Subcellular Distribution of p53 by the p53-Responsive lncRNA NBAT1 Determines Chemotherapeutic Response in Neuroblastoma

Neuroblastoma (NB), a tumor of the peripheral nervous system is the most common extra cranial tumor of childhood. It is the most frequently diagnosed neoplasm during infancy and accounts for around 12% to 15% of pediatric cancer related mortality (1, 2). The appearance of NB tumors is mainly due to the improper differentiation of neural crest cells into mature neurons of the sympathetic nervous system, and the tumors often arise from the sympathetic nerve ganglia in abdomen, mostly from adrenal gland. NB is highly heterogeneous and therapeutically challenging (3), where patients can be broadly categorized into low-, intermediate-, and high-risk subgroups depending on risk factors such as age of diagnosis, disease stages, and nonrandom chromosomal aberrations (4, 5). High-risk tumors have a high chance of recurrence even after intensive multimodal treatment, including chemotherapy, surgery, radiotherapy, and immunotherapy, thus necessitating the need for more therapeutic opportunities.

High-risk NB tumors are highly aggressive and harbor a variety of nonrandom chromosomal alterations such as MYCN oncogene amplification, 1p and 11q deletion, and 17q gain (6, 7). In addition to the nonrandom chromosomal alterations, a significant fraction of NBs possess recurrent somatic mutations in the RAS–MAPK pathway genes, including tyrosine kinase domain of anaplastic lymphoma kinase (ALK; refs. 8, 9), PTPN11 (10), chromatin remodeling genes (11, 12), and neuritogenesis-specific genes (13, 14). Recently, telomerase activation has been shown to play a critical role in defining high-risk NBs, which is achieved by recurrent genomic rearrangements at 5p15.33, proximal to the TERT gene (7, 15). The majority of familial NBs is characterized by heritable mutations in ALK or PHOX2B genes (16, 17), which are highly penetrant. Recently, alternative personalized therapies are being considered for NBs with recurrent somatic and germ line mutations.

Interestingly, the p53 tumor suppressor gene, which is mutated in approximately 60% of all human cancers, is rarely mutated in NB tumors (18–20). It is a well-studied transcription factor and suppresses malignant transformation by promoting DNA damage repair, cell-cycle arrest, and apoptosis in response to cellular stress signals. Of note, functional inactivation of p53 can also occur through posttranslational mechanisms, for example the loss of nuclear localization leading to cytoplasmic accumulation of p53 (21–23) and understanding the mechanisms underlying this phenomenon may allow the development of alternative therapeutic strategies targeting NB.

Long noncoding RNAs (lncRNA) have been shown to act as regulators of development and progression of various diseases such as cancer. LncRNAs, by directly interacting with proteins, can regulate gene expression at the transcriptional and posttranscriptional level (24). In cancer therapy, developing resistance to anticancer drugs is a major challenge, which leads to relapse and even mortality. Chemoresistance is a complex phenomenon, which is caused by a multilayered transcriptionally and posttranscriptionally controlled molecular mechanisms. Recent evidence suggests that lncRNAs may also play an important role in this complex phenomenon. For example, HOTAIR lncRNA sensitizes imatinib resistance through regulating multidrug resistance-associated protein 1 (MRP1; ref. 25). XIST, X-inactive-specific transcript, accounts for doxorubicin resistance in colorectal cancer via modulating the miR-124/SGK1 axis (26). Thus, targeting lncRNAs constitute an important therapeutic option for hard-to-treat cancers.

Genome wide association studies (GWAS) have identified the 6p22.3 locus as NB hotspot locus with cluster of SNPs associated with increased risk of NB development and disease aggressiveness (27, 28). The 6p22.3 locus encoded lncRNA NBAT1 regulates NB development and progression via its tumor suppressor functions. Low NBAT1 expression predicts poor prognosis and, more importantly, correlates with increased cell proliferation (29). On the basis of these properties, we proposed that the lower expression of NBAT1 in high-risk patients could contribute to therapeutic resistance. Here we demonstrate that NBAT1 is a p53 responsive lncRNA and that its low expression is associated with decreased sensitivity to clinically relevant genotoxic drugs because of altered p53 nuclear/cytoplasmic/mitochondrial levels and a change in the stability of the nuclear export protein CRM1. Interestingly, blocking the CRM1 transport machinery and increasing the p53 stability via MDM2 inhibition recovers drug sensitivity in NB cells.

All NB cell lines were maintained at 37°C with 5% carbon dioxide. IMR-32 cell line was maintained in EMEM supplemented with 10% FBS, antibiotics, 1% glutamax, 1% sodium pyruvate, and 1% nonessential amino acids. SH-SY5Y (catalog no.: 300154; CLS Cell Lines Service GmbH), IMR-32 (catalog no.: 300148; CLS Cell Lines Service GmbH), SK-N-BE(2) (catalog no.: ACC 632; DSMZ- German Collection of Microorganisms and Cell Cultures GmbH); and SK-N-F1 (94092304; Merck). SK-N-AS (ATCC; CRL-2137), SK-N-DZ (ATCC; CRL-2149), NB69, Kelly (ECACC; HPA Culture Collections) and SHEP (CRL-2269; ATCC) were obtained from Prof. Tommy Martinsson, University of Gothenburg, and they were authenticated using AmpFLSTR Identifiler Kit (Thermo Fisher Scientific; ref. 30). All cell lines were maintained in DMEM supplemented with 10% FBS and antibiotics. The cell lines were tested for Mycoplasma on regular intervals of 4 to 6 weeks using Mycoalert (Lonza LT07–418), and all the cell lines were used for five to seven passages upon thawing.

NB tissues from patients was acquired during surgery and stored in −80°C. Ethical approval was obtained from the Karolinska University Hospital Research Ethics Committee (Approval no. 2009/1369–31/1 and 03–736). Written informed consent for using tumor samples in scientific research was provided by parents/guardians. In accordance with the approval from the Ethics Committee the informed consent was either written or verbal, and when verbal or written assent was not obtained, the decision was documented in the medical record.

To induce genotoxic stress, NB cell lines were treated with various concentrations of the genotoxic drugs (as indicated in the figures)—nutlin-3a (SML0580; Merck), doxorubicin (D1515; Merck), etoposide (E1383; Merck), cisplatin (C2210000; Merck), and JNJ-26854165 (S1172; Selleckchem) and also treated with respective control (adding DMSO to media for nutlin-3a, etoposide, and JNJ treatment, adding PBS to media for cisplatin and adding media only for DOX treatment in a corresponding volumes not exceeding 1%). We harvested cells 24 hours after drug treatment for further analysis. SH-SY5Y treated with leptomycin B (LMB; Merck L2913) at a concentration of 5 nmol/L (2.5 ng/mL) for 4 hours followed by nutlin-3a treatment in two different doses 2.5 and 5 μmol/L and then harvested after 20 hours for immunostaining. SH-SY5Y and IMR-32 cells were also treated with selinexor (S7252, Selleckchem) alone at 50 nmol/L concentration or in combination with nutlin-3a (doses indicated in figure) for 24 hours and then harvested for further analysis. To perform drug synergy assessment, NB cells were plated in 96-well plate and then treated with single or combinations of compounds and subsequently analyzed for cell viability 48 hours posttreatment. To check the synergy between the drugs, Chou–Talalay combination index was calculated using CompuSyn software (31).

Both the siRNA and shRNAs were designed in the third exon of NBAT1 to knockdown NBAT1 transcript. NBAT1 and p53 siRNA (Merck) transfection was performed using Lipofectamine RNAiMAX reagent (Thermo Fisher Scientific) following the protocol mentioned previously (29). Mission lentiviral transduction particles expressing nontarget shRNA control or shRNA against NBAT1 (Lot 12271212 MN) were obtained from Merck and the stable clones was generated in SH-SY5Y and IMR-32 following the protocol described previously (29).

NBAT1 transcript overexpressed stable clones were generated in pcDNA3.1 vector in SH-SY5Y and IMR-32 cells according to protocol described earlier (29) and used for drug treatments followed by growth assay. SH-SY5Y cells were transfected with p53 WT plasmid (Addgene, 69003) and GFP tagged p53 plasmid (Addgene, 12091) and the p53 expression status was measured by qRT-PCR for p53 WT plasmid and by fluorescence microscopy for GFP tagged p53.

ChIP was performed according to the protocol described previously (32). SH-SY5Y cells were treated with nutlin-3a (concentrations are indicated in Figures) for 24 hours and the chromatin was immunoprecipitated using anti-p53 antibody (Cell Signaling Technology, 2524) and was verified by qPCR. The details of primers used for qPCR are provided as Supplementary Table S1. The processed BIGWIG files of p53 ChIP-seq data for U20S and keratinocyte were downloaded from Cistrome public database (33) and p53 enrichment over the NBAT1 promoter was compared in nutlin-3a and cisplatin-treated U2OS and keratinocyte cells, respectively, and visualized using R package Gviz.

Nuclear and cytoplasmic fractionations of cells were performed using NE-PER Nuclear and Cytoplasmic Extraction Reagents Kit (Thermo Fisher Scientific). Protein extraction from nuclear and cytoplasmic fractions and immunoblotting were done on the same way as described below. Mitochondrial fractionation was performed according to the method described previously (34). Mitochondria from 10 million nutlin-3a-treated SH-SY5Y cells were isolated by differential centrifugation using 1 mL of isolation buffer (320 mM sucrose, 10 mmol/L Tris-HCl, pH 7.4 and 1 mmol/L EDTA), supplemented freshly with protease inhibitor (Halt Protease Inhibitor Cocktail; Thermo Fisher Scientific) and 0.1% BSA (Thermo Fisher Scientific). Cells were homogenized using a Potter–Elvehjem homogenizer on ice (35 strokes). Nuclei and cell debris were pelleted twice at 1,000 × g for 10 minutes at 4°C. Resultant supernatant was centrifuged at 14,000 × g for 15 minutes at 4°C. Mitochondrial pellet was then washed once in isolation buffer, suspended in 400 μL of isolation buffer and centrifuged at 14,000 × g for 15 minutes at 4°C in a new tube. Double washed mitochondrial fraction was then lysed in 50 μL of RIPA buffer and stored at −80°C. Mitochondrial fraction was suspended in 1x NuPAGE LDS sample buffer. Mitochondrial proteins were separated by SDS-PAGE (4%–12% Bis-Tris Protein Gels; Invitrogen). Tim23 protein was used as a mitochondrial marker and detected by Anti-Tim23 mouse antibody (BD Transduction Laboratories, Clone 32/Tim23) in 1:1,000 dilution.

Immunofluorescence was performed after drug treatments to the NB cells, following the protocol as described previously (29). The nuclei were stained with DAPI and localization of p53 was detected using anti-p53 antibody (Merck; 05–224). Expression of CRM1 was visualized by using anti-CRM1 antibody (Cell Signaling Technology; 46249S). Paraffin-embedded xenografts derived from IMR-32 cells were analyzed by hematoxylin and eosin (H&E) and staining was performed with the anti- p53 (Merck; 05–224) and anti-Ki67 antibodies (Abcam; ab16667).

Animal studies were performed in accordance to our Ethical permit (5.8.18–02708/2017), approved by the Animal Ethical Review Board, Sweden. Eight-week-old NSG mice were maintained following institute guidelines and subcutaneously injected with 5 × 10⁶ IMR-32 cells in 25% matrigel/PBS. Mice were divided into four treatment groups vehicle, nutlin-3a, Selinexor, and the combination of nutlin-3a and selinexor, N = 5 in each group with equal ratio of male to female. Treatment with inhibitor was started when the xenografted tumor size reached 100 to 150 mm³ of size. Inhibitors were administered by oral gavage once on every alternative day and for total of six doses. Nutlin-3a (S8059; Selleckchem), dissolved in 2% Klucel and 0.2% Tween 80, was administered at 50 mg per kg body weight and selinexor was administered at 10 mg/kg body weight dissolved in 0.6% Pluronic F-68. Mice were followed up to 12 days following the first dose of treatment and tumors size and mouse weight were measured on every third day throughout the treatment period. Animals were sacrificed following institute protocol and tumors were dissected out to take picture and were fixed in 4% formaldehyde in PBS for IHC.

All the raw and processed RNA-seq data generated in this manuscript is available through GEO accession number: GSE137366.

Considering that NBAT1 is a tumor suppressor lncRNA and its depletion in NB cell lines leads to increased cell growth in vitro and in vivo (29), we sought to investigate the impact of NBAT1 depletion on the genotoxic drug sensitivity of NB cell lines with different genetic alterations. Towards this, we have generated stable NBAT1 knockdown cells (NBAT1 sh) along with cells transduced with nontarget control shRNA (Csh), in both SH-SY5Y (non-MYCN amplified) and IMR-32 (MYCN amplified) NB cells (Supplementary Fig. S1A). NBAT1 sh and Csh SH-SY5Y cells were treated with various genotoxic drugs like doxorubicin, etoposide, and cisplatin. Following the drug treatments, cell viability was measured, which indicated that NBAT1 sh cells were resistant to increasing doses of genotoxic drugs compared with the control cells (Fig. 1A). In addition, we have also treated NBAT1 sh and Csh cells (SH-SY5Y and IMR-32) with two MDM2 inhibitors nutlin-3a and JNJ26854165 (Serdemetan), which stabilize the p53 protein. Similar to genotoxic drugs, NBAT1 sh cells showed resistance to increasing doses of the MDM2 inhibitors (Fig. 1B and C; Supplementary Fig. S1B). Conversely, NBAT1 overexpression sensitized SH-SY5Y and IMR-32 cells to different genotoxic drugs and MDM2 inhibitors (Fig. 1D; Supplementary Figs. S1A and S1C). We have analyzed NBAT1 expression in different NB cell lines (which were either p53 wild type or p53 mutant NB cell lines; Supplementary Fig. S1D) and measured their sensitivity to the MDM2 inhibitor nutlin-3a. NB cell lines (SH-SY5Y, IMR-32, NB69, SK-N-DZ) with higher NBAT1 expression levels showed higher sensitivity to nutlin-3a compared with NB cell lines with lower NBAT1 levels [SK-N-BE(2), SK-N-AS, SHEP; Fig. 1E and F]. These observations further corroborate functional correlation between NBAT1 expression levels and sensitivity to different genotoxic drugs. Interestingly, unlike in wildtype p53 NB cell lines (SH-SY5Y and IMR-32), NBAT1 overexpression in p53 mutant NB cell lines [SK-N-AS and SK-N-BE(2)] with lower NBAT1 endogenous levels did not show any change in sensitivity to genotoxic drug or MDM2 inhibitor, suggesting that both wildtype p53 and NBAT1 levels determine the sensitivity to the drug treatments (Supplementary Figs. S1E and S1F). In addition, IMR-32 cells were more resistant to the drug treatments compared with SH-SY5Y cells, and we found that IMR-32 has 30% less total NBAT1 than SH-SY5Y. We have investigated if the subcellular localization of NBAT1 differs in these two cell types, which may also contribute to the observed differential sensitivity to the drugs. The proportion of NBAT1 in the cytoplasmic fraction was higher in IMR-32 when compared with SH-SY5Y cells (Supplementary Fig. S1G). Thus, the relatively lower level of total NBAT1 together with higher NBAT1 cytoplasmic localization in IMR-32 cells may contribute to the increased drug resistance in IMR-32 cells.

As NBAT1 is known to act as a tumor suppressor lncRNA in multiple cancers (35–37) and, of note, the NBAT1 levels in the NB cell lines is associated with the sensitivity to genotoxic drugs, we were interested to know how NBAT1 expression is regulated during genotoxic stress. We found an increased expression of NBAT1 in SH-SY5Y cells treated with various concentrations of genotoxic drugs (Fig. 2A). To check if NBAT1 activation during genotoxic stress is functionally connected to p53, we have performed RNA-seq on nutlin-3a-treated SH-SY5Y cells and the data suggest that NBAT1 was upregulated in nutlin-3a-treated SH-SY5Y cells compared with the untreated cells in a manner similar to known p53 target genes like p21 and MDM2 (Fig. 2B; Supplementary Table S2). We verified the upregulation of NBAT1 and the other p53 direct target genes in nutlin-3a-treated SH-SY5Y cells by RT-qPCR (Supplementary Figs. S2A and S2B). Interestingly, nutlin-3a treatment of mutant p53 cell lines SK-N-AS and SK-N-BE(2) failed to upregulate NBAT1 expression along with known p53 targets. Moreover, p21 is marginally upregulated (two-fold) in SK-N-AS cells, and whereas in SH-SY5Y cells, it is upregulated by several thousand-fold (Fig. 2C; Supplementary Fig. S2B), suggesting that NBAT1 expression could be under the direct regulation of p53.

To understand the functional connection between NBAT1 and p53, we carried out loss- and gain-of-function experiments of p53 and investigated NBAT1 expression. Knockdown of p53 in SH-SY5Y cells resulted in the reduced expression of NBAT1 (Fig. 2D), whereas p53 overexpression induced NBAT1 expression (Fig. 2E). To check if p53 can regulate NBAT1 expression by binding directly to NBAT1 promoter, we performed a ChIP experiment in nutlin-3a-treated SH-SY5Y cells and observed p53 enrichment at the NBAT1 promoter as well as at the p21 and MDM2 gene promoters (Fig. 2F). To further support the ChIP-qPCR data, we analyzed publicly available ChIP-seq datasets from nutlin-3a-treated human U2OS cells and cisplatin-treated human keratinocytes. These analyses revealed the presence of p53 peaks at the NBAT1 promoter, and more importantly, an increased enrichment of p53 was found at the NBAT1 promoter following nutlin-3a and cisplatin treatments (Supplementary Fig. S2C). To identify the sequences that are required for p53 recruitment to the NBAT1 promoter, we scanned for potential p53 motifs in the p53 ChIP-seq peaks at the NBAT1 promoter using JASPAR (38) and identified two different p53 binding motifs matching consensus p53 motif. One of the p53 motifs is located in the first intron of the NBAT1 (intronic) and the other is located upstream of the NBAT1 transcription start site (upstream; Fig. 2G). The upstream p53 motif showed higher promoter activity than the intronic in luciferase assays, and the promoter activity was further enhanced when the cells were treated with nutlin-3a, suggesting the upstream p53 motif drives p53-dependent transcription. Mutation of critical residues or the deletion of the upstream p53 motif, compromised the luciferase activity in nutlin-3a-treated and -untreated cells (Fig. 2G). Taken together, these observations demonstrate NBAT1 as a bona fide p53-regulated lncRNA.

Previously it has been shown that an intrinsic chemoresistance in tumor cells is in part contributed by the deregulation of p53 controlled pathways (39). Therefore, we sought to investigate if the increasing resistance to genotoxic drugs and MDM2 inhibitors in the NBAT1-depleted cells could be due to compromised p53-mediated target gene activation. For this purpose, control (Csh) and stable NBAT1-depleted (NBAT1 sh) SH-SY5Y cells were treated with nutlin-3a and transcriptomic profiling was performed using RNA-seq. Gene set enrichment analysis (GSEA) of the nutlin-3a-treated control and NBAT1-depleted transcriptomes revealed the enrichment of genes, which are known to be regulated by p53 (Fig. 3A; Supplementary Table S3). We observed that the expression of p53 pathway genes including p21, MDM2, GADD45A was compromised in the NBAT1-depleted cells (Fig. 3B). We validated the deregulation of p53 target genes by RT-qPCR in the NBAT1-depleted cells following nutlin-3a treatment (Fig. 3C).

To decipher if the observed effect of NBAT1 depletion on the p53 target genes was due to perturbation in the p53-mediated transcriptional activation, we performed p53 ChIP-qPCR and found that the p53 binding was affected at the promoters of the p53-dependent genes in the NBAT1-depleted cells (Fig. 3D). Consistent with this, we observed a decrease in the p53-dependent promoter activity in the NBAT1-depleted cells in a luciferase-based assay (Supplementary Figs. S3A and S3B).

p53 mutations are very prevalent in many cancers but are rare in pediatric malignancies, including NBs (40). It has been proposed that, in NB, altered p53 subcellular distribution may lead to the inactivation of p53 nuclear functions (21). We have also tested the subcellular distribution of p53 in 21 NBs, from a Swedish cohort, comprising low- and high-risk patients, by IHC. We found that a majority of high-risk NBs with lower NBAT1 expression had both nuclear and cytoplasmic p53 distribution, whereas patients with low-risk NB with higher NBAT1 expression showed preferential nuclear accumulation (Fig. 4A; Supplementary Table S4). On the basis of these results, we hypothesized that the loss of p53-dependent target gene expression and the resistance to genotoxic drugs of the NBAT1-depleted cells could be due to altered p53 subcellular distribution. We first tested the cellular distribution of wild-type p53 in NB cell lines, SH-SY5Y and IMR-32, which show differential NBAT1 expression and subcellular distribution, with the former showing higher NBAT1 expression and preferential nuclear distribution compared with the latter with lower NBAT1 expression and more cytoplasmic distribution (Fig. 1E; Supplementary Fig. S1G). In SH-SY5Y cells, p53 is enriched in the nuclear fraction, whereas in IMR-32 cells, p53 shows a more cytoplasmic distribution (Fig. 4B). In addition, we analyzed the p53 distribution in a mutant p53 cell line, SK-N-AS, with low NBAT1 expression and found that it had a more cytoplasmic p53 distribution (Fig. 4B). These indicate that NBAT1 expression levels may regulate the subcellular distribution of p53 in NB cells. To explore this further, we performed p53 immunostaining on control and NBAT1-depleted SH-SY5Y cells with or without nutlin-3a treatment and found that the loss of NBAT1 expression affects the nuclear enrichment of p53 in nutlin-3a-treated cells (Fig. 4C and D). These findings were further verified with siRNA-mediated knockdown of NBAT1, which also showed enhanced cytoplasmic accumulation of p53 upon nutlin-3a treatment (Fig. 4E–G). By using cell fractionation experiments, we further confirmed the failure of p53 accumulation to the nuclear compartment in the NBAT1-depleted cells following nutlin-3a treatment (Fig. 4H). During stress conditions, cytoplasmic p53 has been shown to enter into mitochondria to induce apoptosis (41). We wanted to check whether cytoplasmic p53 in NBAT1-depleted cells can enter into mitochondria. Interestingly, mitochondrial p53 levels were drastically decreased in NBAT1-depleted cells compared with the control cells (Fig. 4H).

In addition, the nuclear localization of ectopically expressed GFP tagged p53 (GFP-p53) was also compromised in the NBAT1-depleted SH-SY5Y cells, however in these cells, GFP-p53 nuclear localization was restored upon NBAT1 overexpression (Fig. 4I; Supplementary Fig. S4A). Accumulation of endogenous p53 levels in the nuclear compartment of the NBAT1-depleted cells was also enhanced by NBAT1 overexpression, further suggesting that NBAT1 levels determines the nuclear levels of p53 (Supplementary Fig. S4B). Even in the case of IMR-32 cells, where p53 shows preferentially cytoplasmic distribution, nuclear p53 level was restored following NBAT1 overexpression (Fig. 4J). We also tested GFP-p53 localization in another NB cell line SK-N-AS, with low NBAT1 expression. In these cells, GFP-p53 showed cytoplasmic-nuclear localization but NBAT1 overexpression leads to preferential nuclear localization of GFP-p53 (Supplementary Fig. S4C). Taken together these observations suggest that NBAT1 levels influence the cellular localization of p53.

CRM1 or Exportin1, a well-studied nuclear exporter, has previously been implicated in the spatial regulation of p53 primarily from the nuclear to the cytoplasmic compartment (42). We therefore sought to evaluate if the altered subcellular distribution of p53 in the NBAT1-depleted cells was related to CRM1 function. To this end, we treated NBAT1 knockdown cells with CRM1 inhibitor LMB alone or in combination with nutlin-3a and found that LMB or the combination treatment led to p53 accumulation in the nuclear compartment of NBAT1-depleted cells (Fig. 5A and B). Treatment of NB cells with genotoxic drug etoposide or MDM2 inhibitor nutlin-3a led to decreased levels of CRM1 protein but a higher level of CRM1 was still maintained in the NBAT1-depleted cells (Fig. 5C), suggesting that the failure of nuclear accumulation of p53 in the NBAT1-depleted cells could be because of the higher CRM1 protein levels. We checked the CRM1 status in different NB cell lines (both wild type and mutated for p53), and found varied level of CRM1 in the analyzed cell lines, with no obvious relation between p53 mutational status and CRM1 level. Nevertheless, the CRM1 level was higher in IMR-32 cells compared with SH-SY5Y, where the former showed more cytoplasmic p53 (Figs. 5D and 4B). To further check the relation between NBAT1 and CRM1 level, we modulated the NBAT1 level by overexpression or knockdown. NBAT1 overexpression led to decreased CRM1 level whereas its knockdown caused higher CRM1 level in multiple NB cell lines (Supplementary Figs. S5A and S5B). To test if the nuclear accumulation of p53 upon LMB treatment in the NBAT1-depleted cells was due to CRM1 inhibition, CRM1 was downregulated in the NBAT1-depleted cells using siRNA. Consistent with LMB inhibition, siRNA-mediated knockdown of CRM1 in the NBAT1-depleted cells also led to the nuclear accumulation of p53 (Supplementary Fig. S5C). Consistent with NB cell line data, NB tumors showed higher CRM1 level when NBAT1 expression was low (Supplementary Figs. S5D and S5E). Using proximity ligation assay (PLA) on NBAT1 sh and Csh SH-SY5Y cells, we observed that p53 and CRM1 interaction was higher in the NBAT1 sh cells when compared with the Csh cells. p53 and CRM1 interaction was completely abolished in the Csh cells after treatment with nutlin-3a but considerable CRM1 and p53 interaction was still detected in the NBAT1-depleted cells even after nutlin-3a treatment (Fig. 5E). Taken together, these observations suggest that higher CRM1 level in the NBAT1-depleted cells may lead to p53 cytoplasmic accumulation during genotoxic stress.

We next aimed to explore the possible cause of the higher level of CRM1 protein in the NBAT1-depleted cells. We have recently shown that NBAT1 regulates neuronal differentiation of NB cells by interacting with USP36, a de-ubiquitinase (DUB) enzyme (43). We reasoned that higher CRM1 level in the NBAT1-depleted cells could be because of altered USP36 function. Interestingly, we found CRM1 as an interacting partner of the USP36 in a published dataset where USP36 interacting proteins were identified by co-immunoprecipitation (Co-IP) followed by mass spectroscopy (44). Using PLA and Co-IP experiments, we validated CRM1 interaction with USP36 in SH-SY5Y cells (Fig. 5F and G). To determine if USP36 regulates the CRM1 level, we analyzed CRM1 levels following the loss- and gain-of-function of USP36: USP36 loss caused decrease in CRM1 levels whereas its overexpression led to higher CRM1 levels (Fig. 5H–J). We also observed that USP36 knockdown led to the higher ubiquitination of CRM1 protein in MG132-treated cells (Fig. 5K; Supplementary Fig. S5F), suggesting USP36 regulates CRM1 stability via maintaining its ubiquitination levels.

Our initial observations show that CRM1 inhibition leads to p53 nuclear accumulation in the NBAT1-depleted cells. These suggest that combining CRM1 inhibition along with chemotherapeutic drugs could reverse chemoresistance observed in NBAT1-depleted cells. The CRM1 inhibitor LMB, which binds to CRM1 in nonreversible manner, has high cellular toxicity and side effects (45). Currently, there are several less toxic CRM1 inhibitors available that can inhibit CRM1 function in a reversible manner through noncovalent interactions. One such reversible CRM1 inhibitor is selinexor (KPT-330), which is available as an oral drug (46). We opted to test the efficacy of selinexor in combination with nutlin-3a on the NBAT1-depleted cells to check if p53 nuclear accumulation can be rescued or not. We observed that like LMB, combining selinexor with nutlin-3a rescued the nuclear accumulation of p53 in the NBAT1-depleted cells as well as Csh cells (Fig. 6A and B). This combination treatment induced apoptotic marker cleaved PARP (cPARP) much more effectively whereas the level of anti-apoptotic makers (XIAP) was decreased in the control and NBAT1-depleted cells (Fig. 6C). Besides, the restoration of the p21 expression was also observed in the combination treatment (Fig. 6D). Moreover, the combination treatment was more effective in reducing the colony formation of NBAT1-depleted cells compared with the nutlin-3a treatment alone (Supplementary Fig. S6A).

As the combination of selinexor and nutlin-3a was very effective in the nuclear accumulation of p53 and also induces apoptosis in the NBAT1-depleted cells, we wanted to test if this could be a therapeutic approach for NB. We used IMR-32 cell line, where p53 is not mutated and preferentially localized in the cytoplasmic compartment, to test the efficacy of the drug combination (Fig. 4B). IMR-32 cells were treated individually or in combination with selinexor and nutlin-3a. We observed that the selinexor treatment rescued nuclear accumulation of p53, but when used in combination with nutlin-3a it was even more effective in promoting the nuclear accumulation of p53 (Figs. 6E and F). The combination treatment also enhanced apoptotic markers (cPARP, cCASP3) and decreased anti-apoptotic marker XIAP in IMR-32 cells (Fig. 6G). Interestingly, selinexor was not effective in restoring nuclear accumulation of mutant p53 in SK-N-AS cells, suggesting that cytoplasmic accumulation of mutant p53 in SK-N-AS cells could be CRM1 and NBAT1 independent (Supplementary Fig. S6B). Since selinexor and nutlin-3a combination treatment in IMR-32 cells leads to p53 nuclear accumulation, one can expect to see the enhancement of p53-mediated target gene expression. We found that, compared with the treatment with individual drugs, the combination treatment was more effective in inducing p53-mediated target gene expression, including NBAT1 (Fig. 6H and I). We further verified this using RNA-seq, where p53-mediated gene expression was more induced in the combination treatment compared with either selinexor or nutlin-3a alone (Fig. 6J). Furthermore, the p53 pathway was found to be enriched significantly in the GSEA of the RNA-seq data from the combination as well as selinexor treatments (Supplementary Fig. S6C). Consistent with CRM1 functions in NB cell lines, high CRM1 expression predicts poor overall survival in patients with NB, and also higher CRM1 expression was detected in MYCN amplified NBs (Supplementary Figs. S6D and S6E).

We wanted to test if the combination treatment of selinexor and nutlin-3a acts in a synergistic manner in inducing cell death. To this end, we treated IMR-32 and SH-SY5Y cells with the increasing concentrations of selinexor and nutlin-3a individually or in combination. We observed that selinexor and nutlin-3a combination synergistically induced cell death in these two cell types as indicated by the Chou–Talalay combination index (Supplementary Fig. S6F). The ED50 (50% effect) value for IMR-32 in the combination treatment was 1.16 μmol/L for nutlin-3a and 23.28 ng/mL for selinexor. We tested the efficacy of this prediction of the ED₅₀ value calculated using Chou–Talalay index in IMR-32 cells and found that the combination treatment was significantly more effective in inducing cell death than the treatment with individual drugs (Supplementary Fig. S6G). We also tested this drug combination on p53 mutant NB cell lines SK-N-BE(2) and SK-N-AS and we observed that both the individual or combination treatments were not effective in inducing the cell death in these cell types, suggesting that the efficacy of the drug combination is dependent on functional p53 (Supplementary Fig. S6F). We also tested selinexor treatment effects on NB cell viability in combination with genotoxic drug etoposide and found that the combination treatment with genotoxic drug was more effective than the individual treatments, suggesting that selinexor can also enhance the effect of genotoxic drug on p53 positive NB cells (Supplementary Fig. S6H).

We next subcutaneously engrafted IMR-32 cells into NSG mice and the xenografts were treated with either vehicle control or nutlin-3a and selinexor individually or in combination. We observed that the combination treatment was more effective in reducing tumor progression without any adverse effect on animal weight (Fig. 6K; Supplementary Fig. S6I). In IMR-32 derived xenografts, the combination treatment led to a decrease in Ki67 and an increase in p53 staining compared with the vehicle or individual drug treatments (Supplementary Fig. S6J). The effectiveness of selinexor and nutlin-3a combination in reducing the tumor growth of IMR-32 derived xenografts suggests that selinexor and nutlin-3a combination may constitute an effective treatment strategy for patients with high-risk NB with altered subcellular localized wild-type p53.

High-risk NBs are still a matter of concern with less than 50% long-term survival despite intensive multimodal therapy. Chemoresistance is one of the important factors in the relapsed NB tumors, which is in part contributed by the acquisition of p53 inactivating mutations and consequently the aberrant p53 function (39, 47). However, earlier studies using single-strand conformational polymorphism analyses suggested that p53 mutations are rare in NBs (40). Of note, whole genome sequencing of the 208 NBs revealed a p53 mutational rate of 1.9% (4 of 208 tumors), which is low in comparison to other cancer types such as ovarian cancer where the incidence is as high as 90% (15, 48). Functional inactivation of p53 may also occur by altered subcellular localization of the p53 protein. In undifferentiated patients with high-risk NB, preferential localization of wild type p53 in the cytoplasmic compartment with a concomitant loss in the nuclear compartment was observed (21). Supporting these observations, we found that p53 showed cytoplasmic/nuclear distribution in a majority of high-risk NB tumors with lower NBAT1 expression. However, low-risk tumors that showed preferential nuclear staining had higher NBAT1 expression levels. On the other hand, NB cell line-based investigations on p53 cellular localization have been controversial with some investigations supporting cytoplasmic accumulation (22, 23, 49, 50) whereas others support nuclear accumulation with compromised p53-dependent nuclear functions such as DNA damage induced G₁ cell-cycle arrest (51–54). Nevertheless, altered subcellular localization of p53 is not just restricted to NB but also observed in other cancer types such as breast cancer and glioblastoma (55, 56). Moreover, in a recent TCGA analysis it was discovered that p53 inactivating mutations were less prevalent in 7 cancer types (less than 5%; ref. 48). These observations indicate that in the absence of p53 inactivating mutations in these cancers, like in high-risk NBs, p53 cytoplasmic accumulation may constitute an alternate way of p53 inactivation.

In this investigation, we provide evidence that NBAT1 regulates the subcellular distribution of p53. In NBAT1-depleted cells, cytoplasmic accumulation of p53, loss of p53 levels in mitochondrial and nuclear compartments, accompanied by loss of p53-dependent gene expression may collectively contribute to resistance to drug treatments (Fig. 7). Moreover, NBAT1 overexpression in cells with low NBAT1 levels and p53 cytoplasmic accumulation was sufficient to restore p53 nuclear levels and sensitize the cells to drug treatments, indicating that NBAT1 plays a pivotal role in maintaining p53 functions through regulating its nuclear levels. Interestingly, p53 preferential cytoplasmic accumulation accompanied by its loss in the nuclear compartment of NBAT1-depleted cells parallels the situation in high-risk NB patients. In addition, high-risk patients show lower NBAT1 expression compared with low-risk patients. On the basis of these observations, we speculate that lower NBAT1 expression in high-risk patients could contribute to preferential accumulation of p53 in the cytoplasm and loss of p53-dependent nuclear functions.

Although preferential p53 cytoplasmic accumulation has been noted by several studies in NB, molecular mechanisms leading to cytoplasmic accumulation have not been investigated in greater detail. CRM1 or Exportin1 is one of the well investigated nuclear exporters and accounts for the nuclear to cytoplasmic export of the p53 protein along with several other cellular factors (57). CRM1 is also overexpressed in several cancer types (58, 59), including NB. CRM1 shows higher expression in MYCN amplified tumors and its overexpression correlates with poor overall survival. Our data demonstrates that NBAT1 controlled CRM1 achieves its oncogenic potential, in part, through modulating the levels of nuclear/cytoplasmic levels of p53. We observed that inhibiting CRM1 in a MYCN-amplified aggressive NB cell line, such as IMR-32, where p53 shows preferential cytoplasmic accumulation, restores nuclear p53 levels, and p53-dependent gene expression. The functional relationship that is seen between NBAT1 and CRM1 in NB cell lines also exists in patients with NB tumor: patients with NB with low CRM1 had higher NBAT1 and vice versa. Hence, lower NBAT1 expression in NBs leads to high CRM1 levels, which in turn hijacks p53-dependent tumor suppressor pathways through influencing p53 cellular distribution. NBAT1 controls this CRM1 turnover via its association with an ubiquitin-specific protease USP36. Interestingly, NBAT1 depletion also causes drastic decrease in mitochondrial p53 levels. Loss of the mitochondrial p53 levels could be one of the underlying reasons for the increased cell viability in NBAT1-depleted cells. On the basis of these observations, we speculate that NBAT1 could control multiple p53-dependent nuclear and mitochondrial pathways that may have an impact on tumor suppression. This crucial functional connection between NBAT1 and p53 further reinforced by the fact that NBAT1 itself is a bona-fide p53 responsive lncRNA. These observations suggest that p53 regulates the transcription of NBAT1, which in turn ensures p53 controlled nuclear and mitochondrial pathways, and these in part could contribute to spontaneous tumor regression seen in patients with low-risk NB. This study addresses the mechanisms underlying the loss of the nuclear p53 but future investigations would be required to address the mechanisms underlying the mitochondrial p53 loss in NBAT1-depleted cells.

CRM1 inhibitors are proposed as anticancer therapy for multiple cancers (60). Selinexor is currently being used in clinical trials for acute myelogenous leukemia (AML) and has been shown to work primarily by inducing p53 response in AML via inhibiting its cytoplasmic export (61). Our cell culture-based assay in combination with xenograft studies suggest that combination of CRM1 inhibitor selinexor and MDM2 inhibitor nutlin-3a were potent in eliciting a p53-mediated response in aggressive neuroblastoma cells with cytoplasmic p53 accumulation. More importantly, the combination treatment led to the robust activation of p53 pathway genes, highlighting the suitability of selinexor and nutlin-3a combination treatment for NB. On the basis of these observations, we propose that the combined treatment based on CRM1 and MDM2 inhibitors for patients with high-risk NB, with p53 wild-type and lower NBAT1 expression, could be an effective therapeutic strategy.

L. Kurian reports grants from Else Kröner-Fresenius-Stiftung and NRW stem Cell network during the conduct of the study. M. Fischer reports personal fees from Bayer, Janssen-Cilag, Roche, and BMS outside the submitted work. No disclosures were reported by the other authors.

S. Mitra: Formal analysis, investigation, methodology, writing–original draft. S. Veppil Muralidharan: Formal analysis, investigation, methodology, writing–original draft. M. Di Marco: Investigation, methodology. P.K. Juvvuna: Investigation, methodology. S. Thankaswamy Kosalai: Data curation, formal analysis. S. Reischl: Investigation, methodology, writing–review and editing. D. Jachimowicz: Investigation, methodology. S. Subhash: Data curation, writing–review and editing. I. Raimondi: Investigation, methodology. L. Kurian: Investigation, methodology. M. Huarte: Investigation, writing–review and editing. P. Kogner: Investigation, writing–review and editing. M. Fischer: Investigation, writing–review and editing. J.I. Johnsen: Investigation, methodology, writing–original draft. T. Mondal: Conceptualization, formal analysis, investigation, methodology, writing–review and editing. C. Kanduri: Conceptualization, supervision, funding acquisition, writing–review and editing.

This work was supported by the Knut and Alice Wallenberg Foundation [KAW2014.0057], Swedish Foundation for Strategic Research [RB13–0204], Cancerfonden:CAN2018/591; Swedish Research Council [2017–02834], Barncancerfonden [PR2018–0090], Ingabritt Och Arne Lundbergs forskningsstiftelse, and LUA/ALF to C. Kanduri; and Swedish Research Council (2018–02224), Barncancerfonden (TJ2019–0077 and PR2019–0077) to T. Mondal. Computational support provided by UPPMAX high-performance computing (HPC), part of Swedish National Infrastructure for Computing (SNIC).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.