Structural and Mechanistic Insight into DNA Unwinding by Deinococcus radiodurans UvrD

DNA helicases are responsible for unwinding the duplex DNA, a key step in many biological processes. UvrD is a DNA helicase involved in several DNA repair pathways. We report here crystal structures of Deinococcus radiodurans UvrD (drUvrD) in complex with DNA in different nucleotide-free and bound states. These structures provide us with three distinct snapshots of drUvrD in action and for the first time trap a DNA helicase undergoing a large-scale spiral movement around duplexed DNA. Our structural data also improve our understanding of the molecular mechanisms that regulate DNA unwinding by Superfamily 1A (SF1A) helicases. Our biochemical data reveal that drUvrD is a DNA-stimulated ATPase, can translocate along ssDNA in the 3′-5′ direction and shows ATP-dependent 3′-5′, and surprisingly also, 5′-3′ helicase activity. Interestingly, we find that these translocase and helicase activities of drUvrD are modulated by the ssDNA binding protein. Analysis of drUvrD mutants indicate that the conserved β-hairpin structure of drUvrD that functions as a separation pin is critical for both drUvrD’s 3′-5′ and 5′-3′ helicase activities, whereas the GIG motif of drUvrD involved in binding to the DNA duplex is essential for the 5′-3′ helicase activity only. These special features of drUvrD may reflect its involvement in a wide range of DNA repair processes in vivo.


Introduction
Many biological processes, such as DNA replication, transcription, recombination or repair, require access to the genetic information hidden within the duplex DNA of the genome and for this purpose the double-stranded DNA (dsDNA) needs to be transiently unwound. A diverse set of enzymes, known as DNA helicases, is responsible for catalyzing this process [1,2]. DNA helicases are ubiquitous enzymes and many different helicases are found in a single cell due to the diversity of structures adopted by duplexed DNA. Helicases are a subset of the translocase enzyme family that share a number of conserved signature motifs responsible for either NTP binding and hydrolysis, DNA binding or for coupling these two processes. Based on primary structure analyses and extensive biochemical studies, six superfamilies of helicases have so far been described, each of which possesses a different set of conserved signature motifs [3,4]. Three of these superfamilies (SF1, SF2 and SF6) have been further classified according to their polarity 3′-5′ (type A) or 5′-3′ (type B) [4].
UvrD is classified as a SF1A helicase [3] and plays important functions in DNA replication [5], recombinational repair [6][7][8], methyl-directed mismatch repair [9] and nucleotide excision repair [10]. UvrD consists of two RecA-like domains (1A and 2A) that are responsible for nucleotide binding and hydrolysis and two additional domains (1B and 2B) that are involved in dsDNA binding. UvrD has been shown to translocate along single-stranded DNA (ssDNA) as a monomer, while a number of studies indicate that oligomerization, and notably dimerization, of UvrD is required for helicase activity [11][12][13][14][15]. Over the past 15 years, several crystal structures of SF1A helicases have been determined. In 1996, the structure of Geobacillus stearothermophilus PcrA (gsPcrA) was solved in its apo form [16] and in 1997, the first crystal structure of Escherichia coli Rep helicase complexed with ssDNA was solved [17] providing the first insights into the interaction of the protein with DNA. Subsequently, the structures of gsPcrA in complex with 3′-tailed DNA consisting of a 10 base pair DNA duplex and a seven base single-stranded 3′-tail were determined in apo-and AMPPNP-bound forms [18] and in 2006, several structures of E. coli UvrD (ecUvrD) bound to 3′tailed DNA were determined revealing the details underlying DNA unwinding by SF1A helicases [19]. These structures led to the proposal of a combined wrench-and-inchworm mechanism for DNA unwinding [19,20]. In this model, a rotational movement regulated by ATP binding and hydrolysis acting as the 'engine' is combined with alternate tight and loose interactions at four protein-DNA contact points to produce a highly coordinated unidirectional movement along DNA.
In the radiation-resistant bacterium, Deinococcus radiodurans, unlike in E. coli, UvrD is involved in diverse DNA repair pathways [7]. In particular, UvrD has been shown to play a central role in double-strand break (DSB) repair and reconstitution of the genome following chromosome fractionation [7]. In E. coli, the RecQ, RecD and Helicase IV enzymes participate in DSB repair while in D. radiodurans, these three helicases have been shown to be dispensable [7]. The involvement and importance of a helicase in a given cellular pathway are not conserved from one bacterium to another.
Here we present crystal structures of full-length and a Cterminally truncated construct of D. radiodurans UvrD (drUvrD) in complex with 3′-tailed dsDNA. Our structures obtained in apo-and AMPPNP bound states provide us with several snapshots of this essential cellular process and reveal a largescale spiral movement of UvrD around the duplexed DNA. Our structural data and biochemical analysis of wild-type and mutant drUvrD support the previously proposed wrench-andinchworm model and provide further insight into the local conformational changes associated with DNA unwinding. A structural comparison of drUvrD with its E. coli homologue reveals that most of the differences reside in the inter-domain contacts and the ssDNA binding pocket and gating mechanism. Our biochemical studies reveal that drUvrD is an active DNAstimulated ATPase that also possesses ATP-dependent translocase and helicase activities. Further investigations of these in vitro activities demonstrated that drUvrD translocates along ssDNA with a biased 3′-5′ directionality but, despite belonging to the SF1A protein family, can unwind duplexed DNA in both the 3′-5′ and 5′-3′ directions. Interestingly, we find that these translocase and bipolar helicase activities of drUvrD are modulated by the ssDNA binding protein (SSB).

Cloning, expression and purification of drUvrD and drSSB
Full-length drUvrD (drUvrD FL ) and a truncated construct of drUvrD (drUvrD ∆C ), missing residues 666-745 ( Figure 1A), were cloned into pET151d (Invitrogen). drUvrD FL mutants were prepared with the QuikChange mutagenesis kit (Agilent Technologies). All constructs were expressed in BL21 (DE3) cells. Protein expression was induced by 1 mM IPTG at 15°C overnight. Cells were lysed by sonication and the protein was purified by Ni affinity chromatography (Macherey-Nagel) in 50 mM Tris-HCl pH 8.0, 150 mM NaCl, 5% glycerol and 5 mM MgCl 2 , followed by His-tag cleavage with TEV protease and dialysis to remove the imidazole used for eluting the protein from the Ni column. The protein was further purified on a HiTrap Heparin column (GE Healthcare) and eluted in 50 mM Tris-HCl pH 8.0, 400 mM NaCl, 0.1 mM EDTA, 1 mM DTT, 5 mM MgCl 2 and 5% glycerol. The protein was concentrated to 7-8 mg/ml and stored at -80°C. Deinococcus radiodurans SSB (drSSB) was cloned into pET151d (Invitrogen) for expression with a cleavable N-terminal His-tag and expressed in BL21 (DE3) Star at 20°C overnight. drSSB was purified on Ni-NTA (Qiagen) followed by a size-exclusion chromatography (Superdex 75 10/300 GL) in 50 mM Tris-HCl pH 8.0 and 100 mM NaCl and was stored at -80°C.

DNA oligonucleotides
All DNA oligonucleotides used in this study were purchased from Eurofins-MWG and their sequences are presented in Table S1. The DNA used for co-crystallization with drUvrD ∆C was composed of For25 and its complementary strand Rev25, while drUvrD FL was co-crystallized with DNA formed by For28 and Rev28 ( Figure 1B). Annealed DNAs were purified by anion exchange chromatography (MonoQ; GE Healthcare). HPLCpurified oligonucleotides, For25-21F and Rev25-21F, containing a fluorescein-derivatized thymine (Fluo-dT) at position 21 were simply annealed prior to crystallization trials. For the helicase and DNA binding assays, the DNA substrates were made of a 25mer oligonucleotide containing a Fluo-dT at position 12 (H1T12) and a complementary oligonucleotide containing no extension (H4), a 15 nucleotide (nt) or a 7nt polydT ssDNA extension at either the 3′ (H3-15 or H3-7) or 5′ end (H5-15 or H5-7). For the streptavidin-displacement assay, the 3′-tailed DNA substrate consisted of a 3′-fluorescein labeled 25mer oligonucleotide (H1-3F) annealed to its complementary oligonucleotide with a 25nt polydT ssDNA extension at its 3′ end and containing a biotin conjugated thymine in position 49 (H3-25-B49), while the 5′-tailed DNA substrate was composed of a 5′-FAM labeled 25mer oligonucleotide (H1-5FAM) annealed to its complementary oligonucleotide with a 25nt polydT ssDNA extension at its 5′ end and containing a biotin conjugated thymine in position 2 (H5-25-B2).

Data collection and Structure Determination
Diffraction data (Table 1) were collected at the European Synchrotron Radiation Facility (ESRF) in Grenoble, France and were processed with either XDS [21] or iMosflm [22]. The structure of drUvrD ∆C -For25/Rev25 form I was solved by molecular replacement using Mr. Bump [23] and the gsPcrA helicase as a search model (PDB entry 3PJR). After several rounds of substantial rebuilding of the protein chains in Coot [24], the DNA could be built and the AMPPNP molecules docked into the electron density. Subsequently, this model was used to solve the structures of drUvrD ∆C -For25/Rev25 form II and drUvrD FL -For28/Rev28 by molecular replacement with Phaser [25]. The drUvrD ∆C -For25/Rev25 form I and form II models were refined with Refmac [26], while the drUvrD FL -For28/Rev28 model, solved at lower resolution, was refined in Phenix [27] using drUvrD ∆C -For25/Rev25 form I as a reference model (Table 1). Fig.s of structures were prepared with Pymol [28] and the movie of the morph was created with Chimera [29].

ATPase activity
The rate of ATP hydrolysis by 100 nM drUvrD FL and drUvrD ∆C in the presence of a 25mer polydT oligonucleotide was measured using the spectrophotometric method [30] at 25°C in 50 mM Tris-HCl pH 8.0, 100 mM NaCl, 0.1 mM EDTA, 1 mM DTT, 5 mM MgCl 2 and 5% glycerol (buffer A). The K ssDNA was determined by measuring the rate of ATP hydrolysis in the presence of 2 mM ATP as a function of increasing concentrations of ssDNA (0-10 µM). Kinetic parameters (V max , K m and K cat ) were determined by measuring the rate of ATP hydrolysis in the presence of an excess of ssDNA (10xK ssDNA ) at various ATP concentrations (0-1 mM). The measurements were made in triplicate and the average ATPase rates were plotted and fitted to a hyperbola using Origin.

Helicase assay
Helicase activity of drUvrD FL was assayed in 10 mM Tris-HCl pH 8.0, 50 mM NaCl, 1% glycerol, 5 mM MgCl 2 and 0.1 mg/mL BSA (buffer B). 80 µl reactions containing 20 nM DNA and 250 nM wild-type or mutant drUvrD were incubated at 25°C. The duplexed DNA was either blunt or contained 15nt or 7nt ssDNA extensions at either the 3′-or 5′-ends. The reactions were initiated by addition of 2 mM ATP. At indicated time points, 10 µl samples of the reaction were quenched with 2.5 µl of a solution containing 0.8% SDS, 0.08% bromophenol blue, 24% glycerol, 80 mM EDTA and 20 µM unlabeled H1 oligonucleotide. The reactions were carried out in the absence and presence of 250 nM drSSB. Reaction products were run on a 20 % polyacrylamide TBE gel and the DNA bands were visualized and quantified using a ChemiDoc MP imaging system and the Image Lab software (Bio-Rad). Initial reaction rates were estimated using GraphPad Prism6 and averaged data from three independent experiments were plotted in GraphPad Prism6 with standard deviations represented as vertical error bars.

Streptavidin displacement assay
The translocase activity of drUvrD was assayed using the streptavidin-displacement assay [31,32]. DNA oligonucleotides used in this assay consisted of dsDNA duplexes with a 25 nt ssDNA extension at either the 3′-or 5′-end and a biotin label in positions 49 and 2 respectively. The DNA-streptavidin complexes were formed by incubating the biotinylated dsDNA (0.2 µM) with streptavidin (3.2 µM, Sigma) in 10 mM Tris-HCl pH 8.0, 50 mM NaCl, 0.5 mM EDTA at 25°C for 30 min, before addition of 180 µM biotin. Displacement reactions of 80 µl containing 20 nM streptavidin-loaded DNA and 250 nM drUvrD FL were incubated in buffer B at 25°C. The reactions were initiated by addition of 2 mM ATP. At indicated time points, 10 µl samples of the reaction were quenched with 2.5 µl of a solution containing 0.48% SDS, 0.032% bromophenol blue, 20% glycerol, 160 mM EDTA and 20 µM unlabeled H1 oligonucleotide. The reactions were carried out in the absence and presence of 250 nM drSSB. Reaction products were run on a 10 % polyacrylamide TBE gel and the DNA bands were visualized and quantified using a ChemiDoc MP imaging system and Image Lab software (Bio-Rad). Averaged data from three independent experiments were plotted in GraphPad Prism6 with standard deviations represented as vertical error bars.

DNA Binding
Equilibrium DNA binding assays were performed on a Synergy H4 Hybrid Microplate reader (BioTek), fitted with polarization filters to measure fluorescence anisotropy. The binding assays were conducted in 384-well plates at room temperature in 80 µl reaction volumes in buffer A supplemented with 0.05% Tween 20, 0.1 mg/mL BSA and 1 mM AMPPNP. 0 to 8 µM wild-type and mutant drUvrD were titrated into 2.5 nM fluorescently-labeled dsDNA containing 15nt ssDNA extensions at either the 3′-or 5′-end. Averaged data from three independent experiments were fitted to a standard binding equation (y=Bmax*x/(Kd+x)) assuming a single binding site [33] using GraphPad Prism6. The fits were very good, with R 2 values all above 0.98.

Crystal structures of drUvrD-DNA complexes
A ternary complex containing 2 molecules of intact drUvrD FL , a 21-mer DNA duplex with 7nt ssDNA extensions at each of its 3′-ends and the non-hydrolysable ATP analogue, AMPPNP, was crystallized in space group P2 1 with four drUvrD chains and two DNA duplexes per asymmetric unit ( Figure 1B). These crystals diffracted X-rays to 4.0 Å (Table 1). Despite being present in the crystallized protein, residues 663-745 corresponding to the variable C-terminal region could not be traced in the electron density map, confirming that this region is particularly flexible [34]. Crystals containing the C-terminally truncated drUvrD ∆C ( Figure 1A), an 18-mer DNA duplex with 7nt ssDNA extensions at its 3′-ends ( Figure 1B) and AMPPNP diffracted to higher resolution. This complex produced two crystal forms (I and II) diffracting respectively to 2.5 and 3.0 Å resolution (Table 1). Crystal form I appeared very rapidly (<1 day) and belonged to space group P2 1 2 1 2 1 with two protein monomers and one DNA duplex per asymmetric unit, while crystal form II appeared after at least one week and belonged to space group P2 1 with four molecules of protein and two DNA duplexes per asymmetric unit. In all structures, each drUvrD monomer was bound to the ds-ssDNA junction at either end of the DNA duplex, thus forming an assembly of one DNA duplex with two UvrD monomers ( Figure 1B). In the structures of drUvrD FL and drUvrD ∆C form I, each protein monomer contains one bound AMPPNP molecule, whereas in drUvrD ∆C form II each assembly is composed of a DNA duplex with an AMPPNP-bound UvrD on one end and an apo-UvrD on the other.
In all three structures, the quality of the electron density corresponding to the bound DNA varied considerably over the molecule. In contrast to the very well defined map of the ssDNA tails, the duplex regions were less clear and exhibited significantly higher B-factors than the adjacent protein atoms. In drUvrD ∆C form I and drUvrD FL , the nucleotides at the junction between the dsDNA and the ssDNA are poorly defined, indicating that this region is relatively flexible.
As in previous structures of UvrD-like helicases [17][18][19], drUvrD crystallized as a monomer, and no putative dimer interfaces were detected between adjacent protein molecules in our three structures. drUvrD displays 36% sequence identity with E. coli Rep (ecRep) and UvrD (ecUvrD) helicases and 42% sequence identity with G. stearothermophilus PcrA (gsPcrA) helicase, all of which are members of the SF1A helicase family ( Figure S1). The overall structures of the drUvrD monomers are very similar to those observed in the closed conformation of gsPcrA-DNA and ecUvrD-DNA complexes [18,19] formed by domains 1A, 1B, 2A and 2B (Figure 2A and 2B). When present, AMPPNP is bound at the interface between domains 1A and 2A. ssDNA interacts with all four domains, a majority of contacts being with domain 2A, and interactions with dsDNA involve domains 1B, 2A and 2B (Figure 2A).
A close look at the residue conservation pattern ( Figure S1) reveals that most of the non-conserved residues are found at domain interfaces. The buried surface areas and the nature of contacts at domain interfaces are indeed very different in ecand drUvrD (Table S2). In drUvrD, the interface of domains 1B/2B is significantly smaller than in ecUvrD and the interfaces between domains 1A/1B and 1B/2B involve many more ionic interactions. Such differences may impart increased plasticity and flexibility to drUvrD [35][36][37], but may also increase its sensitivity to stress-related changes in its local environment (e.g. pH, temperature, salt concentration).

Large-scale conformational changes of drUvrD
While the relative orientation of the components of the protein-DNA assembly observed in drUvrD ∆C form II ( Figure  3C) is similar to those observed in previous structures of SF1A helicases, our structures of drUvrD FL and drUvrD ∆C form I provide us with two new snapshots of helicase-DNA complexes ( Figure 3A and 3B). In both drUvrD FL and drUvrD ∆C form I, the two drUvrD protomers are located on the same side of the DNA duplex and induce a bend in the DNA (Table S3). In the case of drUvrD FL , one of the two assemblies in the asymmetric unit forms a very bent assembly (25° bend in the DNA axis) where domains 2B of the two protomers come closer together ( Figure  3A). The second protein-DNA assembly obtained in drUvrD FL crystals is more similar to the assembly obtained in drUvrD ∆C form I. In these structures, the two drUvrD protomers are still located on the same side of the DNA duplex ( Figure 3B), but the DNA duplex is less bent (16° bend in the DNA axis; Table  S3). In drUvrD ∆C , loss of one of the AMPPNP molecules leads to the formation of crystal form II in which the apo molecule has rotated 125° around the DNA with respect to the position of drUvrD FL or 105° with respect to crystal form I ( Figure 3D and 3E), resulting in an assembly with one drUvrD on either side of the DNA duplex ( Figure 3C). In this assembly, the apo-drUvrD ∆C molecule has twisted the 3′-ssDNA extension and maintains the ssDNA in a bent orientation with respect to the DNA duplex axis. As a result, the DNA duplex itself shows a reduced helical twist and reduced bending (Table S3). The accompanying movie (Movie S1) presents a morph of drUvrD ∆C as it rotates around the duplexed DNA and undergoes this large spiral movement corresponding to the transition from crystal form I to crystal form II.

Local structural rearrangements of drUvrD
The loss of the bound nucleotide also results in major structural rearrangements within the protein monomer ( Figure  4). Loss of the nucleotide induces a ~15° rotation of domain 2B and an ~8° rotation of domains 1A and 1B relative to domain 2A in the plane formed by the ss-and dsDNA ( Figure 4A and 4B). It also leads to a ~15° twist of domains 1A and 1B relative to domain 2A around the ssDNA axis ( Figure 4C). Similar rotations have been observed previously in the structures of gsPcrA [18] and ecUvrD [19]. In ecUvrD, however, all three domains (1A, 1B and 2B) moved as a single unit as UvrD converted from a nucleotide-bound state to an apo-form and no significant changes in domain interactions were observed. In contrast, in drUvrD domains 1A and 1B move independently of domains 2A and 2B and these movements lead to a number of structural rearrangements and the disruption of several salt bridges between domains 1A/1B and 1B/2B (Table S2).
Most conformational changes (within the protomer) associated with ATP hydrolysis involve a number of conserved sequence motifs ( Figure S1) identified in ecUvrD [19] and described hereafter. In ecUvrD helix α24 from domain 2B was referred to as the gating helix and was proposed to regulate the exiting of ssDNA. In drUvrD, this helix adopts a similar, closed conformation in both the nucleotide-bound and apo forms of drUvrD ( Figure 4D). This was also the case in the gsPcrA-DNA complexes [18] and in several of the nucleotide-bound and apo structures of ecUvrD [19]. In contrast, the linker that follows this helix and connects domain 2B to domain 2A (an unstructured loop in ecUvrD), adopts a different conformation depending on the nucleotide-bound state of drUvrD. In AMPPNP-bound drUvrD, the linker forms a loop (residues 544-548) and a small helix, α25 ( Figures 2B and 4D), while in the nucleotide-free drUvrD, this region is very flexible with part of the chain missing from the electron density maps, most likely to allow the ssDNA to exit the molecule. Helices α24 and α25 from domains 2B and 2A on one side and the conserved sequence motif Ia (β2 and α3) from domain 1A on the other side thus form an ssDNA gateway, which opens and closes like sliding doors ( Figure  4D). In the AMPPNP bound form, the hydroxyl group of Ser546, on the linker between α24 and α25, is only 4.5 Å away from the carbonyl oxygen of Phe65 (motif Ia), thus blocking the ssDNA exit. In this form, the loop preceding α25 closes down on the 3′-end of the ssDNA and interacts directly via Ser546 with the phosphate of the terminal nucleotide. This conformation is stabilized by helix α25, which is missing in ecUvrD and is unstructured in apo-drUvrD. Upon AMPPNP release, rotation of domains 1A, 1B and 2B opens the gateway; in apo-drUvrD, the opening increases to nearly 10 Å to allow a single nucleotide to thread through. Additionally, in AMPPNPbound drUvrD, the ssDNA gateway is plugged by the tip of the β3-α4 loop (Thr91 from motif Ib) and this plug also moves out of the way in apo-drUvrD to let the ssDNA through ( Figure 4D).

Mechanistic insight into DNA unwinding by drUvrD
In drUvrD ∆C form I, four nucleotides (nt21-24) are tightly bound in the ssDNA-binding pocket ( Figure 5A). The terminal nucleotide (nt25) has already exited the helicase and is no longer visible in the electron density maps. Nucleotides 21 and 22 interact with Arg362 and Asn364 (motif IVa) and stack against Phe263 (motif III) that interferes with the regular stacking of the ssDNA bases and forces nucleotides 23 and 24 to adopt an orientation orthogonal to nucleotides 21 and 22. Nucleotide 24 is stabilized in this conformation by π-stacking of the base between Arg264 and Phe196 (motif Id) and of the deoxyribose ring against Phe65 (motif Ia). These residues are in turn stabilized by a series of stacking interactions involving notably Tyr261 (motif III) and His93 (motif Ib). In the AMPPNPbound molecule of drUvrD ∆C form II, nucleotides 20 to 23 are bound in the binding pocket ( Figure 5B), indicating that drUvrD has translocated along the ssDNA by one nucleotide compared to form I and as a result both nucleotides 24 and 25 have become untraceable.
As in ecUvrD, the apo-drUvrD ∆C observed in crystal form II, reveals a fifth nucleotide in the ssDNA binding pocket ( Figure  5C). Nucleotides 19 to 23 are now visible and sliding of drUvrD along the DNA has positioned nucleotides 19 and 20 in the first two binding sites and nucleotides 21 and 22 in the subsequent sites. As a result, nucleotide 23 is now trapped on its way out. To allow the terminal nucleotide to exit, Phe196, Phe65 and His93 from motifs Ia, Ib and Id have moved out of the way and the α24-α25 linker that interacts via Ser546 with the terminal nucleotide in the AMPPNP-bound forms, has maintained its grip on the 3′-end of the ssDNA and pulled it through the opened gateway driven by the rotation of domain 2B ( Figure  5C). Nucleotides 21 and 22 are now stabilized in their new binding sites by interactions with Tyr390 and Arg392 from motif IVb.
Investigation of the dsDNA binding shows that it is also affected by the nucleotide-bound state of drUvrD ( Figure 5D-F). Interactions between drUvrD and dsDNA involve four contact points: one helix-loop-helix (HLH) motif from domain 1B (α5-α6), two of the three HLH motifs from domain 2B (α17-α18 and α19-α20) and the β-hairpin motif (β13-β14) from domain 2A. In the AMPPNP-bound structures, three of these four sites are in contact with dsDNA; two of them are in common and the third differs between the two forms ( Figure 5D and 5E). In both forms, Arg142 from the α5-α6 HLH motif interacts with the unpaired nt19 at the ss-dsDNA junction and the α17-α18 HLH motif containing the conserved GIG sequence (motif IVc, Figure S1) interacts extensively with nt9-12 in form I and nt7-10 in form II ( Figure 5D and 5E). In form I, the third binding site involves Arg459 from the α19-α20 HLH motif, which interacts with the deoxyribose ring of nt13 (opposite strand) in the minor groove of the DNA duplex ( Figure 5D), while in form II, Phe633 from the β-hairpin motif stacks against the first base-pair   Figure 2A. The helices belonging to the HLH motifs and the β-hairpin structure (orange) are shown and labeled according to the secondary structure succession ( Figure S1). The positively charged residues in contact with dsDNA are illustrated in sticks and the GIG motif is indicated. The number of base-pairs formed between the ss-dsDNA junction and the contact point with the drUvrD GIG motif is shown to the left of each panel. This number differs significantly between the two crystal forms. G-I. Schematic representation of drUvrD's DNA binding in the different crystal structures as indicated below the models. The four protein-DNA contact points that are critical for the wrench-and-inchworm unwinding mechanism are indicated with circled numbers in all panels: HLH motifs interact with dsDNA (1), the β-hairpin motif with the ss-dsDNA junction (2), motif III with the ssDNA (3) and the ssDNA gateway with the exiting ssDNA (4). G. In AMPPNP bound Form I, contact points 1, 3 and 4 are tight. H. In AMPPNP bound Form II, drUvrD's GIG motif (1) has slided along the DNA duplex and pushes the DNA junction against the β-hairpin motif (2), which now stacks tightly against the first base-pair. I. In the apo molecule of Form II, the ssDNA gateway (4) has opened and ssDNA exited the helicase. Domains of drUvrD are colored as in Figure 2A. (nt1=nt18) of the duplex ( Figure 5E). In the absence of nucleotide, however, only two of these contacts remain: the GIG sequence in the α17-α18 HLH motif interacts with nt7-10 and Phe633 from the β-hairpin motif stacks against the first base-pair ( Figure 5F).
Analysis of the ss-and dsDNA binding in the different structures ( Figure 5) indicate how local conformational changes and domain rotations are transformed into drUvrD's linear movement along the DNA via alternate loose and tight protein-DNA contact points, as proposed in the wrench-and-inchworm model for DNA unwinding. During ATP-binding-induced domain closing, binding to duplexed DNA through several HLH motifs (contact 1) and to ssDNA (contacts 3 and 4) are tight, while contact with the ss-dsDNA junction (contact 2) is loose ( Figure  5D and 5G). UvrD then slides along the duplex away from the junction and thereby pushes the duplex DNA against the βhairpin (contact 2). Phe633 located at the tip of the β-hairpin now stacks against the first base-pair ( Figure 5E and 5H). Since the ssDNA gateway is closed at this stage, this movement creates a tension on the ss-dsDNA junction, which distorts the first nucleotide at the ss-dsDNA junction (nt19), thus forming a bulge ( Figure 5E and 5H). This tension is then released during ADP and Pi release: domain rotations open drUvrD's ssDNA gateway (contact 4) to allow the ssDNA to exit the helicase. Contacts with the ssDNA (contact 3) remain tight throughout the process (Fig.s 5A-5C) in order to guide and tether the ssDNA through the gateway and straighten the bulged out nucleotide ( Figure 5I). During this step, contacts with the DNA duplex are restricted to the GIG motif in α17-α18 HLH (contact 1 is looser) and the β-hairpin (contact 2) that is stacked against the first base-pair and is now in a position to act as a solid separation pin for subsequent unwinding of the duplex DNA ( Figure 5F and 5I).

drUvrD is an active, DNA-stimulated ATPase and an ATP-dependent helicase
To better understand how a structurally and mechanistically conserved protein such as UvrD may be involved in diverse repair pathways in different species, we investigated drUvrD's catalytic activities in vitro. As other SF1A helicases, drUvrD displays a clear DNA-stimulated ATPase activity ( Figure 6A). Analysis of the ATPase data measured on drUvrD FL and drUvrD ∆C allowed us to determine their apparent turnover rates (K cat ) for ATP hydrolysis, along with their K m for ATP and their K ssDNA (corresponding to the concentration of ssDNA required for half-maximal ATPase rate) ( Figure 6A). These values are in agreement with those measured for wild-type and a Cterminally truncated form of ecUvrD [34,38]. When compared to drUvrD FL ,drUvrD ∆C exhibits a significantly higher turnover rate and reduced apparent affinities for both ATP and ssDNA, indicating that the C-terminal domain may be regulating the DNA binding and ATPase activities of drUvrD.
We then examined the helicase activity of drUvrD FL on 3′tailed, 5′-tailed and blunt dsDNA ( Figure 6B). Our data reveals that drUvrD unwinds all three of these substrates in an ATPdependent manner to varying extents and, as expected for a member of the SF1A helicase family, unwinds preferentially 3′tailed dsDNA ( Figure 6C and D and Figure S2). The length of the ssDNA overhangs did not significantly affect the helicase activity of drUvrD, since very similar initial rates of unwinding were observed for 15nt and 7nt overhangs ( Figure 6C). Although drUvrD shows a preference for unwinding 3′-tailed dsDNA, drUvrD also melts 5′-tailed DNA at a 3-fold lower rate and blunt dsDNA at a 9-fold lower rate. In the presence of drSSB (added at a 12.5-fold excess with respect to the DNA), the helicase activity on 3′-tailed and blunt dsDNA was unaffected, whereas drUvrD's activity on 5′-tailed dsDNA was strongly stimulated to a rate similar to that observed on 3′-tailed dsDNA ( Figure 6C and D and Figure S2). In these conditions, drUvrD could unwind duplexed DNA with the same efficiency in both directions. SSB has previously been reported to directly stimulate the helicase activity of several other helicases [39,40]. Interestingly, we also observed helicase activity on 3′tailed dsDNA containing a fluorescein-conjugated thymine within its ssDNA extension ( Figure S2) and succeeded in crystallizing drUvrD ∆C in complex with such a DNA. Data to 3.0Å resolution were collected on crystals with modified DNA at position 21 and extra electron density could be seen close to the C7 group of thymine 21 ( Figure S3). Several DNA helicases have previously been shown to unwind lesioncontaining DNA [41]. ecUvrD was previously shown to efficiently unwind thymine glycol containing DNA [42] and E. coli Rep can efficiently unwind a DNA substrate harboring a polyglycol linkage in the ssDNA extension [43]. These findings suggest that although UvrD helicases bind tightly to ssDNA, they are sufficiently flexible to allow bases with bulky modifications through their ssDNA gateway.

drUvrD translocates on ssDNA in the 3′-5′ direction only
We then investigated drUvrD's ability to translocate on ssDNA using a streptavidin-displacement assay (Figure 7). We found that drUvrD could efficiently release streptavidin bound to biotin located at the end of 5′-tails in an ATP-dependent manner, but failed to displace streptavidin bound to biotin at the end of 3′-tails ( Figure 7B and Figure S4), indicating that drUvrD translocates along ssDNA in the 3′-5′ direction. Here again, as in the helicase assays, addition of drSSB to the reaction mix did not affect translocation along 3′-tails, but significantly reduced the translocase activity on 5′-tails ( Figure 7C). Interestingly, the addition of drSSB leads to both a reduction in the amount of streptavidin-free 5′-tailed dsDNA (middle band on the gel, corresponding to the product of the translocase activity) and a major increase in the amount of released ssDNA (lower band on the gel, corresponding to the product of the helicase activity). drSSB therefore modulates the activity of drUvrD on 5′-tailed dsDNA. In the absence of drSSB, drUvrD preferentially translocates along the ssDNA in the 3′-5′ direction, whereas in the presence of drSSB that most likely binds to the 5′ ssDNA extension, drUvrD preferentially unwinds 5′-tailed dsDNA in the 5′-3′ direction through an as yet unidentified mechanism.

Mutagenesis study of drUvrD's DNA binding ability and helicase activity
Directionality of SF1 helicases is believed to be determined by preferential binding to either a 3′-or a 5′-ssDNA overhang, which acts as the entry point for the helicase [4]. We carried out fluorescence anisotropy measurements to evaluate the affinity of drUvrD for either 3′-or 5′-tailed dsDNA ( Figure 8A and Figure S5). drUvrD binds to both of these substrates with similar affinity. The binding of drUvrD to 3'-tailed dsDNA is slightly stronger than to 5′-tailed dsDNA (K d for 3′-tailed dsDNA: 0.36 µM, K d for 5′-tailed dsDNA: 0.48 µM). These values are also very close to the estimated affinity of drUvrD FL for ssDNA derived from our ATPase data (K ssDNA : 0.55 µM; Figure 6A). We mutated residues identified in ecUvrD [19] as being essential for DNA unwinding using the wrench-and-inchworm mechanism (Gly424 and Gly426 from the GIG motif and the βhairpin) and tested the DNA binding and helicase activities of these drUvrD mutants on 3′-or 5′-tailed dsDNA (Figure 8). Mutating Gly426 to threonine (G426T) did not significantly affect the binding of drUvrD to 3′-and 5′-tailed dsDNA, whereas mutating Gly424 to threonine (G424T) alone or together with the G426T mutation significantly impaired the binding of drUvrD to both 3′-and 5′-tailed dsDNA (K d values increased by 3-4 fold; Figure 8A and Figure S5). Deletion of the β-hairpin structure, which is known to act as a separation pin and is essential for ecUvrD's helicase activity [19], did not affect drUvrD's binding to 3′-tailed dsDNA and led to a slightly reduced affinity for 5′-tailed dsDNA ( Figure 8A and Figure S5). These mutations, however, had a much more dramatic effect on the helicase activities of drUvrD ( Figure 8B). Deletion of the β-hairpin dramatically reduced DNA unwinding of both 5′-and 3′-tailed dsDNA and this was also the case for the G426T mutant. In contrast, drUvrD-G424T mutants (single and double) showed a highly stimulated helicase activity on 3′-tailed dsDNA, as has previously been observed for ecUvrD [19], and a reduced activity on 5′-tailed dsDNA. These results suggest that drUvrD's 5′-3′ helicase activity relies on both a functional separation pin and tight binding to duplexed DNA via its GIG motif, whereas its 3′-5′ activity only requires the β-hairpin structure.

Discussion
Because of the helical nature of nucleic acids, helicases are expected to translocate along DNA in a spiral movement. For the first time, our structures trap this large-scale spiral movement and reveal how the combination of rotational and translational movements, associated with the positioning of the helicase at an angle relative to the dsDNA axis produce a spiral trajectory along the DNA duplex. In addition, our two higher resolution structures of drUvrD ∆C provide new insight into the detailed mechanisms underlying ATP-dependent DNA unwinding. Although the details of the protein-DNA contacts are not strictly identical in the structures of drUvrD, ecUvrD [19] and gsPcrA [18], taken together, our observations suggest that the molecular mechanisms underlying this complex process are highly conserved within the SF1A helicase superfamily and support the tightly regulated wrench-and-inchworm model. The main differences we observe concern the gating mechanism regulating the exiting of the ssDNA. As in previous crystallographic studies of SF1 helicases [17][18][19], our crystal structures reveal no direct protein-protein contacts between neighboring UvrD monomers, even in the crystal structure of the intact drUvrD FL , in which the duplexed DNA is significantly bent, bringing the two UvrD monomers close to each other ( Figure 3A).
Despite being structurally and mechanistically conserved with ecUvrD and gsPcrA, to our surprise, our biochemical assays revealed that drUvrD differs from its homologues in a number of ways. We found that, unlike ecUvrD, drUvrD could efficiently unwind dsDNA with only short (7nt) ssDNA overhangs. It is clear from our crystal structures that only a single drUvrD monomer can bind to such a short ssDNA tail. Although our data do not allow us to determine the active oligomeric state of drUvrD, our findings suggest that its helicase activity only requires that one UvrD monomer be loaded on the ssDNA tail. Our data also revealed that drUvrD can efficiently translocate along ssDNA with a biased 3′-5′ directionality as observed previously for ecUvrD [44][45][46], but in contrast can melt both 3′-and 5′-tailed DNA duplexes. This is consistent with our finding that drUvrD binds to both types of DNA. drUvrD also displayed a weak helicase activity on blunt DNA. Most members of the SF1A family show a clear 3′-5′ polarity [38,47,48]; there are, however, several examples of enzymes including the PcrA helicase, notably in gram-positive bacteria, that show bipolar helicase activity [49][50][51][52][53][54]. Several UvrD homologues are also known to act on blunt or nicked DNA [34,47,49,54,55]. Our findings now provide further evidence that SF1A helicases vary both in terms of substrate specificity and helicase polarity.
Interestingly, our experiments carried out in the presence of drSSB, which is known to coat and protect nascent ssDNA in vivo, reveal that SSB plays an important role in modulating the balance between helicase and translocase activity on 5′-tailed dsDNA (Figure 9). The presence of SSB strongly favors the helicase versus translocase activity of drUvrD on such a substrate. This effect could be due to a direct regulation of drUvrD's activity by SSB or more likely to a steric effect of SSB binding to the ssDNA extension. In contrast, SSB does not appear to have any effect on drUvrD's activity on 3′-tailed dsDNA ( Figure 9).
Our mutagenesis, DNA binding and helicase activity data indicate that regardless of the DNA substrate, the GIG motif of drUvrD is critical for DNA binding and the β-hairpin structure is essential for DNA unwinding of both 5′-and 3′-tailed DNA substrates. The GIG motif and the β-hairpin separation pin are two essential features of the wrench-and-inchworm mode of unwinding and appear to be involved in both the 3′-5′ and on 5′-3′ helicase activities of drUvrD. However, we also observe that mutating Gly424 from the GIG motif has a very contrasted effect on 3′-5′ and on 5′-3′ helicase activity (stimulated 3′-5′ activity and reduced 5′-3′ activity), indicating that the GIG motif from domain 2B may be regulating these two processes differently. A number of ecUvrD mutants, including GIG mutants, are known to display reduced DNA binding and yet robust 3′-5′ helicase activity as observed for drUvrD [19,56]. This has been proposed to result from an alternative mode of unwinding, known as strand-displacement, in which movement of ssDNA is deregulated due to reduced contacts with dsDNA. This mechanism has been reported, notably in the absence of domain 2B and duplex DNA binding [19,57]. In the case of drUvrD, impaired DNA binding may cause drUvrD to switch from the controlled wrench-and-inchworm to an unregulated strand-displacement mode of unwinding on 3′-tailed DNA. In such a mode, the rotational movement of domain 2B is no longer coupled to ATP binding and hydrolysis and as a result domain 2B is no longer needed and may adopt a more open conformation, as observed in the DNA-free structures of gsPcrA and ecUvrD [16,58].
The targeting and involvement of helicases in distinct cellular repair processes thus appears to be achieved by their abilities to bind and unwind specific structures corresponding to intermediates of these processes. For example, the 5′-3′ unwinding activity of Staphylococcus aureus PcrA helicase is greatly stimulated in the presence of specific DNA structures [50]. drUvrD's ability to unwind 5′-and 3′-tailed DNA duplexes and containing modified bases within the translocating strand may reflect its implication in diverse DNA repair pathways in vivo. In E. coli, recombinational repair has been proposed to involve the 3′-5′ helicases RecQ and Helicase IV and the 5′-3′ helicase RecD, while D. radiodurans cells missing these genes show wild-type radioresistance and DNA repair capacity [7,59]. In contrast, inactivation of drUvrD leads to a significant increase in the sensitivity of cells to γ-irradiation [7]. This phenotype is further enhanced in cells in which both uvrD and recD2 genes have been disrupted, suggesting that the 5′-3′ helicase, drRecD2, may in part back-up drUvrD's function. While further studies will be needed to decipher the detailed molecular mechanisms that regulate the helicase activities of drUvrD, these observations suggest that in vivo both helicase activities of drUvrD are needed. drUvrD may switch between its translocase and helicase activities in response to external stresses, changes in its environment, or, as suggested by our experiments in the presence of SSB, upon interactions with pathway-specific protein partners such as SSB, MutL or UvrAB [9,55,60]. Table S1. Sequences of DNA oligonucleotides used in this study.

Supporting Information
(DOCX) Figure 9. Model of DNA duplex unwinding and ssDNA translocation by drUvrD. Models of drUvrD DNA unwinding and ssDNA translocase activity on 5' tailed dsDNA (top) and 3' tailed dsDNA (bottom) in the absence (left) and presence (right) of drSSB. Using 5' tailed dsDNA, in the absence of drSSB drUvrD has low 5'-3' helicase activity and high 3'-5' translocase activity while, in the presence of drSSB, drUvrD has high helicase activity and low translocase activity. Using 3'-tailed dsDNA, drUvrD has high 3'-5' helicase activity and no 5'-3' translocase activity, regardless of the absence or presence of drSSB. Binding of fluorescein-labeled DNA to drUvrD ∆C . The DNA oligonucleotides contain a fluoresceinconjugated thymine at position 21 within the ssDNA extension. The 2Fo-Fc electron density map (blue) is contoured at 1σ, while the Fo-Fc difference density map (green) is contoured at 2.5σ. The ssDNA is illustrated in sticks. (TIF) Figure S4. Translocase activity on 3′-and 5′-tailed dsDNA. drUvrD (250nM) translocation activity on streptavidin bound 5′and 3′-tailed 25 base-pair dsDNA (20nM) with 25nt ssDNA extensions. Reactions were stopped at the following time points: 0, 30sec, 1min, 2min, 5min, 10 min and 15min, prior to separation on 10% TBE gels. The fluorescein and the biotin labels are illustrated respectively as a star and an open circle in the schematic representation of the DNA. The upper bands correspond to streptavidin-bound dsDNA substrate, the middleband to the released dsDNA (translocase product) and the lower band corresponds to the product of the helicase activity of UvrD, i.e. ssDNA. Biotinylated dsDNA without streptavidin was loaded in the first well. (TIF) Figure S5. DNA binding to 3′-and 5′-tailed dsDNA. Binding of wild-type (WT) and mutant drUvrD to 3′-(A) and 5′-tailed dsDNA (B) was measured by fluorescence anisotropy. The anisotropy measured for the DNA alone was subtracted from all other values and the change in anisotropy (ΔA) is plotted as a function of UvrD concentration (µM). The averaged data points were fitted to a standard binding equation assuming a single binding site using GraphPad Prism6. Standard deviations are shown as vertical bars. (TIFF)