Targeting of the MYCN Protein with Small Molecule c-MYC Inhibitors

Members of the MYC family are the most frequently deregulated oncogenes in human cancer and are often correlated with aggressive disease and/or poorly differentiated tumors. Since patients with MYCN-amplified neuroblastoma have a poor prognosis, targeting MYCN using small molecule inhibitors could represent a promising therapeutic approach. We have previously demonstrated that the small molecule 10058-F4, known to bind to the c-MYC bHLHZip dimerization domain and inhibiting the c-MYC/MAX interaction, also interferes with the MYCN/MAX dimerization in vitro and imparts anti-tumorigenic effects in neuroblastoma tumor models with MYCN overexpression. Our previous work also revealed that MYCN-inhibition leads to mitochondrial dysfunction resulting in accumulation of lipid droplets in neuroblastoma cells. To expand our understanding of how small molecules interfere with MYCN, we have now analyzed the direct binding of 10058-F4, as well as three of its analogs; #474, #764 and 10058-F4(7RH), one metabolite C-m/z 232, and a structurally unrelated c-MYC inhibitor 10074-G5, to the bHLHZip domain of MYCN. We also assessed their ability to induce apoptosis, neurite outgrowth and lipid accumulation in neuroblastoma cells. Interestingly, all c-MYC binding molecules tested also bind MYCN as assayed by surface plasmon resonance. Using a proximity ligation assay, we found reduced interaction between MYCN and MAX after treatment with all molecules except for the 10058-F4 metabolite C-m/z 232 and the non-binder 10058-F4(7RH). Importantly, 10074-G5 and 10058-F4 were the most efficient in inducing neuronal differentiation and lipid accumulation in MYCN-amplified neuroblastoma cells. Together our data demonstrate MYCN-binding properties for a selection of small molecules, and provide functional information that could be of importance for future development of targeted therapies against MYCN-amplified neuroblastoma.


Introduction
The MYC family members c-MYC, MYCN and L-MYC are transcription factors crucial for the regulation of normal cellular functions including proliferation, cell growth, differentiation, metabolism and apoptosis. However, the genes encoding these proteins are also the most frequently deregulated oncogenes in several types of human cancers [1,2]. c-MYC and MYCN (hereafter MYC), exert their functions mainly through transcriptional modulation of their target genes. The C-terminal domain of MYC comprises a basic helix-loop-helix leucine zipper domain (bHLHZip), necessary for the dimerization with its partner MAX and for sequence-specific binding to DNA [3], while the Nterminal transactivation domain interacts with co-factors to regulate transcription [2]. There is a large overlap between the downstream targets of c-MYC and MYCN and insertion of the mycn gene into the c-myc locus can fully rescue the embryonic lethal phenotype of a c-myc knockout mouse [4]. However, in normal tissue the expression pattern of these two proteins differ significantly [5,6]. In the developing embryo, MYCN is expressed in certain tissues including the central and peripheral nervous systems, lung and spleen, whereas in adults its expression is very low or absent. In contrast, c-MYC is expressed in all proliferating cells in adults [6][7][8][9].
In human tumors, oncogenic alterations in MYC are common and include point mutations that increase protein stability, gene amplification, gene translocation, and enhanced translation [1,2]. MYCN is amplified in cancers such as neuroblastoma (NB), medulloblastoma, lung cancer and glioma [1,[10][11][12]. In NB, a pediatric cancer of the sympathetic nervous system, MYCNamplification is strongly correlated with poor prognosis and advanced tumor stage, and these tumors are often resistant to multimodal therapy [11,12]. MYC is therefore an attractive target for cancer therapy [13,14]. It has been shown that downregulation of MYC leads to cancer cell growth arrest, senescence, enhanced apoptosis, differentiation and/or tumor regression in mouse models of human cancer [15]. Importantly, even transient downregulation of MYC has been reported sufficient to diminish the tumor burden in animal models [15], and the effects of MYC inhibition on normal tissue has been shown to be well tolerated and reversible in adult mice [16][17][18][19][20][21][22][23]. Several groups have made efforts to target MYC at different levels, including its transcription, translation, hetero-dimerization with MAX, as well as targeting its direct or indirect downstream targets [10,14,[24][25][26]. A number of small molecular compounds inhibiting c-MYC-MAX dimerization have been identified [27,28] and among them 10058-F4 is by far the most studied. Biophysical experiments have shown that it interacts with the C-terminal bHLHZip region of c-MYC [28][29][30][31][32]. A fluorescence polarization assay was used to determine the affinity as well as to identify the binding site of 10058-F4 on c-MYC using different deletions, point mutations and synthetic peptides [30]. NMR measurements confirmed that 10058-F4 binds to amino acid residues 402-412 in the bHLHZip domain of c-MYC [30]. Furthermore, metadynamic molecular simulations and an ion mobility mass spectrometry using a peptide corresponding to the identified binding site, indicated that the compound binds to an inactive, disordered conformation of c-MYC [31,32]. Together these studies suggest that 10058-F4 inhibits the function of c-MYC in a direct manner by preventing c-MYC/MAX hetero-dimerization. Importantly, several reports have shown that 10058-F4 affects c-MYC expression and induces cell cycle arrest, inhibits cell growth, promotes apoptosis and confers chemo-sensitivity in a c-MYC specific manner in various cancer cell types [28,[33][34][35]. In addition, treatment of acute myeloid leukemia (AML) cells with 10058-F4 leads to myeloid differentiation [33]. The effect of 10058-F4 treatment in vivo has been investigated in xenograft models of prostate cancer but no significant antitumor activity could be observed, probably due to its rapid clearance and low potency [36]. In contrast, we have recently demonstrated anti-tumorigenic effects of 10058-F4 in two tumor models of MYCN-amplified neuroblastoma, suggesting that direct MYC inhibition using a small molecule is achievable in vivo [36].
The structurally unrelated small molecule 10074-G5 was identified simultaneously as 10058-F4 as another substance that inhibits the c-MYC/MAX interaction. This molecule also decreased c-MYC protein levels and inhibited cell growth [28], but failed to show any antitumor activity in a xenograft model using a Burkitt's lymphoma cell line [37]. The cognate binding site for 10074-G5 on c-MYC was found to be distinct from that of 10058-F4, spanning amino acid residues 363-381 [30,38] (see Figure 1). Both molecules were found to bind independently of each other, and probably induce only local conformation changes in the bHLHZip domain of c-MYC preventing its interaction with MAX [30,38]. In order to identify more potent compounds, several analogs of 10058-F4 have been synthesized, some of which, including #015, #474 and #764, exhibited improved growth inhibition of c-MYC expressing cells [39].
Since c-MYC and MYCN share structural similarity in the bHLHZip domain we hypothesized that 10058-F4 also targets MYCN. We have previously shown that this compound interferes with the MYCN/MAX interaction leading to cell cycle arrest, apoptosis, and neuronal differentiation in MYCN-overexpressing NB cell lines [40]. In addition, using 10058-F4 as a tool, we found that inhibition of MYCN results in mitochondrial dysfunction leading to lipid accumulation. Importantly, 10058-F4 treatment furthermore increased the survival of TH-MYCN transgenic mice and showed anti-tumor effects in established aggressive NB xenografts [40].
Here, we determined the direct binding of 10058-F4 and additional selected c-MYC-targeting compounds to MYCN by surface plasmon resonance (SPR) (see Figure S1 for the structures of the compounds used). We found that all molecules previously reported to bind to c-MYC also bound to MYCN. Treatment with the small molecules furthermore interfered with the MYCN/ MAX interaction and caused protein degradation, apoptosis, differentiation and lipid formation to different extents in MYCNamplified NB cells.

Folding of the human bHLHZip domain of MYCN and c-MYC
The bHLHZip domains of the MYC proteins are intrinsically disordered with little secondary structure without binding to MAX [41]. In order to verify the folding of the protein domains following purification under denaturing conditions and refolding, circular dichroism (CD) was performed. Comparison of the CD spectra of the purified and refolded bHLHZip domains with a model spectrum of a a-helical protein, indicated the presence of some ahelical structure in both c-MYC and MYCN ( Figure S2). Analyses of the obtained spectra using the CONTIN, SELCON and the K2D algorithms [41][42][43][44][45][46][47] predicted the unstructured content to be very similar in both proteins, approximately 58% for c-MYC and 55% for MYCN. The a-helical structure was predicted to be 26% for c-MYC whereas MYCN had a slightly higher a-helical content, 33%. These values are in agreement with the 27% ahelical content documented for the v-MYC bHLHZip [48]. We found approximately 16% b-sheets in c-MYC and 13% in MYCN, respectively (Table S1). These results are indicative of two incompletely folded protein domains with some a-helical structure, as previously reported for the purified bHLHZip of c-MYC [30,31]. Since the secondary structure content of the  [38]. Each small molecule is positioned under their reported or assumed binding site [38,39]. For the 10058-F4 analogs #474 and #764 as well as its potential metabolite C-m/z 232 the binding sites have not been determined experimentally [39]. Through the similarity of their chemical structure to 10058-F4, it has been assumed that these compounds bind to the same site as indicated. doi:10.1371/journal.pone.0097285.g001 Targeting of MYCN with c-MYC Inhibitors PLOS ONE | www.plosone.org bHLHZip domains confirmed results from previous reports, we next performed surface plasmon resonance (SPR) experiments in order to examine the binding of the selected compounds to the purified bHLHZip domains of both proteins.

Binding of small molecules to the bHLHZip of MYCN and c-MYC
The protein sequences of the human c-MYC and MYCN bHLHZip domains share 56% similarity ( Figure 1). Given the highly structural and functional similarity of the C-terminal regions of c-MYC and MYCN it is very likely that molecules binding to c-MYC also bind to MYCN. In addition to 10058-F4 we included three of its structural analogs (Figure 1; Figure S1) in our study. The analogs #474 and #764 have previously been shown to have greater potency compared to the parent compound as judged by enhanced growth inhibition of c-MYC expressing leukemic cells [39]. In contrast, the analog 10058-F47RH (referred to as 7RH throughout) was found to be inactive in a c-MYC binding assay [39], and was thus included as a negative control. We also tested the binding to MYCN of a predicted in vivo metabolite of 10058-F4, C-m/z 232, in order to examine if the modified molecule still possesses some of the capacities of 10058-F4 [36]. Furthermore we included the structurally unrelated compound 10074-G5, previously shown to bind to c-MYC, in order to test the conservation of binding to a second site in the bHLHZip domain of MYC [28,30,38] (Figure 1).
For all SPR binding measurements the compounds were injected at increasing concentrations. After protein immobilization on the CM5 chip surface most of the c-MYC protein appeared to be active, since the expected maximal response (R max , the binding signal at saturation) was reached after injection of increasing concentrations of 10058-F4 ( Figure S3). However for MYCN, only one fourth of the theoretical R max was reached, indicating that not all protein molecules were able to bind to the analytes after immobilization ( Figure S3). However, despite some of the MYCN protein being inactive, increased binding of the molecules was still detected in a dose-dependent manner and K D values could be calculated for most of the compounds (Table 1, Figure 2 and Figure S3). Surprisingly, the obtained R max values for C-m/z 232 to both c-MYC and MYCN were twice as high as those for 10058-F4, and double those of the theoretical R max value for a single site binding to c-MYC, thus suggesting a possible second binding site. The analog #764 as well as 10074-G5 showed especially poor solubility in aqueous buffers and could not be analyzed at concentrations above 50 mM. Hence the R max for c-MYC and MYCN could not be obtained for these molecules. Some unspecific binding, which was visible in the sensorgrams by a continuous, slightly upward trend of the curves, especially at higher concentrations, was detected for 10058-F4, C-m/z-232 and #474 ( Figure S3), probably reflecting weak non-specific binding to other sites in c-MYC [30,32] and MYCN, respectively. As expected, the 10058-F4 analog 7RH previously reported not to bind to c-MYC [39], was also a non-binder in our assays and was thus regarded as a negative control ( Figure S3).
In order to determine the affinity of the compounds to c-MYC and MYCN, the response for each small molecule was plotted against the analyte (compound) concentration. The K D was defined as the concentration of analyte at 50% of the experimental maximal Response (R max ). Approximate K D -values could be determined for all compounds except for #764. Similar affinities were determined for binding of 10058-F4 to both c-MYC and MYCN, with a mean K D value of 39.768.1 mM and 41.9610.6 mM, respectively (Table 1 and Figure 2). The affinities determined for the potential metabolite, C-m/z 232, were the lowest of all the molecules tested with a mean K D of 82.365.5 mM for c-MYC and a mean K D of 87.164.9 mM for MYCN. Compared to 10058-F4 the analog #474 analog showed an enhanced affinity both to c-MYC (mean K D value of 16.661.4 mM) and to MYCN (mean K D = 26.968.1 mM). Importantly, we demonstrated binding of the structurally unrelated compound 10074-G5 to both MYC proteins, and determined the affinity to both MYCN (K D = 19.2611.5 mM) and to c-MYC (K D = 31.7624.9 mM). Although the sensorgrams for #764 indicated binding to both proteins ( Figure S3), we were not able to acquire data reliable enough to calculate its K D . As expected, binding of 7RH could not be detected to either of the two MYC proteins, even at the highest concentrations used ( Figure S3). In summary, all active c-MYC inhibitors tested were also found to bind directly to MYCN.

Targeting the MYCN/MAX interaction in NB cells
In our previous study we observed a reduction in MYCN/MAX interaction after treatment with 10058-F4 [40]. To investigate if the other small molecules also have an effect on MYCN/MAX dimerization we performed a proximity ligation assay (PLA) using MYCN-amplified SMS-KCN69n NB cells. Compared to DMSO treated cells, 10058-F4, its analogs #474 and #764 as well as the structurally unrelated compound 10074-G5 reduced the MYCN/ MAX interaction following 6 hours treatment ( Figure 3). In contrast, the metabolite C-m/z 232 and the non-binder 7RH did not affect MYCN/MAX dimer formation ( Figure 3). No signal was detected in any of the experimental negative controls ( Figure  S4). In summary, all molecules that compared to 10058-F4 showed approximately similar or higher affinity to MYCN were also able to reduce the interactions between MYCN and MAX in NB cells.

Decrease of MYCN protein levels after treatment with small molecules
It has previously been reported that 10058-F4 treatment leads to down-regulation of c-MYC protein levels in c-MYC overexpressing cell lines and we observed similar results for MYCN levels in MYCN-amplified NB cells [33,34,40]. In order to determine the effect of the other compounds analyzed in this study on the MYCN protein levels, SK-N-BE(2) NB cells were treated with as a high concentration of each compound as possible without resulting in excessive cell death. While 10074-G5 decreased the MYCN protein levels at lower concentrations than 10058-F4, the putative metabolite C-m/z 232 showed only a slight reduction in the MYCN levels at high concentrations, and the 10058-F4 analogs #474 and #764 did not cause any changes in the MYCN protein levels at the concentrations which were possible to use without apoptosis induction ( Figure 4). In summary, only 10058-F4 and 10074-G5 were able to efficiently decrease the MYCN protein levels.

Effect of small-molecules on growth inhibition and apoptosis in MYCN-amplified cells
The small molecules 10058-F4, 10074-G5, as well as #474 inhibit cell proliferation and induce apoptosis in a variety of c-MYC expressing cells [28,33,34,39], and we have shown that 10058-F4 induces apoptosis and inhibits cell growth in MYCN overexpressing NB cells [40]. Here we examined the effect on cell growth (and apoptosis) of the other compounds in MYCNamplified Kelly NB cells. As shown in Figure 5A, all molecules tested inhibited cell growth in a concentration dependent manner. Compared to 10058-F4, the analog #474 and 10074-G5 were more potent whereas the metabolite C-m/z 232 caused less growth inhibition ( Figure 5A). Importantly, the calculated IC 50 values from the cell growth inhibition assays and the K D -values from the SPR experiment for binding to MYCN strongly correlated to each other ( Figure 5B).
We next investigated the level of apoptosis induction in response to treatment with the molecules using two different MNA NB cell lines ( Figure 5C, D). As previously shown, Kelly cells were more sensitive to 10058-F4 treatment than the SK-N-BE(2) cells [40]. This effect was also seen after treatment with 10074-G5, while the 10058-F4 analogs #474 and #764 were equally efficient in inducing apoptosis in the two cell lines ( Figure 5C, D). The 10058-F4 analogs #474 and #764 as well as 10074-G5 were more potent than 10058-F4 in inducing cell death in both cell lines analyzed ( Figures 5C, D). To verify the MYCN specificity in apoptosis induction, we next used Tet21N NB cells with doxycycline regulated MYCN expression. As shown previously [40], cells with low MYCN-levels were less sensitive to 10058-F4 treatment compared to cells with high MYCN protein levels. The MYCN dependent induction of apoptosis was also observed after treatment with 10074-G5, #474 and #764. In contrast, no apoptosis was observed after incubation with C-m/z 232 independently of MYCN status ( Figure S5). As previously reported for c-MYC expressing cells [39], the effect of #764 on apoptosis induction was similar or slightly stronger than that of #474 in Tet21N cells ( Figure 5C, D). In conclusion, the affinity of the molecules to MYCN correlated with their ability to inhibit cell growth and to selectively induce apoptosis in MYCN overexpressing NB cells ( Figure S5).

NB cells undergo neuronal differentiation in response to 10074-G5
MYCN counteracts neuronal differentiation whereas inhibition of MYCN promotes differentiation of NB cells [49,50]. We have previously shown that 10058-F4 stimulates NB differentiation, induces expression of the nerve growth factor (NGF) receptor TrkA and hence renders MYCN-amplified NB cells sensitive to NGF-mediated differentiation [40]. Similarly, treatment of SK-N-BE(2) cells with sub-lethal concentrations of 10074-G5 for 14 days also resulted in evident neurite outgrowth ( Figure 6). As previously shown for 10058-F4 treatment, NGF potentiated the differentiated morphology of 10074-G5 treated cells by increasing both the number and the length of neurite protrusions ( Figure 6). The same effect was observed after treatment with C-m/z 232, whereas incubation with neither #474 nor #764 resulted in morphological changes ( Figure S6).

Treatment of NB cells induces lipid accumulation
We have previously shown that 10058-F4 treatment as well as genetic down-regulation of MYCN using shRNA, leads to accumulation of lipid droplets in MYCN-amplified NB cells. We also demonstrated that 10058-F4 treatment results in downregulation of proteins involved in b-oxidation of fatty acids as well as in the respiratory chain, and that these changes are likely to play a role in the observed lipid accumulation [40]. To explore the effect on lipid accumulation of the small molecules studied here cells were treated with sub-lethal concentrations of the compounds and stained with Oil Red O in order to detect lipids. As seen in Figure 7, treatment with C-m/z 232 at concentrations under 80 mM did not resulted in lipid accumulation, whereas at 90 mM the amount of lipid droplets was similar to after incubation with 10058-F4. Interestingly, 10074-G5 treatment resulted in smaller but more numerous lipid droplets compared to 10058-F4 and Cm/z 232. We also observed lipid accumulation, although less pronounced, after treatment with #474, whereas #764 led to only a minor increase in the amount of lipid droplets (Figure 7). Taken together, we have demonstrated that although the analyzed compounds all bind to MYCN, they induce lipid droplet accumulation to different extents.

Discussion
To date several research groups have focused on developing compounds that target c-MYC for cancer therapy, whilst there have been only a few publications that have explored the possibility of targeting MYCN [27,28,39,51,52] The small molecule 10058-F4 has been extensively studied in the context of targeting c-MYC in cancer cells, although more information is needed regarding its true specificity [28,30,33,34,40]. Recently, we have shown that 10058-F4 also reduces MYCN/MAX interaction in addition to c-MYC/MAX and that it induces selective apoptosis and cell growth arrest in MYCN-amplified compared to non-MYCN amplified NB cells. Furthermore we demonstrated significantly delayed tumor growth in a NB xenograft and increased survival in a transgenic mouse model of NB [40]. These results are in contrast with a previous report that showed no significant antitumor activity of 10058-F4 in a xenograft model of prostate cancer [36]. The differences might be due to the targeting of MYCN as well as c-MYC by 10058-F4 in NB, as well as to a potential greater reliance of NB cells on MYCN for their survival.
Inspired by our earlier findings, we explored whether additional compounds shown to modulate c-MYC function could also bind MYCN and inhibit its function in NB cells. Here, extensive characterization of five compounds has been carried out including monitoring of their binding to MYCN as well as their effects in cellular assays used in our previous study.
Comparison of 10058-F4 and three closely related structural analogs Three of the selected compounds, known as #474, #764 and 7RH, are close structural analogs of 10058-F4 [39] (Figure S1). Both #474 and #764 have been previously reported to show improved interaction with c-MYC compared to 10058-F4, whereas 7RH does not bind to c-MYC and was therefore included as a negative control [39]. Our SPR analysis showed binding to both c-MYC and MYCN for all compounds tested except for the non-binder 7RH. The estimated K D values showed a similar affinity of 10058-F4 to both proteins. The approximate affinities we could determine were higher than the originally reported K D -values [30], which might be due to the fact that in the fluorescence polarization assay the bHLHZip domain is free in solution, while in the SPR-based assay it is immobilized on the chip surface by amine coupling and thus physically constrained. However we were encouraged by the fact that 10058-F4 also bound to MYCN with an approximately equal affinity as to c-MYC.
10058-F4 is proposed to bind preferentially to Tyr402 and the hydrophobic region between residues 401 and 406 in c-MYC [32]. As shown in Figure 1, MYCN also contains a Tyr at the analogous position (residue 429) as well as a highly homologous hydrophobic region (residues 428-433). This suggests that 10058-F4 can adopt a similar binding mode for MYCN leading to the equivalent binding of 10058-F4 to c-MYC and MYCN. Whereas it proved difficult to assess the binding of #764 to c-MYC or MYCN due to lack of compound solubility, our studies indicated that #474 bound more strongly to c-MYC than 10058-F4. As expected, 7RH did not bind to c-MYC or to MYCN.
10058-F4, #474 and #764 were also able to decrease the interaction of MYCN and MAX as analyzed in a proximity ligation assay, whereas 7RH did not. Both #474 and #764 were more potent than 10058-F4 in the growth inhibition and apoptosis assays. Importantly in terms of assessing compound specificity, 10058-F4, #474 and #764 all induced apoptosis more effectively in MYCN expressing compared to non-MYCN expressing Tet21N cells. Surprisingly, the analogs #474 and #764 did not cause any decrease in the MYCN protein levels and probably therefore also did not induce neuronal differentiation. In addition, #764 only induced a few and very small lipid droplets compared to 10058-F4 and 10074-G5 whereas #474 treatment resulted in very little lipid accumulation. Taken together, these experiments suggest that the full biological effects of MYC inhibiting compounds are most apparent not only when the binding between MYCN and MAX is inhibited but when the levels of MYCN protein are reduced as well. Residual MYC activity may thus be sufficient to permit continued mitochondrial function, thus preventing the accumulation of neutral lipids.  Analysis of the 10058-F4 metabolite C-m/z 232 In a previous study it was shown that C-m/z 232 is a major metabolite of 10058-F4 resulting from conversion of the thiocarbonyl group in 10058-F4 to the corresponding carbonyl group [36]. We compared 10058-F4 and C-m/z 232 across our series of assays to assess whether the activity observed for 10058-F4 in cells and in our animal model of NB could in fact be due to the presence of C-m/z 232. In general, C-m/z 232 retained some of the activities displayed by 10058-F4 but to a significantly reduced extent. More specifically, C-m/z 232, was able to bind to c-MYC and MYCN but with much lower affinity compared to 10058-F4, and whilst neuronal differentiation and the formation of lipid droplets were seen for cells treated with C-m/z 232, the phenotype was considerably less strong than that observed after 10058-F4 treatment.
Importantly, C-m/z 232 was not able to effect the MYCN/ MAX interaction in the proximity ligation assay even at a concentration of 112.5 mM. In addition, treatment of cells with Cm/z 232 only slightly reduced the MYCN protein levels at a high concentration, and it only caused low levels of growth inhibition and apoptosis. Furthermore, we did not observe any induction of apoptosis in MYCN or non-MYCN expressing Tet21N cells. Taken together these results suggest that when 10058-F4 is used in cells or animal models, the contribution of the metabolite C-m/z 232 to the outcome of the experiment is only minor.

Comparison of the 10058-F4 and 10074-G5 molecules
The 10074-G5 compound which is structurally unrelated to 10058-F4 has previously been reported to modulate c-MYC activity by preventing the c-MYC/MAX interaction. Importantly the regions of the bHLHZip domain of c-MYC that interact with these two molecules have been shown to be distinct [29,36] ( Figure 1). Here, we demonstrate that 10074-G5 also binds to MYCN. However due to the relatively large standard deviation associated with the data no more conclusions can be drawn other than that its affinity is approximate the same as that for 10058-F4.
In addition, a good correlation between the affinities of both 10058-F4 and 10074-G5 for MYCN and their inhibitory effects on MYCN-amplified NB cells was observed. Importantly, 10074-G5 reduced the MYCN/MAX interaction in NB cells as demonstrated by PLA. In general the observed activity of 10074-G5 across the wide range of cell-based assays was very robust, and consistent with its previously reported rapid intracellular accumulation during cell culture [53]. In fact, 10074-G5 mirrored the effects of 10058-F4 but was more effective at reducing MYCN protein levels and was more potent for both growth inhibition and   [52,53] and overall our data suggest that 10074-G5 is at least as promising as 10058-F4 for targeting of MYCN. Potential concerns with regard to a drug discovery however, relate to the presence of a nitro functional group and the benzofurazan core structure of this molecule.
In conclusion, we have carried out a comparison of several small molecules previously shown to bind directly to the bHLHZip of c-MYC in the context of direct binding to MYCN and modulation of MYCN activity. We found that all c-MYC binding molecules analyzed also bind to the corresponding region of MYCN and results in inhibitory activity against NB cells in vitro. Importantly our work highlights 10074-G5 as being of potential interest as a starting point for the development of next-generation inhibitors of the MYCN/MAX interaction.

Subcloning of the bHLHZip domain of human c-MYC and human MYCN
The gene fragments comprising human c-MYC residues 353-437 and human MYCN residues 380-464 were obtained by PCR amplification of cDNA encoding human c-MYC and MYCN. The forward primers used were 59GTTGTCATATGAATGTCAA-GAGGCGAACACACA39 and 59GTTGTCATATGAGT-GAGCGTCGCAGAAACCACA39 as well as the reverse primers 59CAGCTACGGAACTCTTGTGCGTAA39 and 59TTGAA-CACGCTCGGACTTGCTAG39 for c-MYC and MYCN, respectively. The PCR fragments were cloned into the pET28a vector (Novagen) under the control of a T7 promotor and introducing an N-terminal hexa-histidine tag after digestion with NdeI and BamHI. The identity of the plasmids was verified by sequencing.   Targeting of MYCN with c-MYC Inhibitors PLOS ONE | www.plosone.org coli BL21 (DE3) bacteria (Stratagene) at 37uC in LB media with kanamycin. Cultures were induced with Isopropyl b-D-1-thiogalactopyranoside (0,1 mg/ml) at an OD of 0.6 and harvested after 4 hours. After sonication of the proteins in 50 mM NaH 2 PO 4 , 100 mM NaCl and 10 mM Tris-HCl (pH = 8,0) for 6630 seconds, the pellet was resuspended in 100 mM NaH 2 PO 4 , 10 mM Tris-HCl, 10 mM imidazole and 8 M urea (pH = 8,0) prior to additional sonication for 6630 seconds. The proteins were then loaded onto a Ni-Sepahrose High Performance His-Trap column (GE Healthcare), washed with ten column volumes of 100 mM NaH 2 PO 4 , 10 mM Tris-HCl, 40 mM imidazole and 8 M urea (pH = 8,0) and finally eluted with two column volumes 100 mM NaH 2 PO 4 , 10 mM Tris-HCl, 500 mM imidazole and 8 M urea (pH = 8,0). Fractions without visible higher molecular mass contaminations, as judged by SDS-page, were pooled and refolded by dialysis in 50 mM NaH 2 PO 4 , 100 mM NaCl, 10 mM Tris-HCl (pH = 8,0), 5 mM EDTA, 2 mM DTT and decreasing concentrations of urea (6 M, 4 M, 2 M and 0 M). Finally the buffer was exchanged to either PBS (Hyclone) for use in circular dichroism measurements or 20 mM Na 2 HPO 4 *2H 2 O, 300 mM NaCl, 4 mM KH 2 PO 4 , 0,05% Surfactant P20 (GE Healthcare) and 5% DMSO (pH = 7,5) for use in Surface Plasmon Resonance analysis. Afterwards the proteins were concentrated up to a concentration of 5 mg/ml. Protein concentrations were determined by SDS-page comparing purified proteins to a standard curve of BSA loaded on the gel.

Circular Dichroism
The wavelength scan were performed with a protein concentration of 1 mg/ml in PBS at 20uC in a 1 mm path-length cell with a JASCO J-810 spectrometer equipped with a thermoelectric temperature controller. The residual secondary structure of the bHLHZip of c-MYC and MYCN was estimated by means of spectral deconvolution with CONTIN, SELCON and K2D on the Dichroweb server using reference sets 4 and 7. Reported values represent the average result of CONTIN, SELCON and K2D algorithms in percent as well as the relative standard deviation.

Surface Plasmon Resonance
Surface Plasmon Resonance (SPR) experiments were performed using a Biacore T200 (GE Healthcare) instrument at 25uC. The proteins were immobilized on a CM5 (GE Healthcare) chip through amine coupling using the amine coupling kit (GE Healthcare) resulting in immobilization levels between 1900-2100 RU. For each immobilization a new purified batch of protein was used. The samples were flowed over the surface with 30 ml/min for 60 seconds with a regeneration time of 600 seconds. The experiments were performed in the same batch of running buffer as used for dialysis. After each injection the flow delivery system was washed with 50% DMSO to avoid possible binding of molecules. To remove all remaining analyte the surface was regenerated with 10 mM NaOH. The obtained data was analyzed by the Biacore T200 Evaluation Software 2.0 (GE Healthcare). R max is the binding signal at saturation. The K D -values were obtained from steady state fitting of equilibrium curves by plotting the response against the concentration of the analyte (compound). The K D value is calculated as 50% of the experimental R max . The theoretical Rmax = (MW analyte/MW ligand) x immobilized ligand level on the chip (RU).

Cell culture
Human MYCN-amplified neuroblastoma SK-N-BE(2) and SMS-KCN69n cells were grown in MEME:F12-Ham medium (1:1, v/v) supplemented with 10% FBS, 1% glutamine, 1% non-essential amino acids, and 1% penicillin/streptomycin. Human MYCN-amplified neuroblastoma Kelly cells were grown in RPMI 1640 medium supplemented with 10% FBS, 1% glutamine, and 1% penicillin/streptomycin. Tet21N cells derived from the human non-MYCN amplified NB cell line SH-EP, contain a repressible (Tet-Off) MYCN gene and were cultured as previously described [51]. Doxycycline was added as indicated to repress MYCN expression at a final concentration of 1 mg/ml. Proximity Ligation Assay (PLA) 40610 3 SMS-KCN69n cells were seeded on collagen coated coverslips and grown for 24 hours and treated for 6 hours the next day. After 6 hours the cells were washed in PBS and treated as previously described [40]. The PLA assay was performed after manufacturer's instructions with the Duolink kit (Olink, Uppsala, Sweden). The cells were incubated with primary antibody against MYCN (NCMII100, St Cruz Bio) 1:500 and Max (C-17, St Cruz Bio) 1:500 for 1 hour in a humidity chamber at 37uC, which were diluted in antibody diluent provided by the manufacturer. MYCN (1:500) and FAS (A-20, St Cruz Bio) 1:500 as well as MAX (1:500) and GFP (C-2, St Cruz Bio) 1:500 antibodies were used in combination in order to test for any unspecific binding of the secondary antibodies or unspecific signals. All of the following procedures were performed according to manufacturer's instructions. During the amplification step an additional secondary antibody (Alexa fluor 568, goat anti-mouse; Invitrogen) was added 1:1000 to the amplification solution to counterstain for MYCN protein.

Microscopy platform
The confocal images of the cells were acquired at 636 magnification using a motorized axiovert fluorescence microscope 200 M LCI (Zeiss GmBH, Göttingen, Germany) including a CSU10 Yokogawa head; a multilens/multipinhole array aperture (Yokogawa, Japan), an ORCA ER cold CCD camera; detector array 134461024px (Hammamatsu, Japan) and two continuous wave diode-pumped solid-state (DPSS) lasers for fluorescence excitation (Cobolt, Sweden); Cobolt Calypso 491 nm DPSS (excitation of green PLA fluorophore) and Cobolt Jive 561 nm DPSS (excitation of Alexa Fluor 568). Emission filters 525/50 and 607/45 were used respectively. Open lab software (Perkin Elmer, USA) was used to navigate the microscope. Image capture parameters were set to pixel binning 1, the space between each z-layer to 0.6 mm and the z-depth of each capture was on average 25 layers per picture. A total number of 5 to 10 confocal image stacks were captured per compound and per experiment. A total amount of at least 120 cells were analyzed for all compounds. The PLA signals were analyzed using the open source image analysis program Image J (NIH, Bethseda, USA) and the amount of PLA signal per cell was calculated. Two persons analyzed the images independently in a blind fashion.

Image analysis
In order to quantify the PLA signal from the total z-depth of each cell confocal z-stacks was projected in z and identically processed regarding background subtraction. The PLA signal was then quantified blindly and automatically using a find maxima-based algorithm in ImageJ (plugin routine developed by Dr. E. Flaberg and Professor L. Szekely, KI and available upon request). The images for visualization of the results were identically processed regarding background subtraction.

Immunoblotting
Whole cell extracts were prepared and lysed using RIPAbuffer. Western blot analysis was performed as previously described [54]. In brief, membranes were probed with a mouse anti-MYCN antibody (B8.4.B, St Cruz Bio), MAX antibody (C-2, St Cruz Bio) and an HRP-conjugated anti-mouse antibody (DAKO) was used as secondary antibody. The membranes were developed using enhanced chemiluminescence substrate (ECL, Amersham). The membranes were subsequently re-probed with antibodies specific for GAPDH (6C5, St Cruz Bio) in order to confirm equal loading.

Apoptosis detection
Small molecule-induced apoptosis was scored by quantifying the sub-G0/G1 fraction of the cell cycle. Treated neuroblastoma cells, were fixed in 70% ice-cold methanol and stored at 220uC. Cells were stained with propidium iodide solution (5 mg/ml propidium iodide and 25 mg/ml RNase A in PBS) for 30 min at 37uC and analyzed in a FACScan flow cytometer (Becton Dickinson). Apoptotic cells were identified within the PI-stained population by virtue of exhibiting an apparent sub-diploid DNA content. Tet21N cells were pre-treated with doxycycline for 4 days to turn MYCN off or grown without doxycycline. Both cell sets were incubated with the respective compounds followed by FACS analysis.

Crystal violet staining
Cell viability was assayed by crystal violet staining. Kelly cells treated for 48 h with the indicated compound or DMSO were fixed in 1% trichloroacetic acid (TCA) (Merck-Schuchardt) for 1 h in 4uC and stained with 0.04% crystal violet (Sigma) for 30 min at room temperature. Following washing of unbound stain in water, the bound crystal violet was dissolved in 1% SDS. The absorbance was measured at 570 nm using 650 nm as reference wavelength. The IC 50 was determined after scaling the percentage of viable cells in log using a nonlinear regression with graph pad prism 6.

Neuronal differentiation of NB cells
SK-N-BE(2) cells were grown for 15 days in 6 well-plates and treated twice per week with DMSO or compound, with and without NGF. Differentiation was assessed based on number and length of neuronal outgrowths using a phase contrast microscope (Axiovert 40 CFL, Zeiss) and the Axiovision Rel. 4.8 software.
Oil Red O staining of lipids SK-N-BE(2) cells grown on cover slips were fixed in 4% PFA, washed in PBS and then in 60% isopropanol before staining with filtered Oil Red O solution (2 mg/ml) in 60% isopropanol. Cells were washed briefly in distilled water and mounted onto glass slides using water-soluble mounting medium (Aqua Pertex, Histolab products AB). For visualization, bright field images were captured at 40X magnification using an Axiovert 40 CFL inverted fluorescence microscope (Zeiss) and Axiovision Rel. 4.8 software.  bHLHZip domains of c-MYC and of MYCN into the pET28a expression vector, to Dr. E. Flaberg for expert analysis of the PLA slides, to Professor M. Schwab (Cancer Research Center, Heidelberg, Germany) for the Tet21N cells and all present and past members of the Arsenian Henriksson laboratory for helpful discussions. We acknowledge GE Healthcare for a fruitful collaboration and access to the Biacore T200.

Supporting Information
We dedicate this article in the memory of our friend and colleague Dr. Anna McCarthy, a dedicated chemist who generously shared expert advice. She tragically passed away on January 31 st 2014 after battling with acute illness.