An atypical AAA+ ATPase assembly controls efficient transposition through DNA remodeling and transposase recruitment

The IstB AAA+ domain is a close structural paralog of bacterial helicase loaders that can form right-handed helical oligomers

IstB consists of a small N-terminal domain of unknown function and a C-terminal AAA+ domain (Fig. 1C and Fig. S2A). To begin to dissect IstB structure and mechanism, we purified and crystallized the isolated ATPase domain (residues 63–251, termed IstB_AAA+) of the Geobacillus stearothermophilus IS21 family member, IS5376, with the non-hydrolyzable ATP-analog, ADP·BeF₃. Single-wavelength anomalous dispersion (SAD) phasing from selenomethionine-labeled protein was used to determine the structure; during model building, clear electron density was evident for bound nucleotide (Fig. S2B). The final IstB_AAA+ model was refined to a resolution of 2.1 Å, with an R_work and R_free of 18.2/20.3% (Table S1).

IstB_AAA+ adopts a typical AAA+ ATPase fold, in which five parallel β strands are sandwiched on both sides by α helices (Fig. 1D). Examination of the structure reveals that IstB possesses one of the characteristic signatures of the DNA replication initiator and helicase loader group of AAA+ proteins, namely the presence of an extra α-helix in its ATPase fold, denominated the Initator/loader Specific Motif (ISM) (Fig. 1D) (Dueber et al., 2007; Iyer et al., 2004). The existence of this helix defines IstB as a member of the initiator clade of AAA+ proteins, which includes the bacterial helicase loader DnaC/DnaI, and the bacterial DnaA and archaeal/eukaryotic Cdc6/Orc1 replication initiator proteins (Iyer et al., 2004). Despite relatively limited sequence identity (~15–22%), computational searches of the PDB (using DALI, (Holm and Sander, 1995)) show that the AAA+ domain of IstB is most closely related to that of DnaC/I, followed by DnaA (Fig. 1E).

Phylogenetic studies had indicated previously that IstB was likely to possess AAA+ family ATPase motifs involved in nucleotide binding and hydrolysis (Koonin, 1992). The structure reveals that the active site indeed assumes a canonical AAA+ configuration (Fig. 1F). The Walker-A residues of IstB_AAA+ form a phosphate-binding loop (P-loop) in which Lys111 directly contacts the β-phosphate of ADP, while the Walker-B amino acids Asp166 and Glu167 interact with the water coordination shell of an associated Mg²⁺ ion. The “Sensor II” arginine (Arg237) of the protein lies within the most C-terminal α-helix, which forms a truncated variant of the “lid” subdomain found in a majority of AAA+ proteins (Fig. S2C).

AAA+ proteins frequently assemble into large ring-shaped or helical homo-oligomers (Erzberger and Berger, 2006). Inspection of intermolecular contacts seen in the crystal structure reveals that IstB_AAA+ protomers self-associate around a crystallographic 6₁ axis, creating a right-handed filament reminiscent of those adopted by the ATPase domains of DnaC and the MuB transposase regulator (Fig. 1G) (Mizuno et al., 2013; Mott et al., 2008). Formation of this spiral allows a conserved arginine from one promoter (Arg221) to interact with the active site of a neighboring AAA+ domain (Fig. 1F), implicating this residue as a candidate “arginine finger” motif that typically participates in both ATP hydrolysis and inter-subunit communication. Interestingly, analysis of the electrostatic potential of the helical IstB_AAA+ assembly shows that numerous residues, located mainly in the ISM region, create a highly positively charged surface on one side of the ATPase domain oligomer (Fig. S2D).

Full-length IstB forms discrete oligomers

The observation that the IstB AAA+ region could adopt a structural state mirroring that of bacterial initiator/helicase loading proteins and ATP-bound MuB was intriguing. However, in some instances the crystallization of protein fragments can result in the formation of oligomeric states that only partially capture the quaternary arrangement of full-length subunits. For instance, although the ATPase domain of DnaC can crystallize as a continuous 6₁ helix on its own (Mott et al., 2008), it is restricted to forming a six-subunit helix in the presence of its target helicase, DnaB (Arias-Palomo et al., 2013).

To test whether an IstB_AAA+ filament was captured because its N-terminal domain was removed for crystallization, we used analytical size-exclusion chromatography to study the nucleotide-dependent oligomerization of the IstB ATPase domain, full-length IstB, and several IstB active-site mutants. The IstB_AAA+ protein elutes as a monomeric species, both in the presence of ADP and ATP, establishing that the domain on its own is unable to oligomerize efficiently under more dilute conditions than those used during crystallization (Fig. 2A). By comparison, full-length IstB formed an apparent dimer in the presence of ADP, suggesting that its N-terminal domain might support protomer-protomer interactions (Fig. 2B). Interestingly, when mixed with ATP, full-length IstB further shifted to a much larger size, eluting as a well-defined high-molecular weight peak with an apparent mass of ~350 kDa (Fig. 2B). To test whether this higher-order assembly depended on the AAA+ domain of IstB, we analyzed a number of mutants predicted to have defects in ATP binding and hydrolysis (e.g., Walker-A (K111A), Walker-B (E167Q), arginine finger (R221A), and sensor-II (R237A)). Following purification, each of the proteins formed dimers in the presence of ADP as per wild-type IstB (not shown). By contrast, none the mutants were able to shift to a higher molecular-weight species in the presence of ATP, with the exception of the Walker-B mutant, which formed a mix of higher and lower (~150 kDa) molecular mass complexes (Fig. 2C). Overall, these results indicate that full-length IstB does not form continuous filaments as has been seen for MuB and DnaC, and that both ATP and an intact active site are required for IstB to self-assemble into a discrete macromolecular complex.

IstB is a DNA binding protein

Although the exact function of IstB has been enigmatic, its relationship to known DNA binding proteins such as DnaA, DnaC, MuB, and TnsC suggested that it might associate with nucleic-acid substrates. To test this idea and analyze the role that nucleotide coordination and self-oligomerization might have on this activity, we used fluorescence anisotropy to assess the ability of IstB, IstB_AAA+, and our panel of IstB active-site mutants to engage DNA substrates. We first examined DNA binding using two 60 bp duplexes, one containing a random sequence, and another having the characteristic repeat sequence that flanks the transposon (the size of the DNA was chosen after screening different lengths of oligonucleotides for optimal affinity and binding behavior). As expected based on the behavior of MuB and TnsC, IstB did not show any significant sequence specificity (not shown); nevertheless, we continued using the transposon-related DNA for all subsequent analyses.

Treatment with nucleotide was found to stimulate duplex DNA binding by both IstB_AAA+ and the full-length protein (Fig. 2D). However, the apparent K_d of the full-length protein for DNA in the presence of ATP was ~30-fold higher compared to that of the AAA+ domain alone (K_d,app=40 nM vs. 1.5 µM, Table S2). Interestingly, the binding curves manifest by full-length IstB also displayed evidence of positive cooperativity, whereas the curves seen for the AAA+ region did not (Hill coefficient 2.5 vs. 0.85); this result, coupled with the relatively smaller dependency of the full-length protein for ATP for binding, indicates that the N-terminal domain contributes directly to interactions between IstB and DNA (Fig. 2D). Consistent with this idea, ATPase active-site mutations that significantly compromised DNA binding in the context of the isolated AAA+ domain had a much less pronounced effect when introduced into the full-length protein (Fig. 2E and 2F).

IstB homo-oligomerizes into a pentamer of dimers to remodel dsDNA

Intrigued by the ability of IstB to both bind DNA and form a discrete higher-order oligomer, we turned to single-particle cryo-electron microscopy (cryo-EM) to determine the architecture of the ATP-bound IstB complex with and without a client dsDNA substrate. To ensure complete assembly, we first dialyzed full-length IstB into a buffer containing ATP and subsequently ran the sample over an analytical S200 pre-equilibrated in dialysis buffer. Fractions containing the high-molecular weight IstB oligomer were then collected, diluted into a low salt buffer, and incubated with or without a dsDNA 60mer prior to EM analysis.

The cryo-EM structure of the ~290 kDa ATP-IstB complex bound to DNA was determined to subnanometer resolution (~9 Å), using an initial model derived from a random conical tilt reconstruction generated from a negatively-stained sample (Methods). The consistency of the structure is supported by the agreement between reference-free 2D class averages and forward reprojections of the cryo-EM model (Fig. S3). In contrast to other known AAA+ systems, IstB neither assembled into a closed/cracked ring nor an elongated filament, but rather formed a distinctive crescent-shaped particle composed of two C-shaped structures connected by five protruding “spikes” of globular density (Fig. 3A). Close inspection of the model reveals that the IstB oligomer is formed by five repeating elements (Fig. 3B), each consisting of two large lobes that connect to one another by two small rods which emanate from a single, small globular domain (Fig. 3B). A semicircular channel traverses the interior of the IstB oligomer, within which continuous, rod-like density is evident.

To determine which regions of the DNA-bound ATP-IstB complex correspond to specific domains of the protein and the nucleic acid, we fit available crystal structures into the cryo-EM reconstruction. To date, there is no atomic information of the ~7 kDa IstB N-terminal domain, although PSIPRED predicts it to be α-helical. We therefore attempted to localize the AAA+ fold using our IstB_AAA+ crystal structure. Notably, five copies of the ATPase domain, assembled in the same general helical conformation as found in the crystal structure, could be unambiguously fitted into each C-shaped region of density (Fig. 4A and 4B). Only a slight (<7°) rotation of each protomer was necessary to optimize the fit, suggesting that the oligomerization state seen crystallographically closely approximates that used by the full-length assembly.

The global fitting of tw0 pentamers of AAA+ ATPase domains, together with the close agreement between the crystal structure and the secondary structure elements resolved in the EM density, provide strong validation for the cryo-EM reconstruction (Fig. 4C). Given the location of the ATPase elements, the small lobes that protrude from the complex likely correspond to the N-terminal domain of IstB (Fig. 4D and 4E). Each of the five lobes are connected to the AAA+ domains through two short rods of density of the correct dimensions expected for α-helices; together with the two-fold symmetric organization seen in the model, this arrangement indicates that one role of the N-terminal region is to dimerize IstB protomers. This observation is in agreement with the gel-filtration data showing that full-length IstB forms dimers in the presence of ADP, and that the N-terminal region is required for this dimerization (Fig. 2A and 2B).

With the domains of IstB assigned, the only part of the reconstruction left unmodeled was the rod-like density located within the central channel. Notably, this region exhibits clear helical features, with alternating sets of wide and narrow grooves that identify it as duplex DNA (Fig. 4F). Interestingly, to fit within the central groove of the IstB decamer, the DNA must assume a highly bent 180° U-turn structure (Fig. 4G and S4A). To validate the placement of DNA, we determined a cryo-EM reconstruction of ATP-IstB in the absence of the duplex substrate; the difference map between the two structures confirms our assignment (Fig. 4H). Interestingly, in the context of the decameric complex, the highly positively-charged surface of the AAA+ pentamers is oriented towards the internal channel, forming a portion of the nucleic acid binding area (Fig. S4B). Moreover, inspection of the protein-DNA interface revealed that, in agreement with the fluorescence anisotropy data, some residues in the N-terminal region appear capable of participating in DNA binding (Fig. S4C).

Overall, the cryo-EM reconstruction establishes that, in response to ATP binding, IstB assemblies into a pentamer of dimers that engages and markedly reshapes target DNAs. This architectural arrangement is quite distinct compared to other known AAA+ ATPase systems characterized at this time, and was surprising given the close relationship of IstB to proteins such as DnaA and MuB, which form helical filaments. Inspection of the structure reveals why subunit assembly is limited to five IstB dimers: the inverted arrangement of two AAA+ domain pentamers, whose DNA binding surfaces face one another due to the molecular two-fold axes of their associated N-terminal domains, sterically prevents the binding of a sixth ATPase subunit to the complex (Fig. S4D). Thus, the N-terminal region is not only responsible for dimerizing IstB protomers, but also for preventing the formation of contiguous IstB oligomers.

IstB introduces positive supercoils in a nucleotide-dependent manner

Multiple biochemical studies have indicated that the supercoiling state of nucleic acid substrates can have important effects on the action of different transposition systems, including both IS21 and bacteriophage Mu (Harshey and Jayaram, 2006; Reimmann and Haas, 1990). Given extensive DNA deformation imparted by IstB in the cryo-EM reconstruction, we set out to investigate whether the protein might also alter DNA superstructure in solution. Using a topology-footprint assay and native agarose-gel electrophoresis (Methods), we found that IstB introduces DNA supercoiling in a dose-dependent manner and that ATP stimulates this activity (Fig. 5A and S5).

To determine whether IstB introduces positive or negative supercoils into DNA, we analyzed the reaction products by two-dimensional gel electrophoresis. Consistent with the right-handed solenoidal wrap seen in the EM reconstruction, the resultant topoisomer distribution revealed that wild-type IstB introduces positive supercoils in the DNA (Fig. 5B). Moreover, ATPase site mutants that lead to assembly defects of the IstB oligomer are highly compromised for supercoil stabilization, with the arginine-finger mutation displaying the most severe defect (Fig. 5C). Of the mutants tested, the sensor II alteration is the lone exception to this trend; however, we note unlike the arginine finger, which plays a key role in stabilizing trans interactions between subunits, the sensor II arginine acts in cis on its own active site. This configuration may impact inter-subunit interactions to a lesser degree than the arginine-finger, such that DNA binding can overcome sensor II-dependent oligomerization defects, allowing the IstB supercoiling signal to be detected. Overall, these experiments support a role for ATP-dependent self-association of IstB in creating an oligomer that not only bends but also overtwists target DNA duplexes.

ATP is required for IstB-IstA interactions

Several studies have demonstrated that helper proteins like TnsC and MuB not only interact with client DNA segments, but that they also directly engage their partner transposases to promote efficient strand cutting and pasting (Craig et al., 2002). To test whether IstB might interact with IstA in an analogous manner, we performed amylose resin pull-down assays using a transposase construct bearing a C-terminal MBP tag as bait and untagged IstB as prey. Experiments were first conducted with different nucleotides to test whether any interaction might be dependent on the identity of the nucleotide present in the reaction. Analysis of the pull-downs by SDS-PAGE revealed that when incubated with ADP, IstB did not associate with IstA-MBP (Fig. 6A). By contrast, IstB and IstA displayed a clear interaction when ATP was present, with the robustness of the association increasing when the slowly-hydrolyzable nucleotide analog, ATPγS, was added.

Having established that IstA binds to IstB in an ATP-dependent manner, we next sought to determine if this interaction required the formation of higher-order IstB oligomers. Using the isolated IstB AAA+ domain, we found no evidence of binding to IstA-MBP even in the presence of ATP, indicating that the ATPase region alone is insufficient for stable transposase engagement (Fig. 6A and 6B). Next, we assessed our panel of ATPase-site mutants in the context of full-length IstB. Neither the Walker A (K111A), arginine finger (R221A), nor sensor II (R237A) mutants showed an ability to interact with IstA; however, the Walker B (E167Q) mutant showed a strong interaction with the transposase when ATP was present, similar to that seen for the wild type protein in the presence of non-hydrolysable nucleotide. These results demonstrate that IstA directly engages IstB in an ATP-dependent manner, likely by recognizing the nucelotide-assembled form of the regulator.

IstA stimulates IstB ATPase activity and triggers decamer disassembly

Previous reports have established that the direct interaction of either TnsC or MuB with their cognate transposases stimulates the ATPase activity of the helper proteins. This stimulation, which is essential for strand transfer and target immunity, triggers DNA release and – in the case of MuB – filament dissociation (Craig et al., 2002). To analyze IstB ATPase activity and determine what effects IstA and DNA might have on nucleotide hydrolysis, we incubated full-length IstB with [γ-³²P]ATP and unlabeled ATP, and examined phosphate release using thin layer chromatography. IstB alone, as well as the panel of active-site mutants, showed relatively little ATPase activity overall either in the presence or absence of DNA (Fig. 6C and S6). However, the inclusion of IstA appreciably enhanced nucleotide turnover. Interestingly, the combined presence of DNA and the transposase also resulted in increased P_i release, although to a lesser extent than with IstA alone. All catalytic mutants of IstB were relatively resistant to stimulation by IstA, indicating that the observed hydrolysis activity came from the IS21 helper protein, and not from a potential contaminating ATPase.

Given that IstB oligomerizes when bound to ATP, and that IstA stimulates ATP turnover by IstB, we reasoned that transposase binding might trigger the dissociation of IstB assemblies. To test this idea, we first incubated IstB with ATP, with or without IstA, and then used an analytical gel filtration column, equilibrated in a buffer without nucleotide, to analyze complex formation. When IstB was mixed with ATP, the protein eluted as a single peak consistent with the formation of a decamer species (Fig. 6D). However, when the transposase was added to the reaction, the peak shifted significantly, overlaying with dimeric, ADP-bound IstB. Overall, these findings demonstrate that the IstA transposase stimulates IstB ATPase activity both in the presence and absence of DNA, and that this stimulation triggers decamer dissociation.

IstB utilizes a distinctive arrangement of ATPase domain/substrate interactions in binding compared to other AAA+ proteins

Our investigations into IS21 began with the determination of the crystal structure of the ATPase domain of the transposase chaperone, IstB, bound to a non-hydrolyzable analog (Fig. 1D). In accord with previous sequence predictions (Koonin, 1992; Mott et al., 2008), the structure demonstrates that IstB is a AAA+ ATPase, one that likely emerged from a common ancestor shared by DnaC/DnaI replicative helicase loading proteins and DnaA-family replication initiators (Fig. S1C). Using the IstB_AAA+ structure as a guide, mutagenesis-guided biochemical and biophysical experiments show that conserved ATP-coordinating residues of IstB are important for oligomerization, DNA binding/remodeling, and interactions with the IstA transposase (Figs. 2, 4–6). Inspection of inter-subunit contacts shows that – as with DnaC, DnaA and the ATP-dependent phage Mu transposase protein, MuB – the AAA+ domains of IstB possess an ability to self-assembly into right-handed helical oligomers upon ATP binding (Duderstadt et al., 2011; Erzberger et al., 2006; Mizuno et al., 2013; Mott et al., 2008). Strikingly, however, full-length IstB does not polymerize into elongated filaments, but rather uses two pentameric half-spirals to form a decameric, clamshell-shaped structure (Fig. 3 and 4). This organization is unique among AAA+ ATPases that have been characterized to date.

Numerous oligomeric AAA+ enzymes form rings or spirals that accommodate client factors such as DNA within a central pore to promote the detection, translocation, or remodeling of the associated substrate (Duderstadt et al., 2011; Enemark and Joshua-Tor, 2006; Kelch et al., 2011). Interestingly, despite using a common ATP-binding fold and relative positions of secondary structural elements to bind substrate, different AAA+ systems show a remarkable degree of plasticity in how they engage and reshape nucleic-acid segments (Fig. 7A). Surprisingly, we find that IstB does not bind the DNA duplex through the central pore of the AAA+ filament: instead, although the protein interacts with nucleic acid segments using its structurally conserved ISM signature motif, it does so with a different surface as compared to proteins that also possess such an element, such as DnaA or archaeal Orc1 (Fig. S7). Together with its distinctive assembly mechanism, this work reveals that even closely related AAA+ ATPases can undergo an extensive level of higher-order structural rearrangements and divergent adaptations throughout evolution, and that these modifications can lead to significant differences in overall mechanism.

A proposed role for IstB in IstA-mediated transposition

ATP-dependent transposase co-chaperones such as IstB, MuB and TnsC use nucleotide binding and hydrolysis to promote appropriate DNA strand cutting and pasting by their cognate transposases. As such, these three factors fall within the category of molecular matchmakers, proteins that enforce the correct assembly of multisubunit complexes to regulate essential biological processes such as DNA replication, transcription, and repair (Sancar and Hearst, 1993). Although a wide variety of nucleotide-coupled transposase systems have been identified and studied to date, how dedicated NTPase factors interact with DNA and help control transposase activity has not been well understood. For example, in Tn7, the TnsC ATPase is critical both for transposome assembly and modulating transposition (Gamas and Craig, 1992; Stellwagen and Craig, 1997), but how this element physically coordinates such tasks is not known. Similarly, while the MuB ATPase can form filaments on duplex DNA (Greene and Mizuuchi, 2002a, 2002b; Mizuno et al., 2013), how this protein engages a substrate duplex and interfaces with MuA to promote or repress transposition has yet to be defined.

The structure/function studies of IstB presented here provide new insights into the role of ATP-dependent transposase helper proteins, and into the action of AAA+-type co-chaperones for IS21 elements in particular. We find that full-length IstB forms dimers, and that in the presence of ATP, these dimers self-assemble into a decamer that sequesters and bends ~50 base pairs of DNA into a 180° U-turn (Fig. 4). Interestingly, alterations in nucleic acid structure, curvature, and supercoiling are known to be critical for many transposition and retroviral systems (Craigie and Mizuuchi, 1986; Davies et al., 2000; Hallet et al., 1994; Harshey and Jayaram, 2006; Hickman et al., 2014; Kuduvalli et al., 2001; Maertens et al., 2010; Montaño et al., 2012; Müller and Varmus, 1994; Pribil and Haniford, 2003; Pruss et al., 1994; Reimmann and Haas, 1990; Schmid et al., 1999). Although it is unclear why IstB should bend and overwind DNA upon binding, effects on local twist might help counteract the innate, negatively-supercoiled state of target DNA to render binding more energetically favorable. Alternatively, IstB’s action may help preferentially direct the protein to certain regions of the genome that have a desirable local topology. DNA bending, in turn, could be used to facilitate access of the active site of the transposase to the scissile bonds, as well as to help control the directionality of strand transfer, an approach that has been exploited by other transposition systems (Davies et al., 2000; Hickman et al., 2014; Maertens et al., 2010; Montaño et al., 2012). Indeed, the DNA configuration adopted in the presence of the IstB decamer is remarkably similar to ones seen for target DNAs bound to both the MuA transposase and the PFV (prototype foamy virus) retroviral integrase (Maertens et al., 2010; Montaño et al., 2012) (Fig. 7B). Consistent with these concepts, the absence of IstB drastically reduces transposition efficiency of IS21 and generates atypical duplications and/or deletions in the target DNA, indicating that IstB action is required for the correct alignment between the donor-transposase complex and the nucleic acid substrate (Schmid et al., 1999).

Given the duplex reshaping activities of IstB, together with the discovery that IstA specifically recognizes ATP-assembled IstB oligomers, we propose that IstB serves a landing pad that both marks and prepares DNA for IstA-mediated transposition of IS21 elements (Fig. 7C). In this model, an IstB decamer would first associate with a target DNA, which would then recruit IstA, along with a pair of pre-cleaved inverted IS21 repeats, to generate a synaptic complex. The inability of DNA to promote ATP turnover by IstB would impart sufficient stability to the complex to allow time for recognition by IstA. A potential problem arising from the direct recognition of a DNA bend by IstA is that IstB masks the target duplex; however, the capacity of IstA to stimulate ATP hydrolysis and IstB dissociation would be expected to promote clearing of the DNA, thereby allowing the transposase to access the target nucleic acid segment. Future studies aimed at imaging and biochemically characterizing higher-order IstA•IstB complexes in conjunction with DNA will be necessary to help establish the details of these interactions and events further.

Protein expression and purification

All proteins were overexpressed as N or C-terminal His₆-MBP fusion proteins in E. coli, using BL21codon-plus (DE3) RIL (Stratagene) or C41 (Lucigen) cells. Following lysis by sonication, proteins were purified by a combination of affinity chromatography, proteolysis to remove the fusion tags, and size-exclusion chromatography. Purified samples were concentrated and flash-frozen for subsequent studies.

Crystallization, data collection, structure solution and refinement

Se-Met IstB_AAA+ was concentrated to 8 mg/mL and spiked with 2 mM ADP·BeF₃. IstB_AAA+ crystals were grown by mixing 0.65 µL of protein solution with 0.65 µL of well solution containing 100 mM Bis-Tris propane pH 6.5, 300 mM ammonium sulfate, and 18.5% PEG 3350. Following harvesting and cryo-cooling, diffraction data were collected from a single crystal on Beamline 8.3.1 at the Advanced Light Source. Data were phased by SAD, and model building and refinement were performed with COOT and PHENIX (Adams et al., 2010; Emsley and Cowtan, 2004).

Analysis of IstB oligomeric state

IstB, IstB_AAA+ and IstB active site mutants were dialyzed over-night at 4°C into 20 mM HEPES pH 7.5, 300 mM NaCl, 10 mM MgCl₂, 10% glycerol, 1 mM β-mercaptoethanol, with either 1 mM ATP or ADP. 40 µL at 1.5 mg/mL of each protein were then run over a Superdex 200 5/150 GL analytical gel filtration column (GE Healthcare) in the same buffer and monitored by A₂₈₀.

DNA binding assays

Binding of a 5’-FAM-tagged dsDNA 60mer oligonucleotide was monitored at different IstB concentrations by fluorescence anisotropy using a Clariostar (BMG Labtech) microplate reader. Data were plotted and analyzed using PRISM.

EM and image analysis

IstB•ATP samples, both in the presence or absence of DNA, were either stained with 2% (w/v) uranyl acetate or frozen into liquid ethane for cryo-EM analysis. All EM data were collected using LEGINON (Suloway et al., 2005) and preprocessed with APPION (Lander et al., 2009). The initial random conical tilt reconstruction was obtained using XMIPP (Scheres et al., 2008) and refined with EMAN2 and SPARX (Hohn et al., 2007; Tang et al., 2007). Fitting of the atomic structure of the ATPase domain, and segmentation and rendering of the cryo-EM densities was done in CHIMERA (Pettersen et al., 2004).

DNA supercoiling assays

Nicked pSG483 plasmid, a derivative of pUC19, was used as a DNA substrate. Different amounts of IstB were incubated with the DNA for 20 min at 37 °C. Following ligation with T4 ligase, reactions were quenched and run for 18 h on 1% (w/v) TAE agarose gels. To analyze the effect that ATP and ADP might have on supercoiling activity, the plasmid was ligated with the ATP-independent E. coli ligase.

For two-dimensional gels, purified DNA fractions were run for 18 h at 4°C on a 1.2% (w/v) TPE agarose gel (36 mM Tris–HCl pH 7.9, 30 mM NaH₂PO₄, 1 mM EDTA pH 8.0). The gel was then washed three times for 30 min with TPE buffer supplemented with 4.5 µg/ml chloroquine, rotated 90 degrees, and run for 18 h at 4°C in TPE buffer plus 4.5 µg/mL chloroquine.

IstB and IstA pull-down experiments

Coprecipitation studies were performed by incubation of amylose beads (New England Biolabs) with bait and prey proteins in binding buffer. Following washing, samples were analyzed by SDS-PAGE.

ATPase assays

50 µL reactions containing different combinations of 10 µM IstB, 1 µM IstA (0–5 µM for IstA titration assay), and 10 µM dsDNA (random 60mer) were incubated during 1 h at 37 °C in a buffer containing 1 mM cold ATP and 5 nM [γ³²P]ATP (1.125 µCi). Quenched reactions were analyzed using thin-layer chromatography.

IstB disassembly

IstB (32 µM) was incubated in the presence or absence of IstA (17 µM) in 40 µL of reaction buffer (50 mM HEPES pH 7.5, 300 mM NaCl, 10 mM MgCl₂, 10% glycerol, 1 mM β-mercaptoethanol), supplemented with either 1 mM ATP or ADP, for 30 min at 37 °C. Samples were subsequently run in over a Superdex 200 5/150 GL analytical gel filtration column (GE Healthcare), pre-equilibrated in reaction buffer.