Abstract
Purpose: Deregulated MYC drives oncogenesis in many tissues yet direct pharmacologic inhibition has proven difficult. MYC coordinately regulates polyamine homeostasis as these essential cations support MYC functions, and drugs that antagonize polyamine sufficiency have synthetic-lethal interactions with MYC. Neuroblastoma is a lethal tumor in which the MYC homologue MYCN, and ODC1, the rate-limiting enzyme in polyamine synthesis, are frequently deregulated so we tested optimized polyamine depletion regimens for activity against neuroblastoma.
Experimental Design: We used complementary transgenic and xenograft-bearing neuroblastoma models to assess polyamine antagonists. We investigated difluoromethylornithine (DFMO; an inhibitor of Odc, the rate-limiting enzyme in polyamine synthesis), SAM486 (an inhibitor of Amd1, the second rate-limiting enzyme), and celecoxib (an inducer of Sat1 and polyamine catabolism) in both the preemptive setting and in the treatment of established tumors. In vitro assays were performed to identify mechanisms of activity.
Results: An optimized polyamine antagonist regimen using DFMO and SAM486 to inhibit both rate-limiting enzymes in polyamine synthesis potently blocked neuroblastoma initiation in transgenic mice, underscoring the requirement for polyamines in MYC-driven oncogenesis. Furthermore, the combination of DFMO with celecoxib was found to be highly active, alone, and combined with numerous chemotherapy regimens, in regressing established tumors in both models, including tumors harboring highest risk genetic lesions such as MYCN amplification, ALK mutation, and TP53 mutation with multidrug resistance.
Conclusions: Given the broad preclinical activity demonstrated by polyamine antagonist regimens across diverse in vivo models, clinical investigation of such approaches in neuroblastoma and potentially other MYC-driven tumors is warranted. Clin Cancer Res; 22(17); 4391–404. ©2016 AACR.
This article is featured in Highlights of This Issue, p. 4275
Translational Relevance
Hyperactive MYC signaling is an oncogenic driver for a large proportion of human tumors, yet to date no MYC-directed therapeutic has been approved for clinical use. MYC drives myriad signaling pathways that link cell-cycle kinetics with the creation of cell biomass. Polyamines are cationic chaperones that support such MYC activities through ionic and covalent mechanisms, and their homeostasis is critical to both initiating and maintaining the cancer phenotype. Here, we demonstrate in complementary models of the highly lethal childhood tumor neuroblastoma that therapeutics that deplete tumoral polyamines (such as DFMO, celecoxib, and SAM486) synergize to block tumor initiation and regress established tumors. DFMO and celecoxib are FDA-approved drugs with activity in regressing diverse high-risk neuroblastoma subtypes in vivo, alone and in combination with diverse chemotherapy regimens. These data strongly support the testing of such approaches in the clinic.
Introduction
MYC genes coordinate transcriptional programs to promote cell proliferation, biomass production, self-renewal, and numerous other oncogenic attributes (1). Not surprisingly, MYC activity is tightly regulated in normal cells, while deregulated expression is frequent in human cancers (2). Interest in pharmacologically antagonizing Myc is high as it acts at a network node governing growth signals and such antagonism might have broad clinical utility (3). However, Myc functions through protein:protein interactions generating a family of heterodimeric transcription factors with competing activities across thousands of sites in the genome (4). Although direct antagonism of these complexes remains intractable, an alternative approach to antagonize Myc is to inhibit downstream pathways necessary for tumorigenesis. Polyamines represent a family of essential polycations that support Myc functions through ionic and covalent activities. Reduced levels of intracellular polyamines activate checkpoints that constrain proliferation, as seen in senescent and post-mitotic cells, while enhanced polyamine synthesis accompanies oncogenic proliferation (5). Because Myc regulates numerous polyamine enzymes, this pathway has attracted attention as a therapeutic target in cancer (6), although specific testing for synthetic-lethal activity in tumors with deregulated MYC has been lacking.
The ornithine decarboxylase gene, ODC1, is a direct Myc target and bona fide oncogene that can substitute for MYC to transform cells in vitro (7) and in vivo (8). ODC1 encodes the rate-limiting enzyme in polyamine synthesis that decarboxylates ornithine to putrescine (Supplementary Fig. S1). Stepwise conversion to the higher-order polyamines spermidine and spermine occurs via the aminopropyltransferases, SRM and SMS, respectively. The aminopropyl donor for these conversions is derived from the activity of the second rate-limiting enzyme in polyamine synthesis, S-adenosylmethionine decarboxylase (encoded by AMD1). Both Odc1 and Amd1 have the shortest half-lives of any mammalian enzymes (10–30 minutes). Odc activity is further regulated posttranslationally by Odc antizymes that mediate its degradation (9), which are themselves regulated by antizyme inhibitors (10). Catabolism is regulated through polyamine acetylation by the spermidine/spermine-N-acetyltransferase, SAT1, which controls flux through the pathway (11), whereas spermine can also be oxidized via SMOX activity to spermidine. Acetylated polyamines can be exported from the cell or back-converted by polyamine oxidase (PAOX) to spermidine or putrescine, respectively. Finally, an energy-dependent polyamine transport system can import polyamines from the microenvironment. That homeostatic control over the repertoire of polyamines is so highly regulated at the transcriptional, translational, and posttranslational level (9, 12–14) underscores their importance to cell function and provides numerous opportunities for therapeutic intervention (Supplementary Fig. S1).
Neuroblastoma is a lethal childhood tumor in which MYCN is the principal oncogenic driver (15). Its deregulation by genomic amplification correlates with aggressive disease and poor outcome (16). Gene expression profiles also identify strong MYC signatures in poor outcome neuroblastomas without MYCN amplification suggesting this may be a requisite pathway for the high-risk phenotype (17). Because MYC extensively regulates the polyamine pathway, we and others have sought a role for polyamines in supporting neuroblastoma initiation and progression (18, 19). We previously used a transgenic mouse model of neuroblastoma to show that difluoromethylornithine (DFMO, Eflornithine), an irreversible Odc1 inhibitor and FDA-approved drug for the treatment of Trypanosomiasis, delayed tumor initiation in a preemptive therapy model and synergized with chemotherapy to extend survival of mice with established tumors (18). Indeed, DFMO at doses up to 3 g/m2/day were recently shown to be tolerable in children with neuroblastoma, and correlative studies supported depletion of systemic polyamine levels at this exposure (20). Here, we sought to augment the efficacy of DFMO through synergistic targeting of multiple steps in polyamine homeostasis, and to extend these findings to complementary neuroblastoma xenograft models. We show that combined inhibition of Odc1 and Amd1 using DFMO and SAM486 prior to tumor initiation profoundly reduces tumor penetrance in TH-MYCN neuroblastoma-prone mice, validating this pathway as downstream of MYC. In the treatment of established tumors, we show that the combination of DFMO and celecoxib, which induces Sat1, provides synergistic antitumor activity across models harboring highest risk genomic lesions such as MYCN amplification, ALK mutation, and TP53 mutation with multidrug resistance. Furthermore, we show augmented antitumor activity with multiple chemotherapy backbones including combinations widely used in neuroblastoma clinical protocols. Such broad preclinical data are required to prioritize agents for clinical investigation, as those with more limited preclinical testing often fail to demonstrate activity in human trials (21). Our findings support clinical testing of polyamine antagonist regimens in children with neuroblastoma, and potentially additional MYC-driven malignancies in which polyamine homeostasis is pivotal.
Materials and Methods
Cell lines
Human neuroblastoma cell lines with MYCN amplification (IMR5, NLF, SMS-KAN, BE2C) and without (SK-N-SH and SK-N-AS) were obtained from Garrett Brodeur (Children's Hospital of Philadelphia, PA). All are identity-confirmed by our group every 6 months using short tandem repeat (STR)-based genotyping (AmpFISTR, Applied Biosciences) and matched to the COG cell line database (www.cogcell.org). Murine neuroblastoma cell lines 844+/+ and 282+/− were established from tumors arising in a homozygous and hemizygous TH-MYCN mouse, respectively. All cells were grown in RPMI1640 media (Life Technologies) supplemented with 10% FBS, 2 mmol/L l-glutamine, 100 U/mL of penicillin, and 100 mcg/mL gentamicin. Tissue culture was at 37°C in a humidified atmosphere of 5% CO2. Polyamines were added to the culture media where indicated at 1 μmol/L concentration each for putrescine, spermidine, and spermine (Sigma-Aldrich).
Colony formation assay
Neuroblastoma cells were exposed to DFMO at 5 mmol/L for 72 hours, with or without supplemental polyamines (1 μmol/L each of putrescine, spermidine, and spermine), in routine tissue culture media. Thereafter, 1,000 cells were plated in triplicate wells and stained by crystal violet after 14 days. For assessing drug interactions, BE2C cells were plated at 500 cells/well and SK-N-SH cells at 1,000 cells/well, and 5 hours later, cells were exposed to DFMO, celecoxib, or mafosfamide for 72 hours. Starting concentrations for BE2C cells were: 110 μmol/L DFMO, 14.8 μmol/L celecoxib, and 0.51 μmol/L mafosfamide; and for SK-N-SH cells were: 40 μmol/L DFMO, 12 μmol/L celecoxib, and 0.14 μmol/L mafosfamide. Drugs were kept at a constant ratio for three additional concentrations increasing by 1.5-fold. Colonies were fixed and stained after 10 days for BE2C and 14 days for SK-N-SH using 0.5% crystal violet/50% methanol, and counted using Quantity One 1-D Analysis software (Bio-Rad). Each experiment was performed in triplicate. Drug combination effects were determined using the CalcuSyn software (Biosoft), which generates a combination index (CI) as a measure of the combined drug interaction. CI values were generated over a range of fractional cell kill (Fa) levels from 0.05 to 0.90 (5%–90% growth inhibition), and a Fa level of 0.75 was chosen to reflect the higher levels of growth inhibition needed when studying anticancer agents (22).
TH-MYCN mouse model
129×1/SvJ mice transgenic for the TH-MYCN construct (23) were originally provided by Bill Weiss (Department of Neurology, University of California, San Francisco, San Francisco, CA). All murine studies were approved by the Institutional Animal Care and Utilization Committee at The Children's Hospital of Philadelphia (Philadelphia, PA) and the Animal Care and Ethics Committee of the University of New South Wales (Kensington, Sydney). TH-MYCN hemizygous mice were bred and litters randomized to therapy as previously (18). Mice were genotyped from 1-cm tail-snip–isolated DNA using qPCR (18). Mice were screened by experienced animal personnel and sacrificed for pathologic signs of tumor burden. All mouse experiments were performed with a contemporaneous control arm. Transcriptome analyses were performed using available microarray datasets (ref. 24; GEO GSE17740) from the same TH-MYCN colony used herein.
Preemptive therapy trials.
Preemptive DFMO therapy consisted of 1% DFMO added to water ad libitum from birth through day 70 [added to maternal drinking water days 0–21; after day 21 wean, added to pup drinking water; ref. 25]. SAM486 (Sardomozide, Novartis) therapy was given as 5 mg/kg i.p. 3 days per week from day 21 through day 70. Celecoxib was given daily by gavage at 100 mg/kg beginning day 21. In all studies, vehicle was provided by similar route (intraperitoneal or gavage) to control animals. Palpation for tumors was performed thrice weekly. Animals with tumors underwent serial abdominal ultrasonography under isoflurane sedation to determine in situ tumor volume (Vevo660; VisualSonics). Mice without evidence of tumor were sacrificed at day >120 and a necropsy was performed to assess for the occult presence of tumor. No chemotherapy was used in preemptive trials and outcomes were analyzed by genotype (TH-MYCN +/+ or +/−) with the time to tumor detection (tumor-free survival) and time to sacrifice due to signs of tumor progression (survival) as endpoints.
Established tumor trials.
TH-MYCN mice were randomized at the time palpable tumors (∼75–175 mm3) were identified, to regimens including chemotherapy with or without polyamine antagonists: DFMO 1% added to water ad libitum; SAM486 at 5 mg/kg i.p. 3 days per week for up to 6 weeks; celecoxib at 100 mg/kg daily by gavage; started at enrolment. For the SAM486 + DFMO trial, both TH-MYCN hemizygous and homozygous mice were included. For all other trials, only TH-MYCN+/+ mice were included as they have less variable tumor progression kinetics (26). Chemotherapy regimens were temozolomide at 50 mg/kg i.p. (+/+ mice) or 10 mg/kg (+/− mice) for 5 days, irinotecan at 20 mg/kg i.p. for 5 days, topotecan at 2 mg/kg i.p. for 5 days, or topotecan at 0.5 mg/kg i.p. and cyclophosphamide at 10 mg/kg i.p., or cyclophosphamide alone at 20 mg/kg for 5 days, as described. For all studies, vehicle was provided by identical route and regimen to control animals. Animals were followed until sacrifice due to tumor progression (survival).
Murine xenograft models
Cells used for xenograft studies were pathogen free by IMPACT I PCR screening (RADIL, University of Missouri). Xenografts were established in the flank of NCR or BALB/c nu/nu athymic mice (Jackson Laboratories; ref. 27). When tumor volume exceeded 200 mm3, mice were randomly assigned to therapy arms. DFMO was given as 1% DFMO in drinking water ad libitum, SAM486 at 5 mg/kg i.p. three times per week for 2 weeks, and celecoxib was 100 mg/kg daily gavage for 5 days each week. Chemotherapy regimens included cyclophosphamide (30 mg/kg i.p. day 1 and 3) with topotecan (0.25 mg/kg i.p. day 1 and 3), and cyclophosphamide alone (30 mg/kg i.p. day 1 to 3; 90 mg/kg total). Animals were sacrificed when tumor volumes exceeded 2,000 mm3 or if symptoms of tumor progression were apparent and this endpoint used for survival analyses.
Tumor histopathology
Tissues were harvested at sacrifice and fixed in 10% neutral buffered formalin and paraffin embedded for histologic studies, and flash frozen in liquid nitrogen for metabolic assays. Sections were hematoxylin and eosin (H&E) stained and assessed for differentiation, necrosis, and mitotic/karyorrhectic cells. For IHC, 5 μm sections were deparaffinized, hydrated, and treated with appropriate antibodies. Caspase-3 antibody (R&D Systems AF835) and Ki-67 (Santa Cruz Biotechnology SC-7846) staining were performed on an Autostainer Plus (DAKOCytomation). For caspase-3, slides were incubated with caspase-3 antibody at a 1:1,000 dilution for 30 minutes at room temperature, rinsed, and then incubated with biotinylated anti-Rabbit IgG (Vector Laboratories) for 30 minutes. For Ki-67, slides were incubated with Ki-67 antibody at a 1:1,000 dilution overnight at 4°C, rinsed, then incubated with biotinylated anti-Goat IgG (Vector Laboratories) for 30 minutes. After rinsing, slides were incubated with avidin–biotin complex (Vector Laboratories) for 30 minutes, followed by rinsing and incubation with DAB (DAKO Cytomation) for 10 minutes at room temperature, then rinsed and counterstained with hematoxylin. Both caspase-3 and Ki-67 staining was scored as the percentage of stained tumor nuclei.
Polyamine content assays
Tumor tissues harvested at necropsy were flash-frozen in liquid nitrogen, ground to a fine powder, and stored at −80°C. For cell line assays, subconfluent cells were cultured with or without additional polyamines added to the culture media (1 μmol/L each putrescine, spermidine, spermine; Sigma) as indicated, spun and rinsed three times and pelleted and frozen. For polyamine analyses, ground tissues were homogenized in 0.2 N perchloric acid and incubated at 4°C overnight. Dansylated polyamines were separated on a reversed-phase C18 HPLC column (28). Polyamine values were normalized to the amount of DNA in the tissue extracts.
Radiolabeled spermidine transport assays
Radioactive spermidine (Net-522 Spermidine Trihydrochloride, [Terminal Methylenes-3H(n)]) from PerkinElmer was used (specific activity 16.6 Ci/mmol). Cells were plated and grown to approximately 70% confluence. Three plates were used per condition: control, DFMO (5 mmol/L), SAM486 (5 μmol/L), celecoxib (10 μmol/L), or combinations; treated for 48 hours. After repeated washing with PBS, 3H-SPD was added at 1 μmol/L and incubated for 60 minutes at 37°C. Specimens were washed with cold PBS containing 5 μmol/L spermidine and incubated in 500 μL of 0.1 N NaOH at 37°C for 30 minutes to allow cells to dissolve. Contents were sonicated for 10 seconds to ensure uniform dissolution of cell debris. Then, an equal volume of 0.1 N HCL was added to neutralize pH. Specimens were then aliquoted for scintillation counting. Of note, 50 μL of sample was then aliquoted for protein using mini-Bio-Rad procedure. Results were expressed as CPM/400 μL, then protein μg/400 μL was calculated, and finally CPM/μg was calculated.
Statistical analyses
RNA expression levels between tumor and normal tissues were compared using the Wilcoxon exact test. Tumor-free survival and survival analyses (defined above) were performed according to the method of Kaplan and Meier (29) with SEs according to Peto and Peto (30). Comparisons of outcome between subgroups were performed by a two-sided log-rank test. Neuroblastoma differentiation status was compared using the Fisher exact test, and synergy studies were assessed as defined above. All other statistical comparisons were performed using the Student t test for independent sample sets and two-tailed design.
Results
Polyamine homeostasis is faithfully recapitulated in the TH-MYCN neuroblastoma model
Human neuroblastomas with MYC activation coordinately deregulate polyamine enzymes to promote polyamine sufficiency and support Myc functions (18). We therefore tested whether similar polyamine deregulation accompanies tumor progression in the TH-MYCN mouse model. TH-MYCN mice have neural crest-targeted MycN and develop tumors comparable with human neuroblastoma at the genetic (31), histologic (32), and ultrastructural level (23). The model is highly aggressive with 100% tumor penetrance in transgene homozygous (TH-MYCN +/+) mice with lethality by 7 weeks of age. Penetrance is approximately 40% with extended latency in hemizygous (TH-MYCN +/−) mice (26).
Gene expression profiles were derived from nascent tumors through large invasive tumors (24) and compared with nontumor cervical ganglia from both wild-type and TH-MYCN+/+ mice (Fig. 1). Like human neuroblastomas, the polyamine pathway was coordinately deregulated to expand polyamine capacity. ODC1 was upregulated while its antizyme, OAZ2, was downregulated. Prosynthetic spermidine and spermine synthases (SRM and SMS) were also upregulated. Conversely, catabolic SAT1 and spermine oxidase, SMOX, were downregulated, whereas polyamine oxidase, PAOX, which cleaves acetylated polyamines to restore spermidine and/or putrescine was elevated. This pattern closely recapitulates the transcriptional changes found in human neuroblastomas and credentials the TH-MYCN model for preclinical polyamine antagonist studies (Supplementary Table S1).
Coordinate deregulation of polyamine enzymes accompanies tumor progression in the TH-MYCN model. Polyamine gene expression levels for non-tumorous cervical ganglia (CG) from wild-type mice (white bars, n = 3) and TH-MYCN+/+ mice [gray bars, n = 9]; compared with expression from neuroblastomas (NB) arising in TH-MYCN+/+ mice (black bars, n = 6 per group) at four stages of progression. NB-S, NB-M, NB-L, and NB-UL represent small, medium, large, and ultra-large neuroblastomas, respectively, as defined in ref. (24); *, Wilcoxon exact P values <0.05 in comparison with the wild-type CG. SD error bars are shown. WT, wild-type.
Polyamine antagonists collaborate to block neuroblastoma initiation
We previously used the TH-MYCN model to show that preemptive DFMO therapy delayed tumor onset in homozygous mice; however, tumor penetrance remained complete (18). Tumors harvested from DFMO-treated mice had reduced putrescine levels reflecting Odc inhibition (on-target DFMO activity); however, spermidine and spermine were not reduced. We hypothesized that augmented Amd1 activity induced following DFMO exposure in vitro (33) and in vivo (19) might contribute to the rescue of Odc-inhibited cells. Because transcriptional changes were seen at the earliest stage of tumorigenesis, and because polyamine antagonists have chemopreventive activity in individuals at-risk for colon cancer (34), we reasoned that polyamines were required for tumor initiation. We tested the effect of preemptively inhibiting both Odc1 and Amd1 on tumor penetrance and latency in this model.
SAM486 (Sardomozide, Novartis) is an Amd1 inhibitor that demonstrates variable cytotoxicity against human neuroblastoma cells and lesser activity against murine TH-MYCN tumor cells (Supplementary Fig. S2). We randomized TH-MYCN mice to receive SAM486 alone or with DFMO. SAM486 was initiated at day 21 when mice were of sufficient size for serial intraperitoneal injections, relatively late with respect to tumor onset around day 28 (26, 35). Still, treatment with SAM486 reduced tumor penetrance in homozygous mice (P = 0.02; Fig. 2 and Supplementary Table S2). Notably, the combination of DFMO and SAM486 extended tumor latency (P < 0.001) and reduced penetrance to approximately 60% in homozygous mice (n = 67; P < 0.01; Fig. 2A and B) and from 66% to 17% in hemizygous mice (n = 213; P < 0.001; data not shown). Although therapy was stopped at day 70, tumors were rarely detected after this time point. Tumors that arose under SAM486 (P = 0.06), or SAM486 + DFMO (P < 0.05), showed neural differentiation, an effect typical of cytotoxic chemotherapy (36) that was absent from control tumors (Fig. 2C). Tumors did not differ in hemorrhage, necrosis, mitosis/karrhyorhexis, or proliferative and apoptotic indices at the time of terminal tumor progression (Fig. 2D). On-target activity was supported by reduced putrescine in DFMO and SAM486+DFMO–treated tumors, while the addition of SAM486 to DFMO led further to reduced spermidine and spermine, biomarkers of Amd1 inhibition (Fig. 2E).
Dual polyamine antagonism blocks neuroblastoma initiation. A, tumor-free survival for TH-MYCN+/+ (n = 68) treated with vehicle (n = 19), SAM486 (n = 25), or SAM486+DFMO (n = 23); P values by the method of Kaplan and Meier; all pairwise comparisons are P < 0.05 by log-rank test. Two mice were censored (depicted by a black circle) at the time of procedure-related death in the absence of tumor. B, relative change in tumor penetrance in TH-MYCN+/+ and +/− mice treated with DFMO, SAM486, or both DFMO and SAM486 compared with genotype-matched contemporaneous vehicle-treated mice (numbers of mice/group are shown). C, representative H&E stains of tumors demonstrating neural differentiation (arrowheads) induced by polyamine antagonist therapy; D, tumors arising under these conditions did not differ in necrosis, hemorrhage, or proliferative (Ki-67 IHC) or apoptotic (activated caspase-3 IHC) indices. E, evidence for polyamine depletion in TH-MYCN tumors by dual antagonism with DFMO and SAM486. Ctrl, control; D, DFMO treated; S, SAM486 treated; D+S, DFMO and SAM486 treated. *, P < 0.05 by two-tailed t test; **, P < 0.005 by two-tailed t test; NS, not significant. Note: for panels B–E, DFMO-treated tumors were obtained from a prior DFMO trial in a manner identical to the other groups herein (18).
Therapy with SAM486 or SAM486+DFMO was well tolerated. SAM486-treated mice grew at the same rate as control mice (day 42 weight 23.9 ± 2.6 g vs. 25.2 ± 2.3 g; P = 0.35). Mice treated with SAM486+DFMO were smaller at day 21 (17.1 ± 2.9 g vs. 21.3 ±1.8 g; P = 0.01), reflecting DFMO effects since SAM486 therapy started at day 21, but through day 42, mice treated with SAM486+DFMO, like DFMO-only treated mice, trended toward catch-up growth (Supplementary Fig. S3). Thus, dual inhibition of the rate-limiting enzymes in polyamine biosynthesis reduced neuroblastoma initiation and this was correlated with augmented putrescine and spermidine deprivation, supporting a potent role for polyamines in MYCN oncogenesis. Tumors that did arise under polyamine deprivation stress had increased neural differentiation, a favorable histologic finding, compared with polyamine-sufficient tumors.
DFMO, but not SAM486, synergizes with chemotherapy in treating established tumors
We previously showed that DFMO treatment of TH-MYCN mice with palpable neuroblastomas extended overall survival (P < 0.01) and enhanced chemotherapy activity, increasing survival when combined with vincristine, cyclophosphamide, or cisplatinum, and increasing durable complete regressions when combined with cisplatinum (18). We extended this testing additional regimens used for neuroblastoma therapy at diagnosis or following relapse and show that the addition of DFMO also improves survival when given with topotecan (P = 0.02), topotecan and cyclophosphamide (P < 0.02), temozolomide (P = 0.03), or irinotecan (P < 0.01), and induces durable complete regressions when given with irinotecan (P < 0.05; Table 1; Supplementary Fig. S4). Therefore, we confirmed that the addition of DFMO improved outcomes across diverse chemotherapy classes (camptothecins, alkylators, platinators, and microtubule poisons) without evident antagonism.
Summary of preclinical efficacy data treating established neuroblastomas in complementary TH-MYCN and human xenograft models with polyamine antagonists
We next tested the addition of SAM486 to DFMO in the treatment of established tumors. TH-MYCN mice with palpable tumors were randomized to DFMO+SAM486, or vehicle, with or without cyclophosphamide (Fig. 3A). Mice treated with DFMO+SAM486 (without chemotherapy) had survival increased >40% compared with vehicle-treated mice (median 22 days vs. 15 days; P = 0.006) similar to effects of DFMO alone (18). The addition of SAM486 and DFMO to cyclophosphamide therapy did not improve outcome over cyclophosphamide alone but there was instead a trend toward reduced survival (P = 0.11). SAM486+DFMO was similarly tested in neuroblastoma xenograft models. Mice harboring IMR5 xenografts (MYCN amplified) had extended survival when DFMO (P < 0.05) or DFMO+SAM486 (P < 0.03) was added to cyclophosphamide. However, the addition of SAM486 to DFMO and cyclophosphamide did not further improve survival over DFMO and cyclophosphamide alone (P = 0.40; Supplementary Table S2). Mice harboring BE2C tumor xenografts (MYCN amplified, TP53 mutant, multidrug resistant) were treated with cyclophosphamide and topotecan, a regimen often used after relapse, and again DFMO+SAM486 failed to improve survival (P = 0.62; Fig. 3B). Thus, SAM486 does not potentiate the activity of DFMO with chemotherapy in treating established murine or human neuroblastomas despite showing synergy in the preemptive therapy setting.
Effects of combined polyamine antagonist therapy. A, TH-MYCN mice (hemizygous and homozygous; total n = 126) with palpable tumors were randomized to receive DFMO+SAM486 (n = 33), DFMO+SAM486 with cyclophosphamide (n = 35), cyclophosphamide alone (n = 35), or vehicle (n = 23). DFMO+SAM486 extended survival over control mice (P = 0.006), as did cyclophosphamide (P < 0.001). DFMO+SAM486 did not improve survival on a backbone of cyclophosphamide, however, but instead trended toward antagonism of efficacy (P = 0.11). The same findings were seen when analyzed after stratifying by TH-MYCN genotype (data not shown). B, BE2C xenografts established in NCR nu/nu mice (n = 12 per arm) were treated with DFMO + SAM486 on a backbone of cyclophosphamide and topotecan, and compared with cyclophosphamide/topotecan alone or vehicle control. The addition of DFMO+SAM486 did not improve survival when added to chemotherapy when compared with chemotherapy alone (P = 0.62). Treatment annotation is provided above the Kaplan–Meier curve and detailed in Materials and Methods. C, polyamine uptake is functionally relevant in neuroblasts as the effects of DFMO on colony formation activity are completely (844) or partially (IMR5) rescued through the provision of supplemental polyamines to the culture media. D, spermidine uptake for murine (TH-MYCN 282 and 844) and human (IMR5, BE2C, SK-N-AS) neuroblastoma cell lines, with or without pre-exposure to DFMO and/or celecoxib, as indicated. DFMO-mediated polyamine depletion induces increased spermidine uptake. Although celecoxib by itself reduces basal uptake it does not attenuate the increased uptake following DFMO exposure. E, polyamine antagonists alone or in combination (48-hour exposure) deplete neuroblastoma cells of polyamines, with profound reductions in putrescine and spermidine in DFMO and DFMO+celecoxib–treated cells, whereas SAM486 rescues the spermidine depletion. For C–E, DFMO, 5 mmol/L; SAM (SAM486), 5 μmol/L; CEL (celecoxib), 10 μmol/L; PAs, all polyamines added at 1 μmol/L each. *, P < 0.05 by two-tailed t test compared with control, error bars show SEM. F, DFMO and celecoxib (top), and celecoxib and mafosfamide (bottom) synergize to inhibit colony formation in BE2C cells. A combination index (CI) of <0.9 indicates synergy: for DFMO and celecoxib, CI = 0.73; for celecoxib and mafosfamide, CI = 0.87. Wells shown are representative of those generated by exposure of cells to the highest concentrations of each drug used to generate the median effect plot, as described in Materials and Methods.
Disabling polyamine synthesis leads to compensatory increases in polyamine import
Augmented uptake of polyamines from the microenvironment may rescue polyamine-depleted tumor cells (37). The mammalian polyamine transporter remains poorly characterized but spermidine uptake is a biomarker of this activity. Following Odc1 and/or Amd1 inhibition, both murine and human neuroblastoma cells increased spermidine import up to 8-fold in response to polyamine depletion (Supplementary Fig. S5). Adding polyamines to the culture media rescued DFMO-mediated inhibition in colony formation supporting polyamine uptake as functionally relevant (Fig. 3C). Spermidine/spermine-N1-acetyltransferase (SAT1) activity acetylates polyamines to promote their export from the cell and regulates flux through the pathway (13). We reasoned that SAT1 induction might further deplete nonacetylated polyamine pools. COX inhibitors induce SAT1 and synergize with DFMO in reducing colorectal tumor recurrence in humans and in mouse models (38, 39). Because cox-2 is expressed in neuroblastoma and its inhibition by celecoxib has antitumor effects (40), we assessed celecoxib for synergy with DFMO.
Neuroblastoma cells modestly upregulated SAT1 mRNA and protein following celecoxib exposure (Supplementary Fig. S6A and S6B). Although this led to reduced net polyamine import under basal conditions, it could not attenuate the increase in uptake following DFMO-mediated polyamine depletion (Fig. 3D). However, polyamine acetylation and export are not reflected by spermidine uptake, so we measured nonacetylated polyamine levels to show DFMO alone or with celecoxib markedly reduced putrescine and spermidine, while spermine levels were preserved (Fig. 3E). In contrast, the addition of SAM486 to DFMO prevented putrescine and spermidine depletion by inhibiting Amd1. Neither the addition of supplementary polyamines nor extended exposure times altered these changes (Supplementary Fig. S7). Celecoxib itself had no cytotoxicity against murine neuroblastoma cells, and had activity for two of five human neuroblastoma cell lines but only at exposures ≥100 μmol/L not achieved with therapeutic dosing (Supplementary Fig. S6C). At these higher concentrations, we showed that supplementation with additional polyamines could partially rescue survival, supporting polyamine depletion as contributing to celecoxib effects (Supplementary Fig. S6D). We next tested for synergy between DFMO and celecoxib in BE2C cells across concentrations achievable in vivo. DFMO and celecoxib showed synergy in reducing colony formation (combination index, CI, of 0.73) with virtual elimination of colonies at higher concentrations (Fig. 3F). Synergy was not observed in SK-N-SH cells (absent MYCN amplification) when a CI was calculated from all four drug concentrations studied; however, synergy was observed (CI = 0.56) when the highest concentration was excluded. We next assessed celecoxib with mafosfamide (the activated metabolite of cyclophosphamide) and demonstrated colony formation inhibition for both BE2C (CI = 0.87) and SK-N-SH (CI = 0.81) cells (Fig. 3F).
Celecoxib synergizes with DFMO in vivo
Preemptive therapy of homozygous TH-MYCN mice with celecoxib alone from day 21 onward extended tumor-free survival similar to DFMO (P = 0.01; and Supplementary Table S2). We next tested celecoxib in the treatment of established tumors, combined with DFMO and/or chemotherapy. Unlike DFMO, celecoxib did not provide a survival benefit in the absence of cytotoxic agents; however, the combination of DFMO and celecoxib did (P = 0.03; Table 1). With cyclophosphamide, the addition of either celecoxib (P < 0.10), DFMO (P < 0.10), or both celecoxib and DFMO (P < 0.08) led to a trend toward improved survival, while mice treated with any polyamine antagonist (combined) had improved survival (P < 0.02; Fig. 4A and B). We next studied DFMO and celecoxib with two additional chemotherapy combinations, showing markedly extended survival when added to cyclophosphamide and topotecan (P < 0.0005; Fig. 4C) or irinotecan and temozolomide (P < 0.0001; Fig. 4D), compared with chemotherapy alone.
Impact of DFMO and celecoxib on established neuroblastomas in TH-MYCN mice. A, homozyous TH-MYCN mice (n = 9–12 per arm) with palpable tumors were randomized to receive cyclophosphamide alone, or with the addition of DFMO, celecoxib, or both DFMO and celecoxib. Each arm that included a polyamine antagonist trended toward improved survival compared with cyclophosphamide alone (P < 0.10 for all comparisons, as indicated). B, analysis of the same experimental data in A with the three polyamine antagonist arms combined (any polyamine antagonist) illustrates the survival benefit provided by a DFMO and/or celecoxib-based polyamine antagonist regimen when combined with cyclophosphamide chemotherapy (P < 0.02). C, homozyous TH-MYCN mice (n = 10 per arm) with palpable tumors were randomized to the combination of DFMO + celecoxib on a backbone of cyclophosphamide and topotecan, a regimen commonly used for relapsed or refractory neuroblastoma, or chemotherapy alone. The addition of DFMO and celecoxib improved survival (P = 0.0005). D, similar results were obtained using an alternative salvage chemotherapy backbone, irinotecan and temozolomide, as the addition of DFMO and celecoxib improved survival compared with chemotherapy alone (P < 0.001); P values by the method of Kaplan and Meier. Treatment annotation is provided above the Kaplan–Meier curve and detailed in Materials and Methods.
To evaluate the impact of DFMO and celecoxib in treating established human neuroblastomas, we used xenografted tumors with high-risk genetic alterations, using IMR5 cells (MYCN amplified) and SK-N-SH cells (MYCN nonamplified, ALK F1174L mutant) established from tumors at diagnosis, and multidrug-resistant BE2C cells (MYCN amplified and TP53 C135F mutant) established at the time of relapse (Table 1). The addition of DFMO to cyclophosphamide extended survival in mice harboring IMR5 tumors (P < 0.05), including a durable complete response (culled at day 65), whereas all cyclophosphamide alone–treated mice were culled for tumor progression by day 26. The addition of celecoxib to DFMO and cyclophosphamide did not further extend survival (P = 0.33; data not shown). Treatment of mice harboring SK-N-SH xenografts showed that even in the absence of chemotherapy, the combination of DFMO and celecoxib extended survival (P < 0.02; Fig. 5A). One DFMO- and celecoxib-treated mouse was sacrificed for poor weight gain following complete tumor regression (no tumor at necropsy). Another had delayed tumor growth with sacrifice at >275 days. When combined with cyclophosphamide, the addition of DFMO and celecoxib increased survival (P < 0.04). No mouse treated only with cyclophosphamide had a durable regression. In contrast, one mouse treated with cyclophosphamide and DFMO had slow tumor growth through day approximately 160 followed by regression until sacrifice at day 293, and 3 of 8 mice (37%) treated with cyclophosphamide, DFMO, and celecoxib had complete and durable tumor regressions despite receiving no additional chemotherapy retreatment (Fig. 5B).
Impact of DFMO and celecoxib on neuroblastomas xenografts, alone and with chemotherapy. A, BALB/c nu/nu mice (n = 7–8 per arm) harboring SK-N-SH xenografts were randomized to receive DFMO, celecoxib, or both (in the absence of chemotherapy). Neither DFMO nor celecoxib alone significantly extended survival; however, DFMO + celecoxib did (P < 0.02) with a >2-fold extension in median survival. B, the same model was used to test these combinations with cyclophosphamide: again, neither DFMO nor celecoxib extended survival when added to cyclophosphamide; however, the combination of DFMO + celecoxib did (P < 0.04). All cyclophosphamide-only treated animals were culled for tumor progression, one DFMO- and cyclophosphamide-treated mouse had a late tumor regression with no recurrence, and 3 mice (of 8) treated with both DFMO + celecoxib and cyclophosphamide had durable complete regressions with no tumor at necropsy. C, NCR nu/nu mice harboring therapy-resistant BE2C xenografts were randomized to receive cyclophosphamide and topotecan with or without DFMO or DFMO + celecoxib, or control (n = 12 per arm). Mice receiving DFMO had a trend toward extended survival (P < 0.10), whereas mice receiving both DFMO + celecoxib had markedly extended survival (P < 0.02). D, tumor volume over time is shown for mice in C receiving either cyclophosphamide/topotecan alone or combined with DFMO + celecoxib. As with SK-N-SH tumors, some mice receiving both DFMO + celecoxib had late and complete tumor regressions, despite receiving only a single chemotherapy course (days 1–3); P values by the method of Kaplan and Meier are shown. Treatment annotation is provided above the Kaplan–Meier curve and detailed in Materials and Methods.
The postrelapse BE2C xenograft was tested with two separate chemotherapy backbones, to define whether relatively therapy-resistant tumors might have similar vulnerabilities. Remarkably, DFMO alone, celecoxib alone, or DFMO + celecoxib all extended survival of xenograft bearing mice >50% compared with control-treated mice (P < 0.01, P = 0.02 and P < 0.01, respectively), a survival advantage similar to that achieved by cyclophosphamide (Supplementary Fig. S8). When added to cyclophosphamide, the combination of DFMO and celecoxib extended median survival approximately 40%, though all mice were eventually culled for tumor progression (P < 0.01; data not shown). Using a similarly noncurative cyclophosphamide and topotecan regimen, the addition of DFMO alone trended toward extended survival (P < 0.10), whereas the combination of DFMO and celecoxib together extended survival (P < 0.02) and led to durable complete regressions in 2 of 12 mice (Fig. 5C and D). At the time of sacrifice, no residual tumor could be identified nor could a cell line be propagated from the tumor residua.
Discussion
Myc proteins are attractive drug targets as they are among the most frequent somatically activated genes in cancer (2) and drive diverse neoplasia-enabling transcriptional programs (41). Systemic genetic inhibition of MYC has been shown to provide antitumor efficacy with tolerable toxicity (42) supporting that a therapeutic index exists for MYC-directed therapies. Unfortunately, MYC operates through a network of broad, flat protein–protein interactions and pharmacologic inhibition has remained challenging. An alternative to directly antagonizing MYC is to target the downstream pathways necessary for its oncogenic activity. MYC drives resource utilization toward biomass production, including a dramatic increase in protein synthetic capacity (43). This evolutionarily conserved role in protein synthesis is critical to Myc oncogenesis and its inhibition has synthetic-lethal consequences (44). Among the genes highly regulated by MYC are polyamine homeostatic enzymes, constituting a core expression program downstream of MYC in transformation (45).
That polyamine sufficiency is required to support Myc, oncogenesis was initially revealed in the Eu-MYC lymphoma model where genetic (ODC +/−) or biochemical (DFMO) polyamine antagonism was shown to extend tumor latency (46). We therefore explored its requirement in neuroblastomas, in which hyperactivation of MYCN through genomic amplification is present in 40% of high-risk tumors (47) and indirect deregulation of MYCN or MYC occurs in a large fraction of the remainder (17). Furthermore, up to 20% of MYCN-amplified tumors have coamplification of ODC1, demonstrating targeted deregulation of an oncogenic transcription factor and its oncogenic target gene (18, 48, 49). Indeed, high-risk neuroblastomas showed coordinate deregulation of polyamine enzymes in a direction to support polyamine sufficiency (18), whereas high ODC1 expression was independently prognostic for poor outcome, supporting its potential value of this oncogene as a drug target.
We previously showed that preemptive DFMO therapy extended tumor latency in the TH-MYCN model, and that DFMO improved tumor-free and overall survival when added to chemotherapy treatment of established tumors. Here, we provide extended preclinical testing using robust complementary models that support transitioning optimized polyamine antagonist therapies to the clinic. In strong support that polyamine sufficiency is required for MYC-mediated oncogenesis, we show that inhibiting Odc1 and Amd1 with DFMO and SAM486 reduces tumor penetrance by approximately 75% in hemizygous mice and approximately 40% in homozygous mice (where lethal tumor penetrance is 100% without treatment). Tumors rarely arose following withdrawal of therapy suggesting a sustained chemopreventive effect. It is not clear whether this preemptive therapy eradicates nascent tumors or has specific effects that inhibit initiation or tumor stem cell viability. The latter is indirectly supported by the lack of SAM486 activity when added to DFMO in the treatment of established tumors, the finding that ODC1 expression is enriched in neuroblastoma tumor-initiating cell populations (50), and that AMD1 is both essential for stem cell self-renewal and downregulated concurrent with differentiation to the neural lineage (51).
Treating established tumors failed to show an additive benefit for combining Amd1 inhibition with Odc1 inhibition, in contrast with the potent activity seen in the preemptive setting. A metabolomic approach applied to colorectal cancer models led to the hypothesis that reduced cellular thymidine pools contribute to DFMO-induced cytostatic activity (52). DFMO-mediated polyamine depletion leads to compensatory hyperactivation of Amd1 with futile S-adenosylmethionine consumption, followed by depletion of folate-dependent metabolites and thymidine. Thymidine replacement rescued the cytostatic effect of DFMO without restoring polyamine levels. This is notable as the addition of SAM486 to DFMO is predicted by this model to antagonize the activity of DFMO by attenuating thymidine depletion. This is consistent with our findings but will require direct evaluation, including assessing these metabolites in both preemptive and established tumor models.
In seeking additional opportunities to antagonize polyamine metabolism, the cox-2 inhibitor celecoxib seemed particularly promising. Celecoxib induces SAT1, has intrinsic Odc-inhibiting activity (53), and has been studied extensively in cancer trials together with chemotherapy, including in pediatric trials (54). Moreover, strong proof of concept for combining COX inhibition with DFMO was demonstrated by the striking reduction in recurrent colon adenoma in at-risk individuals through the preventive use of DFMO and sulindac (34). Consistent with this, the addition of celecoxib to DFMO extended survival when treating both murine (TH-MYCN) and human (xenograft) neuroblastomas, in combination with numerous clinically relevant chemotherapy regimens, and using genomically distinct high-risk tumor types (MYCN amplification, ALK mutation, TP53 mutation). It is plausible that celecoxib provides additional benefit to DFMO through the effects on spermidine depletion (an effect antagonized when SAM486 was added to DFMO), as this may be a biomarker for protein translation inhibition induced by polyamine deprivation via effects on eIF5A activities. However, pleiotropic COX activities have been postulated to impact tumor phenotype, including in neuroblastoma (55, 56), and warrant further consideration. Indeed, celecoxib has antitumor activity alone and combined with chemotherapy in neuroblastoma xenograft studies in which polyamine depletion stress is absent (40).
Although our preclinical studies support a cancer cell intrinsic activity for polyamine depleting agents, it does not exclude the contribution of cancer cell–extrinsic activities. Indeed, polyamines (via the arginine-ornithine-polyamine axis) contribute to a tumor-permissive microenvironment through myriad effects on immunosurveillance mechanisms (57–59), so efforts to study polyamine depletion in this context are warranted. Indeed, COX activity in tumors contributes to immune evasion and biochemical inhibition of these pathways restores antitumor immunity (60), providing an additional potential mechanism for the enhanced antitumor activities seen with DFMO and celecoxib in our models. Overall, demonstrating activity across a spectrum of preclinical models and in multiagent combinations, as done here, has been recommended in efforts to improve the rate of successful translation into human cancer trials (21). Our data support the integration of DFMO and COX inhibitors such as celecoxib into chemotherapeutic regimens for neuroblastoma and potentially other MYC-driven embryonal cancers (61).
Disclosure of Potential Conflicts of Interest
M.D. Hogarty reports receiving commercial research support from Cancer Prevention Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: N.F. Evageliou, M. Haber, J. Murray, S.K. Gilmour, M.D. Norris, M.D. Hogarty
Development of methodology: N.F. Evageliou, M. Haber, D.S. Ziegler, M.D. Hogarty
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.F. Evageliou, M. Haber, A. Vu, T.W. Laetsch, J. Murray, L.D. Gamble, N.C. Cheng, K. Liu, K.A. Corrigan, D.S. Ziegler, H.T. Webber, C.S. Hayes, B.R. Pawel, G.M. Marshall, S.K. Gilmour, M.D. Hogarty
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.F. Evageliou, M. Haber, J. Murray, L.D. Gamble, N.C. Cheng, K.A. Corrigan, G.M. Marshall, H. Zhao, S.K. Gilmour, M.D. Norris, M.D. Hogarty
Writing, review, and/or revision of the manuscript: N.F. Evageliou, M. Haber, J. Murray, L.D. Gamble, D.S. Ziegler, G.M. Marshall, S.K. Gilmour, M.D. Norris, M.D. Hogarty
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.F. Evageliou, M. Haber, K.A. Corrigan, B.R. Pawel, M.D. Hogarty
Study supervision: M. Haber, M.D. Hogarty
Other (ran experiments): M. Reese
Grant Support
This work was financially supported by grants from the US Department of Defense W81XWH-10-1-0145 (to M.D. Hogarty and S.K. Gilmour), and the Richard and Sheila Sanford Chair in Pediatric Oncology (to M.D. Hogarty), the Children's Neuroblastoma Cancer Foundation (to N.F. Evageliou), The National Health and Medical Research Council (Australia) and Cancer Institute New South Wales (to M. Haber, G.M. Marshall, and M.D. Norris). Children's Cancer Institute Australia for Medical Research is affiliated with the University of New South Wales and Sydney Children's Hospital Randwick, Sydney, Australia and is a member of the Kid's Cancer Alliance.
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.
Acknowledgments
The authors thank Pat Woster (Medical University of South Carolina) for DFMO, Robert Cozens (Novartis, Basil, Switzerland) for SAM486, William Weiss (UCSF) for TH-MYCN mice, Naomi Balamuth and John Maris (University of Pennsylvania) for TH-MYCN model transcriptome datasets, Ashleigh Clark and Michelle Ruhle for expert technical assistance, and Andre Bachmann (Michigan State University) for helpful discussions.
Footnotes
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
- Received October 20, 2015.
- Revision received February 26, 2016.
- Accepted March 15, 2016.
- ©2016 American Association for Cancer Research.
References
Volume 22, Issue 17
