Nuclear Respiratory Factor-1, a Novel SMAD4 Binding Protein, Represses TGF-β/SMAD4 Signaling by Functioning as a Transcriptional Cofactor

SMAD4, a key regulator of transforming growth factor-β (TGF-β) signaling, plays a major role in cell growth, migration, and apoptosis. In particular, TGF-β/SMAD induces growth arrest, and SMAD4 induces the expression of target genes such as p21WAF1 and p15INK4b through its interaction with several cofactors. Thus, inactivating mutations or the homozygous deletion of SMAD4 could be related to tumorigenesis or malignancy progression. However, in some cancer types, SMAD4 is neither mutated nor deleted. In the current study, we demonstrate that TGF-β signaling with a preserved SMAD4 function can contribute to cancer through associations with negative pathway regulators. We found that nuclear respiratory factor-1 (NRF1) is a novel interaction SMAD4 partner that inhibits TGF-β/SMAD4-induced p15INK4b mRNA expression by binding to SMAD4. Furthermore, we confirmed that NRF1 directly binds to the core region of the SMAD4 promoter, thereby decreasing SMAD4 mRNA expression. On the whole, our data suggest that NRF1 is a negative regulator of SMAD4 and can interfere with TGF-β/SMAD-induced tumor suppression. Our findings provide a novel perception into the molecular basis of TGF-β/SMAD4-signaling suppression in tumorigenesis.


Introduction
In vertebrates, the transforming growth factor-β (TGF-β) pathway regulates the expansion of epithelial and neural tissues and the immune system and functions in wound repair. As a consequence, malfunctions in TGF-β pathway signaling often result in tumorigenesis [1]. In the canonical TGF-β signaling pathway, exogenous signaling molecules (ligands) activate the plasma membrane-bound serine/threonine kinase receptors, which further activate the pathway via intracellular SMAD mediators. The TGF-β receptor complex phosphorylates the transcription factors SMAD2 and SMAD3, which then bind to SMAD4, translocate into the nucleus, and associate with diverse DNA-binding cofactors to target genes for regulation. SMAD4, a key downstream component in this system, was first isolated as a tumor-suppressor gene in human pancreatic ductal carcinomas [2,3]. The homozygous deletion or inactivating mutations of SMAD4 play a crucial role in the malignant progression of certain cancer types [4,5]. By contrast, SMAD4 mutations are rarely observed in other cancer types. The function and significance of positive or negative regulators of SMAD4 are being studied in many cancers [6].
Nuclear respiratory factor 1 (NRF1) is a transcription factor that regulates a myriad of cellular functions, including mitochondrial biogenesis, DNA replication and repair, proliferation, and apoptosis. NRF1 was originally identified as a cytochrome-c activator [7]. NRF1 and NRF2 regulate a multitude of genes essential for the expression of proteins implicated in mitochondrial functions and biogenesis [8], mitochondrial replication, gene expression, and protein import and assembly [9][10][11][12]. Besides its role in the regulation of mitochondrial functions, NRF1 is also a crucial player in histone gene expression and acts as a regulator of cell growth and proliferation [13,14]. Accumulating evidences also implicate that NRF1 expression and its transcription factor activity may contribute to the pathogenesis of breast cancer, glioblastoma, and neuronal dysfunction [15][16][17].
Modulation of the transcription factor function through protein-protein interactions is a crucial process in the activation or repression of signal transduction pathways [1,5]. NRF1 binds to the promoter region of E2F downstream signaling molecules, modulating their transcription and thereby controlling cell cycle progression [13]. The interaction between NRF1 and other cofactors is also presumed to have a dramatic and diverse effect in different cell types. NRF1 transmits extracellular physiological changes in a tissuespecific manner via its interaction with the PGC-1 family members [18][19][20]. Employing a computational analysis to predict the tissue-specific combinatorial gene regulation, Yu et al. identified NRF1, SMAD3, and E2F as the top three signaling hubs for interactions between transcription factors in the cervix [21].
Here, we demonstrate that NRF1 is a novel SMAD4-binding partner and that the interaction between SMAD4 and NRF1 can repress TGF-β/SMAD4-induced tumor-suppressor functions. Besides, NRF1 can function as a transcription factor and influence SMAD4 expression. We also identified p15INK4b, a cell cycle inhibitor, as an important target for TGF-β/SMAD4-induced tumor-suppressor functions. These results suggest a possible tumorigenic role for NRF1 and may also explain the paradoxical lack of SMAD4 mutations in some types of cancers.

NRF1 Is a Novel SMAD4-Binding Partner
To explore SMAD4 regulation, we searched for SMAD4-binding partners using a baculovirus ProtoArray; the screen identified NRF1 and other novel interaction proteins (Table 1). Both endogenously and exogenously expressed SMAD4 and NRF1 interacted in vitro, as confirmed by immunoprecipitation ( Figure 1A,B). We also assessed NRF1 and SMAD4 binding using a BiFC assay [22,23] and in situ PLA to exclude the possibility of binding artifacts ( Figure 1C). The BiFC assay not only detects protein-protein interactions in living cells but also identifies the intracellular location where these interactions occur. The plasmids used for the BiFC assay were constructed as previously described [24] and were transfected into HeLa cells. The binding of NRF1to SMAD4 was observed, particularly within the nucleus. In situ PLA, like BiFC, not only confirms the cellular location of proteinbinding interactions but can also confirm the binding patterns of proteins expressed at low levels and identify transient binding interactions. As with BiFC, the in situ PLA results demonstrated that SMAD4 and NRF1 interact within the nucleus. Our multiple experimental results confirm that NRF1 is a novel SMAD4-binding partner.  For the BiFC analysis, HA-SMAD4-VC155 and FLAG-NRF1-VN173 constructs were co-transfected into HeLa cells. After 24 h, bound proteins were detected by visualizing the fluorescence signal. For in situ PLA assays, cultured HeLa cells were fixed and incubated with NRF1 or SMAD4 antibodies and the PLA probe. Fluorescence was imaged with a confocal microscope system. DIC: differential interference contrast. Scale bar, 5 µm.

SMAD4 MH1 and MH2 Domains Interact with the NRF1 Dimerization Domain
To identify the specific binding domains responsible for NRF1 and SMAD4 binding, we first generated six deletion constructs, each containing one functional domain of either NRF1 or SMAD4 (schematics in Figure 2A,B). Using the BiFC assay, we observed binding between full-length (FL) SMAD4 and each of the NRF1 deletion constructs and between FL NRF1 and each of the SMAD4 deletion constructs. Similar to WT, nuclear binding interactions were observed for SMAD4 MH1 and MH2 with FL NRF1 (Figure 2A) and for NRF1 (1-108) with FL SMAD4 (Figure 2A). Interestingly, the other NRF1 domain constructs (NRF1 (108-304) and NRF1 (304-503)) interacted with FL SMAD4 in the cytoplasm ( Figure 2A). We observed cytoplasmic and nuclear binding for SMAD4 MH1 and FL NRF1, suggesting that NRF1 and SMAD4 could initially interact in the cytoplasm and then translocate into the nucleus. The binding specificity between SMAD4 and NRF1 was confirmed through BiFC competition assays with the transfection of tagged and nontagged proteins, in which nontagged NRF1 and SMAD4 were used as competitors (Supplementary Figure S1). The fluorescence signal from tagged NRF1 and SMAD4 was counted with and without the nontagged competitors, and the results revealed that each binding interaction was specific. NRF1 and SMAD4 binding was further confirmed by protein IP from HeLa cells ( Figure 3A,B).

NRF1 Inhibits mRNA Expression of p15INK4b via SMAD4-NRF1 Interactions
We hypothesized that NRF1 binding to SMAD4 influences the interaction between SMAD4 and SMAD2/3. As shown in Figure 4A, NRF1 overexpression slightly increased p-SMAD2/3 binding to SMAD4, even in the presence of TGF-β. We then asked whether NRF1 affects TGF-β-induced SMAD4 target gene expression. Using qRT-PCR, we quantified the transcription levels of a well-known SMAD4 target gene, p15INK4b, in HeLa and SiHa cells ( Figure 4B). The qRT-PCR results showed that TGF-β/SMAD4-induced p15INK4b transcription was decreased when NRF1 was overexpressed in both HeLa (22.5%) and SiHa (24.5%) cells. Furthermore, we confirmed that the mRNA expression of p15INK4b, which was further increased by the overexpression of SMAD4, was dramatically inhibited by NRF1 in SiHa (47.5%) cells ( Figure 4B). We performed a ChIP analysis to confirm the observed effects of NRF1 on the direct interaction of SMAD4 to the SMAD-binding element (SBE) of the p15INK4b promoter in SiHa cells. SMAD4 directly binds to the two SBE regions (ChIP I and ChIP II) of the p15INK4b promoter, and this binding is inhibited by NRF1 overexpression ( Figure 4C). DNA-pulldown assays support the ChIP assay results (Supplementary Figure S2). The NRF1-binding sites on the SMAD4 and p15INK4b promoters were mutated, and the binding of NRF1 was decreased in comparison to the wild-type promoter sequences. Taken together, the results suggest that, although the binding between SMAD4 and NRF1 moderately strengthened the interaction between SMAD4 and phosphorylated SMAD2/3, NRF1 inhibits TGF-β/SMAD4-induced p15INK4b mRNA expression by interfering with SMAD4 binding to the p15INK4b promoter, thus inhibiting the TGF-β/SMAD4-induced tumor-suppressor function.

NRF1 Regulates SMAD4 Expression through Its Function as a Transcription Factor
NRF1 is a well-known transcription factor that regulates the expression of a plethora of metabolic genes related to cellular growth and development. Hence, we examined whether NRF1 directly regulates SMAD4 transcription, in addition to the regulation we observed as a direct SMAD4-binding partner. Changes in the SMAD4 mRNA expression, measured by qRT-PCR, were observed after NRF1 overexpression in HeLa, SiHa, and MCF7 cells ( Figure 5A). The SMAD4 mRNA levels were decreased by over 60% after NRF1 transfection in all three cell lines. We identified four putative NRF1-binding sites within the SMAD4 promoter −500 bp to the transcription start site using MatInspector professional software and the internet-based TFSEARCH database (Supplementary Figure S3). We used a luciferase assay and generated three constructs, each containing a partial deletion mutant in the SMAD4 promoter (Luc-371, -216, and -41), to identify the core region enabling NRF1mediated transcriptional repression at the SMAD4 promoter ( Figure 5B, left). The transient co-transfection of NRF1 and each SMAD4 deletion mutant revealed changes in SMAD4 promoter activity ( Figure 5B, right). The 3 -truncated constructs Luc-216 and -41 were 30% and 60% inhibited, respectively. The inhibition of SMAD4 transcription by NRF1 was also confirmed using a dose-dependent NRF1 luciferase assay ( Figure 5C). The inhibition of SMAD4 promoter activity by NRF1 was concentration-dependent. Furthermore, the ChIP assay results confirmed that the two NRF1-binding elements (NBEs) are core sites for the regulation of SMAD4 expression by NRF1 ( Figure 5D). NRF1 directly binds to the SMAD4 promoter at ChIP sites I and II, confirming that NRF1 inhibits SMAD4 expression by directly binding to the SMAD4 promoter. In addition to its effect on SMAD4 transcription, NRF1 binding to the SMAD4 promoter also slightly inhibited the SMAD4 protein expression (Supplementary Figure S4).

Discussion and Conclusions
In this study, we found that the amino-terminal domain of NRF1 promotes NRF1 binding to SMAD4 in the nucleus based on multiple assays, including BiFC, in vivo proximity ligation assays, and protein immunoprecipitation from HeLa cells. Much to our surprise, the DNA-binding domain of NRF1, a domain necessary for its interaction with other binding partners [25,26], was not necessary for NRF1 and SMAD4 binding in the nucleus. The BiFC fluorescence signal detected for NRF1:SMAD4 binding may result from functional nuclear localization sequence 88-116 of SMAD4 MH1 and the SMAD4 MH2 domain mainly in the nucleus. A luciferase reporter assay was used to investigate the regulatory role of NRF1 as a transcription factor for SMAD4. Serial SMAD4 promoter deletions were generated based on the results from a manual search for putative NRF1-binding sites in the SMAD4 promoter region. Significant inhibition of SMAD4 transcriptional activity was observed in HeLa cells upon the ectopic expression of NRF1. Three pairs of promoter regions were derived from a minimal proximal promoter region (−371 bp to −41 bp) of SMAD4 to identify the region necessary for NRF1 binding more specifically.
Interestingly, the construct with the NRF1-binding site deletion (Luc-41) rescued the promoter activity. Therefore, we concluded that this NRF1-binding site (−186 bp to −175 bp) is crucial for the regulation of SMAD4 promoter activity. Moreover, the ectopic expression of NRF1 significantly decreases SMAD4 mRNA transcription and moderately reduces the protein expression (Supplementary Figure S4).
In the genome, a heterodimer of two transcription factors can frequently influence other binding partners, which are likely to be common DNA elements. NRF1, as a regulator of diverse E2F targets, controls cell cycle progression [13,27]. In addition, the transcription factor activity of NRF1 is highly active in human cancer. Recently, Bhawe et al. reported that aberrant NRF1 activity and its regulated genes, including TGF-β1 and p15INK4b, are overexpressed in glioblastoma (high-grade astrocytoma) [15]. p15INK4b, an E2F target gene, also functions as a downstream effector of SMAD4-mediated antiproliferative effects. This prompted us to hypothesize that p15INK4b is a common downstream target of NRF1 and SMAD4. SMAD complexes can occupy the SMAD-binding region (SBR) element located in the distal region of the p15INK4b promoter and influence the transcriptional activation of p15INK4b in response to TGF-β [28,29]. Our promoter analysis identified one putative NRF1-binding site (GGCGCATGCGTC) residing approximately 240 bp downstream of the SBR2 element on the p15INK4b promoter (Supplementary Figure S5). Although this site is not the perfect palindrome ((T/C)GCGCA(T/C)GCGC(A/G)) usually required for NRF1 recognition, the sequence GCGCRYGCGY is an alternative binding element consensus sequence that is preferred by the chicken NRF1 homolog initiation binding receptor (IBR) [14]. Our ChIP assays confirmed the binding sites of both NRF1 and SMAD4 in the p15INK4b promoter region. NRF1 regulates p15INK4b transcription in two manners: (1) When NRF1 is ectopically expressed in the presence of SMAD4, it dramatically decreases the transcription of p15INK4b, regardless of the SBE function ( Figure 4B). This suggests that NRF1 directly interacts with the p15INK4b promoter either alone or as an NRF1-SMAD4 heterodimer. (2) Alternatively, without a functional NBE, NRF1 can minimize the promoter activity in the presence of SMAD4, indicating an indirect influence likely relying on the NRF1 regulation of SMAD4. In the former scenario, the DNA may form a bent structure, as observed for the binding of transcription regulators, including E2F, to the promoter region [13]. Cyclin D1 and Cyclin D1-associated kinases, whose kinase activity can be inhibited by p15INK4b [30], interact directly with NRF1 in the breast cancer cell line MCF-7 and inhibit NRF1 through phosphorylation [31].
Together with our findings, this suggests that NRF1 is involved in the regulation of p15INK4b transcription via direct or indirect interactions with SMAD4 ( Figure 6). In the absence of NRF1, SMAD4 stimulates p15INK4b transcription through the recruitment of a SMAD protein complex; however, the presence of NRF1 in the p15INK4b promoter region prevents p15INK4b induction by negatively regulating SMAD4 transcription. Furthermore, the loss of p15INK4b and SMAD4 activity, mediated by NRF1, can induce an alternative TGF-β signaling pathway in cancer cells, thereby promoting proliferation and tumor progression. Cyclin D1 abundance has been proposed to correlate with the reduced mitochondrial activity due to the reciprocal regulation of NRF1 by Cyclin D1 [31]. Consequently, this decrease in the mitochondrial function might be responsible for the preference of glycolysis over the TCA cycle in cancer cells, a phenomenon named the Warburg effect [32]. Earlier reports of the gene module analysis showed an association between increased nuclear mitochondrial genes, rather than the presence of NRF1 itself, and a poor survival prognosis in breast cancer patients [33]. NRF1, as a hub for p15INK4b regulation, is essential in oncogenesis through its crosstalk with SMAD4 in the TGF-β signaling pathway. To understand the NRF1 and SMAD4 protein-protein interaction more completely, an extensive analysis of each protein and/or structural biology studies will be required. Recently, Barbagallo, D et al. and Broggi, G et al. reported that the RNA-binding motif responsible for the interaction between Serine and Arginine Rich Splicing Factor 1 (SRSF1) and circSMARCCA5 (a 269-nucleotide-long circRNA) are coupled with increased amounts of total Vascular Endothelial Growth Factor A (VEGFA) mRNA secretion and the recommended use of SRSF1 as a diagnostic immunomarker in gliomas [34,35]. In the case of NRF1, it appears to be both a negative regulator and an interaction partner of SMAD4. As such, by forming a heterodimer with SMAD4, NRF1 influences the transcriptional activity of p15INK4b by interfering with SMAD4 expression, with a possible dependence on direct contact with the NBE in the p15INK4b promoter region. This novel regulation mechanism might be an indicator of the oncogenic role of NRF1. Additionally, the functional loss of SMAD4, caused by an NRF1-induced decrease in p15INK4b transcriptional induction, provides a rationale for the absence of SMAD4 mutations or deletions in some tumor types.

Cell Culture, Transient Transfection, and Treatments
HeLa, SiHa, and MCF7 cells were purchased from the Korean Cell Line Bank (KCLB) and were authenticated by a short tandem repeat (STR) analysis at the cell line core facility at Abion (Abion, Inc., Seoul, Korea). HeLa and SiHa cells were cultured using RPMI 1640, and Dulbecco's modified Eagle's medium was used for MCF7 cells. Culture medium was supplemented with 10% fetal bovine serum (Gibco, ThermoFischer Scientific, Seoul, Korea) and 1% penicillin/streptomycin (Gibco, ThermoFischer Scientific, Seoul, Korea). Cells were incubated at 37 • C at 5% CO 2 atmosphere and were regularly tested for mycoplasma infection using the Myco VALiD Mycoplasma PCR Detection Kit (Intron Biotechnology, Kyung-Gi, Korea). Cultured cells were transiently transfected using Genefectine Reagent (Genetrone Biotech, Gyeonggi-do, Korea). TGF-β1 plasmid (Komabiotech, Seoul, Korea) was transfected at 10 ng/mL.

The Proximity Ligation (PLA) Assay
HeLa cells were used for the PLA experiments following the manufacturer's protocol (O-LINK Bioscience, Uppsala, Sweden). Paraformaldehyde fixed cells on a 4-well cell culture slide (SPL Life Sciences Co., Gyeonggi-do, Korea) by the overnight incubation of were incubated with NRF1 and SMAD4 antibodies at 4 • C overnight, after 1 h of blocking with 5% nonfat milk to avoid nonspecific binding. Cells were washed twice with TBST for 5 min and incubated with the PLA probe solutions for 60 min at 37 • C. After incubation, cells were washed with TBST for 5 min and incubated for half an hour with the ligase solution at 37 • C. Samples were washed twice with TBST for 2 min, followed by amplification using polymerase solution for 100 min at 37 • C. The samples were washed with SSC buffers (prepared according to the manufacturer's recipe) before mounting [24]. The slides were visualized, and images were taken using an LSM 700 ZEISS laser scanning confocal microscope (Carl Zeiss, Jena, Germany).

Quantitative Real-Time Polymerase Chain Reaction (qRT-PCR) Analysis
Total RNA was extracted from transfected cells using the Hybrid-RTM total RNA Kit (GeneAll Biotechnology, Seoul, Korea). Following cDNA synthesis using the Superscript II First-Strand Synthesis System (Life Technologies), qRT-PCR was executed with a dual system LightCycler (Roche Diagnostics). The SYBR Green-based comparative CT method was used to analyze the target gene expression in relation to HPRT expression (relative fold-change = 2 −∆∆CT ) [36]. Primers used are listed in Supplementary Table S1. All PCR primers were purchased from Cosmo Genetech (Seoul, Korea).

Luciferase Reporter Gene Assay
Luciferase reporter constructs were cloned using the restriction map of the BAC729G3 bacterial artificial chromosome (BAC) clone from the RPCI-11 human BAC library (Invitrogen, Carlsbad, CA, USA), which covers the alternative promoter region of SMAD4 (Luc-371) [24,36]. All SMAD4 constructs (Luc-371, Luc-216, and Luc-41) were confirmed using sequencing. Twenty-four hours after the transfection, cells were lysed with luciferase assay buffer. Luciferase activity was measured using the dual-luciferase reporter assay system following the manufacturer's protocols (Promega, Madison, WI, USA), and luminescence was measured using a GENios Pro Microplate Reader (Tecan Trading AG, Mannedorf, Switzerland).

Screening Analysis of Transcription Factor Binding Sites (TFBS)
Screening of TFBS was carried out as described previously [24,36]. Briefly, regulatory elements in the core promoter region were identified using MatInspector (Genomatix Software GmbH, Munich, Germany; http://www.genomatix.de, accessed on 1 January 2010). The TFBS program (http://www.cbrc.jp/research/db/TFSEARCH.html, accessed on 1 January 2010) was used to identify the putative NRF1 transcription factor binding sites within the 5 -flanking region of the SMAD4 promoter.

Statistical Analysis
The results were statistically validated using one-way ANOVA analysis followed by Tukey's test for multiple comparisons. GraphPad prism software was used for all statistical analyses (GraphPad Software Inc., San Diego, CA, USA). Results were considered significant at * p < 0.05, ** p < 0.001, *** p < 0.001, or **** p < 0.0001. All the data with error bars represent means ± SD for at least three independent experiments.