Therapeutic targeting of circ‐CUX1/EWSR1/MAZ axis inhibits glycolysis and neuroblastoma progression

Abstract Aerobic glycolysis is a hallmark of metabolic reprogramming in tumor progression. However, the mechanisms regulating glycolytic gene expression remain elusive in neuroblastoma (NB), the most common extracranial malignancy in childhood. Herein, we identify that CUT‐like homeobox 1 (CUX1) and CUX1‐generated circular RNA (circ‐CUX1) contribute to aerobic glycolysis and NB progression. Mechanistically, p110 CUX1, a transcription factor generated by proteolytic processing of p200 CUX1, promotes the expression of enolase 1, glucose‐6‐phosphate isomerase, and phosphoglycerate kinase 1, while circ‐CUX1 binds to EWS RNA‐binding protein 1 (EWSR1) to facilitate its interaction with MYC‐associated zinc finger protein (MAZ), resulting in transactivation of MAZ and transcriptional alteration of CUX1 and other genes associated with tumor progression. Administration of an inhibitory peptide blocking circ‐CUX1‐EWSR1 interaction or lentivirus mediating circ‐CUX1 knockdown suppresses aerobic glycolysis, growth, and aggressiveness of NB cells. In clinical NB cases, CUX1 is an independent prognostic factor for unfavorable outcome, and patients with high circ‐CUX1 expression have lower survival probability. These results indicate circ‐CUX1/EWSR1/MAZ axis as a therapeutic target for aerobic glycolysis and NB progression.


Introduction
Neuroblastoma (NB), a malignant tumor arising from primitive neural crest, accounts for 15% of cancer-related mortality in childhood (Brodeur, 2003). For high-risk NB, the clinical outcome remains poor in despite of multimodal therapeutic approaches (Brodeur, 2003). To support their tumorigenecity and aggressiveness, tumor cells uptake and convert a large amount of glucose into lactic acid even in the presence of adequate oxygen, which is known as aerobic glycolysis or Warburg effect (Hanahan & Weinberg, 2011). Activation of oncogenes (c-Myc) or inactivation of tumor suppressors (p53) contributes to aberrant expression of transporters and metabolic enzymes of aerobic glycolysis (Shim et al, 1997;Bensaad et al, 2006;Yang et al, 2014). However, identification of transcriptional regulators for aerobic glycolysis in NB still remains to be determined. Circular RNAs (circRNAs) are a novel class of endogenous noncoding RNAs that are generated from exons or introns, and may function as microRNA (miRNA) sponges, regulators of transcription and splicing, or partners of RNA-binding protein (RBP) (Lasda & Parker, 2014;Li et al, 2015b). For example, circRNA antisense to cerebellar-degeneration-related protein 1 (Cdr1as) harbors 70 binding sites for miR-7 to regulate its transport in neurons (Piwecka et al, 2017). A special class of exon-intron circRNAs, such as circEIF3J and circPAIP2, is predominantly localized in the nucleus and enhance transcription of their parental genes (Li et al, 2015b). In addition, intronic circRNAs, such as ci-ankrd52, are able to regulate transcription efficiency of parental genes by binding to RNA polymerase II (Zhang et al, 2013). However, the roles of circRNAs in aerobic glycolysis during tumor progression remain largely elusive.
In this study, we identify CUT-like homeobox 1 (CUX1) as a transcription factor facilitating aerobic glycolysis and tumor progression in NB. We also reveal the oncogenic functions of a CUX1-generated intron-containing circular RNA (circ-CUX1) in tumorigenesis and aggressiveness. Elevated circ-CUX1 promotes the aerobic glycolysis, growth, and aggressiveness of NB cells by binding to EWS RNA-binding protein 1 (EWSR1) and facilitating its interaction with MYC-associated zinc finger protein (MAZ), resulting in MAZ transactivation and transcriptional alteration of CUX1 and other genes associated with tumor progression, suggesting circ-CUX1/EWSR1/MAZ axis as a therapeutic target for aerobic glycolysis and NB progression.

CUX1 facilitates aerobic glycolysis and tumor progression
Comprehensive analysis of a microarray dataset (GSE16476) (Molenaar et al, 2012) of 88 NB cases identified 8 differentially expressed glycolytic genes (fold change > 2.0, P < 0.05) that were consistently associated with death, advanced international neuroblastoma staging system (INSS) stages, and clinical progression (Fig 1A). Similarly, we also found 52 transcription factors consistently associated with these clinical features (Fig 1A), which were subjective to further overlapping analysis with potential transcription factors regulating all of 8 glycolytic genes revealed by Genomatix program (http://www. genomatix.de/solutions/genomatix-software-suite.html). The results indicated CUX1 as the top transcription factor ranking by number of potential targets (Fig 1A). Higher transcript levels of CUX1 isoform p200 were noted in NB cell lines SH-SY5Y, IMR32, and SK-N-AS, while p75 (Goulet et al, 2002) was expressed at very low levels (Appendix Fig S1A). Consistently, elevated levels of p200 CUX1 and its proteolytically processed isoform p110 were noted in these NB cells, cervical cancer HeLa cells, colon cancer LoVo cells, and prostate cancer PC-3 cells, than those of non-transformed normal MCF 10A cells (Appendix Fig S1A). However, both transcript and protein levels of CDP/cut alternatively spliced cDNA (CASP) (Gillingham et al, 2002) were not differently expressed between normal and tumor cells (Appendix Fig S1A). In an independent cohort of 54 primary NB tissues, the transcript levels of p200 CUX1, but not of CASP, were higher than those in normal fetal adrenal medulla (P < 0.05, Appendix Fig S1A), especially in cases with poor stroma (P = 0.0002) or advanced INSS stages (P = 0.007), without association with MYCN amplification (P = 0.56, Appendix Fig S1B).

Circ-CUX1 is up-regulated in NB tissues and cell lines
The copy number of CUX1 gene, locating at chr7: 101460882-101901513, was neither significantly altered in NB (Appendix Fig  S4A) nor associated with death, MYCN amplification, INSS stages, or survival of NB cases derived from Oncogenomics database (Appendix Fig S4A and B). There were no genetic variants of CUX1 gene in 563 NB cases of public datasets (Appendix Fig S4C). Among 37 potential circRNAs generated from CUX1 gene in circBase (Glazar et al, 2014), 15 had more than 2 read scores, while further RT-PCR validation revealed 7 detectable circRNAs in IMR32 cells (

Circ-CUX1 exerts an oncogenic role in tumor progression
We further observed the potential effects of circ-CUX1 on biological features of tumor cells. The ECAR was increased and decreased in IMR32, SH-SY5Y, LoVo, and PC-3 cells stably transfected with circ-CUX1 or sh-circ-CUX1, along with reduced and enhanced OCR, while transfection of circ-CUX1-Mut did not affect these features ( Fig EV2E). Notably, ectopic expression of circ-CUX1 increased the glucose uptake, lactate production, and ATP levels of IMR32 cells, which was attenuated by 2-DG treatment (Appendix Fig S6A). Stable over-expression or knockdown of circ-CUX1 increased and decreased the anchorage-independent growth and invasion of IMR32 and SH-SY5Y cells, respectively (Fig 3A and B). Consistently, stable transfection of circ-CUX1 or sh-circ-CUX1 #1 into IMR32 cells resulted in a significant increase or decrease in growth, tumor weight, Ki-67 proliferation index, CD31-positive microvessels, glucose uptake, lactate production, and ATP levels of subcutaneous xenograft tumors in nude mice (Fig 3C-E). Athymic nude mice treated with tail vein injection of IMR32 cells with stable overexpression or knockdown of circ-CUX1 displayed more or less lung metastatic colonies, with lower or greater survival probability, respectively ( Fig 3F). These results indicated that circ-CUX1 exerted an oncogenic role in tumorigenesis and aggressiveness.

Circ-CUX1 directly interacts with EWSR1 protein in NB cells
To explore the protein partner of circ-CUX1, RNA pull-down was performed using biotin-labeled probes generated by ligation of linear transcript in vitro (Petkovic & Muller, 2015) or synthesized as oligonucleotides targeting junction site ( Fig 4A). Mass spectrometry revealed 47 proteins consistently pulled down by exogenous circ-CUX1 and antisense probe targeting endogenous circ-CUX1, but not by linear transcript or sense probe, and 18 of them were RBPs defined by RBPDB (http://rbpdb.ccbr.utoronto.ca). Further comprehensive analysis of protein interacting with transcription factors of CUX1 promoter revealed by Genomatix and BioGRID database ▸ Figure 2. Circ-CUX1 is up-regulated and enhances CUX1 expression in NB.

A
Schematic illustration showing the generation of circ-CUX1 from CUX1. B RT-PCR or PCR assay revealing the amplification of circ-CUX1 from cDNA or genomic DNA (gDNA) of IMR32 and HeLa cells, with validation by Sanger sequencing. C, D Real-time qRT-PCR (C, normalized to b-actin, n = 6) and Northern blot (D) indicating the circ-CUX1 levels in cell lines and IMR32 cells transfected with empty vector (mock) or circ-CUX1 and treated with RNase R (3 U lg À1 ). One-way ANOVA, *P < 0.05 versus HEK293. E, F Real-time qRT-PCR (E, normalized to b-actin) and RNA-FISH with antisense junction probe and RNase R (3 U lg À1 ) treatment (F) showing the distribution and localization (arrowheads) of circ-CUX1 in IMR32 cells stably transfected with mock or circ-CUX1 (n = 5), using GAPDH and U1 as controls. Scale bar: 10 lm. G Real-time qRT-PCR assay indicating circ-CUX1 expression (normalized to b-actin) and its correlation with CUX1 levels (Pearson's correlation coefficient) in tumor tissues, normal fetal adrenal medulla (FAM), or normal counterparts. Student's t-test, **P < 0.01 versus FAM or normal. H Western blot showing the CUX1 levels in tumor cells stably transfected as indicated.
Notably, higher MAZ levels were observed in NB tissues than those in normal fetal adrenal medulla (P < 0.0001), especially in those with poor stroma (P = 0.0205) or advanced INSS stages (P = 0.0097), without association with MYCN amplification (P = 0.6445, Appendix Fig S7A). Among 60 MAZ target genes derived from RNA-seq results and ChIP-X analysis, the expression of CUX1, S100 calcium-binding protein A9 (S100A9), mucin 4 (MUC4), Kruppel-like factor 10 (KLF10), or thioredoxin-interacting protein (TXNIP) was significantly correlated with that of MAZ in 54 NB cases (Appendix Fig S7B). In addition, higher expression of EWSR1, MAZ, S100A9, or MUC4 and lower expression of KLF10 or TXN1P were associated with poor survival of NB patients (GSE16476, Appendix Fig S7C). In RNA pull-down and chromatin isolation by RNA purification (ChIRP) (Chu & Chang, 2016) assays using biotinlabeled circ-CUX1 junction probe, circ-CUX1 was associated with EWSR1 and MAZ protein, and promoters of target genes (CUX1, S100A9, MUC4, KLF10, or TXNIP), but not with transcripts of downstream genes in SH-SY5Y cells (Appendix Fig S8A). Ectopic expression or knockdown of circ-CUX1 enhanced and reduced the binding of MAZ to these target gene promoters in IMR32 and SH-SY5Y cells, while silencing or over-expression of EWSR1 abolished these effects ( Fig 5E and Appendix Fig S8B). The activity of wild-type CUX1 promoter, but not of that with mutant MAZ-binding site, was increased and decreased by ectopic expression or knockdown of circ-CUX1 ( Fig 5F and Appendix Fig S8C). In addition, the levels of CUX1, S100A9, MUC4, KLF10, or TXNIP were significantly altered in IMR32 and SH-SY5Y cells stably transfected with circ-CUX1 or shcirc-CUX1 #1 (Fig 5G and H  A, B Soft agar (A) and Matrigel invasion (B) assays showing the anchorage-independent growth and invasion capability of IMR32 and SH-SY5Y cells stably transfected with empty vector (mock), circ-CUX1, scramble shRNA (sh-Scb), or sh-circ-CUX1 (n = 5). Scale bars: 100 lm. Student's t-test, one-way ANOVA, *P < 0.05 versus mock or sh-Scb. C Representative fluorescence images, in vivo growth curve, and weight at the end points of subcutaneous xenograft tumors formed by IMR32 cells stably transfected as indicated in nude mice (n = 5 for each group). Student's t-test, one-way ANOVA, *P < 0.05 versus mock or sh-Scb. D, E Immunohistochemical staining showing the expression of Ki-67 and CD31 (D) and glucose uptake, lactate production, and ATP levels (E) within subcutaneous xenograft tumors formed by IMR32 cells stably transfected as indicated (n = 5 for each group). Scale bars: 100 lm. Student's t-test, **P < 0.01 versus mock or sh-Scb. F Representative images, HE staining (arrowheads), quantification of lung metastatic colonization, and Kaplan-Meier curves of nude mice treated with tail vein injection of IMR32 cells stably transfected as indicated (n = 5 for each group). Scale bar: 100 lm. Student's t-test, **P < 0.01 versus mock or sh-Scb. Log-rank test for survival comparison.
Data information: Data are presented as mean AE SEM. Exact P values are specified in Appendix Table S4. Therapeutic peptide blocking the circ-CUX1-EWSR1 interaction Based on the importance of RRM domain (especially 394-397 or 406-410 aa) of EWSR1 in interacting with circ-CUX1, we further designed a cell-penetrating peptide, named as EWSR1 inhibitory peptide of 22 amino acids (EIP-22), that might potentially block circ-CUX1-EWSR1 interaction (Fig 6A). Treatment of SH-SY5Y cells with EIP-22 resulted in its obvious aggregation within the nucleus (Fig 6B). Biotin-labeled peptide pull-down assay revealed the binding of EIP-22 to endogenous circ-CUX1 in SH-SY5Y cells (Appendix Fig S9B). In addition, EIP-22 treatment reduced the interaction between circ-CUX1 and EWSR1, but not that of pri-miR-222 and EWSR1 (Ouyang et al, 2017) or circACC1 and AMP-activated protein kinase beta 1 (AMPKb1) (Li et al, 2019; Fig 6C and D). Administration of EIP-22 inhibited the viability, anchorage-independent growth, and invasion of SH-SY5Y cells (Appendix Fig S9C,  Fig 6E and F), with alteration of circ-CUX1 downstream gene expression (Appendix Fig S9D). In contrast, EIP-22 treatment resulted in no significant alteration in the viability of MCF 10A, non-transformed normal cells with very low circ-CUX1 expression (Fig 2C and Appendix Fig S9C). Notably, EIP-22 treatment synergized the suppressing effects of glycolysis inhibitors, 2-DG and 3-bromopyruvate (3-BP) (Cardaci et al, 2012;Zhang et al, 2014), on the viability, growth, and invasion of IMR32 and SH-SY5Y cells (Appendix Fig  S9E-G). Intravenous administration of EIP-22 significantly reduced the growth, tumor weight, Ki-67 proliferation index, and CD31-positive microvessels, altered circ-CUX1 target gene expression, and decreased the glucose uptake, lactate production, and ATP levels in subcutaneous xenograft tumors formed by injection of SH-SY5Y cells (Fig 6G and Appendix Fig S10A-C). Moreover, administration of EIP-22 via tail vein reduced the lung metastatic colonies and prolonged the survival time of athymic nude mice received tail vein injection of SH-SY5Y cells (Fig 6H). These data suggested that EIP-22 suppressed NB progression by blocking circ-CUX1-EWSR1 interaction.
Therapeutic lentivirus-mediated circ-CUX1 knockdown in vivo We further explored the therapeutic efficiencies of circ-CUX1 knockdown on athymic nude mice bearing xenograft tumors formed by subcutaneous or tail vein injection of IMR32 cells. Lentivirusmediated knockdown of circ-CUX1 significantly reduced the growth, tumor weight, Ki-67 proliferation index, and CD31-positive microvessels of subcutaneous xenograft tumors (Appendix Fig S11A  and B), with altered expression of circ-CUX1 target genes (Appendix Fig S11C). The glucose uptake, lactate production, and ATP levels were significantly decreased in xenograft tumors of nude mice received tail vein injection of lentivirus carrying sh-circ-CUX1 (Appendix Fig S11D). In addition, administration of lentivirus carrying sh-circ-CUX1 #1 decreased the lung metastatic colonies and prolonged the survival time of nude mice (Appendix Fig S11E). These results indicated that lentivirus-mediated circ-CUX1 knockdown inhibited aerobic glycolysis and NB progression in vivo.

Discussion
Recent studies show that although LDHA and LDHB promote tumorigenicity, they are dispensable for aerobic glycolysis in NB (Dorneburg et al, 2018), suggesting the involvement of other glycolytic genes in this process. So far, interrogative screening of transcriptional regulators of aerobic glycolysis in NB remains unknown.
In this study, we identified CUX1 as a transcription factor facilitating the expression of glycolytic genes ENO1, GPI, and PGK1 in NB. We demonstrate that circ-CUX1 interacts with EWSR1 protein to increase MAZ transactivation, which subsequently regulates the transcription of CUX1 and other genes associated with tumor progression in cis and in trans (Fig 6I), such as S100A9 ( . The discovery of circ-CUX1/ EWSR1/MAZ axis represents a promising step for therapeutic intervention against tumors. CUX1 is a transcription factor involved in embryonic development (Michl et al, 2005;Harada et al, 2008) and regulates cellular proliferation, migration, and epithelial-to-mesenchymal transition, suggesting its emerging roles in tumorigenesis and aggressiveness (Michl et al, 2005). Elevated CUX1 expression has been documented in many tumors and is associated with poor survival of patients (Liu et al, 2013). Full-length p200 CUX1 binds rapidly but only transiently to DNA (Liu et al, 2013), while its proteolytic product p110 isoform activates gene transcription (Harada et al, 2008;Kedinger et al, 2009). In this study, our results indicated that CUX1 was an independent prognostic marker for progression and poor outcome of NB. In addition, p110 CUX1 promoted the expression of target genes ▸ Figure 4. Circ-CUX1 directly interacts with EWSR1 protein in NB cells.
A Schematic illustration, Coomassie Blue staining, and Venn diagram showing the differential proteins pulled down by biotin-labeled linear or circular exogenous circ-CUX1, sense (S) or antisense (AS) probe targeting junction site of endogenous circ-CUX1 from IMR32 cells, and over-lapping analysis with RBP and proteins interacting with potential transcription factors of p200 CUX1 revealed by Genomatix program and BioGRID database. B Western blot (upper panel) and RT-PCR (lower panel) assays indicating the proteins and transcripts pulled down by biotin-labeled linear or circular exogenous circ-CUX1, sense (S) or antisense (AS) probe targeting junction site of endogenous circ-CUX1 from IMR32 cell lysates, using AS probes of p200 CUX1 or CASP as controls. C RIP and real-time qRT-PCR assays revealing the interaction of EWSR1 with circ-CUX1, p200 CUX1, or CASP in SH-SY5Y cells and those stably transfected with empty vector (mock) or circ-CUX1 (n = 5). Student's t-test, *P < 0.05 versus IgG; D P < 0.01 versus mock. D 3D confocal images of dual RNA-FISH and immunofluorescence staining assay showing the co-localization of circ-CUX1 and EWSR1 in IMR32 cells stably transfected with mock or circ-CUX1. Scale bar: 10 lm. E RNA EMSA determining the interaction between biotin-labeled circ-CUX1 probe and EWSR1 protein within nuclear extracts of SH-SY5Y cells (arrowheads). F In vitro binding assay depicting the recovered circ-CUX1, p200 CUX1, or CASP detected by RT-PCR (upper panel) after incubation with GST-tagged recombinant EWSR1 protein validated by Western blot (lower panel).
Data information: Data are presented as mean AE SEM. Exact P values are specified in Appendix Table S4. Source data are available online for this figure. ENO1, GPI, and PGK1 in NB cells. As a glycolytic enzyme, ENO1 acts as a metabolic tumor promoter by contributing to Warburg effect . GPI is a housekeeping cytosolic enzyme responsible for catalytic interconversion between glucose-6-phosphatase and fructose-6-phosphate, and plays a key role in glycolytic pathway ( Zdralevi c et al, 2017). During the glycolytic process, PGK1 contributes to ATP generation and participates in tumor progression . Our gain-and loss-of-function studies indicated that CUX1 promoted the aerobic glycolysis, growth, and invasiveness of NB cells, suggesting its oncogenic roles in NB progression. Human CUX1 gene locates at chromosome 7q22, a region associated with copy number gain that contributes to multidrug resistance in NB (Mazzocco et al, 2015). However, we found no alteration of copy number or genetic variants of CUX1 in NB cohorts, indicating other mechanisms facilitating its over-expression. CircRNAs play important roles in regulating gene expression at post-transcriptional or transcriptional levels (Hansen et al, 2013;Li et al, 2015b). For example, ciRS-7 and circSry serve as sponges of miR-7 and miR-138 in the cytoplasm (Hansen et al, 2013). Exonic circRNAs also exert regulatory functions in the cytoplasm by forming a ribonucleoprotein complex with miRNA and AGO protein (Lasda & Parker, 2014). Meanwhile, exon-intron circRNAs are predominantly localized in the nucleus and regulate their parent gene expression in a cis-acting manner through specific RNA-protein interaction (Li et al, 2015b). In this study, circ-CUX1 was identified as an intron-containing circRNA up-regulated in tumor tissues and cells. Circ-CUX1 enhanced the expression of CUX1 at transcriptional level, and tumor-promoting functions of circ-CUX1 were mediated, at least in part, through interacting with EWSR1 protein in NB cells.
As one member of EWS family of RNA-binding proteins, EWSR1 participates in gene transcription, splicing, and miRNA processing (Luo et al, 2015). Chromosomal translocation of ESWR1 has been discovered in Ewing sarcoma (Sohn et al, 2010). However, our results revealed no EWS-Fli1 gene fusion in NB cells. Due to lack of DNA-binding domain, EWSR1 usually acts as a potent transcriptional cofactor in tumor progression via interacting with transcription regulatory proteins, such as CREB-binding protein and p300 (Chakravarti et al, 1996). In this study, we found that RRM domain of EWSR1 was necessary for its interaction with ZNF domain of MAZ. As a ubiquitously expressed transcription factor, MAZ binds to GC-rich cis-elements through its C2H2-type ZNF motif (Parks & Shenk, 1996) and activates transcription of KRAS and vascular endothelial growth factor (VEGF) in pancreatic cancer, cervical cancer, and glioblastoma cells (Smits et al, 2012;Cogoi et al, 2013). Our evidence indicated that through interplay with its cofactor EWSR1, MAZ regulated the transcription of CUX1, S100A9, MUC4, KLF10, or TXNIP in NB cells. Notably, circ-CUX1 bound to RRM region of EWSR1, resulting in EWSR1-mediated MAZ transactivation, suggesting the oncogenic roles of circ-CUX1/EWSR1/MAZ axis in aerobic glycolysis and tumor progression.
In summary, we demonstrate that elevated CUX1 and its generated circ-CUX1 are associated with poor outcome of NB patients, ◀ Figure 5. Circ-CUX1 facilitates EWSR1-mediated MAZ transactivation in NB cells.

A
Volcano plots indicating RNA-seq results of 781 up-regulated and 434 down-regulated genes in IMR32 cells upon stable circ-CUX1 over-expression (fold change > 1.5, P < 0.05). B ChIP-X analysis (left panel) showing top five transcription factors regulating the altered genes, and Venn diagram (right panel) indicating the identification of MAZ by over-lapping analysis of five transcription factors and EWSR1-interacting proteins derived from BioGRID database. C Co-IP and Western blot assays showing the interaction between EWSR1 and MAZ in IMR32 and SH-SY5Y cells stably transfected with circ-CUX1 or sh-circ-CUX1 #1, respectively. D BiFC assay revealing the interaction (arrowheads) of EWSR1 and MAZ in IMR32 and SH-SY5Y cells stably transfected as indicated, with nuclei stained by DAPI. Scale bars: 10 lm. E ChIP assay showing the binding of MAZ (normalized to input, n = 5) to target gene promoters in IMR32 cells stably transfected as indicated. One-way ANOVA, *P < 0.05 versus mock+sh-Scb. F Dual-luciferase assay revealing the relative activity of p200 CUX1 promoter with wild-type or mutant MAZ-binding site in IMR32 cells stably transfected as indicated (n = 5). One-way ANOVA, *P < 0.05 versus mock+sh-Scb. G, H Real-time qRT-PCR (G, normalized to b-actin, n = 5) and Western blot (H) assays showing the expression of EWSR1, MAZ, and their target genes in IMR32 cells stably transfected as indicated. One-way ANOVA, *P < 0.05 versus mock+sh-Scb.
Data information: Data are presented as mean AE SEM. Exact P values are specified in Appendix Table S4.
Data information: Data are presented as mean AE SEM. Exact P values are specified in Appendix Table S4. and exert oncogenic roles in aerobic glycolysis and tumor progression. Mechanistically, circ-CUX1 binds to EWSR1 protein to facilitate MAZ transactivation, resulting in transcriptional alteration of CUX1 and other genes associated with NB progression. An inhibitory peptide (EIP-22) blocking circ-CUX1-EWSR1 interaction or lentivirus-mediated circ-CUX1 knockdown suppresses the aerobic glycolysis, tumorigenesis, and aggressiveness of NB cells. Combinational administration of EIP-22 and glycolysis inhibitors (2-DG or 3-BP) targeting HK2, GPI, or GAPDH (Cardaci et al, 2012;Zhang et al, 2014) exerts synergistic effects in suppressing growth and aggressiveness of NB cells. Due to limited size of cohort, the prognostic value of circ-CUX1 and CUX1 and their association with MYCN amplification in NB warrant further investigation. This study extends our knowledge about the regulation of aerobic glycolysis by transcription factor and its generated circRNA, and suggests that circ-CUX1/EWSR1/MAZ axis may be a potential therapeutic target for NB.

Materials and Methods
Cell culture   Table S1). The transcript levels were analyzed by 2 ÀDDCt method. De novo RNA synthesis was blocked by ActD (5 lg ml À1 ) treatment, while mRNA stability was examined by transcript levels at indicated time points.

Northern blot
The non-junction and junction probes specific for circ-CUX1 were synthesized and labeled by digoxigenin (DIG, Appendix Table S2). For Northern blot, 20 lg of total RNA was separated on 3-(N-morpholino)propanesulfonic acid-buffered 2% (w/v) agarose gel containing 1.2% (v/v) formaldehyde under denaturing condition at 80 V for 4 h, and transferred to Hybond-N+ membrane (Pall Corp., Port Washington, NY). Hybridization was performed at 65°C for 16-18 h in DIG Easy Hyb solution (Roche) and detected by anti-DIG antibody (1:500 dilution) and chemiluminescence substrate CSPD (Roche).

Rescue of target gene expression
To rescue circ-CUX1 knockdown-altered target gene expression, EWSR1 or MAZ was transfected into stable cell lines. To restore target gene expression altered by circ-CUX1, shRNAs against EWSR1 or MAZ (Appendix Table S3) were transfected into tumor cells using GeneSilencer Transfection Reagent (Genlantis, San Diego, CA).

RNA-seq assay
Total RNA of tumor cells (1 × 10 6 ) was extracted using TRIzol â reagent (Life Technologies, Inc.). Library preparation and transcriptome sequencing on an Illumina HiSeq X Ten platform were carried out at Novogene Bioinformatics Technology Co., Ltd. (Beijing, China). Fragments per kilobase of transcript per million fragments mapped (FPKM) of each gene were calculated.

RNA pull-down and mass spectrometry
Biotin-labeled oligonucleotide probes targeting junction sites of circRNAs were synthesized (Invitrogen). Linear circ-CUX1 was in vitro transcribed using Biotin RNA Labeling Mix (Roche) and T7 RNA polymerase, incubated with guide oligonucleotides (Appendix Table S2), circularized using T4 RNA ligase I, treated with RNase R, and purified with RNeasy Mini Kit (Qiagen Inc.). RNA pull-down was performed as previously described . Retrieved protein was detected by Western blot or mass spectrometry (Wuhan Institute of Biotechnology, Wuhan, China), while recovered transcripts were measured by RT-PCR using primers (Appendix Table S1). In ChIRP assay, cells were harvested, cross-linked, sonicated, hybridized with probes, and mixed with streptavidin magnetic beads (Chu & Chang, 2016). The retrieved DNA was detected by PCR using primers (Appendix Table S1).

RNA-FISH
Biotin-labeled antisense or sense probe for circ-CUX1 junction was synthesized (Appendix Table S2). The probes for GAPDH and U1 were generated by in vitro transcription of PCR products (Appendix Table S1) using DIG Labeling Kit (MyLab Corporation, Beijing, China). Cells were incubated with 40 nM FISH probe in hybridization buffer (100 mg ml À1 dextran sulfate, 10% formamide in 2 × SSC) at 37°C for 16 h, with or without RNase R (3 U lg À1 ) treatment. The signals of circ-CUX1 were detected by Fluorescent In Situ Hybridization Kit (RiboBio, Guangzhou, China), with nuclei staining by 4 0 ,6-diamidino-2-phenylindole (DAPI).

Design and synthesis of inhibitory peptides
Based on interacting region of EWSR1 revealed by mutagenesis and in vitro binding assays, wild-type and mutant inhibitory peptides blocking circ-CUX1 and EWSR1 interaction were designed and synthesized by linking with biotin-labeled 11 amino acid cell-penetrating peptide (YGRKKRRQRRR) from Tat protein transduction domain at the N-terminus and conjugating with fluorescein isothiocyanate (FITC) at the C-terminus (ChinaPeptides Co. Ltd, Shanghai, China), with purity larger than 95%.

Biotin-labeled peptide pull-down
Total RNA was isolated using RNeasy Mini Kit (Qiagen Inc.) and incubated with biotin-labeled peptide at 4°C overnight. Then, incubation of RNA-peptide complex with streptavidin-agarose was undertaken at 4°C for 2 h. Beads were extensively washed, and circRNAs pulled down were measured by real-time qRT-PCR.

Aerobic glycolysis and seahorse extracellular flux assays
Cellular glucose uptake, lactate production, and ATP levels were detected as previously described (Ma et al, 2014). ECAR and OCR were measured in XF media under basal conditions and in response to glucose (10 mM), oligomycin (2 lM), and 2-deoxyglucose (50 mM) using a Seahorse Biosciences XFe24 Flux Analyzer (North Billerica, MA).
In vivo growth, metastasis, and therapeutic assays All animal experiments were carried out in accordance with NIH Guidelines for the Care and Use of Laboratory Animals, and approved by the Animal Care Committee of Tongji Medical College (approval number: Y20080290). For in vivo tumor growth and experimental metastasis studies, tumor cells (1 × 10 6 or 0.4 × 10 6 ) were injected into dorsal flanks or tail vein of blindly randomized 4-week-old female BALB/c nude mice (National Rodent Seeds Center, Shanghai, China) breeding at specific pathogen free (SPF) condition (n = 5 per group) (Zhang et al, 2012;Zhao et al, 2016;Li et al, 2018a,b). For in vivo therapeutic studies, tumor cells (1 × 10 6 or 0.4 × 10 6 ) were injected into dorsal flanks or tail vein of nude mice, respectively. One week later, mice were blindly randomized and treated by tail vein injection of synthesized cellpenetrating peptide (ChinaPeptides, Shanghai, China) or lentivirus (1 × 10 7 plaque-forming units) as indicated. The in Vivo Optical Imaging System (In-Vivo FX PRO, Bruker Corporation, Billerica, MA) was applied to acquire fluorescent images of xenograft tumors in nude mice.

Patient tissue samples
Human tissue study was approved by the Institutional Review Board of Tongji Medical College (approval number: 2011-S085). All procedures were conformed to principles set forth by Declaration of Helsinki and Department of Health and Human Services Belmont Report. Written informed consent was obtained from all patients without preoperative chemotherapy or other treatment. Fresh tumor tissues were collected at surgery, validated by pathological diagnosis, and stored at À80°C. Total RNAs of normal fetal adrenal medulla were purchased from Clontech (Mountain View, CA).

Statistical analysis
All data were shown as mean AE standard error of the mean (SEM). Cutoff of gene expression was defined by average values. Two-sided unpaired Student's t-test and one-way ANOVA were used to The paper explained Problem Neuroblastoma (NB) is the most common extracranial tumor in childhood. Although countless efforts have been made to improve the therapeutic efficiency, the outcome of patients suffering from high-risk NB still remains poor. Aerobic glycolysis is a hallmark of metabolic reprogramming that contributes to tumor progression. Further in-depth investigation of mechanisms regulating aerobic glycolysis during NB progression is vital for resolution of these issues.

Results
By integrating analysis of public datasets, we identify that CUX1 and CUX1-generated circular RNA (circ-CUX1) facilitate aerobic glycolysis and NB progression. Mechanistically, transcription factor p110 CUX1, a proteolytic product of p200 CUX1, promotes expression of glycolytic genes, while circ-CUX1 facilitates EWSR1-mediated MAZ transactivation to alter expression of p200 CUX1 and other genes associated with tumor progression. High expression of circ-CUX1 or p200 CUX1 is associated with poor outcome of NB patients. Administration of an inhibitory peptide blocking circ-CUX1-EWSR1 interaction or lentivirus mediating circ-CUX1 knockdown suppresses aerobic glycolysis, growth, and aggressiveness of NB cells.

Impact
Our results extend the knowledge about regulation of aerobic glycolysis by transcription factor and its generated circRNA. First of all, transcription factor CUX1 is essential for glycolytic gene expression during NB progression. Secondly, EWSR1 interacts with MAZ to facilitate its transactivation and regulate downstream gene expression. Thirdly, circ-CUX1 facilitates the interaction of EWSR1 with MAZ. Finally, blocking circ-CUX1-EWSR1 interaction might be a novel therapeutic strategy for NB and other tumors.
ª 2019 The Authors EMBO Molecular Medicine 11: e10835 | 2019 compare difference. Pearson's correlation coefficient assay was used to analyze expression correlation. Log-rank test and Cox regression models were used to assess survival difference and hazard ratio. All statistical tests were considered significant when P < 0.05. Randomization and blinding strategies were used whenever possible. Experimental sample size was determined on the basis of power analysis assuming a significance level (alpha) of 0.05 and a power of 80%. Animal cohort sizes were determined on the basis of similar studies. The exact P-values and number of replicates were indicated in Appendix Table S4.
Expanded View for this article is available online.