supplemented drinking water, dexamethasone (0.6 mg/kg/6 hours), and mannitol (100 mg/kg/6 hours). Note that 54 Gy of radiotherapy was delivered to appropriately anesthetized mice as 2 Gy/day fractions via an orthovoltage irradiator or image-guided rodent irradiator (SARRP, Xstrahl). Drugs were delivered via tail vein bolus injections or using Alzet pumps (2001D, mean pumping rate ∼ 8.0 μL/h; loaded with 150 mg/mL gemcitabine solution prepared in 50:50 PEG300: propylene glycol; Supplementary Methods; Supplementary Table S2). For combination surgery–radiotherapy or surgery–radiotherapy–chemotherapy studies, mice were rested for 72 hours in between therapeutic modalities.

Pharmacokinetic studies are described in detail in the Supplementary Methods. Briefly, blood samples were collected from euthanized mice via cardiac stick into tubes containing tetrahydrouridine (THU, final concentration 150 μg/mL). Plasma was separated and samples were stored at −80°C until analysis. Intracranial microdialysis studies were performed as described previously (19). A guide cannula (MD-2255, BASi) and allografted tumor cells were implanted stereotactically in the cortex of immunocompromised mice. Once tumors formed, a precalibrated microdialysis probe (MD-2211, BASi; 38 KDa MWCO membrane) was implanted through the microdialysis guide cannula and perfused with artificial cerebrospinal fluid (0.5 μL/min). Mice were dosed with gemcitabine and plasma samples collected via retro-orbital bleeds. Drug levels were measured using a validated high-performance liquid chromatography–mass spectrometry method. Tumor cell proliferation and apoptosis were assessed by immunohistochemical quantification of Ki67 and Caspase 3, respectively (Supplementary Methods).

High-throughput screens were performed by seeding tumor cells in 384-well plates as described in the Supplementary Methods and previously (19). Each plate included dilution series of test compounds (8.3 μmol/L to 0.5 nmol/L), DMSO-only negative controls and cycloheximide or bortezomib single point (0.5 μmol/L) and dose-response (0.5 μmol/L to 0.01 nmol/L)–positive controls. Cell number was determined in each well using the Cell Titer Glo reagent (Promega). All assays were conducted in triplicate. Wash-out studies were similarly performed to assess the minimum time–concentration exposure required to inhibit cell growth by 50% by replacing drug-containing medium with fresh medium 1, 3, 6, 10, 24, or 72 hours after dosing. Tumor cell apoptosis was assessed by fluorescence-activated cell sorting to detect Annexin V staining (apoptosis) and DAPI staining (DNA integrity).

We recruited from our clinical MDT, a pMDT comprising statisticians, biologists, chemists, pharmacologists, and clinicians. The pMDT met weekly to design, conduct, and review preclinical studies that closely recapitulate multimodality clinical trials (Fig. 1). Trial statisticians ensured appropriate randomization of tumor-bearing animals and statistical powering of study arms; neurosurgeons performed all mouse neurosurgery; radiation oncologists prescribed and delivered fractionated radiotherapy to mice; clinical pharmacologists and oncologists guided trial drug doses and schedules; and radiologists and small animal imaging specialists evaluated treatment response. The pMDT adhered to strict, pre-agreed, standard operating procedures that dictated the progress of therapies through the preclinical pipeline (Fig. 1). Preclinical trial data were accessible to all pMDT members in real time via a centralized electronic mouse medical record.

Using cross-species functional genetic screens, we previously generated a series of orthotopic, genetic mouse (m), and human xenograft (x) models that recapitulate the histology, transcriptome, and growth of SEP-CR(+), SEP-CR(–), and CPC tumors (mSEP-CR[+], xSEP-CR[+], mSEP-CR[-]^RTBDN, mSEP-CR[-]^EPHB2, mCPC; Supplementary Table S1; refs. 15–18). Because clinical trials frequently employ MRI to assess treatment response, we first confirmed that bioluminescence (our preferred method of imaging) and MRI provide equivalent measures of tumor volume in our mouse models (R² = 0.96, P < 0.0001; Fig. 2A and B). Armed with these data and the survival rates of 294 tumor-bearing mice, pMDT statisticians then employed the Wilcoxon rank-sum test and Noether's power formula to design studies with a >83% power to detect a significant survival difference between animals receiving test or control treatment.

To test the therapeutic value of surgery in our models, we established cohorts of mice harboring mSEP-CR(+), xSEP-CR(+), mSEP-CR(–)^RTBDNa, mSEP-CR(–)^RTBDNb, or mCPC as described previously (15, 16, 18). Mice bearing equivalent sized tumors were then randomized to undergo microscope-guided tumor resection by neurosurgeons or anesthesia alone (Fig. 2C). Gross total resection (≤10% residual postoperative bioluminescence) was achieved in 100% (n = 14/14), 64% (n = 9/14), 51% (n = 24/47), 71% (n = 15/21), and 43% (n = 13/30) of mice harboring mSEP-CR(+), xSEP-CR(+), mSEP-CR(–)^RTBDNa, mSEP-CR(–)^RTBDNb, or mCPC, respectively, recapitulating the total resection rates of these tumors in children (Fig. 2D–H; refs. 11–14). Surgical resection of mSEP-CR(+) and xSEP-CR(+) produced only transient, significant reductions in tumor volume, and these tumors regrew rapidly following total resection, resulting in no overall survival advantage (Fig. 2D and E). In contrast, total resection produced sustained, significant reductions in the volume of mSEP-CR(–)^RTBDNa and mSEP-CR(–)^RTBDNb and increased the survival of mice harboring these tumors, curing some animals (Fig. 2F and G). Thus, the relatively poor prognosis of patients with SEP-CR(+) may in part reflect the failure of surgery to control these tumors (7). Gross total resection is generally regarded as optimal therapy of CPC, although this has not been demonstrated definitively because the disease is so rare (12, 13). In support of this notion, total resection significantly reduced tumor burden for around 1 week and marginally, but significantly, extended the survival of mice with mCPC (Fig. 2H).

Postoperative cranial irradiation has been a mainstay of ependymoma therapy for decades and is used to treat some patients with CPC (7–9, 11–13). To test the efficacy of radiotherapy in our mouse models, we randomized mice with equally sized mSEP-CR(+), xSEP-CR(+), mSEP-CR(–)^RTBDNa, mSEP-CR(–)^RTBDNb, or mCPC to receive 27 daily fractions of 2 Gy cranial irradiation (mimicking that given to patients) or mock treatment (Fig. 3A). In stark contrast with surgery, radiotherapy significantly impaired the growth of all SEP-CR(+) and SEP-CR(–) models relative to controls for between 2 and 10 weeks, resulting in a significant survival advantage for treated mice (Fig. 3B–E). Notably, regrowth of mSEP-CR(+) and mSEP-CR(–)^RTBDN was observed before the end of radiotherapy, suggesting the emergence of resistant clones, potentially explaining why this treatment ultimately failed. Conversely, and in agreement with the limited radiosensitivity of infant CPC, radiotherapy only transiently impaired mCPC growth and had no therapeutic efficacy against this tumor (Fig. 3F).

Having evaluated the efficacy of surgery and radiotherapy independently, we conducted a series of combination studies to determine the benefit of combining these modalities (Fig. 3G). Although surgery alone did not benefit mice harboring mSEP-CR(+), postoperative irradiation significantly prolonged tumor control in these animals relative to radiotherapy alone, resulting in cures for almost half of all treated mice (Fig. 3H). In contrast, surgical resection of xSEP-CR(+) did not prolong the survival of mice with this tumor relative to those treated with irradiation alone, possibly reflecting the rapid regrowth of these tumors following surgical debulking, resulting in a shorter period of tumor control overall (Fig. 3I). However, combination surgery and irradiation significantly impaired the growth of mSEP-CR(–)^RTBDNa relative to surgery alone and extended the survival of mice with these tumors beyond that achieved with either therapy alone (Fig. 3J). Combination surgery and radiotherapy were not attempted in mCPC because this tumor resisted both treatments.

Having established the value of surgery and radiotherapy among our models of SEP and CPC, and shown that the pattern of response to these treatments approximates that observed in patients, we looked to see if our models might be useful for developing chemotherapies. Using an integrated in vitro and in vivo screen that we deployed previously to identify potential brain tumor treatments for clinical trial, we screened 114 drugs that are FDA approved or currently in clinical trial (19, 20). mSEP-CR(–)^RTBDNb and mCPC cells were chosen for these studies because they represent relatively responsive and resistant tumor types, respectively. In line with their relative resistance to treatments, 40 drugs inhibited the proliferation of mSEP-CR(–)^RTBDNb cells by 50% (IC₅₀) at concentrations ≤1 μmol/L after 72 hours in vitro compared with only 26 drugs against mCPC cells: 22 of these drugs had IC₅₀ ≤ 1 μmol/L against both cell types (P < 0.0001, Fisher exact for overlap; Fig. 4A).

Thirteen drugs with IC₅₀ values ≤ 1 μmol/L at 72 hours were then subjected to “washout” studies to determine the minimum concentration–time exposure required to inhibit cell proliferation (Fig. 4B). Exposure to < 1 μmol/L of cabazitaxel, pralatrexed, gemcitabine, panobinostat, carfilzombib, or vosaroxin for just 1 hour inhibited the proliferation of both mSEP-CR(–)^RTBDNb and mCPC cells by >50%; chaetocin was similarly active against mCPC, whereas the IC₅₀ of acivicin after 1-hour exposure almost achieved 50% inhibition in the ≤1 μmol/L range. Published pharmacokinetic data indicated that cabazitaxel, gemcitabine, pralatrexate, and pemetrexed would penetrate the central nervous system (CNS) and provided appropriate doses and scheduling for vosaroxin, chaetocin, and acivicin (20–32). Therefore, to select which of these drugs might be suitable for further preclinical development, the pMDT designed and conducted a series of monotherapy preclinical trials in mice harboring mSEP-CR(–)^RTBDNb or mCPC. The goal of these studies was to look for any evidence of antitumor activity (growth and/or survival). Doses and schedules of each drug were designed to mimic those achievable in patients. We also conducted a monotherapy study of cladribine as a “negative control” compound because this drug was relatively inactive in vitro and was predicted not to penetrate the CNS. Of all agents tested, only gemcitabine (120 mg/kg intravenous bolus) displayed significant activity against both mSEP-CR(–)^RTBDNb and mCPC: this treatment was the only monotherapy to significantly impair the growth of mSEP-CR(–)^RTBDNb and to prolong the survival of mice with these tumors; therefore, this drug was selected for further repurposing studies (Fig. 4C and D). Vosaroxin—a topoisomerase II inhibitor causing site-selective DNA damage—also produced a modest but significant survival advantage for mice harboring mSEP-CR(-)^RTBDNb tumors (Fig. 4C). These data underscore that drugs with relatively potent activity in vitro may lack efficacy in vivo when administered at clinically relevant doses. In light of the considerable activity of gemcitabine, we selected this drug for further preclinical development.

Gemcitabine can be administered as an intravenous bolus or infusion, resulting in very different pharmacokinetic profiles (33). Therefore, we treated mice bearing mSEP-CR(–)^RTBDNb or mCPC with various gemcitabine regimens and simultaneously measured concentrations of the drug in plasma and brain tumor extracellular fluid (tECF) using intratumoral microdialysis. Mice were treated initially with two clinically relevant gemcitabine regimens: 60 mg/kg i.v. bolus that is active against a mouse model of group 3 medulloblastoma; or continuous 3-hour infusion via subcutaneous Alzet pumps (19, 20). Note that 60 mg/kg i.v. bolus gemcitabine produced a plasma AUC_0–6hr of 25.9 μmol/L*hr that is equivalent to that observed in children treated with 1,200 mg/m² (Fig. 5A; ref. 34). The tumor-to-plasma partition coefficient for unbound gemcitabine (K_p,uu) at this dose was 0.51 and 0.18 for mice bearing mSEP-CR(–)^RTBDNb and mCPC tumors, respectively. The gemcitabine concentration in tECF produced by this regimen only remained above the in vitro washout IC₅₀ of each tumor type for less than 3 hours (compare Figs. 4B and 5A). In contrast, 3-hour infusions of gemcitabine produced plasma exposures of 95.1 ± 24.1 μmol/L*hr—equivalent to treating children with 2,000 mg/m²—and in both models maintained a tECF concentration above the IC₅₀ in washout studies for ≥7 hours (compare Figs. 4B and 5B). To determine if these in vivo exposures produce the antitumor cell effects predicted in vitro, we harvested tumors from mice at 3, 8, 24, and 48 hours following initiation of gemcitabine therapy and estimated levels of tumor cell proliferation and apoptosis. Three-hour infusions of gemcitabine induced significantly greater and more sustained levels of tumor cell apoptosis in mSEP-CR(–)^RTBDNb and mCPC than did 60 mg/kg i.v. bolus treatment, and 3-hour infusions produced a more significant and sustained reduction in tumor cell proliferation, although this was only observed in mSEP-CR(–)^RTBDNb (Fig. 5C; Supplementary Fig. S1).

As a final step to select the optimal dose and schedule of gemcitabine for preclinical assessment, we further expanded the repertoire of gemcitabine regimens to assess the relative activity of 200 mg/kg bolus and 6-hour infusions (n ≥ 10 mice per cohort; Fig. 5D–F). Of all regimens tested, 3- and 6-hour infusions of gemcitabine were most efficacious, producing similar degrees of tumor growth suppression and enhanced overall survival; however, 3-hour infusions proved the least toxic. Additional 3-hour gemcitabine infusion monotherapy trials identified significant active against mSEP-CR(+), xSEP-CR(+), and a mSEP-CR(–) model driven by EPHB2 (Fig. 5G–I; ref. 17). Three-hour gemcitabine infusions were more efficacious than combination cisplatin/cyclophosphamide or cisplatin/etoposide/vincristine that approximates “standard of care” chemotherapy regimens that have been tested against ependymoma and CPC, respectively, in the clinic (Fig. 5J and K; refs. 7, 35). Therefore, we selected 3-hour infusions of gemcitabine for our final phase of preclinical repurposing.

The efficacy of gemcitabine monotherapy in our model systems suggests it may have value as an adjuvant therapy in the clinic. Therefore, the pMDT designed a combination study aimed at testing the value of adding 3-hour gemcitabine infusions to “standard-of-care” surgery and fractionated radiotherapy (Fig. 6A). With regard to ependymoma, we focused on SEP-CR(+) disease because this tumor type is the most aggressive form of SEP. The mSEP-CR(+) rather than xSEP-CR(+) model was chosen for these studies because these models displayed similar responses to surgery, radiotherapy, and gemcitabine as monotherapies, but the more rapid growth profile of mSEP-CR(+) enabled completion of these large combination studies in a timely manner. Mice were treated with GTR followed by 54 Gy fractionated irradiation and then 3 weeks of consecutive 3-hour gemcitabine infusions. This treatment was tolerated remarkably well. Although the average tumor burden of mice receiving gemcitabine was lower than that of animals treated with surgery and irradiation alone, this difference was not significant (Fig. 6A); however, the addition of gemcitabine doubled the median survival (183 days) of mice relative to those treated with surgery and irradiation alone (96 days), and cured 50% of animals (P < 0.00001; Fig. 6B). These data underscore the need to assess both tumor volume and animal survival as response metrics to preclinical therapy because tumor imaging in small animals may operate at the limits of resolution. We next assessed the value of adding serial postoperative, 3-hour gemcitabine infusions to the treatment of mCPC (Fig. 6C and D). As observed previously, gross total tumor resection alone produced a modest but significant survival advantage for mice harboring mCPC (median survival surgery = 34 days vs. control = 22 days; P < 0.003; Fig. 6D); and gemcitabine therapy alone markedly extended the survival of mice with these tumors (median survival gemcitabine = 42 days vs. control = 22 days; P < 0.0001; Fig. 6D). Notably, no significant difference in survival was observed between mice undergoing surgery or gemcitabine therapy alone; however, surgical resection followed by gemcitabine significantly extended survival above that of animals receiving surgery alone (median survival surgery alone = 44 vs. surgery+gemcitabine = 46.5 days; P < 0.0001; Fig. 6D). Together, these data suggest that 3-hour infusions of gemcitabine may add therapeutic value to “standard-of-care” surgery and radiation in the treatment of SEP and may improve the results of surgical resection of CPC. We propose that these regimens should be tested in the clinic.

The past decade has witnessed a revolution in our understanding of human cancer. The integration of genomic and developmental biology has shown that morphologically similar cancers comprise subtypes, driven by different genetic alterations, which likely arise within distinct cell lineages (36). These data help explain why cancers once regarded histologically as homogeneous diseases display discrepant behaviors. For example, medulloblastoma and ependymoma are now known to include subtypes with extraordinarily good (e.g., WNT-medulloblastoma and SEP-CR[-]) or bad (Group 3-medulloblastoma with MYC amplification and SEP-CR[+]) prognosis (11, 37). This knowledge could pinpoint patients who might be cured with less toxic therapy, as well as poor prognosis patients who need new treatments. Indeed, clinical trials of decreased radiotherapy are ongoing among patients with WNT-medulloblastoma (NCT01878617). But integrating understanding of tumor biology into established clinical practice is enormously challenging and requires a number of assumptions that are often made without knowledge of subtype-specific treatment efficacy. For example, reducing radiotherapy for children with WNT-medulloblastoma assumes that this therapy, rather than surgery or chemotherapy, is relatively redundant. And trials of new treatments for ‘poor prognosis’ tumors often assume that relatively ineffective conventional therapies should be retained; this approach runs the risk of increasing toxicity unnecessarily.

So how can we integrate new understanding of cancer biology and therapy into empirical treatment regimens that have developed over decades? It is unlikely that we will achieve this through clinical trials alone, especially among patients with rare disease variants: the small populations of patients with these tumors limit the number of drugs and regimens that can be tested in a timely manner. The preclinical platform described here provides an evidence-based approach to guide clinical trials for rare brain tumor subtypes. It is important to note that this platform is not designed to replace or reduce the clinical trial platform; but rather to better triage drugs so that clinicians can focus on novel regimens with the greatest potential to cure. Key features of our approach include the use of accurate mouse models of human brain tumors and the coordinated engagement of clinical and research professionals in regular pMDT discussions, greatly facilitating the codevelopment of clinically relevant preclinical trials.

Maximal surgical resection of ependymoma followed by irradiation is consistently associated with a better patient outcome regardless of primary tumor site (8, 10, 11, 14). This observation has led to the widespread notion that SEPs have a high probability of cure with surgery alone, and underpins Arm 1 of an ongoing Children's Oncology Group study in which children achieving a gross total resection of classic histology SEP receive no further treatment (NCT01096368; ACNS0831). But in our studies, surgery alone had no therapeutic value in the treatment of mSEP-CR(+) or xSEP-CR(+), only benefiting mice with mSEP-CR(–)^RTBDNa. However, total resection of mSEP-CR(+) did markedly improve the efficacy of irradiation, curing a significant number of animals and also improved the survival of mice with mSEP-CR(–)^RTBDNa. These data underscore the important point that treatments can display surprising interactions, producing high cure rates when used in combinations that not are not apparent when the treatments are used individually. Our data also support the notion that irradiation is a highly effective treatment of SEP and suggest that avoiding radiotherapy for all patients with totally resected SEP, regardless of subtype, may be inappropriate. Rather, as a minimum, we recommend the prospective evaluation of SEP molecular subtype in ACNS0831 to ensure that SEP-CR(+) patients are not undertreated. Reverse translation of these clinical data will be critical to validate the predictions made by our preclinical system. This later point is particularly important because our model system is likely to be predictive but not infallible. Indeed, in contrast with mSEP-CR(+), total resection of xSEP-CR(+) did not improve the efficacy of radiotherapy, indicating that further iteration between preclinical and clinical work will be required to understand the ependymoma-subtype–specific relevance of combination treatment.

The effectiveness of preirradiation surgery in our models also supports the widely held notion that cytoreductive surgery increases radiosensitivity and chemotherapy by removing therapy-resistant, hypoxic, and highly-proliferative tumor cores (38). These data might also explain why resection and irradiation are more effective than partial resection and irradiation among patients with posterior fossa subtype-A ependymoma—another aggressive disease subtype (14). Thus, our models provide an opportunity to explore the biological basis of cytoreductive surgical efficacy. Our models may also facilitate the identification, isolation, and study of radiation resistant tumor clones, because our imaging studies revealed regrowth of mSEP-CR(+), xSEP-CR(+), and mSEP-CR(-) prior to completion of radiotherapy.

In contrast with our SEP models, radiotherapy proved ineffective against mCPC. The radioresistance of mCPC may reflect the Tp53-null status of these tumors, because this gene mediates cell death mechanisms in irradiated cells (39). Notably, 60% of human CPCs contain mutant Tp53: these tumors also tend to be radioresistant, clinically aggressive, and to develop in infants (40, 41). It is interesting that our mCPC model is initiated in embryonic choroid plexus; therefore, these tumors likely model radioresistant, aggressive, and TP53-mutant infant CPC (18).

Although chemotherapy has been evaluated in ependymoma and CPC, its role remains controversial with only limited benefit reported (7, 13). These data are in keeping our observations that most drugs displaying potent activity against our models in vitro failed to produce therapeutic benefit in vivo. This notion is also supported by our observation that mouse models of SHH-medulloblastoma—a more chemosensitive disease—responded to treatments that were ineffective against mSEP and mCPC, e.g., pemetrexed and 60 mg/kg bolus gemcitabine (20, 42). Our preclinical in vitro and in vivo pipeline did identify 3-hour infusions of gemcitabine as a potential new treatment of SEP and CPC. This regimen generated tECF concentrations above the in vitro IC₅₀ for ≥7 hours and proved more effective against mSEP-CR(+) and mCPC than combination conventional chemotherapy regimens with reported activity in patients (7, 35). Thus, we suggest that 3-hour infusions of gemcitabine should be tested in patients with SEP and CPC.

Fewer than 150 adults and children with all variants of SEP and CPC are available for enrollment on clinical trials in the United States each year, severely limiting studies of new treatments (43). Indeed, it is widely agreed that the rarity of CPC poses an almost insurmountable hurdle to the efficient development of new treatments through clinical trial (13). For example, the only multicenter CPC clinical trial conducted to date was initiated in 2000 (CTP-SIOP-2000), but 17 years later, the results of this trial are yet to be published. Our preclinical system provides an alternative, evidence-based approach to prioritize combination regimens for the clinic, potentially avoiding years of trials of ineffective therapies. Of particular note, by recapitulating surgery, irradiation, and chemotherapy, our approach allows for preclinical trials of multiple doses, delivery routes, and schedules of novel chemotherapies in the context of standard-of-care treatment. In this regard, sequential total tumor resection, fractionated radiotherapy, and 3-hour gemcitabine infusions doubled the median survival of mice with mSEP-CR(+) relative to surgery and irradiation alone, curing half of all animals. Our studies also provide evidence that combination surgery and gemcitabine infusion therapy may benefit the treatment of CPC. These data are in keeping with the activity of gemcitabine in other chemoresistant cancers including pancreatic cancer (44, 45). We therefore recommend that gemcitabine infusions might prove effective as postsurgery and irradiation chemotherapy. Furthermore, because gemcitabine may also serve as a radiosensitizer, we are currently exploring the timing of gemcitabine treatment relative to irradiation and whether gemcitabine may be added to conventional treatment regimens in younger patients.

Although our model system provides a promising tool to prioritize complex combination treatment regimens for clinical trial, the accuracy of these predictions remains to be assessed. It is a hard reality that most cancer treatments that are effective in animal models fail in patients (46, 47). Indeed, although our models closely replicate the morphology and transcriptome of the corresponding human tumors, they are maintained in immunocompromised hosts and therefore cannot account for contributions of the host immune system to tumor biology and treatment. Thus, preclinical platforms such as the one presented here require careful iterative study with clinical translation to be validated and refined. This important ongoing process further underscores the value of convening pMDT teams comprising laboratory and clinical oncology professionals.

B.B. Freeman is an employee of and has ownership interests (including patents) at KinDynaMet LLC. M.F. Roussel is a consultant/advisory board member for Cold Spring Harbor Laboratories and the National Cancer Institute, and reports receiving commercial research support from Eli Lilly. No potential conflicts of interest were disclosed by the other authors.

Conception and design: B.V. Nimmervoll, N. Boulos, A. Shelat, A. Gajjar, Y.T. Patel, R.K. Guy, T.E. Merchant, K.D. Wright, R.J. Gilbertson

Development of methodology: B.V. Nimmervoll, N. Boulos, B. Bianski, J. Dapper, A. Shelat, S. Terranova, A. Gajjar, Y.T. Patel, B.B. Freeman, A. Onar-Thomas, C.F. Stewart, R.K. Guy, T.E. Merchant, C. Calabrese, K.D. Wright, R.J. Gilbertson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.V. Nimmervoll, N. Boulos, B. Bianski, J. Dapper, M. DeCuypere, Y.T. Patel, B.B. Freeman, C.F. Stewart, T.E. Merchant, C. Calabrese, K.D. Wright, R.J. Gilbertson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.V. Nimmervoll, N. Boulos, J. Dapper, A. Shelat, Y.T. Patel, B.B. Freeman, C.F. Stewart, R.K. Guy, T.E. Merchant, C. Calabrese, K.D. Wright, R.J. Gilbertson

Writing, review, and/or revision of the manuscript: B.V. Nimmervoll, N. Boulos, A. Shelat, A. Gajjar, A. Onar-Thomas, C.F. Stewart, M.F. Roussel, T.E. Merchant, K.D. Wright, R.J. Gilbertson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.V. Nimmervoll, N. Boulos, A. Gajjar, B.B. Freeman, K.D. Wright, R.J. Gilbertson

Study supervision: B.V. Nimmervoll, N. Boulos, A. Gajjar, K.D. Wright, R.J. Gilbertson

Other (wrote the manuscript): R.J. Gilbertson

This work was supported by grants from the NIH, P01CA96832 (R.J. Gilbertson, B.V. Nimmervoll, N. Boulos, A. Gajjar, C.F. Stewart, and M.F. Roussel) and R0CA1129541 (R.J. Gilbertson and N. Boulos); the American Lebanese Syrian Associated Charities (R.J. Gilbertson, B.V. Nimmervoll, N. Boulos, A. Gajjar, C.F. Stewart, M.F. Roussel, B. Bianski, J. Dapper, A. Shelat, Y.T. Patel, B.B Freeman, A. Onar-Thomas, R.K. Guy, T.E. Merchant, C. Calabrese, and K.D. Wright); Cancer Research UK (R.J. Gilbertson, B.V. Nimmervoll, and S. Terranova); the Mathile Family Foundation (R.J. Gilbertson, B.V. Nimmervoll, and S. Terranova); and Cure Search (R.J. Gilbertson, B.V. Nimmervoll, and S. Terranova).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.