Single-molecule level structural dynamics of DNA unwinding by human mitochondrial Twinkle helicase

Knowledge of the molecular events in mitochondrial DNA (mtDNA) replication is crucial to understanding the origins of human disorders arising from mitochondrial dysfunction. Twinkle helicase is an essential component of mtDNA replication. Here, we employed atomic force microscopy imaging in air and liquids to visualize ring assembly, DNA binding, and unwinding activity of individual Twinkle hexamers at the single-molecule level. We observed that the Twinkle subunits self-assemble into hexamers and higher-order complexes that can switch between open and closed-ring configurations in the absence of DNA. Our analyses helped visualize Twinkle loading onto and unloading from DNA in an open-ringed configuration. They also revealed that closed-ring conformers bind and unwind several hundred base pairs of duplex DNA at an average rate of ∼240 bp/min. We found that the addition of mitochondrial single-stranded (ss) DNA–binding protein both influences the ways Twinkle loads onto defined DNA substrates and stabilizes the unwound ssDNA product, resulting in a ∼5-fold stimulation of the apparent DNA-unwinding rate. Mitochondrial ssDNA-binding protein also increased the estimated translocation processivity from 1750 to >9000 bp before helicase disassociation, suggesting that more than half of the mitochondrial genome could be unwound by Twinkle during a single DNA-binding event. The strategies used in this work provide a new platform to examine Twinkle disease variants and the core mtDNA replication machinery. They also offer an enhanced framework to investigate molecular mechanisms underlying deletion and depletion of the mitochondrial genome as observed in mitochondrial diseases.

Mitochondria produce Ͼ90% of cellular ATP through the process of electron transport and oxidative phosphorylation (1). The mitochondrial genome encodes 13 subunits of the electron transport chain and is essential for cellular respiration (2). Human mitochondrial DNA (mtDNA) 5 is a double-stranded, circular, 16,569-base pair genome with a purine-rich heavy (H)strand and a pyrimidine-rich light (L)-strand. Depletion of mtDNA or accumulation of point mutations and deletions in mtDNA directly leads to mitochondrial dysfunction. Inherited defects in mtDNA maintenance result in neurodegenerative and neuromuscular diseases and are a major driving factor in the aging process (3)(4)(5)(6).
Twinkle is essential for maintenance of mtDNA, because knockdown of TWNK (previously designated C10orf2) expression by RNA interference in cultured human cells reduces mtDNA copy number (27), and conditional knockout of TWNK in mice results in severe and rapid mtDNA depletion in targeted tissues and embryonic lethality (28). Transgenic mice expressing an autosomal dominant progressive external ophthalmoplegia allele of TWNK accumulate multiple mtDNA deletions during aging and develop progressive respiratory dysfunction (29) through a mechanism driven by stalled mtDNA replication (30). Disease alleles of the human TWNK gene are associated with improper mtDNA maintenance and cosegregate with severe heritable neuromuscular disorders, including infantile onset spinocerebellar ataxia, progressive external ophthalmoplegia with multiple mtDNA deletions, hepatocerebral mtDNA depletion syndrome, and Perrault syndrome (17,31). Genetic analysis of affected patients has identified at least 49 missense mutations in TWNK that conspicuously cluster in the linker region and C-terminal helicase domain of Twinkle (31). Biochemical analyses of many of these disease variant proteins indicate diminished DNA-unwinding activity and altered capacity to form stable hexameric conformations (32)(33)(34)(35), consistent with expectations from structural models predicting that the C-terminal domain is essential for hexamerization (36,37).
Knowledge of the molecular mechanisms governing mtDNA replication is crucial to understanding the origins of human disorders known collectively as "mitochondrial diseases." Although proteomic analysis of mitochondrial nucleoids and biochemical studies with purified proteins have identified core protein components of the mtDNA replication fork (38), specific information on the composition, stoichiometry, conformation, and dynamics of protein-protein and protein-DNA interactions remains less well-defined. We wanted to assess the mechanism of Twinkle helicase-binding and unwinding DNA in vitro. We were concerned that the varied stability of WT Twinkle helicase and certain disease variants would limit the utility of bulk biochemical assays because of ensemble averaging effects, so we selected a single-molecule technique. Atomic force microscopy (AFM) is a single-molecule technique that does not require labeling or chemical modifications of biomolecules including proteins and DNA (39 -42). AFM is particularly advantageous to examine Twinkle function on long DNA substrates while accounting for structural heterogeneity. Accordingly, we employed direct imaging by AFM in air and liquids to investigate the structure and dynamics of assembly, DNA binding, and unwinding activity of individual Twinkle hexamers.
Here, we found that Twinkle exists as a heterogeneous population of oligomeric states that self-assemble into hexamers capable of switching between open-and closed-ring configurations in the absence of DNA. Twinkle appears to load onto and to unload from circular gapped plasmid DNA in the open-ring configuration. Closed-ring conformers unwind several hundred base pairs of duplex DNA when ATP is present, and Twinkle disassociates from the DNA upon completion of unwinding. Displaced ssDNA products fold into higher-order structures at extended reaction times. The addition of mtSSB stimulates DNA unwinding by Twinkle by modulating the ways Twinkle loads onto DNA and by stabilizing the unwound ssDNA product. This study provides a new platform to examine Twinkle disease variants and the core mtDNA replication machinery. Our single-molecule approach will facilitate investigation of molecular mechanisms underlying deletion and depletion of the mitochondrial genome as seen in mitochondrial diseases.

AFM imaging in air reveals the heterogeneity of Twinkle oligomers
Hydrodynamic analysis indicates that, like other members of the superfamily 4 DNA helicases, purified recombinant human Twinkle exists principally as a hexamer in solutions with moderate ionic strength (34,43). Low-salt conditions promote disassociation of oligomeric structures and protein aggregation, but addition of divalent metal ions, nucleotides, and nonionic detergents facilitate assembly and stabilization of higher-order oligomers as large as heptamers (34,44). Examination of oligomeric structure by two-dimensional EM of adsorbed Twinkle complexes showed coexistence of hexamer, heptamer, and open-ring structures and suggested a dynamic interconversion between the three forms (44). A recent analysis of Twinkle in solution by small angle X-ray scattering also revealed a mixture of hexameric and heptameric conformations, and construction of a three-dimensional averaged molecular model from cryo-EM images of glutaraldehyde cross-linked Twinkle hexamers revealed two distinctly stacked layers with 6-fold rotational symmetry (37). Recombinant human Twinkle helicase was purified to Ͼ95% homogeneity, and its robust DNA helicase activity in vitro was validated using a forked oligonucleotide substrate (Fig. S1), as described previously (34) and under "Experimental procedures." We then utilized AFM to assess the volume distribution of individual Twinkle complexes in buffer containing 0.4 M NaCl without chemical cross-linking or prior selection for the hexameric subpopulation. The average AFM volume (552 Ϯ 347 nm 3 ) confirmed that the majority of Twinkle complexes occupied higher oligomeric states (volume Ͼ 450 nm 3 ) at this salt concentration in the absence of DNA (Fig. S2).
We also applied AFM imaging to assess conformational heterogeneity of Twinkle oligomers at 0.1 M NaCl to test conditions more favorable to DNA binding and enzymatic activity. When Twinkle helicase (100 nM) was deposited onto a mica surface in the presence of ATP (4.5 mM) and imaged using the AC AFM imaging mode in air, Twinkle oligomers were identified as spatially resolved particles in both AFM topographic and phase images, with phase images providing superior contrast (Fig. 1). Various oligomers with clearly identifiable subunits were visually classified as monomeric (26%), dimeric (19%), trimeric (20%), tetrameric (9%), pentameric (12%), and hexameric (14%) complexes (n ϭ 162). AFM volumes of complexes were measured (Fig. S3A) and compared with globular proteins with known molecular weights (Fig. S3B). AFM volumes of globular proteins increase linearly with their molecular weights (39,45), and this trend holds for the various Twinkle oligomers (Fig.  S3B), although AFM volumes increase more steeply because of Single-molecule dynamics of Twinkle DNA helicase extra space between Twinkle subunits ( Fig. 1) relative to globular reference proteins of matched molecular weights (45). To control for image selection bias, the AFM volumes of all Twinkle complexes were also plotted. The data revealed that the majority (82.0%) of Twinkle complexes were dimers, trimers, or tetramers (Ͻ450 nm 3 ), whereas a minority (18.0%) occupied higher oligomeric states such as pentamers, hexamers, or heptamers (volume Ͼ 450 nm 3 ) in the absence of DNA (Fig. S3C). Among hexameric Twinkle complexes (n ϭ 83), 35% existed in the closed-ring configuration (Fig. 1F), and 65% existed as open rings (Fig. 1G). Collectively, these results indicate that Twinkle exists in solution as a heterogeneous mixture of oligomers in the absence of DNA.

Twinkle oligomers self-assemble on DNA
Twinkle is able to bind and unwind a closed circular DNA template in vitro, indicating that neither a free DNA end nor a specialized helicase loader are required for the helicase to function (18). Structural studies indicate that ring helicases load onto such substrates either by assembling from monomers into a ring around the DNA or by cracking open a preassembled ring to allow entry of ssDNA (46). Twinkle may behave like other hexameric helicases and adopt a nonplanar conformation during translocation or convert to a hexameric right-handed helix when loading onto ssDNA (47,48). The observed distribution of the Twinkle oligomer population suggests that active Twinkle hexamers may assemble from lower-order oligomers such as dimers and trimers, as proposed for Escherichia coli DnaB or bacteriophage T7 gene product 4 (49,50). Preincubation of Twinkle with ssDNA oligonucleotides (108 nt) prior to AFM imaging (ratio ϭ 6 subunits/oligonucleotide) shifted the fraction of higher-order Twinkle oligomers (AFM volume Ͼ 450 nm 3 ) from 18.0% to 33.3% (Fig. S3C), suggesting the assembly of Twinkle hexamers is facilitated by the presence of ssDNA. To begin to decipher the mechanism by which Twinkle helicase loads onto DNA, we preincubated Twinkle with a circular DNA substrate containing site-specific 37-nt ssDNA gaps ( Fig. 2A). Two ssDNA gaps were generated by removing short ssDNA segments from one DNA strand between closely spaced nick sites (51), and the presence of the ssDNA gaps was validated by restriction digestion (see "Experimental procedures"). Individual protein-DNA complexes were examined by AFM imaging in air. Various lower order Twinkle oligomers were observed to bind the DNA in the absence of ATP (Fig. S4, A-D), although the specificity for single-stranded gapped regions cannot be determined because the circular substrate lacks a DNA end to be used as a reference point. Higher-order Twinkle oligomers also bind the DNA, and a subpopulation of Twinkle oligomers binds in an open-ring configuration with DNA clearly visible (Fig. S4E). Taken together, our observations support a model in which Twinkle monomers spontaneously assemble into higher-order structures that admit DNA through an opening in the helicase ring.

Twinkle binds ssDNA gaps with high specificity
Twinkle helicase binds single-stranded, double-stranded, and forked DNA substrates with high affinity (K d(DNA) ϭ 5-10 nM) as determined by electrophoretic mobility shift assay and fluorescence anisotropy (34,52). However, these methods cannot distinguish the binding position of Twinkle on DNA ends or on ssDNA regions of artificial replication forks, nor can they identify the specific oligomeric structure of Twinkle bound to DNA. To refine our understanding of Twinkle binding to more complex DNA structures, we also produced a linear DNA substrate containing a site-specific 37-nt ssDNA gap flanked by dsDNA arms (see "Experimental procedures"). AFM imaging shows a contour length of 663.8 Ϯ 19.4 nm for this substrate ( Fig. 2A). The ssDNA gap is located 470 bp (23%, ϳ160 nm assuming 0.34 nm/bp) from one blunt end of the 2030-bp DNA fragment. Specific binding of mtSSB confirms the position and single-stranded nature of this gapped region (Fig. 2, B and D). The DNA-binding specificity of Twinkle helicase was determined by AFM analysis of Twinkle bound to the linear gapped DNA substrate in the absence of ATP (Fig. 2C). The position distribution of bound Twinkle indicated preferential binding to a position 20.7 Ϯ 4.9% from the end, consistent with the position of the ssDNA gap (Fig. 2D). Although some nonspecific DNA binding was observed on this gapped substrate, bound Twinkle was distributed randomly along the nongapped linear DNA (Fig. 2D). Analysis of the fractional occupancies (39) of Twinkle along the linear gapped DNA establishes that Twinkle has a high binding specificity (S ϭ K SP /K NSP, ϳ2.5 ϫ 10 3 ) for the ssDNA gap. In comparison, mtSSB displays a higher binding specificity (S ϭ K SP /K NSP ϭ 1.45 ϫ 10 4 ) determined by the same method (53). Highly specific binding of Twinkle to the ssDNA gap suggested that this defined DNA substrate is ideally suited to assess DNA unwinding at the single-molecule level.

Single-molecule dynamics of Twinkle DNA helicase Twinkle unwinds DNA with 5-3 polarity in a closed-ring configuration
To evaluate DNA unwinding by Twinkle, we imaged the linear gapped DNA substrate (663.8 Ϯ 19.4 nm) following incubation with Twinkle in the presence of ATP. Inclusion of ATP caused the helicase to move away from the position of the gap, and structures resembling short segments of ssDNA were visible adjacent to Twinkle-DNA complexes in AFM images (Fig.  S5A). The complexity of these structures prevents accurate assessment of the length of the unwound ssDNA segments, so instead we determined the contour length of the remaining duplex DNA following unwinding reactions. The average length of the remaining linear dsDNA (486 Ϯ 84 nm) was 73.2 Ϯ12.8% of the total length of the original linear gapped DNA substrate (Fig. S5B), strongly suggesting that Twinkle unwinds the short arm of this linear gapped DNA substrate. Consistent with the strand-specific structure of the ssDNA gap ( Fig. 2A) and the obligate 5Ј-3Ј polarity of translocation for superfamily 4 DNA helicases, we infer that Twinkle binds the substrate at the position of the ssDNA gap and translocates along the intact 2030-nt DNA strand to unwind the shorter DNA arm. Loss of the displaced DNA strand and compaction of ssDNA segments after unwinding complicates the analysis of ssDNA structures on this linear DNA substrate. Accordingly, we shifted our focus to a circular gapped DNA sub-strate to assess the configurations and actions of Twinkle helicase.
Although superfamily 4 helicases may load and slide along DNA as open-ringed hexamers (46), the active form of the Twinkle is predicted to be a closed-ring hexameric configuration (17,36). AFM imaging shows Twinkle in the closed-ring configuration (n ϭ 101) bound to the circular gapped DNA substrate (Fig. 3A) in the presence of ATP. In some cases, AFM phase images clearly reveal DNA passing through the center channel of the Twinkle ring (Fig. 3A, right panels). Inclusion of ATP in the DNA-binding reactions permits direct visualization of individual Twinkle complexes unwinding DNA, because unwound ssDNA products exhibit an AFM height (200 Ϯ 104 pm, Gaussian peak) notably shorter than that of the dsDNA (371 Ϯ 116 pm, Gaussian peak) substrate (Fig. 3, B and C). Control reactions with supercoiled or nicked circular DNA produced no detectable DNA unwinding events.
Importantly, the capacity of AFM imaging to unambiguously identify the ssDNA product has permitted us to discern the functional configuration of Twinkle during DNA unwinding. The circumference of active Twinkle complexes in our AFM images was 79.7 Ϯ 12.8 nm (Gaussian peak Ϯ S.D.), which corresponds to a ring with a diameter of 25.4 Ϯ 4.1 nm (Fig. 3D). This value agrees with the previously reported 24.0-nm maximum particle dimension for Twinkle in solution that was cal-

Single-molecule dynamics of Twinkle DNA helicase
culated from the small angle X-ray scattering pattern (37). Taken together, AFM imaging directly shows that Twinkle unwinds DNA in a closed-ring hexameric conformation.

Quantifying DNA unwinding by Twinkle at the single-molecule level
Typical DNA substrates utilized to assess DNA unwinding by Twinkle in vitro contain forked structures with short duplex regions and are constructed from synthetic oligonucleotides or by annealing oligonucleotides (Յ60 nt) to single-stranded M13 DNA (19,34,52). Biochemical assay of Twinkle DNA helicase activity on such substrates bearing free 5Ј-tails (Fig. S1B) provided an average steady-state unwinding rate of 2.16 Ϯ 0.19 bp/min for Twinkle hexamers (34). Because this value is normalized to total protein, the true unwinding rate is likely much higher because not all Twinkle monomers are assembled into higher-order oligomers and bound to the DNA substrate in an active conformation. Accordingly, we sought to minimize ensemble averaging effects by quantifying the DNA-unwinding activity of Twinkle at the single-molecule level. Unwinding reactions were assembled as before with the circular gapped DNA substrate, and regions of unwound ssDNA were identified by their AFM height (as in Fig. 3, B and C). The length of ssDNA product generated in a 1-min incubation was 121 Ϯ 98 nm (Gaussian peak Ϯ S.D., n ϭ 221). Assuming that one nucleotide of ssDNA spans 0.5 nm (54), the average rate of individual unwinding events is 242 Ϯ 195 bp of dsDNA unwound per minute (Fig. 4, A and C). We note that the pSCW01 plasmid utilized to construct the DNA helicase substrate is a tandem duplication of 2030 bp ("Experimental procedures"), so circular gapped substrates derived from pSCW01 possess two independent unwinding sites adjacent to ϳ2-kb segments of duplex DNA available for unwinding. Helicase action on circular DNA molecules (n ϭ 76) displayed either single (80%) or double (20%) unwinding events (Fig. 4A). Calculated rates of unwinding were not affected by having one or two independent

Single-molecule dynamics of Twinkle DNA helicase
unwinding events on the same DNA molecule. The DNA unwinding rate determined for single molecules correlates well with the steady-state rate of ATP hydrolysis for Twinkle in the presence of calf thymus DNA (349 Ϯ 32 ATP hydrolyzed per min per Twinkle hexamer), as previously reported for bulk biochemical reactions (34). The similar rates are broadly consistent with the notion that processive translocation is closely coupled to ATP hydrolysis in our single-molecule Twinkle reactions, as seen earlier for the homologous T7gp4 helicase (55)(56)(57).
Longer incubation times were expected to cause proportionately longer regions of the DNA substrate to be unwound; however, the length of ssDNA products reached an apparent limit in longer reactions (Fig. 4, B and C). Regions of ssDNA generated in 20-min reactions displayed lengths of 30.8 Ϯ 26.8 nm, or 61.7 Ϯ 53.7 nt (Gaussian peak Ϯ S.D., n ϭ 101), which reduced the apparent unwinding rate to only 3.08 Ϯ 2.68 bp/min. Because this value is approaching the unwinding rate observed in ensemble reactions, we searched for barriers to unwinding at extended reaction times. Twinkle is both an unwinding and an annealing helicase, and transient stalling and reannealing of complementary ssDNA directly interferes with efficient translocation (58,59). Interestingly, longer incubation times (20 min) for Twinkle on the circular gapped DNA substrate also converted a significant fraction of DNA unwinding events (95% of n ϭ 53 circular DNA molecules) into extended DNA tracts with unexpectedly taller AFM heights (Fig. 4B). These structures exhibited an average AFM height of 664 Ϯ 177 pm (Fig.  3C) and an average length of 47.8 Ϯ 35.5 nm (n ϭ 37). We interpret these structures as extended segments of folded DNA, although the mechanism of forming these higher-order secondary structures in open circular DNA is unclear.
Furthermore, because intact Twinkle hexamers are not always visible at the ssDNA-dsDNA junctions in unwound circular DNA molecules (Fig. 4, A and B), we suspected that stalling of translocation promotes disassociation of protein-DNA complexes. To address this question, we examined AFM images to assess the fraction of DNA unwinding sites with Twinkle remaining bound at increasing times of incubation. Although Twinkle subunits were retained at the majority of unwinding events (92.8%, n ϭ 125) after 1 min of incubation, Twinkle binding at unwinding sites was reduced to 56.7, 43.0, and 23.4% after 5-(n ϭ 141), 10-(n ϭ 100), and 20-min (n ϭ 149) incubations, respectively (Fig. 4D). Retention of protein complexes smaller than hexamers also suggested partial disassembly of Twinkle hexamers over time. This decrease in Twinkle occupancy fits well to a function of single exponential decay, suggesting that disassociation/disassembly of Twinkle occurs as a distinct transition. Treating these events as one population (n ϭ 515) permits calculation of a disassociation rate constant of 0.138 Ϯ 0.029 events/min (Fig. 4D). Similarly, combining the rates of DNA unwinding (242 Ϯ 195 bp/min) and disassociation (0.138 Ϯ 0.029 events/min) established by our single-molecule method provides an estimation of translocation processivity (1750 Ϯ 1480 bp/disassociation) for isolated Twinkle helicase on the circular gapped DNA substrate.

Single-molecule dynamics of Twinkle DNA helicase mtSSB stimulates DNA unwinding by Twinkle
The ability of mtSSB to stimulate the unwinding activity of Twinkle helicase in ensemble reactions in vitro has been documented with a number of DNA substrates (19,20,23,25). The means and specificity of this stimulation are debated, and proposed mechanisms underlying stimulation by mtSSB include binding the displaced DNA strand to suppress reannealing of unwound ssDNA, direct stabilizing interaction with the helicase, and binding the DNA substrate in a manner that facilitates loading of Twinkle. To further understand the role of mtSSB protein in DNA unwinding by Twinkle at the single-molecule level, we repeated unwinding reactions with the circular gapped DNA substrate in the presence of mtSSB. AFM imaging in air allowed us to visualize extensive DNA-unwinding products as longer ssDNA regions both within the closed circular structure and as frayed ssDNA "tails" (Fig. 5A). Such ssDNA structures would be generated simultaneously by initiation of a single unwinding event at the position of the gap, followed by strand separation during translocation of the helicase. Images also revealed increased lengths of DNA-unwinding products for 1-min incubations containing 71 nM Twinkle with 35 nM mtSSB (Fig. 5A). The stimulatory effects of mtSSB were assessed by calculating rates of unwinding for separate reactions with a fixed concentration of Twinkle and varied concentrations of mtSSB. Whereas the unwinding rate in the absence of mtSSB was 242 Ϯ 195 bp/min, reactions containing 17.8, 35.0, 54.0, or 142 nM mtSSB tetramers exhibited unwinding rates of 287 Ϯ 226, 466 Ϯ 453, 1145 Ϯ 539, and 1265 Ϯ 541 bp/min (Gaussian peaks Ϯ S.D.), respectively (Fig. 5D). Stimulation of unwinding was generally proportional to added mtSSB, and maximal stimulation reached a plateau with apparent unwinding rates ϳ5-fold higher than observed for reactions lacking mtSSB. Control samples showed no observable unwinding of the gapped DNA substrate upon binding mtSSB in the absence of Twinkle helicase (Fig. 2, B and D), although passive unwinding of a few nucleotides may be inferred (see "Discussion").
Analysis of single-molecule unwinding reactions offers hints at the mechanisms by which mtSSB could stimulate DNA unwinding by Twinkle. ChIP-seq experiments have shown that mtSSB binds the displaced H-strand of mtDNA replication intermediates in vivo (26). Binding of SSB to freshly separated ssDNA strands would suppress reannealing and appear to stimulate the helicase in vitro, both in ensemble and single-molecule reactions. Previous work has demonstrated that both human and Drosophila mtSSB, as well as the heterologous T7 gp2.5 SSB, can stimulate Twinkle helicase activity on DNA substrates containing short duplex regions in vitro (23,58). The lack of species specificity suggests nonspecific stimulation caused by stabilization of ssDNA products (analogous to the inclusion of excess complementary oligonucleotide to trap a displaced DNA strand). Specific stimulation of helicase activity would be expected for binding factors that interact with the helicase to diminish pausing or disassociation. The time scale required to monitor helicase pausing is not readily accessible through AFM. However, the presence of mtSSB substantially increased the average length of DNA-unwinding products in our single-molecule images, which is consistent with the hypothesis that mtSSB reduces pausing or disassociation of the helicase. Twinkle preferentially binds ssDNA with high affinity (34), and the helicase loads onto DNA by assembling a hexameric structure around a covalently closed circular ssDNA substrate or by threading onto the single-stranded 5Ј-tail of a forked DNA substrate (18). Utilizing a plasmid-based DNA substrate with strand-specific, site-specific 37-nt ssDNA gaps, we have shown that Twinkle binds at the position of a gap and

Single-molecule dynamics of Twinkle DNA helicase
initiates unwinding of the duplex portion of the substrate (Figs. [2][3][4][5]. Although this circular gapped DNA substrate has two unwinding sites, the majority of DNA molecules imaged by AFM exhibit single unwinding events (Figs. 4A and 5A). Whether Twinkle initiates translocation on the covalently closed circular DNA strand or at a 5Ј-end of the gapped DNA strand in the presence of mtSSB is not clear. However, inspection of AFM images offers useful insight, because the obligate 5Ј-3Ј polarity of the helicase constrains double unwinding events to certain predictable structures. For example, unwinding initiated at both gaps by Twinkle translocating on the covalently closed strand would generate two expanding singlestranded regions with complementary free 3Ј-tails at the expense of two shrinking double-stranded regions on opposite sides of the DNA circle. Initiation of unwinding at both 5Ј-ends of the gapped DNA strand would generate the same structures with free 5Ј-tails. Nonetheless, these general configurations were not observed in AFM images of more than 500 DNA molecules. Similarly, one single-stranded region flanked by both 5Јand 3Ј-tails would be an indication of two Twinkle complexes initiating divergent unwinding at just one gap site, but this configuration also was not observed. In contrast, Twinkle hexamers initiating translocation on the covalently closed strand at one gap and on the 5Ј-end at the opposite gap would converge. Incomplete unwinding would generate two single-stranded regions between the gap sites and both 5Ј-and 3Ј-tails on the same half of the DNA substrate, and continued unwinding would result in complete removal of an open strand between the two gap sites. Structures matching these descriptions were readily apparent in AFM images (Fig. 5, B and C, respectively, and as depicted in Table S1). Also, the measured length of the dsDNA region in the unaffected half of the DNA substrate remained essentially constant for all double unwinding configurations (Table S1). This length (594 Ϯ 63 nm, n ϭ 174) was consistent with the distance between the two ssDNA gaps (2030 bp) on the substrate, lending further support to the existence of converging double unwinding events on this DNA substrate.

Real-time imaging of the dynamics of Twinkle assembly and DNA binding
Although AFM imaging in air creates remarkable static images of protein-DNA complexes, dynamic interactions such as conformational changes and DNA binding are more difficult to interpret. Therefore, we also utilized AFM imaging in liquids, which permits serial scanning of individual complexes at high resolution (Fig. 6). First, we monitored conformation dynamics of Twinkle applied to an aminopropyl silane (APS)-treated mica surface in the presence of ATP. Successive images display the apparent sequential recruitment of Twinkle subunits to form higher-order Twinkle oligomers (Fig. 6A, n ϭ 28). A separate series of images show closed-ring Twinkle hexamers appearing to switch spontaneously to open-ring configurations (Fig. 6B, n ϭ 27  observations). When circular gapped DNA was deposited onto APS-treated mica (Fig. 6C) and Twinkle was introduced directly to the solution, interaction of Twinkle with the DNA substrate could be observed over time. In a representative example (Fig. 6D), opening of the Twinkle ring (green arrow becomes open arrow) coincides with disappearance of the DNA molecule (red arrow) from the mica surface, which implies opening of the ring structure correlates with disassociation of Twinkle from the DNA substrate. In summary, time-lapse AFM imaging in liquids enables near real-time assessment of several dynamic behaviors of Twinkle helicase, such as ring assembly, conformational changes, and disassociation from DNA.

Discussion
AFM imaging in air has demonstrated that Twinkle exists in solution as a heterogeneous mixture of varying oligomeric states and that ssDNA facilitates self-assembly of higher-order Twinkle oligomers. AFM imaging in liquids suggests that Twinkle hexamers can be completed by the sequential recruitment of subunits and that Twinkle spontaneously switches between open-and closed-ring hexameric configurations. Collectively, our imaging results support a model in which Twinkle loads onto DNA in an open-ring configuration, unwinds DNA as a closed-ring structure, and unloads from DNA upon reopening of the ring structure. Quantifying DNA-unwinding activity in such a complex system required a single-molecule approach to avoid underestimation caused by ensemble averaging. Direct visualization of DNA-unwinding products by AFM imaging showed that Twinkle is capable of unwinding several hundred base pairs of duplex DNA in a single unwinding event. When we restricted reactions times to 1 min to minimize ssDNA product reannealing and enzyme disassociation, the average rate for individual Twinkle-unwinding events was ϳ240 bp/min. Dividing this rate by the calculated disassociation rate (0.138 min Ϫ1 ) allowed us to estimate the average translocation processivity for Twinkle to be 1750 bp/disassociation event, which is ϳ10% of the length of mtDNA (16569 bp). However, titration of mtSSB into Twinkle reactions greatly increased lengths of DNA-unwinding products, thereby raising the apparent unwinding rate by a factor of 5. Utilizing the disassociation rate determined for isolated Twinkle, we estimate a translocation processivity for Twinkle helicase in the presence of mtSSB to be 9150 Ϯ 4360 bp/disassociation event. Even though helicase products can exceed 1 kb, processivity may be undervalued because the calculation relies on rates derived from 1-min unwinding reactions. Any reduction of the Twinkle disassociation rate elicited by mtSSB would further enhance helicase processivity. Nevertheless, this first report of Twinkle helicase processivity suggests that over half the mitochondrial genome could be translocated during a single Twinkle binding event under optimal conditions in vitro. Similarly, Twinkle helicase does not function alone at the mtDNA replication fork in vivo. The natural abundance of mtSSB in cultured HeLa cells is sufficient to bind the entire mitochondrial genome (26,60), suggesting that helicase processivity may approach the unit length of the mitochondrial genome in vivo.
Several processes likely contribute to the ability of mtSSB to enhance both the apparent efficiency and processivity of DNA unwinding by Twinkle helicase. Binding the displaced DNA strand to suppress product reannealing has been shown to stimulate helicase activity in ensemble reactions in vitro. In one study, mtSSB enhanced Twinkle-unwinding activity more than 2-fold, although mtSSB sufficient to saturate both strands in the M13 ssDNA-based substrate was utilized (19). In the current study, mtSSB also stimulated unwinding of the circular gapped plasmid-based DNA substrate, and initial stimulation was roughly proportional to added mtSSB. Previously, we utilized AFM to examine the binding modes of E. coli SSB and human mtSSB to ssDNA. Unlike E. coli SSB, mtSSB does not form nucleoprotein filaments that are a hallmark of cooperative DNA binding but instead binds noncooperatively to extended ssDNA structures at physiological concentrations of salt (53). The current study utilizes similar solution conditions, and nucleoprotein filaments were not observed in any of our AFM images. This absence is noteworthy and implies that mtSSB binds noncooperatively to freshly generated ssDNA when Twinkle helicase unwinds the circular gapped DNA substrate. We observed maximal stimulation of Twinkle helicase activity at a mtSSB concentration far below the quantity needed to saturate unwound ssDNA products (Fig. 5D), which is consistent with stabilization of the unwound DNA product by noncooperative binding. Protein binding partners that reduce pausing of the helicase would serve to stimulate Twinkle activity. As mentioned above, including mtSSB in unwinding reactions increased the length of ssDNA products in our single molecule images, suggesting that mtSSB may facilitate an increased rate of translocation or a decreased rate of disassociation for Twinkle. However, quantifying the relative contributions of these possible effects is not feasible, because the inability to easily differentiate mtSSB tetramers (61.3 kDa) and Twinkle monomers (73.5 kDa) by AFM prevents determination of the Twinkle disassociation rate in the presence of mtSSB.
An additional important scenario is the possibility that mtSSB contacts Twinkle or modifies the DNA substrate in a manner that helps to load the enzyme onto the gapped substrate in an active conformation. Knowledge of Twinkle's substrate specificity informs this discussion. Twinkle prefers forked DNA substrates with free 5Ј-tails, and the enzyme cannot efficiently unwind fully hybridized DNA substrates or duplex substrates with blunt or recessed 5Ј-ends (19,59). However, Twinkle actively unwinds both 5Ј-tailed and 3Ј-tailed oligonucleotides hybridized to M13 ssDNA in vitro, albeit with different efficiencies (18,58). This paradoxical result stems from Twinkle utilizing two modes for loading onto DNA substrates. Intact Twinkle hexamers are able to thread directly onto the free end of the 5Ј-tailed substrate, whereas hexamers must assemble around the covalently closed circular strand of the 3Ј-tailed substrate. The reduced efficiency of unwinding the 3Ј-tailed substrate can be explained by the additional steps required to assemble the active enzyme (18). The circular gapped DNA substrate used in the current study has one covalently closed circular strand, and the complementary open strand bears two 37-nt gaps in symmetrical positions on opposite sides of the plasmid (Fig. 2). Twinkle complexes bind and initiate unwinding from one or both of the ssDNA gaps. As described above, the pattern of utilizing the two gaps changes when mtSSB is included in unwinding reactions. In the absence of mtSSB, 80% of circular gapped DNA substrate molecules unwound by Twinkle were initiated at a single gap site (Fig. 4A).
In the presence of mtSSB, this fraction was reduced to 44%, and the remaining molecules appeared to utilize two initiation sites in double unwinding events ( Fig. 5 and Table S1). The ability of mtSSB to stimulate the apparent rate of translocation (Fig. 5D) is inadequate to explain this significant increase in double unwinding structures when mtSSB is included in helicase reactions. The critical point is that converging double unwinding events require the Twinkle complexes to employ different modes for loading onto each of the two ssDNA gaps. It is tempt-

Single-molecule dynamics of Twinkle DNA helicase
ing to speculate that the presence of mtSSB at the singlestranded gaps modulates assembly of Twinkle around the covalently closed circular DNA strand. Alternatively, mtSSB may partially fray the fully duplex 5Ј-ends of the open DNA strands to facilitate loading of Twinkle over a freshly exposed free 5Ј-tail. Either or both options could enhance the opportunity to form the double unwinding events seen on circular DNA substrates in our AFM images. In summary, mtSSB appears to stimulate Twinkle helicase unwinding the circular gapped DNA substrate by a combination of effects, including stabilization of the unwound ssDNA product and changing the way Twinkle loads onto DNA.
The bacteriophage T7 DNA replisome is an extensively studied model system for strand displacement DNA replication (61). Rapid quench kinetics measurements indicate that the T7 DNA helicase unwinds duplex DNA slower than the rate of DNA synthesis catalyzed by the DNA polymerase, making DNA synthesis the driving force behind progression of the replication fork (62). Although the T7 DNA polymerase and DNA helicase work on separate DNA strands, an overall effect is that the enzymes remain functionally and physically coupled at the DNA replication fork (48,63). Mitochondrial DNA replication also follows a strand displacement model in which Twinkle helicase translocates on the displaced parental mtDNA Hstrand and DNA Pol ␥ utilizes the parental L-strand as the DNA template for synthesis of the nascent H-strand. Pre-steadystate kinetics measurements show that the reconstituted Pol ␥ holoenzyme can synthesize DNA at ϳ45 nt/s under optimal conditions in vitro (16), which predicts that one strand of the 16,569-bp mitochondrial genome could be replicated in just over 6 min. Our AFM study suggests the maximal rate of DNA unwinding by Twinkle in the presence of mtSSB is ϳ21 bp/s. Taken together, these values suggest that WT Pol ␥ and Twinkle also remain coupled at the mtDNA replication fork. Missense disease variants of TWNK and POLG disrupt the function and stability of the enzymes, which interferes with coordinated action of the enzymes and likely promotes the formation of mtDNA deletions during replication and repair of mtDNA in vivo (35,64,65). For example, persistent pausing of the helicase, such as at DNA secondary structures, would stall the mtDNA replication fork and promote disassociation of Twinkle from the unwinding site. Similarly, stalling of the polymerase caused by enzyme dysfunction or miscoding lesions in the DNA template could uncouple the replisome and prompt Twinkle to advance ahead of the polymerase and generate unusually long regions of ssDNA. We have indirect evidence for excess ssDNA in the mitochondrial genome of Saccharomyces cerevisiae strains bearing replication-defective, disease-associated missense mutations in mip1, the yeast homolog of human POLG. When we exposed cells to the exogenous base-alkylating agent methyl methanesulfonate, mutant strains exhibited a 30-fold increase in mutagenesis of their mtDNA relative to similarly treated WT strains. Because base excision repair cannot remove methyl methanesulfonate-damaged bases from ssDNA, the increased vulnerability to mutagenesis suggested irreparable damage to mtDNA that occurred during DNA replication (66).
In summary, we utilized AFM imaging in air and real-time imaging in liquids to observe and report the configurations and dynamic steps of Twinkle oligomer assembly, DNA binding, and unwinding. Our findings address fundamental questions regarding the ways Twinkle loads onto DNA and the mechanisms underlying stimulation of unwinding by mtSSB. This work provides a new framework to study the mtDNA replication machinery, the molecular defects of protein variants identified in heritable mitochondrial genome maintenance disorders, and the impact of mtDNA lesions on mtDNA replication.

DNA substrates
Synthetic oligodeoxyribonucleotides were obtained from Integrated DNA Technologies. Linear dsDNA containing a 37-nt ssDNA gap was prepared by treating pSCW01 plasmid DNA (duplication of 2030 bp) provided by Peggy Hsieh (NIDDK, National Institutes of Health) (51) with Nt.BstNBI to introduce two clusters of four closely spaced nicks in one strand, incubating with complementary oligonucleotides (oligonucleotides:plasmid ratio ϭ 10:1) at 68°C for 30 min followed by slow cooling to room temperature, and removing excess unbound oligonucleotides and short duplex DNA using 100K MW Amicon Ultra filtration (51,70). Circular gapped DNA was linearized by digestion with ScaI in buffer 3.1 (NEB), which positioned the 37-nt ssDNA gap 470 bp (23%) from one end of the blunt-ended 2030-bp linear DNA fragment. Diagnostic restriction digestion of the gapped region indicated typical DNA gapping efficiencies of 85-95% (53,71). Linear dsDNA lacking a ssDNA gap was prepared by directly linearizing pSCW01 plasmid DNA with ScaI restriction endonuclease. A forked DNA helicase substrate was assembled by hybridizing two 40-nt synthetic oligonucleotides (5Ј-ATGCTAGCTTGGCTG-TGACTTTAAACCTGTCGTGCCAGCT-3Ј and 5Ј-AGCTG-GCACGACAGGTTTCCCGACTGGAAAGCGGGCAGTG-3Ј) to form a structure with 18 complementary base pairs and two 22-nt single-stranded tails. This substrate was utilized in ensemble biochemical reactions to assess and validate the DNA Single-molecule dynamics of Twinkle DNA helicase helicase activity of purified Twinkle in vitro (Fig. S1B), as described previously (34).

AFM sample preparation
To assess the oligomeric states of Twinkle helicase, protein samples were diluted in helicase buffer containing 20 mM HEPES (pH 7.5), 100 mM NaCl, 7.5 mM MgCl 2 , 7% glycerol, 0.2% Nonidet P-40, and 4.5 mM ATP to a final concentration of 100 nM hexamers. The samples were immediately deposited onto freshly cleaved mica surfaces at room temperature, followed by washing with deionized water and drying with streams of nitrogen gas prior to AFM imaging. Twinkle complexes in helicase buffer containing 0.4 M NaCl were also examined by AFM, as indicated. To study the effects of ssDNA on the oligomeric state of Twinkle, 100 nM Twinkle hexamers were premixed with 100 nM single-stranded 108-nt oligonucleotides in helicase buffer and incubated at room temperature for 10 min prior to deposition for AFM imaging. The DNA binding specificity of mtSSB was determined by incubating 73.5 nM mtSSB (tetramers) with 3.6 nM linear gapped DNA. The DNA binding specificity of Twinkle was determined by incubating 71.2 nM Twinkle hexamers with 3.6 nM linear gapped DNA in helicase buffer without ATP. DNA unwinding was examined in reactions containing helicase buffer, 3.6 nM circular gapped DNA, 71.2 nM Twinkle helicase (hexamers), and 0 -142 nM mtSSB (tetramers, as indicated). The samples were incubated at room temperature for 1, 5, 10, or 20 min (as indicated), further diluted in helicase buffer, and deposited onto a mica surface for AFM imaging as mentioned before. For real-time imaging of Twinkle dynamics in liquids, 0.7 nM circular gapped DNA was deposited onto a freshly prepared APS-treated mica surface, incubated for 15 min, and washed with helicase buffer. Twinkle hexamers (71.2 nM) were incubated in helicase buffer for 5 min at room temperature and added to the DNA sample on the APS-mica surface with a pipette tip while scanning.

AFM imaging and image analysis
AFM imaging in air was carried out using the AC mode on a MFP-3D-Bio AFM (Asylum Research, Oxford Instruments) with Pointprobe PPP-FMR probes (Nanosensors, spring constants at ϳ2.8 N/m). All images were captured at a scan size of 1-3 m ϫ 1-3 m, a scan rate of 1-2 Hz, and a resolution of 512 ϫ 512 pixels. AFM imaging in liquids was accomplished using a BL-AC40TS (BioLever Mini) probe (Asylum Research), and images were captured at a scan rate of 2-4 Hz with a resolution of 512 ϫ 512 pixels. Lengths of dsDNA and ssDNA segments were measured with MFP-3D software (Asylum Research). AFM volumes of Twinkle oligomers in the absence and presence of DNA oligonucleotides were measured with Gwyddion software. The volume data are reported as means Ϯ S.D.

Data availability
AFM data are available upon request to Parminder Kaur, North Carolina State University (pkaur3@ncsu.edu) or William Copeland, NIEHS, National Institutes of Health (copelan1@niehs.nih.gov).