Previous crystallographic and biochemical studies have suggested that the exceptional stability of UvrB preincision complexes results from insertion of a flexible β-hairpin element between the two DNA strands at an as-yet undetermined locus within a ~6 base-pair (bp) unpaired duplex segment “bubble” containing the lesion [6]. Under physiological conditions, the energetically demanding process of wresting open duplex DNA at the lesion and inserting the β-hairpin is fueled by ATP hydrolysis by both UvrA and UvrB [21]. The same energy-intensive conditions that load UvrB onto a lesion promote dissociation of the two (vide infra), and this has thwarted all previous attempts to crystallize a UvrB preincision complex. We reasoned that it might be possible to synthesize a UvrB preincision complex through a stepwise assembly process that would eliminate the need for ATP expenditure and kinetically trap the complex from dissociation, while restricting the roaming range of UvrB on DNA. In previous studies, we [22–25] and others [26, 27] have employed a synthetic crystallography approach centered upon the ability of intermolecular disulfide-crosslinking (DXL) to trap otherwise transient intermediates in protein-DNA recognition [28]. Our studies build upon the groundbreaking work of van Houten, Kisker, and Goosen, which revealed that UvrB extrudes from the DNA a nucleoside in the strand clamped by the β-hairpin loop (hereafter called ‘inner’ strand) [17, 29]; the extruded nucleobase inserts into the highly conserved hydrophobic pocket. We therefore introduced a mutant Cys residue into the hydrophobic pocket and disulfide-crosslinked the resulting UvrB(T251C) protein to a single-stranded DNA substrate containing a thiol-tethered nucleobase, a cytosine. We purified this species by ion exchange and then annealed the complementary strand (the outer strand) to produce UvrB bound to duplex DNA (Figures 2(a) and 2(b)). The complimentary strand is designed to contain two pyrimidine-pyrimidine mismatches on the immediate 5′ side of the modified cytosine; these were introduced to destabilize the duplex in the region expected to form a bubble, thereby facilitating maintenance of the helix-penetrated state for the β-hairpin. The corresponding DNA duplex containing no mismatched base-pairs failed to yield diffraction-quality crystals.

To establish the physiologic relevance of the preincision complex trapped by this approach, we subjected UvrB(T251C) synthetically bound to 5′ end labeled 32-mer DNA duplex to UvrC incision assay. Although incision efficiency is low, presumably because DNA tethering prevents UvrB from properly positioning incision sites for endonuclease activity, the UvrB-bound DNA duplex was recognized and acted upon by the downstream endonuclease UvrC, as judged by UvrC having made an ATP-dependent 5′-incision in the inner strand tethered to UvrB (Figure 2(c)). This experiment establishes that stepwise complex assembly governed by DXL yields a viable, physiologically relevant preincision complex. Furthermore, as incision is known to occur exclusively on the lesion-containing strand, this result shows definitively that the inner strand in our complex is bound in such a way as to mimic a bona fide lesion; hence the inner strand is the lesion-containing strand. Incision is known to take place at the 8th phosphodiester bond 5′-to the lesion, thus allowing us to infer that the enforced extrahelical cytosine in our structure is being presented by UvrB to UvrC as though it were a bona fide lesion, and UvrC is recognizing it as such (Figure 2(d)). Further biochemical evidence in favor of these assertions is given below.

UvrB is classified as a member of SF2 helicases on the basis of conserved helicase motifs (I-VI) [30] (Supplementary Figure 1). It contains the classical RecA-like domains (domains 1a and 3), in addition to three auxiliary domains (domains 1b, 2, and 4), plus the element aforementioned β-hairpin that projects from domain 1a. We determined crystal structures of UvrB disulfide-crosslinked to a 20-mer duplex substrate, in both the presence and absence of ATP, at 2.6 Å and 2.8 Å resolution, respectively (Figure 3 and Table S1). The structures presented herein contain UvrB domains 1a through 3 (residues 1-593); domain 4, which lacks electron density in the previous structure due to structural polymorphism [16], was omitted for the structural studies but retained for biochemical experiments. For those crystals grown in the presence of ATP, we clearly observe a nucleotide bound to the ATP binding site located at the interdomain cleft within helicase domain. However, the electron density corresponds to ADP plus inorganic phosphate (Figure 3(e)); hence we conclude that the ATPase site retains activity in our crosslinked complex.

As expected on the basis of the synthetic crystallography approach used to trap these structures, the 13th nucleobase on the inner strand, a cytosine, is extruded from the duplex and inserted into the hydrophobic pocket between domains 1a and 1b (Figure 3(f)). As predicted, in our structures the β-hairpin loop was found to be fully inserted between the two DNA strands, clamping the inner strand between the β-hairpin and domain 1b (Figures 3(a)–3(c)). Though insertion of the β-hairpin loop requires disruption of only one or two base-pairs at the insertion site, our structures reveal disruption of base-pairing throughout the entire 6 bp duplex region comprising the crosslinked cytosine. This disruption of intact Watson-Crick base-pairs appears to be accomplished by UvrB-induced unwinding of the duplex substrate immediately upstream of the lesion. The size of the induced bubble observed in our complex is fully consistent with the bubble size as determined via biochemical experiments [31].

Previous DNA-bound structures of UvrB bore only fragmentary nucleic acids, leaving uncertain the overall disposition of the DNA duplex with respect to the protein. In our structures of a lesion-mimetic recognition complex, the duplex runs along the entire long axis of UvrB, contacting every domain except for domain 2 (Figures 3(a)–3(c)). The backbone phosphates of the inner DNA strand make extensive contacts to the domains 1a and 3 (Figures 3(d) and 4(b)). The outer strand also makes contacts to UvrB, but these comprise almost exclusively van der Waals interactions (Figure 3(d)). In agreement with previous EM studies [32], the UvrB-bound DNA contains a sharp bend (~60°) localized at the junction between the bubble and the fully base-paired segment flanking on its left side (“left flank,” Figure 3(d)). Bending appears to increase the accessibility of the 5′-incision site experimentally determined here, which is located on the left flank, in addition to drawing the 5′-incision site closer in space to the 3′-incision site (Figure 3(a)). Flanking the bubble and hairpin insertion site on the right side is the fully base-paired duplex segment “right flank” containing the inferred 3′-incision site (5th phosphodiester bond 3′ to the lesion); this too is almost completely exposed (Figure 3(c)). Thus, our structures reveal that UvrB grasps the duplex so as not only to recognize the lesion, but, as hypothetically proposed previously [33], also to present the scissile phosphates indeed on the same face of the DNA duplex and with optimal accessibility and geometry.

The β-hairpin loop is a structural motif that is commonly employed by DNA helicases as a “pin” to separate two DNA strands [34–36]. In the current structures, we observed that the β-hairpin unwinds the duplex DNA in the 5′ to 3′ direction. The contacts between the β-hairpin loop and the inner strand are almost identical to the previously determined structure of UvrB bound to DNA containing a simple stem-loop structure [17]. The DNA in that structure contains a single-stranded extension projecting from the stem, and the first nucleoside of that extension adopts an extrahelical conformation and is inserted into the same slot on UvrB as is the extrahelical lesion-mimetic cytosine in our structures. A triad of three strictly conserved and functionally critical tyrosine residues located at the inner face of the β-hairpin (Y92, Y93, and Y96) occupy the space vacated by the extrahelical nucleoside; Y92 and Y93 form hydrogen bonds to phosphates neighboring the extruded nucleotide, and Y96 forms π-stacking interaction with the base directly 5′ to the extruded nucleotide on the inner strand (Figure 3(d) and Supplementary Figure 2). The extruded nucleobase undergoes ~90° rotation around its glycosidic bond to be fully embedded within the hydrophobic pocket constricted by type-conserved residues I306 and F249 (Figure 3(f)). With respect to the outer strand, helix penetration by the β-hairpin results in three extrahelical bases. The first two of them are stacked over Y95, which is also important for formation of a stable preincision complex [37].

Previous crystallographic studies on SF2 helicases demonstrated that ATP binding and hydrolysis are associated with rigid-body movements within the RecA-like domains, alternating between “open” and “closed” states and that this movement repositions two nucleic-acid binding surfaces, thereby driving translocation of the enzyme along the nucleic acid [38]. On the basis of its conserved helicase motifs, the RecA-like domains of UvrB (domains 1a and 3) are likely to undergo similar conformational changes in response to ATP-binding and ATP-hydrolysis. If so, the observed contacts between the helicase motifs engaged in DNA binding and the inner strand phosphates suggest that our structures obtained in both the presence and absence of ATP (hydrolyzed to ADP and Pi in crystallo) represent “open” states (Figure 4(a)). The hallmark of the open state is that, in it, the successive string of phosphates contacted by the two RecA-like domains is separated by one phosphate (Figure 4(b) and Supplementary Figures 3(a) and 3(c)) [34, 39], while in the closed state the respective string of phosphates forms a continuous array of contacts [39–41] (Supplementary Figure 3(b)). Even though our structures represent the open state, we note that the domains 1a and 3 from these structures could be individually superimposed on their respective counterparts in closed structures of SF2 helicases with no major steric clash, suggesting that UvrB can also adopt the closed state on DNA (Figure 4(a)).

Unlike other helicases studied to date, UvrB catalyzes the translocation of a “bubble” along a stretch of otherwise fully paired duplex DNA, becoming positionally fixed only upon encounter with a lesion. This unique interaction mode of UvrB is apparently enabled by extensive protein contacts made to both strands of duplex DNA (Figure 3(d)). Of particular note is the loop preceding motif VI (F-loop, residues G526-S530) that approaches but does not snugly contact the outer strand at the left flank (Figure 4(c)). A key feature of the F-loop is the functionally critical residue F527 [33], which in our structure is lodged deeply in the minor groove, but not so deeply as to intercalate between DNA bases. If the UvrB helicase domain underwent rigid-body movements similar to those seen for other helicases as proposed previously [33], ATP-dependent domain closure would force the F-loop toward DNA (Figure 4(d)), perhaps causing F527 to intercalate into the duplex stack, which may function as a “ratchet” to facilitate unidirectional translocation upon ATP-hydrolysis. Indeed, in a structure we determined of UvrB (S91C) bound to a fully base-paired DNA duplex, F527 is intercalated into a sharply bent DNA duplex (Supplementary Figures 4(a) and 4(c)); furthermore, the crystal structure of the analogous complex containing a F527A mutation revealed substantially reduced DNA bending (Supplementary Figure 4(b)) and increased DNA disorder presumably due to the loss of “ratchet” that seems to lock the duplex DNA in position (Supplementary Figures 4(d) and 4(e)). That F527 intercalation is important functionally suggested by the reduced incision efficiency observed for the F527A mutant UvrB [33].

The ATPase activity of UvrB is known to be stimulated by DNA binding [42], and the present structures suggest a mechanism for such activation. Crystal structures of several related helicases have revealed that helicase motif V binds to both nucleic acid and ATP, thereby coordinating ATPase-driven translocation along the nucleic acid strand [39, 40]. However, in all UvrB structures lacking duplex DNA, R506 in motif V forms hydrogen bonds with S477 in motif IVa, resulting in disengagement of motif V from the ATP binding site (Figures 4(e) and 4(f)) [16, 17, 43]. In our structures, however, these interactions are disrupted by the inner DNA strand binding into the shallow depression formed between motifs IVa and V. This severs the connection between motifs IVa and V, releasing motif V to contact directly all three phosphate moieties of ATP (or of ADP and Pi) through the main chain of G508 and the side chain of D510, while simultaneously stabilizing the catalytically essential residues R540 and R543 of motif VI to form hydrogen bonds with the β- and γ-phosphates (Figures 4(e) and 4(g)). These newly observed contacts facilitate the proper positioning of the γ-phosphate with respect to E339 of the DExH motif (motif II), the side chain carboxylate of which is believed to activate a water molecule for nucleophilic attack on the γ-phosphate. Consistent with our observations, mutagenesis studies showed that all motifs V and VI residues mentioned above are critical for ATPase activity or the formation of a preincision complex or both [44, 45].

In order to provide additional biochemical confirmation of our assignment for the lesion-containing DNA strand, we incorporated a well-established UvrB substrate, fluorescein-adducted thymine (Flu-dT), into a duplex DNA disulfide-crosslinked to UvrB (T251C). The lesion was positioned downstream of the β-hairpin either on the inner or outer DNA strand. These poised complexes were then treated with DTT to release the enzyme from its staging point, and ATP was included to power duplex translocation (Figure 5(a)). Under these conditions, a persistent UvrB-DNA complex was observed only when the lesion was positioned on the inner strand, indicating that UvrB can recognize damage on the inner strand, but not the outer strand, while translocating along the inner strand in the 5′ to 3′ direction (Figure 5(b)). Next, we subjected UvrB(T251C) to a UvrA-assisted DNA binding assay with duplex DNA substrates containing a cystamine crosslinker (XL) and a Flu-dT at various positions either on the same strand or opposite strand (Figure 5(c) and Supplementary Figure 5(a)). Although substrates amongst each set showed similar UvrB binding as judged by native PAGE, efficient crosslinking was observed only when the lesion was positioned on the 5′-side of the crosslinker on the same strand (Figures 5(d) and 5(e) and Supplementary Figures 5(b) and 5(c)). Since crosslinking is dependent upon extrusion of the XL nucleobase from DNA, these data are consistent with the notion that (i) UvrB catalyzes nucleobase extrusion and hence presumably bubble formation while scanning DNA for damage and (ii) the lesion-containing strand is the inner strand.

B. caldotenax UvrA, UvrB, and UvrC proteins were cloned into pET-28a(+) vector (Novagen) and expressed in E. coli BL21 (DE3) pLysS using standard protocol. Harvested cells were lysed using sonication, followed by centrifugation at 35,000 g to clear the lysate. Subsequently, all proteins were purified using nickel-nitrilotriacetic acid (Ni-NTA) agarose (Qiagen), Heparin agarose (GE Healthcare), and size-exclusion chromatography using HiLoad 16/600 Superdex 200 (GE Healthcare). Cysteine residues for disulfide crosslinking (T251C or S91C) and a stop codon to generate domain 4 deletion construct (Δ594-658) were introduced to the UvrB expression vector using a QuikChange II site-directed mutagenesis kit (Agilent Technologies). To obtain homogeneous UvrB-DNA disulfide-crosslinked complexes, all native cysteine residues in B. caldotenax UvrB (C144, C211, and C303) were mutated to serine for both structural and biochemical studies. Only for crystallographic studies, His-tag was cleaved from UvrB by overnight incubation with TEV protease at 4°C prior to size-exclusion chromatography. UvrA and UvrC were frozen in 25 mM Tris pH 7.4, 250 mM NaCl, 20% (v/v) glycerol, and 5 mM βME at -80°C until use. UvrB was stored in the same buffer but was lacking βME.

All site-specifically modified single-stranded oligonucleotides were synthesized on ABI 392 DNA synthesizer (Applied Biosystems) using standard protocol and reagents in 1 μM scale, purified by urea-PAGE, and desalted by Sep-Pak columns (Waters). Oligonucleotides containing a disulfide-bearing tether at the N⁴-position of a cytosine or at the nonbridging position of DNA backbone were synthesized and functionalized as described previously [48, 49]. Oligonucleotide modification was confirmed by MALDI-TOF. All unmodified oligonucleotides were purchased from Eurofins Genomics.

For assembly of UvrB(T251C)-dsDNA complexes for crystallization, a single-stranded 20-mer oligonucleotide containing a thiol-tether attached to a cytosine (100 μM, 5′-GCTCTAGATTTTC\astATACGGC-3′, where C\ast denotes a disulfide-tethered cytosine) was mixed with a truncated UvrB mutant (40 μM, Δ594-658) in 25 mM Tris pH 7.4, 200 mM NaCl, at 16°C for 3-5 days. UvrB disulfide-crosslinked to a single-stranded oligonucleotide was purified by Mono-Q anion exchange chromatography (GE Healthcare) using a linear salt gradient of 95% Buffer A (25 mM Tris pH 7.4) to 25% Buffer B (25 mM Tris pH 7.4, 2.0 M NaCl). Fractions containing the complex were pooled and mixed with 1.5 molar excess of the complementary oligonucleotide containing two pyrimidine-pyrimidine mismatches (5′-GCCGTATGCCAATCTAGC-3′, where mismatched bases are underlined) for annealing at 4°C for 18 hrs. UvrB(S91C)-dsDNA and UvrB(S91C/F527A)-dsDNA complexes were obtained similarly by assembling 16-mer oligonucleotides (5′-TCTCCATCG_CGCTACC-3′, where the underlined position indicates the site of backbone modification to a phosphoramidite with a cystamine and 5′-GGTAGCGCGATGGAGA-3′). Purification of UvrB disulfide-crosslinked to the duplex DNA was performed on a Mono Q using a linear salt gradient of 95% Buffer A to 30% Buffer B, followed by size exclusion chromatography on Superdex75 10/300 (GE Healthcare) using buffer (25 mM Tris pH 8.0, 150 mM NaCl).

All crystals were obtained by the sitting-drop vapor diffusion method at 20°C. UvrB(T251C)-dsDNA complex was crystallized in 100 mM Bis-Tris (pH 6.5), 100 mM ammonium acetate, and 17-19% (w/v) polyethylene glycol (PEG) 8000 at 6.5 mg/ml complex concentration. Crystals of UvrB(T251C)-dsDNA bound to ADP and Pi were grown in similar conditions, except the reservoir that contained 2 mM ATP and 5 mM MgCl₂. UvrB(S91C)-dsDNA-ADP crystals were obtained in 0.1 M HEPES pH 7.5, 0.02 M MgCl₂, and 20-22% poly(acryl acid sodium salt) 5100 at 5 mg/ml complex concentration. UvrB(S91C/F527A)-dsDNA crystals were grown in 0.2 M potassium sulfate, 18-20% PEG3350, 2 mM ADP, and 5 mM MgCl₂ at 5 mg/ml complex concentration, but the electron density corresponding to ADP was not observed. All crystals were transferred to a reservoir solution supplemented with 20% (v/v) glycerol and then frozen in liquid nitrogen for data collection. Diffraction datasets were collected at -170°C at the 24-ID-C and 24-ID-E beamlines (NE-CAT) of the Advanced Photon Source. Datasets were processed with the HKL program suites [50]. Initial molecular replacement solutions were obtained by PHASER in the CCP4 suite [51], using the coordinates of previously determined UvrB structures as search models. UvrB-dsDNA models were built through iterative cycles of manual model building in COOT [52] and structure refinement using REFMAC5 [53] and PHENIX [54]. The Ramachandran plots, calculated by MolProbity [55], contain no residues in disallowed regions for all structures. All the structure model figures in the paper were prepared using PyMol (The PyMOL Molecular Graphics System, Version 1.3, Schrödinger, LLC.). Full details on the data collection and structure refinement are contained in Supplementary Table 1. The coordinates and structure factors have been deposited in the Protein Data Bank under accession codes 6O8E, 6O8F, 6O8G, and 6O8H.

UvrB(T251C)-dsDNA complexes for UvrC incision assays were prepared via disulfide-crosslinking and stepwise strand assembly using a crosslinker-containing 32-mer single-stranded oligonucleotide (5′-TATGATGGAGACTCTTTTTTC\astGCGGCAATTTT-3′, where C\ast denotes a disulfide-tethered cytosine) and 32-mer (5′-TAAAATTGCCGCGAAAAAAGAGTCTCCATCAT-3′) complementary oligonucleotides. In order to label both strands of the duplex DNA bound to UvrB(T251C), 10 fmol of the purified UvrB(T251C)-dsDNA complexes were added to 1 μCu γ-[³²P]-ATP (7,000 Ci mmol⁻¹, Perkin Elmer) and 5 units of T4 polynucleotide kinase in 1X PNK buffer (New England Biolabs) for 1 hour at 37°C. The 5′ end labeled UvrB-dsDNA complexes (2 nM) were then incubated with 5-100 nM UvrC in 50 mM Tris pH 7.4, 100 mM NaCl, 0.1 mg/ml BSA, 1 mM ATP, 10 mM MgCl₂, and 5% (v/v) glycerol at 55°C for 30 minutes. Reactions were terminated by adding an equal volume of a stop buffer (20 mM EDTA, 95% formamide, 0.05% bromophenol blue, and 0.05% xylene cyanol). The reaction mixtures were analyzed on 15% or 23.5% denaturing polyacrylamide gels and visualized via phosphor-imaging using a Typhoon phosphor imager and ImageQuant software (GE Healthcare).

UvrB(T251C) bound to a DNA duplex containing a single fluorescein-adducted thymine on the inner strand (5′ TACCCCGTTGCATCC\astTAG(Flu-dT)TCCCCAAAATTTT 3′ paired with 5′-AAAATTTTGGGGAACTAGGATGCAACGGGGTA-3′, where C\ast and Flu-dT denote a disulfide-tethered cytosine and fluorescein-adducted thymine, resp.), DNA duplex containing a fluorescein lesion on the outer strand (5′ TACCCCGTTGCATCC\astTAGATCCCCAAAATTTT 3′ paired with 5′-AAAATTTTGGGGT(Flu-dT)FCTAGGATGCAACGGGGTA-3′), or DNA duplex containing no lesion (5′ TACCCCGTTGCATCC\astTAGTTCCCCAAAATTTT 3′ paired with 5′-AAAATTTTGGGGAACTAGGATGCAACGGGGTA-3′) were prepared via disulfide crosslinking and stepwise strand assembly and 5′ ends labeled as described above. Complexes (2 nM) were then incubated in 50 mM Tris pH 7.4, 100 mM NaCl, 0.1 mg/ml BSA, 1 mM ATP, 10 mM MgCl₂, 5% (v/v) glycerol, and 5 mM DTT for 15 minutes at 55°C. The reaction products were analyzed on 6% native polyacrylamide gel containing 1 mM ATP and 10 mM MgCl₂ and visualized via phosphor-imaging using a Typhoon phosphor imager and ImageQuant software (GE Healthcare).

50-mer single-stranded oligonucleotides containing a thiol-tether (Supplementary Figure 5) were radioactively labeled at 5′ ends and then annealed with their respective complementary oligonucleotides in a 1.2-fold molar excess in buffer (10 mM Tris pH 8.0). The labeled duplex substrates (2 nM) were incubated with 2.5 nM UvrA, 100 nM UvrB in 50 mM Tris pH 7.4, 100 mM NaCl, 0.1 mg/ml BSA, 1 mM ATP, 10 mM MgCl₂, and 5% (v/v) glycerol for 30 minutes at 55°C. Each reaction mixture was split into two and analyzed by both 6% native-PAGE and 6% SDS-PAGE in nonreducing conditions and visualized as describe above. The relative radioactivities of the bands were evaluated employing the ImageQuant software (GE Healthcare). All reactions were completed in triplicate and the data are presented as average ± standard deviation.