Targeting Ornithine Decarboxylase Impairs Development of MYCN-Amplified Neuroblastoma

Neuroblastoma is a pediatric malignancy that arises from the neural crest, and patients with high-risk neuroblastoma, which typically harbor amplifications of MYCN, have an extremely poor prognosis. The tyrosine hydroxylase (TH) promoter-driven TH-MYCN transgenic mouse model faithfully recapitulates many hallmarks of human MYCN-amplified neuroblastoma. A key downstream target of Myc oncoproteins in tumorigenesis is ornithine decarboxylase (Odc), the rate-limiting enzyme of polyamine biosynthesis. Indeed, sustained treatment with the Odc suicide inhibitor α-difluoromethylornithine (DFMO) or Odc heterozygosity markedly impairs lymphoma development in Eμ-Myc transgenic mice, and these effects are linked to the induction of the cyclin-dependent kinase (Cdk) inhibitor p27^Kip1, which is normally repressed by Myc. Here, we report that DFMO treatment, but not Odc heterozygosity, impairs MYCN-induced neuroblastoma and that, in this malignancy, transient DFMO treatment is sufficient to confer protection. The selective anticancer effects of DFMO on mouse and human MYCN-amplified neuroblastoma also rely on its ability to disable the proliferative response of Myc, yet in this tumor context, DFMO targets the expression of the p21^Cip1 Cdk inhibitor, which is also suppressed by Myc oncoproteins. These findings suggest that agents, such as DFMO, that target the polyamine pathway may show efficacy in high-risk, MYCN-amplified neuroblastoma. [Cancer Res 2009;15(4):547–53]

Neuroblastoma is a childhood malignancy that accounts for nearly 8% of all childhood cancers and 15% of pediatric cancer–related deaths (1). This tumor arises from sympathetic nervous tissue, most commonly in the adrenal gland, and its onset is age-dependent, with an incidence, during the first year of life, twice that of the second. Furthermore, neuroblastoma is the most common cancer during infancy, with an incidence nearly twice that of pediatric leukemia. Tragically, children with stage IV, high-risk neuroblastoma have a <40% long-term survival (2).

The best-described genetic alteration in neuroblastoma is the amplification of the proto-oncogene MYCN, which occurs in ∼20% of all neuroblastoma and is associated with the high-risk phenotype (3, 4). MYCN is a member of the Myc family of transcription factors, which function as master regulators that coordinate cell growth (mass), cell metabolism, and division and which are activated in up to 70% of rapidly dividing malignancies. Their selective activation in cancer likely reflects their regulation of a large cast of genes involved in cell metabolism and proliferation, as well as in tumor angiogenesis and metastasis (5). Indeed, expression profiling and genome-wide chromatin binding analyses suggest that Myc can regulate the transcription of up to 15% of the genome (6, 7). Enforced expression of Myc in transgenic mouse models is sufficient to provoke a wide array of malignancies that accurately phenocopy human malignancies (8–11). This is particularly true of the tyrosine hydroxylase (TH) promoter-driven expression of N-Myc in neural crest progenitors that gives rise to a malignancy that faithfully recapitulates MYCN-amplified stage IV neuroblastoma, including shared syntenic chromosomal alterations (9).

A hallmark of cancer is that sustained expression or activity of oncogenic lesions are necessary to maintain the malignant state, and this especially applies to tumors driven by Myc, wherein even rather brief inactivation of Myc usually leads to complete tumor regression (12, 13). Thus, targeting Myc directly (e.g., blocking its transcription or translation or augmenting its turnover) could be therapeutic, yet Myc oncoproteins are also required for the growth of most normal cell types (14, 15), raising concerns of acquiring a suitable therapeutic index. Furthermore, the complexity of Myc's transcriptional response raises questions regarding which targets downstream of Myc to choose and whether intervention through any single target will show efficacy.

Ornithine decarboxylase (Odc), the first and a rate-limiting enzyme in the polyamine biosynthesis pathway, is a well-characterized, direct Myc transcription target (16, 17) clearly amenable to therapeutic intervention. Odc converts ornithine into putrescine, which is then converted into the more abundant polyamines spermidine and spermine (Supplementary Fig. S1), which control various aspects of cell biology, including replication, translation, and cell growth and survival (18, 19). Accordingly, polyamine levels are tightly maintained in cells, and these controls include the catabolic arm of the pathway that allows for back conversion (Supplementary Fig. S1) and, for Odc, also include regulation of its translation and turnover, which is directed by a dedicated inhibitor, coined antizyme, that directly shuttles Odc to the proteasome for destruction. In turn, an Odc decoy, coined antizyme inhibitor, dampens antizyme activity (reviewed in ref. 19).

Elevated Odc levels and accompanying increases in polyamine pools are common in cancer (18), and enforced Odc expression in the skin of transgenic mice leads to increased tumor incidence and rate in response to chemical carcinogens (20, 21). Furthermore, like Myc, Odc cooperates with activated Ras in transformation in vitro and in vivo (22, 23). Notably, Odc heterozygosity or treatment with the difluoromethylornithine (DFMO), a suicide inhibitor of Odc, impairs Myc-induced lymphomagenesis in Eμ-Myc transgenic mice (24), a model of human B-cell lymphoma (8). In B cells, targeting Odc selectively impairs the proliferative response of Myc by disabling its ability to suppress the expression of the cyclin-dependent kinase (Cdk) inhibitor p27^Kip1 (24). Importantly, recent clinical trials of colon and prostate cancer, two malignancies with known Myc involvement (e.g., ref. 11), have shown remarkable efficacy of DFMO as a chemoprevention agent (25, 26). Here, we tested the hypothesis that targeting Odc would also show efficacy in MYCN-driven neuroblastoma and report that DFMO, but not Odc heterozygosity, impairs the proliferative response of N-Myc and delays tumor incidence and onset. Interestingly, targeting Odc in this context affects the expression of a second arbiter of the proliferative response of Myc, the Cdk inhibitor p21^Cip1. Thus, targeting Odc disables Myc-induced tumorigenesis via distinct effectors, depending upon tumor type, yet this typically involves Cdk inhibitors that disable the proliferative response of Myc.

Array analyses. The GSE3960 Series Matrix File was downloaded from National Center for Biotechnology Information Gene Expression Omnibus database. This file summarizes the expression profiles of 101 primary human neuroblastoma using Affymetrix U95Av2 arrays (27). Z scores were used in GeneSpring 7.3 (GS) for hierarchical clustering and visualization of microarray data. Z scores were calculated by subtracting the average gene signal in all arrays from the signal for each gene and dividing the result by the SD of all measured signals. Pearson correlation similarity measures and average linkage clustering algorithms were used in GS for hierarchical clustering of samples, which segregated the two major tumor groups. GS was also used for Student's t test between the two tumor groups. Genes with P values of <0.05 were identified as those that were significantly differentially expressed between the two tumor groups.

RNA preparation and analyses. Tumors were collected from TH-MYCN mice (9) at the time of sacrifice and were snap frozen. An aliquot of each tumor was homogenized. RNA from tumor samples and cultured cells was prepared using the RNeasy kit (Qiagen). The iScript cDNA synthesis kit (Bio-Rad) and 1 μg of RNA were used to prepare cDNA for quantitative real-time PCR (qRT-PCR). qRT-PCR was performed with the iTaq SYBRGreen kit and an iCycler machine (Bio-Rad). Data analyses were performed with the ΔΔCt method, wherein ubiquitin served as the internal control.

To assess potential effects of DFMO on the turnover of p21^Cip1 mRNA in MYCN-amplified neuroblastoma, IMR-32 and CHP-134 cells were either mock-treated or DFMO-treated for 3 d. Actinomycin D (10 μg/mL) was then added, cells were harvested at specific intervals, RNA was isolated, and p21^Cip1 mRNA levels were determined by qRT-PCR. Expression was standardized to the expression of ubiquitin; experiments were performed in duplicate.

Western blot analyses. Cell culture lysates and homogenized tumor aliquots from TH-MYCN mice were disrupted in lysis buffer [50 mmol/L HEPES (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA, 2.5 mmol/L EGTA, and 0.1% Tween 20 with 1 mmol/L phenylmethylsulfonyl fluoride, 10 mmol/L β-glycerophosphate, 1 mmol/L NaF, 1 mmol/L NaVO₄, and complete mini tablet protease inhibitor (Roche)] by sonication, as described (24). For analyses of p21^Cip1 levels in neuroblastoma cell lines, nuclear extracts were prepared, as described by Andrews and Faller (28). Protein (40–50 μg/lane) was separated on 10% SDS-polyacrylamide gels, transferred to polyvinylidene difluoride membranes (Immobilon-P, Millipore), and blotted for antibodies specific for N-Myc (OP13, Calbiochem), ODC (from Dr. Lisa Shantz, Pennsylvania State University School of Medicine), p21^Cip1 (for mouse, sc-6246, Santa Cruz; for human, sc-397, Santa Cruz), p27^Kip1 (610242, BD Transduction Labs), p53 (for mouse, 1C12, Cell Signaling; for human, sc-6243, Santa Cruz), actin (AC-15, Sigma), and α-tubulin (B-5-1-2, Sigma).

Cell culture. Human neuroblastoma cell lines SK-N-MC, SK-N-SH, IMR-32, and CHP-134 were obtained from American Type Culture Collection (ATCC) and were maintained in RPMI 1640 (Life Technologies) with 10% fetal bovine serum, 1× l-glutamine, 1× MEM nonessential amino acids, and 1% penicillin/streptomycin at 37°C, 5% CO₂. Cells were harvested for protein and RNA analyses. For treatment with DFMO, cells were split at 2.5 × 10⁶ into 15-cm plates for protein analyses and 4 × 10⁵ into 6-cm plates for RNA and fluorescence-activated cell sorting (FACS) analyses 4 h before 5 mmol/L DFMO treatment. Cells were cultured in the presence of DFMO for 72 h and then harvested for protein, RNA, or FACS analyses. For growth curve analysis, 1 × 10⁵ cells were plated into six-well plates 4 h before 5 mmol/L DFMO or mock treatment. Each day, three wells of mock-treated and three wells of DFMO-treated cells were individually collected and counted. Cells were counted using a Cellometer Auto T4 cell counter (Nexcelom Bioscience).

Mice and tumor analyses. Mice expressing the human MYCN gene under the control of the rat tyrosine hydroxylase promoter (TH-MYCN; ref. 9) were maintained on a 129x1/SvJ background. TH-MYCN littermates were given either water, water containing 1% DFMO from weaning (∼21 d) to sacrifice, or water containing 1% DFMO from weaning to 120 d of age. TH-MYCN were also bred to Odc^+/− mice (ref. 29; on a 94% 129x1/SvJ background) to generate TH-MYCN;Odc^+/+ and TH-MYCN;Odc^+/− littermates (97% 129x1/SvJ background). Mice were monitored daily for illness and tumor development. Sick animals were sacrificed, and tumors were collected, snap frozen, and stored at −80°C.

Flow cytometry. Cell cycle analysis was performed by using the FITC BrdU flow kit (559619, BD Biosciences). DFMO-treated or mock-treated (72 h) cultured cells were incubated with 10 μmol/L BrdUrd for 30 min and collected by trypsinization. Cells were stained for BrdUrd incorporation and cell cycle, as previously described (24). Stained cells were analyzed using a BD FACSCanto II; doublets were excluded based on pulse width/height and separated based on FITC staining (BrdUrd incorporation) versus 7-AAD (cell cycle).

MYCN-amplified neuroblastomas express elevated levels of ODC. High-risk, stage IV neuroblastoma is often accompanied by MYCN amplification (3, 4). To initially canvass the potential relationships of amplified MYCN in neuroblastoma with the polyamine pathway, we performed hierarchical clustering of microarray data of 101 primary human neuroblastomas from Gene Expression Omnibus Series GSE3960, wherein ∼20% of the tumors are MYCN-amplified (27). Several polyamine biosynthetic genes were up-regulated in MYCN-amplified versus non–MYCN-amplified tumors, including ODC, AMD1 (S-adenosylmethionine decarboxylase), SMS (spermidine synthase), and SRM (spermine synthase; Fig. 1A). In addition, there were significantly reduced levels of the ODC antagonist antizyme 2 (OAZ2) in MYCN-amplified neuroblastoma, whereas other components of the pathway were not significantly different between the two tumor cohorts (Fig. 1A). Collectively, these data suggest an increase in polyamine pools in MYCN-amplified neuroblastoma.

The expression of the entire polyamine pathway was also interrogated in the SK-N-MC, SK-N-SH, IMR-32, and CHP-134 human neuroblastoma cell lines. SK-N-MC and SK-N-SH are non–MYCN-amplified neuroblastoma cells, whereas IMR-32 and CHP-134 contain 25 and 100 copies of the MYCN gene per cell, respectively (30). qRT-PCR and Western blot analyses confirmed that IMR-32 and CHP-134 cells expressed much higher levels of MYCN RNA and protein than SK-N-MC and SK-N-SH cells (Fig. 1B and C). The levels of ODC were also increased in the MYCN-amplified cell lines, IMR-32 (1.8-fold) and CHP-134 (2.2-fold), compared with SK-N-MC (1.0-fold) and SK-N-SH cells (1.2-fold; Fig. 1B). Similarly, ODC protein levels were elevated in the MYCN-amplified versus non–MYCN-amplified neuroblastoma cell lines (Fig. 1C). Therefore, ODC expression is selectively elevated in MYCN-amplified primary neuroblastoma and neuroblastoma cell lines.

The expression levels of the other genes in the polyamine pathway were also assessed in these neuroblastoma cell lines by qRT-PCR. Here, although the expression of spermidine synthase was elevated in the MYCN-amplified lines (Fig. 1B), the expression of others that were significantly different in primary neuroblastoma were either unchanged between the two cell line cohorts (e.g., AMD1 and SSAT) or were regulated in an inverse fashion (e.g., SRM and OAZ2; Supplementary Fig. S2). Thus, the expression of the ODC and SMS polyamine biosynthetic enzymes pathway are concordant between MYCN-amplified primary neuroblastoma and MYCN-amplified neuroblastoma cell lines.

DFMO selectively impairs the proliferation of MYCN-amplified neuroblastoma. The increased levels of ODC in MYCN-amplified neuroblastoma and MYCN-amplified neuroblastoma cell lines suggested that inhibiting ODC might selectively affect neuroblastoma with MYCN involvement. To test this notion, we initially assessed the effects of DFMO, an agent that was originally developed as a suicide inhibitor of the active ODC dimer (19), on the growth and/or survival of MYCN-amplified neuroblastoma cells. DFMO had no appreciable effects on the survival of any of the four neuroblastoma cell lines (data not shown). By contrast, DFMO had selective effects on the proliferation of MYCN-amplified versus non–MYCN-amplified neuroblastoma cell lines (Fig. 2A). Here, there were marked increases in the percentages of IMR-32 and CHP134 cells in G₁ and corresponding reductions in the percentage of these cells in S phase (Fig. 2A and Supplementary Fig. S3). By contrast, DFMO had only modest effects on the proliferation of SK-N-MC cells and had no effect on the growth of SK-N-SH cells (Fig. 2A and Supplementary Fig. S3). Accordingly, growth curve experiments ± DFMO showed that DFMO treatment effectively abolished the growth of MYCN-amplified IMR-32 and CHP-134 neuroblastoma cells but had more modest effects on the rates of cell growth of SK-N-MC cells and none whatsoever on SK-N-SH cells (Fig. 2B). Therefore, the growth of MYCN-amplified neuroblastoma cells is dependent upon ODC.

DFMO, but not Odc heterozygosity, impairs N-Myc–driven neuroblastoma. In Eμ-Myc transgenic mice, either DFMO or Odc heterozygosity impairs the proliferative response of c-Myc in B cells and markedly delays lymphomagenesis (24). Given the selective effects of DFMO on the proliferation of MYCN-amplified neuroblastoma cells, we reasoned that similar responses would be manifest in the rat tyrosine hydroxylase promoter-driven TH-MYCN transgenic model, which overexpress N-Myc in the sympathetic nervous tissue derived from the neural crest and which develop neuroblastoma with ∼70% penetrance. Notably, hemizygous TH-MYCN mice develop tumors that are histologically similar to those arising in pediatric stage IV, MYCN-amplified neuroblastoma, including syntenic gain and loss of chromosomes (9). Beginning at 3 weeks of age, these transgenic mice were treated with 1% DFMO in their drinking water, a dose that effectively blocks Odc enzyme activity in vivo and leads to marked reductions in putrescine and spermidine pools (24). These mice were then followed for neuroblastoma development and compared with the control transgenic cohort that received normal drinking water. As expected (31), 74% (20 of 27) of untreated TH-MYCN mice developed neuroblastoma by 1 year of age. By contrast, only 48% (12 of 25) of TH-MYCN mice receiving DFMO-drinking water developed neuroblastoma by 1 year of age (Fig. 3A). Furthermore, DFMO also delayed disease, wherein the median survival of the DFMO-treated cohort that developed neuroblastoma was 139 days and that of the untreated cohort was 84 days (Fig. 3A). Therefore, DFMO delays both the onset and the incidence of neuroblastoma.

In Eμ-Myc transgenics, continuous DFMO treatment is required to prevent lymphoma development, and when treated transgenics are deprived of the drug, they develop lymphoma with kinetics nearly identical to that of untreated Eμ-Myc mice (24). To address whether this was also the case for MYCN-driven neuroblastoma, a cohort of TH-MYCN mice received DFMO-treated water from 3 weeks of age until they were 120 days old, at which time they were then withdrawn from the drug. Surprisingly, only 38% (8 of 21) of this cohort of mice succumbed to disease (Fig. 3A). Thus, only transient exposure to DFMO is needed for chemoprevention, a finding which suggests that Odc activity is only required early during neuroblastoma development.

To test if Odc heterozygosity would similarly impair neuroblastoma development, we backcrossed Odc^+/− mice on to a 129x1/SvJ background until they were ∼94% 129x1/SvJ, and these mice were then crossed to TH-MYCN mice (129x1/SvJ) to generate TH-MYCN;Odc^+/+ and TH-MYCN; Odc^+/− littermates (∼97% 129x1/SvJ), which were followed for 1 year for neuroblastoma development. On this mixed background, incidence of disease was reduced to 54% (22 of 41) for the TH-MYCN;Odc^+/+ cohort. In this context, Odc heterozygosity did not significantly affect disease incidence (P = 0.6241), wherein 44% (16 of 36) of TH-MYCN;Odc^+/− mice developed neuroblastoma (Fig. 3B). Furthermore, there was no significant difference in the median survival between the two cohorts (Fig. 3B). Therefore, Odc heterozygosity is not sufficient to prevent Myc-driven cancer in this context. However, the reductions in penetrance of neuroblastoma from 74% on a pure 129x1/SvJ background (Fig. 3A), where disease penetrance and severity is most manifested (9), to that observed for TH-MYCN;Odc^+/+ mice (54% penetrance; Fig. 3B) could potentially mask modest effects of Odc heterozygosity on N-Myc–driven tumorigenesis.

DFMO targets p21^Cip1 to impair the growth of MYCN-amplified neuroblastoma. In Eμ-Myc B cells, targeting Odc disables Myc-mediated suppression of p27^Kip1 (24). We, therefore, tested whether DFMO-induced growth arrest in MYCN-amplified neuroblastoma cells was also specifically associated with increased levels of p27^KIP1. As expected, there were no effects of DFMO on the expression of N-MYC protein or ODC in MYCN-amplified neuroblastoma, but surprisingly, there were also no effects of DFMO on p27^KIP1 mRNA or protein levels (Fig. 4A and B). Myc also suppresses the transcription of the Cdk inhibitor p21^Cip1 (32, 33); thus, we tested whether DFMO disables this response in MYCN-amplified neuroblastoma. Indeed, qRT-PCR and Western blot analyses established that DFMO induced marked and selective increases in p21^CIP1 mRNA levels in the IMR-32 and CHP-134 MYCN-amplified cell lines (Fig. 4A). p21^CIP1 is a known transcription target of p53 (34) but p21^CIP1induction by DFMO was not associated with increased levels of p53 (Fig. 4B), which is wild type in this cast of neuroblastoma cells (35).¹

However, the induction of p21^CIP1 mRNA by DFMO did seem to be due to effects on transcription, as DFMO had no effects on the t_½ of p21^CIP1 mRNA in actinomycin-D experiments, which measured their turnover (Supplementary Fig. S4). Importantly, the induction of p21^CIP1 by DFMO was also manifested at the protein level, wherein there were selective increases in p21^CIP1 protein in DFMO-treated IMR-32 and CHP-134 MYCN-amplified neuroblastoma (Fig. 4B). Therefore, the selective effects of DFMO in impairing the proliferative response of MYCN-amplified neuroblastoma are associated with increases in p21^CIP1 expression.

DFMO targets p21^Cip1 in TH-MYCN neuroblastoma. The findings that DFMO selectively targeted p21^CIP1 in human neuroblastoma suggested that this pathway would also be involved and targeted in vivo in neuroblastoma arising in TH-MYCN transgenics. One prediction was that p21^Cip1 expression would be generally suppressed in such TH-MYCN tumors. Indeed, only 3 of 11 neuroblastomas from TH-MYCN transgenics showed detectable levels of p21^Cip1 protein (Fig. 5A). In contrast, nearly all neuroblastoma expressed p27^Kip1 (Fig. 5A) and again in contrast to Myc-driven lymphomas where p27^Kip1 protein levels are suppressed (24). Therefore, p21^Cip1 expression is suppressed in N-Myc–driven neuroblastoma.

A second prediction was that DFMO treatment would selectively affect p21^Cip1 expression in the delayed-onset tumors that eventually arose in the DFMO-treated cohort of TH-MYCN mice. Indeed, levels of p21^Cip1 mRNA and protein were higher in most DFMO-treated tumors compared with the untreated cohort (Fig. 5A and B). Therefore, DFMO also induces the expression of p21^Cip1 in MYCN-driven neuroblastoma in vivo. In contrast, levels of MYCN RNA and protein and of p27^Kip1 were unaffected by DFMO (Fig. 5A and B).

ODC amplification has been described as a mechanism of DFMO resistance (36); thus, we also assessed whether Odc was induced in the DFMO-treated cohort. Indeed, there were marked increases in Odc protein in the DFMO-treated cohort of neuroblastoma (Fig. 5A). However, the Odc gene was not amplified in these tumors (Supplementary Fig. S5) and levels of Odc transcripts were similar between the two cohorts (Fig. 5B). Therefore, up-regulation of Odc in the DFMO-treated cohort seems due to either alterations in Odc translation or turnover.

Myc oncoproteins augment the rate of cell proliferation (37), and in Eμ-Myc transgenic mice this response is limiting for tumor onset and survival. Specifically, in B cells enforced c-Myc expression increases the number of cells in cycle by repressing p27^Kip1 protein levels, and this occurs through the Myc-mediated regulation of E2f1 and Cks1, which together are required for activating the SCF^SKP2 complex that targets p27^Kip1 for destruction by the proteasome (38, 39). Accordingly, loss of p27^Kip1 accelerates rates of Myc-induced lymphoma development (40), whereas loss of either E2f1 or Cks1 compromises Myc's proliferative response, suppression of p27^Kip1, and onset and incidence of Myc-drive lymphoma (38, 39). Quite strikingly, the Myc-to-p27^Kip1 pathway is also controlled by the polyamine pathway, wherein Odc expression is greatly elevated in Eμ-Myc B cells and targeting Odc with DFMO or Odc heterozygosity blocks the ability of Myc to repress p27^Kip1 (24).

The findings presented herein show some parallels of targeting Odc in MYCN-amplified neuroblastoma compared with Myc-overexpressing B cells. Specifically, akin to Myc-driven B-cell lymphomas of mice and man, ODC is also overexpressed in stage IV, MYCN-amplified pediatric neuroblastoma, findings consistent with ODC being a direct transcription target of Myc oncoproteins (16, 17). Furthermore, DFMO also selectively cancels the proliferative response of MYCN-amplified neuroblastoma and dramatically impairs the onset and incidence of N-Myc–driven neuroblastoma in vivo. However, interesting and context-specific differences are also evident, wherein DFMO effects on the proliferative response driven by N-Myc are linked to effects on p21^Cip1 in both human and mouse neuroblastoma and only transient treatment with DFMO is necessary to achieve chemoprevention. Furthermore, Odc heterozygosity has no effect on tumorigenesis in this tumor model, whereas in Eμ-Myc transgenics and in carcinogen-driven papillomas, Odc heterozygosity has marked effects on tumor onset, incidence, and overall survival (24, 41). Cell type and/or tumor context-specific control of polyamine homeostasis could account for these differences and allow neural crest progenitors to overcome effects of reductions in Odc. Indeed, polyamine homeostasis is different between Myc-driven lymphoma and neuroblastoma, as all of the biosynthetic enzymes are up-regulated in Myc-induced B-cell lymphoma (24) whereas only ODC and spermidine synthase are increased in MYCN-amplified neuroblastoma (Fig. 1A and B and Supplementary Fig. S2). Finally, in neural crest progenitors, there may be no substantial effects of loss of one Odc allele on overall levels of Odc protein, on Odc enzyme activity, and/or on polyamine pools.

Loss of the neurofibromatosis-1 (Nf1) or retinoblastoma (Rb) tumor suppressors accelerates disease in TH-MYCN mice (9). These studies are consistent with apparent alterations of NF1 in pediatric MYCN-amplified neuroblastoma (42, 43) and with the reduced expression of p21^Cip1 in most TH-MYCN neuroblastomas (Fig. 5), which express hyperphosphorylated (i.e., inactive) pRb (data not shown). They are also consistent with reduced p21^CIP1 expression in MYCN-amplified versus nonamplified human neuroblastoma (P < 0.005, Gene Expression Omnibus Series GSE3960; data not shown). They also underscore the unique genetics of this peculiar malignancy versus those driven by Myc in other scenarios, wherein there seems to be little to no contribution of p21^Cip1 on tumor development (40). Finally, these findings support the interesting notion that Myc activation in different tumor types selects for the inactivation of distinct tumor suppressor pathways, which contribute to malignant progression.

The mechanisms by which targeting Odc induces the p27^Kip1 and p21^Cip1 inhibitors to impair the proliferative response of Myc in lymphoma versus neuroblastoma, respectively, are not fully understood. In B cells, targeting Odc disables the ability of Myc to trigger p27^Kip1 protein degradation, and here, effects appear via Cks1,²

which is induced by Myc and is a required component of the SCF^Skp2 complex that directs p27^Kip1 destruction (39, 44). By contrast, in MYCN-amplified neuroblastoma there are elevated levels of ODC and low levels of p21^CIP1 mRNA and protein, and DFMO seems to cancel the ability of N-Myc to repress p21^Cip1 transcription, both ex vivo and in vivo. We propose that MYCN-amplified cells are, similar to Myc-expressing B cells, “addicted” to ODC for their proliferative response and that, in neuroblastoma, this hinges upon proper regulation of p21^CIP1. Furthermore, we suspect that the effects of DFMO on p21^Cip1 expression that have been observed in some tumor cell lines (45, 46) reflect Myc involvement in the tumors from which they were derived, which renders these cells addicted to ODC. How DFMO may affect p21^CIP1 transcription in MYCN-amplified neuroblastoma is not yet clear. However, because this response seems to be p53-independent and does not involve changes in the t_½ of p21^CIP1 mRNA, we hypothesize that DFMO somehow interferes with Myc-mediated transcriptional repression of p21^CIP1, which occurs by binding of Myc to Miz-1 (32).

The findings that DFMO is an effective prevention agent in neuroblastoma were quite surprising, because the tyrosine hydroxylase promoter driving the MYCN transgene is active from the genesis of the neural crest early in development. Thus, one would predict more benefit if DFMO was given earlier in life. However, DFMO cannot be given during gestation, where this drug or total loss of Odc causes early embryonic lethality (37, 47), nor after birth, as there are side effects on neuronal development (48). Thus, there are valid concerns whether DFMO could be given in a safe manner to pediatric patients suffering from stage IV MYCN-amplified neuroblastoma. Nonetheless, with the low doses of DFMO that have been given in the colon and prostate cancer chemoprevention trials, wherein there were no noticeable side effects (25, 26), it may be wise to revisit this issue for stage IV neuroblastoma because transient treatment with DFMO here is also sufficient to provide chemoprevention and especially because it may show benefit for this otherwise lethal malignancy. Notably, recent findings that ODC overexpression is a superior indicator of poor prognosis in neuroblastoma (49) suggests that even those patients lacking MYCN amplicons but overexpressing ODC may experience benefit from DFMO treatment. Regardless, our studies do suggest that targeting the polyamine pathway should perhaps be considered in combination with more conventional agents to treat this devastating disease.

No potential conflicts of interest were disclosed.

Grant support: NIH grants R01-CA100603 (J.L. Cleveland) and F32-CA115075 (R.J. Rounbehler) and monies from State of Florida to the Scripps Research Institute.

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.

We thank the Animal Resource Center of Scripps-Florida for their assistance, Dr. William Weiss for providing the TH-MYCN transgenic mice, Dr. Lisa Shantz for the kind gift of the anti-ODC antibody, Dr. Meredith Steeves for assistance with generating the Odc Southern blot probe, and members of the Cleveland laboratory for their help and review.