Research ArticleCancer

Inhibition of the ALDH18A1-MYCN positive feedback loop attenuates MYCN-amplified neuroblastoma growth

See allHide authors and affiliations

Science Translational Medicine  19 Feb 2020:
Vol. 12, Issue 531, eaax8694
DOI: 10.1126/scitranslmed.aax8694

Interrupting a vicious cycle

Neuroblastoma is a relatively common pediatric neuroendocrine tumor. Although some neuroblastomas spontaneously regress, others can be deadly and difficult to treat, especially when they are associated with MYCN amplifications. MYCN itself has proven difficult to target, but Guo et al. have found that it may be possible to intervene indirectly, by targeting aldehyde dehydrogenase family 18 member A1 (ALDH18A1). The authors identified a positive feedback loop between ALDH18A1 and MYCN, as well as a chemical inhibitor of ALDH18A1 that can interrupt the cycle and effectively treat neuroblastoma in mouse models.

Abstract

MYCN-amplified neuroblastoma (NB) is characterized by poor prognosis, and directly targeting MYCN has proven challenging. Here, we showed that aldehyde dehydrogenase family 18 member A1 (ALDH18A1) exerts profound impacts on the proliferation, self-renewal, and tumorigenicity of NB cells and is a potential risk factor in patients with NB, especially those with MYCN amplification. Mechanistic studies revealed that ALDH18A1 could both transcriptionally and posttranscriptionally regulate MYCN expression, with MYCN reciprocally transactivating ALDH18A1 and thus forming a positive feedback loop. Using molecular docking and screening, we identified an ALDH18A1-specific inhibitor, YG1702, and demonstrated that pharmacological inhibition of ALDH18A1 was sufficient to induce a less proliferative phenotype and confer tumor regression and prolonged survival in NB xenograft models, providing therapeutic insights into the disruption of this reciprocal regulatory loop in MYCN-amplified NB.

INTRODUCTION

Neuroblastoma (NB) is a neuroendocrine tumor that arises in the developing sympathetic nervous system and is the most common extracranial solid tumor in childhood, with a median age at diagnosis of 18 months (1). NB has a broad spectrum of clinical behavior and heterogeneous biological features, ranging from spontaneous regression to progressive disease and metastasis (2, 3). A number of genetic aberrations have been extensively investigated, among which amplification of MYCN (encoding N-MYC) was found in about 20% of patients and was identified as one of the strongest determinants with prognostic value, specifically for predicting unfavorable clinical outcomes and poor survival of high-risk patients with NB (46).

As a transforming gene and oncogenic driver, N-MYC plays multiple roles in NB malignancy and functions as a master transcriptional regulator that can activate genes involved in self-renewal, proliferation, pluripotency, angiogenesis, and metastasis and suppress the expression of genes that promote differentiation, cell cycle arrest, and immune surveillance (7). Despite difficulties in targeting transcription factors, previous attempts have used small molecules or low–molecular weight compounds, such as 10058-F4 (8), NY2267 (9), IIA6B17 (10), and Tz-1 (11), to target members of the MYC family. However, because of c-MYC selectivity or impaired functional efficiency in vivo, the use of these therapeutic approaches has proven challenging in overcoming MYCN-amplified NB, and thus, accumulating reports have considered N-MYC to be “undruggable,” with limited treatment options (7). Therefore, an urgent need to develop alternative therapeutic strategies for targeting MYCN in NB remains, and exploring the upstream regulators of MYCN expression at genetic and epigenetic levels is imperative.

Tumorigenesis is dependent on metabolic reprogramming, during which alterations in intracellular and extracellular metabolites exhibit potentially long-ranging impacts on epigenomic and transcriptomic regulation (12, 13). Specifically, glutamine is the most abundant amino acid in tissue culture medium, and in the blood, it is used to excrete excess nitrogen from amino acid catabolism (14). Because of an increased need for glutathione (15), nucleotides, and amino acids (14), deregulated glutamine metabolism is critical for the establishment and maintenance of the tumorigenic state, resulting in aberrantly activated oncogenes and loss of tumor suppressors (12). Aldehyde dehydrogenase family 18 member A1 (ALDH18A1) is the key enzyme for the synthesis of proline from glutamate, which catalyzes the coupled phosphorylation and reduction conversion of glutamate to β-pyrroline-5-carboxylate (P5C) and plays a critical role in regulating glutamine metabolism (16). ALDH18A1 functions as a component in regulating the cell proliferation of melanoma via proline biosynthesis (17). However, the molecular basis of ALDH18A1’s effects on tumorigenesis, especially on MYCN-amplified NB, remains elusive and warrants further investigation.

Previous reports implicate a variety of mechanisms, including multiple genetic and epigenetic events, in the carcinogenesis of high-risk NB (18). MicroRNA (miRNA) deregulation is an important component of this landscape via both oncogenic and tumor-suppressive mechanisms (19). Studies have shown that transcription factor SP1 is a direct target of miR-29b (20), and a negative correlation between SP1 and miR-29b expression has been observed across acute myeloid leukemia (AML) patient cohorts (21). SP1 interacts with nuclear factor κB (NF-κB) and histone deacetylases (HDACs) to further inhibit miR-29b expression, thus forming a regulatory SP1/NF-κB/HDAC/miR-29b network that regulates KIT expression in AML (21). In addition, miR-29b is a direct MYCN-targeting miRNA capable of suppressing the endogenous N-MYC protein in MYCN-amplified human NB cell line Kelly (22).

In the present study, we provide mechanistic insights into the impact of ALDH18A1 activation on NB cell behavior and demonstrate that both genetic down-regulation of and pharmacological interference with ALDH18A1 by the inhibitor YG1702 attenuate the growth of MYCN-amplified NB and down-regulate MYCN. In addition, N-MYC reciprocally modulates ALDH18A1 expression by direct transcriptional activation, setting up a positive feedback loop between ALDH18A1 and MYCN.

RESULTS

MYCN amplification score (MYCN-AMP6) is correlated with the prognosis of patients with NB

By comparing the gene expression (GE) profiles of patients with NB harboring MYCN amplification and a normal copy number in four cohorts of 88, 649, 30, and 18 individuals [GSE16476 (23), GSE45547 (24), GSE13136 (25), and E-MEXP-669 (26), respectively], we obtained lists of differentially expressed (DE) genes: 72, 45, 106, and 309 genes in these respective cohorts exhibited >2-fold differences in expression (P < 0.01). The overlap of these DE genes suggested that the nine most common DE genes are associated with MYCN amplification (Fig. 1A).

Fig. 1 Analysis of NB GE identifies a six-gene prognostic signature, in which ALDH18A1 correlates with MYCN amplification.

(A) Venn diagram depicting the overlap profile of DE genes between patients with NB with and without MYCN amplification in the GSE16476, GSE45547, GSE13136, and E-MEXP-669 cohorts. (B) Heatmap showing the GE patterns of the eight overlapping DE genes (excluding MYCN) as in (A), with the 88 patients in the GSE16476 cohort ranked by MYCN-AMP6 scores. The six signature genes are depicted in red on the right. (C) Kaplan-Meier estimates of the OS (top) and EFS (bottom) according to the MYCN-AMP6 score calculated using data from the GSE16476 cohort. The OS and EFS of patients with scores above and below the cutoff (determined by the Youden index derived from ROC curve) are shown by the red and blue lines, respectively. (D) Kaplan-Meier estimates of the OS (top) and EFS (bottom) according to the MYCN-AMP6 scores calculated using data from the E-MTAB-1781, E-MTAB-161, and E-MTAB-179 cohorts. For all panels, the OS and EFS of patients with MYCN-AMP6 scores above and below the cutoff (determined by the Youden index derived from the ROC curve) in each cohort are shown by the red and blue lines, respectively. (E) GSEA plots depicting enrichment of the MYCN target gene signature, the MYCN amplification–associated gene signature, and the genes coamplified with MYCN in the transcriptional profiling of samples with high ALDH18A1 expression in the GSE16476 cohort. FDR, false discovery rate; NES, normalized enrichment score. (F) Immunoblot analysis of ALDH18A1 and N-MYC protein expression in the SK-N-BE(2) and IMR32 (MYCN-amplified) or SH-SY5Y and SK-N-SH (normal MYCN copy number) NB cell lines. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as the loading control. Representative results from four replicates are shown. MW, molecular weight. (G) IHC analysis of ALDH18A1 protein expression in tumor samples from patients with NB with or without MYCN amplification. Scale bars, 25 μm. The MYCN status was determined by FISH analysis in which MYCN was labeled with green and 2q11 (served as the internal control) was labeled with orange. Scale bars, 10 μm. (H) Quantitative analysis of ALDH18A1 protein expression in tumor samples with (n = 6) or without (n = 43) MYCN amplification calculated as integrated optical density (IOD) values using Image-Pro Plus software 6.0. Student’s t test, ***P < 0.001.

To determine the core transcriptional components of MYCN amplification that relate to clinical outcomes across a broad spectrum of patients with NB, we applied a statistical regression algorithm based on the least absolute shrinkage and selection operator (LASSO) (2729) to relate the eight overlapped DE genes (excluding MYCN) to patient survival in the GSE16476 cohort. An optimal six-gene signature was yielded, and the weighted combined GE of these six genes was calculated for each patient, generating the MYCN-AMP6 score (Fig. 1B). High MYCN-AMP6 scores were associated with poor overall survival (OS) and event-free survival (EFS) in this training cohort (Fig. 1C), reflecting a link between MYCN amplification–associated GE programs and clinical outcomes.

We next evaluated the association of the MYCN-AMP6 score with survival in three large datasets of NB cohorts: E-MTAB-1781 (30), E-MTAB-161 (31), and E-MTAB-179 (32). Patients with high MYCN-AMP6 scores had shorter OS and EFS times than patients with low scores in these three tested cohorts (Fig. 1D). High MYCN-AMP6 scores were also associated with poor prognosis in stage 4 patients (fig. S1), who have the worst survival rates and need improved therapy. Together, these results demonstrate the potential prognostic value of the MYCN-AMP6 score in patients with NB.

ALDH18A1 overexpression is associated with the genetic amplification of MYCN

Among the six genes in the MYCN-AMP6 score, ALDH18A1 encodes P5C synthase, the key enzyme of glutamic acid metabolism (16). Thus, we used mRNA profiling to examine a potential relationship between ALDH18A1 and MYCN amplification. A positive relationship between ALDH18A1 mRNA and MYCN amplification and the corresponding MYCN overexpression was observed in the GSE16476 (fig. S2A), GSE13136 (fig. S2B), GSE12460 (33) (fig. S2C), E-MEXP-669 (fig. S2D), and GSE45547 (fig. S3) cohorts and confirmed in the E-MTAB-1781 (fig. S4A), E-MTAB-179 (fig. S4B), and E-MTAB-161 (fig. S4C) cohorts. In addition, we obtained similar results from the analysis of 16 NB cell lines with different MYCN statuses derived from the Cancer Cell Line Encyclopedia (fig. S4D). To determine whether ALDH18A1 correlates with specific MYCN-associated molecular programs, we performed gene set enrichment analysis (GSEA) of the GSE16476 cohort and found enriched MYCN target genes (NMYC_01, HALLMARK_MYC_TARGETS_V1, and HALLMARK_MYC_TARGETS_V2), genes positively correlated with MYCN amplification (KIM_MYCN_AMPLIFICATION_TARGETS_UP), and genes coamplified with MYCN (LASTOWSKA_COAMPLIFIED_WITH_MYCN), all of which are molecular features associated with MYCN amplification, in samples with relatively high ALDH18A1 expression (Fig. 1E and fig. S5).

In addition, we examined ALDH18A1 expression in the MYCN-amplified human NB cell line IMR32, MYCN-amplified human NB cell line SK-N-BE(2), human NB cell line SH-SY5Y (normal MYCN copy number), and human NB cell line SK-N-SH (normal MYCN copy number) by immunoblotting. The results showed overexpression of ALDH18A1 in the MYCN-amplified NB cell lines (Fig. 1F). To better understand the relationship between ALDH18A1 and MYCN amplification, we performed immunohistochemical (IHC) analysis for ALDH18A1 protein expression across tissue samples from 49 patients with NB with MYCN statuses identified by fluorescence in situ hybridization (FISH) and confirmed a positive relationship between ALDH18A1 expression and MYCN amplification (Fig. 1, G and H), underscoring for the connection between ALDH18A1 overexpression and MYCN amplification in NB.

ALDH18A1 expression correlates with the clinicopathological characteristics of patients with NB

We next sought to determine whether aberrant ALDH18A1 expression might also be related to clinically relevant features of patients with NB. First, we examined 49 previously described sporadically occurring primary NBs and 5 normal reference tissues in which ALDH18A1 protein expression was determined by IHC staining, and the results showed ALDH18A1 overexpression in NBs (Fig. 2A). High ALDH18A1 protein expression was associated with established risk factors, including the International Neuroblastoma Staging System (INSS) stage (Fig. 2, B to D) and MYCN amplification (Fig. 1H). In addition, Affymetrix mRNA analysis of the GSE16476 cohort showed that ALDH18A1 mRNA expression was associated with age, INSS stage, and progression or relapse status (Fig. 2, E to H). Similar results were obtained in the three larger cohorts E-MTAB-161, E-MTAB-179, and E-MTAB-1781 (fig. S6A and table S1).

Fig. 2 ALDH18A1 is associated with clinical factors and the prognosis of patients with NB.

(A) Quantitative analysis of ALDH18A1 protein expression in NBs (n = 49) versus normal peripheral nerve tissue (n = 5). Mann-Whitney U test, ***P < 0.001. (B) Clinical characteristics of 49 patients with NB and low or high ALDH18A1 expression. P values were calculated using Pearson’s χ2 test and Fisher’s exact test. (C) Representative images of the IHC analysis of ALDH18A1 protein expression in patients with NB at different INSS stages shown in (B). Scale bars, 100 μm (top) and 25 μm (bottom). (D) Quantitative analysis of ALDH18A1 protein expression in NB at INSS stages 1, 2, and 4s versus stages 3 and 4. Student’s t test, **P < 0.01. (E) Correlation between ALDH18A1 mRNA expression and key clinical characteristics of patients with NB. The data are derived from the GSE16476 cohort. P values were calculated using Pearson’s χ2 test and Fisher’s exact test. Prog./Relap., progression/relapse; Non., no progression/relapse. (F to H) Quantitative analysis of ALDH18A1 mRNA expression in patients with NB as shown in (E), comparing those ≤18 months of age versus >18 months of age (F), INSS stages 1, 2, and 4s versus stages 3 and 4 (G), and without progression or relapse versus with progression or relapse (H). Student’s t test, **P < 0.01, ***P < 0.001. (I) Kaplan-Meier estimates of the OS and EFS of patients with NB with high or low ALDH18A1 expression in the GSE1476 cohort. Patients with ALDH18A1 expression above and below the cutoff (determined by the Youden index derived from the ROC curve) are shown by the red and blue lines, respectively. (J) GSEA of the correlation of the down-regulated genes in NB samples with poor survival and low ALDH18A1 expression.

Moreover, Kaplan-Meier analysis of the GSE16476 cohort and all three tested cohorts showed that high ALDH18A1 expression was associated with poor OS and EFS, whereas low ALDH18A1 expression was correlated with favorable clinical outcomes (Fig. 2I and fig. S6B). High ALDH18A1 expression was also associated with poor prognosis in stage 4 patients (fig. S6C). GSEA showed that low ALDH18A1 expression was correlated with down-regulation of genes associated with a poor survival prognosis in patients with NB (ASGHARZADEH_NEUROBLASTOMA_POOR_SURVIVAL_DN) (34) (Fig. 2J), suggesting that ALDH18A1 may be a predictor of patient survival.

In the multivariate survival analysis using Cox proportional hazards (CPH) models, high ALDH18A1 expression remained a risk factor for adverse outcomes with respect to both EFS and OS in the GSE16476 cohort (tables S2 and S3). Similarly, ALDH18A1 retained prognostic value for NB risk estimation in CPH models of E-MTAB-1781 based on EFS considering the established prognostic factors (table S4). In a multivariate CPH model of MYCN-amplified NB subcohorts from E-MTAB-1781, which included all the above clinical parameters except the MYCN and chromosome 11q statuses, ALDH18A1 expression was a risk factor based on both EFS and OS (tables S5 and S6).

ALDH18A1 promotes the proliferation, self-renewal, and tumor-initiating capacity of NB cells

To determine whether aberrant ALDH18A1 expression is involved in multiple biological pathways or processes, we initially performed Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis to assess gene correlation (Pearson r cutoff = 0.5; P < 0.05) in the GSE16476 cohort based on the R2 platform. We determined that genes correlated with ALDH18A1 were enriched in pathways related to several biosynthetic and proliferative aspects of cellular physiology, including spliceosomes, RNA transport, ribosome biogenesis, DNA replication, and cell cycle (fig. S7A). Similarly, the DNA replication, cell cycle, and mRNA processing pathways, as annotated by WikiPathways, were also up-regulated in samples with relatively high ALDH18A1 expression as compared to those with relatively low expression (fig. S7B). Amino acid metabolism and epigenetic modification–associated metabolic pathways (trans-sulfuration and one-carbon metabolism and folate metabolism) were up-regulated, with differentiation-related pathways being down-regulated (fig. S7B). Moreover, in all seven tested cohorts, genes involved in DNA replication and the cell cycle identified by KEGG pathway analysis showed hyperactivation in patients with high ALDH18A1 expression as compared to patients with low ALDH18A1 expression (fig. S7C).

Upon ALDH18A1 overexpression in SK-N-SH cells, we observed changes of metabolites, including glutamic acid, in the cellular metabolic profile (fig. S8, A to D, and data file S1). Metabolic pathway analysis revealed that biochemical pathways including alanine, aspartate, and glutamate metabolism (P = 2.61 × 10−4, impact = 0.55); aminoacyl–transfer RNA (tRNA) biosynthesis (P = 2.50 × 10−6, impact = 0.23); arginine and proline metabolism (P = 4.35 × 10−6, impact = 0.33); and d-glutamine and d-glutamate metabolism (P = 8.36 × 10−5, impact = 0.11) were significantly changed (fig. S8E and data file S2).

We also knocked down the expression of ALDH18A1 in SK-N-BE(2) cells and found that glutamic acid and other metabolites such as serine and valine were changed (fig. S9, A to D, and data file S3). In addition, we identified significantly altered biochemical pathways, including alanine, aspartate, and glutamate metabolism (P = 0.01, impact = 0.58); aminoacyl-tRNA biosynthesis (P = 1.83 × 10−4, impact = 0.17); purine metabolism (P = 4.15 × 10−4, impact = 0.14); cysteine and methionine metabolism (P = 6.7 × 10−4, impact = 0.20); and pantothenate and CoA biosynthesis (P = 0.01, impact = 0.25) (fig. S9E and data file S4).

On the basis of these results, we speculated that interfering with ALDH18A1 expression could induce metabolomic changes including alterations in glutamate metabolism, which has been suggested to stimulate a large diversity of intracellular signaling pathways and to regulate cell proliferation and tumor development (35, 36). Together, these data suggested a functional role of deregulated ALDH18A1 expression in cellular growth and proliferation-related processes.

Thus, we examined how the knockdown and overexpression of ALDH18A1 affect the proliferation of NB cell lines with different MYCN statuses using 3-(4, 5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTS) assays. We first stably transfected short hairpin RNA (shRNA) for ALDH18A1 into SK-N-BE(2) cells and confirmed effective knockdown of ALDH18A1 expression (fig. S10). In IMR32 and SK-N-BE(2) cells, both of which showed relatively high ALDH18A1 expression, we observed that down-regulating ALDH18A1 expression reduced cell proliferation relative to that of scramble control or wild-type (WT) cells (Fig. 3A). Conversely, forced ALDH18A1 overexpression enhanced the proliferation of SK-N-SH and SH-SY5Y NB cells (Fig. 3A), which have comparatively low ALDH18A1 expression. Collectively, these findings highlight a functional dependence of NB cell proliferation on ALDH18A1 signaling.

Fig. 3 ALDH18A1 affects the proliferation, self-renewal, and tumorigenicity of NB cells.

(A) MTS assay of viability in SK-N-BE(2) and IMR32 cells transfected with ALDH18A1 shRNA (shALDH18A1) or the scramble control and in SH-SY5Y and SK-N-SH cells transfected with the ALDH18A1 ORF (ALDH18A1OE) or empty vector. The bars indicate the means ± SEMs of six replicates. Student’s t test, ***P < 0.001. OD, optical density. (B) Quantitative analysis of tumorsphere formation in SK-N-BE(2) and IMR32 cells transfected with shALDH18A1 or the scramble control and in SH-SY5Y and SK-N-SH cells transfected with the ALDH18A1 ORF (ALDH18A1OE) or empty vector. The bars indicate the means ± SEMs of six replicates. Student’s t test, ***P < 0.001. (C) Representative images of the symmetric division of SK-N-BE(2) cells and the asymmetric division of SK-N-SH cells showing the expression of Numb and Nestin during the late stage of mitosis. Numb is red, Nestin is green, and DNA [stained with 4,6-diamidino-2-phenylindole (DAPI)] is blue. Scale bars, 10 μm. (D) Quantification of the composition of ACD, SCD, and an ambiguous phenotype in SK-N-SH cells transfected with the ALDH18A1 ORF (ALDH18A1OE, n = 748) or empty vector (n = 361) and in SK-N-BE(2) cells transfected with shALDH18A1 (shALDH18A1-1, n = 317; shALDH18A1-2, n = 162; shALDH18A1-3, n = 147) or the scramble control (n = 280). SCD, symmetric cell division; ACD, asymmetric cell division. (E) Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis of relative mRNA expression of self-renewal–related genes after ALDH18A1 knockdown in SK-N-BE(2) cells (ALDH18A1KD). The bars indicate the means ± SEMs of four replicates. Student’s t test, **P < 0.01, ***P < 0.001. ns, no significant difference. (F) Subcutaneous NOD/SCID mouse xenograft assay of SK-N-BE(2) cells transfected with ALDH18A1 shRNA [SK-N-BE(2)KD] or the scramble control [SK-N-BE(2)NC] (n = 6), IMR32 cells transfected with ALDH18A1 shRNA (IMR32KD) or the scramble control (IMR32NC) (n = 8), and SK-N-SH cells transfected with the ALDH18A1 ORF (SK-N-SHOE) or empty vector (SK-N-SHNC) (n = 7). Images were acquired on day 42 [SK-N-BE(2) and IMR32] or day 28 (SK-N-SH) after inoculation. Because the SK-N-BE(2)KD group did not develop palpable or detectable tumor, we only presented images of the anatomical structure where transplantations occurred, highlighted as the red dashed rectangle. The tumor volume was analyzed by a scatterplot. The bars show the means ± SEMs. Student’s t test, ***P < 0.001.

Moreover, we investigated the effects of ALDH18A1 expression on self-renewal by quantitatively analyzing tumorsphere formation. ALDH18A1 knockdown inhibited the self-renewal potential and decreased the number of tumorspheres, whereas ALDH18A1 overexpression in NB cells enhanced self-renewal and increased tumorsphere formation relative to that in the empty vector group (Fig. 3B), illustrating an oncogenic function of ALDH18A1 in promoting the cells’ self-renewal potential.

Because the balance between asymmetric cell division (ACD) and symmetric cell division (SCD) may be a key regulator of self-renewal, differentiation, and tumorigenesis in NB cells (37), we investigated whether ALDH18A1 expression may affect the division fate and self-renewal capacity of NB cells. By observing the distribution of Numb, a well-established indicator of ACD/SCD status (38, 39), and Nestin, a protein marker for neural stem cells (NSCs), we noted that SK-N-SH cells showed frequent ACD, whereas in SK-N-BE(2) cells, ACD was hardly detectable (Fig. 3, C and D). Our findings showed that the number of cells with the SCD phenotype was higher in SK-N-SH ALDH18A1OE cells than in control cells, and we observed suppression of SCD in SK-N-BE(2) ALDH18A1KD cells (Fig. 3D). Thus, ALDH18A1 knockdown evoked phenotypic changes opposite to those induced by ALDH18A1 overexpression. These results suggest that ALDH18A1 expression could affect the ACD/SCD phenotype, indicating that ALDH18A1 may be a regulator controlling the self-renewal potential of NB cells. To determine whether the self-renewal of NB cells involves differential GE patterns, we next compared the relative expression of genes in the self-renewal category defined by Soucie et al. (40), revealing decreased expression of nine self-renewal genes, including AKT, BMI1, BRCA1, CIRH1A, LIN28B, OCT4, SOX2, SSEA-1, and MYCN in SK-N-BE(2) ALDH18A1KD cells (Fig. 3E).

Last, given that high ALDH18A1 expression correlates with poor clinical outcomes, we further assessed in vivo tumorigenicity by implanting 1.5 million NB cells into nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice. SK-N-BE(2) and IMR32 cells transfected with shRNA constructs targeting ALDH18A1 produced no detectable tumors [SK-N-BE(2) group] or smaller tumors (IMR32 group) 6 weeks after inoculation, whereas all mice in the scramble control group developed subcutaneous tumors (Fig. 3F). In addition, larger xenografts were formed from SK-N-SH ALDH18A1OE cells than those from cells transfected with empty vector (Fig. 3F). Together, these observations highlight the oncogenic potential of ALDH18A1 expression, imply a rate-limiting role for ALDH18A1 in NB pathogenesis, and support the hypothesis that ALDH18A1 promotes the proliferation, self-renewal, and tumor-initiating capacity of NB cells.

ALDH18A1 transcriptionally regulates the expression of MYCN via the miR-29b/SP1 regulatory loop

Given that targeted ALDH18A1 knockdown decreased the expression of endogenous MYCN mRNA in SK-N-BE(2) cells (Fig. 3E), we hypothesized that aberrant ALDH18A1 expression may transcriptionally modulate the expression of MYCN. To support this hypothesis, we evaluated the phenotypic effect of ALDH18A1 on MYCN expression upon knockdown of ALDH18A1 in MYCN-amplified NB cells and observed comparable suppression of MYCN mRNA and protein expression in SK-N-BE(2) and IMR32 cells (Figs. 3E and 4, A and B). Moreover, forced overexpression of ALDH18A1 increased N-MYC protein expression in MYCN-nonamplified NB cells (SK-N-SH and SH-SY5Y) (Fig. 4B). To investigate whether the effects of ALDH18A1’s modulation on MYCN expression are dependent on cell cycle progression, we applied nocodazole, which chemically interferes with the organization of microtubules in cells and induces cell cycle arrest in G2-M phase (41). After treatment with nocodazole for 16 hours, the elevated N-MYC protein expression in SH-SY5Y ALDH18A1OE cells remained unchanged (fig. S11). It suggested that ALDH18A1’s regulatory impact on MYCN expression was not dependent on cell cycle progression.

Fig. 4 ALDH18A1 regulates MYCN expression at the transcriptional level.

(A) qRT-PCR analysis of relative ALDH18A1 and MYCN mRNA expression in IMR32 cells transfected with ALDH18A1 shRNA (ALDH18A1KD) or the scramble control. The bars indicate the means ± SEMs of four replicates. Student’s t test, ***P < 0.001. (B) Immunoblot analysis of ALDH18A1 and N-MYC protein expression in SK-N-BE(2) and IMR32 cells transfected with ALDH18A1 shRNA (ALDH18A1KD) compared to that in scramble control-transfected (Scramble) or untransfected (WT) cells and in SK-N-SH and SH-SY5Y cells transfected with the ALDH18A1 ORF (ALDH18A1OE) compared to that in empty vector–transfected (Vector) or untransfected (WT) cells. GAPDH was used as the loading control. Representative results from four replicates are shown. (C) Relative luciferase activity of reporter constructs containing the MYCN promoter in SK-N-BE(2) cells transfected with ALDH18A1 shRNA (ALDH18A1KD) compared to that in scramble control-transfected (Scramble) or untransfected (WT) cells and in SK-N-SH cells transfected with the ALDH18A1 ORF (ALDH18A1OE) compared to that in empty vector–transfected (Vector) or untransfected (WT) cells. The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001. (D) Immunoblot analysis of ALDH18A1, SP1, and E2F1 protein expression in SK-N-SH and SH-SY5Y cells transfected with the ALDH18A1 ORF (ALDH18A1OE) compared to that in cells transfected with empty vector and in SK-N-BE(2) cells transfected with ALDH18A1 shRNA (ALDH18A1KD) compared to that in cells transfected with the scramble control. GAPDH was used as the loading control. Representative results from four replicates are shown. (E) Immunoblot analysis of SP1 and N-MYC protein expression in SK-N-BE(2) cells transfected with three different siRNAs (siSP1-1 to siSP1-3) targeting SP1 or with siControl (siNC) for 72 hours. GAPDH was used as the loading control. Representative results from four replicates are shown. (F) qRT-PCR analysis of relative miR-29b expression in SK-N-BE(2) cells transfected with ALDH18A1 shRNA (ALDH18A1KD) compared to that in cells transfected with the scramble control and in SK-N-SH cells transfected with ALDH18A1 ORF (ALDH18A1OE) compared to that in empty vector–transfected cells. The bars indicate the means ± SEMs of three replicates. Student’s t test, *P < 0.05, ***P < 0.001. (G) Relative luciferase activity of reporter constructs containing the MYCN 5′-promoter in SK-N-SH cells cotransfected with ALDH18A1 ORF (SK-ALDH18A1OE) and either miR-29b mimic or the negative control (NC) mimic (Mock). The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001. (H) qRT-PCR analysis of relative MYCN mRNA expression in SK-N-SH cells cotransfected with the ALDH18A1 ORF (SK-ALDH18A1OE) and either miR-29b mimic or the NC mimic (Mock) and in SK-N-BE(2) cells transfected with miR-29b mimic or NC mimic (Mock). The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001. (I) Immunoblot analysis of SP1 and N-MYC protein expression from cells treated as shown in (H). GAPDH was used as the loading control. Representative results from four replicates are shown.

To determine the precise role of ALDH18A1 in regulating MYCN, we cotransfected SK-N-BE(2) and SK-N-SH cells with luciferase reporter constructs containing the human MYCN 5′ promoter and either ALDH18A1 shRNA or ALDH18A1 open reading frame (ORF) constructs. The relative luciferase activity in ALDH18A1KD cells was reduced upon transfection of the MYCN 5′ promoter compared to that of scramble control. Conversely, ectopic ALDH18A1 expression resulted in induction of reporter activity from the MYCN 5′ promoter (Fig. 4C). The positive relationship between ALDH18A1 expression and MYCN promoter activity raised the possibility that ALDH18A1 may regulate the transcription factors SP1 and E2F1, which are important for driving basal MYCN expression (31, 42, 43). Western blot analysis revealed that the exogenous expression and knockdown of ALDH18A1 up-regulates and attenuates SP1 protein expression, respectively. However, no obvious changes in E2F1 expression were detected (Fig. 4D). These data suggest that ALDH18A1 transcriptionally modulates MYCN expression by controlling SP1 expression. Thus, we hypothesized that a miR-29b/SP1 autoregulatory loop (21) might contribute to the ALDH18A1-mediated transcriptional regulation of MYCN expression and that miR-29b targeting of MYCN (22) is implicated in the posttranscriptional modulation of MYCN expression.

To support this hypothesis, we first examined the effect of SP1 on MYCN expression and observed decreased N-MYC protein expression upon small interfering RNA (siRNA)–mediated knockdown of SP1 (Fig. 4E), indicating that the miR-29b/SP1 autoregulatory loop regulates MYCN expression. Then, we evaluated miR-29b expression in SK-N-BE(2) cells with targeted ALDH18A1 silencing and observed miR-29b induction upon ALDH18A1 knockdown (Fig. 4F). Conversely, forced ALDH18A1 expression suppressed miR-29b in SK-N-SH cells (Fig. 4F). These findings confirm an inverse relationship between ALDH18A1 and miR-29b in NB cells, suggesting that the modulation of MYCN expression by ALDH18A1 is possibly carried out via the miR-29b/SP1 autoregulatory loop (21). Next, we sought to test the phenotypic effects of miR-29b by transfecting miR-29b mimics into SK-N-SH ALDH18A1OE cells. Ectopic miR-29b expression produced a decrease in the luciferase activity of the MYCN promoter construct (Fig. 4G) and subsequent suppression of MYCN mRNA and protein expression (Fig. 4, H and I), along with decreased SP1 protein expression (Fig. 4I, left), implicating the miR-29b/SP1 autoregulatory loop as a component of ALDH18A1-induced MYCN expression. Similar results were obtained upon transfection of miR-29b mimics in SK-N-BE(2) WT cells (Fig. 4, H and I). These observations demonstrate that ALDH18A1 can transcriptionally support MYCN expression via the miR-29b/SP1 autoregulatory loop in the absence of MYCN amplification and that relatively high endogenous ALDH18A1 expression may also contribute to the high MYCN expression in MYCN-amplified NBs through regulation of the miR-29b/SP1 loop.

ALDH18A1 posttranscriptionally regulates the expression of MYCN via a miRNA network

We further assessed whether miR-29b up-regulation caused ALDH18A1KD-mediated MYCN repression. Ablation of miR-29b partially rescued MYCN promoter activity in SK-N-BE(2) ALDH18A1KD cells (Fig. 5A) and increased MYCN mRNA expression in SK-N-SH WT cells (Fig. 5B). However, no appreciable derepression of N-MYC protein expression was detected upon miR-29b silencing (Fig. 5C), suggesting a contributing role for posttranscriptional regulation in controlling the N-MYC protein expression. To explore the underlying mechanism, we initially performed luciferase assays by cotransfecting SK-N-BE(2) and SK-N-SH cells with luciferase reporter constructs containing the MYCN 3′ untranslated region (3′UTR) and either ALDH18A1 shRNA or ALDH18A1 ORF. Neither ALDH18A1 knockdown nor overexpression induced detectable changes in luciferase activity from the MYCN 3′UTR compared with that in control cells (Fig. 5D).

Fig. 5 ALDH18A1 regulates MYCN expression at the posttranscriptional level.

(A) Relative luciferase activity of reporter constructs containing the MYCN promoter in SK-N-BE(2) cells cotransfected with ALDH18A1 shRNA [SK-N-BE(2) ALDH18A1KD] and either miR-29b ASO or scramble ASO. The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001. (B) qRT-PCR analysis of relative MYCN mRNA expression in SK-N-SH cells transfected with miR-29b ASO or scramble ASO. The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001. (C) qRT-PCR analysis of relative miR-29b expression and immunoblot analysis of N-MYC protein expression in SK-N-SH cells transfected with miR-29b ASO or scramble ASO. The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001. (D) Relative luciferase activity of reporter constructs containing the MYCN 3′UTR in SK-N-BE(2) cells transfected with ALDH18A1 shRNA (ALDH18A1KD) compared to those in scramble control-transfected (Scramble) or untransfected (WT) cells and in SK-N-SH cells transfected with the ALDH18A1 ORF (ALDH18A1OE) compared to that in empty vector–transfected (Vector) or untransfected (WT) cells. The bars indicate the means ± SEMs of three replicates. Student’s t test. ns, no significance. (E) Venn diagram depicting the overlap between sets of potential MYCN-targeting miRNAs predicted by TargetScan, microRNA.org, miRDB, and PicTar. (F) qRT-PCR analysis of relative expression of 20 predicted potential MYCN-targeting miRNAs in SK-N-BE(2) and SK-N-SH cells. The bars indicate the means ± SEMs of three replicates. Student’s t test, **P < 0.01, ***P < 0.001. (G) Immunoblot analysis of SP1 and N-MYC protein expression in SK-N-SH cells transfected with mixed miR-193b, miR-29a, and miR-29b ASOs (miRNA-ASO mix) or scramble ASO (left) and in SK-N-BE(2) cells cotransfected with ALDH18A1 shRNA (BE2-ALDH18A1KD) and either mixed miR-193b, miR-29a, and miR-29b ASO (miRNA-ASO mix) or scramble ASO. GAPDH was used as the loading control. Representative results from four replicates are shown. (H) qRT-PCR analysis of relative miR-29a and miR-193b expression and immunoblot analysis of N-MYC protein expression in SK-N-SH cells transfected with individual miRNA (miR-29a and miR-193b) ASO or scramble ASO. The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001.

Considering the concordant transcriptional and translational suppression of MYCN expression, we hypothesized that on the one hand, robust induction of miR-29b upon ALDH18A1 loss plays a critical role in MYCN down-regulation via the miR-29b/SP1 autoregulatory loop at the transcriptional level. On the other hand, because epigenetic dysregulation contributes to NB tumorigenesis (44), a miRNA regulatory network at the posttranscriptional level might also be responsible for MYCN depression despite no appreciable overall changes in the MYCN 3′UTR luciferase activity. To this end, we used four bioinformatic software programs (TargetScan, microRNA.org, miRDB, and PicTar) to predict the potential miRNAs that would bind the MYCN 3′UTR. Among these predicted miRNAs, we found 20 that overlapped between the four algorithms (Fig. 5E). Next, we examined the expression of these 20 miRNAs in SK-N-BE(2) and SK-N-SH cells, which exhibit relatively high and low ALDH18A1 and MYCN expression, respectively. We observed overexpression of miR-29a and miR-193b in SK-N-SH cells relative to that in SK-N-BE(2) cells, as well as miR-29b (Fig. 5F). We also identified several other DE miRNAs that showed a positive correlation with ALDH18A1 (Fig. 5F). Together, these findings further support our idea that apart from transcriptional regulation, a miRNA network might be implicated in the posttranscriptional modulation of MYCN expression. Therefore, we used a loss-of-function assay to show that depletion of miR-193b, miR-29a, and miR-29b induced the luciferase activity of reporter constructs containing the MYCN 3′UTR in SK-N-BE(2) ALDH18A1KD cells (fig. S12), partially rescued SP1 and N-MYC protein expression in SK-N-BE(2) ALDH18A1KD cells, and up-regulated SP1 and N-MYC protein expression in SK-N-SH WT cells (Fig. 5G). Collectively, these results suggest that the disruption of the internal balance in the miRNA network might account for the targeted inhibition of MYCN expression. Individual miRNA (miR-29b, miR-29a, and miR-193b) antisense oligonucleotides (ASOs) could not induce appreciable N-MYC protein expression (Fig. 5, C and H).

MYCN reciprocally enhances ALDH18A1 expression via direct transcriptional activation

Liu et al. (45) reported that c-MYC could up-regulate ALDH18A1 in human Burkitt lymphoma model P493 and PC3 human prostate cancer cells. Structurally and biologically, c-MYC and N-MYC are highly homologous and show prominent functional redundancy (7). We therefore sought to determine whether the transcription factor N-MYC is involved in the modulation of ALDH18A1 expression in NB. Our previous results revealed that ALDH18A1 is included in the NMYC_01, which is a gene set with at least one occurrence of N-MYC binding site in the regions spanning up to 4 kb around the transcription start site (TSS) (fig. S5A), suggesting that ALDH18A1 might be a direct target of N-MYC and implying a positive feedback loop between ALDH18A1 and MYCN.

Overexpression and knockdown of MYCN increased and decreased ALDH18A1 protein expression, respectively (Fig. 6A). In addition, Omomyc, a dominant-negative allele of MYC family members (including c-MYC and N-MYC) that carries point mutations in the leucine zipper domain, disrupts MYC/MAX heterodimerization, and interferes with MYC-induced transactivation (46, 47), also reduced ALDH18A1 expression (Fig. 6A), further supporting the idea that N-MYC positively modulates ALDH18A1 expression.

Fig. 6 N-MYC reciprocally transactivates ALDH18A1 transcription.

(A) Immunoblot analysis of N-MYC and ALDH18A1 protein expression in SH-SY5Y cells transfected with the MYCN ORF (MYCNOE) or empty vector and in SK-N-BE(2) cells transfected with three different siRNAs (siMYCN-1 to siMYCN-3) targeting MYCN or siControl (siNC) or with Omomyc for 72 hours. GAPDH was used as the loading control. Representative results from four replicates are shown. (B) Schematic diagram showing the location of potential N-MYC binding sites on the human ALDH18A1 gene promoter. TSS, transcription start site. (C) Relative luciferase activity of constructs containing fragments of the ALDH18A1 promoter region in SK-N-BE(2) and IMR32 cells. The bars indicate the means ± SEMs of three replicates. (D) ChIP assay showing the N-MYC enrichment statuses of the ALDH18A1 gene promoter in SK-N-BE(2) and SH-SY5Y cells. (E) EMSA was performed with nuclear extracts from SH-SY5Y and SK-N-BE(2) cells incubated with biotin-labeled double-stranded oligonucleotides containing the putative N-MYC binding site 2 on the ALDH18A1 promoter. The presence of N-MYC on the proximal ALDH18A1 promoter was demonstrated by the position of the antibody supershift upon incubation with anti–N-MYC. (F) Relative luciferase activity of constructs containing the intact ALDH18A1 promoter (WT-Luc) or the fragment with deletion of N-MYC binding site 2 (Mut-Luc) in SK-N-BE(2) cells. The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001. (G) Summary diagram describing the positive feedback loop between ALDH18A1 and MYCN in NB cells.

To begin unraveling the regulatory mechanisms of ALDH18A1 expression, we examined the ALDH18A1 promoter region for transcription factor binding sites and identified two potential N-MYC binding sites (site 1, 5′-CACGTGGTGGAA-3′; site 2, 5′-CGCCACGTGGAGC-3′) located −1117 to −1105 base pairs (bp) and −40 to −27 bp upstream of the TSS, respectively, in the human ALDH18A1 gene promoter (Fig. 6B). Next, to identify potential N-MYC binding sites, we transfected IMR32 and SK-N-BE(2) cells harboring MYCN amplification and high MYCN expression with a series of luciferase constructs containing fragments of the ALDH18A1 promoter region (Fig. 6C). Constructs containing the −1529- to +148-bp, −993- to +148-bp, and – 443- to +148-bp regions upstream of the TSS, all of which encompassed the putative N-MYC binding site 2 (5′-CGCCACGTGGAGC-3′, −40 to −27 bp), exhibited prominently increased reporter activity compared with that in the pGL4 control. However, no appreciable change in reporter activity from the −993- to +148-bp region of the ALDH18A1 promoter region, which contains only site 2 but not site 1, was observed when compared with the −1529- to +148-bp region containing both putative binding sites. Moreover, a construct containing the +3- to +148-bp region lacking N-MYC binding sites 1 and 2 exhibited decreased reporter activity, similar to those in the control group (Fig. 6C). These observations suggest that the proximal ALDH18A1 promoter is likely to be activated by MYCN via the putative N-MYC binding site 2 (5′-CGCCACGTGGAGC-3′, −40 to −27 bp) (Fig. 6B).

Subsequently, we performed chromatin immunoprecipitation (ChIP) and found N-MYC enrichment on the ALDH18A1 promoter (Fig. 6D) in SK-N-BE(2) cells compared with that in SH-SY5Y cells or with the immunoglobulin G (IgG) control (Fig. 6D), further supporting our hypothesis. This result was confirmed by electrophoretic mobility shift assay (EMSA) on nuclear extracts from SK-N-BE(2) and SH-SY5Y cells with biotin-labeled probes spanning N-MYC binding site 2 on the ALDH18A1 promoter [referred to as the Hot Probe (WT)] or mutant versions of the labeled N-MYC binding site 2 (Hot Probes Mut-1 and Mut-2) and a specific competitive unlabeled probe sharing the same sequence as that of the labeled probes (referred to as the Cold Probe) (Fig. 6E). The N-MYC-probe complex was attained when the Hot Probe (WT) was incubated with SK-N-BE(2) nuclear extracts, whereas no apparent N-MYC–probe complex were attained when the Hot Probes (Mut-1 and Mut-2) were incubated with SK-N-BE(2) nuclear extracts or the Hot Probe (WT) was incubated with SH-SY5Y nuclear extracts, supporting N-MYC enrichment on the ALDH18A1 promoter. The specific binding of N-MYC at the ALDH18A1 promoter was demonstrated by the abrogation of binding in the presence of the Cold Probe. In addition, binding of anti–N-MYC reduced the gel mobility of the protein-DNA binding complexes, producing a secondary mobility shift compared with the IgG control group (Fig. 6E). Moreover, on the basis of deletion analysis, deleting the N-MYC binding site 2 fragment suppressed the increase in the luciferase activity induced by the intact ALDH18A1 promoter (Fig. 6F). Thus, we confirmed that MYCN directly targets the ALDH18A1 promoter by specifically associating with N-MYC binding sites and transcriptionally activating ALDH18A1 expression. Collectively, these observations provide a model of a reciprocal activation loop between ALDH18A1 and MYCN, suggesting a scenario in which aberrant ALDH18A1 expression along with amplification and corresponding overexpression of the MYCN oncogene dictates the observed ALDH18A1 phenotypes, which may explain the marked correlation between ALDH18A1 expression and NB patient survival (Fig. 6G).

YG1702, an ALDH18A1-specific inhibitor, is identified by molecular docking

Therapeutic strategies harnessing the altered metabolic properties of cancer cells are currently focused on developing specific inhibitors against key metabolic enzymes (48). To this point, we used molecular docking to screen for specific ALDH18A1 inhibitors (Fig. 7A). Before molecular docking, a target library (>200,000 compounds) was prepared by removing salt ions, retaining only the largest molecular fragments, enumerating tautomers, and protonation. The three-dimensional (3D) structures were generated by energy minimization. Using high-throughput docking, the top 11,180 compounds were further submitted to flexible docking, and the docked poses were refined with a force field. Thereafter, absorption, distribution, metabolism, excretion, and toxicity (ADMET) descriptors of the top 1000 compounds were calculated, and those with bad ADMET properties were filtered. During the ADMET filtering process, all compounds with Absorption_T2_2D > 9.6 and log(Sw) > −4.0 were filtered. Log(Sw) represents the solubility of a molecule. The remaining 588 compounds were divided into structural clusters using fingerprint-based clustering, and a diverse subset with 30 hits was ultimately identified (Fig. 7B). Among these 30 hits, the compound YG1702 [C23H30N2O7S; molecular weight = 478.57; Fig. 7C, structural formula; Fig. 7D, identification result from proton nuclear magnetic resonance (H-NMR)] was identified with low cytotoxicity in normal human NSCs, human brain glial (HEB) cells, and human colonic mucosal epithelial NCM460 cells. In addition, the sensitivity to YG1702 was dependent on MYCN expression, and its median inhibitory concentration in NB cell lines with high MYCN expression was lower than that in NB cell lines with low MYCN expression (Fig. 7E and fig. S13). Therefore, YG1702 was selected for further investigation. The binding mode of YG1702 in the ALDH18A1 active site is shown in Fig. 7F (3D mode) and Fig. 7G (2D mode). Next, we used molecular docking to calculate the ALDH18A1 docking score of 36 analogs (32 compounds derived from the Specs library and 4 clinical drugs) with structures similar to that of YG1702. The docking score of YG1702 with ALDH18A1 was the highest among the 37 compounds, suggesting that this particular chemical scaffold, YG1702, was indeed promising (table S7). In addition, the four YG1702 analogs (clinical drugs: C18H19NO4Cl2, C18H20N2O6, C21H26N2O7, and C17H18N2O6) had comparable cytotoxic effects in HEB and IMR32 cells (fig. S14A). The structural formulas and binding sites of four Specs compounds and four clinical drugs are shown in figs. S14B and S15.

Fig. 7 Molecular docking was applied to screen the ALDH18A1 inhibitor.

(A) 3D diagram of the ALDH18A1 structure. (B) Flowchart of the virtual screening workflow. PDB, Protein Data Bank. (C) Chemical structure of YG1702. (D) Identification of YG1702 by H-NMR. ppm, parts per million. FT, Fourier transform. (E) Cell viability analyses of MYCN-amplified [SK-N-BE(2) and BE2C] and MYCN-nonamplified (SH-SY5Y and SK-N-AS) NB cells, SK-N-SH cells transfected with the ALDH18A1 ORF (SK-OE) or empty vector (SK-NC), and normal cell lines (HEB, NCM460, and NSC) treated with increasing concentrations of YG1702 for 72 hours. The inhibition rates relative to those of PBS-treated cells are shown (inhibition rate = 1 − relative cell viability). The results indicate the means ± SEMs of three replicates. (F) A 3D diagram of YG1702 binding in the ALDH18A1 active site. (G) A 2D diagram of the binding of YG1702 to ALDH18A1. (H) An ITC assay was performed to assess the specific binding of YG1702 to ALDH18A1 recombinant protein.

Then, we investigated the interaction between YG1702 and the recombinant ALDH18A1 protein using isothermal titration calorimetry (ITC). As shown in Fig. 7H, the obtained thermodynamic parameters were as follows: number of binding sites per monomer (N) = 0.949 ± 0.0659 sites, association constant (Ka) = 5.89 × 107 ± 3.00 × 107 M−1, enthalpy change (ΔH) = −3.366 × 105 ± 3.499 × 104 J/mol, and entropy change (ΔS) = −980 J/mol per degree. These data indicated that YG1702 physically interacts with ALDH18A1 with a high affinity and might potentially affect its enzymatic activity.

YG1702 inhibits N-MYC expression and attenuates the growth of human NB

To further assess whether YG1702 treatment can inhibit ALDH18A1 function to depress the expression of N-MYC and influence the biological characteristics of NB cells, we first treated NB cells [SK-N-BE(2) and SK-N-SH ALDH18A1OE] with YG1702 in vitro and observed MYCN down-regulation at the mRNA and protein levels (Fig. 8, A to D). Moreover, ALDH18A1 expression was also down-regulated, consistent with a reciprocal activation loop between ALDH18A1 and MYCN (Fig. 8, B to D). Moreover, the tumorsphere formation (Fig. 8E), viability (Fig. 7E), and proliferation (Fig. 8F) of different NB cell lines were reduced after YG1702 treatment.

Fig. 8 Inhibition of ALDH18A1 activity drives MYCN-overexpressing NB cells to acquire the MYCN-low phenotype.

(A) qRT-PCR analysis of relative MYCN mRNA expression in SK-N-BE(2) cells treated with the indicated doses of YG1702 for 48 hours. The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001. (B) qRT-PCR analysis of relative ALDH18A1 and MYCN mRNA expression in SK-N-SH ALDH18A1OE cells treated with the indicated doses of YG1702 for 48 hours. The bars indicate the means ± SEMs of three replicates. Student’s t test, ***P < 0.001. (C) Immunoblot analysis of N-MYC and ALDH18A1 protein expression in SK-N-BE(2) cells treated with the indicated doses of YG1702 for 48 hours. GAPDH was used as the loading control. Representative results from four replicates are shown. (D) Immunoblot analysis of N-MYC and ALDH18A1 protein expression in SK-N-SH ALDH18A1OE cells treated with the indicated doses of YG1702 for 48 hours. GAPDH was used as the loading control. Representative results from four replicates are shown. (E) Quantitative analysis of tumorsphere formation in SK-N-BE(2) cells and SK-N-SH ALDH18A1OE cells treated with the indicated doses of YG1702 or PBS. The bars indicate the means ± SEMs of four replicates. Student’s t test, **P < 0.01, ***P < 0.001. (F) MTS assay of proliferation in SK-N-BE(2) and SK-N-SH ALDH18A1OE cells treated with the indicated doses of YG1702. The bars indicate the means ± SEMs of six replicates. Student’s t test, ***P < 0.001. (G) Body weights of NOD/SCID mice treated with YG1702 [intraperitoneal injection (I.P.), 45 mg/kg per day, three times] or PBS after the SK-N-BE(2) xenograft tumors reached 25 to 50 mm3. Student’s t test. (H and I) SK-N-BE(2) and IMR32 xenograft tumor size in mice treated with YG1702 or PBS after the tumors reached 25 to 50 mm3. Day 0 indicates the day of cell implantation. After 14 days, the mice were divided into two groups: PBS (control) treated and YG1702 (I.P. 45 mg/kg per day) treated. (H) The bars show the tumor volumes of six [SK-N-BE(2)] or seven (IMR32) mice per group (means ± SEMs). Student’s t test, **P < 0.01, ***P < 0.001. (I) Images were acquired on day 29 after inoculation. (J) Kaplan-Meier curves showing the survival of SK-N-BE(2) (n = 6) and IMR32 (n = 7) tumor–bearing mice upon continuous treatment with YG1702 or PBS. The statistical significance of the difference between the two treatment groups was evaluated using a log-rank test.

To evaluate the therapeutic potential of YG1702, we next investigated whether YG1702 inhibits NB development in subcutaneous xenograft mouse models [SK-N-BE(2) and IMR32]. When the tumors reached 25 to 50 mm3, the mice were divided into two groups receiving an intraperitoneal injection of either YG1702 (45 mg/kg per day) or phosphate-buffered saline (PBS) (negative control) every 3 days for a total of three times. No apparent weight loss was observed in any of the YG1702-treated mice over the course of the experiment (Fig. 8G), and YG1702 inhibited the growth of the xenografts in tumor-bearing mice compared with those in mice of the negative control group (Fig. 8, H and I). The Kaplan-Meier survival curves showed an OS advantage for tumor-bearing mice receiving continuous treatment with YG1702 compared with mice receiving the PBS control (Fig. 8J). IHC analysis of N-MYC protein expression in SK-N-BE(2) xenografts confirmed the attenuation of N-MYC and ALDH18A1 protein expression in YG1702-treated tumors compared to tumors from the PBS control group (fig. S16). These functional studies demonstrate that the pharmacological inhibition of ALDH18A1 by YG1702 can lead to attenuated MYCN expression and inhibition of self-renewal, cell proliferation, and in vivo tumorigenicity of NB cells, similar to the effects observed after genetic ALDH18A1 knockdown.

DISCUSSION

In this study, we identified ALDH18A1 as a potential risk factor with prognostic value in patients with NB, especially those harboring MYCN amplification. Moreover, by incorporating bioinformatic analysis with functional studies, we revealed ALDH18A1 as a regulator of NB cell proliferation, self-renewal, and tumorigenicity. Mechanistic investigations showed that ALDH18A1 could both transcriptionally and posttranscriptionally regulate MYCN expression, the latter of which was then validated to reciprocally trigger the direct transactivation of ALDH18A1 transcription, thus forming a positive feedback loop between ALDH18A1 and MYCN. Last, we identified the ALDH18A1-specific inhibitor YG1702 by molecular docking and screening and confirmed that it could drive MYCN-overexpressing NB cells to acquire a less proliferative phenotype.

Targeted MYCN silencing induces differentiation and triggers apoptosis in NB (49, 50). Unfortunately, the development of clinically suitable, safe, and effective drug delivery vehicles hinders the widespread use of RNA interference therapeutics for MYCN control and disease treatment (51). Recent efforts have focused on the chemical screening and synthesis of potent small molecules as direct approaches to targeting the transcription factor MYC (5254). Considering the functional dependence on heterodimerization with MAX, small-molecule antagonists that hamper the transactivation of MYC-MAX heterodimers have been shown to inhibit the MYC-induced transcriptional program (8, 10, 55). However, the direct pharmacological inhibition of MYCN has thus far proven elusive, mainly because these approaches lack N-MYC selectivity or in vivo efficiency (10, 56). Thus, to date, MYCN is not amenable to pharmacological inhibition (57). In addition, controlling the MYCN downstream transcriptional program is complex not only because of difficulties in precisely reaching the magnitude of repression necessary for suppressing transformation and tumorigenesis without inducing detrimental effects on normal counterparts but also because of the alternative pathways and mechanisms that might bypass and compensate for MYCN downstream signaling (54). In general, controlling MYCN-dependent expression patterns or activities has achieved only modest progress (52).

We report that ALDH18A1 substantially affects MYCN expression. First, we preliminarily concluded that ALDH18A1 positively modulates MYCN mRNA and protein expression in IMR32 and SK-N-BE(2) cells. Next, our findings indicated that ALDH18A1 might induce changes in the reporter activity driven by MYCN 5′ promoter via the previously reported miR-29b/SP1 autoregulatory loop (21), suggesting that ALDH18A1 regulates MYCN at the transcriptional level. SP1 can directly target the MYCN promoter and control its transcriptional activity (58). In addition, MYC down-regulates miR-29b expression (59), and miR-29b is a direct MYCN-targeting miRNA (22). Then, we determined that ALDH18A1 knockdown–induced increase in miR-193b and miR-29a together with miR-29b contributes to the prominent depression of MYCN expression, and individual transfection with miR-29a, miR-29b, or miR-193b ASO did not induce detectable N-MYC protein expression. These results highlight a miRNA network that allows targeted inhibition of MYCN protein translation.

By performing a dual luciferase assay, EMSA, and ChIP analyses, we demonstrated that MYCN directly targets the ALDH18A1 promoter by specifically associating with N-MYC binding sites and transcriptionally activating ALDH18A1 expression. These observations agree with evidence that c-MYC, an N-MYC homolog, can amplify the protein expression of ALDH18A1 (45). In addition, our findings identified the −40- to −27-bp region (5′-CGCCACGTGGAGC-3′) as the proximal ALDH18A1 promoter for direct N-MYC binding. These results further emphasize the importance of a positive feedback loop between ALDH18A1 and MYCN, consistent with the idea that the oncogenic MYC family plays a pivotal role in diverse facets of metabolic reprogramming (60).

Tumorigenesis is dependent on metabolic reprogramming as both direct and indirect consequences of oncogenic driver mutations, and the alterations in intracellular and extracellular metabolites that can accompany cancer-associated metabolic reprogramming exert profound impacts on the expression of oncogenes and tumor suppressors (12). Glutamine metabolism provides a basic element for tumor cell metabolism and participates in intracellular signaling pathways involved in tumorigenesis (12). However, tumors may exhibit increased conversion of glutamine to glutamate relative to normal tissue (14). Metabolite tracing data have shown a minimal contribution of glutamine carbon to the tricarboxylic acid (TCA) cycle in Kras-driven lung tumors, suggesting that glutamine metabolism mainly participates in amino acid catabolism rather than supporting the production of TCA cycle intermediates (14). In our study, we observed that expression changes in ALDH18A1, the key enzyme in glutamine and proline metabolism, affects cell proliferation, self-renewal potential, and tumorigenicity, indicating that ALDH18A1 may lock NB cells in a more proliferative or tumorigenic state. KEGG pathway analysis indicated that metabolic pathways including pyrimidine and purine metabolism are associated with ALDH18A1 expression. In addition, amino acid metabolism and epigenetic modification–associated metabolic pathways (trans-sulfuration and one carbon metabolism and folate metabolism) were up-regulated in patients harboring relatively high ALDH18A1 expression. Metabolic reprogramming permits cancer cells to reshape the surrounding microenvironment and maintain an undifferentiated state for deregulated self-renewal and uncontrolled aggressiveness (12). Accumulating evidence has demonstrated that alterations in global GE patterns, especially multiple mutations affecting epigenetic regulators involved in DNA methylation, histone acetylation, methylation, and phosphorylation as well as the chromatin-remodeling complex, could alter epigenetic modifying enzymes or substrates that are particularly vulnerable to cancer-related metabolic changes (61). Epigenome remodeling may further regulate cellular signaling and transcriptional programs to enable increased nutrient acquisition and biosynthesis, ultimately shaping cancer cell metabolism (13). Our metabolomics study identified that modulation of ALDH18A1 expression could affect glutamate metabolism along with the biosynthesis of pantothenate and coenzyme A (CoA), the latter of which is the product of glycolytic metabolism and a donor to acetylation reactions such as H3K9 acetylation (13). Therefore, ALDH18A1 might regulate the transcription of growth-related genes such as the MYCN oncogene, at least partially through modulating acetyl-CoA and histone acetylation.

We used docking and virtual screening to identify the compound YG1702, a specific inhibitor of ALDH18A1. Our results showed the predicted binding pattern of this compound with respect to its target and indicated the selected binding site in the overall structure of the ALDH18A1 protein used for docking. The selection of this binding site was also empirically validated by calculating the ALDH18A1 docking score of 36 analogs (32 compounds derived from the Specs library and 4 clinical drugs) with structures similar to that of YG1702. Such validation is essential to exclude false-positive results, especially when the compounds are from external sources. Then, we found that YG1702 might not only affect ALDH18A1 in vitro by specific binding but also recapitulate its knockdown phenotype at the molecular and functional level both in vitro and in vivo. YG1702 has no direct effect on ALDH18A1 expression, but our data showed that YG1702 treatment decreased MYCN expression at the transcriptional and posttranscriptional levels in a dose-dependent manner, resulting in corresponding changes in ALDH18A1 expression, further validating the positive feedback and reciprocal activation loop between ALDH18A1 and MYCN.

YG1702 inhibited MYCN expression, demonstrating in vitro efficacy at low micromolar concentrations only in MYCN-amplified NB cells. Moreover, we achieved in vivo results using multiple xenograft models. These results were similar to those of previous investigations showing that MYCN-amplified cell lines were particularly sensitive to CD532, a potent MYCN inhibitor that conformationally disrupts Aurora A and completely dissociates the MYCN–Aurora A complex (62) and ABT-199, a selective B cell leukemia/lymphoma 2 (BCL-2) inhibitor (63) that triggers apoptosis in MYCN-amplified NB cells with up-regulated NOXA expression induced by MYCN amplification (57). Similarly, amplification of MYCN and the corresponding MYCN overexpression may confer sensitivity to YG1702 in NB cells and thus provide potential therapeutic opportunities for patients with MYCN-amplified NB by disruption of the positive feedback loop between ALDH18A1 and MYCN. Whereas CD532 potentiates MYCN degradation via canonical phosphorylation and ubiquitination pathways downstream of the MYCN transcriptional program (62), and ABT-199 induces the apoptosis-primed state by elimination of the prosurvival BCL-2 signal to indirectly target MYCN (33), our use of the specific ALDH18A1 inhibitor YG1702 suggests a mechanism by which pharmacologically targeting upstream regulators or disrupting the autoregulatory and reciprocal feedback loop could enhance the efficacy of transcription factor blockades.

Our study has some limitations. Despite the observed phenotypic effects, further investigations may be needed to identify whether ALDH18A1’s regulatory impact on MYCN expression is independent of the MYCN status. Moreover, further exploration is warranted to identify the mechanism underlying ALDH18A1’s ability to modulate the miR-29b/SP1 autoregulatory loop and to assess whether ALDH18A1 itself could be a component of this regulatory network. Our current study did not precisely define the interplay between the miRNA network and MYCN expression or the contribution of transcriptional and posttranscriptional regulation to ALDH18A1-mediated modulation of MYCN expression. Multifaceted mechanistic studies need to be integrated to understand the complex and rapid changes occurring upon ALDH18A1-induced miRNA network oscillation and regulation of MYCN expression. In addition, before YG1702 was used in vivo, our experiments did not include validation of its structure-activity relationship or its chemical stability. A more robust experimental validation of the chemical structure of YG1702, such as testing YG1702 analogs in vitro, is warranted to show that this compound is indeed a lead drug candidate. The in vitro inhibitory activity of this compound against ALDH18A1 should also be quantitatively demonstrated. YG1702 has two ester groups that typically exhibit low stability and are often hydrolyzed. Further studies verifying that this compound is indeed stable enough to be active in vivo, along with investigating its pharmacokinetic profile, solubility, and chemical stability, should be prioritized. Furthermore, the dose-dependent effects of YG1702 on tumorigenicity in preclinical xenograft models need to be tested to achieve broad applicability in MYCN-amplified NB. Combination-based therapeutic approaches also require further evaluation.

Recent studies have documented that the oncogenic driver c-MYC can also regulate specific pathways involved in the aggressiveness of MYCN-nonamplified NB cells (64, 65), suggesting a functional redundancy of the MYC family members in NB. Moreover, c-MYC was identified as a core transcription factor that contributes to deregulated cell growth and proliferation, stem cell fate, neoplastic transformation, and tumorigenesis (52). It is also broadly involved in diverse types of malignancies. Increased or deregulated c-MYC expression has been linked to aggressive human prostate cancer (66), triple-negative breast cancer (67), Burkitt lymphoma (68), and multiple myeloma (69) and is estimated to be frequently altered in as many as 70% of human cancers (53). On the basis of the homology of structures shared by c-MYC and N-MYC, our findings may provide insight into the mechanisms between c-MYC and intermediate metabolites and suggest a potential therapeutic approach for c-MYC–dominated carcinogenesis.

MATERIALS AND METHODS

Study design

The aim of the study was to provide mechanistic insights into the impact of genetic manipulation of ALDH18A1 and pharmacological interference with ALDH18A1 on NB cell behavior and MYCN expression in vitro and in vivo. We evaluated a cohort of newly diagnosed patients with NB for ALDH18A1 expression by IHC staining and MYCN amplification by FISH. All the analyzed samples were molecularly characterized as part of the routine diagnostic workup. The control and treatment groups, the number of biological replicates (sample sizes), and statistical analyses for each experiment are specified in the figure legends. For in vivo tumor xenograft studies, experimental groups were composed of six to seven animals each to ensure statistical power. Animals were randomly assigned to the control and treatment groups and housed together to minimize environmental differences and experimental bias. Investigators were not blinded during the follow-up of the mice and evaluation of the in vivo experiments. All the original data are in data file S5.

Animal procedures

NB cells were washed and resuspended in PBS containing Matrigel (BD Biosciences). Cell suspensions (50 μl) containing 1.5 × 106 cells were inoculated into the subcutaneous tissues of the left flanks of 5- to 6-week-old female NOD/SCID mice purchased from the animal center of the Third Military Medical University (Army Medical University). The length (L) and width (W) of the resulting tumors were determined three times weekly using a digital caliper, and the tumor volumes (cubic millimeters) were calculated as (L × W2)/2. Mice were euthanized when the maximum tumor length reached 2.0 cm. Six weeks [for SK-N-BE(2) and IMR32 cells] or 4 weeks (for SK-N-SH cells) after the injection, tumors from each group were excised. All mouse experiments were performed in accordance with relevant institutional and national guidelines and approved by the Institutional Animal Care and Use Committee of the Third Military Medical University (Army Medical University). To estimate the therapeutic potential of YG1702, we subcutaneously injected 1.5 × 106 SK-N-BE(2) or IMR32 cells into the left flanks of NOD/SCID mice to form primary tumors. When the primary tumor size reached 50 mm3, the tumor was excised and divided into equal parts for a second inoculation into the subcutaneous tissues of the left flanks of another group of NOD/SCID mice. Then, when the tumor sizes reached 25 to 50 mm3, the SK-N-BE(2) or IMR32 tumor-bearing mice were randomly divided into two groups [SK-N-BE(2), n = 6; IMR32, n = 7]. PBS and the inhibitor YG1702 were intraperitoneally administered to the mice at a dosage of 45 mg/kg per day at 3-day intervals (days 14, 17, and 20). The body weights and tumor volumes of the mice were monitored.

Tissue samples

Paraffin-embedded samples (n = 24) were obtained from patients with NB undergoing surgery at the Southwest Hospital, Daping Hospital of the Third Military Medical University, and Pediatric Hospital of Chongqing Medical University. The tissue array (US Biomax Inc., catalog no. MC602) contained 25 NB patient samples and 5 normal peripheral nerve samples. The tumor size was determined as the maximum tumor diameter, and lymph node metastasis was histologically diagnosed. All patients had sufficient samples for IHC analysis. Written informed consent was obtained from all patients. The Institute Research Medical Ethics Committee of the Third Military Medical University (Army Medical University) granted approval for this study.

Statistical analysis

Statistical analyses were performed using IBM SPSS statistics software version 22. Results are expressed as means ± SEMs, and P < 0.05 was considered significant. Student’s t test (two-tailed) was used to determine the significance of the difference between two groups of data. The Mann-Whitney U test was used to compare the difference in ALDH18A1 protein expression between NB tissue and normal peripheral nerve tissue. Survival curves were estimated according to the Kaplan-Meier product-limit method, and differences between the survival curves were assessed using the log-rank test. Cutoff values for MYCN-AMP6 scores and ALDH18A1 expression were determined by the Youden index derived from the receiver operating characteristic (ROC) curve. Univariate survival analysis was performed using the Kaplan-Meier and CPH models, with comparisons performed using the Mantel-Cox log-rank test. Multivariate analyses were based on the CPH models; the covariates in the CPH models included ALDH18A1 expression and established risk factors, including the age at diagnosis, INSS stage, MYCN status, chromosome 1p status, chromosome 11q status, and Shimada classification, all of which were dichotomized to assess the OS and EFS predictions.

SUPPLEMENTARY MATERIALS

stm.sciencemag.org/cgi/content/full/12/531/eaax8694/DC1

Materials and Methods

Fig. S1. High MYCN-AMP6 scores are associated with poor prognosis in patients with stage 4 NB.

Fig. S2. ALDH18A1 correlates with MYCN amplification and corresponding MYCN overexpression in four cohorts.

Fig. S3. ALDH18A1 correlates with MYCN amplification and corresponding MYCN overexpression in an additional cohort.

Fig. S4. ALDH18A1 correlates with MYCN amplification and corresponding MYCN overexpression in three additional cohorts and in cell lines.

Fig. S5. ALDH18A1 correlates with the MYCN-dependent transcriptional program.

Fig. S6. ALDH18A1 expression is associated with pathological parameters and survival in multiple NB cohorts.

Fig. S7. Aberrant ALDH18A1 expression correlates with the proliferative potential of NB cells.

Fig. S8. Overexpression of ALDH18A1 induces changes in the metabolic profile and biochemical pathways in SK-N-SH cells.

Fig. S9. ALDH18A1 knockdown triggers changes in the metabolic profile and biochemical pathways in SK-N-BE(2) cells.

Fig. S10. The shRNA-mediated ALDH18A1 knockdown is confirmed in MYCN-amplified SK-N-BE(2) cells.

Fig. S11. The regulatory impact of ALDH18A1 on MYCN expression was not dependent on cell cycle progression.

Fig. S12. Depletion of three miRNAs increased the luciferase activity of the MYCN 3′UTR reporter construct.

Fig. S13. Cell sensitivity to YG1702 is MYCN dependent.

Fig. S14. YG1702 analogs have comparable cytotoxicity in HEB and IMR32 cells.

Fig. S15. YG1702 analogs share a similar binding mode at the ALDH18A1 active site.

Fig. S16. YG1702 suppresses N-MYC and ALDH18A1 protein expression in vivo.

Table S1. The relationship between the expression of ALDH18A1 and the clinicopathological features.

Table S2. Univariate and multivariate analyses of different prognostic parameters in the GSE16476 cohort (EFS).

Table S3. Univariate and multivariate analyses of different prognostic parameters in the GSE16476 cohort (OS).

Table S4. Univariate and multivariate analyses of different prognostic parameters in the E-MTAB-1781 cohort (EFS).

Table S5. Univariate and multivariate analyses of different prognostic parameters in patients with MYCN amplification in the E-MTAB-1781 cohort (EFS).

Table S6. Univariate and multivariate analyses of different prognostic parameters in patients with MYCN amplification in the E-MTAB-1781 cohort (OS).

Table S7. The docking scores of YG1702 and 36 analogs.

Table S8. ShRNA, siRNA, and oligonucleotide sequences.

Data file S1. Changes in the cellular metabolic profile after ALDH18A1 overexpression in SK-N-SH cells.

Data file S2. Altered biochemical pathways after ALDH18A1 overexpression in SK-N-SH cells.

Data file S3. Changes in the cellular metabolic profile after ALDH18A1 knockdown in SK-N-BE(2) cells.

Data file S4. Altered biochemical pathways after ALDH18A1 knockdown in SK-N-BE(2) cells.

Data file S5. Original data.

Data file S6. Primers for polymerase chain reaction and quantitative reverse transcription polymerase chain reaction.

References (7078)

REFERENCES AND NOTES

Acknowledgments: We thank G.-M. Yan (ZhongShan School of Medicine, Sun Yat-sen University) for the HEB cell line (a human glial cell line). We thank Y.-Y. Wu (Department of Gastroenterology, XinQiao Hospital of the Third Military Medical University) for the normal colonic mucosal epithelial NCM460 cell line. We thank H. Xiao (Daping Hospital of the Third Military Medical University) and B. Wang (Pediatric Hospital of Chongqing Medical University) for the tissue samples. We thank S. Nasi for plasmid pCS Omomyc. Funding: This study was supported by grants from the National Key Research and Development Program of China (2016YFA0202104 to S.-C.Y.), the National Natural Science Foundation of China (81572880 to S.-C.Y. and 81821003 to X.-W.B.), the Postgraduate Education Foundation of Chongqing (YJG153062 to S.-C.Y.), and the Key Clinical Research Program of Southwest Hospital (SWH2016ZDCX1005, SWH2017ZDCX1003, and SWH2019TD-01 to S.-C.Y.). Author contributions: Y.-F.G., J.-J.D., J.W., L.L., and D.W.: acquisition of data, analysis and interpretation of data, and drafting of the manuscript. X.-Z.L., J.Y., H.-R.Z., J.L., Y.-J.Y., Z.-Y.Y., J.C., X.-M.L., T.T., T.-T.H., and F.W.: acquisition of data. X.-Y.Y. and Q.W.: statistical analysis. X.-W.B.: analysis and interpretation of data, critical revision of the manuscript for important intellectual content, and study supervision. S.-C.Y.: study conception and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, critical revision of the manuscript for important intellectual content, statistical analysis, funding acquisition, and study supervision. All authors have read the manuscript and approved it for publication. Competing interests: Part of the work described in the study was included in pending patent applications [2018105309109, application of aldehyde dehydrogenase 18A1 gene and its encoded product in MYCN-amplified NB (S.-C.Y., Y.-F.G., and J.-J.D.); 2018105296876, application of YG1702 in the preparation of ALDH18A1-specific inhibitor (S.-C.Y., Y.-F.G., and J.-J.D.)]. All other authors declare that they have no competing interests. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article