p113 isoform encoded by CUX1 circular RNA drives tumor progression via facilitating ZRF1/BRD4 transactivation

Human NB cell lines SH-SY5Y (CRL-2266), SK-N-SH (HTB-11), SK-N-BE(2) (CRL-2271), BE(2)-C (CRL-2268), and IMR-32 (CCL-127), cervical cancer HeLa (CCL-2) cells, prostate cancer PC-3 cells (CRL-1435), nontransformed mammary epithelial MCF 10A (CRL-10317), and embryonic kidney HEK293T (CRL-3216) were obtained from American Type Culture Collection (Rockville, MD). Cell lines were authenticated by short tandem repeat profiling, and used within 6 months after resuscitation. Mycoplasma was regularly examined with Lookout Mycoplasma PCR Detection Kit (Sigma, St. Louis, MO). Tumor cells and HEK293T cells were cultured in RPMI1640 supplied with 10% fetal bovine serum (Gibco, Grand Island, NY), while MCF 10A cells were cultured in DMEM/F12 medium containing 5% horse serum (Invitrogen, Carlsbad, CA) and 20 ng/ml epidermal growth factor (Peprotech, Rocky Hill, NJ). For metabolic SD stress, cells were maintained in RPMI1640 containing 1% fetal bovine serum.

Nuclear, cytoplasmic, and total RNAs were extracted using RNA Subcellular Isolation Kit (Active Motif, Carlsbad, CA) or RNeasy Mini Kit (Qiagen Inc., Redwood City, CA). For circRNA detection, RNase R (3 U/ug, Epicenter, Madison, WI) treatment was undertaken at 37 °C for 15 min, while reverse transcription kit (TakaRa, Dalian, China) was used for cDNA synthesis. Quantification of mRNA and circRNA was performed using a SYBR Green PCR Master Mix (Applied Biosystems, Carlsbad, CA) and primers (Additional file 1: Table S1).

Proteins were extracted with RIPA lysis buffer (Thermo Fisher Scientific, Inc., Waltham, MA). Anti-p113 polyclonal antibody was prepared by immunizing rabbits with synthesized peptide corresponding to C-terminus of p113 (EQQLSAKNSTLKGRRD; ABclonal Biotechnology Co., Ltd., Wuhan, China). Western blot analysis was performed as previously described [14‐16], with antibodies specific for CUX1 (ab230844), transcription factor 3 (TCF3, ab69999), ALDH3A1 (ab186726), NDUFA1 (ab249923), NDUFAF5 (ab240971), β-actin (ab179467, Abcam Inc., Cambridge, MA), CUX1 (sc-514,008), histone H3 (sc-517,576), glyceraldehyde 3-phosphate dehydrogenase (GAPDH, sc-47724), glutathione S-transferase (GST, sc-33614, Santa Cruz Biotechnology, Santa Cruz, CA), Flag-tag (14793S), ZRF1 (12844S), BRD4 (13440S), hemagglutinin (HA)-tag (3724S), or His-tag (12698S, Cell Signaling Technology Inc., Danvers, MA).

Linear ecircCUX1 (hsa_circ_30402) was obtained from NB tissues by PCR (Additional file 1: Table S2) and inserted into pLCDH-ciR (Geenseed Biotech Co., Guangzhou, China). The ecircCUX1-3Flag construct was established by inserting a 3 × Flag-coding sequence upstream of TGA codon within linear ecircCUX1 (Additional file 1: Table S2). Mutation of ecircCUX1 was performed with GeneTailor™ Site-Directed Mutagenesis System (Invitrogen) and primers (Additional file 1: Table S2). Synthesized p113-3Flag sequence was subcloned into pcDNA3.1-mini (Addgene, Cambridge, MA), while p113-3Flag with circular frames was ligated into CV186 (Genechem Co., Ltd., Shanghai, China) or pCMV-HA (Addgene). Human ZRF1 cDNA (1866 bp) was obtained from Genechem Co., Ltd., while BRD4 cDNA (4089 bp) was provided by Dr. Guosong Jiang [17]. Truncations of ZRF1 or BRD4 were obtained by PCR amplification (Additional file 1: Table S2) and subcloned into pCMV-3Tag-1A (Stratagene, Santa Clara, CA), pCDNA4-His (Invitrogen), pGEX-6P-1 (Addgene), or pMal-c4X (Addgene), respectively. The pGEX-6P-1 or pMal-c4X constructs were transformed into E. coli to produce GST-tagged ZRF1 or maltose-binding protein (MBP)-tagged BRD4 proteins. Single guide RNAs (sgRNAs) targeting downstream region of gene transcription start site (Additional file 1: Table S3) were inserted into dCas9-BFP-KRAB (Addgene) [18]. Oligonucleotides specific for short hairpin RNAs (shRNAs, Additional file 1: Table S3) were inserted into GV298 (Genechem Co., Ltd). Lentiviral plasmids were co-transfected with psPAX2 and pMD2G into HEK293T cells to generate infectious lentivirus. Stable cell lines were obtained with puromycin selection for 3-4 weeks.

The internal ribosome entry site (IRES) reporter of ecircCUX1 was amplified with primers (Additional file 1: Table S2) and subcloned into pGL3-Basic (Promega, Madison, WI). Human ZRF1 activity reporter was established by inserting oligonucleotides containing four canonical binding sites (Additional file 1: Table S2) into pGL3-Basic (Promega). Promoter fragment of ALDH3A1 (1179 bp), NDUFA1 (1077 bp), or NDUFAF5 (1079 bp) was amplified from genomic DNA with primers (Additional file 1: Table S2) and subcloned into pGL3-Basic (Promega). Dual-luciferase assay was performed according to the manufacturer’s instructions (Promega).

Genomic insertion of a 3 × Flag coding sequence or mutation of ecircCUX1 was performed by homology-directed repair of a CRISPR-Cas9-induced DNA break. The sgRNA oligonucleotides targeting exon 9 of CUX1 were designed using online program (http://​crispr.​mit.​edu), and ligated into LentiCRISPR V2 vector (Addgene). Donor DNA carrying 1200 bp of homologous sequences upstream and downstream exon 9 of CUX1 with 3 × Flag insertion or open reading frame (ORF) mutation was synthesized. Cells were co-transfected with sgRNA and donor DNA, selected with puromycin for 2-3 weeks, and validated by PCR amplification using primers (Additional file 1: Table S1) and Sanger sequencing.

Total RNA of tumor cells (1 × 10⁶) was isolated using TRIzol™ reagent (Life Technologies, Inc., Gaithersburg, MD). Library preparation and transcriptome sequencing on a BGIseq500 platform were carried out at Beijing Genomics Institute (BGI-Shenzhen, China).

Co-immunoprecipitation (co-IP) was performed as previously described [14‐16], with antibodies for p113 (ABclonal Biotechnology Co., Ltd), GST-tag (sc-33614, Santa Cruz Biotechnology), Flag-tag (14793S), ZRF1 (12844S), BRD4 (13440S), HA-tag (3724S), His-tag (12698S), MBP-tag (2396S, Cell Signaling Technology Inc., MA). Bead-bound proteins were analyzed by Coomassie blue staining, western blot, or mass spectrometry (Wuhan Institute of Biotechnology, Wuhan, China).

ChIP assay was undertaken using EZ-ChIP kit (Upstate Biotechnology, Temacula, CA) [14‐16], with antibodies specific for p113 (ABclonal Biotechnology Co., Ltd), ZRF1 (12844S), or BRD4 (13,440, Cell Signaling Technology Inc.) and primers (Additional file 1: Table S1). ChIP-seq was undertaken at Wuhan Seqhealth Technology Co., Ltd. (Wuhan, China).

Cells were seeded on coverslips and incubated with antibody specific for p113 (ABclonal Biotechnology Co., Ltd), Flag-tag (14793S, Cell Signaling Technology Inc.), 4-hydroxynonenal (4-HNE, MAB3249, Novus Biologicals, Ltd., Centennial, CO), or ALDH3A1 (ab186726, Abcam Inc.) at room temperature for 2 h. Then, coverslips were incubated with Alexa Fluor 594 goat anti-mouse IgG or Alexa Fluor 488 goat anti-rabbit IgG, and stained with 4′,6-diamidino-2-phenylindole (DAPI, 300 nmol·L^− 1, Sigma).

Human p113 ORF (342 bp), ZRF1 cDNA (1866 bp), and BRD4 cDNA (4089 bp) were subcloned into pBiFC-VN173 or pBiFC-VC155 (Addgene), and co-transfected into tumor cells with Lipofectamine 2000 (Invitrogen) for 24 h. The fluorescence was observed with a confocal microscope (Nikon, Japan) [15, 16, 19].

Inhibitory peptides for blocking interaction between p113 and ZRF1 were designed. The 11-amino acid long peptide (YGRKKRRQRRR) from Tat protein transduction domain served as a cell-penetrating peptide. Thus, inhibitory peptides were chemically synthesized by linking with N-terminal biotin-labeled cell-penetrating peptide and C-terminal fluorescein isothiocyanate (FITC, ChinaPeptides Co. Ltd., Shanghai, China), with purity larger than 95%.

Cellular proteins were isolated using 1× cell lysis buffer (Promega), and incubated with biotin-labeled peptide and streptavidin-agarose at 4 °C for 2 h. Beads were extensively washed, and protein pulled down was measured by western blot.

Sedimentation of polysomal fractions was performed as previously described [20]. Briefly, tumor cells were treated with 100 μg/ml of cycloheximide (Sigma) for 5-10 min. Cell extracts were layered on top of 15-30% (w/v) linear sucrose gradient and centrifugated (40,000×g) at 4 °C for 2 h. Polysome-bound transcripts in collected gradient fractions were detected by real-time qRT-PCR.

Target metabolic profiling was performed by gas chromatography-mass spectrometry (GC-MS) at Wuhan Metware Biotechnology Co, Ltd. (Wuhan, China). Briefly, samples were quickly frozen in liquid nitrogen, ultrasonically crushed, and filtered through SPE column. Then, SPE column was extracted with n-hexane and detected by GC-MS.

Cells were seeded in XF cell culture microplates (30,000/well) at 37 °C for 24 h, cultured in FAO assay medium (pH 7.4, 2.5 mmol·L^− 1 glucose, 0.5 mmol·L^− 1 carnitine) without CO₂ for 1 h, and incubated with bovine serum albumin (BSA) or oleic acid-BSA (200 μmol·L^− 1, Sigma). Then, oligomycin (1 μmol·L^− 1), carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP, 2 μmol·L^− 1), and rotenone/antimycin A (0.5 μmol·L^− 1, Sigma) were added, while oxygen consumption rate (OCR) was detected using a Seahorse Biosciences XFe24 Flux Analyzer (North Billerica, MA).

For imaging mitochondrial mass or morphology independent of membrane potential [21], live cells were stained by MitoTracker Green (100 nmol·L^− 1, Invitrogen) at 37 °C for 30 min or transfected with MitoRFP reporter (Addgene) for 48 h, with nuclei staining by Hochest 33,342 (Sigma). Mitochondrial structure was observed by transmission electron microscopy [22]. Cells were treated with JC-1 (10 μg/ml, Sigma) at 37 °C for 20 min, while red fluorescence intensity reflecting Δψm was assessed with a fluorescence microscope. Mitochondrial complex I activity was determined by Mitochondrial Complex I Activity Assay Kit (Sigma).

Cellular nicotinamide adenine dinucleotide (NAD⁺)/nicotinamide adenine dinucleotide (NADH) ratio or ATP levels were measured using NAD⁺/NADH Assay Kit (ab65348, Abcam Inc.) and ATP Determination Kit (A22066, Invitrogen), respectively.

In vitro viability, growth, and invasive capabilities of tumor cells were detected by 2-(4,5-dimethyltriazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT; Sigma) colorimetric [23, 24], soft agar [15, 16, 19], and matrigel invasion [15, 16, 19] assays.

Four-week-old male BALB/c nude mice (National Rodent Seeds Center, Shanghai, China) were reared at specific pathogen free condition. For in vivo tumor growth, metastasis, and peptide therapeutic studies, tumor cells (1 × 10⁶ or 0.4 × 10⁶) were injected into dorsal flanks or tail vein of blindly randomized nude mice (n = 5 per group), respectively. For therapeutic studies, 1 week after injection, mice were blindly randomized and treated by tail vein injection of synthesized cell-penetrating peptides (ChinaPeptides, Shanghai, China) as indicated [15, 16, 19].

All procedures were conformed to principles set forth by Declaration of Helsinki. Written informed consent was obtained from all legal guardians of patients without a history of preoperative chemotherapy or radiotherapy. Human normal dorsal root ganglia were collected from therapeutic abortion. Fresh tumor tissues were collected at surgery, validated by pathological diagnosis, and stored at − 80 °C.

Immunohistochemical staining and quantitative evaluation were performed as previously described [15, 16, 19], with antibody specific for p113 (ABclonal Biotechnology Co.), Ki-67 (ab92742, Abcam Inc.), or CD31 (ab28364, Abcam Inc.).

RNA-seq and ChIP-seq data have been deposited in Gene Expression Omnibus (GEO) database (https://​www.​ncbi.​nlm.​nih.​gov/​geo/​, accession number GSE182329 and GSE182402). Public datasets are available from GEO database (GSE62564, GSE25066, GSE17679, GSE65904, GSE2658) or The Cancer Genome Atlas (TCGA) database (https://​cancergenome.​nih.​gov).

All data were shown as mean ± standard error of the mean (s.e.m.). Cutoff values were determined by average gene expression levels. Student’s t test, analysis of variance (ANOVA), and χ² analysis were applied to compare difference. Fisher’s exact test was applied to analyze statistical significance of overlapping. Log-rank test was used to assess survival difference. All statistical tests were two-sided and considered significant when false discovery rate (FDR)-corrected P values were less than 0.05.

To investigate crucial factors regulating metabolic reprogramming, NB cell lines were treated by serum deprivation (SD), a condition causing metabolic stress [25], while altered proteins were analyzed by isobaric tags for relative and absolute quantification (iTRAQ). There were 381 and 145 differentially expressed proteins in SH-SY5Y and SK-N-BE(2) cells upon SD for 24 h, respectively (Additional file 1: Table S4), and overlapping with Genomatix database (http://​www.​genomatix.​de/​) indicated that expression of four transcription factors was significantly altered (Fig. 1a). Among them, CUT-like homeobox 1 (CUX1) and TCF3 were consistently upregulated in both NB cell lines. Validating western blot assay indicated that expression of major CUX1 isoforms [p200, p110, or CDP/CUT alternatively spliced product (CASP)] and TCF3 remained unaffected, while a new CUX1 isoform with a molecular weight of 13 kDa was unexpectedly upregulated in a time-dependent manner upon SD treatment (Fig. 1b). The amino acid sequence of this new isoform was deduced from exons 10 and 11 of CUX1 (Fig. 1c). Analysis from circRNADb database (http://​reprod.​njmu.​edu.​cn/​circrnadb) identified a spanning junction ORF-containing circRNA (hsa_circ_30402) derived from exons 9-11 of CUX1 (referred as ecircCUX1), which might encode a conserved 113-amino acid protein (p113) in primates (Fig. 1d and Additional file 1: Fig. S1a). Accurate circularization of ecircCUX1 was validated in BE(2)-C cells (Fig. 1a), with cytoplasmic abundance (Additional file 1: Fig. S1b). Notably, SD treatment did not affect the ecircCUX1 levels (Additional file 1: Fig. S1c), but increased their binding to polysomes in NB cells (Additional file 1: Fig. S1d). Using prepared p113 antibody or Flag-tag antibody, western blot assay revealed the increase of p113 levels in NB cells transfected with Flag-tagged wild-type ecircCUX1, but not with ORF mutant vector (Fig. 1e). In addition, ectopic expression of ecircCUX1 or SD treatment led to upregulation of p113, but not of p200, p110 or CASP, in NB cells (Fig. 1f). Next, CRISPR-Cas9-mediated knockin of Flag-tagged ecircCUX1, but not of mutant ecircCUX1, resulted in p113 expression in HEK293T and SK-N-BE(2) cells (Fig. 1g, h and Additional file 1: Fig. S1e, f). Importantly, IRES of ecircCUX1 induced a remarkable luciferase activity (Additional file 1: Fig. S1g), indicating its 5′ cap-independent translation pattern. These results demonstrated that SD stress promoted the expression of p113 derived from a circRNA of CUX1 in NB cells.

Then, we observed the expression profiles of p113 in tumor tissues and cells. Higher p113 levels were noted in NB specimens and cell lines, than those of normal dorsal root ganglia (Fig. 2a). Endogenous p113 expression was noted in nuclei of tumor cells (Additional file 1: Fig. S2a), which was detected in 28/42 (66.7%) NB cases, with higher expression in those with elder age (P = 0.001), poor differentiation (P = 0.001), higher mitosis karyorrhexis index (P = 0.002), or advanced international neuroblastoma staging system (INSS) stages (P = 0.003, Additional file 1: Table S5). In these patients, high p113 expression was associated with poor survival probability (P < 1.0 × 10^− 4, Additional file 1: Fig. S2b). Elevated p113 expression was also detected in a variety of cancer tissues (Additional file 1: Fig. S2c). Using SH-SY5Y, SK-N-SH, BE(2)-C, and IMR32 cells (with low or high p113 levels) as models, stable transfection of wild-type ecircCUX1 or two independent shRNAs targeting junction site of ecircCUX1 (sh-ecircCUX1) resulted in over-expression or silencing of p113, but not of p200 or p110 (Fig. 2b and Additional file 1: Fig. S2d), mainly localizing at nuclei (Fig. 2c, d). No alteration of p113 expression was observed in NB cells stably transfected with ORF mutant form of ecircCUX1 (Fig. 2b).

Since previous studies implicate the potential impacts of SD on lipid metabolic reprogramming [26], lipid metabolite profiling assay was further undertaken, which revealed the increase of long-chain fatty acid production in SH-SY5Y cells stably transfected with ecircCUX1 (Fig. 2e). Of note, over-expression of ecircCUX1 or p113, but not of ecircCUX1 with ORF mutation (ecircCUX1 Mut), increased the levels of palmitic acid (C16:0), palmitoleic acid (C16:1n7), oleic acid (C18:1n9c), stearic acid (C18:0), and arachidonic acid (C20:4n6), while knockdown of ecircCUX1 exerted the opposite effects (Fig. 2f). To explore the potential roles of ecircCUX1 in FAO, NB cells were treated with oleic acid under fatty acid-restricted conditions. There was elevated OCR, mitochondrial membrane potential, complex I activity, NAD⁺/NADH ratio, and ATP production in SH-SY5Y and BE(2)-C cells stably transfected with ecircCUX1 or p113, but not with ecircCUX1 Mut (Fig. 2g, h and Additional file 1: Fig. S2e, f). Meanwhile, silencing of ecircCUX1 abolished the increase of OCR, mitochondrial membrane potential, complex I activity, NAD⁺/NADH ratio, and ATP production induced by exogenous oleic acid (Fig. 2g, h and Additional file 1: Fig. S2e, f). However, ectopic expression of ecircCUX1 or p113 did not affect mitochondrial mass in NB cells, with intact two-layer membrane and inner structures (Additional file 1: Fig. S3a-c). These results suggested that p113 was upregulated and facilitated fatty acid oxidation and mitochondrial activity in NB.

We further explored the biological impacts of ecircCUX1 on growth and aggressiveness of NB. Stable over-expression of ecircCUX1 or p113, but not of ecircCUX1 Mut, facilitated the growth and invasion capability of SH-SY5Y and SK-N-SH cells (Additional file 1: Fig. S4a-d). Meanwhile, the growth and invasion of BE(2)-C and IMR32 cells were significantly reduced by knockdown of ecircCUX1 (Additional file 1: Fig. S4a-d). Consistently, stable transfection of ecircCUX1 or sh-ecircCUX1 #2 into NB cells resulted in a significant increase or decrease of growth, tumor weight, long-chain fatty acid levels, complex I activity, NAD⁺/NADH ratio, ATP production, Ki-67 proliferation index, and CD31-positive microvessels of subcutaneous xenografts in athymic nude mice (Additional file 1: Fig. S4e, S5a, S5b). In experimental metastasis assay, athymic nude mice treated with tail vein injection of SH-SY5Y or BE(2)-C cells stably transfected with ecircCUX1 or sh-ecircCUX1 #2 displayed more or fewer lung metastatic colonies, with lower or greater survival probability, respectively (Additional file 1: Fig. S5c, d). These results suggested that ecircCUX1 promoted the growth and aggressiveness of NB cells via encoding p113.

To characterize downstream targets of p113, RNA-seq assay was performed, which indicated 460 upregulated and 1695 downregulated genes in SH-SY5Y cells upon ecircCUX1 over-expression (Fig. 3a). Co-IP and subsequent mass spectrometry assays revealed 761 and 847 protein pulled down by p113 or Flag antibody in SH-SY5Y cells stably transfected with 3Flag-tagged p113 (Additional file 1: Table S6 and Table S7), and overlapping analysis with transcription factors and epigenetic factors derived from ChIP-X program [27] and EpiFactors database [28] identified seven potential p113-interacting proteins (Fig. 3b). Among them, ZRF1 was the only transcription factor, while BRD4 might serve as a transcriptional co-factor. Endogenous interaction of p113 with ZRF1 and BRD4 was observed in NB cells, which was respectively increased or decreased by ectopic expression or knockdown of ecircCUX1, but not affected by transfection of ecircCUX1 with ORF mutation (Fig. 3c and Additional file 1: Fig. S6a). Two-step immunoprecipitation assay revealed respective binding of p113 to ZRF1 or BRD4 (Fig. 3d). Especially, SANT domain (450-621 aa) of ZRF1 as well as bromodomain 2 (BD2) domain (250-470 aa) of BRD4 were essential for their interaction with p113 (Additional file 1: Fig. S6b-d). In BiFC assay [29], p113 physically interacted with ZRF1 and BRD4, whereas no interaction was noted between ZRF1 and BRD4 (Fig. 3e). Then, CRISPR interference (CRISPRi) approach [30] was applied for knockdown of ZRF1 or BRD4 (Fig. 3f, g). Ectopic expression of ecircCUX1 increased the interaction of p113 with ZRF1 or BRD4, which was individually abolished by silencing of ZRF1 or BRD4, respectively (Fig. 3h), indicating the mediating roles of p113 in forming trimer complex with ZRF1 and BRD4 (Fig. 3i). Notably, ectopic expression or knockdown of ecircCUX1 facilitated and attenuated the transactivation of ZRF1 in NB cells, respectively (Fig. 3j). Meanwhile, ectopic expression of ecircCUX1 with ORF mutation did not affect transcriptional activity of ZRF1 (Fig. 3j). These results indicated that p113 physically interacted with ZRF1 and BRD4 as a trimer complex in NB cells.

To further investigate target genes of p113/ZRF1/BRD4 complex, ChIP-seq assay was performed using ZRF1- or BRD4-specific antibody, while results were overlappingly analyzed with ecircCUX-regulated genes derived from RNA-seq. Comprehensive analysis revealed 46 target genes regulated by transcriptional trimer complex p113/ZRF1/BRD4, especially those involved in metabolic pathway or complex I biogenesis, including ALDH3A1, NDUFA1, and NDUFAF5 (Fig. 4a, b). Knockdown of ZRF1 abolished the increased enrichment of p113/ZRF1/BRD4 complex on promoters of these target genes induced by over-expression of ecircCUX1, while silencing of BRD4 did not affect the binding of p113 or ZRF1 to target gene promoters (Additional file 1: Fig. S7a-c). Ectopic expression of ecircCUX resulted in increase of promoter activity, transcript and protein levels of ALDH3A1, NDUFA1, and NDUFAF5 in SH-SY5Y and SK-N-SH cells, which were eliminated after knockdown of ZRF1 or BRD4, respectively (Additional file 1: Fig. S7a-c and Fig. 4c). Consistent with the roles of ALDH3A1 in converting lipid peroxidation product (4-HNE) into fatty acids [31] (Fig. 4d), immunostaining assay indicated upregulation of ALDH3A1 and downregulation of 4-HNE in NB cells stably over-expressing ecircCUX, which was rescued by silencing of ZRF1 or BRD4 (Additional file 1: Fig. S8a). Notably, silencing of ZRF1 or BRD4 abolished the increase of long-chain fatty acid production, OCR levels, mitochondrial membrane potential, complex I activity, NAD⁺/NADH ratio, and ATP production in SH-SY5Y cells stably over-expressing ecircCUX1 (Fig. 4e, f and Additional file 1: Fig. S8b). Moreover, ectopic expression of ZRF1 increased the expression of ALDH3A1, NDUFA1, and NDUFAF5, and led to increase in OCR levels, mitochondrial membrane potential, complex I activity, NAD⁺/NADH ratio, ATP production, growth, and invasion of BE(2)-C cells, while knockdown of ALDH3A1, NDUFA1, or NDUFAF5 abolished these effects (Additional file 1: Fig. S9a-f). Taken together, these results suggested that p113/ZRF1/BRD4 complex promoted lipid metabolic reprogramming and mitochondrial complex I activity in NB cells.

We further explored the cooperative roles of p113 and ZRF1 in NB progression. Knockdown of ZRF1 by CRISPRi significantly prevented the increase of anchorage-independent growth and invasion of NB cells induced by stable ecircCUX1 over-expression (Fig. 5a, b). There were increased growth, volume, weight, Ki-67 proliferation index, and CD31-positive intratmoral microvessels of subcutaneous xenografts in athymic nude mice formed by injection of SH-SY5Y cells stably transfected with ecircCUX1, and these effects were eliminated after knockdown of ZRF1 (Fig. 5c, d). Athymic nude mice treated with tail vein injection of SH-SY5Y cells stably transfected with ecircCUX1 presented more lung metastatic colonies and lower survival probability, which was abolished by silencing of ZRF1 (Fig. 5e, f). These results suggested that p113 facilitated the tumorigenesis and aggressiveness of NB cells via interacting with ZRF1.

Based on the importance of SANT domain of ZRF1 in interacting with p113, we designed cell-penetrating peptides using Peptiderive server [32], named as ZRF1 inhibitory peptide with 12 amino acids in length (ZIP-12) (Fig. 6a). Incubation of NB cells with ZIP-12, but not with mutant control peptide, led to obvious aggregation within nucleus (Fig. 6b). ZIP-12 was able to bind to endogenous p113 protein (Additional file 1: Fig. S10a) and abolish the interaction between p113 and ZRF1 in NB cells, without impact on that of p113 and BRD4 (Fig. 6c). In addition, treatment with ZIP-12, but not of mutant control peptide, inhibited the ZRF1 enrichment, promoter activity, and expression of ALDH3A1, NDUFA1, and NDUFAF5 in NB cells (Additional file 1: Fig. S10b-e). As a result, there was reduced long-chain fatty acid levels, mitochondrial membrane potential, complex I activity, NAD⁺/NADH ratio, ATP production in ZIP-12-treated BE(2)-C cells (Additional file 1: Fig. S10f and Fig. 6d). Administration of ZIP-12 suppressed the viability of BE(2)-C and IMR32 cells in a dose- and time-dependent manner, without impact on that of non-transformed normal MCF 10A cells (Additional file 1: Fig. S11a, b). Moreover, ZIP-12 repressed anchorage-independent growth and invasion of BE(2)-C and IMR32 cells (Additional file 1: Fig. S11c). Intravenous administration of ZIP-12 suppressed the fluorescence signals, volume, weight, Ki-67 proliferation index, and CD31-positive microvessels of subcutaneous xenograft tumors formed by BE(2)-C cells in nude mice (Fig. 6e and Additional file 1: Fig. S11d). Moreover, intravenous administration of ZIP-12 led to less lung metastatic colonies and prolonged the survival time of athymic nude mice treated with tail vein injection of BE(2)-C cells (Fig. 6f and Additional file 1: Fig. S11e). Collectively, these results demonstrated that blocking p113-ZRF1 interaction suppressed NB progression.

In 498 well-defined NB cases (GSE62564), Kaplan–Meier survival plots revealed that higher expression of p113 host gene CUX1 (P = 6.6 × 10^− 3), ZRF1 (P = 3.4 × 10^− 16), BRD4 (P = 9.3 × 10^− 4), ALDH3A1 (P = 5.7 × 10^− 5), NDUFA1 (P = 2.5 × 10^− 14), or NDUFAF5 (P = 3.0 × 10^− 3) was associated with lower overall survival probability of patients (Fig. 7a). High expression of ZRF1 or BRD4 was also associated with poor survival of patients with breast cancer, colon cancer, Ewing sarcoma, glioma, melanoma, myeloma, or renal cancer (Additional file 1: Fig. S12). There was a positive expression correlation between ZRF1 and ALDH3A1 (R = 0.131, P = 3.4 × 10^− 3), NDUFA1 (R = 0.334, P = 2.0 × 10^− 14), or NDUFAF5 (R = 0.315, P = 6.6 × 10^− 13) in 498 NB specimens (Fig. 7b). These results indicated that expression of CUX1, ZRF1, BRD4, or target genes was associated with poor outcome of tumors.