Development of Chromosome 1q+ Specific Treatment for Highest Risk Pediatric Posterior Fossa Ependymoma Free

Ependymoma (EPN) is an aggressive pediatric brain tumor with poor survival and significant morbidity, particularly in the context of multiple recurrences (1). Most pediatric EPN arise in the posterior fossa (PF) and are predominantly group A EPN (PFA). Over 70% of children with PFA tumors will relapse, and once relapsed, effective therapies are lacking (2, 3). Specific negative prognostic risk factors include PFA and gain of chromosome 1q (1q+) based on multiple national and international study cohorts. Children's Oncology Group (COG) ANCS0121 trial demonstrated a 5-year progression-free survival rate for PF EPN patients with 1q+ of 47.4% and 1q wild-type (WT) of 82.8% (P = 0.189; ref. 2). In a more recent multicenter study of matched primary and recurrent pediatric EPN tumors, 25% had gain of 1q+; however, this number doubled at recurrence with 50% of tumors at first recurrence having 1q+ (4).

The role of chemotherapy in the treatment of EPN remains uncertain. Several ongoing clinical trials in the United States and Europe are evaluating standard chemotherapy agents, agents that are known to cross the blood–brain barrier, as opposed to drugs that have a biological rationale specific to EPN. Although the COG ACNS0831 trial, which randomized PF EPN patients to a multiagent chemotherapy regimen or observation following radiation, has completed accrual, data analysis is ongoing. In the meantime, several groups have conducted high-throughput drug screens in models of pediatric EPN, including PFA cell lines with 1q+, to identify rational chemotherapeutic agents to combine with or follow standard radiotherapy (5, 6). These preclinical screening studies identified purine antimetabolite 5-fluorouracil (5FU) as an EPN-selective therapy. 5FU is a well-established chemotherapeutic agent that rapidly enters cells and is converted to several active metabolites causing downstream DNA and RNA damage. 5FU is used for standard treatment of cancers, such as colorectal adenocarcinoma, hepatoblastoma, and nasopharyngeal carcinoma. Of specific relevance to PFA, a recent report found that gain of 1q+ was associated with significantly higher sensitivity to 5FU compared with WT 1q tumor cells of multiple tumor types (7).

A pharmacokinetic study demonstrated optimal drug concentrations in the brains of murine models of EPN with intravenous bolus administration of 5FU, as opposed to continuous infusion methods prevalent among cancer regimens (8). Preclinical pharmacokinetic–pharmacodynamic analysis from this study confirmed the dosage selected for intravenous bolus 5FU achieved target exposure for antitumor effect (8). A St. Jude phase I study was subsequently designed to establish the safety profile of single-agent weekly intravenous bolus 5FU in a cohort of recurrent EPN (9). Twenty-three relapsed pediatric EPN patients (16 PF; 6 supratentorial; 1 spinal tumor location) were enrolled. Of the enrolled PF patients, 5 (31%) patients experienced partial responses (duration 6–54 weeks, median 12 weeks), and patients who demonstrated objective tumor responses were treated at dosage 500 mg/m² or higher (9). Although this study was encouraging, there was no clear mechanism for 5FU sensitivity in pediatric EPN patients, and further clinical trials were not explored.

Considering the known radiosensitization properties of 5FU in multiple tumor types, the efficacy of combining 5FU during radiation in in vitro and in vivo preclinical models of 1q+ PFA was tested. Additionally, we identify the mechanism by which 5FU sensitivity is enhanced in 1q+ PFA cells. Further, we identify retinoid tretinoin (ATRA), which was previously identified as a PFA-selective FDA-approved oncology drug (5), as a rational and effective maintenance strategy for combination with 5FU and chemotherapy, which in vivo showed significant improvement of long-term survival of mice with 1q+ PFA. This manuscript provides a biologically tested therapy approach in both in vitro and in vivo preclinical models of highest-risk pediatric EPN subtype, which combines already FDA-approved agents with known safety profiles in children, making it readily translatable to large-scale clinical trials.

EPN cell lines, MAF-811 and MAF-928, were established from recurrent pediatric PFA that have been well characterized (10). Both cell lines harbor the highest risk phenotype of chromosome 1q gain (1q+) and loss of chromosome 6q (6q−). These cell lines are established from flank tumors propagated in NSG mice and viably cryopreserved in Bambankers (BB001, GC Lymphotec). Cell lines are routinely screened for Mycoplasma, and short tandem repeat sequencing to ensure cell line identities. For all in vitro experiments, cells are cultured as monolayers in Opti-MEM with 15% FBS and 1% penicillin/streptomycin (pen/strep; O15; Gibco). HEK 293 FT cells (RRID:CVCL_6911, ATCC) were cultured in DMEM supplemented with 10% FBS and 1% pen/strep. All in vitro experiments were performed using cells below passage 5 following thawing.

5FU (Sigma PHR1227) and ATRA (Sigma PHR-1187) were dissolved in DMSO and stored in small aliquots at −80°C. For in vitro experiments, 5FU and ATRA stocks were further diluted in serum-free Opti-MEM. For in vivo experiments, 5FU stocks were diluted in sterile PBS, and ATRA was made fresh daily by resuspending the powder in sterile corn oil (Sigma C8267). All work involving chemotherapy agents is performed under institutionally approved safety protocol (IBC-00001304).

MAF-811 and MAF-928 cells were seeded at 4,000 cells per well in six 96-well plates (Corning) in O15. Cells were cultured at 37°C and 5% CO₂ and monitored using an IncuCyte Zoom (RRID: SCR_019874, Essen BioScience). After 24 hours, plates were treated with radiation on a cesium source irradiator. Each dose was delivered in 1.25 Gy fractions, rotating the plates at each fraction until the final radiation dose was achieved. Immediately following radiation CellEvent Caspase-3/7 Green Detection Reagent (Life Technologies) and 5FU curve starting at 10 μmol/L was added to the plates. For combination experiments using ATRA, radiation and 5FU treatments were repeated as described and cells incubated for 48 hours before 10 μmol/L ATRA was added. Images were captured at 4-hour intervals from four separate regions per well using a 10× objective over 10 days. Each experiment was done in triplicate and the accumulation of caspase-3/7 over time was normalized to the confluence of cells.

Following the 10 days of incubation on the IncuCyte, cell viability was measured using CellTiter-Glo (Promega) following the manufacturer's protocol. The optical density of each well was measured using CellTiter-Glo protocol template on GloMax Explorer (RRID:SCR_015575, Promega). The proportion of cells per treatment group was normalized to control wells of DMSO control-treated cells.

Gene expression of MDM4 and UCK2 was obtained from existing RNA-seq transcriptomic data published as part of a recent study (GSE226961) in accordance with local and federal human research protection guidelines and institutional review board regulations (COMIRB 95-500). These data were supplemented with new RNA-seq data from 16 “normal” brain samples (GSE244124) obtained predominantly from autopsy material that is collected after parent approval and deidentified in accordance with HIPPA regulations. RNA was isolated from autopsy samples and sequenced using the Illumina Novaseq6000. Sequencing data were aligned using STAR (RRID: SCR_004463) and quantification of RNA expression as fragments per kilobase per million and derived by Cufflinks (RRID: SCR_014597) as described previously (11).

Total RNA was isolated from cells using DNA/RNA All Prep Kit (Qiagen, cat. #80204) and analyzed for gene expression using qRT-PCR performed on a StepOne Plus Real-Time PCR System with TaqMan Gene Assay Reagents (Applied Biosystems), according to the manufacturer's protocols. cDNA was generated from total RNA using a High-Capacity cDNA Kit (Applied Biosystems). The resulting cDNA was loaded at 50 ng per well for qRT-PCR (four replicates), along with TaqMan gene-specific probe for UCK2 (Hs00989900_m1) and CDKN1A (p21; Hs00355782_m1). Relative gene expression was calculated using the ΔΔC_t method with 18S (Hs03003631_g1) as endogenous control.

To investigate the mechanism of cell death from 5FU and radiation-treated cells, we used the Human Apoptosis Array Kit (RRID: AB_3086729, R&D ARY009). The apoptosis array is a nitrocellulose membrane that detects 35 apoptosis-related proteins including positive controls. To generate lysates, MAF-811 and MAF-928 cells were plated 10⁶ cells in 10-cm dishes. Cells were irradiated 24 hours later, and 5FU was added immediately after radiation. Cells were harvested per the manufacturer's protocol 72 hours after treatment with radiation and 5FU. Briefly, cells were washed in PBS and then homogenized in lysis buffer using a rotorstator. Cell lysates were gently rocked for 30 minutes at 4°C and centrifuged at 14,000 × g for 5 minutes (4°C). The supernatants were removed and stored at −80°C until use. A total of 250 μg of protein, as determined by the BCA assay, was used for each array and developed using the chemiluminescent reagent provided. Positive signals on the developed film were analyzed using a transmission mode scanner and collected at pixel density using ImageJ software (RRID:SCR_003070, https://imagej.nih.gov/ij).

Genetic knockdown of UCK2 (TRC number TRCN0000037849 and TRCN00000380355) was performed using a Sigma-Aldrich pLKO system. shRNA plasmids were packaged into the virus by transfection of HEK 293FT cells with TransIT-LT1 Reagent (Mirus) and pMD2G (RRID:Addgene_12259) and psPax (RRID:Addgene_12260) packaging plasmids. Viral supernatant was harvested and centrifuged to remove debris. Polybrene was added to the virus and to EPN cell lines prior to viral transduction. Virus was removed after 24 hours, and transduced cells were positively selected with puromycin. Knockdown validation was performed using qRT-PCR.

We have previously published the characteristics of MAF-928_XC patient-derived orthotopic xenograft models (PDX; ref. 12). Briefly, 250,000 cells from viably frozen stocks of disaggregated flank tumor cells were implanted into the fourth ventricle (lat: 0.000, ant: −6.000, ventr: −4.3000; with respect to the bregma), under stereotactic control, of 6- to 8-week-old NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (RRID:IMSR_JAX:005557, bred in-house) mice. Sequential T2-weighted magnetic resonance imaging (T22-MRI, field strength 9.4 Tesla) to confirm tumor engraftment was performed between 6 and 8 weeks after surgery with a 90% to 100% take rate for this PDX model. All mouse work was approved by Colorado University (CU) Anschutz Institutional Animal Care and Use Committee (IACUC 00784). All mice were housed in CU Anschutz Vivarium. Mice were maintained at five per cage under humidity and temperature-controlled conditions with a light/dark cycle set at 12 hours. Mice were given plastic huts, cotton pads, and autoclaved cardboard toilet paper rolls to provide enrichment. Handling was performed with universal sterile precautions and experienced personnel performed all procedures. There were two full-time veterinarian technicians and a staff of animal caretakers. Animals were monitored for discomfort daily by laboratory animal technicians. Due to the severe stress that can be created by tumor development, mice were sacrificed as soon as signs were evident in consultation with our veterinarians.

Mouse numbers for each group were determined a priori using a 25% increase in survival with a power of 90% and an alpha of 0.05. Six mice were needed per treatment arm to achieve significance with three experimental replicates. Following confirmation of tumor formation by T2w-MRI scan, cages were arranged in numerical order and mice numbered 1 to 36. Treatment assignment was given based on order in which drugs were administered [vehicle control (n = 5), 5FU only (n = 4), 5FU + ATRA (n = 5), 10 Gy + vehicle (n = 12), 10 Gy + 5FU (n = 12), and 10 Gy + 5FU + ATRA (n = 12)]. For chemo-only groups in the initial experiment, cage floods resulted in death unrelated to tumors. With the numbers that survived, chemotherapy showed little effect. This prompted the elimination of chemotherapy-only groups in repeat experiments to prevent unnecessary harm. MRI analysts, who provided read-outs on tumor volumes, were blinded to treatment group assignment. High-resolution T2w-MRI scans were performed on a Bruker 9.4 Tesla BioSpec scanner, and all analyses were carried out using ParaVision NEO v.2.2 as previously described (12).

Irradiation was performed under IACUC protocol 00784. Using the X-RAD SmART image-guided irradiator (RRID:SCR_021996, Precision X-Ray Inc) at 225 kVp with a 0.3-mm copper filter, mice received 10 Gy in five daily fractions at a dose rate of 5.9 Gy/minute. Under isoflurane anesthesia, each mouse was positioned in prone orientation and aligned to the isocenter in two orthogonal planes by fluoroscopy (12). Half the dose was delivered from each side in opposing lateral beams. Monte-Carlo simulation in SmART-ATP (SmART Scientific Solutions B.V.) showed 95% coverage of the target volume (fourth ventricle) receiving the prescribed dose.

5FU dose of 75 mg/kg was previously established in the preclinical study for EPN (8). However, this dose proved to be too toxic in our 1q+ PFA PDX models when combined with radiation, and we reduced the dose moving forward (Supplementary Fig. S1). 5FU (30 mg/kg) or equivalent PBS volume was administered once weekly via intraperitoneal injections. Mice receiving radiation in combination with 5FU were given one dose a day prior to radiation and one dose on the final day of irradiation. This was to simulate a human trial in which weekly 5FU injections are given during 52 weeks of fraction radiation. Mice were given a two-week recovery period following radiation before weekly cycles of 5FU or PBS were continued. Cycles were comprised of weekly 5FU for 6 weeks followed by a 2-week break and repeated until mice reached an endpoint. ATRA (40 mg/kg) or equivalent volume of corn oil was administered through oral gavage 5 days per week. ATRA treatments began after the two-week recovery from radiation and were given in six-week cycles with a two-week break to coincide with the 5FU schedule until mice reached the study endpoint. Three MRI scans were performed following radiation during rest periods to monitor tumor growth. Mice were provided moist chow or nutrigel (S5769, Bio-Serv) during chemotherapy cycles to support weight loss due to treatment. Mice were monitored every day and weights were documented twice weekly.

Survival was the primary outcome measurement for preclinical studies. Chemotherapy toxicities (for example, weight loss) and tumor volume were secondary outcome measurements. Endpoint criteria for this study were approved through IACUC 00784. When mice were moribund and/or had lost more than 15% of their initial body weight, they were euthanized. Mice were euthanized by CO₂ asphyxiation followed by neck dislocation by lab personnel. Following neck dislocation, brains were removed and preserved in 10% buffer formalin for 24 hours before being transferred to 70% ethanol. Fixed tumor was processed for immunohistochemistry analysis as previously described (12).

Survival analysis was performed using the Kaplan–Meier analysis. All data were pooled from replicate experiments and reported as the mean with standard deviation for error bars. Quantification of tumor burden was achieved by hand-free drawing of the region of interest (ROI on axial and sagittal T2w-MRI scans, by two analysts). The total of ROI areas on all affected slices was multiplied by slice thickness to achieve a 3D assessment of tumor burden, using Bruker ParaVison NEO v.2.2. software. Statistical significance was defined as P < 0.05 using the Student t test (RRID:SCR_002798, GraphPadv10).

The normal brain gene-expression data generated in this study are publicly available in Gene-Expression Omnibus (GEO, RRID: SCR_005012) at GSE 244124 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE244124). All other data generated in this study are available upon request from the corresponding author.

Our original work evaluating an FDA-approved drug screen panel identified a 5FU analogue, floxuridine, as the top PFA-selective inhibitor (5). However, 5FU, also a top PFA-selective compound in this study, was chosen as the agent to move forward based on existing pharmacokinetic and safety studies available in children (8, 9). Two pediatric PFA cell lines harboring 1q+/6q−, MAF-811, and MAF-928, were seeded in 96-well plates and treated with a dose curve of fractioned radiation (Fig. 1A). Immediately following radiation treatment, a range of 5FU doses for dose curve modeling was added to each plate. We measured cell proliferation by phase confluency (Fig. 1B) and caspase-3/7 cleavage (Supplementary Fig. S2A) every 4 hours using the IncuCyte live-cell imaging system. MAF-928 was much more sensitive to both 5FU and radiation alone with IC₅₀ values of 2.5 μmol/L and 2.5 Gy, respectively. MAF-811, which has a more differentiated expression profile (10), was sensitive to higher doses of 5FU and similar radiation doses with IC₅₀ values of 5 μmol/L and 2.5 Gy, respectively. These IC₅₀ doses are in the range of those found in previous published studies that identified 5FU as a candidate for EPN treatment and correlate with in situ pharmacokinetic–pharmacodynamic simulation model for 5FU dosing in pediatrics (6, 8). When 5FU was combined with radiation, we found significant growth impairment compared with single treatment alone (Fig. 1C; Supplementary Fig. S2B). This corresponded with greater loss of cell viability after 10 days, particularly in MAF-928 cells in which radiation alone had little effect on cell viability (Fig. 1D). We did not see a significant difference in cleaved caspase-3/7 over time (Supplementary Fig. S2A).

Gain of 1q in pediatric PFA results in trisomy chromosome 1q. Girish and colleagues recently reported that trisomy 1q mimics p53 deactivating mutations in part due to the third copy of MDM4, known to be a negative regulator of p53 (7). PFA do not harbor p53 mutations nor have any other known mutational drivers been identified in PFA. We investigated the expression of MDM4 across our PFA tumor bank to determine whether 1q+ PFA tumors had increased expression. Both 1q WT and 1q+ PFA tumors have higher MDM4 gene expression compared with normal pediatric brain (Fig. 2A). Consistent with Girish and colleagues’ findings, PFA tumors with a gain of chromosome 1q have significantly higher gene expression of MDM4 compared with WT PFA tumors (FPKM mean difference 4.885; 95% CI, 1.644–8.127, P = 0.0018; Fig. 2A). Interestingly, CDKN1A (p21) gene expression was significantly increased in 5FU-treated cell lines in a dose-response manner as measured by qRT-PCR (Fig. 2B). As CDKN1A is upregulated by p53 activity, this would suggest that 5FU might restore p53 activity in 1q+ PFA cell lines.

We, therefore, sought to measure p53 protein activity and other apoptosis pathway effectors using an apoptosis proteomic array in both MAF-811 and MAF-928 cell lines following 5FU treatment (Fig. 2C; Supplementary Fig. S3). In both cell lines, there was an increase in phosphorylated p53 on all three serine sites in response to 5FU treatment (Fig. 2C). Consistent with qRT-PCR data (Fig. 2B), we also observed an increase of p21 protein in 5FU-treated cells (Fig. 2C). We found p53 protein phosphorylation was further increased with radiation treatment (Fig. 2C). Although we did not see a significant change in the real-time imaging of cleaved caspase-3/7, we observed higher levels cleaved caspase-3 in the proteomic array analysis with radiation treatment alone and with combination 5FU and radiation (Fig. 2C). We performed MDM4 qRT-PCR on both cell lines following 5FU treatment and found MDM4 increased with 5FU (Supplementary Fig. S4). These data suggest that 5FU treatment, especially in combination with radiation, appears to overcome trisomy MDM4-mediated suppression of p53 activity, but this effect is independent of MDM4 gene expression. Although increased phosphorylation of p53 does not mean all p53 functions are restored, our findings do provide a partial mechanism for the significant decrease in proliferation and viability in combined 5FU and radiation-treated 1q+ PFA cell lines.

Another therapeutically relevant trisomy gene in 1q+ tumors identified by Girish and colleagues is Uridine-cytidine kinase 2 (UCK2; ref. 7). UCK2 is a rate-limiting protein that facilitates the incorporation of uracil during DNA replication, and in 1q+ tumors were shown to confer sensitivity to anticancer nucleotide analogues that include 5FU (7). We, therefore, hypothesized reduction of UCK2 would decrease 1q+ PFA cell lines’ sensitivity to 5FU. Consistent with the findings of Girish and colleagues, 1q+ PFA tumors have significantly higher UCK2 gene expression compared with WT 1q PFA tumors (FPKM mean difference 8.905; 95% CI, 5.245–12.56, P < 0.0001; Fig. 2A). We used qRT-PCR to compare UCK2 gene expression following the stable introduction of shRNA targeting UCK2 and compared these levels with two 1q WT PFA tumors (UPN 758 and UPN 911). Interestingly, while MAF-811 and MAF-928 are trisomies for UCK2, we found relative gene expression by qRT-PCR was 2-fold higher than both 1q WT PFA tumors (Fig. 2D). Both shRNA constructs were able to reduce UCK2 gene expression to levels of 1q WT PFA tumors (Fig. 2D). Following the selection of shUCK2 and shNT MAF-811 and MAF-928 with puromycin, we performed proliferation assays with the same dose curve of 5FU as previous experiments. In cells containing nontargeting control, we saw effects on growth with 5 μmol/L and higher of 5FU in MAF-811 and 2.5 μmol/L and higher of 5FU in MAF-928 (Fig. 2D). Knockdown of UCK2 gene expression reversed 5FU sensitivity with growth suppression starting to be seen only with the highest 5FU dose of 10 μmol/L in both MAF-811 and MAF-928 (Fig. 2D). This would suggest that pediatric patients with 1q+ PFA would be more sensitive to 5FU treatments given trisomy UCK2 levels.

The standard of care for pediatric EPN is maximal surgery followed by radiation treatment. We, therefore, evaluated the efficacy of 5FU treatment alone and in combination with fractionated radiation in vivo using our novel 1q+ PFA patient-derived xenograft models (12). 5FU was given twice during fractionated radiation to the PF in tumor-bearing mice (Fig. 3A). 5FU alone was not effective, and experiments without radiation were humanely discontinued (Fig. 3B). Overall survival in mice treated with 10 Gy radiation combined with 5FU was prolonged compared with radiation alone though these data did not reach significance (Fig. 3B). Tumor volume from MRI scans also showed a trend in response to 10 Gy + 5FU compared with radiation alone; more numbers would be needed to achieve significance. Immunohistochemistry analysis of tumor removed from radiation alone mice showed the occurrence of metastatic tumor lesions in the olfactory bulb and lateral ventricles (Supplementary Fig. S5).

Experiments in our second PDX model were significantly hindered by the COVID-19 pandemic. MAF-811_XC mice were only given one 6-week cycle of 5FU maintenance therapy, and this was not sufficient to prolong survival in combination with radiation (Supplementary Fig. S6).

Although we observed a significant decrease in tumor growth in vitro with 5FU in combination with radiation, we wanted to evaluate whether efficacy could be improved by adding a second chemotherapy agent. ATRA was identified as a top PFA-specific inhibitor from our original FDA-drug screen (ref. 5; Supplementary Fig. S7A). ATRA is a standard component of certain pediatric leukemia therapy regimens. To simulate using ATRA as maintenance chemotherapy, we treated MAF-811 and MAF-928 with 5FU and radiation as before, and 48 hours later treated with a single dose of ATRA (Fig. 4A). ATRA alone reduced cell proliferation and viability in both MAF-811 and MAF-928 (Fig. 4B). Combining 5FU and ATRA resulted in a similar loss of cell proliferation as 5FU alone though still significant compared with ATRA alone (Fig. 4B, top). We also see a similar pattern when ATRA was added following just radiation (Fig. 4B, middle). However, when ATRA was added following 2.5 Gy radiation and 5FU treatment, we observed a greater loss of proliferation for all doses 5FU (Fig. 4B, bottom). For MAF-811, we saw decreased cell viability after 10 days of cells treated with combination treatment; however, viability was relatively similar for MAF-928 (Supplementary Fig. S7B). This is likely due to the number of cells left after 10 days in MAF-928 combination treated wells (Fig. 4B). As we observed in previous experiments, there was no significant difference in real-time imaging of cleaved caspase-3/7 (Supplementary Fig. S7C).

We next evaluated the efficacy of adding an ATRA maintenance chemotherapy regime to our in vivo model. Following recovery from radiation treatment, mice were given both weekly 5FU and daily ATRA treatment for 6 weeks with a 2-week break (Fig. 5A). For ATRA alone, we did see a significant survival advantage over vehicle alone (Fig. 5B). However, there did not seem to be a benefit to adding 5FU with ATRA without radiation. When the combination of 5FU and ATRA was given following radiation treatment, there was a significant increase in mouse survival (95% CI, 1.868–14.87, P = 0.0017). Additionally, a decrease in tumor growth was evident with combination treatment and was nearly significant at 32 weeks following radiation (average tumor volume 10 Gy + vehicle = 14.55; average tumor volume 10 Gy + 5FU + ATRA = 4.8; 95% CI, −22.06 to 2.556, P = 0.055; Fig. 5C).

Immunohistochemistry analysis of tumors removed from mice postmortem shows significant tumor burden in radiation-alone mice (Fig. 5D). Ki67 counts for combination chemotherapy and radiation were significantly lower than with radiation alone (Fig. 5D, P < 0.05). However, there was no significant difference in caspase counts between treatment conditions (Fig. 5D).

We also performed these experiments with MAF-811 xenograft models (Supplementary Fig. S8). Again, these mice were only given a single 6-week cycle of chemotherapy. We did, however, observe significance in overall survival in combination 5FU + ATRA (Supplementary Fig. S8A). MAF-811 is a much slower-growing xenograft compared with MAF-928, which is likely the reason we observed a response with just chemotherapy alone. However, when combined with radiation, we did not see a survival advantage (Supplementary Fig. S8B).

Chromosome 1q gain (1q+) PFA has poor outcomes, and most patients experience recurrence or death within 2 years. The incidence of 1q+ at primary diagnosis in PFA is approximately 25%, and this number doubles at first recurrence (4). Time to disease recurrence and death is considerably shorter for children with 1q+ PFA than children with 1q WT PFA (4). There is an urgent need for therapy designed specifically for this high-risk pediatric population with 1q+ PFA both at diagnosis and recurrence.

Treatment advances in PF EPN have stalled due to a lack of relevant preclinical models. We were the first to publish in vitro and in vivo models of PFA with both models harboring a high-risk phenotype of 1q+ (10, 12). In this study, we utilized these models to test the efficacy of 1q+ specific chemotherapy strategy that provides a rationale for clinical testing. 5FU and ATRA were identified as top PFA cell line–specific inhibitors from an FDA-drug screen (5). Here we show that combining 5FU with radiation leads to significant loss in cell proliferation in vitro, which was further enhanced when ATRA was added as a maintenance chemotherapy approach. A limitation of the in vitro data is that true synergy was unable to be measured with the analyses performed. However, in vivo studies of combination radiation and 5FU and ATRA led to significant survival advantage compared with just radiation alone. This also corresponded to less tumor burden and decreased Ki67 counts on postmortem histology slides.

Although our 1q+ PFA xenograft models are important for the preclinical evaluation of any treatment strategy, there are still limitations to in vivo studies. Timing of treatment initiation in vivo was shown to affect efficacy, and we have found that, like the human patient disease, no therapy was effective when there was a significant tumor burden. The early deaths in the 5FU + radiation arm were not seen in the replicate cohort when we started treatment when tumor burden was small on MRI scans. Additionally, the COVID pandemic severely affected in vivo experiments done in MAF-811 with lockdown delaying the initiation of treatment and staff illness preventing the continuation of chemotherapy cycles.

5FU is a highly potent and widely available chemotherapy agent. Its use in pediatric EPN was first proposed in 2011 by Atkinson and colleagues in which a high-throughput in vitro and in vivo drug screen was performed to identify possible inhibitors (6). 5FU was the top inhibitor identified from their screen, which prompted a phase I study evaluating safety and dosage strategy for children with relapsed EPN (9). 5FU and other 5FU analogues were also identified in our drug screen as top PFA-selective inhibitors (5). More recently, a plausible mechanism for 5FU sensitivity was identified. Girish and colleagues found the gene UCK2, which is located on the long arm of chromosome 1q, is amplified in cancers that have a trisomy 1q phenotype (7). UCK2 is a DNA synthesis rate-limiting protein that facilitates the incorporation of uracil into newly synthesized DNA. Loss of one copy of UCK2 in trisomy 1q cell lines reversed the sensitivity of the trisomy cells to 3-Deazauridine, a UCK2-mediated nucleotide analogue (7). We found 1q+ PFA also had elevated levels of UCK2 compared with 1q WT PFA tumors. We found reducing the expression of UCK2 in our 1q+ PFA cells to levels seen in 1q WT PFA was sufficient to reduce the sensitivity of the cells to 5FU treatment, providing a mechanism for 5FU as a selective inhibitor in 1q+ PFA. A limitation of our in vitro and in vivo models is both 1q gain and 6q loss high-risk chromosomal abnormalities are present. We are therefore unable to determine the extent 6q loss has on the 5FU effect that was observed. However, the loss of sensitivity to 5FU with UCK2 knockdown would suggest that loss of 6q does not affect response to 5FU. There are no known models of PF EPN with chromosome 1q gain and WT chromosome 6q to validate findings or to determine whether ATRA sensitivity correlates with the loss of 6q.

There are no known mutational drivers in PFA. In a recent report, chromosome 1q+ behaves like a p53 mutation through trisomy expression of MDM4, a negative regulator of p53 downstream activation (7). We have shown 1q+ PFA tumors have significantly higher MDM4 gene expression compared with 1q WT PFA and normal brain. Interestingly, 5FU treatment, especially when combined with radiation, induced activation of the p53 signaling pathway. This finding suggests that 5FU treatment may abrogate the effect of trisomy MDM4 in 1q+ PFA tumors.

We, along with other researchers, have reported PFA tumors are a mixture of molecular subpopulations, including undifferentiated, stem-like cells (UEC), mesenchymal cells (MEC), and ependymally differentiated ciliated (CEC), and transportive EPN cells (TEC; refs. 13, 14). We hypothesized identifying an inhibitor that would target differentiated UEC toward less aggressive CEC and TEC cells would enhance the effect of 5FU and radiation on PFA cell lines. From our FDA-drug screen, ATRA was identified as the second most potent PFA-selective FDA-approved oncology drug (5). Retinoids are potent differentiation agents, and it is therefore plausible that ATRA is working by differentiating UEC cells toward low-risk ependymally differentiated CEC and TEC subpopulations. Further studies need to be completed to determine the mechanism by which ATRA is working in PFA.

Both 5FU and ATRA are agents that are commonly used in pediatric oncology, making them readily translatable into combination clinical trials for high-risk EPN. Based on this work, this multimodal therapy is most promising for the highest-risk EPN patients, specifically those children with PFA tumors that harbor 1q gain. Multiple consortia in the United States and Europe are planning to develop clinical trials with this combination therapy, as there is an urgent need to improve clinical outcomes for this patient population.

J. Steiner reports grants from NIH during the conduct of the study. N.J. Serkova reports grants from NIH during the conduct of the study. H.B. Lindsay reports personal fees from Day1 outside the submitted work. No disclosures were reported by the other authors.

A.M. Griesinger: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration. A.J. Calzadilla: Data curation, validation, writing–review and editing. E. Grimaldo: Data curation, writing–review and editing. A.M. Donson: Conceptualization, data curation, methodology, writing–review and editing. V. Amani: Conceptualization, data curation, writing–review and editing. A.M. Pierce: Data curation, methodology, writing–review and editing. J. Steiner: Data curation, visualization, writing–review and editing. S. Kargar: Data curation, formal analysis, methodology. N.J. Serkova: Formal analysis, visualization, methodology, writing–review and editing. K.C. Bertrand: Conceptualization, methodology, writing–review and editing. K.D. Wright: Conceptualization, methodology, writing–review and editing. R. Vibhakar: Conceptualization, supervision, methodology. T. Hankinson: Conceptualization, resources. M. Handler: Conceptualization, resources. H.B. Lindsay: Conceptualization, methodology, writing–review and editing. N.K. Foreman: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing. K. Dorris: Conceptualization, supervision, writing–original draft, project administration.

The genomics core is funded through the University of Colorado Cancer Center (P30CA046934). All MRI scans were acquired at the Animal Imaging Shared Resources and by P30CA046934 and the NIH high-end shared instrumentation grant (S10OD023485). Animal studies were funded by The Tanner Seebaum Foundation and Department of Defense (CA190494). This study was supported by NIH/NCATS Colorado CTSA Grant Number UM1 TR004399. Contents are the authors’ sole responsibility and do not necessarily represent official NIH views. The authors would like to thank The Morgan Adams Foundation for their continued support. The authors would like to thank the veterinarians, animal care team, and vet technicians at University of Colorado Anschutz Vivarium.