Targeted delivery of napabucasin with radiotherapy improves outcomes in diffuse midline glioma

Abstract

Background

Diffuse midline glioma (DMG) is the most aggressive primary brain tumor in children. All previous studies examining the role of systemic agents have failed to demonstrate a survival benefit; the only standard of care is radiation therapy (RT). Successful implementation of radiosensitization strategies in DMG remains an essential and promising avenue of investigation. We explore the use of Napabucasin, an NAD(P)H quinone dehydrogenase 1 (NQO1)-bioactivatable reactive oxygen species (ROS)-inducer, as a potential therapeutic radiosensitizer in DMG.

Methods

In this study, we conduct in vitro and in vivo assays using patient-derived DMG cultures to elucidate the mechanism of action of Napabucasin and its radiosensitizing properties. As penetration of systemic therapy through the blood-brain barrier (BBB) is a significant limitation to the success of DMG therapies, we explore focused ultrasound (FUS) and convection-enhanced delivery (CED) to overcome the BBB and maximize therapeutic efficacy.

Results

Napabucasin is a potent ROS-inducer and radiosensitizer in DMG, and treatment-mediated ROS production and cytotoxicity are dependent on NQO1. In subcutaneous xenograft models, combination therapy with RT improves local control. After optimizing targeted drug delivery using CED in an orthotopic mouse model, we establish the novel feasibility and survival benefit of CED of Napabucasin concurrent with RT.

Conclusions

As nearly all DMG patients will receive RT as part of their treatment course, our validation of the efficacy of radiosensitizing therapy using CED to prolong survival in DMG opens the door for exciting novel studies of alternative radiosensitization strategies in this devastating disease while overcoming limitations of the BBB.

Key Points.

• • Identified a potent reactive oxygen species-inducer as a therapeutic radiosensitizer in pediatric diffuse midline glioma.

• • Confirmed the efficacy and survival benefit of targeted delivery of Napabucasin using convection-enhanced delivery (CED) with radiation therapy (RT).

• • The first study to show the feasibility and efficacy of CED concurrent with RT in the brainstem.

Importance of the Study.

In this manuscript, we explore the potency of Napabucasin in diffuse midline glioma (DMG) and its robust radiosensitizing effects. Although exciting, one of the major obstacles in the treatment of DMG is the intact blood-brain barrier (BBB), which limits the ability to provide clinical benefit. We utilize our expertise in targeted drug delivery technology to explore (FUS)-mediated BBB disruption as well as convection-enhanced delivery (CED) to circumvent the BBB. This is the first study to show the feasibility of CED concurrent with radiation therapy (RT) in the brainstem, which is paramount to advancing care in those with DMG given RT is required for nearly every patient. Furthermore, this publication serves as a proof-of-concept for our unique drug discovery and delivery pipeline to bring promising drug candidates forward in synergy with the standard of care RT in DMG, agnostic to BBB permeability.

Diffuse midline glioma (DMG) is the most aggressive primary brain tumor in children, carrying a median overall survival (OS) of less than 1 year.¹ Among pediatric high-grade gliomas, DMG encompasses a group of aggressive tumors arising in midline structures of the brain including the midbrain, brainstem, and spinal cord. In the most recent World Health Organization (WHO) classification of central nervous system (CNS) tumors, DMGs are defined by molecular alterations leading to common downstream epigenetic dysregulation (H3 K27-altered).² In contrast to other high-grade gliomas for which surgery may be feasible, the location of DMG in midline structures makes surgical resection unsafe.³ Studies examining the role of alternative systemic therapies and/or the addition of radiosensitizers have failed to demonstrate a survival benefit to date.⁴ This is in part due to the heterogeneous nature of DMGs, making them quite challenging to treat. DMGs are intrinsic with an extensive microscopic infiltrative component extending to areas of the brain where the blood-brain barrier (BBB) remains intact; this prevents the permeation of many systemic therapies into the brain tissue.⁵ Thus, therapeutic approaches to targeting DMG must consider both the agent’s mechanism of action and ways to maximize drug delivery to the tumor.

The mainstay of treatment for DMG is conventionally fractionated radiotherapy (RT) delivered over 6 weeks,⁶ though this appears only to provide transient relief of symptoms while offering limited survival advantage. Since RT remains the cornerstone of therapy in DMG care, the identification of a therapeutic strategy that works safely and in synergy with RT will be paramount to providing optimal benefit in this devastating disease. NAD(P)H quinone dehydrogenase 1 (NQO1) is one of the downstream targets of nuclear factor erythroid 2-related factor 2 (NRF2), a well-known regulator of an array of downstream enzymes with important detoxifying and antioxidant functions that are activated upon exposure to oxidative stresses, such as RT. The NQO1 enzyme facilitates the 2-electron reduction of quinone to hydroquinone, protecting cells against the oxidative damage of quinones.⁷ Certain agents (known as NQO1 bioactivatable reactive oxygen species [ROS]-inducing agents); however, upon reduction via NQO1, form an unstable hydroquinone that spontaneously regenerates to the parent compound in a 2-step oxygenation process. This futile recycling produces elevated ROS concentrations that lead to substantial DNA damage and cell death.⁷ Interestingly, preclinical data reveals an RT-mediated increase in NRF2 and downstream NQO1 at the transcriptional and translational levels,^8,9 which begs the question as to whether RT can prime or enhance the efficacy of these agents.

Napabucasin (BBI608) is a small molecule of naphthoquinone under clinical investigation in various cancer types, including bladder and gastrointestinal malignancies.^10–12 With its known favorable toxicity profile and well-understood pharmacokinetics in humans,¹³ there is an efficient avenue for clinical translation. In pancreatic cancer cells, Napabucasin is bioactivated by NQO1, resulting in futile redox cycling and ROS generation, leading to cell death.¹⁴ We, therefore, tested the efficacy of Napabucasin in DMG both in vitro and in vivo, and hypothesized that the cytotoxic effects of Napabucasin would be further augmented in the presence of RT. We also explore the use of focused ultrasound (FUS) and convection-enhanced drug delivery (CED) to overcome the limitations of the BBB and maximize therapeutic benefit.

Methods

The Full Text of the Methods, Including Sources of Reagents and Equipment

Cell culture.—

Patient-derived cell lines were a kind gift from Dr. Michelle Monje’s lab at Stanford University School of Medicine (DIPG36, DIPG17, DIPG6, and SF8628). The murine DMG spheres expressing H3.3K27M (DIPG-KAPP and DIPG4423) were kind gifts from Dr. Nada Jabado’s lab at McGill University and Dr. Oren Becher’s lab at Icahn School of Medicine at Mount Sinai, respectively, Supplementary Methods.

DMG mouse models.—

All animal experiments were performed under Columbia University Institutional Animal Care and Use Committee–approved protocols and complied with the ethical regulations and humane endpoint criteria according to the NIH Guide for the Care and Use of Laboratory Animals. For the subcutaneous xenograft model, DIPG36 cells were injected into the flank of 6-week-old athymic nude mice. For the subcutaneous syngeneic DMG flank model, DIPG-KAPP cells were injected into the flank of 6-week-old mice (B6[Cg]-Tyrc-2J/J, Jackson). When tumors reached 50 mm³, mice were randomized to vehicle, drug monotherapy with Napabucasin 10 mg/kg via intraperitoneal (i.p.) injection daily for 5 consecutive days, RT monotherapy with 2 Gy daily for 5 consecutive days using the Small Animal Radiation Research Platform (SARRP), or combination therapy with 2 Gy daily and concurrent Napabucasin 10 mg/kg daily for 5 consecutive days.

For the orthotopic syngeneic DMG mouse model, DIPG-KAPP cells were injected into the right pons as previously described.¹⁵ After confirming successful implantation, mice were randomized and treated as above for 5 consecutive days with either vehicle, 10 mg/kg Napabucasin via i.p. injection, 2 Gy radiation, or combination therapy.

Generation of stable cell lines.—

Generation of Cas9-expressing cell lines and NQO1-depleted cells was performed via infection with lentivirus packaged using LentiV-Cas9-puro vector (a kind gift from Dr. Christopher R. Vakoc at Cold Spring Harbor Laboratory) and puromycin selection as previously described. NQO1 sgRNAs and non-targeting controls were cloned into LRG2.1T lentiviral vector (gift from Dr. Christopher R. Vakoc at Cold Spring Harbor Laboratory). Virus package and infection were performed as previous described.^16,17

Cell viability assays.—

To determine cell viability after treatment with Napabucasin, cells were seeded into 96-well plates containing either DMSO or serial dilutions of Napabucasin. After treatment for 72 hours, CellTiter-Blue reagent was added, and fluorescence was measured by GloMax explorer.

Colony formation assays.—

Cells were seeded at low density into 6-well plates coated with Matrigel. Twenty-four hours later, cells were irradiated using the MultiRad 350 (Precision X-Ray) at increasing doses from 0 to 8 Gy. Immediately following, cells were treated with either DMSO or 0.2 µM Napabucasin. Fourteen days later, colonies were stained with 0.1% crystal violet.

ROS quantification.—

DMG cells were treated with either vehicle control, 1 µM Napabucasin, 4 Gy of irradiation, or combination treatment with 4 Gy and 1 µM Napabucasin. Twenty-four hours later, cells were stained using DCFDA/H2DCFDA Cellular ROS Assay Kit (Abcam, 113851). Flow cytometry was immediately performed using an Attune NxT Flow Cytometer.

RNA-seq analysis.—

For bulk RNA-seq analysis, datasets from 76 patient tissue samples (all H3K27M mutant) were obtained from Suzy Baker, Alan Mackay,¹⁸ and Selin Jessa.¹⁹ Counts for the 76 patient samples and 246 normal caudate brain tissue samples from Genotype-Tissue Expression (GTEx) portal²⁰ were normalized and adjusted for batch effect using ComBat-seq function.²¹ For single-cell RNA-seq, raw FASTQ files for 2 DMG tumor-bearing mice (DIPG4423) and one normal mouse brainstem were obtained from Fernandez et al. (bioRxiv 2003;2024[2024]:2017–585370).

Small animal magnetic resonance imaging.—

Ten days after tumor implantation in the brainstem, magnetic resonance imaging [MRI] was performed to confirm tumor formation and once weekly thereafter to monitor tumor volume. A Bruker 94/20 Magnetic Resonance Imager was used as previously described (Fernández et al. [bioRxiv 2003;2024(2024):2017–585370]).^15,22 All DICOM images were exported and analyzed using the 3D Slicer image computing platform.

In vivo ultrasound imaging.—

For the subcutaneous xenograft model, 4 weeks after tumor implantation, weekly ultrasound imaging was conducted using Visulasonics VEVO 3100 High-Frequency Ultrasound Imaging System. Tumor volumes were calculated using Vevo LAB software.

Cell and small animal irradiation.—

Mice were irradiated using the Small Animal Radiation Research Platform (SARRP) as previously described.²³ For the radiation of cells and for use in mice when the SARRP was unavailable, a MultiRad350 (Precision X-Ray) was used at a dose rate of 1.22 Gy/min.

Convection-enhanced drug delivery.—

On day 10 after mouse intracranial tumor implantation, CED pumps were implanted into the pons as previously described.^24,25 Briefly, 80 µM Napabucasin was dissolved in PBS and loaded into a 7-day ALZET micro-osmotic pump. The pump was connected to the Alzet Brain Infusion Kit 2. The infusion cannula was inserted 4.7 mm deep into the skull surface and secured onto the skull using adhesive glue. The drug was delivered at a rate of 0.5 µL/h for 7 days. At that time, pumps were removed as per the manufacturer’s recommendation.

Focused ultrasound.—

FUS was conducted using a single-element, spherical-segment FUS transducer driven by a function generator (Agilent) through a 50-dB power amplifier (E&I) as previously described.¹⁵

Results

NQO1 Is Expressed in DMG Patient Tissue and Tumor-Derived Cell Cultures

Recent studies have reported that NQO1 expression is upregulated in several solid malignancies including cervical, breast, and liver, among others.^26–28 A correlation has been described between high levels of NQO1 expression and increased radioresistance in glioma mouse models, and NQO1 upregulation has unsurprisingly been associated with poorer outcomes in glioblastoma (GBM) patients.^29–31 The role of NQO1 in DMG, however, is not well-elucidated. We first analyzed bulk RNA-sequencing (RNA-seq) data from 76 DMG patient tissue samples.¹⁹ When comparing gene expression from DMG patient tissue to normal brain tissue (caudate) from GTEx, NQO1 was found to be expressed at higher levels (mean log₁₀TPM 1.40 and 1.15 in DMG versus normal brain, respectively; Figure 1A). We also analyzed single-cell RNA-seq data obtained 4 weeks after stereotactic implantation of mouse DMG tumor cells (DIPG4423) into the pons, and found NQO1 expression to be significantly higher in DMG cells when compared to normal mouse brainstem (Figure 1B). We then confirmed the presence of NQO1 protein expression in a majority of our DMG tumor-derived cell cultures using Western blot analysis and observed lower expression in normal mouse brain tissue (Figure 1C). Thus, NQO1 expression in DMG offers a promising avenue of efficacy for NQO1-bioactivatable ROS-generating agent therapy.

Figure 1.

NQO1 is expressed in DMG patient tissue and tumor-derived cell cultures. (A) Box and Whisker plot of bulk RNA-sequencing data from 76 DMG patient tissue samples showing NQO1 expression in log₁₀ (transcripts per million [TPM]) compared to normal brain tissue from 246 normal caudate tissue samples from GTEx (1.40 and 1.15 in DMG and normal brain tissue, respectively, ***= P < .001). (B) Violin plot with single cell normalized expression of NQO1 generated using Seurat SCTransform from single cell RNA-sequencing data from two syngeneic DMG mouse model tumors (combined) as compared to normal mouse brainstem (***= P < .001). (C) Western blot analysis of NQO1 and α-Tubulin in several tumor-derived cell cultures, mouse brain tissue lysate, and human embryonic kidney (293T cells). Error bars represent median +/− IQR. P value for A was calculated using a Wilcoxon test, comparing log₁₀TPM after removing batch effects. The P value for B was calculated using a Wilcoxon test comparing normalized gene expression between DMG tumor versus normal mouse brainstem in Seurat with P-values corrected by the Bonferroni method to adjust for multiple comparisons across all genes.

Napabucasin Is a Potent ROS-Inducer in DMG

To determine the potency of Napabucasin in vitro, 4 patient-derived DMG cell cultures (DIPG36, DIPG6, DIPG17, and SF8628) were treated with increasing concentrations of Napabucasin for 72 hours. Calculated IC50 values for DIPG36, DIPG6, DIPG17, and SF8628 were 0.80, 1.05, 0.84, and 1.44 µM, respectively (Figure 2A). Using the PLATE-Seq microfluidic automation platform,²² we generated RNA-seq profiles in DIPG6 at 24 hours following treatment with several drugs, including Napabucasin. Upon treatment with Napabucasin at its highest sublethal concentration (48 hours EC₂₀) for 24 hours, gene ontology (GO) analysis of the top 200 upregulated genes prioritized those involved in ROS production as third based on adjusted P-value (Figure 2B). Enrichment analysis using VIPER-inferred protein activity³² also revealed significant upregulation of ROS pathways with Napabucasin compared to vehicle control (Figure 2C). These results are consistent with the hypothesis that Napabucasin induces ROS production in DMG.

Figure 2.

Napabucasin is a potent ROS-inducer in DMG. (A) Dose–response curves for cell viability of DMG cell cultures, as measured by MTT ([3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide]) using CellTiter-Blue after 72-hour exposure to Napabucasin (N = 3 for each cell line). (B) GO analysis of the top 200 upregulated genes after 24 hours of Napabucasin treatment in DIPG6 in vitro (top 10 pathways shown based on adjusted P-value). (C) Pathway enrichment analysis using VIPER-inferred protein activity (top 10 pathways shown based on adjusted P-value). (D) Graph of DCFDA assay showing relative ROS levels (compared to vehicle) after treatment with 1 or 2 µM Napabucasin for 6 hours (N = 3 per group, *= P < .05) in DIPG-KAPP and DIPG36. Error bars represent mean +/− SD, P values for (D) were calculated using a two-tailed Student’s t-test.

Next, using a cell-permeant reagent, 2ʹ,7ʹ-dichlorofluorescin diacetate (DCFDA), we quantitatively assessed ROS production in DMG cells treated with Napabucasin. We treated DIPG-KAPP and DIPG36 cells with increasing doses of Napabucasin (1 and 2 µM) and observed a dose-dependent increase in ROS production (Figure 2D). Taken together, these findings validate Napabucasin as a potent agent in vitro, leading to significant ROS production and activation of downstream oxidative response pathways.

Napabucasin Acts as a Radiosensitizer

It is well-established that RT involves the induction of DNA damage directly by ionizing radiation or indirectly by the generation of ROS.³³ This process activates downstream antioxidant enzymes, including superoxide dismutase, glutathione reductase, as well as NQO1. NQO1 serves to detoxify quinones and maintain intracellular antioxidants.^27,34–36 To recapitulate RT-induced NQO1 expression, we treated DIPG-KAPP and DIPG6 cells with a single dose of 5 and 10 Gy of RT using the MultiRad 350 (Precision X-Ray). Twenty-four hours later, cells were harvested for western blot analysis and RT-qPCR. Treatment with 5 Gy in DIPG-KAPP led to a significant increase in NQO1 gene expression when normalized to GAPDH (Figure 3A), which was confirmed on western blot quantification (Figure 3B). In DIPG6, significant changes at the gene and protein expression levels were noted after 10 Gy of treatment, but not after 5 Gy, suggesting the presence of variation in dose–response to radiation across cell lines.

Figure 3.

Napabucasin is a radiosensitizer in vitro. (A) Effects of RT on NQO1 expression detected by RT-qPCR 24 hours post-RT (N = 3 per group, *= P < .05). (B) Western blot analysis of NQO1 and quantification of band intensities (relative to control α-Tubulin) from cell lysate 24 hours post-RT (*= P < .05). (C). Clonogenic survival assay using colony formation in DIPG-KAPP and DIPG36 cells. Representative 6 well plates after treatment of DIPG36 with 4 Gy of RT in the presence and absence of 0.2 µM Napabucasin (left). On right, plots show surviving fraction at 14 days after increasing doses of RT in the presence and absence of 0.2 µM Napabucasin for DIPG-KAPP and DIPG36 cells. (D) Bar graph showing normalized ROS levels detected by DCFDA assay in DIPG36 (left) and DIPG-KAPP (right) after treatment with 1 µM Napabucasin for 24 hours, a single dose of 4 Gy RT, combination treatment, or no treatment (control; N = 3 per group). Error bars represent mean +/− SD. P values for A, B, and D were calculated using a two-tailed Student’s t-test comparing mean values between each treatment group. For C, P values were calculated comparing differences in fit using nonlinear regression with the linear quadratic cell death function.

We next hypothesized that treatment with Napabucasin, an NQO1-bioactivatable ROS-inducer, would have enhanced efficacy in the setting of RT and serve as a radiosensitizing modality. To evaluate radiosensitization, we conducted clonogenic assays in DIPG36 and DIPG-KAPP cells. Cells were treated with a single dose of RT ranging from 0 to 8 Gy in the presence or absence of Napabucasin. Fourteen days later, colonies were fixed and counted for each condition and normalized to non-irradiated conditions to generate surviving fractions (Figure 3C). Nonlinear regression using the linear quadratic cell death function for both lines shows significant differences in fit for both lines. Data were then imported to SynergyFinder Plus to assess for synergy.³⁷ Summary bliss synergy score was calculated at 16.7 and 17.36 for DIPG36 and DIPG-KAPP, respectively. The summary synergy scores can be interpreted as the average excess response due to drug interactions (ie, a synergy score of 15 corresponds to 15% of response beyond expectation).³⁸

We then conducted the DCFDA assay to quantify ROS levels using combination treatment. DIPG36 and DIPG-KAPP cells were treated with 1 µM Napabucasin or vehicle with and without a single dose of 4 Gy of RT. Twenty-four hours later, flow cytometry was performed with the bar graph representing the relative ROS levels compared to non-treated controls. Cells treated with Napabucasin and RT monotherapy had increased ROS production, with a significant increase in ROS production when the therapies were combined (Figure 3D). Based on these findings, the in vitro radiosensitizing effects of DMG are apparent and likely due to enhanced combinatorial ROS-induction.

Napabucasin-Mediated Effects Are Dependent on NQO1

We next sought to determine if the downstream cytotoxicity, ROS production, and radiosensitizing properties generated by Napabucasin are indeed mediated by NQO1. NQO1 was successfully depleted in Cas9-expressing patient-derived cell line DIPG36 using 2 independently designed sgRNAs (NQO1-1 and NQO1-2; Figure 4A). NQO1-depletion led to a 2.47-fold and 2.36-fold increase in IC50 for sgNQO1-1 and sgNQO1-2, respectively, compared to cells infected with lentivirus expressing sgRNA targeting the ROSA26 locus control (sgNeg), supporting the role of NQO1 in Napabucasin-mediated cell death (Figure 4B).

Figure 4.

The effects of Napabucasin are dependent on NQO1. (A) Western blot analysis of NQO1 and control α-Tubulin after Cas9-expressing DIPG36 cells were infected with lentivirus expressing sgNQO1-1, sgNQO1-2, and sgRNA targeting the ROSA26 locus control (sgNeg). (B) Dose–response curves for cell viability as measured by MTT ([3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide]) using CellTiter-Blue after 72-hour exposure to Napabucasin (N = 3 for each cell line). IC50 values increase by more than 2-fold with NQO1 depletion. (C) Effects of Napabucasin on gene expression of 3 downstream oxidative response genes; NQO1 KO and WT DIPG36 treated with 1 µM Napabucasin or vehicle control for 6 hours followed by RT-qPCR. The bar graph shows the relative change in gene expression, normalized to Actin and vehicle (N = 3 per group, *= P < .05). (D) Bar graph showing normalized ROS levels detected by DCFDA assay after treatment of NQO1-depleted DIPG36 cells with 1 µM Napabucasin for 24 hours, a single dose of 4 Gy RT, or combination treatment (top). Bar graph showing comparison in normalized ROS levels detected by DCFDA assay between WT and NQO1-depleted cells (bottom; N = 3 per group, * = P < .05). (E) Clonogenic survival assay using colony formation in NQO1 KO and WT DIPG36 cells. Plots show surviving fraction at 14 days after increasing doses of RT in the presence and absence of 0.2 µM Napabucasin for each cell line. Error bars represent mean +/− SD. P values for C and D were calculated using a two-tailed Student’s t-test comparing mean values between each treatment group. For E, P values were calculated comparing differences in fit using nonlinear regression with the linear quadratic cell death function.

Based on the apparent dependency, at least in part, of NQO1 on Napabucasin cytotoxicity, we sought to confirm that NQO1-mediated bioactivation of Napabucasin indeed triggers ROS production and downstream effects. Sulfiredoxin-1 (SRXN1), Heme oxygenase-1 (HMOX1), and Glutamate-cysteine ligase modifier subunit (GCLM) are well-established downstream genes that are upregulated in response to oxidative stress.^39–41 We treated DIPG36 wild-type (WT) and NQO1-depleted (KO) cells with 1 µM Napabucasin for 6 hours. Using RT-qPCR, Napabucasin treatment led to a marked increase in expression of these oxidative stress response genes (increased SRXN1, HMOX1, and GCLM expression by 3.4-, 23.7-, and 4.0-fold, respectively), and NQO1 depletion significantly abrogated this transcriptional effect (Figure 4C).

We then repeated the DCFDA assay to quantify ROS production in NQO1 KO cells; although RT monotherapy induced a significant increase in fluorescence, treatment with Napabucasin did not significantly increase ROS production (Figure 4D). To evaluate changes in radiosensitizing properties, clonogenic survival assays were performed in both NQO1-depleted cells (sgNQO1-1 and sgNQO1-2) and WT cells (sgNeg). Surviving fractions at increasing doses of RT in the presence and absence of 0.2 µM Napabucasin were plotted. Nonlinear regression using the linear quadratic cell death function confirmed the loss of significant differences in fit in both NQO1-depleted cells compared to WT cells (Figure 4E). Representative images from the assay can be found in Supplementary Figure S2. Taken together, Napabucasin and RT independently induce ROS production, and there appears to be an additive effect when combining treatments. Furthermore, Napabucasin-mediated ROS production and radiosensitization are dependent on NQO1.

Napabucasin Has a Radiosensitizing Effect in Subcutaneous Xenograft Models

Next, we moved to study whether the in vitro cytotoxicity of Napabucasin could be recapitulated in vivo. As the BBB may selectively inhibit localized drug delivery, we first used an in vivo subcutaneous xenograft mouse model to avoid the potential need to circumvent the BBB. Previous studies show successful antitumor effects with a wide range of intraperitoneal doses of Napabucasin,⁴² ranging from 10 to 40 mg/kg.^12,43

To determine the maximum tolerated dose (MTD), we treated 3 healthy 6-week-old mice with 5, 10, 15, and 20 mg/kg Napabucasin via i.p. injection daily for 1 week. Mice were then observed and weighed daily for 14 days to assess for signs of toxicity. Doses up to 10 mg/kg were tolerated. 15 and 20 mg/kg led to significant morbidity and mortality with significant weight loss due to gastrointestinal (GI) toxicity (diarrhea; Supplementary Figure S3). Of note, in several clinical trials using Napabucasin, GI adverse events frequently led to dose hold, modification, or drug discontinuation.^11,44,45 As such, we determined the MTD as 10 mg/kg, and this dose was used for subsequent in vivo studies.

Six-week-old athymic nude mice were subcutaneously injected with DIPG36 cells to create a subcutaneous DMG xenograft mouse model. Additionally, mice with intact immune systems were injected subcutaneously with mouse DIPG-KAPP cells. When flank tumors met the threshold for treatment initiation at a volume of 50 mm³, mice were randomized to 4 treatment groups: (1) vehicle control (10% DMSO), (2) Napabucasin 10 mg/kg via i.p. injection for 5 consecutive days, (3) daily RT to a dose of 2 Gy for 5 consecutive days, or (4) combination RT and Napabucasin treatment. After the 5-day treatment, tumor volume was monitored weekly using ultrasound imaging. In both models, Napabucasin monotherapy using this dose regimen did not significantly reduce tumor volume compared to vehicle control. RT alone had significantly better tumor control compared to both the vehicle and Napabucasin-only groups. The combination treatment of RT and Napabucasin had statistically significant improved local tumor control starting at 4 and 8 weeks post-treatment initiation for DIPG-KAPP and DIPG36, respectively (Figure 5A-D). Therefore, without consideration of the BBB, Napabucasin is ineffective as monotherapy in vivo, but does provide a local control benefit in the presence of RT.

Figure 5.

Napabucasin has a radiosensitizing effect in subcutaneous xenograft models. After flank injection with DMG cells, treatment was initiated when tumors reached 50 mm³. Mice received either vehicle, Napabucasin monotherapy (10 mg/kg i.p. injection daily for 5 days), 2 Gy RT daily for 5 days, or combination therapy. Plots showing fold-change in tumor volume weekly starting from week 1 (2 days following completion of a treatment regimen) in DIPG-KAPP (A) and DIPG36 (B) (N = 6 mice per condition). Error bars represent mean +/− SD, * = P < .05 when comparing RT monotherapy to combination therapy at specified timepoint. (C) Representative ultrasound image of flank tumor showing contoured gross tumor volume. (D) Representative image of nude mice with DIPG36 flank tumors at week 8. P values for A and B were calculated using a two-tailed Student’s t-test comparing mean values between RT monotherapy and combination therapy groups.

Targeted Drug Delivery of Napabucasin Using CED in Combination With RT Prolongs Overall Survival in Orthotopic DMG Mouse Models

We then moved to a syngeneic orthotopic DMG mouse model, testing the effects of these therapies with consideration of BBB penetrability. We stereotactically implanted DIPG-KAPP cells into the right pons, and MRI images were obtained 10 days post-implantation to confirm tumor formation (experiment schema can be found in Supplementary Figure S4). Mice were then randomized to the same 4 groups as above. The addition of Napabucasin monotherapy did not prolong survival in our model, like the pattern observed in the subcutaneous xenograft models. Furthermore, Napabucasin did not provide additional benefits when added to RT (Supplementary Figure S5). Given the relatively low MTD of Napabucasin via i.p. injection as compared to other published experiences,^12,43 we hypothesized this discrepancy between the subcutaneous and orthotopic models may be due to inadequate intratumoral drug concentration, especially since Napabucasin monotherapy did not produce a robust effect in subcutaneous xenograft models, where the BBB was not present as an additional layer of selectivity.

To further characterize the distribution of Napabucasin in the mouse model, we quantified serum and brainstem drug concentrations using LC/MS after a single i.p injection of 15 mg/kg Napabucasin. Although the MTD was 10 mg/kg when administering multiple doses, we previously confirmed the safety of up to 15 mg/kg for a single-dose experiment. A recent phase I study on the pharmacokinetics of a single oral dose of radiolabeled Napabucasin showed peak serum concentration of Napabucasin at 2.75 hours, with a half-life of 7.92 hours.¹³ As i.p. administration of pharmacological agents results in faster and more complete absorption compared to oral administration,⁴⁶ we quantified serum and brainstem concentration of Napabucasin at several intervals over 90 minutes. We found serum concentration peaked at 30 minutes post-injection, with peak brainstem concentrations at around 60 minutes (Supplementary Figure S6). Thus, the 60-minute timepoint was chosen for further quantification studies. 10 days post-tumor implantation, an MRI was conducted to confirm the tumor location. Mice were then treated with 15 mg/kg Napabucasin via i.p. injection. At 60 minutes post-injection, plasma and brainstem tissue were harvested for Napabucasin quantification. Mean serum and brainstem concentrations of Napabucasin were 538 and 11.8 ng/mL, respectively (Supplementary Figure S7).

To test the hypothesis that targeted drug delivery to the tumor would enhance therapeutic efficacy, we utilized FUS to noninvasively and reversibly disrupt the BBB. FUS uses non-ionizing acoustic waves similar to diagnostic ultrasound, and studies have shown that optimization of FUS delivery with the combined use of microbubbles can achieve local and reversible BBB-opening and increase drug delivery in multiple preclinical animal models, including DMG mouse models.^15,47,48 Our group has demonstrated successful BBB opening and enhanced drug delivery of multiple agents including etoposide¹⁵ and anti-PD1.⁴⁹ We also showed FUS-mediated BBB opening is immediate, and the BBB remains open for approximately 72 hours after a single treatment session.¹⁵ After confirming successful DMG tumor formation 10 days post-injection using MRI, we performed a single 2-minute 4-point FUS session. Immediately after treatment, mice were treated with 15 mg/kg Napabucasin via i.p. injection. At 60 minutes post-injection, plasma and brainstem tissue were harvested for Napabucasin quantification. Mean serum concentration was 498 and 538 ng/mL in the FUS and no FUS groups, respectively (P = ns). Mean brainstem concentration in the FUS group was 7.7 and 12.1 ng/mL in the no FUS group (P = ns, N = 3 matched serum and brainstem samples per group; Figure 6A). As such, FUS does not appear to enhance the brainstem delivery of Napabucasin using these established parameters.

Figure 6.

Targeted delivery of Napabucasin using CED in combination with RT is feasible in orthotopic DMG mouse models and prolongs overall survival. (A) Bar graph showing mean brainstem concentration of Napabucasin’s most common in vivo metabolite (dihydro‐napabucasin [M1]) after a single 2-minute FUS session with microbubbles or a 4-day osmotic pump infusion of 80 µM Napabucasin directly into the brainstem (CED; N = 3 matched serum and brainstem samples per group). Napabucasin concentration in the brainstem was 625, 7.7, and 12.1 ng/mL after CED, FUS, and no BBB disruption/bypass, respectively. * = P < .05. (B) Representative T2-weighted MRI images at day 9 post-implantation confirming T2 edema (left). After pump implantation, CT images were obtained (middle) and overlayed with the T2-weighted MRI images taken prior to implant to confirm correct pump placement. (C) Schematic overview of steps and timing for stereotactic implantation of DIPG-KAPP cells and subsequent CED pump placement (created with BioRender). (D) Representative T2-weighted MRI images over time in each experimental group (D = days post-tumor implantation; top). The bar graph below shows the mean change in tumor volume at day 24 post-implantation (N = 6 mice per group). (E) Kaplan–Meier curve showing overall survival of DMG orthotopic model with targeted drug delivery using CED pumps. Combination treatment with Napabucasin and RT had the longest survival benefit, with median survival 46 versus 33 days (P < .05), 26 days (P < .05), and 29 days (P < .05) in RT only, Napabucasin only, and vehicle groups, respectively (N = 9 mice per group). Error bars represent mean +/− SD, P values for A and D derived using 2-tailed Student’s t-test and P values in E derived by log-rank test (Mantel-Cox test),

We then turned our attention to BBB circumvention using a convection-enhanced drug delivery (CED) system. CED involves the use of catheters that are surgically implanted into the brain tumor and surrounding tissue.^25,50 CED operates via pressure gradient differentials that allow a homogeneous distribution of molecular agents in the target tissue. CED into the murine brainstem has been proven safe and feasible in prior preclinical studies.^51,52 Six-week-old mice were stereotactically injected with DIPG-KAPP cells into the right pons. Seven-day ALZET infusion pumps were then implanted containing 100% PBS to first determine the feasibility of this technique. After the 7-day infusion, pumps were removed, and mice were monitored daily for signs and symptoms of neurologic toxicity and weight loss greater than 20% as surrogate endpoints for OS. We defined mortality a priori as operative-related if mice met criteria for death during or within 5 days after pump placement; 3 of the 15 (20%) mice experienced operative-related mortality. No morbidity or mortality was noted during or after surgery for pump removal (Supplementary Figure S8). To visualize delivery and distribution with this technique, additional mice were selected to receive 100 µL gadodiamide (GE Healthcare) diluted in PBS to a final concentration of 2.87 mg/mL via the infusion pump. After pump implantation, CT images were obtained and overlayed with the T2-weighted MRI image taken prior to implant to confirm correct pump placement (Figure 6B). Seven days after pump implantation, pumps were removed, and a T1 post-contrast weighted MRI sequence was obtained to confirm the presence of gadolinium contrast enhancement (Supplementary Figure S9).

Upon confirmation of the feasibility of this technique in the mouse brainstem, a dose escalation study was conducted to determine the MTD of Napabucasin via CED. Ten days after stereotactic intracranial injection of DIPG-KAPP cells, 7-day ALZET osmotic pumps were placed containing increasing concentrations of Napabucasin from 0 to 100 µM (3 mice per condition). After pump implantation, mice were monitored closely for signs and symptoms of neurological toxicity and weighed daily. Napabucasin was tolerated up to a dose of 80 µM (Supplementary Figure S10) and was therefore used for the following experiments.

Using the MTD of 80 µM, we then used LC/MS to confirm localized drug delivery to the brain. Ten-day post-tumor implantation in the right pons using DIPG-KAPP cells, a 4-day ALZET infusion pump was implanted containing 80 µM Napabucasin. On day 5 post-pump placement, serum and right brainstem samples were collected. The mean brainstem concentration of the most common in vivo metabolite of Napabucasin (dihydro‐napabucasin [M1]) was 625 ng/mL, significantly elevated when compared to brainstem samples of mice after receiving FUS or no BBB modulation (P < .05; Figure 6A).

For formal survival analysis using Napabucasin via CED and RT, DIPG-KAPP cells were stereotactically injected into the right pons of 6-week-old mice. MRI was conducted 9 days post-injection to confirm tumor formation. Mice then underwent a second surgery on day 10 to implant a 7-day ALZET micro-osmotic pump using the same burr hole used during tumor implantation (schematic in Figure 6C). Mice were randomized to 4 groups with 100 µL reservoirs placed in subcutaneous pocket between the shoulder blades of each mouse containing either: (1) 80 µM Napabucasin, or (2) vehicle. One day following, mice underwent CT imaging using a Quantum FX microCT imaging system. CT images were fused with T2-weighted MRI images from the prior day to confirm placement of the pump within 1–2 mm of T2 enhancement. If the pump was not within 2 mm of the T2 enhancement on MRI, the mouse was removed from the study. Mice were then randomized to receive either 2 Gy RT daily to the entire brainstem for 5 consecutive days (total of 10 Gy), or no RT. After 5 days of RT, mice had one rest day, and pumps were removed on day 7 (from the date of initial pump placement). Mice were then monitored with daily weights, and weekly T2-weighted MRI images were acquired to measure tumor growth over time (Figure 6D).

For evaluating the safety of combination therapy with CED and RT, morbidity–mortality was defined as treatment-related if mice met criteria for death during the 5-day RT treatment course or within 48 hours after completion of RT. Of the 20 mice receiving CED concurrent with RT (10 with Napabucasin + 10 with vehicle), 2 mice (1 from each treatment arm) met the criteria for death during concurrent RT (10% mortality). After accounting for these toxicities, Napabucasin monotherapy did not prolong survival (median OS 29 and 26 days in Napabucasin monotherapy and vehicle control groups, respectively, P = ns). RT monotherapy produced a significant survival benefit compared to vehicle and Napabucasin monotherapy groups (median OS 33 days, P < .05), but combination treatment with Napabucasin and RT had the largest survival benefit compared to all other treatment groups (median OS 46 days, P < .05; Figure 6E). Thus, CED of Napabucasin in combination with RT is feasible and prolongs overall survival in DMG mouse models.

Discussion

Despite ongoing preclinical and clinical research efforts, the prognosis for DMG patients remains dismal. Penetration of systemic therapy through the BBB is a concern in several CNS malignancies, particularly in DMG where the BBB remains relatively intact.⁵³ As we were only able to quantify extremely low levels of Napabucasin in mouse DMG tumors at several time points after administration of the drug at the MTD via i.p. injection, we suspected inadequate intratumoral drug concentration was responsible for the lack of synergy observed with RT in our orthotopic model. Excitingly, in our preclinical DMG mouse model, we were able to successfully confirm the efficacy of Napabucasin-mediated CED. Furthermore, to the best of our knowledge, this is the first study to assess the role of concurrent RT with CED in the brainstem. Although the safety of CED concurrent with RT in the brain cortex has been reported,⁵⁴ to date, its feasibility in the brainstem had not yet been published prior to this study. In our study, we show the addition of concurrent RT to CED leads to a relatively small increase in operative-related morbidity–mortality of approximately 10%. It is important to highlight that surgical techniques in animal models differ extensively from those of humans, demonstrating the importance of early phase I clinical studies to translate these findings to clinical practice. Regardless, the relative safety of concurrent RT and CED in mouse DMG is an essential step forward, as nearly all DMG patients will receive upfront RT as a standard of care treatment. In our preclinical mouse model, CED not only appears feasible with the addition of concurrent RT, but we validate the efficacy of radiosensitizing therapy using CED with Napabucasin to prolong survival in DMG mouse models, which opens the door for novel studies of alternative radiosensitization strategies in this devastating disease while overcoming limitations of the BBB.

Our current interdisciplinary team of scientists, clinicians, and biomedical engineers each plays a crucial role in providing a hopeful future for patients with DMG. In this study, we demonstrate the sequential steps of our drug development and delivery pipeline (Supplementary Figure S12), removing BBB permeability as a restricting factor when considering potential therapeutic targets. Using a DMG regulatory network from 122 publicly available RNA-seq profiles, our collaborators created RNA-seq profiles following perturbation with ~300 oncology drugs and used this to identify agents that invert patient master regulatory protein activity profiles using the New York State Department of Health-approved OncoTreat algorithm⁵⁵; Napabucasin was one of the top predicted target agents for DMG sensitivity,^56,57 aiding in its selection as proof-of-concept for our drug delivery optimization pipeline.

Several methods to circumvent the BBB are under active investigation, including non-invasive methods such as FUS, lipid nanoparticle technology, virus-mediated delivery, and chemical modifications of systemic agents.⁵⁸ Our group has shown feasibility and safety of BBB opening using FUS in DMG mouse models,¹⁵ which has led to the opening of two phase I clinical trials at our institution assessing the safety and feasibility of FUS with concurrent Etoposide (NCT05762419) and Panobinostat (NCT04804709). Although we did not see increased drug delivery of Napabucasin with FUS using our well-established parameters,¹⁵ there are several reasons that may explain this finding. One such reason is that small molecule inhibitors such as Napabucasin do not have a specific receptor-ligand interaction. These molecules may be washed out from the brain parenchyma at a faster rate than anticipated in the setting of FUS-induced vasodilation, requiring finer adjustments to the timing and frequency of FUS sessions.⁵⁹ Pharmacokinetic studies in our laboratory are ongoing to further elucidate the patterns of brain uptake and clearance of Napabucasin and other small molecules to help establish a more optimal time point for FUS with systemic drug administration. Regardless, several preclinical studies in DMG mouse models from our group and others show increased intratumoral drug concentration with FUS,⁶⁰ and recent phase I/II studies demonstrated that repeat BBB-opening using an implantable ultrasound device with concurrent cytotoxic chemotherapy improved targeted drug delivery and survival in patients with recurrent GBM.^61–63 As such, continuing to explore FUS in our drug delivery pipeline will be essential moving forward as new therapeutic targets are tested.

Using a more invasive technique such as CED, we were able to directly bypass the BBB and increase the delivery of Napabucasin. CED has been safely performed in preclinical mouse DMG models,⁶⁴ and multiple clinical phase I studies have confirmed the safety of CED in the pons with aqueous Panobinostat⁶⁵ and radiolabeled antibody [¹²⁴I]-8H9.^65,66 Our group has also recently completed accrual for a phase I trial testing MTX-110 (a water-soluble Panobinostat nanoparticle formulation) in newly diagnosed DMG (NCT04264143). Going forward, as RT in clinical DMG can last upwards of 6 weeks, it will be important to understand the feasibility of prolonged infusion via CED to maximize radiosensitizing effects. Of note, in our institution’s phase I feasibility study of CED in DMG, the device was permanently implanted, with 2 48-hour pulse infusions of MTX110 given using a wireless clinical programmer separated by 7 days.⁶⁷ As such, there is a clinical path forward for prolonged drug infusion using CED in the DMG space.

Indeed, overexpression of NQO1 has been a documented poor prognostic factor in multiple solid malignancies including gastric, hepatocellular, and cervical cancers.^26,68,69 These preclinical findings have led to multiple clinical trials in advanced and metastatic GI malignancies using Napabucasin given its mechanism of action as a ROS generator bioactivated by NQO1.^14,70 Early phase I data show a tolerable adverse event profile.⁷¹ Unfortunately, subsequent randomized studies did not show a clinical benefit in survival.^11,45 Importantly, these trials did not assess the ability of Napabucasin to act as a radiosensitizer, as these patients did not receive concurrent RT on trial. Given RT’s ability to form ROS and induce NQO1 expression, we hypothesized these 2 therapies may act synergistically. We excitingly showed radiosensitizing properties in vitro and saw the addition of Napabucasin to a course of targeted RT leads to improved local tumor control in vivo. It is important to highlight the variability in NQO1 expression across DMG cell lines based on western blot analysis. Although the calculated IC50 values across lines were quite similar (around 1 µM), further experiments are needed to identify whether NQO1 expression may be a prognostic marker or biomarker for response to Napabucasin and/or RT. We must also acknowledge the little to no effect of Napabucasin monotherapy when using either orthotopic or subcutaneous xenograft models. Unlike in cell culture, animal models have a tumor microenvironment that may lead to compensatory mechanisms for ROS production induced by Napabucasin. Furthermore, the in vivo active M1 metabolite has been shown to have 12.57-fold less activity than the parent compound,¹³ which likely plays a role in the lack of in vivo efficacy with monotherapy. It is possible that RT has a multi-modal benefit in vivo, inducing further ROS production and increasing NQO1 expression, allowing Napabucasin to exert its therapeutic benefit. It is also interesting to note the kinetics of tumor growth and treatment response in our 2 subcutaneous xenograft models appeared quite different. As one mouse model was immunocompetent (DIPG-KAPP) and the other immunocompromised (DIPG36), studies of the effects of RT and Napabucasin therapy on the tumor microenvironment are ongoing.

With our findings that CED of Napabucasin concurrent with RT prolongs survival in DMG, we help move the needle forward in our mission to bring promising drug candidates forward in synergy with the standard of care RT, agnostic to BBB permeability. We plan to use this drug discovery and delivery platform to pave the way for a brighter, hopeful future for patients and families suffering from DMG, with a tangible avenue for clinical translation at our fingertips.

Supplementary material

Supplementary material is available online at Neuro-Oncology (https://academic.oup.com/neuro-oncology).

Conflict of interest statement

Dr. Califano is founder, equity holder, and consultant of DarwinHealth Inc., a company that has licensed some of the algorithms used in this manuscript from Columbia University. Columbia University is also an equity holder in DarwinHealth Inc. US patent number 10,790,040 has been awarded related to this work, and has been assigned to Columbia University with Dr. Califano as an inventor. Lorenzo Tomassoni is an employee of DarwinHealth Inc.

Funding

American Society of Clinical Oncology (ASCO) Young Investigator Award, Radiological Society of North America (RSNA) Resident Research Grant, and Columbia Cancer Training Program for Resident-Investigators (CAPRI) Grant to MG. NCI Outstanding Investigator Award (R35 CA197745) and 2 NIH Shared Instrumentation Grants (S10 OD012351 and S1 0OD021764) to AC. Postgraduate studies in North America and the Asia-Pacific region Fellowship by “La Caixa Foundation” to ECF. Hope and Heroes, Fegel Family Foundation, Sebastian Strong Foundation, St. Baldrick Foundation, Swim Across America, Matheson Foundation, and Focused Ultrasound Foundation funding to CCW. Hyundai Hope on Wheels Hope Scholar Award, Swim Across America, Rally Foundation, StacheStrong and Musella Foundation to RDG. Two NIH grants (R01 NS132344-01 and R01 CA277605-01A1) to ZZ.

Author contributions

Conceptualization: M.G., Z.Z., C.C.W., J.P., R.W.R., S.K.C., and X.Z. Methodology: M.G., C.C., P.C., J.B., Y.W., L.T., G.S., and S.Z. Investigation: M.G., E.C.F., N.Y., D.K., J.C.A., Y.W., E.W., C.J.K., X.B., F.Z., P.C., and J.B. Visualization: M.G., S.D., E.C.F., E.W., L.T., G.S., and S.Z. Funding acquisition: M.G., C.C.W., Z.Z., A.C., and J.P. Project administration: C.C.W., Z.Z., A.C., J.P., R.W.R., J.G., and S.K.C. Supervision: C.C.W., Z.Z., A.C., J.P., R.W.R., J.G., S.K.C., and S.Z. Writing—original draft: M.G., X.Z., G.D., H.W., and C.C.W. Writing—review & editing: All authors

Data availability

All data that support the findings of this study are available upon request from the corresponding author. All supporting data and pre-processing/analysis code are available to editors and peer reviewers upon request at any time.