The MNT transcription factor autoregulates its expression and supports proliferation in MYC-associated factor X (MAX)-deficient cells

The MAX network transcriptional repressor (MNT) is an MXD family transcription factor of the basic helix-loop-helix (bHLH) family. MNT dimerizes with another transcriptional regulator, MYC-associated factor X (MAX), and down-regulates genes by binding to E-boxes. MAX also dimerizes with MYC, an oncogenic bHLH transcription factor. Upon E-box binding, the MYC–MAX dimer activates gene expression. MNT also binds to the MAX dimerization protein MLX (MLX), and MNT–MLX and MNT–MAX dimers co-exist. However, all MNT functions have been attributed to MNT–MAX dimers, and no functions of the MNT–MLX dimer have been described. MNT's biological role has been linked to its function as a MYC oncogene modulator, but little is known about its regulation. We show here that MNT localizes to the nucleus of MAX-expressing cells and that MNT–MAX dimers bind and repress the MNT promoter, an effect that depends on one of the two E-boxes on this promoter. In MAX-deficient cells, MNT was overexpressed and redistributed to the cytoplasm. Interestingly, MNT was required for cell proliferation even in the absence of MAX. We show that in MAX-deficient cells, MNT binds to MLX, but also forms homodimers. RNA-sequencing experiments revealed that MNT regulates the expression of several genes even in the absence of MAX, with many of these genes being involved in cell cycle regulation and DNA repair. Of note, MNT–MNT homodimers regulated the transcription of some genes involved in cell proliferation. The tight regulation of MNT and its functionality even without MAX suggest a major role for MNT in cell proliferation.

Consistent with MNT functioning as a MYC transcriptional antagonist, enforced MNT expression inhibits cell proliferation and impairs MYC-dependent transformation (5,11). The deficiency or down-regulation of MNT in fibroblasts leads to increased proliferation (i.e. similarly to MYC overexpression) and partially rescues the proliferative arrest caused by MYC deficiency (10,11,16,17). MNT ablation in vivo leads to breast and T-cell tumors (10,12,17), and according to the Cancer Genome Atlas, about 10% of human tumors show deletion of a MNT allele (1).
Partial or total MNT deficiency in mouse models impairs MYC-dependent tumorigenesis (18,19), and MNT knockout in some cell models inhibits proliferation and promotes apoptosis (10,16,17). Thus, MYC and MNT proteins co-exist in proliferating cells, and depending on the model, MNT acts as a MYC antagonist or a cooperator of MYC.
However, there is scarce information about MNT transcriptional regulation, and it is unknown whether MNT exerts functions without MAX. In this work, we studied possible MAXindependent functions of MNT using UR61 cells as the main model. These rat pheochromocytoma cells do not express a functional MAX protein but a truncated form (termed MAX PC12 ) that lacks the second helix and leucine zipper region of the bHLH-LZ domain, which are the regions responsible for dimerization with MYC and MNT (20). Here, we describe a change of MNT subcellular localization depending on MAX expression and the repression of its own promoter in the presence of MAX. In addition, we show the first examples of MNT functions that are independent of MAX. In the absence of MAX, 1) MNT regulates gene transcription by binding to DNA through the formation of MNT-MNT or MNT-MLX complexes, and 2) MNT is required for optimal cell proliferation.

MNT levels depend on MAX
To explore the effect of MAX on MNT expression, we first compared MNT and MAX levels in proliferating cells from 13 cell lines derived from different tissues and species, including two cell lines lacking MAX: UR61 and the human small-cell lung carcinoma H1417 cells (21). The results showed that although MNT expression varies among the cell lines, the two MAX-deficient cell lines and others with low MAX levels (HeLa and CEM) expressed high MNT levels, whereas other cells with high MAX levels expressed low MNT levels (293T, K562, and Ramos) (Fig. 1A). MNT was expressed in all cell lines as a protein doublet, due to a slower-migrating phosphorylated MNT form (22). MAX was also expressed as a doublet of 21 and 22 kDa (23). Although the correlation was not universal, the high MNT expression in some cell lines with null or low MAX levels led us to explore whether MAX influenced MNT levels.
For this purpose, we transfected UR61 cells with a construct carrying human MAX cDNA driven by the metallothionein promoter, which is activated by Zn 2ϩ cations (24). Several clones were isolated, and two of them with robust MAX induction were mixed, and the resulting cells were termed URMax34. We also generated a cell line transfected with the empty vector, termed URMT, which is a pool of five transfected clones. The induction of MAX in response to Zn 2ϩ in URMax34 cells was confirmed by immunoblot (Fig. 1B). We examined the effect of MAX induction on MNT levels in URMax34 cells, and we found that MNT was down-regulated upon MAX induction by Zn 2ϩ (Fig. 1B). As expected, the treatment of URMT cells with Zn 2ϩ did not change the MNT levels (Fig. S1A). To confirm this result and rule out effects potentially related to the generation of stably transfected clones (as the URMax34 system), UR61 cells were transiently transfected with a MAX expression vector, and the results showed a decrease in MNT protein levels in MAX-transfected cells (Fig. 1C). It is noteworthy that the down-regulation of MNT provoked by the re-expression of MAX in these cells also occurred at the mRNA levels, as determined by RT-qPCR (Fig. 1D). We then analyzed the effect of MAX-enforced expression in three human small-cell lung cancer cell lines deficient in MAX (21). MAX-enforced expression was achieved by lentiviral transduction, and the levels of MNT were examined by immunoblot in Lu134, Lu165, and H1417 cell lines (Fig. 1E). In all cell lines, MAX expression resulted in lower MNT protein levels. The MNT mRNA levels were also down-regulated in Lu165 and in H1417 cells upon MAX expression (Fig. 1F).
We next sought to confirm this in a different cell type with endogenous MAX. A MAX expression vector was transfected into human myeloid leukemia K562 cells. The immunoblot results showed lower MNT levels in the cells with MAX overexpression (Fig. 1G). Next, we used the Kmax12 cell line (24), a K562 derivative carrying a MAX transgene in which expression is induced by Zn 2ϩ . Induction of MAX expression in Kmax12 cells resulted in a concomitant MNT down-regulation (Fig.  1H). We also used the opposite approach, i.e. depleting cells of MAX and analyzing the expression of MNT. As shown in Fig.  1I, MNT protein expression was up-regulated when MAX was silenced with siRNA in K562. Interestingly, MNT was also upregulated in MAX-silenced cells at the mRNA level (Fig. 1J). Altogether, the results showed that low MAX levels result in MNT up-regulation at the mRNA and protein levels.

MNT binds and represses its own promoter in the presence of MAX
The above results showing that MNT down-regulation took place at the mRNA level prompted us to investigate whether MNT impairs the activity of its own promoter. Bioinformatic analysis of human, mouse, and rat MNT promoter regions revealed that there are two E-box sequences within 1 kb upstream from the transcriptional start site of MNT (one canonical E-box, CACGTG (E-box 1) and one noncanonical CATGTG (E-box 2)) that are conserved among these three different species (Fig. 2A). We then constructed a luciferase reporter carrying the 850-bp upstream region of the transcription start site from the human MNT gene. The construct was termed MNT-Luc (Fig. 2B). HEK293T cells (which express MAX) were transfected with the MNT-Luc and MNT expression vectors (or the corresponding empty vectors). The results showed that MNT overexpression led to a reduction in the luciferase activity (Fig. 2C, left panel), suggesting that MNT-MAX negatively regulates the MNT promoter. To determine the contribution of the two E-boxes in the MNT-mediated negative autoregulation, we constructed two reporters containing each of the E-boxes, termed E-box 2 MNT-Luc (containing the last 220 bp of the MNT-Luc reporter, which includes E-box 2) and E-box 1 MNT-Luc (containing the first 570 bp of the MNT-Luc reporter which includes E-box 1) (Fig. 2B). The luciferase assays in HEK293T cells suggested that the MNT promoter down-regulation depended on the E-box 2 (Fig. 2C,  left panel).
We also investigated the activity of the MNT promoter in UR61 cells, which do not express MAX. UR61 cells were transfected with the MNT-Luc vector together with MNT and MAX expression vectors. The results also showed a decrease in the luciferase activity although less than in HEK293T cells ( Fig.   Figure 1. High MNT expression in MAX-deficient cells. A, cell lysates of the indicated cell lines were analyzed by immunoblot to determine the levels of MNT and MAX. ␤-Actin levels were determined as a protein-loading control. The MAX-deficient cell lines analyzed were UR61 (rat pheochromocytoma) and H1417 (human small cell lung carcinoma). The rest are MAX-expressing cells: HEK293T (human embryonic kidney, 293T); K562 (human chronic myeloid leukemia); HeLa (human cervical cancer); Ramos (human B-cell lymphoma); SH-SY5Y (SHSY, human neuroblastoma); T98G (human glioblastoma); mouse embryonic fibroblasts; Neuro-2a (N2a, mouse neuroblastoma); C6 (rat glioma); Jurkat (human T lymphoma); and CEM (human T leukemia). B, control URMT and URMax34 cells were treated for 24 h with 50 and 100 M Zn 2ϩ , and the MNT and MAX protein expression was determined by immunoblot. ␣-Tubulin levels were determined as a protein-loading control. C, levels of MNT and MAX determined by immunoblot in UR61 cells 24 h after transfection with a MAX expression vector or the empty vector pCEFL (EV). D, mRNA expression determined by RT-qPCR in URMT and URMax34 cells treated for 24 h with 100 M Zn 2ϩ . Data represent the mean Ϯ S.D. from three independent experiments; **, p Ͻ 0.05. E, protein levels of MNT by immunoblot in Lu165, Lu134, and H1417 cells. The lysates were prepared 72 h after infection with MAX-expressing lentivirus or empty lentivirus (EV).

MNT functions in the absence of MAX
2C, right panel). The expression of MNT alone also led to a decrease in the luciferase activity, suggesting that MNT can down-regulate the MNT promoter in UR61 cells in the absence of MAX. The repressive effect of MNT was stronger in HEK293T cells than in UR61 cells, which may be explained by the limited overexpression of MNT protein achieved in transfected UR61 cells (shown below). Co-transfection of MNT and MAX resulted in a decrease in the activity of both promoter constructs (MNT-Luc and E-box 2 MNT-Luc). We next constructed a reporter with a deletion of the Ϫ788 E-box (⌬E-box 2 MNT-Luc), and the results showed that in UR61 cells MNT had no effect on the activity of the mutant reporter (Fig. 2C, right panel), confirming that E-box 2, mapping at Ϫ788, was critical for MNT-mediated down-regulation of its own promoter.
Because MNT-MAX bind E-boxes in the promoters to repress transcription, we analyzed the ChIP-seq data published by the ENCODE project. The data revealed two regions bound by MAX in the human MNT promoter that encompass the two E-boxes (Fig. 2D). Interestingly, according to ENCODE, ChIPseq for MNT also showed that MNT binds to the region encompassing E-box 2 in human K562 cells (Fig. 2D). We also analyzed the ChIP-seq peaks for MAX, MYC, and MXI1 proteins on the MNT promoter as published by the ENCODE project. We looked at other human cell lines like H1ES, HeLa, NB4, A549, GM78, HEPG, SKSM, and IMR90. The analysis revealed that the three proteins presented peaks at the same positions near the transcription start site of the human MNT gene (Fig. S1B), suggesting that members of the MYC and MXD family bind and possibly regulate the promoter of MNT.
As MNT was significantly down-regulated when MAX was re-expressed, we hypothesized that MNT-MAX heterodimers might bind to the MNT promoter and down-regulate its own expression. To explore this hypothesis, a ChIP assay was performed with anti-MNT and anti-MAX antibodies in URMT and URMax34 cells treated with Zn 2ϩ , which induces MAX in the URMax34 cells. We studied the Ϫ593 E-box of the rat MNT promoter, as well as a region of the FBXO32 (Atrogin-1) promoter as a positive control for MNT binding (25). A region mapping at 1369 bp upstream on MNT transcription start site with no E-boxes was used as negative control. As shown in Fig.  2E, MNT and MAX bound to the MNT promoter in URMax34 cells. However, in the MAX-deficient control URMT cells, the binding of MNT to its promoter was very weak, but above that of the IgG controls. This weak binding could be responsible for the decreased promoter activity detected in the results shown of Fig. 2C in MNT-transfected UR61 cells. MNT and MAX were bound to the positive control FBXO32/Atrogin-1 promoter in URMax34 but not in URMT control cells (Fig. 2E). These data show that MNT most likely binds to its own promoter as a heterodimer with MAX and suggest a possible negative regulation of MNT's own expression.
Then, to explore the binding of MNT to other regions of the genome in the absence of MAX, we performed ChIP-seq experiments with an anti-MNT antibody in MAX-deficient URMT cells. We sequenced the immunoprecipitated chromatin and the inputs of three independent experiments. The results showed first that MNT bound very weakly to the promoter of MNT in the absence of MAX, while stronger binding was detected in other genes like CCNG2 (Table S1). This indicates that MNT binds to DNA even in the absence of MAX. The peaks of MNT bound to the regulatory regions of some of these genes (MNT, FBXO32/Atrogin-1, CCNG2, CDK12, and ERCC6) are shown on Fig. S2A. To evaluate the possible transcriptional effect of MNT binding to these genes, we knocked down MNT in URMT and URMax34 cells through short-hairpin constructs and checked their expression by RT-qPCR. The results showed that upon MNT knockdown, the mRNA expression of these genes increased (Fig. S2B), indicating that MNT binds to their promoters to repress their expression even in the absence of MAX.
Ontology analysis of the genes obtained in our ChIP-seq experiments reveals that MNT-bound genes are involved in cell cycle, DNA replication, and regulation of gene expression (Fig.  S3A). We then analyzed whether there were any DNA motifs that were over-represented in the regions bound by MNT. The results revealed that MNT binds to regions with E-boxes in the absence of MAX but also to regions containing DNA-binding motifs for other transcription factors such as forkhead factors, SMAD, VDR, and TBXT (Fig. S3B).

MNT knockdown impairs cell proliferation in MAX-deficient cells
We next asked for a possible biological effect of MNT on cell proliferation in the UR61 MAX-deficient cells. We decided to knock down MNT through siRNAs. For this, we used expression vectors for two short-hairpin constructs against the rat MNT gene. These shMNT constructs efficiently reduced MNT levels and were termed shMNT-1 and shMNT-2 (Fig. 3A). The vectors also carried a puromycin-resistance gene. UR61 cells were transiently co-transfected with the shMNT constructs (or the empty vector) and a GFP expression vector in a proportion of 5:1 (shMNT/GFP vector) to ensure that the GFP-positive cells also incorporated the shMNT plasmid. Six days after transfection, the fraction of GFP-positive cells was analyzed by flow cytometry. The results showed that the fraction of GFPpositive cells was clearly reduced in cells with depleted MNT (Fig. 3B), suggesting that MNT loss resulted in impaired cell proliferation. In a second approach, we transfected the UR61 cells with the shMNT constructs and counted viable cells after 3, 7, and 12 days of transfection. As shown in Fig. 3C, MNTsilenced UR61 cells grew slower than controls. We also performed clonogenic assays in UR61 transfected with the shMNT and/or a MAX expression vector (in 1:3 proportions to ensure that the shMNT-containing cells had also incorporated the MAX vector) as well as the empty vectors. Then, 24 h after transfection puromycin was added and after selection, the colonies were stained with crystal violet, and the dye was solubilized and quantified. The data showed that MNT depletion provoked a dramatic growth inhibition (Fig. 3D). We conclude that MNT depletion impairs cellular proliferation in MAX-deficient cells. This was a striking result as it demonstrates a MAX-independent biological function of MNT. We also observed that UR61 cells overexpressing the MAX protein grew slower than the control (Fig. 3D), confirming the effect previously reported for the MAX-deficient PC12 pheochromocytoma cells (UR61 parental cells) (20). In addition, when MNT silencing was accompanied by MAX-enforced expression, the inhibition of UR61 cell proliferation was stronger (Fig. 3D). Consistent with the anti-proliferative effects of MNT depletion, we failed to generate stable MNT-silenced UR61 cell lines. To generate UR61 cells with MNT overexpression, we transfected UR61 with MNT expression vectors, but we were not able to detect a significant increase in MNT protein levels upon transfection. However, MNT protein was readily detected by immunoblot when the transfected cells were treated with the proteasome inhibitor bortezomib (Fig. S4A).
This could be explained if in the MAX-deficient cells the MNT levels are already high and enforced expressed MNT is degraded by the proteasome. In fact, some increase in the levels of MNT was detected when MNT was transfected in MAXexpressing URMax34 cells (Fig. S4B) or when MNT was cotransfected with MAX in UR61 cells (Fig. S4C).
Because the depletion of MNT impairs cell proliferation of UR61 cells, we wondered whether apoptosis is involved in this process. We found that the fraction of cells with a sub-G 0 amount of DNA was higher in cells transfected with the shMNT vector (Fig. 3E) and an increase in annexin V-positive MNT functions in the absence of MAX cells (Fig. 3F). The levels of cyclin A (a marker of cell proliferation) and survivin/BIRC5 (a marker of both proliferation and apoptosis) in URMT and URMax34 were analyzed by immunoblot. The results showed a decrease in survivin and in cyclin A in cells with depleted MNT, both in MAX-deficient cells (URMT) and in MAX-expressing cells (URMax34) (Fig. 3G). However, we did not detect cleavage of PARP1, a marker of apoptosis (data not shown). These results suggest that the depletion of MNT leads to cell proliferation arrest even in the absence of MAX, which can be partially due to the induction of apoptosis. In addition, restoration of MAX expression also impairs proliferation The results are summarized in Fig. 3H.

MNT regulates gene expression in MAX-deficient cells
Because the data shown above indicated that MNT knockdown in UR61 cells leads to growth arrest, we set out to study whether MNT regulates the transcriptional program in the absence of MAX. For this purpose, we determined the transcriptomes upon MNT depletion in cells with and without MAX. URMT and URMax34 cells were transfected with the short-hairpin RNA construct against the rat MNT gene (shMNT-1) or the empty vector as control in two different biological replicates. We first checked the depletion of MNT mRNA in the two replicates used for RNA-seq upon transfection of the shMNT-1 construct in URMT and URMax34 cells. URMT and URMax34 cells were transfected with the shMNT-1 construct, and 2 days after transfection, cells were treated with Zn 2ϩ for 24 h. The results indicated the induction of MAX mRNA expression in URMax34 as well as significant MNT mRNA depletion in the two independent transfections (Fig.  4A). These RNA preparations were submitted for next generation sequencing, RNA-seq . The data were then processed bioinformatically as described under "Experimental procedures" to obtain the expression values of each experimental replicate. For all of the different comparisons, we grouped the genes that were up-regulated and down-regulated in both biological replicates considering a log2 RPKM fold change higher than 0.7 or smaller than Ϫ0.7 (1.6ϫ or 0.6ϫ fold change, respectively) of the corresponding control and a p value Ͻ0.1. The heat maps of the gene expression signatures clearly showed that the depletion of MNT in both URMT and URMax34 induced gene expression changes (Fig. S5) indicating that MNT can be involved in transcriptional regulation without the concourse of MAX. Specifically, 281 genes were regulated upon MNT depletion in MAX-deficient URMT cells (Table S2) and 537 genes in MAX-expressing cells (i.e. URMax34 treated with Zn 2ϩ ) (Table S3). Of those, ϳ62% of the genes were downregulated upon MNT depletion in both cell lines (Fig. 4B). In the presence of MAX (URMax34 cells), MNT regulates more genes than in the absence of MAX (URMT cells): 537 versus 281 genes (Fig. 4B). In addition, the URMax34 versus URMT gene expression signature heat maps also showed that the expression of MAX induced gene transcriptional changes independently of MNT depletion (Fig. S5). Specifically, 377 genes were found to be differentially expressed when comparing URMax34 and URMT cells (Table S4) and 361 genes when comparing URMax34 and URMT cells depleted of MNT (Table S5). Among MAX-regulated genes, roughly half of the genes were down-regulated (Fig. 4B). Importantly, the comparison between the lists of differentially expressed genes upon MNT depletion in cells without MAX (URMT) and with MAX (URMax34) revealed 158 shared genes (27% up-regulated and 73% down-regulated) in the two biological replicates (Fig. 4C). These common genes are listed in Table S6. Thus, 56% (158/ 281) of the genes regulated by MNT in URMT (without MAX) cells are also regulated in URMax34 cells (when MAX is expressed). However, the 30% (158/537) of the genes regulated in URMax34 cells (with MAX) are also regulated in the MAXless URMT cells (Fig. 4C). Furthermore, we generated a dot plot graph with the RPKM values of the differentially expressed transcripts obtained in our RNA-seq data in URMT and URMax34 cells upon MNT knockdown (Fig. S6). The graph

MNT functions in the absence of MAX
showed that the transcriptional changes induced by MNT depletion followed the same pattern in the URMT and URMax34 cells suggesting that the presence of MAX did not affect dramatically the direction of MNT-transcriptional regulation in the UR61 cells. The lists of differentially-regulated genes upon MNT knockdown were compared with the gene sets derived from the biological process gene ontology (based on MsigDB platform, http://software.broadinstitute.org/gsea/ msigdb), 11 and the comparison showed the enrichment in cell cycle-related pathways in both cell lines (Fig. S7A). In addition, the analysis with the Gene Set Enrichment Analysis (GSEA) platform showed that the gene expression signature in URMT cells upon MNT depletion was associated with the cell cycle as the two most enriched pathways (Fig. 4D). A list of these genes is shown in Fig. S7B. Altogether, the functional analysis of the gene transcriptional signatures is consistent with the inhibition of cell proliferation elicited by MNT silencing (Fig. 3).
Thus, to explain and confirm the effects of MNT knockdown on UR61 cell proliferation, we selected several genes involved in cell cycle and DNA replication and repair processes in which expression was changed according to the RNA-seq data to analyze their expression by RT-qPCR. URMT and URMax34 cells were transfected with the shMNT-2 construct and the empty vector following the same conditions as used for the RNA-seq experiment. Fig. 4E shows that genes involved in the cell cycle, DNA replication, and DNA repair were down-regulated upon silencing of MNT except CDKN1C (p57, a cell cycle inhibitor), which was up-regulated. Altogether, the results are consistent with the negative effect of MNT depletion on cell proliferation in UR61 cells.

MNT localizes in the cytoplasm of MAX-deficient cells
In an attempt to explain the higher MNT levels observed in UR61 cells as compared with the MAX-expressing counterpart, we next examined the subcellular localization of the excess MNT present in MAX-deficient cells. URMT and URMax34 cells were treated with Zn 2ϩ to induce MAX, and we performed a cytoplasmic and nuclear fractionation to analyze MNT levels. In control URMT cells, MNT protein was found in the nuclear fraction, as expected, but also at similar levels in the cytoplasm. In contrast, in URMax34-expressing MAX, MNT protein was found only in the nucleus (Fig. 5A), which is the localization typically described for MNT (26). We were not able to use untreated URMax34 cells as a control as they express some MAX even in the absence of Zn 2ϩ (Fig. 1B). Cytoplasmic MNT was also observed in H1417 cells, human cells deficient in MAX (Fig. 5A). As a control, the localization of MNT was also analyzed in HEK293T cells, which express the MAX protein. Cytoplasmic/nuclear fractionation revealed that MNT and MAX were localized in the nucleus but not in the cytoplasm of HEK293T cells (Fig. 5A). As controls for nuclear and cytoplasmic proteins, we used SIN3B and RhoGDI, respectively (Fig.  5A). To test whether the localization of MNT depended on MAX, we silenced the MAX protein in K562 cells with siRNA, and we carried out cytoplasmic/nuclear fractionation. In con-trol K562 cells, MNT and MAX were mainly localized in the nucleus. In contrast, in MAX-depleted cells, a significant amount of MNT was found in the cytoplasm of K562 cells. As controls for nuclear proteins, we used MYC and CTCF (Fig.  5B). It is noteworthy that nuclear MNT levels were similar in URMT and URMax34, as well as in control K562 and MAXdepleted K562 cells. Taken together, the data indicate that the absence of MAX leads to an up-regulation of MNT protein and that the excess of MNT appears to be accumulated in the cytoplasm.

MNT interacts with MLX in the absence of MAX
Besides MAX, MNT can also bind the HLH protein MLX (6,27). Therefore, MNT would have two partners in UR61-expressing MAX but only one in the absence of MAX, and alternatively, MNT could form homodimers, as schematized in Fig.  5C and previously described in vitro (5). To investigate whether the MNT-MLX interaction also takes place in UR61, we transfected MLX-Flag into URMT cells, and 48 h later lysates were immunoprecipitated with anti-MLX and anti-MNT antibodies. The immunoprecipitates were analyzed by immunoblot, and the data demonstrated that MNT interacted with MLX in UR61 cells (Fig. 5D). We next studied this interaction in URMax34 cells treated with Zn 2ϩ , i.e. cells expressing MAX. The immunoblot results showed that MNT and MLX also interacted between them, but the interaction was weaker when MAX expression was induced by Zn 2ϩ . Thus, the data suggest that, at least in our experimental conditions, MNT-MAX dimers were formed preferentially than MNT-MLX dimers (Fig. 5E). To confirm the MNT-MLX interaction in URMT cells, we prepared a HA-tagged MNT mutant with a deletion of the HLH domain of mouse MNT, termed ⌬bHLH MNT-HA (Fig. 5F). This mutant and the WT counterpart were transfected into URMT and the corresponding lysates were immunoprecipitated with anti-HA. The results showed that HLH domain is required for the MNT-MLX interaction (Fig. 5G). As a proof of concept, we performed the same co-immunoprecipitation in HEK293T, obtaining the same result (Fig. 5H).
In contrast to MAX, MLX is found both in the nucleus and cytoplasm (28). By immunoblot analysis, it was found that MLX was present in the cytoplasm and, at a lesser extent, in the nucleus of URMT cells (Fig. 5I). Next, we prepared nuclear and cytoplasmic fractions of URMT cells and studied by immunoprecipitation with anti-MNT and anti-MLX antibodies whether MNT and MLX interacted in these cell compartments. The results showed that MNT and MLX co-immunoprecipitated in both nuclear and cytoplasmic fractions (Fig. 5J).
As mentioned above, it is described that MNT can form homodimers in vitro and in two-hybrid experiments in yeasts (4,5). Thus, another possibility is that MNT acts as a transcription factor by forming homodimers. This homodimerization has not been demonstrated in animal cells, so we tested this in HEK293T and UR61 cells. We first co-transfected HEK293T cells with GFP-MNT and Flag-MNT constructs and immunoprecipitated with anti-GFP antibody. The immunoblot analysis demonstrated the presence of the smaller Flag-MNT protein in the material immunoprecipitated with anti-GFP indicating that MNT forms homodimers in the cell (Fig. 6A). As expected, both

MNT functions in the absence of MAX
big and small MNT forms were detected when the immunoblots were analyzed with anti-MNT antibody. The results suggested the presence of homodimers between GFP-MNT and Flag-MNT in HEK293T cells. As a control, the lysates were immunoprecipitated with anti-MAX antibody, and both MNT and GFP-MNT were found to be bound to MAX (data not shown).
Next, we investigated the MNT homodimerization in the UR61 system. URMT cells were infected with lentiviral particles containing the GFP-MNT construct, immunoprecipitated with the anti-GFP antibody, and the immunoprecipitates analyzed by immunoblot with anti-MNT antibody and anti-GFP antibodies. The results showed that endogenous MNT was present in the immunoprecipitates with anti-GFP (Fig. 6B), suggesting that MNT forms homodimers in human HEK293T cells and in MAX-deficient rat URMT cells. As MNT dimerization in yeast two-hybrid assays depends on the bHLH-LZ (4, 5), we wondered whether HLH was involved in the homodimerization of MNT in our system. We transfected HEK293T cells with HA-tagged MNT constructs that lack the bHLH region (⌬bHLH MNT) (Fig. 5F) or the WT version together with GFP-MNT constructs, and we carried out immunoprecipitations with anti-HA antibodies that should only recognize the HAtagged proteins. The immunoblot analysis of the immunopre-cipitates revealed that ⌬bHLH MNT was unable to interact with GFP-MNT (Fig. 6C). The same experiment was performed in URMT cells, and the same result was observed, i.e. GFP-MNT bound to WT MNT-HA but not to the ⌬bHLH form (Fig. 6D). We conclude that MNT homodimerizes through the bHLH domain, as expected. We also wondered whether MNT preferentially binds to MAX or to MLX than to MNT itself. For this, we immunoprecipitated with the antibodies anti-GFP and lysates of URMax34 cells transfected with GFP-MNT that were untreated or treated with Zn 2ϩ (to induce MAX). The levels of MAX and MLX after the immunoprecipitation assay were determined by immunoblot, and the results showed that when MAX and MLX are expressed, MNT preferentially binds to MAX than to MLX or MNT (Fig. 6E). Altogether, the results indicate that in URMT cells, MNT can form homodimers or heterodimers with MLX, whereas in URMax34 (in the presence of Zn 2ϩ ), MNT mainly forms heterodimers with MAX.
The former results showing that MNT can form homodimers open the possibility that in MAX-less cells, MNT can regulate gene expression either as a homodimer or as a heterodimer with MLX. To investigate this, we knocked down both MNT and MLX each with two short-hairpin constructs, in URMT cells (MAX-less) and URMax34 cells (with MAX) and  7A) and protein levels (Fig. S8A). MLX knockdown led to retarded proliferation (Fig. S8B) to a similar extent as in MNT depletion. The RT-qPCR analysis showed that in MAX-deficient cells, upon MNT knockdown, CDKN1C/p57 was up-regulated, whereas E2F6, PARPBP, and BIRC5 were down-regulated, as already shown in Fig. 4. However, the expression changes of these four genes were not significantly modified

MNT functions in the absence of MAX
when MLX was depleted or when both MNT and MLX were concomitantly depleted, suggesting that MNT could regulate these genes as homodimers (Fig. 7A). In contrast, CDK1 and BRCA1 were down-regulated when MNT was silenced but also when MLX was silenced (Fig. 7A). However, the expression of CDK1 and BRCA1 was not down-regulated further when both MNT and MLX were silenced. This suggests that both MNT-MNT and MNT-MLX dimers could be regulating the transcription of these genes.
To check whether the transcriptional changes observed upon MNT and MLX silencing were caused by a direct regulation driven by MNT and/or MLX, we performed ChIP-PCR assays in URMT and URMax34 cells after 24 h of stimulation with Zn 2ϩ (i.e. cells in the absence or presence of MAX). From the genes analyzed by RT-qPCR (Fig. 7A), we chose to analyze the genes that showed a peak for MNT binding in the K562 ChIPseq published in the ENCODE project: E2F6, BIRC5, CDK1, BRCA1, and CDKN1C. Then, primers for amplifying these regions in the rat genome were designed, and the presence of E-boxes in the amplicons was confirmed in all but BRCA1 (Fig.  7B, bottom). The coordinates of the E-boxes in human and rat genomes are shown in Fig. S9A. The ChIP-PCR results showed that in the absence of MAX (in URMT cells), MLX and MNT were bound to the promoter of BRCA1 and BIRC5 (Fig. 7B).
Together with the results of Fig. 7A, the data suggest that these genes are regulated by MNT homodimers and possibly MNT-MLX dimers in the case of BRCA1, as the expression of this gene is affected by MLX knockdown. In the presence of MAX (URMax34 treated with Zn 2ϩ ), both MNT and MLX were bound to the promoters of the selected genes. MNT showed a much stronger signal in URMax34 cells that in the absence of MAX (Fig. 7B), suggesting a higher affinity of MNT-MAX to DNA and/or higher stability of the heterodimers versus MNT homodimers, as already shown in Fig. 4. We also performed a ChIP-PCR with anti-MAX in the same genes and amplicons in URMax34 cells and (as a negative control) in URMT. The results showed that MAX was bound to all promoters, but the signal in BIRC5 and BRCA1 was not statistically significant (Fig. S9B). Altogether, the data identified genes involved in cell cycle regulation and survival that are directly regulated by MNT-MNT or MNT-MLX dimers.

Discussion
In this study, we report several novel findings. (i) MNT is required for optimum proliferation even in the absence of MAX. (ii) In the absence of MAX, MNT expression is elevated, and a significant fraction localizes in the cytoplasm. (iii) MNT represses its own transcription in a MAX-dependent manner. (iv) MNT is able to regulate the expression of genes in the absence of MAX. (v) MNT forms homodimers in the cell. Previous reports suggest that MNT functions as a "MYC buffer," curbing excessive MYC activity that would lead to cell transformation. For instance, in vivo MNT knockdown antagonizes MYC-driven lymphomagenesis (18,19), whereas MNT silencing leads to MYC-like phenotypes (10,11,16). Our data show tight control of MNT expression by which MNT limits its own mRNA expression. These controls suggest that MNT plays a critical function in cell biology. Given the relevance of MNT to modulate MYC activity and its central position between the MYC-MAX and MLX-MONDO networks, the activities and regulation of MNT are key issues for MYC-dependent oncogenesis.
MNT expression was high in some MAX-deficient cells like the rat UR61 cells and human lung carcinoma cell lines deficient in MAX. The absence of MAX was in part responsible for this effect because (i) MAX re-expression in UR61, Lu165, Lu134, and H1417 cells results in decreased MNT protein and mRNA (UR61, H1417, and Lu165), and (ii) MAX overexpression in K562 cells results in MNT down-regulation, whereas MAX knockdown results in MNT up-regulation. The mechanism for this MAX effect on MNT levels depended, at least partially, on the auto-repression of MNT expression by MNT-MAX dimers. Luciferase reporter experiments showed that MNT represses its own promoter (which depends on a conserved E-box), and ChIP assays showed that MNT strongly binds to its own promoter when MAX is ectopically expressed in UR61 cells. MNT binds weaker to its promoter in the absence of MAX, in agreement with a the lack of binding to DNA of MNT in vitro (5), and a recent report shows reduced MNT binding to chromatin in MAX-deficient lymphocytes (29). In agreement with these results, the excess levels of MNT localize to the cytoplasm of MAX-deficient cells, whereas MNT is predominantly nuclear in MAX-expressing cells. This would explain why MAX re-expression leads to a decrease in MNT levels; MNT-MAX dimers would be formed to repress MNT expression in the nucleus. On the contrary, in the absence of MAX, MNT is expressed at higher levels because there would be no negative MNT autoregulation. We explored possible MAX-independent effects of MNT using the UR61 model. Although originally MAX was defined as an obligate dimerization partner of MYC, work carried out in the PC12 model and in Drosophila indicated that MYC can function in a MAX-independent manner, for example in inhibition of differentiation (30 -32). MYC overexpression blocks RAS-induced differentiation of UR61 cells (31), although we did not detect a significant effect of MNT depletion on differentiation (data not shown). In contrast, depletion of MNT in MAX-deficient UR61 cells impairs cell proliferation. To our knowledge, this is the first report on a MAX-independent function of an MXD protein.
The fact that MNT depletion impairs UR61 proliferation in a MAX-independent manner adds complexity to the MNT-MYC functional interactions. Conversely to our results in UR61 cells, ablation of Mnt in rodent fibroblasts leads to increased proliferation and transformation capacities (10,16). Actually, in human gastric cancer, high MNT expression correlates with shorter survival, whereas MYC overexpression has the opposite effect (http://kmplot.com/analysis/). 11 Because MNT is a transcription factor, we compared the transcriptomes of parental UR61 cells versus cells with depleted MNT. Upon MNT depletion, a number of genes involved in cell cycle progression and DNA damage response were down-regulated in cells with and without MAX. Also, genes involved in cell cycle arrest such as CDKN1C/p57 were up-regulated. These gene regulations are concordant with the decrease in proliferation exerted by MNT silencing in UR61. It is open to discussion how MNT regulates genes in the absence of MAX. MNT can form heterodimers with the HLH protein MLX (6,27). We have confirmed the ability of MNT to bind MLX in UR61 cells, but when MAX is expressed, the dimers MNT-MAX are favored versus MNT-MLX. As a relevant fraction of MLX is cytoplasmic and we have shown that MNT and MLX also interact in the cytoplasm, this interaction could help to explain the increase in cytoplasmic MNT observed in MAX-depleted cells. The model is depicted in Fig. 8A.
Moreover, we have shown that MNT homodimerizes in UR61 cells, and therefore it is likely that MNT regulates genes as a homodimer as well. In fact, co-depletion of MLX and MNT does not modify the effect of MNT depletion on genes as CDKN1C/p57, BIRC5/Survivin, CDK1, E2F6, BRCA1, and PARPB. Furthermore, ChIP-seq and ChIP-PCR analyses show that MNT binds to some genes in MAX-deficient cells that are regulated by MNT (BIRC5/Survivin, CDK1, BRCA1, ERCC6, and FBXO32), supporting the possibility of a direct regulation by MNT homodimers or an indirect regulation in other cases where we did not detect MNT binding to the regions assayed (CDKN1C and E2F6). CDK1 is the only essential CDK protein

MNT functions in the absence of MAX
for cell cycle progression in animal cells (33). BIRC5/Survivin is also a critical protein for cell survival (34), and we have confirmed its regulation by MNT in URMT cells at the protein level. Thus, the MNT-dependent regulation of both genes helps to explain the antiproliferative effect of MNT knockdown in MAX-deficient cells. Altogether, the data suggest that MNT can regulate transcription as a homodimer or heterodimer with MLX or another not yet identified protein. MNT homodimerization has already been shown in yeast two hybrids (4) and in vitro (5), although recombinant MNT homodimers did not bind DNA (5). The analysis of MNT-bound regions in our ChIP-seq experiments also revealed the presence of sites for forkhead factors (FOXP1, FOXA2, and FOXO3) in agreement with the reported coordinated regulation between MNT and FOXO of some cell cycle control genes (25). Further work is required to clarify this point. According to our results, MNT homodimers or MNT-MLX heterodimers would be regulating cell cycle and DNA repair checkpoint genes in a MAX-independent manner for correct survival and proliferation of the cells. The model based on these observations is presented in Fig. 8B.
In summary, the results reported here show a strict autoregulation of MNT, supporting a pivotal role of MNT in the control of cell proliferation even in the absence of its canonical dimerization partner MAX.

Immunoblot and immunoprecipitation
Cell lysis, immunoblots, and immunoprecipitations were performed as described (37). Each immunoblot shows a representative experiment out of at least two biological replicates with similar results. The antibodies used are described in Table S7.

Luciferase reporters and assays
To generate the MNT-luc, E-box 1 MNT-luc, and E-box 2 MNT-luc reporter vectors, two pairs of primers were designed for each construct (Table S8), targeting sequences of the human genome corresponding to 850 bp upstream from the transcription start site of the MNT gene. The amplified DNA was inserted into the pBV-luc reporter vector at the EcoRV and HindIII sites (38). The sequence of the MNT promoter was from the UCSC genome browser. Cells were transfected with jetPEI reagent (Polyplus), and luciferase assays were performed, as described previously (37).

Cell proliferation, cell cycle, and annexin V-binding analysis
Cell proliferation was monitored with a cell counter (Nucleo-Counter NC-100, Chemometec) or a cytometer (Guava PCA, Merck Millipore). For the clonogenic assays, 1-2 ϫ 10 6 cells/ml were seeded in a 6-well or 60-mm plate after transfection by electroporation. 48 h post-transfection, cells were selected with puromycin at 0.2-0.5 g/ml final concentration. After 8 -17 days, the cells were stained with crystal violet, and the dye was measured by absorbance at 595 nm as described (39). To determine the sub-G 0 -G 1 population, cells were transfected, and 7 days post-transfection, the DNA concentration was analyzed by propidium iodide staining and flow cytometry as described (40). Apoptosis was assayed by annexin V-positive binding in MACSQuant VYB (Miltenyi Biotec). Roughly, 10 6 cells were harvested and washed twice with previously filtered PBS and 3 mM EDTA. Cells were resuspended in 10 mM HEPES/NaOH, pH 7.4, and stained with 5 l of annexin V-FITC (BD Biosciences) for 1 h. Cells were washed twice and resuspended in 500 l of binding buffer with 5 l of 7-aminoactinomycin D (Immunostep apoptosis detection kit). The results were analyzed with Flow Logic software (Miltenyi Biotec).

RNA analysis and RNA-seq
For qPCR, total RNA was isolated using the TRI Reagent Solution (Invitrogen). cDNA was generated by reverse transcription (RT) using the iScript (Bio-Rad). Quantitative PCR (qPCR) was performed with specific primers (Table S8) using the iTaq Universal SYBR Green Supermix (Bio-Rad) and CFX Connect Real-Time PCR Detection System (Bio-Rad). RNA was converted into cDNA and analyzed as described (41). Levels of mRNA were normalized against actin and RPS14 mRNA levels. For RNA-seq , total RNA was isolated with the RNeasy kit (Qiagen) from two independent experiments of MNT silencing in URMT and URMax34 cells, both treated with 100 M ZnSO 4 for 24 h. mRNA libraries were prepared using the Illumina TruSeq RNA Sample Prep kit version 2 (kit RS-122-2002, Illumina). A minimum of 40 million 50-base single-end reads per sample were obtained. TopHat algorithm (42) was used to align the data using a set of gene model annotations and/or known transcripts of rat genome obtained from RefSeq database. Cufflinks software (43) was run to estimate transcript abundances represented in RPKM units (reads per kilobase per million reads) as described (44). The gene expression of the genes are represented as RPKM values. The log2 RPKM values of the transcripts that were differentially expressed in URMT and URMax34 cells were represented in a dot graph with the log2 RPKM values of the control condition (pLKO) on the x axis and the log2 RPKM values of the experimental condition (shMNT) on the y axis.

ChIP and ChIP-seq
Total-cell extracts were first lysed with a hypotonic buffer (described under "Nuclear/cytoplasmic fractionation") for purifying the nuclear compartment. Then, nuclear lysis and ChIP were performed essentially as described previously (37). Immunoprecipitated DNA was purified with the QIAquick PCR purification kit (Qiagen) and analyzed by qPCR. The antibodies and primers used are described in Tables S7 and S8, respectively. For ChIP-PCR, chromatin was prepared from URMT and URMax34 cells (both treated with 100 M ZnSO 4 for 24 h) and immunoprecipitated with anti-MNT antibody and unspecific IgG. For ChIP-seq, we used MNT-bound chromatin and the corresponding inputs from URMT cells. Singleend 51-bp ChIP-seq data from three replicate experiments and three input samples of URMT cells were generated by HiSeq. Alignment and peak detection were performed using the ENCODE (phase-3) transcription factor ChIP-seq pipeline specifications (45). Reads were aligned to the rat reference genome (assembly Rnor_6.0) using BWA (46), removing duplicates with PICARD (Picard Toolkit 2018, http://broadinstitute. github.io/picard) 11 and filtering all reads with a quality score Ͻ30. Peaks were called using SPP (47) and input samples as background samples. Enrichment and quality measures were computed with Phantompeakqualtools (48). Reproducibility of peaks identified from the three replicate experiments was measured using IDR with a threshold of 0.1. Peak annotation was performed with Homer (http://homer.ucsd.edu/homer/motif/ motifDatabase.html) 11 (49).
The SYDH ENCODE project was chosen for representation of the binding peaks in the promoter of the human MNT gene. The sequences corresponding to those peaks were analyzed confirming the presence of the E-boxes identified in the human MNT promoter. Enriched motifs on immunoprecipitated regions were identified by HOMER (49) FindMotifsGenome algorithm using the rat genome Rnor_6.0 version as reference, masking common repeats and with a region size of 200 bp (http://homer.ucsd.edu/homer/motif/motifDatabase.html). 11 The RNA-seq and ChIP-seq data have been deposited to the European Nucleotide Archive with the accession number PRJEB23604.

Statistical analysis
Student's t test was used to evaluate the significance of differences between control and experimental groups. A p value of less than 0.05 was considered as significant. The threshold for expression changes in the RNA-seq analysis was set as log2(RPKM fold change) Ն0.7 or Յ Ϫ0.7 and a p value Ͻ0.1.