Research ArticleNeuroblastoma

Inhibition of polyamine synthesis and uptake reduces tumor progression and prolongs survival in mouse models of neuroblastoma

See allHide authors and affiliations

Science Translational Medicine  30 Jan 2019:
Vol. 11, Issue 477, eaau1099
DOI: 10.1126/scitranslmed.aau1099

A double attack on neuroblastoma

MYCN-amplified neuroblastoma (NB) is associated with an aggressive phenotype and poor outcome. Polyamine expression, critical for cell growth and survival, is increased in MYCN-amplified tumors. However, inhibition of polyamine synthesis failed in clinical trials. Now, Gamble et al. showed that, in NB cells, MYCN directly modulates polyamine synthesis, catabolism, and transport. In patient-derived cells and rodent models, MYCN directly modulated the expression of genes involved in polyamine catabolism, synthesis, and transport. Combined inhibition of polyamine synthesis and uptake increased survival in animal models of MYCN-amplified NB. These results suggest that the combined inhibition of polyamine synthesis and transport might be an effective strategy for treating MYCN-amplified NB.


Amplification of the MYCN oncogene is associated with an aggressive phenotype and poor outcome in childhood neuroblastoma. Polyamines are highly regulated essential cations that are frequently elevated in cancer cells, and the rate-limiting enzyme in polyamine synthesis, ornithine decarboxylase 1 (ODC1), is a direct transcriptional target of MYCN. Treatment of neuroblastoma cells with the ODC1 inhibitor difluoromethylornithine (DFMO), although a promising therapeutic strategy, is only partially effective at impeding neuroblastoma cell growth due to activation of compensatory mechanisms resulting in increased polyamine uptake from the surrounding microenvironment. In this study, we identified solute carrier family 3 member 2 (SLC3A2) as the key transporter involved in polyamine uptake in neuroblastoma. Knockdown of SLC3A2 in neuroblastoma cells reduced the uptake of the radiolabeled polyamine spermidine, and DFMO treatment increased SLC3A2 protein. In addition, MYCN directly increased polyamine synthesis and promoted neuroblastoma cell proliferation by regulating SLC3A2 and other regulatory components of the polyamine pathway. Inhibiting polyamine uptake with the small-molecule drug AMXT 1501, in combination with DFMO, prevented or delayed tumor development in neuroblastoma-prone mice and extended survival in rodent models of established tumors. Our findings suggest that combining AMXT 1501 and DFMO with standard chemotherapy might be an effective strategy for treating neuroblastoma.


Neuroblastoma (NB) is the most common solid tumor among infants and arises from the primitive sympathetic nervous system. Amplification of the MYCN oncogene occurs in 40 to 50% of high-risk NBs and is a powerful marker of aggressive disease and poor outcome (1). Members of the MYC family of transcription factors play critical roles in many cellular processes required for tumorigenesis including cell growth and differentiation, metabolism, and genome stability. MYCN expression targeted to developing neural crest tissue in transgenic mice directly induces NB tumor formation that faithfully recapitulates the human form of the disease (2). However, MYCN is not currently directly targetable, and hence, considerable effort has been directed toward understanding the critical MYCN target genes and pathways that are important in driving tumor progression (35).

Polyamines are ubiquitous small basic molecules that bind to DNA, RNA, and proteins and are essential for cell viability, cell growth, and differentiation (6). The polyamine pathway is also a key metabolic pathway implicated in cancer growth and progression, and these molecules support many processes regulated by the MYC family of oncoproteins (7). It is well established that there are elevated polyamines in cancer cells compared to normal cells, making the polyamine pathway an attractive therapeutic target (6, 8).

Steady-state concentrations of polyamines are maintained through regulation of synthesis and catabolism, as well as polyamine uptake and export (fig. S1). The key rate-limiting enzyme in polyamine synthesis is ornithine decarboxylase 1 (ODC1), which converts ornithine to putrescine and is tightly regulated at the levels of transcription, translation, and degradation (9). ODC1 has oncogenic and transforming abilities and is one of the most well-characterized direct transcriptional targets of c-MYC/MYCN (10, 11). High ODC1 expression is found in MYCN-amplified NB tumors, and high expression predicts poor outcome, whereas disabling ODC1 inhibits NB proliferation in vitro and in a transgenic mouse model of NB (5, 12).

Clinical trials in other cancer types using single agents targeting the polyamine pathway have generally proven to be disappointing (13, 14). In this regard, use of difluoromethylornithine (DFMO), an irreversible inhibitor of ODC1, can lead to activation of compensatory mechanisms resulting in increased polyamine uptake from the microenvironment (15, 16). This suggests that targeting multiple steps in the pathway may be a more effective therapeutic approach than monotherapy. There are several ongoing and completed clinical trials for the use of DFMO in NB, including one for relapsed or refractory high-risk NB incorporating inhibition of polyamine synthesis using DFMO, in combination with a stimulator of the polyamine catabolic pathway, celecoxib, and conventional chemotherapeutic agents (NCT02030964).

A previous study has shown that combination of DFMO with the polyamine transport inhibitor, AMXT 1501, synergistically reduced cell viability in NB cells (17). Here, we have shown that MYCN coordinately regulates expression of the entire polyamine pathway to drive increased polyamines in high-risk NB through increased polyamine synthesis, decreased polyamine catabolism, and increased polyamine uptake. We show that polyamine metabolism acts as a rheostat in NBs, with strength of signaling highly correlated with tumor phenotype and outcome. We also identified SLC3A2 as the key transporter of polyamines into NB cells, serving as a mechanism to rescue tumor cells from DFMO-mediated polyamine depletion, and showed that inhibition of polyamine uptake by AMXT 1501 in combination with DFMO was effective in treating NB in mouse models.


Polyamine pathway gene expression and clinical outcome

The assessment of prognostic factors is a major objective in cancer research to gain insights into the biology of a disease (18). To further investigate the role of the polyamine pathway in NB biology, we therefore examined an expression array database of 649 primary NB samples (649 cohort) for their prognostic impact. High expression of each polyamine biosynthetic gene [ODC1, spermidine synthase (SRM), spermine synthase (SMS), adenosylmethionine decarboxylase (AMD1), and antizyme inhibitor (AZIN1)] and spermidine oxidase (SMOX), which returns a highly bioactive polyamine, was prognostic of poor event-free survival (EFS) and overall survival (OS) (Fig. 1A, fig. S2A, and Table 1). Coordinately, low expression of the genes driving catabolic flux or antagonizing ODC1 activity [spermine/spermidine N1-acetyltransferase (SAT1), polyamine oxidase (PAOX), antizyme 1 (OAZ1), OAZ2, and OAZ3] was prognostic of poor EFS and OS (Fig. 1B, fig. S2B, and Table 1). High expression of each biosynthetic gene and low expression of each catabolic gene were also associated with MYCN amplification and unfavorable tumor stage (Table 2). In the same 649 cohort, comparison of MYCN-amplified and nonamplified patients revealed that expression of each biosynthetic gene was significantly higher and that expression of each catabolic gene was significantly lower in MYCN-amplified versus nonamplified samples (P < 0.0001 in all cases; Fig. 1C). When established clinical markers for NB, MYCN amplification status, International Neuroblastoma Staging System (INSS) tumor stage, and age at diagnosis were included with polyamine gene expression (high versus low) in a Cox proportional model, several genes retained independent prognostic significance including ODC1 (EFS, P = 0.001; OS, P = 0.003), SMS (EFS, P = 0.050; OS, P = 0.020), AZIN1 (EFS, P = 0.039; OS, P = 0.011), SRM (EFS, P = 0.016), OAZ2 (EFS, P = 0.004), and OAZ1 (OS, P = 0.012) (Table 1 and table S1).

Fig. 1 Prognostic significance of polyamine pathway gene expression and association with MYCN gene amplification in NB.

Kaplan-Meier survival curves of EFS for 628 patients with NB used in this study, according to expression of (A) the biosynthetic genes, and (B) the catabolic genes of the polyamine pathway. P values were calculated using two-sided log-rank tests. (C) Association between MYCN gene amplification status and expression of the polyamine pathway genes in the 649-patient cohort. The red bar is the median value, and two-tailed Mann-Whitney tests were used to determine significance. ***P < 0.0001.

Table 1 Univariate and multivariate Cox regression analysis of factors prognostic for outcome in a cohort of patients with NB.

Relative hazards were calculated as the antilogs of the regression coefficients in the proportional hazards regression. Univariate: log-rank tests; Multivariate: Cox proportional hazards analysis adjusted for MYCN status (amplified versus nonamplified), tumor stage (favorable versus unfavorable), and age at diagnosis (<18 months versus ≥ 18 months). Each row is a separate model. See table S1 for the P values and hazard ratios associated with MYCN status, tumor stage, and age in each model after multivariate analysis. 95% CI, 95% confidence interval.

View this table:
Table 2 Clinical and molecular characteristics of NB tumors from a cohort of 649 patients.

Categories are overlapping. For ODC, SMS, SRM, AMD1, AZIN1, SMOX, and SLC3A2, expression was considered high or low in relation to the upper quartile, and for SAT1, PAOX, OAZ1, OAZ2, and OAZ3, expression was considered high or low in relation to the lower quartile. Expression data was available for all 649 study patients.

View this table:

For those patients with tumors lacking MYCN amplification, most of the polyamine genes were prognostic of EFS (6 of 11) and OS (8 of 11) (table S2), a finding that is consistent with reports suggesting that both c-MYC (19, 20) and MYCN (21) frequently drive NB progression in high-risk patients with no MYCN amplification. These results are consistent with the hypothesis that targeting the polyamine pathway both in patients with and without MYCN-amplified tumors could be an effective treatment strategy.

Identification of the membrane transporters involved in polyamine uptake

Polyamine transport is an important mechanism of polyamine homeostasis. Although a number of transporters have been implicated in polyamine transport (table S3), those involved in uptake in mammalian cells remain poorly defined; hence, we sought to define the specific transporter(s) involved in polyamine uptake in NB. Expression of candidate transporters (2237) was determined in 14 patient-derived xenografts (PDXs) of NB after RNA sequencing (Fig. 2A). Expression of only four of the transporters was detectable across all samples [solute carrier family 7 member 1 (SLC7A1), SLC7A2, SLC3A2, and Glypican 1 (GPC1)], and a high expression of three of these four transporters (SLC7A1, SLC3A2, and GPC1) was corroborated in the 649 cohort (Fig. 2B). Last, in a MYCN transgenic NB mouse model (TH-MYCN+/+), where tumors that recapitulate human NB spontaneously arise with 100% penetrance as a result of overexpression of the MYCN transgene, again, the expression of SLC7A1, SLC3A2, and GPC1 was higher compared to that of other transporters (Fig. 2C).

Fig. 2 Expression of putative polyamine transporters and their prognostic impact in NB.

Expression of all potential polyamine transporters implicated in the literature in (A) a panel of 14 PDX models of NB, determined by RNA sequencing (TPM, transcripts per million); (B) the tumors of the 649-patient cohort of NB; and (C) the tumors that spontaneously arise in the TH-MYCN mouse model of NB at 6 weeks of age, determined by microarray (tumors from four mice). (B and C) Box and whisker plots display the lower quartile, median, and upper quartile within the box, and lines indicate the variability outside of the upper and lower quartiles, extending to the maximum and minimum measurements. (A to C) Differences between transporter expressions were assessed by one-way analysis of variance (ANOVA) and Tukey’s multiple comparison tests. NS, not significant. (D) Kaplan-Meier survival curves of EFS in 628 of 649 patients in the study cohort according to expression of SLC3A2, SLC7A1, SLC7A2, and GPC1. Patients were dichotomized around the upper quartile for SLC3A2, SLC7A1, and SLC7A2 and the lower quartile for GPC1. P values were calculated using two-sided log-rank tests. (E) Associations between MYCN gene amplification status and expression of the transporter genes in the same cohort. The median value is indicated by the red bar, and two-tailed Mann-Whitney tests were used to determine significance. *P < 0.05, **P < 0.01, and ***P < 0.001.

We examined whether increased expression of the transporter genes potentially responsible for polyamine uptake would also be predictive of patient outcome. Of the transporters listed in table S3, high expression of SLC7A1, SLC7A2, and SLC3A2 was associated with poor survival (Fig. 2D), as was high expression of SLC12A8 and SLC22A16 (fig. S3). However, SLC12A8 and SLC22A16 were not differentially expressed in each dataset (Fig. 2, A to C) and are therefore less likely to be playing a key role in polyamine uptake in NB. In addition, low rather than high expression of GPC1 was prognostic of poor outcome (Fig. 2D). As might be anticipated given these results, there was an increased expression of SLC7A1, SLC7A2, and SLC3A2 in MYCN-amplified patients and decreased GPC1 expression in the 649 cohort (P < 0.0001 in all cases; Fig. 2E).

DFMO treatment leads to up-regulation of polyamine uptake via transporter activity (8, 15). Treatment of MYCN-amplified BE(2)-C and KELLY NB cell lines with DFMO resulted in increased SLC3A2 protein expression, whereas no effect on SLC7A1, SLC7A2, and GPC1 was observed (Fig. 3, A and B, and fig. S4). Increased SLC3A2 protein expression after DFMO treatment was also observed in MYCN+ SHEP Tet-21/N cells (Fig. 3C) and in tumors from DFMO-treated TH-MYCN+/+ mice (fig. S5). While the exact mechanism of up-regulation of SLC3A2 after DFMO treatment remains to be established, the DFMO-induced increase in SLC3A2 expression was abrogated after both short interfering RNA (siRNA)–mediated MYCN knockdown in BE(2)-C and KELLY cells (fig. S6) and MYCN down-regulation in the MYCN-inducible Tet-21/N NB cell line (Fig. 3C).

Fig. 3 Identification of SLC3A2 as the key membrane transporter involved in polyamine uptake in NB.

(A) MYCN-amplified BE(2)-C and KELLY human NB cells were treated with 1, 5, or 10 mM DFMO for 24, 48, and 72 hours, and the effect on SLC3A2, SLC7A1, SLC7A2, and GPC1 protein expression was determined by Western blot. For SLC3A2, both bands correspond with the predicted band sizes of 80 to 105 kDa, whereas for SLC7A1, the top band corresponds with the correct size of 68 kDa. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (B) Quantification of SLC3A2 from (A). Replicates were normalized to control. Two-way ANOVA with Dunnett’s multiple comparisons tests were used to determine significance between treatment and control for each time point. (C) Treatment of SHEP Tet21/N cells with 1, 5, and 10 mM DFMO for 72 hours and the effect on SLC3A2 protein expression, with quantification shown, in the absence (MYCN+) or presence (MYCN−) of doxycycline (Dox; 2 μg/ml). Statistical analysis was as described for (B). (D) Knockdown of SLC3A2 protein using two independent siRNAs (#1 and #2) in BE(2)-C and KELLY cells and the effect on intracellular accumulation of radiolabeled spermidine (60-min incubation time). Two-tailed t tests were used to determine significance. Each replicate was normalized to control. Each experiment was performed in triplicate, and error bars represent SEM. Control, nontargeting control siRNA. *P < 0.05, **P < 0.01, and ***P < 0.001.

To directly assess the effect of SLC3A2 on polyamine uptake, radiolabeled spermidine was measured after SLC3A2 knockdown using two independent siRNAs. As shown in Fig. 3D, a significant decrease in radiolabeled spermidine uptake was observed after SLC3A2 knockdown [BE(2)-C, P < 0.01; KELLY, P < 0.01], whereas depletion of SLC7A1 and GPC1 had no effect (fig. S7, A to D). Although SLC7A2 knockdown also decreased spermidine uptake (fig. S7, A to D), expression of this transporter is low in NB, and there was no observable increase in SLC7A2 protein expression after DFMO treatment (Fig. 3A and fig. S4), suggesting that it is unlikely to be the key transporter for polyamine uptake. No effect on spermidine uptake was observed after silencing of SLC12A8 and SLC22A16 in BE(2)-C cells, in agreement with their negligible expression in NB tumors (fig. S7, A and B).

To confirm the positive association of SLC3A2 protein expression with MYCN, a tissue microarray (TMA) consisting of tumors from 10 PDXs and 7 cell line xenografts was stained and scored for SLC3A2 and MYCN expression. Clear membrane staining was observed for SLC3A2, and this was markedly stronger in the MYCN-amplified samples (Fig. 4). Thus, using gene expression profiles from human and transgenic murine NBs, as well as associations with patient outcome and MYCN amplification status, and direct genetic and biochemical functional evidence, we identified SLC3A2 as a major contributor to polyamine uptake in NB cells. The importance of SLC3A2 in this role was further highlighted by the finding that high expression of this gene was independently predictive of poor outcome in patients with NB [EFS: P = 0.031, hazard ratio (HR) = 1.40 (1.03 to 1.91); OS: P = 0.005, HR = 1.65 (1.16 to 2.35); table S1].

Fig. 4 MYCN and SLC3A2 protein expression determined by immunohistochemical analysis of a NB TMA.

A TMA consisting of primary patient-derived NB xenografts (n = 10) and xenografts of human NB cell lines (n = 7) was stained with antibodies to either MYCN or SLC3A2. (A) MYCN-nonamplified samples include two NB cell line xenografts, SH-SY5Y and SK-N-AS, and a primary NB PDX, RA-13. MYCN-amplified samples include two NB cell line xenografts, KELLY and CHP-134, and a primary NB PDX, A6912. Photos were taken at 600× magnification. H&E, hematoxylin and eosin. (B) MYCN and SLC3A2 expression were scored for all samples in the TMA and compared in MYCN-nonamplified and MYCN-amplified samples. Mann-Whitney tests were performed to determine significance.

MYCN regulation of polyamine pathway genes

Given the strong association between MYCN and genes within the polyamine pathway, we conducted a series of experiments to investigate the possible regulation of these genes by MYCN. Examination of the relationship between MYCN expression or activity with expression of the polyamine pathway genes demonstrated positive correlations of both MYCN gene expression and MYCN activity with all the biosynthetic genes and with SLC3A2, and negative correlations with the catabolic genes, in the 649 cohort (fig. S8). Overall, correlations with MYCN activity were stronger than correlations observed with MYCN expression (P = 0.025, two-tailed paired t test, for the comparison between all R values for MYCN activity and polyamine gene correlations and those for MYCN expression and polyamine gene correlations; fig. S8, B and C, respectively). In MYCN-inducible Tet-21/N cells, down-regulation of MYCN led to reduced expression of biosynthetic ODC1, SRM, SMS, and AMD1 and the SLC3A2 transporter and an up-regulation of catabolic SAT1 (Fig. 5A and fig. S9A). A similar pattern of expression was observed after silencing of MYCN using siRNA in BE(2)-C cells (fig. S9B), although MYCN knockdown resulted in a large increase in SMOX expression. A reduction in SLC3A2 protein expression was also observed after 48 hours of MYCN knockdown in BE(2)-C and KELLY cells (fig. S10).

Fig. 5 Regulation of biosynthetic SRM, catabolic SAT1, and the polyamine transporter SLC3A2 genes by MYCN in human NB.

(A) The effect of MYCN down-regulation in Tet-21/N NB cells by addition of tetracycline (Tet-off system) on SRM, SAT1, and SLC3A2 expression at 12, 24, and 48 hours after tetracycline administration. (B) Polyamine gene promoters for SRM, SAT1, and SLC3A2 were cloned upstream of the pGL3basic-Luciferase vector, and luciferase activity was monitored in Tet-21/N MYCN(+/−) cells. RLU, relative luciferase units. (A and B) Significance was determined using two-tailed unpaired t tests (*P < 0.05, **P < 0.01, and ***P < 0.001). (C) Dual cross-linking ChIP in BE(2)-C cells. For catabolic SAT1, Sp1 binding at the promoter in conjunction with MYCN was determined. Experiments were performed in triplicate, and error bars represent SEM.

To determine whether the close associations between MYCN expression/activity and polyamine pathway gene expression observed above were likely to be due to direct regulation of these genes by MYCN, promoter regions of the polyamine genes were cloned into a luciferase reporter vector and analyzed for activity after transfection into Tet-21/N cells. After induction of MYCN expression, increased promoter activity of ODC1, SRM, SMS, and AMD1 was observed (Fig. 5B and fig. S11), suggesting that MYCN transcriptionally activates these genes. Of the catabolic genes, reduced promoter activity of SAT1, OAZ1, and OAZ2 was observed after MYCN induction (Fig. 5B and fig. S11), suggesting that MYCN transcriptionally represses these genes. Despite seeing a large increase in PAOX and OAZ3 upon MYCN down-regulation in the same Tet-21/N cells, there was no change in promoter activity in MYCN(+) compared to MYCN(−) cells (fig. S11). A large increase in SMOX promoter activity was seen in MYCN(−) cells, again suggestive of negative regulation by MYCN. These findings are consistent with the hypothesis that MYCN is regulating the expression of the polyamine pathway genes.

Further evidence in support of this hypothesis was obtained from chromatin immunoprecipitation (ChIP) analysis. Bioinformatic scanning of the polyamine gene promoters revealed the presence of canonical and noncanonical E-boxes, the characteristic MYCN/ Myc-associated factor X (MAX) binding motif (fig. S12 and table S4). To determine whether MYCN can directly target these core promoter regions, standard and dual cross-linking ChIP was performed in Tet-21/N, BE(2)-C, and LAN-1 cells. The MYCN/MAX complex specifically and directly bound each biosynthetic gene promoter, the SLC3A2 promoter (Fig. 5C and fig. S13), and the promoters of the catabolic genes SAT1, PAOX, OAZ1 [BE(2)-C cells only], and OAZ2 (Tet-21/N and LAN-1 cells only) (fig. S13). We have previously reported that MYCN can repress gene transcription indirectly, by interacting with the transcription factor Sp1 (38). We investigated Sp1 binding in the catabolic genes that are negatively regulated by MYCN and found that Sp1 robustly bound DNA at the same regions occupied by MYCN for four of the five genes (Fig. 5C and fig. S13). There was little binding of the OAZ3 promoter by MYCN in any cell line. These findings were confirmed using a publicly available ChIP sequencing dataset (39) for three NB cell lines with varying amounts of MYCN (40). The results demonstrated that MYCN bound each polyamine gene promoter most strongly in BE(2)-C cells, which express the highest amounts of MYCN, and with a much lower intensity in the MYCN-nonamplified SHEP cells (fig. S14). MYCN binding to the promoters of several polyamine pathway mouse ortholog genes in the TH-MYCN mouse cells was also observed (fig. S14D).

We next investigated whether c-MYC also regulated SLC3A2, since ODC1, SRM, and AMD1 are known c-MYC target genes (10, 11). After down-regulation of c-MYC in P493-6 cells, which express this gene under a TET-off promoter, SLC3A2 was down-regulated, as well as expression of most of the biosynthetic genes (fig. S15A). As expected, increased expression of most of the catabolic genes was observed in response to c-MYC loss. ChIP in both P493-6 cells and c-MYC–overexpressing H1299 cells confirmed binding of c-MYC at the SLC3A2 promoter (fig. S15, B and C).

To support a functional role for MYCN/c-MYC in driving polyamine production in NB, we confirmed that chromatin at the promoter regions of the polyamine genes was in a favorable conformation for MYCN binding and transactivation and confirmed the absence of methylation in these promoter regions. ChIP was performed using antibodies recognizing acetylated histone H3 and dimethylated lysine 4 on histone 3 (active chromatin markers) and trimethylated lysine 9 on histone 3 (inactive marker). Favorable chromatin conformation markers were detected for all polyamine genes, including SLC3A2, in two NB cell lines, with the exception of SAT1 and OAZ3 in the Tet-21/N cells only, where low or no MYCN binding was also detected (Table 3 and fig. S16). In addition, silencing of gene promoters by methylation was assessed by methyl DNA immunoprecipitation (Table 3 and fig. S17), and again, no promoter methylation was seen for any biosynthetic gene or SLC3A2. Of the catabolic genes, the SAT1 promoter was methylated in Tet-21/N cells only, suggesting a cell line–specific feature, which instead favors a lack of active chromatin formation and weak MYCN promoter binding (fig. S17 and Table 3). Although MYCN and Sp1 binding and a favorable chromatin conformation were detected at the PAOX promoter, there was methylation in the promoter region in all three cell lines for this gene (fig. S17 and Table 3). Because PAOX expression is also negatively affected by MYCN down-regulation, it is unlikely that, in this case, methylation affects the binding of MYCN to the PAOX promoter. Overall, these data, which are summarized in Table 3, highlight the coordinate regulation of the polyamine pathway by MYCN/c-MYC, supporting our hypothesis that these oncoproteins drive high amounts of polyamines in MYCN/c-MYC–driven cancer.

Table 3 Summary of clinical and molecular data of each of the genes involved in regulating intracellular polyamines.

Blue shading indicates that high expression is prognostic of poor outcome or positive regulation by MYCN. Green shading indicates that low expression is prognostic of poor outcome or negative regulation by MYCN. Gray shading is indicative of an active chromatin conformation or no promoter methylation. For ChIP studies, Sp1 binding in conjunction with MYCN was only investigated for catabolic genes that are negatively regulated by MYCN. Prognosis data (EFS) and gene expression data are from the 649-patient cohort (for which EFS data were available for 628 patients). H, high; L, low; +ve, positive; −ve, negative; Y, yes; N, no.

View this table:

Targeting intracellular polyamines

We next generated a polyamine signaling score for each tumor in the 649 cohort expression array database, as a weighted average of expression decile (high for high expression of biosynthetic genes and SLC3A2 and low for high expression of catabolic genes). The polyamine signaling score was highly prognostic for outcome with worsening EFS or OS with each increasing quartile (Fig. 6A). Given that ODC1 is rate limiting for polyamine synthesis and that SLC3A2 is the predominant regulator of polyamine transport or uptake, we assessed whether these genes alone held prognostic value. Multivariate analysis including ODC1 and SLC3A2 as the only two variables in the Cox model revealed that both genes were independently predictive of poor EFS and OS (table S5). Furthermore, these two genes together were superior in predicting outcome by comparison with all 12 genes involved in polyamine regulation combined (Fig. 6A), as demonstrated by the reduced Akaike information criterion (AIC) of 2587.47 for ODC1 and SLC3A2 compared to 2619.28 for all 12 genes (P < 0.001 in both cases). This was also the case in the clinically relevant subset of patients with stage 4 disease (AIC of 1224.47 for ODC1 and SLC3A2, compared to 1248.20 for the 12-gene signature; Fig. 6A).

Fig. 6 Combined inhibition of polyamine synthesis and uptake in NB cell lines.

(A) Association between polyamine score, where a high score is indicative of high production of polyamines, and EFS, with all biosynthetic genes, catabolic genes, and SLC3A2 included in the model (12-gene signature, left) and for ODC1 and SLC3A2 only (ODC1 and SLC3A2 signature, right) in the overall cohort (top) and in stage 4 patients (bottom). Kaplan-Meier plots were stratified by expression quartiles of the polyamine score. The AIC was calculated, where a lower AIC is indicative that the combination is more predictive of outcome. (B) Colony assays after treatment with the ODC1 inhibitor, DFMO, and/or the polyamine transport inhibitor, AMXT 1501, in BE(2)-C and KELLY cells. Synergy was determined using CalcuSyn, where a CI < 0.9 is indicative of a synergistic effect. (C) Effect of the combination of DFMO and AMXT 1501 for 72 hours on radiolabeled spermidine uptake (60-min incubation time). Two-way ANOVA was used to test for an interaction between DFMO and AMXT 1501 (DFMO: F2,24 = 14.28, P < 0.001; AMXT: 1501 F3,24 = 48.82, P < 0.001; DFMO/AMXT 1501 interaction: F6,24 = 3.55, P = 0.012, followed by Tukey’s multiple comparison tests). Experiments were performed in triplicate, and error bars represent SEM. *P < 0.05, **P < 0.01, and ***P < 0.001.

Given these results, we investigated the potential for synergism in combining inhibition of polyamine synthesis and transport. The polyamine transport inhibitor AMXT 1501 (41) completely blocked radiolabeled spermidine uptake in multiple NB cell lines tested (fig. S18), whereas the combination of AMXT 1501 and DFMO was synergistic in reducing colony formation in BE(2)-C and KELLY cells, as shown by a combination index (CI) of less than 0.9 for each of the cell lines [BE(2)-C, CI = 0.51 ± 0.10; KELLY, CI = 0.19 ± 0.01; Fig. 6B]. Consistent with these findings, knockdown of SLC3A2 or of MYCN (fig. S10) similarly synergized with DFMO in reducing the number of colonies, in both BE(2)-C and KELLY cells, below that observed with either treatment alone (fig. S19). We also showed that DFMO treatment led to a compensatory increase in uptake of radiolabeled spermidine but that 0.1 μM AMXT 1501 was able to completely abolish this increased uptake (Fig. 6C).

We next tested DFMO and AMXT 1501 in the TH-MYCN mouse model, where we have found that ganglia from wild-type mice display lower quantities of polyamine metabolites by comparison with tumors from littermates homozygous for the MYCN transgene (fig. S20), further supporting a requirement for increased polyamine production during NB development. This combination was given both preemptively and in a treatment setting. Not only was tumor formation prevented in 40% of mice when given prophylactically but also tumor-free survival was significantly extended in those mice developing tumors (P < 0.001 compared to vehicle control and single agents; Fig. 7A). The combination also significantly extended survival when given to mice with established tumors that are lethal within 10 days in the absence of effective therapy (P < 0.005 compared to vehicle control and single agents; Fig. 7B). By comparison with the tumors of vehicle-treated mice, both putrescine and spermidine concentrations were significantly reduced in the tumors of mice receiving the combination (P < 0.001), with spermidine depleted to a greater extent than after treatment with DFMO alone (P < 0.001), consistent with on target activities (Fig. 7C).

Fig. 7 Effect of combined inhibition of polyamine synthesis and uptake in the TH-MYCN+/+ mouse model of NB.

(A to B) Kaplan-Meier survival curves showing the survival of TH-MYCN+/+ transgenic mice after treatment with DFMO and AMXT 1501 at weaning (3 weeks of age), before detectable tumor formation (A), or after detection of a small palpable tumor (B). (C) Concentration of putrescine, spermidine, and spermine present in the tumors of mice at the time of sacrifice after treatment with DFMO and/or AMXT 1501 from the time of detection of a small palpable tumor. Two mice on DFMO/AMXT 1501 were excluded from analysis because they were no longer receiving AMXT 1501 at the time of cull. One-way ANOVA and Tukey’s multiple comparison tests were used to determine significance. (D) Kaplan-Meier survival curves showing the survival of TH-MYCN+/+ transgenic mice after treatment with either cyclophosphamide/topotecan (Cyclo/Topo), the four-drug combination currently in clinical trial (Cyclo/Topo/celecoxib/DFMO) or with the “clinical trial” protocol with AMXT 1501 either added or substituted for celecoxib. (E) Kaplan-Meier survival curves showing the survival of TH-MYCN+/+ transgenic mice after treatment with either irinotecan/temozolomide (Irino/Tmz), the four-drug combination Irino/Tmz/celecoxib/DFMO, or this four-drug combination with AMXT 1501 either added or substituted for celecoxib. (A, B, D, and E) Mice received AMXT 1501 for 42 days, and the gray arrow indicates the time point when AMXT 1501 treatment was stopped. DFMO (1%) was given continuously in drinking water. Two-tailed log-rank tests were used to test for significance between treatment groups. *P < 0.05, **P < 0.01, and ***P < 0.001.

We also tested DFMO with a dicaprate salt formulation of AMXT 1501 in a PDX model of NB. This formulation behaves similarly to the AMXT 1501 tetrahydrochloride salt formulation used in the previous experiments (fig. S21, A and B) but has improved bioavailability and is the salt used in initial clinical trials. In Balb/c nude mice xenografted with the postrelapse human NB PDX, COG-N-440x, the DFMO/AMXT 1501 combination significantly extended survival (P < 0.01; fig. S21C). When DFMO and AMXT 1501 were given in combination with clinically used chemotherapy in TH-MYCN+/+ mice (cyclophosphamide and topotecan or the alternative chemotherapy backbone of irinotecan and temozolomide), the increase in survival was significantly improved compared to cyclophosphamide/topotecan/celecoxib/DFMO (which is currently in clinical trials for children with NB) (P = 0.031; Fig. 7D) or irinotecan/temozolomide/celecoxib/DFMO (P = 0.014; Fig. 7E). Celecoxib has been combined with DFMO because it induces SAT1, the enzyme that acetylates polyamines for export from the cell and is rate limiting for catabolic flux of polyamines (42). Here, we show that the addition of AMXT 1501 to these four-drug combinations extends survival with less than 50% of mice displaying any tumor progression and that the rest remains tumor free when the experiment was terminated at 20 weeks [P = 0.005 (Fig. 7D) and P < 0.001 (Fig. 7E)]. In the COG-N-440x PDX model, the combination of cyclophosphamide/topotecan/AMXT 1501/DFMO significantly extended survival compared to cyclophosphamide/topotecan (P < 0.0001) and was no different from animals treated with cyclophosphamide/topotecan/celecoxib/DFMO (fig. S21D). DFMO and AMXT 1501 gave the same magnitude of antitumor benefit as the chemotherapy regimen (Fig. 7, D and E, and fig. S21D).

No weight loss, or any other signs of toxicity, was observed in any treatment group with the exception of the cyclophosphamide/topotecan/AMXT 1501/DFMO or cyclophosphamide/topotecan/AMXT 1501/celecoxib/DFMO groups, where some weight loss was evident, requiring euthanasia of a small number of mice (1 of 20 in the four-drug combination and 3 of 20 in the five-drug combination) (fig. S22).


The results in this study indicate a high level of orchestration of the polyamine pathway by MYCN in NB. We demonstrated that the entire metabolic pathway is under transcriptional control of the MYCN oncoprotein, including SLC3A2 for which we provide evidence for it being the major transporter involved in polyamine uptake in NB. This study also demonstrates the clear benefit of combining inhibition of polyamine synthesis with blockade of polyamine transport in transgenic and PDX models of NB, an approach that is highly effective at depleting polyamines from the cell, because polyamine uptake from the microenvironment provides a rescue pathway for polyamine-dependent tumor cells. Current clinical trials have used cyclooxygenase inhibitors to induce SAT1 to augment polyamine acetylation and export in the absence of a credentialed polyamine transport inhibitor, and future trials may benefit from combination with agents such as AMXT 1501.

Despite promising preclinical results, DFMO as a monotherapy for established cancers has met with limited success in clinical trials and led to a focus on the use of DFMO as a chemopreventive agent where it has been shown to be effective in preventing colorectal cancer (when given in combination with SAT1-inducing nonsteroidal anti-inflammatory drugs), prostate cancer, and skin cancer (4345). Prophylaxis treatment of TH-MYCN+/+ mice, a NB model with 100% lethal tumor penetrance, with DFMO and AMXT 1501 prevented NB formation in 40% of mice, evidence that polyamines are critical for NB initiation and that polyamine depletion in this childhood disease may have a chemopreventive role.

Unlike prokaryotes and unicellular eukaryotes (46, 47), the polyamine transport system in mammalian cells is still poorly defined. Our results provide strong evidence that SLC3A2 is the key player in polyamine uptake in NB cells. SLC3A2 forms the heavy chain of the CD98 glycoprotein and functions as a heterodimer with partners that transport different molecules. It is widely overexpressed in cancer cells, where it is also associated with poor prognosis and progressive or metastatic tumors (4852). CD98 contributes to cell survival and proliferation through well-studied roles in amino acid transport, which requires interaction with a CD98 light chain, and by amplifying integrin signaling, which results in reduced anchorage dependence (48, 53, 54). Other prosurvival functions may involve modulation of the tumor microenvironment and of adaptive immunity by supporting lymphocyte clonal expansion (48, 55). Previous studies identified a role for SLC3A2 as part of an arginine-putrescine antiporter in colon cancer–derived cells where it interacts with a Y+ L-type amino acid transporter light chain, such as SLC7A5, and exports putrescine under normal conditions (30). However, under specific concentration gradients of low intracellular putrescine, this polyamine was taken up (31), suggesting a bidirectional role in polyamine transport and a similar mechanism for the increased spermidine uptake that we observed after DFMO treatment may also be likely.

Our data support a role for SLC3A2 in polyamine uptake and indicate that this might therefore be another important mechanism by which SLC3A2 promotes tumor formation. Because of its multiple roles in promoting cell proliferation and tumor formation, blocking this protein has potential clinical value, and studies using CD98 antibodies have led to recent clinical trials in acute myeloid leukemia (56). Both SLC7A5 and SLC7A6 dimerize with SLC3A2 (57, 58) and are highly prognostic of poor outcome in NB [Kocak dataset, R2: Genomics Analysis and Visualization Platform (]. Further studies are required to determine whether either of these is the light chain dimerization partner of SLC3A2 required for polyamine uptake. AMXT 1501 is a lipophilic polyamine mimetic, which potently inhibits cellular spermidine uptake in the low nanomolar concentration range without crossing the cell membrane (41). A previous study has found that AMXT 1501 and DFMO were synergistic in vitro in NB cells (17), a finding we have not only confirmed but also extended into NB mouse models. When administered as a monotherapy, AMXT 1501 had no antitumor effect, consistent with the notion that only in a setting of polyamine depletion by DFMO is the rescue pathway by the transporter important. When given to mice with established tumors, DFMO and AMXT 1501 together were as effective at delaying tumor formation as chemotherapy alone (cyclophosphamide/topotecan and temozolomide/irinotecan) in complementary mouse models of NB. Whereas DFMO treatment resulted in decreased concentrations of putrescine and spermidine in tumors of treated mice, the combination of DFMO and AMXT 1501 further depleted spermidine by comparison with DFMO alone.

Among the polyamines, spermidine retains a unique function beyond that of a chaperone in that it is solely required for the hypusine modification necessary to activate eukaryotic translation initiation factor 5A (eIF5A), a key mammalian protein translation factor (59). Depletion of spermidine thereby uniquely inhibits a pathway-mediating protein translation downstream of MYC signaling. In addition, DFMO-induced lowering of putrescine and spermidine has been shown to markedly increase dcAdoMet, the cofactor used by the enzymes SRM and SMS to produce spermidine and spermine (60). The amount of spermine remained unchanged, and spermine has been shown to be more tightly bound to acidic sites on macromolecules compared to spermidine and putrescine, and has been postulated to act as a storage molecule for quick transfer to spermidine (61). Consistent with this, high expression of SMOX, a highly inducible enzyme responsible for catalyzing the oxidative cleavage of spermine to spermidine, was prognostic of poorer survival, suggesting that the intratumoral concentration of spermidine may be more important than spermine in promoting aggressive disease. SMOX was increased in MYCN-amplified tumors but appeared to be negatively regulated in cell lines and may therefore work to maintain a fine balance between spermine and spermidine that may not be as critical in an artificial system.

In a similar fashion to MYCN, c-MYC expression has been shown to be highly correlated with basal polyamine transport activity (62, 63), and previous studies have shown that ODC1, SRM, and AMD1 are direct transcriptional targets of c-MYC (10, 11). We have now shown that SLC3A2 is also a target gene of c-MYC, providing further evidence that the combination of DFMO and AMXT 1501 might have therapeutic potential in adult c-MYC–driven cancers. In addition, c-MYC has been shown to be a potent transforming gene in a subset of high-risk NB cases and predominantly promotes NB progression in stage 4 MYCN-nonamplified NB (19, 20), suggesting that this subgroup might also be sensitive to this combination and demonstrating the wide clinical potential of this specific polyamine depletion therapy.

There is an urgent need for new and effective therapies for the treatment of patients with high-risk NB, where survival rates remain below 50% (64). The current clinical trial protocol of DFMO, celecoxib, cyclophosphamide, and topotecan (NCT02030964) depletes polyamines by targeting two steps in the pathway but does not prevent uptake from the microenvironment, indicating that this protocol can be further optimized. As demonstrated in this study, combining inhibition of polyamine synthesis and transport, together with conventional chemotherapy, resulted in an increased survival benefit for MYCN-driven NB-prone mice by comparison with the polyamine depletion protocol currently in clinical trial. Thus, this treatment approach warrants clinical investigation in children with MYCN/c-MYC–driven high-risk NB and potentially other cancers with activation of either of these oncoproteins.

Future studies will need to address some of the limitations of this study. Although evidence suggests that the combination of DFMO and AMXT 1501 would be effective in any MYC-driven cancer, it would be necessary to test this combination in a c-MYC–driven cancer. In addition, although we have shown that SLC3A2 is the main transporter of polyamines in NB, we have not directly identified an interaction between SLC3A2 and AMXT 1501, which would confirm SLC3A2 as the key transporter blocked by AMXT 1501. We also know that SLC3A2 has various binding partners and that the key binding partner required for polyamine uptake should be identified. Although safety studies for AMXT 1501 have not yet been completed in adults, the first clinical trial has recently opened (NCT03077477), bringing the future use of this promising compound for the treatment of children one step closer.


Study design

This study was designed to identify the key transporter involved in polyamine uptake in NB and to evaluate the association of expression of each polyamine pathway gene with outcome in NB, the role of MYCN in regulating the polyamine pathway, and the effect of combined inhibition of polyamine synthesis and transport in NB mouse models.

A dataset of 649 patients was used to demonstrate the clinical potential of each gene of the polyamine pathway and to look for an initial association between expression of these genes and MYCN. This association was further evaluated using molecular studies to look at direct regulation by MYCN of the polyamine pathway genes. To identify the key transporter in polyamine uptake, expression of transporters implicated in polyamine uptake in the literature was explored in PDX tumors by RNA sequencing, in tumors taken from TH-MYCN mice, in a mouse model that strongly replicates MYCN-driven NB, and in the 649-patient NB cohort. Further experiments to identify the key transporter included the effect of DFMO treatment on expression, because we know that DFMO enhances polyamine uptake and the effect of knockdown on radiolabeled spermidine uptake. To evaluate the effect of combined inhibition of polyamine synthesis and transport, the TH-MYCN mouse model and a human PDX model of NB was used. We conducted preclinical testing of combined DFMO and AMXT 1501 in conjunction with currently used chemotherapeutics for NB and compared the effectiveness to the polyamine depletion protocol currently in clinical trial. In vitro studies were conducted with multiple technical and biological replicates to ensure reproducibility of data. Sample sizes for animal models were determined on the basis of extensive previous experience with similar experimental types, and animals were randomly allocated to groups. Raw data are shown in table S10.

Statistical analysis

EFS and OS were determined, and univariate and multivariate analyses were performed using SPSS version 24, as previously described (65). Tumors were categorized as having high or low gene expression based on quartile cut points, as previously described (65), with biosynthetic genes and SLC3A2 cut at the upper quartile and catabolic genes at the lower quartile. P values and hazard ratios were determined using Cox proportional hazard models.

All in vitro experiments were repeated at least three times, and the means ± SEM were calculated. Associations among gene expressions, MYCN status, and tumor stage were determined using two-sided Fisher’s exact tests. Differences between two groups were determined with two-tailed Student’s t tests (unpaired or paired where specified) or a Mann-Whitney test where specified. Differences between three or more groups were determined using one-way or two-way ANOVA followed by Tukey’s (for comparison of all means) or Dunnett’s (for comparison to control means) multiple comparison tests. To measure the relationship between MYCN activity and expression with polyamine gene expression, the Pearson correlation coefficient was determined, with a value closer to +1 indicative of a stronger positive relationship and −1 indicative of a stronger negative relationship.

In vivo studies were carried out using multiple animals (7 to 20 per group, specified in figures), and Kaplan-Meier curves were generated from survival data. Log-rank tests were used to look for statistical differences between treatment groups. In all cases, P < 0.05 was regarded as statistically significant.



Fig. S1. A schematic overview of the polyamine pathway.

Fig. S2. Prognostic significance of expression of the genes involved in the polyamine pathway in NB.

Fig. S3. Prognostic impact of SLC22A16 and SLC12A8 expression in NB.

Fig. S4. Quantification of SLC7A1, SLC7A2, and GPC1 protein expression from Western blots in Fig. 3A.

Fig. S5. Expression of SLC3A2 in tumors from TH-MYCN+/+ mice after DFMO treatment.

Fig. S6. Suppression of MYCN and the effect on DFMO-induced SLC3A2 up-regulation.

Fig. S7. Knockdown of potential polyamine transporters and the effect on radiolabeled spermidine uptake.

Fig. S8. Correlation between polyamine gene expression profiles and MYCN activity or MYCN expression in the 649-patient NB dataset.

Fig. S9. MYCN regulation of genes involved in regulating intracellular polyamines.

Fig. S10. Knockdown of MYCN results in a decrease in SLC3A2 protein expression.

Fig. S11. Promoter activity of the genes involved in regulating intracellular polyamines in MYCN(+) and MYCN(−) Tet-21/N cells.

Fig. S12. Bioinformatic scanning of the polyamine pathway gene promoters for canonical and noncanonical E-boxes.

Fig. S13. ChIP to determine MYCN and Sp1 binding to the promoters of the genes involved in regulating polyamines.

Fig. S14. Binding of MYCN to polyamine gene promoters in the Zeid ChIP-Seq dataset.

Fig. S15. Regulation of SLC3A2 and the polyamine pathway genes by c-MYC.

Fig. S16. Chromatin epigenetic markers at the promoters of the genes involved in regulating polyamines.

Fig. S17. DNA methylation at the promoters of the genes involved in regulating polyamines.

Fig. S18. Radiolabeled spermidine uptake in a panel of NB cells treated with AMXT 1501.

Fig. S19. Sensitization of NB cells to DFMO treatment after knockdown of SLC3A2 or MYCN.

Fig. S20. Basal polyamines and major polyamine metabolites are increased in TH-MYCN mice homozygous for the MYCN transgene compared to wild-type mice.

Fig. S21. Dicaprate formulation of AMXT 1501 in combination with DFMO.

Fig. S22. Weight loss and gain of TH-MYCN mice treated upon detection of a small palpable tumor.

Table S1. Multivariate analyses for polyamine pathway genes in the 649-patient cohort of NB.

Table S2. Univariate analysis of MYCN-nonamplified patients in the 649-patient cohort of NB.

Table S3. Putative polyamine transporters implicated in the literature.

Table S4. Bioinformatic scanning for canonical and noncanonical E-boxes within the promoters of polyamine regulatory genes.

Table S5. Multivariate Cox regression analysis of ODC1 and SLC3A2 expression as prognostic factors in the 649-patient cohort of NB.

Table S6. siRNA used for transporter knockdown.

Table S7. Primer pairs used for polyamine gene expression analysis.

Table S8. Primer pairs used for polyamine promoter cloning in the pGL3basic vector.

Table S9. Primer pairs used for quantitative ChIP.

Table S10. Raw data (provided as separate Excel file).

References (6677)


Acknowledgments: The Children’s Cancer Institute is affiliated with UNSW Australia and the Sydney Children’s Hospitals Network. Funding: This work was supported by grants from the National Health and Medical Research Council Australia (APP1132608 and APP1125036), Cancer Institute New South Wales (14/TPG/1-13), Cancer Council New South Wales (PG 16-01), Kids Cancer Alliance, Bright Blue Charity for Sick Children, a U.S. Department of Defense Research Award (W81XWH-1010145), a Wipe Out Kids Cancer Award, and the Italian Association for Research on Cancer (AIRC; IG-11400 and IG-15182 to G.P.). G.M.M. is a recipient of a three-year fellowship from AIRC (18170). Author contributions: M.H. and M.D.N. conceived the project. M.H., M.D.N., and G.P. supervised the project. M.R.B., M.D.H., D.S.Z., G.M.M., T.N.T., B.B.C., J.I.F., and W.B.L. were advisors on the project. A.O. and M.F. supplied data for the project. L.D.G., S.P., J.M., J.I.F., G.P., M.D.N., and M.H. designed the experiments. S.P., E.V., G.M., and S.D.G. performed MYCN regulation studies (MYCN knockdown, Tet-21/N experiments, promoter activity, and ChIP). L.D.G. optimized transporter knockdown. L.D.G., D.M.T.Y., and C.L.F. performed radiolabel experiments. J.M., G.E., S.A., and S.S. carried out animal work. L.D.G., K.M.H., and L.X. performed colony assays and Western blots after DFMO treatment or knockdown in cell lines and DFMO treatment in mice. A.J.G. performed immunohistochemical experiments. D.R.C. prepared samples both for measuring polyamine transporters in TH-MYCN mice by microarray and measuring polyamine metabolites in these mice. M.R.B. measured amounts of polyamines in mouse tumors. A.K. prepared samples for RNA sequencing. L.D.G., S.P., J.M., L.X., and A.J.G. analyzed the experiments. L.D.G. and A.J.R. performed analyses with the 649-patient cohort. L.D.G., C.M., B.L., and F.M.G. did statistical analyses for the project. L.D.G. wrote the manuscript with contributions from M.D.N., M.H., G.P., and S.P. All authors reviewed the manuscript and provided input. Competing interests: M.R.B. is employed as president and chief security officer of Aminex Therapeutics Inc. and also has ownership interest (including patents) in the same company. The original U.S. patent covering AMXT 1501 composition of matter is USRE43327E1 “Hydrophobic polyamine analogs and methods for their use” (originally issued as U.S. patent no. 6,963,010). A PCT WO patent application covers AMXT 1501 dicaprate salt forms (WO2017165313 “Bioavailable Polyamines”). Data and materials availability: The COG PDX models were obtained under a material transfer agreement (MTA) between the Children’s Oncology Group and the Children’s Cancer Institute, and AMXT 1501 was provided under an MTA between Aminex Therapeutics Inc. and the Children’s Cancer Institute. The microarray data for the 649-patient NB cohort is available at the Gene Expression Omnibus database (accession no. GSE45480) along with age, stage, and MYCN amplification status. Data for the TH-MYCN tumor microarray are available in the ArrayExpress database under accession no. E-MTAB-3247. All the remaining data used for the study are present in the main text or in the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article