The RNA Binding Complexes NF45-NF90 and NF45-NF110 Associate Dynamically with the c-fos Gene and Function as Transcriptional Coactivators*

Background: NF45-NF90 and NF45-NF110 possess double-stranded RNA binding domains. Results: NF45-NF90 and NF45-NF110 occupy the c-fos gene dynamically and enhance its transcriptional induction by extracellular stimuli. Conclusion: NF45-NF90 and NF45-NF110 function as coactivators and are involved in various steps of gene expression. Significance: Analysis of RNA binding coactivators is important for understanding the coordination among different steps of gene expression. The c-fos gene is rapidly induced to high levels by various extracellular stimuli. We used a defined in vitro transcription system that utilizes the c-fos promoter to purify a coactivator activity in an unbiased manner. We report here that NF45-NF90 and NF45-NF110, which possess archetypical double-stranded RNA binding motifs, have a direct function as transcriptional coactivators. The transcriptional activities of the nuclear factor (NF) complexes (NF45-NF90 and NF45-NF110) are mediated by both the upstream enhancer and core promoter regions of the c-fos gene and do not require their double-stranded RNA binding activities. The NF complexes cooperate with general coactivators, PC4 and Mediator, to elicit a high level of transcription and display multiple interactions with activators and the components of the general transcriptional machinery. Knockdown of the endogenous NF90/NF110 in mouse cells shows an important role for the NF complexes in inducing c-fos transcription. Chromatin immunoprecipitation assays demonstrate that the NF complexes occupy the c-fos enhancer/promoter region before and after serum induction and that their occupancies within the coding region of the c-fos gene increase in parallel to that of RNAPII upon serum induction. In light of their dynamic occupancy on the c-fos gene as well as direct functions in both transcription and posttranscriptional processes, the NF complexes appear to serve as multifunctional coactivators that coordinate different steps of gene expression to facilitate rapid response of inducible genes.

the possibility that heterogeneous nuclear ribonucleoprotein R serves as a coactivator for the c-fos gene through a novel mechanism that differs from those of p300/CBP, TORCs (transducers of regulated CREB activity), and Mediator.
To further clarify the molecular identity of the TREF activities, we purified another component of TREF and identified it as the complex of NF45 and NF90. NF90 and its splicing variant NF110 contain double-stranded RNA binding motifs (dsRBMs), which have been demonstrated experimentally to bind dsRNAs (11)(12)(13). Consistently, NF90 binds the adenylate uridylate-rich elements present in a number of mRNAs to regulate the stability, nuclear export, and cellular distribution (14 -22) of these mRNAs; moreover, NF90 is also known to modulate the rate of translation (17,20,23). NF90 and NF110 bind to genomic RNAs of various viruses and are not only involved in cellular defense against viral infection but also are utilized as host factors for viral replication (24,25). In addition to the functions involving their dsRNA binding activities, NF90 and NF110 have been implicated in regulating transcription. Indeed, several studies suggested that the NF45-NF90 complex binds specific DNA sequences (26 -29) and activates transcription in cell-based assays (30 -35). Given the role for the NF complexes in mRNA stabilization, however, it has been unresolved if the NF complexes have a direct transcriptional function to increase mRNA levels.
Using a highly purified in vitro transcription system, we show that the NF complexes have a direct transcriptional function as a coactivator. This coactivator activity does not require dsRNA binding activities, which are essential for the mRNA stabilizing activity of the NF complexes. Knockdown of the endogenous NF90/NF110 in mouse cells shows that the NF complexes play an important role in rapid induction of c-fos transcription. The NF complexes are present on the c-fos enhancer/promoter region before serum induction, and their occupancies within the coding region increase in parallel to that of RNAPII upon serum induction. Consistent with their occupancy on the c-fos enhancer/promoter and their role as a coactivator, the NF complexes interact with the activators and the general transcriptional machinery. Given their dynamic occupancy on the c-fos gene and well known functions in posttranscriptional processes, the NF complexes may coordinate multiple steps in gene expression by dynamically associating with diverse machineries that regulate gene expression.

Experimental Procedures
Plasmids-Based on the published DNA sequences (33), the coding regions of human NF45, NF90, and NF110 were amplified from HeLa cell mRNA using a PCR-based method (10). The cDNA clones were tagged with the triple (3ϫ) FLAG at their C termini by PCR and then subcloned into pVL1392 (BD Biosciences) for baculovirus expression or pcDNA3 (Invitrogen) for transient expression in mammalian cells. The cDNAs encoding NF45, NF90, and NF110 tagged with GST at their N termini were created by a PCR-based method and subcloned into pVL1392. The baculovirus vector for expressing firefly luciferase was created by using the cDNA amplified from pGL3-basic vector (Promega). PCR-based mutagenesis was used to create the cDNAs encoding NF90(FA) and NF110(FA) by mutating Phe-432 and Phe-559 into alanine, and those encoding NF90(AP) and NF110(AP) by mutating Ala-458 and Ala-588 into proline. The cDNAs of general transcription factors were tagged N-terminally with the FLAG epitope using corresponding cDNAs (36), and the derived cDNAs were subcloned into pVL1392. RNGTT (RNA guanylyl transferase and 5Ј-phosphatase), Oct-1/POU2F1, and USF (upstream transcription factor) were similarly tagged with the FLAG epitope using cDNAs amplified by PCR using HeLa cell mRNA and subcloned into pVL1392. All the cDNA clones were sequenced entirely to exclude PCR-derived errors. The vectors for expressing FLAG-tagged SRF, Elk-1, CREB, and ATF1 were described previously (10).
Protein Purification-Baculovirus transfer vectors were cotransfected into Sf9 cells with linearized baculoviral DNA (BD Biosciences) to produce the recombinant viruses, which were used for infecting High Five cells. The cell extracts were prepared essentially as described (10). For expressing the complexes of NF45-NF90 and NF45-NF110, High Five cells were coinfected with two different baculoviruses expressing each subunit. All the purification procedures used BC buffer, which consists of 20 mM Hepes-KOH, pH 7.9, 1 mM EDTA, and 10% glycerol, with the number in the parentheses indicating the concentration (mM) of KCl. The extract was loaded onto a HiTrap Q column equilibrated with BC(100) containing 0.5 mM PMSF and 1 mM DTT, and the bound proteins were eluted in a linear gradient from 0.1 M KCl to 1.0 M KCl. The NF45, NF90, NF110, and their complexes were eluted at 300 -500 mM KCl. The fractions containing the NF proteins were then subjected to a Sephacryl S300 gel filtration column (GE Healthcare) in BC(500) containing 0.5 mM PMSF. Because all the NF proteins contained a 3ϫFLAG tag, appropriate fractions were bound to anti-FLAG M2-agarose (Sigma) in BC(300) containing 1 mM PMSF and 0.01% Triton X-100, and the bound proteins were eluted in the same buffer containing 0.1 mg/ml 3ϫFLAG peptide. Extracts for general transcription factors were prepared similarly. For GST pulldown assays, the extract was purified partially on a HiTrap Q column to remove nucleic acids. Extracts for GST fusion proteins were prepared similarly and used directly for binding the fusion proteins to glutathione-Sepharose 4B (GE Healthcare). Purification of the transcription factors used for in vitro transcription was performed as described (10).
Antibodies-The cDNAs encoding the C-terminal regions of NF45 (residues 162-390), NF90 (residues 353-706), and NF110 (residues 725-898) were amplified and subcloned into pET15b, and the resultant plasmids were transformed into BL21(DE3)pLysS. The recombinant proteins, induced by 1 mM isopropyl 1-thio-␤-D-galactopyranoside, were separated by preparative SDS-PAGE, and after excision of the appropriate portion of the gel the proteins were eluted using the Model 422 electro-eluter (Bio-Rad). The purified proteins were then used to immunize rabbits to obtain anti-sera. The crude antisera were purified through columns in which the C-terminal regions of NF45, NF90, and NF110 were cross-linked to Sepharose 4B by dimethyl pimelimidate (Pierce). The bound antibodies were eluted in 0.2 M glycine-HCl, pH 2.5, and dialyzed against phosphate-buffered saline. The anti-RPB2 antibody was prepared in essentially the same manner using the residues 1-306 of RPB2 as purified antigen.
In Vitro Transcription-In vitro transcription assays were performed essentially as described previously using the templates containing the c-fos promoter, pfMC2AT, and five GAL4-binding sites, pG5HMC2AT (10,37). To allow the inclusion of additional factors in the reactions, however, the volume of transcription assays was increased to 50 l instead of 25 l.
Chromatin Immunoprecipitation (ChIP) Assays-ChIP assays were performed as previously described (38) with some modifications. To avoid inactivating anti-NF antibodies, 1% SDS in the lysis buffer and the ChIP dilution buffer and 0.1% SDS in low salt buffer and high salt buffer were substituted by 1% and 0.1% Nonidet P-40, respectively; moreover, 1% deoxycholate in LiCl Buffer was omitted. C3H10T1/2 cells, grown in a 100-mm culture dish, were starved in DMEM containing 0.5% fetal bovine serum (FBS) for 24 h and then stimulated by the addition of 20% FBS for 30 or 60 min. Then, the cells were fixed, collected after wash, and suspended in 150 l of the lysis buffer containing Protease Inhibitor Mixture (Sigma). The soluble extract, prepared by sonication, was diluted and used for immunoprecipitation, in which antigen-purified anti-NF45, NF90/NF110, NF110, and RPB2 antibodies as well as non-immune rabbit IgG (Sigma) were mixed with the diluted extract. After the addition of recombinant Protein A-Sepharose CL-4B (GE Healthcare), the mixture was rotated for 1 h. The mixture was then washed successively in low salt buffer, high salt buffer, LiCl buffer, and Tris-EDTA buffer, and DNA was isolated for quantitative PCR analyses. The primers used for quantitative PCR were as follows: 5Ј-GACTGAAACGCTATTTCACTGTAC-3Ј and 5Ј-TCCCACCACTACAAAACCTGAGAG-3Ј for the Ϫ3,400 region, 5Ј-CCAACCTCAGTCCTAAAGTTTCTC-3Ј and 5Ј-GGTCATTGTCCAGCAATCTGGAAC-3Ј for the Ϫ600 region, 5Ј-GCGAGCTGTTCCCGTCAATCCCTC-3Ј and 5Ј-AGAC-CTTCCGCGTGTAGGATTTCG-3Ј for the Ϫ300 region, 5Ј-TTGAAAGCCTGGGGCGTAGAGTTG-3Ј and 5Ј-GGCTC-TATCCAGTCTTCTCAGTTG-3Ј for the transcription start site, 5Ј-GGTTAGAAAAACTGCTTCACCGAG-3Ј and 5Ј-TTTCTCTTCCTCTTCAGGAGATAG-3Ј for the ϩ1,600 region, 5Ј-ATTAACCTGGTGCTGGATTGTATC-3Ј and 5Ј-ATTGACGCTGAAGGACTACAGTAC-3Ј for the ϩ3,000 region, 5Ј-TCCTAAGCCAGTGAGTGGCACAGC-3Ј and 5Ј-ACACAACTATGTCAGAAGCAAATG-3Ј for the transcription start site of the ␤-globin gene, and 5Ј-ACCCTTGGACCC-AGCGGTACTTTG-3Ј and 5Ј-AAGGTGCCCTTGAGGCTG-TCCAAG-3Ј for the ϩ400 region of the ␤-globin gene. For PCR reactions, Power SYBR Green PCR master mix and ABI PRISM 7900HT Sequence Detection System (Applied Biosystems) were utilized.
GST Pulldown Assays-GST fusion proteins were expressed by infecting insect cells with the appropriate baculoviruses. When expressing GST-fused NF45-NF90 or NF45-NF110, the cells were co-infected with the baculovirus expressing GST-NF45 and that expressing either NF90 or NF110. GST-VP16 was expressed in Escherichia coli as described (37). GST pulldown assays for purified proteins, the TREF fraction, or HeLa nuclear extract were performed essentially as described (38) except that the bound proteins were eluted from the Sepharose beads prebound with GST activators using an equal volume of BC(1000) containing 0.5 mM PMSF and 1 mM DTT.
RNA Interference (RNAi) Assays-ϳ5 ϫ 10 5 C3H10T1/2 cells in a 35-mm culture dish were transfected with Stealth RNAi TM siRNA for NF90/NF110 (Invitrogen, MSS205464) or a negative control duplex (Invitrogen, medium GC duplex) using Lipofectamine TM 2000 (Invitrogen), and then the cells were cultured in fresh DMEM containing 10% FBS for 48 h to knock down NF90/NF110 expression before cell lysates were prepared. To induce c-fos transcription, the siRNA-transfected cells were starved in DMEM containing 0.5% FBS for 24 h before the addition of the medium containing 20% FBS, 25 ng/ml anisomycin, or 5 ng/ml EGF. The cells were then collected at the indicated time points to prepare total cellular RNAs and cell lysates. The efficiency of the NF90/NF110 knockdown was estimated by immunoblots of the cell lysates with antibodies against NF90/NF110. The isolated total cellular RNAs were used for measuring the amount of the c-fos mRNA by quantitative RT-PCR. The primers for amplifying the c-fos mRNA were 5Ј-GAGGACCTTACCTGTTCGTGAAA-3Ј and 5Ј-CCAGATGTGGATGCTTGCAA-3Ј. The primers for the ␣-tubulin RNA were 5Ј-GGTTCCCAAAGATGTCAAT-GCT-3Ј and 5Ј-CAAACTGGATGGTACGCTTGGT-3Ј. The amounts of the ␣-tubulin mRNA, which were essentially invariable under the tested conditions, were used for normalization.
Cell Transfection and Reporter Assays-3.3 ϫ 10 4 HEK293 cells in a 48-well plate were transfected with 356 ng of plasmid DNA using Lipofectamine TM 2000 (Invitrogen), and the cells were washed and then lysed. The lysates were used for assaying the luciferase and ␤-galactosidase activities according to the manufacturer's instructions (Promega). Each experiment was performed at least three times, and the values of the luciferase activities were normalized with those of ␤-galactosidase activities.

Results
The NF45-NF90 Complex Is a Component of the TREF Activity-To identify the molecular component(s) of the TREF activity, we purified the TREF fraction derived from HeLa nuclear extracts (Fig. 1A) by monitoring its coactivator activity using the c-fos template, pfMC2AT (10). The purification through four different columns revealed that TREF could be separated into three distinct activities termed TREF␣, TREF␤, and TREF␥ (Fig. 1A). TREF␣ was identified as heterogeneous nuclear ribonucleoprotein R, an RNA-binding protein that possesses three RNA recognition motifs and an arginine-glycineglycine-rich box (10). In this study we purified TREF␥ to near homogeneity, and its peak fractions revealed two major polypeptides that were identified as NF45 and NF90 by mass spectrometric analyses (Fig. 1, A and B). To obtain direct evidence of the transcriptional activity of NF45 and NF90, FLAG-tagged NF45 and NF90 were co-expressed in insect cells and purified to near homogeneity (Fig. 1E). Because NF45 and NF90 interact with each other to form a stable complex, we will hereafter refer to their complex as NF45-NF90. Recombinant NF45-NF90 stimulated transcription from the c-fos promoter in a dose-dependent manner and to a similar but slightly lesser extent as compared with native TREF␥ (Fig. 1F), eliciting a ϳ5-fold stim-ulation over the activator-dependent c-fos transcription (lanes 2 and 6).
The Transcriptional Activities of the NF Complexes Vary on Different Promoters-Because TREF stimulates transcription from the c-fos promoter but has little stimulatory activity toward GAL4-VP16-dependent transcription (10), we tested if the NF complexes display a selective activity toward the c-fos promoter as well. As shown in Fig. 2A, in vitro transcription assays showed that 30 and 90 ng of NF45-NF90 stimulated activator-dependent transcription from the c-fos promoter by 1.5-and 4.5-fold over that in the absence of NF45-NF90 (compare lane 2 versus lanes 4 and 6). In contrast, NF45-NF90 stimulated GAL4-VP16-dependent transcription by 2.1-and 1.9fold under the same condition ( Fig. 2A, compare lane 8 versus lanes 10 and 12). Thus, although TREF does not stimulate transcription from the 5xG promoter at all (10), NF45-NF90 does   OCTOBER 30, 2015 • VOLUME 290 • NUMBER 44 do so, albeit to a lesser extent than the c-fos promoter, indicating that some components in TREF other than NF45-NF90 may repress transcription from the 5xG promoter. In cell-based assays using the c-fos promoter fused to the luciferase gene, NF45-NF90 and NF45-NF110 stimulated c-fos transcription by ϳ3.5-fold (Fig. 2B). Using the same assays, we found that other promoters were also stimulated by both NF45-NF90 and NF45-NF110 to different extents (2-5-fold) (Fig. 2C). In contrast, the 5xG promoter fused to the luciferase gene was stimulated by Ͻ2-fold (Fig. 2B). These in vitro and in vivo studies indicate that the coactivator activities of NF45-NF90 and NF45-NF110 are less effective toward GAL4-VP16-dependent transcription than c-fos transcription and that their transcriptional effects vary among different promoters.

NF45-NF90 and NF45-NF110 Function as Transcriptional Coactivators
The NF Complexes Cooperate with PC4 and Mediator-We also tested if NF45-NF90 cooperates with PC4 and Mediator to stimulate c-fos transcription in vitro, as observed for TREF (10). When NF45-NF90, PC4, and Mediator were tested individually, each coactivator stimulated activator-dependent transcription from the c-fos promoter by 3-7-fold, with the strongest stimulation displayed by PC4 (Fig. 2D, compare lane 2 versus lanes [3][4][5]. NF45-NF90 showed a somewhat weaker activity than PC4, yet it had a stronger activity than Mediator (Fig. 2D, lanes 3-5). Although their individual transcriptional stimulation was moderate, the combined effects by two of the coactivators were more pronounced (Fig. 2D, lanes 6 -8), indicating that they cooperate with each other to stimulate c-fos transcription. Indeed, when all three coactivators were included simultaneously in the reaction (Fig. 2D, lane 9), the level of c-fos transcription increased by 54-fold over that in the absence of any coactivator (compare lane 2 versus lane 9) and by 543-fold over its basal level in the absence of activators and coactivators (compare lane 1 versus lane 9). These results demonstrate that both NF45-NF90 and NF45-NF110 cooperate with PC4 and Mediator to stimulate c-fos transcription in vitro.
Double-stranded RNA Binding Activity Is Dispensable for the Coactivator Activity of the NF Complexes in Vitro-Previous in vivo studies showed that the dsRNA binding activity of NF110 is important for increasing the expression of the reporter gene in cell-based assays (31). In these assays, however, the increased expression could be caused by the mRNA stabilization by NF90 and NF110 (14,16,17,19). Moreover, NF110, which has a weaker dsRNA binding activity than NF90, showed a stronger effect on gene expression than NF90 in similar cell-based assays (31,33). To clarify the relationship between dsRNA binding and transcriptional activities of the NF complexes, we created NF90(FA), in which two phenylalanine residues, Phe-432 and Phe-559, were mutated into alanine and NF90(AP), in which two alanine residues, Ala-458 and Ala-588, were mutated into proline as well as the corresponding NF110 mutants described previously (31). As expected, NF90(FA), NF90(AP), NF110(FA), and NF110(AP) showed virtually no binding to 32 P-labeled poly(I)⅐poly(C) in Northwestern assays (Fig. 3A). When introduced into C3H10T1/2 cells together with NF45, the dsRBM mutants, especially NF90(AP) and NF110(AP), showed lower levels of the reporter gene expression than their wild-type counterparts (Fig. 3B), in essential agreement with the previous study (31).
However, when these mutants were tested directly in in vitro transcription assays, in which there is no post-transcriptional effect, NF45-NF90(FA), NF45-NF90(AP), NF45-NF110(FA), and NF45-NF110(AP) stimulated transcription from the c-fos promoter to the levels comparable with those of their wild-type counterparts (Fig. 3C), indicating that the intact dsRBMs are dispensable for the direct coactivator activity of the NF complexes observed in vitro. We also tested the effect of RNA on the coactivator activities of the NF complexes by including poly(I)⅐poly(C) or total cellular RNA. Increasing the amounts of poly(I)⅐poly(C) (Fig. 3D) or total cellular RNA (Fig. 3E) did not augment the coactivator activities of the NF complexes but instead gradually repressed the levels of transcription (Fig. 3D). This repressive effect of poly(I)⅐poly(C) (Fig. 3D) and total cellular RNA (Fig. 3E) is unrelated to the dsRNA binding activity of the NF complexes because the repression was also observed   OCTOBER 30, 2015 • VOLUME 290 • NUMBER 44 regardless of their dsRNA binding activities. Thus, although there remains a possibility that some specific RNAs such as small NF90-associated RNAs (40) regulate the activities of the NF complexes via their dsRBMs, these results indicate that the NF complexes possess a coactivator activity that is independent of their dsRNA binding activity.

NF45-NF90 and NF45-NF110 Function as Transcriptional Coactivators
NF90 and NF110 Are Important for Rapid Induction of c-fos Transcription-Next, to obtain the in vivo evidence that NF45-NF90 and NF45-NF110 are required for c-fos transcription, we knocked down the endogenous NF90/NF110 and analyzed the level of c-fos transcription. When C3H10T1/2 cells were transfected with siRNA specific for the mouse NF90/NF110 gene, the amounts of NF90 and NF110 decreased to undetectable levels, whereas control siRNA did not have any effect (Fig. 4A). We then analyzed induction of the c-fos mRNA in cells stimulated with 20% fetal bovine serum after 24 h of serum starvation. Fig. 4B shows that the knockdown reduced the level of the c-fos mRNA, especially during 15-30 min after serum stimulation and that the level of the c-fos mRNA returned virtually to that in the control cells after 90 min. Because serum may contain various stimuli that can induce c-fos transcription, we tested two well defined stimuli, anisomycin and EGF, whose effects are mediated by the JNK/p38 and Erk1/2 subfamilies of MAP kinases, respectively. As shown in Fig. 4C, RNAi knockdown of NF90/NF110 had no significant effect on anisomycin-induced c-fos transcription. By contrast, EGF induction of the c-fos gene was markedly reduced upon NF90/NF110 knockdown (Fig.  4D). In all the RNAi knockdown experiments, the levels of the ␣-tubulin mRNA remained comparable in cells treated with control or NF90/NF110-specific siRNA (Fig. 4, E-G). Thus, NF90 and NF110 play an important role in rapidly inducing c-fos transcription, especially in response to serum or EGF stimulation.
The NF Complexes Dynamically Occupy the c-fos Gene-In view of the multiple roles for the NF complexes in transcription (Fig. 4) and posttranscriptional processes (39, 41), we wished to know the temporal occupancy of NF45, NF90, and NF110 on the c-fos gene during its transcriptional induction. C3H10T1/2 cells were starved and then treated with 20% FBS for 0, 30, and 60 min before the cells were fixed with formaldehyde, and the chromatin was immunoprecipitated by anti-NF45, anti-NF90/ NF110, or anti-NF110 antibodies (Fig. 1C). The precipitated DNA was then used for quantitative PCR with sets of primers designed to amplify ϳ100-bp regions around Ϫ3,400, Ϫ600, Ϫ300, ϩ1, ϩ1,600, and ϩ3,000 of the c-fos gene (Fig. 5A). In addition, sets of primers for the ϩ1 and ϩ400 regions of the ␤-globin gene, which is not expressed in C3H10T1/2 cells, were used as controls.
Under the starved condition (0 min), the NF proteins were present from the upstream regions (Ϫ600 and Ϫ300) to the transcription start site (ϩ1), whereas much lesser amounts of the NF proteins were observed in the far upstream region (Ϫ3,400) and the downstream regions (ϩ1,600 and ϩ3,000) within the c-fos gene (Fig. 5B). RNAPII also showed a similar occupancy to those of the NF proteins. As expected, 30 min after serum addition when the c-fos gene is induced maximally,    OCTOBER 30, 2015 • VOLUME 290 • NUMBER 44 the RNAPII occupancy increased dramatically at the start site (ϩ1) and in the downstream regions (ϩ1,600 and ϩ3,000) of the c-fos gene. Concomitantly, the occupancies of the NF proteins increased in a similar manner to that of RNAPII in the coding regions (ϩ1,600 and ϩ3,000), raising the possibility of some physical association between RNAPII and the NF proteins. Whereas the occupancies of the NF proteins remained largely unchanged in the upstream regions (Ϫ600 and Ϫ300) and at the start site (ϩ1), slight reductions in the occupancies of NF90 and NF110 were observed in the Ϫ300 and ϩ1 regions, which could possibly reflect transfer of the NF proteins from the promoter to the elongating RNAPII. After 60 min the occupancies of the NF proteins and RNAPII returned rapidly to the pre-induction levels. As the control, no significant occupancies of the NF proteins and RNAPII were observed on the ␤-globin gene in C3H10T1/2 cells. These experiments demonstrate that the NF proteins are associated with the regulatory and core promoter regions of the c-fos gene before induction and that the occupancies of the NF proteins undergo dynamic changes during induction of c-fos transcription. These results suggest that the NF proteins, probably as the complexes of NF45-NF90 and NF45-NF110, may travel along the gene with elongating RNAPIIs, further substantiating their dual roles in transcription and posttranscriptional processes.

NF45-NF90 and NF45-NF110 Function as Transcriptional Coactivators
NF45-NF90 and NF45-NF110 Stimulate Transcription through Both the Upstream Regulatory Elements and the Core Promoter-Our in vitro analyses indicated that the NF proteins stimulate both activator-dependent and basal transcription ( Figs. 1 and 2). In addition, we observed a marked occupancy of the NF proteins on the c-fos gene from the Ϫ600 to ϩ1 region (Fig. 5), which encompasses from the upstream regulatory elements to the core promoter. Thus, to delineate the c-fos promoter region that mediates the transcriptional activity of the NF proteins, we tested a series of deletion mutants of the c-fos regulatory elements (Fig. 6A). Deletion of the upstream region from Ϫ733 to Ϫ500 did not affect the luciferase activity at all. However, further deletion from Ϫ500 to Ϫ55, including the known regulatory elements (SRE, FAP-1, and CRE), reduced the level of the luciferase activity dramatically regardless of the presence of the NF complexes, in accord with the critical roles of these elements in c-fos transcription. Importantly, the 2.5-3.0-fold stimulation by NF45-NF90 or NF45-NF110 observed for the c-fos promoter (from Ϫ733 to ϩ158) decreased gradually as the region downstream from Ϫ500 was removed progressively. These results indicate that the NF complexes act broadly through the regulatory elements between Ϫ500 and Ϫ55, the region that conforms approximately with the occupancy of the NF proteins on the endogenous c-fos gene (Fig. 5). Moreover, introduction of point mutations into the individual regulatory elements (SRE, FAP-1, and CRE) also decreased the luciferase activity while still allowing an ϳ1.8-fold stimulation in the presence of NF45-NF90 and NF45-NF110 (Fig. 6B). Finally, the reporter gene with only the core promoter (Ϫ55 to ϩ158) was stimulated ϳ1.8-fold by NF45-NF110 but not by NF45-NF90, indicating that NF45-NF110 has a direct stimulatory effect on the core promoter as well. These results indicate that the NF complexes stimulate c-fos transcription through   both the upstream regulatory elements and the core promoter, in agreement with the occupancies of the c-fos enhancer/promoter region by the NF proteins in cells.
Multiple Interactions of the NF Proteins with Activators and the Basal Transcriptional Machinery-The coactivator function of the NF complexes and their occupancy on the c-fos gene on its enhancer/promoter region suggested that they might interact directly with activators and the basal transcriptional machinery. We, therefore, expressed and purified recombinant FLAG-tagged NF proteins and their complexes (NF45-NF90 and NF45-NF110) as well as the firefly luciferase protein for a control. The tagged proteins were then allowed to interact with various GST-fused activators pre-bound to glutathione-Sepharose beads, and the bound proteins were eluted and analyzed by immunoblotting with anti-FLAG M2 antibody. As shown in Fig. 7A, NF45-NF90 and NF45-NF110 interacted with CREB, Elk-1, ATF1, and USF to varying degrees, and NF45-NF90 interacted with Oct-1/POU2F1. The observed interactions are specific because the luciferase protein did not interact with any activators, and none of the NF proteins or their complexes interacted with GST alone (Fig. 7A). Consistent with its role as a functional regulator of NF90 (32), NF45 appeared to modulate the interactions of NF90 and NF110 with the activators, although NF45 did not interact with any activators by itself. For instance, NF45 enhanced the interaction between NF90 and Oct-1/POU2F1, whereas it diminished the interaction between NF110 and ATF1 (Fig. 7A). Finally, the NF proteins showed very weak interactions with the VP16 activation domain, which agrees with their minimal stimulatory activity toward GAL4-VP16-dependent transcription (Fig. 2, A and B).
As these assays were performed with purified proteins, we further examined the specificity by testing if the activators interact with the NF complexes in the extracts. Instead of purified NF complexes, we first incubated the TREF fraction ( Fig.  1A) with GST-fused activators bound to glutathione-Sepharose beads. After washing, the bound proteins, which corresponded to 1.9 -2.3% of the proteins contained in the TREF fraction (data not shown), were tested for the presence of the NF complexes by immunoblots using anti-NF90/NF110 or anti-NF45 antibodies. As shown in the left panel of Fig. 7B, the bound fractions contained the NF complexes. We then used HeLa nuclear extract to perform the same experiments, and 1.0 -1.5%   OCTOBER 30, 2015 • VOLUME 290 • NUMBER 44 of the proteins were bound to GST-fused activators (data not shown). Similar to the results with the TREF fraction, the NF complexes were detected in the bound fractions (Fig. 7B, right  panel). These results show that the NF complexes interact specifically with the activators, albeit to different degrees. Next, we tested the interactions between the NF proteins and the components of the basal transcriptional machinery.

NF45-NF90 and NF45-NF110 Function as Transcriptional Coactivators
Because we wished to test if the NF proteins interact with multiprotein complexes including TFIID and TFIIH, we prepared NF45, NF90, and NF110 as well as their complexes (NF45-NF90 and NF45-NF110) as GST fusion proteins. In NF45-NF90 and NF45-NF110, only NF45 was expressed as a GST fusion protein, which formed heterodimers with non-tagged NF90 or NF110. As shown in Fig. 7C, the NF proteins showed interactions with TFIIB, TFIID, and TFIIH but not with TFIIA, TFIIE, and TFIIF. Although NF45 did not interact with any activators (Fig. 7A), it interacted with TFIID and TFIIH as efficiently as did its heterodimeric partners, NF90 and NF110 (Fig. 7C). NF45 reduced the interaction of NF90 with TFIIB, again indicating a possible regulatory role for NF45 in the context of the NF complexes. Collectively, the NF proteins have multiple interactions with both activators and the basal transcriptional machinery, in agreement with their direct transcriptional function and their occupancies on both the upstream regulatory elements and the core promoter (Figs. 1, 2, and 5).
Finally, despite the similar dynamic occupancies of RNAPII and the NF proteins on the c-fos gene (Fig. 5), neither individual NF proteins (NF45, NF90, and NF110) nor their complexes (NF45-NF90 and NF45-NF110) showed any interaction with RNAPII under the condition where the capping enzyme, RNGTT, bound specifically to the phosphorylated IIO (Fig.  7D). This result suggests that interactions between the NF complexes and RNAPII, if any, are likely to be indirect.

Discussion
The NF Complexes Are Multifunctional Coactivators-In this study we purified TREF␥ and ascribed its activity to the NF complexes, which consist of either NF45 and NF90 or NF45 and NF110 (39). The NF complexes were identified initially as a DNA binding factor that binds the ARRE site (26,27), the CCAAT-binding site (28), and an MHC-I site (29). However, independent studies identified the NF proteins by means of their RNA binding activity and named them as ILF3 (42), DRBP76 (13), or TCP80 (23). Indeed, NF90 and NF110 possess two dsRBMs, which, as the name implies, is a modular nucleic acid binding motif that binds duplexed RNA but not ssRNA, dsDNA, or single-stranded DNA (11)(12)(13). Consistent with the presence of dsRBMs, the NF complexes do not bind dsDNA under the conditions where their binding to dsRNA can be demonstrated unambiguously (33,43,44). Given the absence of any consensus DNA sequence among the presumptive NF binding DNA elements described above (26 -29), it remained to be defined if the NF complexes regulate transcription by binding directly to specific DNA sequences (45). Further complicating their role in transcription is that the NF complexes stabilize mRNA through binding to specific RNA sequences (14,19). Thus, the transcriptional stimulation by the NF complexes observed in cell-based assays may be attributed to posttran-scriptional effects (39,41). Thus, despite several reports that the NF complexes increase the level of mRNA in vivo (30 -35), it has been unclear whether they have a direct function in transcription.
Combined with in vitro transcription assays, our unbiased identification and purification of the TREF␥ activity from cell extract demonstrate beyond the presumed role in transcription by earlier studies that the NF complexes possess a direct function in transcription from the c-fos promoter. Earlier (31) as well as our own studies (Fig. 3B) using cell-based assays show that the mutations of dsRBMs drastically diminish the ability of the NF complexes to stimulate transcription as measured by mRNA levels. However, the NF complexes stimulated transcription in vitro even when its dsRBMs were mutated (Fig. 3C), demonstrating that the NF complexes possess a direct transcriptional function that is separate from the posttranscriptional functions that require intact dsRBMs. Consistent with the direct transcriptional function, the NF proteins interact physically with the components of the basal transcriptional machinery, notably with TFIID and TFIIB that are pivotal in transcriptional regulation. Because the NF complexes exert the transcriptional effect through multiple cis-regulatory elements (Fig. 6), probably by interacting with various activators (Fig.  7A), they are likely to function as a transcriptional coactivator. Given that knockdown of NF90/NF110 attenuates the induction of c-fos transcription upon serum or EGF stimulation (Fig.  4D) together with their previously demonstrated roles in posttranscriptional processes (39,41), the NF complexes may coordinate transcription and post-transcriptional processes to elicit rapid response of inducible genes.
Coordination between transcription and posttranscriptional processes has been reported for numerous proteins (46,47). For example, PGC-1 and CoAA, initially identified as transcriptional coactivators, were later shown to function in posttranscriptional processes as well (48,49). PGC-1 possesses an RNA recognition motif and an RS domain, both of which are RNA binding domains that interact with splicing factors and are essential for coupling transcription and mRNA processing (48,50). Similarly, two RS domains in CoAA play a role in linking transcription and splicing in a promoter-preferential manner (50). Furthermore, another coactivator, RNA helicase A, possesses dsRBMs and an arginine-glycine-glycine-rich domain (41,51), which are RNA binding motifs also present in NF90 and NF110. Interestingly, RNA helicase A copurifies with NF45-NF90 through several column chromatographies and interacts directly with NF90 and NF110 (11,33), suggesting a functional partnership between the NF complexes and RNA helicase A. On the whole, the prevalence of RNA binding motifs in these coactivators suggests that the functional coordination between transcription and posttranscriptional processes may entail direct interactions between these RNA binding coactivators and the RNA transcript emerging from RNAPII.
How the NF complexes participate in multiple aspects of RNA metabolism, however, remains to be elucidated. One possibility is that NF90 and NF110 have separate roles in distinct steps of gene expression; for example, NF110 may function at earlier steps than does NF90. For instance, although NF110 and NF90 are distributed in both nucleus and cytoplasm (14, 27, 28, 33,52), NF110 tends to be more abundant in nucleus, whereas NF90 is more readily observed in cytoplasm. In addition, NF110 appears to stimulate gene expression slightly better than NF90 in vivo (31,33) and in vitro (Fig. 1G); by contrast, NF90, but not NF110, is implicated in mRNA export and stabilization (14,19). An alternative, but not mutually exclusive, possibility is that the NF complexes serve as a scaffold that assembles different proteins to acquire functions besides their intrinsic activities. Consistent with this, the NF proteins have been shown to interact with multiple proteins such as Yin Yang 1 (30,53), protein kinase R (13,30,54,55), protein-arginine methyltransferase 1 (11,33,53,56), RNA helicase A (11,33), survival of motor neuron (30), and FUS/TLS RNA-binding protein (30) and have been isolated as components of RNAPII-interacting proteins as well as integral components of H complex (57,58).

NF45-NF90 and NF45-NF110 Function as Transcriptional Coactivators
Although we identified the NF complexes from their transcriptional activity on the c-fos promoter, our own data and those from others (26,32,33,35) indicate that the NF complexes stimulate transcription of other genes as well. In NF90 Ϫ/Ϫ knock-out mice, T cells show severely impaired expression of IL-2, which is regulated at multiple steps of gene expression by the NF complexes (59). Moreover, up-regulation of NF90 in cancer cells increases mRNAs of interferon-regulated genes, which are also prototypical inducible genes (34). These observations make it highly likely that a subset of inducible genes require the NF complexes for their prompt expression, which requires coordination among the multiple steps from transcription to translation.
Dynamic Occupancy of the NF Complexes within the c-fos Gene-The NF complexes have been shown to possess both nuclear export and nuclear localization signals, which permit shuttling between nucleus and cytoplasm (14,52,60,61). In a significant extension of the earlier studies that showed the occupancy of NF90 on the IL-2 gene (59) and their association with chromatin (33, 62), our ChIP analyses revealed that the NF complexes are present on the intragenic and enhancer/promoter regions of the c-fos gene. Moreover, the occupancy of the NF proteins in the intragenic region parallels both temporally and spatially with that of RNAPII during c-fos induction (Fig. 5). This dynamic occupancy implies that the NF complexes move along the c-fos gene in association with transcribing RNAPII, which is entirely consistent with their dual roles in transcription and post-transcriptional processes (Fig. 8).
The dynamic occupancy along the c-fos gene raises the intriguing possibility that the NF complexes may act on the very same transcript whose transcription initiation they stimulated in the first place. This would explain why cell-based assays showed that the mutations of the dsRBMs drastically reduce the stimulatory effect of the NF complexes on transcription in vivo (Fig. 3B) (31). Because the mutated dsRBMs can compromise the posttranscriptional effects of the NF complexes, their stimulatory effect on transcription initiation might have been largely cancelled out in the cell-based assays, giving a misleading impression that the transcriptional activity requires dsRNA binding.
The NF complexes may be recruited to the promoter region via direct interactions with the activators and/or the preinitiation complex (Fig. 8). NF90 and NF110 showed numerous interactions with the activators and the basal transcriptional machinery (Fig. 7), supporting the ChIP data that showed the occupancy of the NF complexes on the enhancer/promoter region (Fig. 5). Although the ChIP data also suggest that the NF complexes may be transferred from the promoter to the elongating RNAPII, if and how this occurs remains unclear. However, the transfer of the NF complexes may be critical for their functions. Indeed, when NF90 is tethered to the promoter region as Gal4-NF90, Gal4-NF90 appears to act like a dominant negative form and inhibits transcription (32), presumably because Gal4-NF90 cannot be transferred from the promoter region to the elongating RNAPII.
Given the parallel occupancies of the NF proteins and RNA-PII, it was somewhat unexpected to find that the NF proteins do not interact directly with purified RNAPII irrespective of the phosphorylation status of its CTD (Fig. 7D). Three lines of evidence, however, strongly indicate an indirect physical association between RNAPII and the NF complexes, probably mediated by their associated proteins and/or RNA transcript. First, a proteomic analysis revealed that the NF complexes are present in H complex together with heterogeneous nuclear ribonucleoproteins and other types of RNA-binding proteins (63). H complex is believed to bind nascent RNA that has just emerged from RNAPII before the assembly of splicing factors (57,58). Second, the NF complexes are co-precipitated with the phosphorylated (and thus elongating) RNAPII and various RNA-binding proteins by an anti-CTD antibody (64). Third, the NF complexes are immunoprecipitated by an anti-Chd1 antibody, which also  precipitates RNAPII and numerous proteins involved in splicing and other RNA processing steps (65). Interestingly, Chd1 binds the H3-K4me3 (66,67), whose presence on the chromatin (68,69) overlaps with the occupancy of the NF complexes (Fig. 5).
In conclusion, this study reports that the NF complexes have a direct function in stimulating transcription as a coactivator. Given their roles in multiple steps of gene expression and the dynamic occupancy on the c-fos gene, it will be essential to elucidate the structure and function of the proteins that associate with the NF complexes during distinct steps of gene expression.
Author Contributions-T. N. and A. F. designed and performed all the experiments. M. S. and K. N. provided technical assistance and experimental materials. T. N., A. F. and K. H. analyzed the data, wrote the paper, and prepared the figures. All authors reviewed the results and approved the final version of the manuscript.