MYC transcription activation mediated by OCT4 as a mechanism of resistance to 13-cisRA-mediated differentiation in neuroblastoma

Despite the improvement in clinical outcome with 13-cis-retinoic acid (13-cisRA) + anti-GD2 antibody + cytokine immunotherapy given in first response ~40% of high-risk neuroblastoma patients die of recurrent disease. MYCN genomic amplification is a biomarker of aggressive tumors in the childhood cancer neuroblastoma. MYCN expression is downregulated by 13-cisRA, a differentiating agent that is a component of neuroblastoma therapy. Although MYC amplification is rare in neuroblastoma at diagnosis, we report transcriptional activation of MYC medicated by the transcription factor OCT4, functionally replacing MYCN in 13-cisRA-resistant progressive disease neuroblastoma in large panels of patient-derived cell lines and xenograft models. We identified novel OCT4-binding sites in the MYC promoter/enhancer region that regulated MYC expression via phosphorylation by MAPKAPK2 (MK2). OCT4 phosphorylation at the S111 residue by MK2 was upstream of MYC transcriptional activation. Expression of OCT4, MK2, and c-MYC was higher in progressive disease relative to pre-therapy neuroblastomas and was associated with inferior patient survival. OCT4 or MK2 knockdown decreased c-MYC expression and restored the sensitivity to 13-cisRA. In conclusion, we demonstrated that high c-MYC expression independent of genomic amplification is associated with disease progression in neuroblastoma. MK2-mediated OCT4 transcriptional activation is a novel mechanism for activating the MYC oncogene in progressive disease neuroblastoma that provides a therapeutic target.


Introduction
Neuroblastoma is a cancer of the sympathetic nervous system mostly occurring in young children 1 . High-risk patients are >18 months old with stage 4 disease, stage 3 with unfavorable histopathology, or any tumor with MYCN gene amplification 2 . Treatment of high-risk neuroblastoma with non-myeloablative (conventional) chemotherapy alone achieves an initial response in most patients, but eventually 80-90% of patients develop progressive disease (PD) refractory to further therapy 3 . Neuroblastoma can spontaneously mature to a benign tumor known as ganglioneuroma and a variety of agents have been shown to induce growth arrest and morphological differentiation (neurite outgrowth) of human neuroblastoma cell lines 4 . All-trans retinoic acid (ATRA) and isotretinoin (13-cis-retinoic acid = 13-cisRA)-induced morphological differentiation 5 , decreased MYCN expression, and decreased cell proliferation in both MYCN geneamplified and non-amplified human neuroblastoma cells in vitro 6,7 .
A randomized Phase III clinical trial showed that intensive myeloablative therapy supported by autologous hematopoietic stem cell transplantation (ASCT) improved outcome for high-risk neuroblastoma relative to conventional chemotherapy [8][9][10] , and that outcome was further improved using 13-cisRA to treat minimal residual disease 11 after ASCT 12 . A subsequent trial further improved survival by adding dinutuximab (a chimeric anti-GD2 antibody) and cytokines to post-ASCT maintenance therapy 13 . However, many high-risk neuroblastoma patients still ultimately die from progressive disease (PD) 1 . In this study, we sought to define molecular mechanisms of 13-cisRA resistance in neuroblastoma. We observed that c-MYC (but not MYCN) expression was consistently increased in neuroblastoma cell lines and patient-derived xenografts (PDX) established from clinical samples of PD neuroblastoma. Here we report a novel pathway of MYC transcriptional activation that confers resistance to 13-cisRA, is frequently activated in PD neuroblastoma, and may serve as a therapeutic target.

MYC is transcriptionally activated in 13-cisRA-resistant neuroblastoma cells
Tumor biopsies are infrequently obtained for research from children with neuroblastoma at time of progressive disease (PD) [14][15][16] . Bone marrow and blood samples are collected by the Children's Oncology Group/ALSF Childhood Cancer Repository (www.CCcells.org) to establish and bank low-passage patient-derived cell lines and PDXs, including lines established from patients who developed PD after therapy with 13-cisRA. These include pre-therapy at diagnosis (Dx) and post therapy PD models established from pre-and post therapy tumor samples from the same patient. Clinical annotation and validation of origin of the cell lines and PDXs used in this study are shown in Supplementary Tables 1 and 2. MYCN and c-MYC protein levels were compared in 19 Dx (Fig. 1a) and 16 PD (Fig. 1b) cell lines. c-MYC protein levels were significantly higher in PD models compared with Dx models (Fig. 1c). To determine whether increased c-MYC in PD cell lines might be due to the selective pressure of 13-cisRA treatment, the SMS-LHN (LHN) human neuroblastoma cell line (MYCN expression without genomic amplification) 17 was treated with 13-cisRA, mimicking the clinical drug exposures obtained in children with neuroblastoma (5 μM for 2 wks of every 4 wks for 6 cycles). LHN responded initially to 13-cisRA with neurite outgrowth and decreased MYCN protein ( Fig. 1d and Supplementary Fig. 1a). After six cycles of 13-cisRA, a resistant variant cell line (LHN-R) grew continuously during 13-cisRA treatment without showing neurite outgrowth (Fig. 1d) or cell cycle arrest (Fig. 1e). In LHN-R treated with 13-cisRA, mRNA and protein expression of c-MYC was increased while MYCN remained suppressed (Fig. 1f, g).
High c-MYC protein is seen in 11% of neuroblastoma at diagnosis (MYC genomic amplification seen in 1%) and has been associated with a poor clinical outcome 18 . Enhancer hijacking and focal enhancer amplification have been suggested as mechanisms for activating MYC expression in neuroblastoma 19 . However, the incidence of MYC transcriptional activation at PD and its molecular mechanisms remain unknown. As c-MYC was elevated in PD neuroblastoma cell lines and in those selected for resistance to 13-cisRA, we sought to demonstrate that c-MYC overexpression confers resistance to 13-cisRA. Stable clones expressing wild-type c-MYC with 439 or 454 amino acids (AUG-and CUG-initiated) as well as deletion (Δ72-209, encompassing MYC box 3) or point mutation (V409D, functionally critical in MAX dimerization) were created by transducing 4-hydroxytamoxifen (4-OHT)-inducible estrogen receptor (ER)-fusion constructs ( Supplementary Fig. 1b) and confirmed exogenous protein levels for wild-type and mutant c-MYC (Supplementary Fig. 1c). Cyclin A, a c-MYC downstream target indicating c-MYC functionality, was detected in the nucleus of cells expressing c-MYC 439 , c-MYC 454 , and the V409D mutant after 13-cisRA treatment ( Supplementary  Fig. 1d). LHN cells expressing wild-type MYC did not respond to 13-cisRA treatment, whereas the deletion mutant showed significant cell cycle arrest by 13-cisRA (p < 0.01, Supplementary Fig. 1e).
To show the direct functional connection between c-MYC and Cyclin A expression, we used CRISPR/Cas9 to knockout MYC in LHN-R. MYC double knockout (KO) was lethal to LHN-R cells, and thus the experiments were conducted in MYC single KO cells. In the MYC KO cells, 13-cisRA decreased Cyclin A expression (Fig. 1h) and induced neurite outgrowth (Fig. 1i). MYC KO increased MYCN expression (Fig. 1h), and MYC overexpression resulted in the decrease in MYCN ( Supplementary Fig.  1f). We noted that these data show that c-MYC overexpression causes resistance to 13-cisRA in neuroblastoma cell lines.

POUs domain of OCT4 is required for MYC activation
Of the three variants of OCT4 21 , the POU domain is highly conserved through evolution 22 and is a bipartite DNA-binding unit composed of two DNA-binding domains, POU-specific domain (POUs) and POU homeodomain (POU HD ), tethered by a flexible linker 21 . Residues on POU HD activates downstream target gene transcription 23   Mass spectrometry analyses identified OCT4-binding proteins and also determined post-translational modification of OCT4 in LHN-R cells transduced with mycDDK-tagged POU5F1 (Fig. 3a). The mass spectrometry and the prediction of kinase interactions with OCT4 generated from PhosphoMotif Finder (Supplementary Table 3) identified two kinases, DNA-PKcs and MAPKAPK2 (MK2), which bind OCT4 and phosphorylate at S93 (DNA-PKcs) and S111 (MK2) (Fig. 3b), and this nuclear binding was confirmed by subcellular fractionation co-IP (Fig. 3c, left). The direct binding of MK2 and OCT4 was verified using purified proteins by Ni-NTA pull down (Fig. 3c, top) or anti-FLAG IP (Fig. 3c, bottom).
MK2 is a downstream substrate of p38MAPK pathway and post-transcriptinally regulates cytokines, indicating its role as a pro-inflammatory mediator 24 . By regulating phosphorylation and mRNA stability, MK2 also involves in actin remodeling, cell migration, cell cycle, and apoptosis [25][26][27] . To determine the role of MK2 in OCT4 activation, MAPKAPK2 (gene encoding MK2) was stably knocked down in several neuroblastoma cell lines, including COG-N-508h a cell line established from a patient that progressed after 13-cisRA treatment. MAP-KAPK2 knockdown resulted in reduced OCT4 and c-MYC (Fig. 3d), and the MAPKAPK2 knockdown cells treated with 13-cisRA showed reduced Cyclin A, increased expression of NeuN ( Fig. 3e), increased neurite outgrowth (Fig. 3f), and arrested cell cycle (Fig. 3g), indicating the restoration of sensitivity to 13-cisRA. Neurite outgrowth in response to 13-cisRA in MAPKAPK2 knockdown cells was also seen in the PD neuroblastoma cell line COG-N-443h ( Supplementary Fig. 5a, b).
Given that MK2 is activated via phosphorylation, the correlation of mRNA or protein levels of MK2 may not indicate the activity of the enzyme. However, in a search of potential biomarkers in existing database, we evaluated the relationship of MAPKAPK2, MYC, or MYCN  expression and patient survival in the NCI TARGET database of mRNA expression for 175 MYCN nonamplified primary neuroblastoma tumors 28 . MAPKAPK2 expression strongly positively correlated with MYC expression, but inversely correlated with MYCN expression ( Fig. 4a, b, p < 0.001). In addition, we observed that high MAPKAPK2 expression was associated with poor survival (Fig. 4c, adjusted p = 0.012, Supplementary Fig.  5c-e). In neuroblastoma primary tumors with low MYCN, tumors with high c-MYC protein had a trend of higher MK2 protein expression relative to those with low MYCN and c-MYC proteins (Fig. 4d).
Expression of pMK2 and pOCT4 activating c-MYC were associated with progressive disease in neuroblastoma We next bacterially expressed the human recombinant proteins of wild-type OCT4 and OCT4 mutants (S93A and S111A, constructs shown in Fig. 4e, verified the expression in Supplementary Fig. 6a, b), and purified them followed by thrombin cleavage (Fig. 4f, g). An in vitro MK2 kinase assay using the recombinant proteins showed that MK2 phosphorylated OCT4 WT but not OCT4 S111A (Fig. 4h).
To demonstrate that S111 residue of OCT4 is phosphorylated in cell lines established from PD patients as well as LHN-R cells, a custom antibody of pOCT4 S111 was produced. The specificity was confirmed in LHN-R cells exogenously expressing OCT4 WT and two mutants. Anti-pOCT4 S111 antibody did not recognize the S111A mutant OCT4 but it reacted with OCT4 WT and the S93A mutant while all three proteins were detected by anti-OCT4 antibody (Fig. 5a). Neutralized antibody (epitope saturation with an excess of pOCT4 S111 peptide) did not detect pOCT4 S111 , whereas the antibody pre-incubated with OCT4 S111 peptide was able to detect pOCT4 S111 ( Fig. 5b), confirming specificity of the anti-pOCT4 S111 antibody.
The constitutive protein expression of c-MYC, OCT4, pOCT4 S111 , MK2, and pMK2 was compared between cell lines established pre-therapy (Dx) (Fig. 5c) and those established at time of progressive disease (PD) (Fig. 5d, quantitation data in Fig. 5e) in non-paired as well as paired (Dx and PD established from the same patient) cell lines (Fig. 5g). c-MYC protein levels in Fig. 5 are from Fig.  1a, b to show the association between c-MYC and OCT4/ pOCT4 S111 /MK2/pMK2. The levels of c-MYC, pOCT4 S111 , and pMK2 were significantly higher in PD cell lines compared with Dx cell lines (p < 0.01) from other patients. OCT4 levels were not significantly different due to one outlier (COG-N-497h). In matched-pair cell lines (Dx and PD from same patient), the seven PD cell lines showed higher expression of c-MYC, OCT4, and pMK2 relative to autologous Dx lines from the same patients. We compared cell lines established from patients who progressed prior to 13-cisRA treatment to cell lines from patients treated with 13-cisRA (Fig. 5f). c-MYC and pMK2 were significantly higher in PD cell lines with no prior 13-cisRA exposure relative to Dx lines and no significant differences were seen between PD lines from patients treated with 13-cisRA and PD lines without 13-cisRA. Thus, pMK2-mediated increased c-MYC expression is not limited to tumors exposed to 13-cisRA.
When the S111 residue of OCT4 was mutated, pOCT4 S111 was not detected (Fig. 6a). S111 mutation did not affect the nuclear translocation of the protein, but it did result in decreased cyclin A (Fig. 6a), reduced c-MYC expression (Fig. 6b), and decreased OCT4 protein stability (half-lives of proteins: >6 vs 3 h, Fig. 6c). When we further treated these cells with chloroquine or bortezomib in the presence of cycloheximide (CHX), bortezomib was able to (see figure on previous page) Fig. 3 OCT4 interacting proteins by protein identification (ID) and post-translational modification (PTM) and verification of the protein expression and the role of the interacting MK2 protein in 13-cisRA resistance. a Left: Coomassie brilliant blue stained SDS-PAGE gel separating immunoprecipitated subcellular fractions of cells with exogenous OCT4 expression using anti-FLAG antibody. M: Marker, C: cytosolic, N: nuclear. For protein ID, the entire lanes were analyzed, and for PTM, the specific OCT4 band was subjected to mass spectrometry analyses. Right: OCT4 protein sequence with PTM status identified by mass spectrometry. Underline: mass spectrometry coverage, purple: not phosphorylated or acetylated, Red: phosphorylated, acetylated, or undetermined. b Predicted PTM sites in NTD and POUs domains of OCT4 interacting with/phosphorylated by MK2 (S 111 ) and DNA-PKcs (S 93 ). c Left: Direct nuclear interaction of MK2 or DNA-PKcs with OCT4 confirmed by immunoprecipitation in LHN-R cells stably transduced with the doxycycline-inducible construct of wild-type OCT4 with mycDDK-tag. Right: The direct interaction between OCT4 and MK2 is also confirmed by Ni-NTA pull down (top panel), and also by immunoprecipitation by FLAG using purified proteins (bottom panel). d Effect of stable MAPKAPK2 knockdown on OCT4 and c-MYC protein expression in LHN-R (selected for resistance to 13-cisRA in the laboratory) and COG-N-508h (established from a PD patient sample after 13-cisRA treatment). The results were reproducible in a repeat experiment. e Protein expression of OCT4, MK2, Cyclin A, and NeuN (mature neuronal marker) in LHN-R cells with MAPKAPK2 knockdown. Cells stably transduced with non-targeting NT-shRNA or MAPKAPK2-shRNA were treated with vehicle control or 5 μM 13-cisRA for 14 days. The results were reproducible in a repeat experiment. f Reversal of 13-cisRA resistance shown as neurite outgrowth in MAPKAPK2 knockdown LHN-R cells. Cells stably transduced with non-targeting NT-shRNA or MAPKAPK2-shRNA were treated with vehicle control or 5 μM 13-cisRA for 14 days. A scale bar: 100 μM. g Reversal of 13-cisRA resistance shown as cell cycle arrest in MAPKAPK2 knockdown LHN-R cells. Cells stably transduced with non-targeting NT-shRNA (S-phase cells: 11 ± 0.5% vs 9 ± 2.2%, p = 0.11) or MAPKAPK2-shRNA (S-phase cells: 8.5 ± 0.1% vs 3.2 ± 0.2%, p < 0.01) were treated with vehicle control or 13-cisRA for 14 days. Left: representative histograms, right: percentage of cells in each phase of cell cycle.
rescue OCT4 S111A from degradation (Fig. 6d), suggesting that the degradation of the mutated OCT4 is expedited via proteasomes. MK2 inhibition by a small molecule MK2 inhibitors resulted in decreased pOCT4 S111 , OCT4, c-MYC, and Cyclin A in CHLA-20, LHN-R, COG-N-289, COG-N-334, and COG-N-415 (Fig. 6e, f)  cytotoxicity testing showed that cell lines with high c-MYC were more sensitive than low c-MYC lines to an MK2 inhibitor in four non-paired cell lines (Fig. 6g) and two pairs of matched-pair cell lines (Fig. 6h). These data suggest that the inhibition of OCT4 phosphorylation at S111 residue affects the stability of OCT4. Protein expression of c-MYC, pOCT4 S111 , and pMK2 T334 was significantly higher in neuroblastoma PDXs (Supplementary Table 2) established at progressive disease compared with PDXs established at diagnosis (Fig.  7a, b). Although the PDX sample size was smaller than cell lines, a strong correlation was seen between c-MYC and pOCT4 S111 as well as c-MYC and pMK2 T334 (Fig. 7c). Also, expression of c-MYC, pOCT4 S111 , and pMK2 T334 from pre-therapy was higher than progressive disease in a matched pair of Dx/PD cell lines (603 h/623 h, Fig. 5g) and the corresponding PDXs (603×/623×) established from the same patient (Fig. 7a). We assessed phosphorylation of both T334 and T222 (Figs. 5 c, d, g, 7a, Supplementary Fig. 7a, b) as both are required for maximal activation of MK2 29,30 . Thus, phosphorylation of OCT4 by MK2 transcriptionally activates MYC and this oncogenic pathway is upregulated in progressive disease relative to pre-therapy neuroblastomas.

Discussion
MYCN genomic amplification is well established as one of the prognostic markers in neuroblastoma 31,32 . More recently, the prognostic significance of c-MYC has been reported in a small group 33 , and then in a large cohort of neuroblastoma patients with undifferentiated/poorly differentiated tumors 18 . Although MYC amplification is rare in neuroblastoma, high c-MYC protein in patients with undifferentiated tumors was as frequent as 11% at diagnosis, suggesting that MYC transcriptional activation rather than gene amplification is a driver of aggressive tumor behavior in most c-MYC-driven neuroblastomas. As tumor biopsies of patients at PD are infrequently obtained from children with neuroblastoma, we have instead utilized low-passage cell lines and PDXs established under uniform conditions from bone marrow or blood of children with neuroblastoma at Dx prior to therapy and also at PD. Our discovery of c-MYC mediating resistance to 13-cisRA and being highly expressed in PD neuroblastoma suggests that MYC transcriptional activation occurs both in a subset of patients at diagnosis and likely more frequently in tumors at time of PD.
Tumors that recurred after PIK3CA inactivation acquired focal amplification of MET or MYC in an in vivo mouse breast cancer model expressing PIK3CA H1047R34 . Also, the MYC/eIF4E axis is a mediator of drug resistance to BEZ-235 in breast cancer cells 35 . In leukemia, MYC activation compromised tyrosine kinase inhibitor sensitivity and functional inhibition of MYC overcame resistance by promoting differentiation 36 . In our experiments, MYCN regulation by 13-cisRA was not attenuated by repeated exposure to the drug in neuroblastoma cells, but escape from sustained suppression of MYCN occurred via MYC transcriptional activation. MYC and MYCN are structurally and functionally homologous 37 , and Mycn can functionally replace Myc in murine development, cell growth and differentiation 38 . We showed that MYC knockdown increased MYCN protein, but the cells did not regain their proliferative properties, suggesting that 13-cisRA-resistant cells are addicted to c-MYC, which functionally replaces MYCN in 13-cisRA-resistant neuroblastoma. Four of the seven matched-pair cell line and xenograft models were established from patients who were not treated with 13-cisRA due to rapid progression of the disease and the PD cell lines from four pre/post therapy pairs along with one non-paired PD cell line showed elevated pMK2 and c-MYC. Thus, the OCT4/c-MYC axis we discovered likely is a mechanism of early disease progression in some patients and is acquired to promote resistance to 13-cisRA in others.
In the current study, we defined MK2 phosphorylation of OCT4 as a novel mechanism for oncogenic activation of MYC in neuroblastoma (Fig. 7d). Oct4 is an important (see figure on previous page) Fig. 4 MK2 as a biomarker for 13-cisRA resistance and prognosis in neuroblastoma. a Positive correlation between mRNA expression of MYC and MAPKAPK2 in 249 neuroblastoma patients. Expression data from the neuroblastoma NCI TARGET database (https://ocg.cancer.gov/programs/ target/data-matrix). b Inverse correlation between MYCN and MAPKAPK2 mRNA expression in 249 neuroblastoma patients (Data: NCI TARGET database https://ocg.cancer.gov/programs/target/data-matrix). c Overall survival of patients by MAPKAPK2 mRNA expression in neuroblastoma from the NCI TARGET database. Of the total patients (n = 247), only patients with MYCN non-amplification (n = 175) were used for the analysis. The data was scanned to identify maximum separation of the curves, and the p-value was adjusted by Bonferroni adjustment. The overall survival using the median MAPKAPK2 expression and the event-free survival are shown in Supplementary Fig. 5c-e. d MK2 protein expression by immunohistochemistry staining in neuroblastoma primary tumors collected at diagnosis with low expression of both MYCN and c-MYC proteins (upper panel) and with high c-MYC protein (lower panel) expression. A white scale bar: 20 μM, a black scale bar: 20 μM. e Constructs (plasmid: pGEX-4T-1) encoding a human wild-type OCT4 or OCT4 mutants (S93A and S111A) with a GST tag and thrombin cleavage site at the NH 2 -terminus. f Human recombinant OCT4 proteins expressed in BL21/DE3 strain of E. coli after IPTG induction, GST column purification, and thrombin cleavage subjected to SDS-PAGE and stained with Coomassie brilliant blue solution. g Proteins in f detected by immunoblotting using anti-OCT4 antibody. h In vitro kinase assay of MK2 on phosphorylating OCT4WT and OCT4 S111A . The results were reproducible in a repeat experiment.
transcription factor that controls self-renewal and pluripotency in the early stages of mammalian embryogenesis 39 , and it cannot be replaced by any other POU family members in reprogramming somatic cells into induced pluripotent stem cells (iPSC) 40 . Ectopic expression of OCT4 blocks progenitor-cell differentiation and causes dysplasia in epithelial tissues 41 . Not only is OCT4 required to maintain the pluripotency of human E A OCT4WT-mycDDK OCT4 S111A -mycDDK OCT4 S93A -mycDDK anti-pOCT4 S111 pOCT4 S111 peptide OCT4 S111 peptide embryonic stem cells, but it is also involved in lineage specification 42,43 . For reprogramming somatic cells, OCT4 along with c-MYC, KLF4, and SOX2 are required to induce pluripotency in both human and mouse somatic cells in vitro 44 . Other studies demonstrated that higher expression of OCT4 occurs in embryonic or cancer stem cells 39,[45][46][47] . One study showed Oct4 is upstream of mych, a Myc family member, in the development of zebrafish 48 .
Our study is the first to show OCT4 as an upstream regulator of c-MYC in human cancer. Previous studies in embryonic stem cells showed that phosphorylation sites in OCT4 POU HD (T235, S236, and S229) are important for pluripotency 49,50 . This is especially important as our data suggest that the function of OCT4 in cancer may be distinguished from its role in pluripotency. The downstream activation of genes by OCT4, in cancer or in pluripotency will depend on kinase-specific phosphorylation of residues. Our data establish the MK2/OCT4/c-MYC axis as a mechanism of progressive disease neuroblastoma regardless of prior to 13-cisRA exposure. This mechanism of MYC transcriptional activation may be operative in other cancer types and could potentially be an important therapeutic target. The p38 MAPK (p38) kinase, upstream of MK2, has been investigated as a therapeutic target in inflammatory diseases due to its role in the regulation of the mRNA stability of TNF-α, along with several other mediators with unknown immune responses 51 . Many p38 inhibitors were tested in clinical trials but none progressed to phase III mainly due to systemic toxicities (hepato-, cardiac toxicities, and CNS disorders) 52,53 , possibly due to p38 being involved in the regulation of more than 60 substrates with various physiological roles 54 . For that reason, MK2 (the first substrate of p38) is being tested as an alternative target to p38 in inflammatory diseases 55 . Regardless, experiments were conducted to confirm p38 as the upstream activator of MK2 in neuroblastoma ( Supplementary Fig. 7c). As our data show that MK2 mediates c-MYC activation, which is an important oncogenic pathway in many cancers, and direct inhibition of c-MYC is challenging 56 , inhibiting MK2 action on OCT4 is potentially a novel therapeutic target in cancers with high c-MYC expression.
It has been reported that the degradation of retinoid receptors is known to induce resistance to retinoids. However, in LHN cells and its resistant clone, LHN-R, no differences in the expression of RAR-β was observed ( Supplementary Fig. 7d), indicating that the degradation of retinoic receptors is not the mechanism of resistance to 13-cisRA in neuroblastoma.
In summary, our studies define the MK2/OCT4/c-MYC axis as a novel mechanism of MYC transcriptional activation in neuroblastoma that mediates resistance to 13-cisRA by rescuing 13-cisRA induced MYCN downregulation and is commonly found in progressive disease neuroblastomas. Regulation by MK2 of OCT4 binding to newly identified binding sites in the promoter/enhancer region of MYC may provide a novel molecular target for regulating the MYC oncogene. This novel mechanism for regulation of c-MYC expression warrants investigation in cancers other than neuroblastoma.

Collections of human neuroblastoma clinical samples and their cell lines and derivatives
All clinical specimens from patients ≥18 months of age with stage 4 neuroblastoma at diagnosis (Dx) and post therapy progressive diseases (PD) were collected from year 2000 through 2016 at the Cancer Center laboratory of C. Patrick Reynolds in the Texas Tech University Health Sciences Center (TTUHSC, Lubbock, TX). The Children's Oncology Group (COG) provides aliquots of these samples to the laboratory of Dr. Reynolds for the specific purposes of patient-derived cell line and xenograft establishment (The COG repository for cell lines and xenografts: www.CCcells.org). For the proposed activities, enrollment of specimen donors was performed by a strict procedure that has been approved by the Institutional Review Boards at participating institutions, with informed consent and the utmost attention to the issues of patient safety, anonymity and confidentiality. In neuroblastoma patients, tumor samples from PD are rarer than Dx samples as tumor re-biopsy at relapse is often not clinically indicated. In contrast, bone marrows and blood samples are often obtained from neuroblastoma patients at relapse. Cell lines established from tumors collected at Dx and at PD and cell lines with matching Dx/PD pairs (established from seven patients) were cultured in IMDM medium supplemented with 2 mM L-Glutamine, human 1× ITS [10 µg/ml Insulin, 5.5 µg/ml Transferrin (ironfree), and 5 ng/ml Sodium selenite], and 20% heatinactivated FBS. They were mycoplasma-free and routinely checked for cell line identify using short tandem repeat genotyping as compared with the original primary sample material within the database: www.CCcells.org. The models used are show in Supplementary Tables 1, 2.

Establishment of 13-cisRA-resistant neuroblastoma models
To provide models for defining the molecular mechanisms of resistance to 13-cisRA, SMS-LHN (LHN) human neuroblastoma cell line (MYCN expressing but lacking MYCN or MYC genomic amplification) 17 was treated with a clinically achievable concentration of 13-cisRA (5 μM) according to the dosing schedule clinically used in high-risk patients (i.e., 2-week on/2-week off/6 cycles). The 13-cisRA resistance was considered established when neurite outgrowth was not seen in the presence of 13-cisRA, and the resistant variant was named SMS-LHN-R (LHN-R). The cells were maintained in RPMI-1640 supplemented with 10% heat-inactivated FBS.

Immunoblotting and immunoprecipitation
The cell lines and seven matched pairs of Dx/PD human neuroblastoma samples were prepared for IP or IB as described previously 57 . Densitometric analysis of protein quantitation was determined by Quantitiy One® 1D software v4.6.
(see figure on previous page) Fig. 6 The role of MK2 phosphrylation at S111 in c-MYC increase and the effect of MK2 inhibition in cell viability. a OCT4 at S111 residue is phosphorylated in LHN-R. OCT4WT and OCT4 S111A mutant were exogenously expressed in LHN-R cells. The expression of pOCT4 S111 , OCT4, c-MYC, and Cyclin A was evaluated in subcellular fractions. b Assessment of DNA-binding ability of wild-type OCT4 and mutant OCT4 S111A using MYC −1209/−1140 /MYC reporter assay system. Empty vector, POU5F1-mycDDK and POU5F1 S111A -mycDDK (4 μg each) were separately co-transfected with reporter gene MYC −1209/−1140 /DDK-MYC-mER TM (4 μg) in HEK293FT cells. After 48 h, the equal amount of protein lysates (20 μg) were run and analyzed by SDS/PAGE and WB using specific antibodies, as indicated. Anti-ERα: DDK-c-MYC-mER TM expression, anti-DDK (FLAG): expression of exogenous mycDDK-tagged wild-type OCT4 and its mutant. c OCT4 stability was decreased in OCT4 S111A mutant relatively to OCT4 WT , shown by cycloheximide (CHX) treatment. The cells were transduced with a doxycycline-inducible system to exogenously express OCT4 WT or OCT4 S111A mutant, treated with doxycycline (Dox) for 48 h, followed by cycloheximide (CHX) incubation for various times before immunoblotting. d Rescue of OCT4 S111A protein degradation by bortezomib. The LHN-R cells were transduced with a Dox-inducible system to exogenously express OCT4 S111A mutant, treated with Dox for 24 h, followed by 10 μg/ml of cycloheximide (CHX) along with Bortezomib (1 μM, proteasome inhibitor) or Chloroquine (50 μM, lysosome inhibitor) incubation for 6 h before immunoblotting. DMSO was served as a vehicle control. e Decreased pOCT4 S111 , OCT4, c-MYC,  Fig. 5g. b Dot plots quantitating immunoblotting data from a. The values were normalized in two ways by the expression of specific proteins in CHLA-20 and GAPDH. c Correlation between c-MYC and pOCT4 S111 , c-MYC and pMK T334 , and pOCT4 S111 and pMK T334 . The results of linear regression analyses with 95% confidence interval (dotted line) are presented (n = 17). d Proposed mechanism of c-MYC transcriptional activation in progressive disease neuroblastoma. MK2 shuttles into the nuclei of cells and phosphorylates OCT4 at S111 residue. There are two OCT4-binding sites in the c-MYC promoter/enhancer region between −1209 and −1140. The binding of pOCT4 S111 increases transcriptional activation of MYC.
CRISPR/Cas9 KO plasmid (h) and MYC HDR plasmid (1 μg each), was added directly to the dilute UltraCruz® Transfection Reagent (150 μl), and incubate for 20 min at room temperature. The MYC CRISPR/Cas9 KO plasmid (h) consists of a pool of three plasmids, each encoding the Cas9 and a target-specific 20-nucleotide gRNA designed for MYC knockout. When cells reach about 70-80% confluency, the Plasmid DNA/UltraCruz® Transfection Reagent Complex was added dropwise and gently mixed. After 96 h incubation, the cells were visually confirmed by detection of the RFP via fluorescent microscopy. Furthermore, the cells with MYC KO (+/− or −/−) were enriched by selection with complete medium containing Puromycin (0.5 μg/ml) for at least 2-3 weeks. The MYC knockout efficiency in the stable cell clones was confirmed by IB.

Protein/DNA arrays hybridization and analysis
Panomics Protein/DNA Combo Array carrying 345 specific DNA-binding elements was used for identification of TFs critical for gene regulation in 13-cisRA sensitive cells LHN and the resistant cells LHN-R, according to the instruction manual (Affymetrix). The signals were densitometrically quantified and normalized to an internal standard using Image Studio TM software v5.x.

Cloning and mutagenesis
The mER TM cDNA was amplified by the Expand High

Luciferase reporter assay
Luciferase reporter assay was performed using Ready-To-Glow TM Secreted Luciferase Reporter System according to manufacturer's instructions (Clontech). In addition, a secreted alkaline phosphatase (SEAP) reporter plasmid was included in all transfections for normalization. The relative fold increase in luciferase activity was normalized with the SEAP activity.
Construction of regulatory sequences of MYCN promoter and MYC enhancer into the DDK-MYC-ER gene reporter system The DDK-MYC 454 -mER TM DNA fragment was amplified and inserted into the Age1/Not1 sites of pMetLuc2 Reporter to replace Luc gene (henceforth DDK-MYC-ER). MYCN promoter (Xho1/BamH1) (−1/−1098 upstream of the AUG start site of MYCN) was amplified with its specific primers using SK-N-BE(2) genomic DNA as a template. PCR products were cut with Xba1/BamH1-HF, EcoR1-HF/BamH1-HF and Xho1/BamH1-HF, respectively, and then subcloned into DDK-MYC-ER or MYCN-mycDDK reporter. 293FT cells were co-transfected with these (4 µg each) along with mock vector (pCMV6-entry-mycDDK) or vector expressing POU5F1 and the indicated mutant forms (4 µg each) for 48 h. The whole-cell lysates were subjected to SDS-PAGE/IB, and then probed with the indicated antibodies. The primers used in the reaction are shown in Supplementary Table 4.
Electrophoretic mobility shift assay (EMSA) EMSA 58 was conducted using the Panomics' EMSA Gel Shift Kit (Affymetrix) according to the manufactures' instructions. The sequences of biotin-labeled double-stranded WT MYC probes are shown in Supplementary Table 4.

Chromatin immunoprecipitation (ChIP) assay
ChIP analysis was performed as previously described 59 . Briefly, the proteins in LHN-R-resistant neuroblastoma cells were cross-linked to DNAs with 1% formaldehyde for 10 min at room temperature. The reaction for protein/ DNA cross-linking was stopped by addition of Glycine (125 mM). Cells were rinsed twice with ice-cold 1× PBS and harvested by a silicone scraper. After centrifugation for 5 min at 1000×g, cells were lysed subsequently with the ice-cold lysis buffer (iL1b and iL2). Cells were then resuspended in complete shearing buffer iS1b with protease inhibitors to shear cross-linked protein/DNA to an average size of 200-1000 bp by sonication using UCD-300 Bioruptor Plus (Diagenode). An aliquot of the sheared chromatin DNAs were used as an input. The chromatin extracts containing protein/DNA fragments were ChIPed at 4°C overnight using ChIP-grade rabbit anti-OCT4-1 or rabbit anti-OCT4-2 polyclonal antibody mixing together with DiaMag protein A-coated magnetic beads. The normal rabbit IgG and CTCF antibody were used as negative and positive control, respectively. The magnetic beads were washed four times with wash buffer W1, W2, W3, and W4, and DNA fragments were eluted with elution buffer iE1 and decross-linked with iE2. After 4 h incubation at 65°C, DNA was washed with Wash buffer 1 and 2 and then enriched by purification using IPure beads. The semi-quantitative real-time PCR was performed using the Mastercycler (Eppendorf). PCR parameters used in the reaction were as follows: 95°C/ 3 min; 40 cycles with denaturation at 95°C/20 s, annealing at 60°C/20 s, and polymerization at 72°C/ 20 s. PCR products were analyzed electrophoretically on 2% agarose gels and amplification was performed in triplicate and repeated twice. Relative occupancy values were calculated by determining the apparent ChIP efficiency (ratios of the amount of immunoprecipitated DNA to that of the input sample) and normalized to the level observed at a control region, which was defined as 1.0. The four sets of primers used to quantify the DNA fragments of MYC from the ChIP-enriched DNA are listed in Supplementary Table 4.

RNAi-mediated gene silencing
Transfection of mammalian cells was carried out as described in the Amaxa's cell line database (Lonza). Briefly, LHN-R-resistant cells were electroporated with 50 or 100 nM of scramble non-targeting, POU5F1 (71264) or TCF3 (s13874) pre-designed siRNA (Ambion) in solution SF using program EN-150 of the Amaxa 4D-Nucleofector TM System (Lonza). Cells were grown in RPMI-1640/10% FBS complete medium for 2-3 days at 37°C in a 5% CO 2 incubator. The protein levels of OCT4 and TCF3 were validated using IB with specific antibodies.
Assay of 13-cisRA resistance by flow cytometric analysis of nuclear DNA contents Effect of 13-cisRA treatment on cell proliferation was determined by examining the distribution of cells in different phases of the cell cycle by flow cytometry. The LHN, LHN-R, LHN-MYC stable clones, and LHN-R-POU5F1 knockdown clones were separately plated in a T75 flask overnight. The cells for testing 13-cisRA resistance were treated with vehicle or 13-cisRA (5 µM) in the complete medium for 14 days with the drugs being replenished every 96 h. To evaluate the cell cycle arrest, cells (10 6 ) were washed once with ice-cold 1× PBS and then fixed and permeabilized in 70% ethanol at −20°C for 1 h. After centrifugation, cells were incubated with 0.5 ml PI/RNase staining buffer (BD Biosciences) in the dark at room temperature. After 15 min, DNA content was analyzed using a LSR II flow cytometer (BD Biosciences), which was operated with DiVa software v4.1.2. DNA content histogram analysis was performed using FlowJo software v7.6.
Protein identification by MALDI-TOF and post-translational modification by Triple Quad mass spectrometry OCT4-mycDDK protein was pulled down by IP using EZview Red anti-FLAG affinity purification, and the protein was separated on SDS-PAGE, stained with the Colloidal blue, and cut out for analysis at the Proteomics Core Facility of University of Texas Southwestern Medical Center, Dallas, TX.

Purification of human recombinant wild-type OCT4 and its mutant proteins
The BL21-DE3 (EMD Millipore) was transformed by heat-shock method with the pGEX-4T1 plasmid encoding a human wild-type OCT4 or OCT4 mutants (S93A and S111A) with a GST tag and a thrombin recognition sequence at the NH 2 -terminus. The expression of GST-OCT4, GST-OCT4 S93A , and GST-OCT S111A recombinant proteins were induced by adding IPTG (1 mM) for 3 h at 37°C with a vigorous shaking until the cell density reads at OD 600 = 0.6-0.8. The E. coli cells were harvested by spinning at 4000 × g for 20 min at 4°C and then resuspended in BugBuster® Master Mix (EMD Millipore) with complete EDTA-free protease and phosphatase inhibitors (Roche). After incubation the bacterial suspensions were completely lysed on a rotating mixer for 20 min at room temperature, and the crude extracts were centrifuged at 16,000×g for 20 min at 4°C. The soluble extracts were loaded directly onto the GST SpinTrap TM columns with Glutathione Sepharose TM 4B (GE Healthcare), which were pre-equilibrated by adding 1× PBS binding buffer (pH = 7.4). The GST-tagged proteins were bound to the Glutathione Sepharose TM 4B. After washing three times, the bound GST-OCT4 or its mutant fusion proteins were cleaved by an in-gel digestion with thrombin protease (GE Healthcare) for 16 h at 22°C, followed by removal of the GST tag by centrifugation. The thrombincleaved OCT4, OCT4 S93A , and OCT4 S111A recombinant proteins were concentrated by Amicon Ultra Centrifugal Filters with M.W. 10 kDa cut-off (EMD Millipore), and the protein concentration was determined by BCA assay (Pierce). The purified recombinant proteins were subjected to SDS-PAGE and confirmed by staining with Colloidal blue solution (Invitrogen).
Custom anti-human phosphoOCT4 S111 rabbit polyclonal antibody Anti-phosphoOCT4 S111 antibody (RRID_AB_2721810) was produced by GenScript. Anti-pOCT4 S111 antibody was prepared by immunizing two New Zealand rabbits four times with a NH 2 -terminal KLH-conjugated phosphopeptide SNSDGA{pS} 57 PEPCTVT as an antigen. The phospho-specific antibody was affinity-purified through a phosphopeptide-conjugated Sepharose CL-4B column. Eluted IgG was then passed through the corresponding nonphosphorylated peptide (SNSDGASPEPCTVT) column to deplete any IgG that were not specific to pOCT4 S111 .

Patient-derived xenograft (PDX) models
The 6-to-8-week-old nu/nu mice (Envigo) were subcutaneously injected with patient-derived xenografts at 10-20 million cells (100 μl) per mouse. Tumor volume was measured by ½ length × width × height. Mice were sacrificed once the tumor volume exceeded 500 mm 3 for the purpose of our experiments. The animal protocol to establish PDX models is reviewed and approved by the TTUHSC Institutional Animal Care and Use Committee (IACUC).

Statistical analysis
Student's t-test was used to determine statistically significant differences on MS Excel. p-values were two-sided and tests were considered significant at p < 0.05. All the experiments were performed in triplicate and were consistently repeatable; for simplicity, one representative experiment for each condition is shown.
NCI TARGET data: Correlation between MYCN, MYC, and MK2 were assessed numerically and graphically. Specifically, scatter plots were generated to demonstrate the relationships between these genes, and the corresponding Spearman correlations were calculated. Overall survival was defined as the time elapsed from study enrollment to death, with those living at the time of last follow-up censored. Median expression of MK2 was used to categorize patients into low and high expression groups. Log-rank test was used for comparing overall survival of these two groups, for MYCN non-amplified patients (n = 175) of all patients (n = 247). Indeed, MK2 expression cut-off that achieved the best separation of patient survival was obtained by scanning the whole expression range, with a Bonferroni adjustment of p-value performed. All analyses were performed by using the SAS software (Windows v9.4).