Biochemical Journal

Research article

The variable N-terminal region of DDX5 contains structural elements and auto-inhibits its interaction with NS5B of hepatitis C virus

Sujit Dutta , Garvita Gupta , Yook-Wah Choi , Masayo Kotaka , Burtram C. Fielding , Jianxing Song , Yee-Joo Tan

Abstract

RNA helicases of the DEAD (Asp-Glu-Ala-Asp)-box family of proteins are involved in many aspects of RNA metabolism from transcription to RNA decay, but most of them have also been shown to be multifunctional. The DEAD-box helicase DDX5 of host cells has been shown to interact with the RNA-dependent RNA polymerase (NS5B) of HCV (hepatitis C virus). In the present study, we report the presence of two independent NS5B-binding sites in DDX5, one located at the N-terminus and another at the C-terminus. The N-terminal fragment of DDX5, which consists of the first 305 amino acids and shall be referred as DDX5-N, was expressed and crystallized. The crystal structure shows that domain 1 (residues 79–303) of DDX5 contains the typical features found in the structures of other DEAD-box helicases. DDX5-N also contains the highly variable NTR (N-terminal region) of unknown function and the crystal structure reveals structural elements in part of the NTR, namely residues 52–78. This region forms an extensive loop and an α-helix. From co-immunoprecipitation experiments, the NTR of DDX5-N was observed to auto-inhibit its interaction with NS5B. Interestingly, the α-helix in NTR is essential for this auto-inhibition and seems to mediate the interaction between the highly flexible 1–51 residues in NTR and the NS5B-binding site in DDX5-N. Furthermore, NMR investigations reveal that there is a direct interaction between DDX5 and NS5B in vitro.

  • auto-inhibition
  • DDX5
  • DEAD-box family
  • hepatitis C virus
  • NS5B
  • N-terminal region

INTRODUCTION

Hepatitis C is an infectious disease affecting an estimated 150–200 million people worldwide. Infection is caused by the HCV (hepatitis C virus), which can often lead to cirrhosis, steatosis and hepatocellular carcinoma. HCV is an enveloped single-stranded positive-sense RNA virus in the family Flaviviridae [1]. The viral genome is encoded by three structural proteins (core, E1 and E2) and seven NS (non-structural) proteins (p7, NS2, NS3, NS4A, NS4B, NS5A and NS5B). NS3, a serine protease and RNA helicase, along with NS5B, an RNA-dependent RNA polymerase, are key enzymes required for HCV replication and therefore common targets for antiviral agents. Viral replication requires interaction between RNA, viral and host proteins [1]. NS5B has been shown to interact with a growing number of host proteins, including eukaryotic initiation factor 4AII [2], cellular vesicle membrane transport protein VAP-33 [3], nucleolar phosphoprotein nucleolin I [4] and the DEAD (Asp-Glu-Ala-Asp)-box RNA helicase DDX5 [5].

DDX5, also referred to as p68, is a nuclear prototypic member of the DEAD-box family of proteins. The DEAD-box family belongs to helicase SF2 (superfamily 2), which also includes DEAH, DExH and DExD families (for recent reviews see [68]). DEAD-box helicases share nine conserved motifs that are clustered in a central region and possess highly variable N- and C-termini. The central region of DEAD-box helicases can be organized into two domains; domain 1 consists of motifs Q, I (Walker A), II (Walker B, DEAD-box), Ia, Ib and III, whereas domain 2 consists of motifs IV, V and VI. Motif Q, which is unique to DEAD-box helicases, forms the nucleotide-binding site together with motifs I and II. Motif III has been implicated in linking ATP binding and hydrolysis with the helical activity. The remaining motifs are involved in RNA binding. The functions of the variable N- and C-terminal regions are not fully characterized, but they are thought to interact with RNA substrates or cofactors so as to confer specificity or fine modulation of function.

DDX5 was first identified by its immunological cross-reactivity with a monoclonal antibody to the large T-antigen of simian virus 40 [9]. Co-purification of DDX5 with splicesomes initially suggested a role in RNA splicing and this was subsequently confirmed when DDX5 was shown to be an essential splicing protein acting at the U1 snRNA (small nuclear RNA) 5′ splice site [10]. DDX5 has also been shown to be involved in RNA export, ribosome assembly, translation and RNA degradation [6,11,12]. However, there is a growing body of evidence suggesting DDX5 has an additional role as a transcriptional co-activator for different genes [13].

DDX5 has been shown to interact with HCV NS5B, and the knockdown of DDX5 by RNA interference caused a reduction in the transcription of negative-strand HCV RNA [5]. Furthermore, the overexpression of NS5B has been shown to result in the redistribution of DDX5 from the nucleus to the cytoplasm [5] and, interestingly, DDX5 was also found to have undergone similar translocation in HCV-infected cells [14]. In addition, single nucleotide polymorphisms in the DDX5 gene have been shown to be significantly associated with increased risk of advanced fibrosis in HCV patients [15]. Although the precise role of DDX5 in HCV infection has yet to be defined, these studies suggest that it may play an important role in HCV replication.

In order to gain an understanding of the mechanism of interaction between DDX5 and NS5B, we expressed DDX5[1–305aa (amino acids)] and obtained the crystal structure of part of the variable NTR (N-terminal region) and domain 1 (residues 52–304) at 2.7 Å (1 Å=0.1 nm) resolution. The residues in DDX5 involved in the interaction with NS5B were also evaluated by using site-directed mutagenesis and co-immunoprecipitation experiments. Furthermore, NMR investigations were performed to determine whether there is a direct interaction between DDX5 and NS5B.

MATERIALS AND METHODS

Mammalian expression constructs

The open reading frame encoding human DDX5 was amplified from a spleen cDNA expression library [5]. DDX5 was cloned into pXJ40myc, a Myc-tagged plasmid derived from pXJ40 [16]. DDX5 deletion and substitution mutants were generated by PCR. NS5B was amplified from the cDNA of HCV-S1 of genotype 1b [17], and cloned into pXJ40flag, a FLAG-tagged plasmid derived from pXJ40 [16].

Transfection, co-immunoprecipitation and Western blot analysis

Huh-7 cells (human hepatoma, JCRB Cell Bank) were cultured in Dulbecco's minimal Eagle's medium supplemented with 10% (v/v) fetal bovine serum (HyClone Laboratories) and antibiotics, 10 units/ml penicillin and 100 μg/ml streptomycin (Sigma), and maintained at 37°C in 5% CO2.

Plasmids were transiently transfected into Huh-7 cells plated in 6-cm2 tissue culture dishes using Lipofectamine™ (Invitrogen), according to the manufacturer's instructions. The cells were harvested ~24 h post-transfection and lysed in 150 μl of RIPA buffer [50 mM Tris/HCl, pH 7.4, 150 mM NaCl, 0.5% NP-40 (Nonidet P40), 0.5% deoxycholic acid, 0.005% SDS and 1 mM PMSF] and subjected to three cycles of freezing and thawing. The cell lysates were spun down at 16 000 g for 20 min at 4°C to remove cell debris.

For co-immunoprecipitation, cell lysates were first incubated with 2 μg of anti-Myc polyclonal antibody (Santa Cruz Biotechnology) for 1 h at room temperature (24°C) before mixing with 20 μl of pre-washed Protein A–agarose beads (Roche). The mixtures were left to incubate overnight with agitation at 4°C. The protein-bound beads were then washed four times with 1 ml of RIPA buffer. Then 5× SDS loading buffer [0.3 M Tris/HCl, pH 6.8, 5% (w/v) SDS, 50% (v/v) glycerol, 0.1 M DTT (dithiothreitol) and 0.1% Bromophenol Blue] was added to the washed beads. The bound proteins were eluted from the beads by boiling the samples at 100°C for 5 min. Similarly, 5× SDS loading buffer was added to aliquots (containing 10 μg of total protein) of cell lysates before immunoprecipitation and the samples were boiled at 100°C for 5 min.

All protein samples were resolved by SDS/PAGE (7.5–12% gel), transferred on to Hybond-C nitrocellulose membranes (GE Healthcare) and blocked with 5% (w/v) non-fat dried skimmed milk powder in PBS with 0.05% Tween 20. The membranes were then incubated with primary antibodies [anti-Myc polyclonal antibody (Santa Cruz Biotechnology), anti-FLAG monoclonal antibody (Sigma) or anti-FLAG polyclonal antibody (Sigma)] followed by secondary antibodies conjugated with horseradish peroxidase. The proteins were then visualized using a SuperSignal® West Pico Substrate Kit (Pierce). The chemiluminescent protein signals were captured on Amersham Hyperfilm™ (GE Healthcare).

Expression and purification of bacterially expressed DDX5(1–305aa)

The N-terminal domain of DDX5 (residues 1–305) was expressed as a GST (glutathione transferase) (pGEX6p1, GE Heathcare) fusion protein. GST–DDX5(1–305aa) was expressed in Escherichia coli BL21-CodonPlus-RIL (Stratagene). Cultures were grown at 37°C in Terrific Broth and, on reaching a D600 of 0.8, cells were cooled to 16°C and induced with isopropyl β-D-thiogalactopyranoside to a final concentration of 0.2 mM. After an incubation period of 24 h, cells were harvested. Bacterial pellets were resuspended in lysis buffer (50 mM Tris/HCl, pH 7.4, 300 mM NaCl and 2 mM DTT) supplemented with Complete™ protease inhibitor (Roche). For purification, cells were subjected to sonication with an ultrasonic liquid processor for 4–8 30 s pulses with a 1 min interval between pulses. The lysate was cleared by centrifugation and loaded on to a 5 ml glutathione–Sepharose column (GE Healthcare) pre-equilibrated with lysis buffer. The column was washed to remove unbound material. Removal of the GST tag from the N-terminus of DDX5(1–305aa) was achieved by proteolytic cleavage using recombinant 3C protease (GE Healthcare). Cleaved protein was further purified by size-exclusion chromatography using a Superdex S200 column (GE Healthcare) pre-equilibrated in lysis buffer. Purified DDX5(1–305aa) was concentrated to 12 mg/ml using an Amicon Ultra 10 kDa cut-off filter (Millipore).

Crystallization and data collection

Crystals of recombinant DDX5(1–315aa) (12 mg/ml) were obtained at 15°C using the sitting-drop vapour diffusion method from 1:1 μl of protein and precipitant containing 2% (v/v) Tacsimate (pH 4.0), 0.1 M Bis-Tris (pH 6.5) and 20% (w/v) PEG [poly(ethylene glycol)] 3350. Crystals were transferred to a reservoir solution containing the precipitant with the addition of 30% glycerol, before flash-freezing in liquid nitrogen.

X-ray diffraction data from a single crystal of DDX5(1–305aa) was collected at the NSRRC (National Synchrotron Radiation Research Center, Taiwan) on beamline 13B1. Raw data were integrated and scaled using the HKL2000 program suite [18].

Structure determination and refinement

The structure of DDX5(1–305aa) was determined by the molecular replacement method, using the structure of domain 1 of DDX3X [19] (PDB code 2I4I) as the search model, with the program MolRep from the CCP4 suite [20]. The model was refined with CNS [21], and multiple rounds of manual fitting with the program O [22] using 2FoFc and FoFc electron density maps. The refined model consists of 57 water molecules, with a final R and Rfree of 27% and 29% respectively. There is one molecule in the asymmetric unit with 253 residues. The stereochemistry of DDX5(1–305aa) was checked with PROCHECK [23]. The refinement statistics are summarized in Table 1.

View this table:
Table 1 Data collection and refinement statistics

Values in parentheses indicate values in the highest-resolution shell.

In vitro translation and GST pull-down assay

pXJ40myc-DDX5(61–614aa) and pXJ40flag-NS5B plasmids were used as templates to produce the 35S-labelled proteins using the TNT® reticulocyte lysate system (Promega), according to the manufacturer's protocol.

Next, 30 μg of GST and GST fusion proteins bound on glutathione–Sepharose beads were washed three times with GST pull-down buffer (PBS with 0.5% Triton X-100, 0.5% NP-40, 1 mM EDTA, 1 mM EGTA and 0.4 mM PMSF). A 20 μl portion of 35S-labelled proteins, diluted with 80 μl of GST pull-down buffer, was added to the beads and the mixtures were incubated at room temperature for 1 h. The beads were then washed four times with GST pull-down buffer. 5× SDS loading buffer was added to the beads and the bound proteins were eluted from the beads by boiling at 100°C for 5 min. The samples were resolved by SDS/PAGE (10% gel). The resolved gel was soaked in a solution of 45% (v/v) methanol and 10% (v/v) acetic acid for 30 min and then placed in Amplify™ solution (GE Healthcare) for 30 min. The gel was placed on a piece of filter paper, covered with cling wrap and dried for 1 h at 80°C in a gel drier. The radiolabelled signals were captured on Amersham Hyperfilm™ (GE Healthcare).

Expression and purification of recombinant proteins for CD and NMR studies

The gene encoding the 570-residue NS5B was amplified by PCR from the pXJ40flag-NS5B plasmid, which contains NS5B of HCV genotype 1b [5], and then cloned into pET32a. The gene encoding DDX5(61–305aa) was excised by restriction enzyme digestion from the mammalian expression vector described above and cloned into pGEX6p1 vector (GE Heathcare). These two bacterial expression vectors were transformed into E. coli BL21 (DE3) Star (Invitrogen) cells. For expression of recombinant proteins, cells were grown in Luria–Bertani medium in the presence of ampicillin (100 μg/ml) at 37°C to reach an absorbance of 0.6 at 600 nm and subsequently induced with respective optimized isopropyl β-D-thiogalactopyranoside concentrations. Harvested cells were resuspended and lysed by sonication with an ultrasonic liquid processor for 4–8 30 s pulses with a 1 min interval between pulses in lysis buffer [50 mM Tris, pH 7.5, 500 mM NaCl, 10% (v/v) glycerol, 20 mM imidazole and 10 mM 2-mercaptoethanol] containing protease inhibitor cocktail (Roche). His-tagged NS5B and GST-tagged DDX5(61–305aa) proteins were purified by Ni2+-affinity chromatography (Qiagen) and glutathione sepharose chromatography (GE Healthcare) respectively under native conditions. The recombinant proteins were released from the fused tags by in-gel cleavage with either thrombin (for NS5B) or 3C protease [for DDX5(61–305aa)] and further purified by size-exclusion chromatography using a Superdex S200 column (GE Healthcare). The production of the isotope-labelled DDX5(61–305aa) protein for NMR studies followed a similar procedure, except that the bacteria were grown in M9 medium with the addition of (15NH4)2SO4 for 15N labelling as described previously [24,25]. The concentration of protein samples was determined by spectroscopy in the presence of denaturant [2426].

CD and NMR experiments

All CD experiments were carried out in a Jasco J-810 spectropolarimeter as previously described [25] at 25°C. The protein concentration was 20 μM in 2 mM phosphate buffer (pH 6.5) for all far-UV CD experiments.

NMR samples were prepared in 10 mM phosphate buffer in the presence of 10 mM DTT (pH 6.5). All NMR data were collected at 25°C on an 800-MHz Bruker Avance spectrometer equipped with a shielded cryoprobe as described previously [24,25]. For HSQC (heteronuclear single quantum coherence) characterization of the 15N-labelled DDX5(61–305aa), samples were prepared at a protein concentration of 100 μM. For NMR characterization of the binding interaction of DDX5(61–305aa) to NS5B, one-dimensional 1H NMR spectra of the DDX5(61–305aa) protein were acquired at a protein concentration of 20 μM in the absence or presence of the unlabelled NS5B at molar ratios of 1:1; 1:2 and 1:3 (DDX5/NS5B). NMR data were processed with NMRpipe [27] and analysed with NMRview [28].

RESULTS

DDX5 contains two independent NS5B-binding sites

To determine which domain(s) in DDX5 are involved in the interaction with NS5B, co-immunoprecipitation experiments were performed using deletion mutants of DDX5. FLAG-tagged NS5B was expressed in Huh-7, a liver cell line that supports HCV replication, together with Myc-tagged DDX5 proteins (Figure 1). The results showed that two non-overlapping fragments of DDX5, DDX5(61–305aa) and DDX5(306–614aa), bound to NS5B, indicating that there are two independent NS5B-binding sites in DDX5. In contrast, DDX5(1–305aa) showed no significant binding to NS5B, indicating that the deletion of the N-terminal 60 residues of DDX5(1–305aa) enhanced its binding to NS5B (Figure 1). DDX5(1–305aa) and DDX5(306–614aa) shall be referred to DDX5-N and DDX5-C in the present study.

Figure 1 Interaction of HCV NS5B with full-length and deletion mutants of DDX5 as determined by co-immunoprecipitation experiments

Huh-7 cells were transfected with cDNA constructs for expressing Myc-tagged GST (negative control), Myc-tagged full-length or deletion mutants of DDX5 and FLAG-tagged HCV NS5B. The cells were harvested at 16 h post-transfection, lysed and subjected to immunoprecipitation with anti-Myc polyclonal antibody and Protein A–agarose beads. The amount of FLAG-tagged NS5B that co-immunoprecipitated (co-IP) with the Myc-tagged proteins was determined by Western blot (WB) analysis with an anti-FLAG monoclonal antibody (top panel). The protein marked with an asterisk represents the heavy chain of the antibody used for immunoprecipitation (top panel). The amounts of Myc-tagged and FLAG-tagged proteins in the lysates before co-IP were determined by subjecting aliquots of the lysates to Western blot analysis (middle and bottom panels). Molecular mass in kDa is shown on the right-hand side.

Structural analyses of numerous DEAD-box RNA helicases revealed that they contain a central region composed of two conserved domains, termed domains 1 and 2, flanked by highly variable N- and C-terminal sequences [19,2933]. Bio-informational analysis revealed that DDX5-N and DDX5-C contain domains 1 and 2 respectively (results not shown).

Domain 1 of DDX5 is similar to that of other DEAD-box RNA helicases

To gain further insights into the interaction between DDX5 and NS5B, DDX5-N and DDX5-C were expressed as N-terminal GST fusions. GST was removed with the 3C protease, but five amino acids (GPLGS) remained fused to the N-terminus of these proteins. Purification of DDX5-C failed, as the protein underwent degradation when expressed in E. coli (results not shown). Purification of DDX5-N was successful and crystallization trials were performed. A GST-pull-down assay was also performed and the results showed that DDX5-N interacts with NS5B (Supplementary Figure S1 at http://www.BiochemJ.org/bj/446/bj4460037add.htm). DDX5-N was found to crystallize in space group I222 and the structure was solved by molecular replacement using the structure of domain 1 of DDX3X (PDB code 2I4I) and refined to the crystallographic R-factor of 0.27 (Rfree 0.29) and deposited in the PDB under accession code 4A4D. One molecule was identified in the asymmetric unit. Owing to lack of density, residues 1–51 were not observed. The refinement statistics are shown in Table 1.

DDX5-N contains eight centrally located β-strands packed against nine α-helices (Figure 2). A structure-homology search on the Dali server (http://www.ebi.ac.uk/dali) reveals a high degree of similarity of domain 1 of DDX5 (residues 79–303) with the equivalent domain in human DDX3X [19], Drosophila Vasa [32] and eIF4a (eukaryotic initiation factor 4A-I) [30] with an rmsd (root mean square deviation) of 1.4 Å, 1.7 Å and 1.9 Å respectively. Structure-based sequence alignment revealed that residues 52–78 of DDX5-N form part of the NTR, whereas residues 79–303 correspond to conserved domain 1 (Figure 3). Residues 52–78 of the NTR contain a loop and an α-helix, whereas the conserved domain 1 displays a RecA-like structure (Figure 3).

Figure 2 Structural features of DDX5-N

Ribbon representation of the structure of DDX5-N (residues 52–304) (PDB code 4A4D). Residues 52–78 (orange) correspond to a segment from the NTR, whereas the remaining residues 79–304 correspond to the conserved domain 1. Conserved motifs are illustrated accordingly. All structural figures were generated with PyMOL (http://www.pymol.org/).

Figure 3 Structure-based sequence alignment of DEAD-box helicases DDX5, DDX3X, VASA and eIF4a

DEAD-box helicases contain two conserved covalently linked globular domains, termed domains 1 and 2, flanked by highly variable N- and C-terminal sequences. Only part of the highly variable N- and C-terminal sequences is shown. Identical residues are shown in white letters with filled boxes, whereas similar residues are boxed. The conserved domains harbour highly conserved motifs. Domain 1 contains conserved motifs Q, I, II, Ia, Ib and III, whereas domain 2 possess motifs IV, V and VI. The conserved motifs are highlighted. Secondary structure elements of DDX5 are shown.

It is clear that domain 1 of DDX5 contains many typical features found in the structures of other DEAD-box helicases. For example, the conserved motifs Q, I (Walker A), II (Walker B, DEAD-box), Ia, Ib and III in domain 1 of DDX5 superimpose with the corresponding regions in DDX3X (Figure 4). The only difference is in the P-loop (found in motif I) because DDX3X is in complex with AMP, whereas DDX5-N is in the apo form. In addition, DDX3X has an insertion between the P-loop and the RNA-binding motif Ia that forms a helix. It has been suggested that this insertion helix is involved in RNA-binding and positioning [19]. This helix is not present in DDX5 (Figure 4).

Figure 4 Structural comparison of domain 1 of DDX5 and DDX3X

The apo form of DDX5 is highlighted in green and red, whereas the AMP-bound form of DDX3X is in blue. The regions highlighted in red (a and b) show the major difference between DDX5 and DDX3X. The P-loop in DDX5 is in a closed conformation (a). Furthermore, DDX5 lacks a 10-aa insertion, which forms a helix between the P-loop and motif Ia in DDX3X (b). Despite these differences, the main chain atoms of domain 1 of both helicases superimposed with an rmsd of 1.4 Å.

Recently, the structure of domain 1 (residues 79–303) of DDX5 in complex with ADP was determined to a resolution of 2.6 Å [33] (PDB code 3FE2). Structural analyses reveal that residues in the Q motif make contact with the adenine moiety of the nucleotide, whereas a phosphate moiety makes contact with the P-loop in motif I [33]. Superimposition of the common core of the apo and bound form reveals an rmsd of 1.2 Å. Overall, small changes are observed upon ADP binding, with the exception of the P-loop. Upon ADP binding, the P-loop shifts by 3.4 Å from the apo form (Figure 5). It is not possible to compare the features in the NTR (residues 1–78) of the two forms, because the NTR is absent from the protein used to solve the structure of ADP-bound DDX5 [33].

Figure 5 Structural comparison of the apo and ADP-bound forms of domain 1 of DDX5

Stereo view of apo (green) and ADP-bound forms of DDX5 (pale blue). The P-loop (found in motif I) for the ADP-bound form is highlighted in purple. The P-loop undergoes significant conformation changes upon nucleotide binding. The bound ADP is shown in sticks.

The ATP-binding site of DDX5-N is not essential for the interaction with NS5B

The crystal structure of DDX5-N showed that there is no nucleotide in the ADP/ATP-binding site. In order to determine whether the ATP-binding site of DDX5-N is important for the interaction between DDX5-N and NS5B in Huh-7 cells, co-immunoprecipitation experiments were performed to determine the interaction of two DDX5 substitution mutants (K144N and D248N) with NS5B. A previous study has shown that the K144N substitution abolished ATP binding, whereas substitution within the DEAD-box did not have any effect on ATP binding [34]. The DDX5-N(K144N) substitution mutant bound NS5B to a similar extent as DDX5-N, suggesting that ATP binding is not essential for the interaction (Figure 6). The DDX5-N(D248N) substitution mutant also showed similar binding to NS5B, suggesting that the DEAD-box is not essential for the interaction.

Figure 6 Co-immunoprecipitation of NS5B with DDX5-N containing a mutation in the ATP-binding or ATP-hydrolysis site

Huh-7 cells were transfected with cDNA constructs for expressing Myc-tagged GST (negative control), Myc-tagged DDX5-N or Myc-tagged DDX5-N containing a substitution mutation and FLAG-tagged HCV NS5B. The cells were harvested at 16 h post-transfection, lysed and subjected to immunoprecipitation with an anti-Myc polyclonal antibody and Protein A–agarose beads. The amount of FLAG-tagged NS5B that co-immunoprecipitated (co-IP) with the Myc-tagged proteins was determined by Western blot analysis with an anti-FLAG monoclonal antibody (top panel). The protein marked with an asterisk represents the heavy chain of the antibody used for immunoprecipitation (top panel). The amounts of Myc-tagged and FLAG-tagged proteins in the lysates before co-IP were determined by subjecting aliquots (containing 10 μg of total protein) of the lysates to Western blot analysis (WB; middle and bottom panels). Molecular mass in kDa is shown on the right-hand side.

The flexible region in the NTR of DDX5-N auto-inhibits its interaction with NS5B

The co-immunoprecipitation experiments showed that DDX5-NΔ60aa [i.e. DDX5(61–305aa)] binds NS5B stronger than DDX5-N (Figure 1), suggesting that the NTR of DDX5-N can auto-inhibit its interaction with NS5B. However, owing to disorder, only part of the NTR (52–78) was observed in our crystal structure. This region forms an extensive loop and supplements the core with an additional α-helix (Figure 2). Unfortunately, the crystal structure does not reveal any contact between the NTR and the rest of DDX5-N. This does not rule out that the NTR can fold back on the rest of DDX5-N, but rather is a reflection of the highly flexible nature of the NTR and the possibly highly dynamic nature of the interaction. Hence, to further investigate the structural elements in NTR that are involved in the auto-inhibition, co-immunoprecipitation experiments were performed using DDX5-N and three mutants of DDX5-N. The first mutant is DDX5-NΔ60aa, in which the N-terminal part of the NTR has been deleted, but the helix in the NTR remains intact. The second mutant is DDX5-NΔ78aa, where the entire NTR has been deleted. The third mutant is DDX5-NΔ70–78aa, where the helix in the NTR has been deleted, but the flexible first 51 residues and the extended loop remains intact (Figure 7A).

Figure 7 Co-immunoprecipitation of NS5B with DDX5-N containing a deletion mutation in NTR

(A) A schematic diagram showing deletion mutants of DDX5-N with modifications in the NTR. The first 51 residues of the NTR are highly disordered, followed by an extended loop structure from amino acids 52 to 69 and an α-helix from amino acids 70 to 78. (B) Huh-7 cells were transfected with cDNA constructs for expressing Myc-tagged GST (negative control), Myc-tagged DDX5-N or Myc-tagged DDX5-N containing a deletion mutation in the NTR and FLAG-tagged HCV NS5B. The cells were harvested at 16 h post-transfection, lysed, and subjected to immunoprecipitation with an anti-Myc polyclonal antibody and Protein A–agarose beads. The amount of FLAG-tagged NS5B that co-immunoprecipitated (co-IP) with the Myc-tagged proteins was determined by Western blot analysis with an anti-FLAG monoclonal antibody (top panel). The protein marked with an asterisk represents the heavy chain of the antibody used for immunoprecipitation (top panel). The amounts of Myc-tagged and FLAG-tagged proteins in the lysates before co-immunoprecipitation were determined by subjecting aliquots (containing 10 μg of total protein) of the lysates to Western blot analysis (middle and bottom panels). Molecular mass in kDa is shown on the right-hand side.

Consistent with the results shown in Figure 1, DDX5-NΔ60aa and DDX5-NΔ78aa showed significantly higher binding to NS5B than DDX5-N (Figure 7B), indicating that the auto-inhibition is abolished when part of, or the complete, NTR is deleted. The internal deletion mutant DDX5-NΔ70–78aa also binds strongly to NS5B (Figure 7B), indicating that residues 1–69 cannot cause auto-inhibition in the absence of the helix formed by residues 70–78. This suggests that this helix may bring the flexible 1–69 residues of NTR close to the NS5B-binding site in DDX5-N, presumably within residues 79–305. Consistently, in vitro translated DDX5(61–614aa) was pulled down by DDX5(1–80aa) fused to GST, but not by GST alone (Figure 8).

Figure 8 GST pull-down of 35S-labelled DDX5(61–614aa) by GST-DDX5(1–80aa)

A 30 μg portion of GST or GST–DDX5(1–80aa) was mixed with in vitro-translated 35S-labelled DDX5(61–614aa). Glutathione–Sepharose-bound fusion proteins (30 μg) used in the pull-down assay were analysed by SDS/PAGE followed by Coomassie Blue staining to ascertain the purity of the proteins (top panel). An autoradiograph shows the amount of 35S-labelled DDX5(61–614aa) protein bound to GST–DDX5(1–80aa) (lane 3) or GST alone (lane 2) (lower panel). One tenth of the binding input was loaded in lane 1 (bottom panel). Molecular mass in kDa is shown on the right-hand side.

DDX5-NΔ60aa interacts directly with NS5B

In order to rule out that the possibility that the auto-inhibition by the NTR of DDX5-N is mediated through the interaction with a third protein, CD and NMR experiments were next performed to characterize the solution conformation of DDX5-NΔ60aa and its in vitro interaction with NS5B. First, NS5B protein was expressed and assessed for its solution conformation. As seen in Figure 9(A), NS5B has a far-UV CD spectrum typical of a helix-dominant protein, with two large negative signals at ~208 nm and ~222 nm respectively, consistent with its three-dimensional structure. The very large positive signal at ~192 nm indicates that it also has a tight tertiary packing. A one-dimensional 1H NMR spectrum for NS5B (Figure 9B) was also collected, but the resonance peaks are very broad, mostly due to the short T2 (transverse relaxation time) resulting from its very large size.

Figure 9 Solution conformation of DDX5(61–305aa) and its binding with NS5B

(A) Far-UV CD spectra of DDX5(61–305aa) (blue) and NS5B (red). (B) One-dimensional 1H NMR spectrum of NS5B. (C) Two-dimensional (1H,15N)-HSQC spectrum of the 15N-labelled DDX5(61–305aa). (D) One-dimensional 1H NMR spectrum of the DDX5(61–305aa) protein. Red arrows indicate several very up-field resonance peaks resulting from the tight tertiary packing. (E) Superimposition of one-dimensional 1H NMR spectra of DDX5(61–305aa) in the absence (blue) and in the presence of NS5B at molar ratios of 1:1 (black); 1:2 (green) and 1:3 (red) (DDX5/NS5B). (F) Superimposition of the simulated (black) and real (red) one-dimensional 1H NMR spectra of the mixture of DDX5(61–305aa) and NS5B at a molar ratio of 1:3 (DDX5/NS5B). The simulated spectrum is generated by simply adding up two separate spectra of DDX5(61–305aa) and NS5B at the corresponding concentrations.

Secondly, DDX5-NΔ60aa was purified and subjected to far-UV CD spectroscopy (Figure 9A). The results show that DDX5-NΔ60aa has a helix-dominant structure with tight tertiary packing, in agreement with its three-dimensional structure determined in the present study. Consistent with the CD results, the one-dimensional 1H NMR spectrum of DDX5-NΔ60aa shows a large dispersion of the amide proton signals as well as many very up-field signals (Figure 9D), indicating that DDX5-NΔ60aa is well folded. However, also probably due to its relatively large size, its proton resonance signals are broad. This is further confirmed by its two-dimensional (1H,15N)-HSQC spectrum acquired on a 15N-labelled DDX5-NΔ60aa sample (Figure 9C), in which although both 1H and 15N spectral dispersions are large, many peaks from amide protons are too broad to be detectable.

As a consequence, the direct binding of DDX5-NΔ60aa to NS5B was assessed by monitoring the one-dimensional 1H NMR spectra of DDX5-NΔ60aa in the absence or presence of NS5B at molar ratios of 1:1, 1:2 and 1:3 (Figure 9E). Interestingly, addition of NS5B into DDX5-NΔ60aa at a molar ratio of 1:1 (DDX5:NS5B) caused a significant decrease in resonance intensities and resulted in resonance broadenings (Figure 9E). This decrease could not be due to a non-specific aggregation because the two proteins alone showed no sign of aggregation at much higher concentrations. In particular, both of them could be crystallized at very high concentrations to determine their X-ray structures. Furthermore, no large change was further detected for the NMR spectra when more NS5B was added (at molar ratios of 1:2 and 1:3). This clearly indicates that DDX5-NΔ60aa binds to NS5B to form a tight complex, because the complex formation was almost completed at a molar ratio of 1:1 (DDX5/NS5B). However, the very large molecular mass of the tight complex will lead to a dramatic shortening of T2, and consequently the NMR signals become very broad, thus leading to the significant reduction of the signal intensity. If no interaction existed between DDX5-NΔ60aa and NS5B, the NMR spectrum of the mixture would be a simple addition of two separate spectra which should have resonance intensities larger than each spectrum (Figure 9F). However, this is not the case, demonstrating that DDX5-NΔ60aa does interact directly with NS5B.

Further biophysical characterization of the interaction between DDX5-NΔ60aa and NS5B was hindered by the fact that the DDX5-NΔ60aa–NS5B complex appears to be prone to aggregation. Although during the NMR titrations, no visible aggregation was observed, the mixture sample would precipitate overnight. Indeed, in order to obtain thermodynamic binding parameters, isothermal titration calorimetry was performed with various protein concentrations, but all of the experiments failed because of sample precipitation. It is likely that the precipitation of the complex was accelerated by rapid spinning of the needle during the isothermal titration calorimetry measurements.

DISCUSSION

Over the past decade, a growing number of DEAD-box RNA helicase structures have been determined [19,2933]. They contain a central region composed of two conserved domains, termed domains 1 and 2, flanked by highly variable N- and C-terminal sequences. Comparison of the known structures of domains 1 and 2 reveal that they have a fold belonging to the RecA superfamily, where five β-strands are surrounded by five α-helices [35]. We have previously shown that DDX5 interacts with NS5B and that the C-terminal of NS5B is essential for this interaction [5]. Co-immunoprecipitation experiments showed that there are two independent NS5B-binding sites in DDX5, one in DDX5-N (corresponding to residues 1–305 of DDX5) and the other in DDX5-C (corresponding to residues 306–614 of DDX5) (Figure 1). Interestingly, DDX5-N and DDX5-C contain domains 1 and 2 respectively.

The crystal structure of the apo form of DDX5-N (residues 52–304) was solved and it reveals that domain 1 of DDX5 contains many typical features found in the structures of other DEAD-box helicases (Figures 2 and 3). Generally, the P-loop in DEAD-box helicases adapts either an open or closed conformation. The open conformation, typically found in NTPases where nucleotide is bound, is observed in the nucleotide-bound form of the DEAD-box helicase eIF4A [30] and UAP56 [U2AF (U2 small nuclear ribonucleoprotein auxiliary factor)-associated protein of 56 kDa] [31], whereas the closed conformation is observed in the unbound structures of eIF4A [30], UAP56 [31] and BstDead (Bacillus strearothermophilus Dead protein) [29]. In DDX5-N, the P-loop clearly adopts an open conformation in the presence of bound nucleotide, whereas the apo form is in a closed conformation (Figures 4 and 5). However, neither the ATP-binding site nor the DEAD motif in DDX5-N is required for its interaction with NS5B (Figure 6).

The NTRs of DEAD-box helicases are highly variable and there is limited structural information [6,7]. For DDX5-N, no density was observed for residues 1–51 within the NTR. Analysis, using the protein disorder server RONN [36], predicted this region to be disordered as is observed with other DEAD-box helicases. The remaining part of the NTR (residues 52–78) comprises a loop and an α-helix located before domain 1. To our knowledge, this is only the second report of structural elements in the NTR of DEAD-box RNA helicases, as most of the structures of DEAD-box RNA helicases were obtained using proteins expressed without the NTR. Indeed, the recently solved structure of ADP-bound DDX5 consists only of domain 1 [33] (PDB code 3FE2). The first report documented an α-helix in the NTR of DDX19 that inserts between domains 1 and 2 to inhibit ATP hydrolysis unless it is displaced by RNA binding [37].

Interestingly, the NTR of DDX5-N inhibits its interaction with NS5B (Figure 7). The α-helix in the NTR is essential for this auto-inhibition and seems to mediate the interaction between the highly flexible 1–51 residues in the NTR and NS5B-binding site in DDX5-N (Figures 7 and 8), presumably located within residues 79–305. Furthermore, DDX5(1–80aa) can bind directly to DDX5(61–614aa) (Figure 8), thus providing further evidence that the flexible NTR of DDX5 folds back to interact with the NS5B-binding site. Consistently, NMR investigations reveal that there is indeed a direct interaction between DDX5-NΔ60aa and NS5B in vitro (Figure 9).

The results of the present study suggest that the DDX5-N and DDX5-C interact with NS5B independently. The high-resolution three-dimensional structure of the apo form of DDX5-N was solved and provides a basis for future studies to define the precise role of DDX5 during HCV infection. Crystallization of full-length DDX5 and the DDX5–NS5B complex is also actively being pursued. Our results also reveal that the interaction between DDX5-N and NS5B is auto-inhibited by the highly flexible NTR of DDX5-N and that the α-helix formed by amino acids 70–78 is necessary for the inhibition to occur. It is expected that solving the structure of full-length DDX5 can help us to ascertain whether this α-helix can insert between domains 1 and 2 as has been observed for DDX19 [37].

Recent studies revealed that several DEAD-box RNA helicases, like DDX3, rck/p54 and RNA helicase A, are required for HCV RNA replication [3840]. It is intriguing that multiple DEAD-box RNA helicases are important for HCV replication and further studies are warranted to determine whether they are involved in the same step(s) of the viral life cycle. It is also possible that these host proteins are involved in more than one viral–host interaction. For example, DDX3 not only binds the HCV core protein, but is probably associated with an HCV NS protein or HCV RNA itself [38,41].

AUTHOR CONTRIBUTION

Sujit Dutta, Garvita Gupta and Yook-Wah Choi designed experiments. Sujit Dutta, Garvita Gupta, Yook-Wah Choi and Masayo Kotaka performed experiments. Sujit Dutta, Yook-Wah Choi, Burtram Fielding, Jianxing Song and Yee-Joo Tan wrote and revised the paper.

FUNDING

This work was supported by grants from the Ministry of Education (MOE) of Singapore [Academic Research Fund Tier 1 Grant R-182-000-170-112 (to Y.-J.T.) and Tier 2 Grant R-154-000-525-112 (to J.S.)]. Initial work was supported by intramural funds from the A*STAR.

Acknowledgments

We thank the personnel at NSRRC for their kind help during data collection.

Footnotes

  • The structural co-ordinates of the N-terminal domain of DDX5 will appear in the PDB under accession code 4A4D.

Abbreviations: aa, amino acids; A*STAR, Agency for Science, Technology and Research; DEAD, Asp-Glu-Ala-Asp; DTT, dithiothreitol; eIF4a, eukaryotic initiation factor 4A; GST, glutathione transferase; HCV, hepatitis C virus; HSQC, heteronuclear single quantum coherence; NP-40, Nonidet P40; NS, non-structural; NTR, N-terminal region; NSRRC, National Synchrotron Radiation Research Center, Taiwan; rmsd, root mean square deviation; T2, transverse relaxation time; UAP56, U2AF (U2 small nuclear ribonucleoprotein auxiliary factor)-associated protein of 56 kDa

References

View Abstract