E. coli gyrase fails to negatively supercoil diaminopurine-substituted DNA

Abstract

Type II topoisomerases modify DNA supercoiling, and crystal structures suggest that they sharply bend DNA in the process. Bacterial gyrases are a class of type II topoisomerases which can introduce negative supercoiling by creating a wrap of DNA before strand passage. Isoforms of these essential enzymes were compared to reveal whether they can bend or wrap artificially stiffened DNA. E. coli gyrase and human topoisomerase IIα were challenged with normal DNA or stiffer DNA produced by PCR reactions in which diaminopurine (DAP) replaced adenine deoxyribonucleotidetriphosphates. On single DNA molecules twisted with magnetic tweezers to create plectonemes, the rates or pauses during relaxation of positive supercoils in DAP-substituted versus normal DNA were distinct for both enzymes. Gyrase struggled to bend or perhaps open a gap in DAP-substituted DNA, and segments of wider DAP DNA may have fit poorly into the N-gate of the human topoisomerase IIα. Pauses during processive activity on both types of DNA exhibited ATP-dependence consistent with two pathways leading to the strand-passage competent state with a bent gate segment and a transfer segment trapped by an ATP-loaded and latched N-gate. However, E. coli DNA gyrase essentially failed to negatively supercoil 35% stiffer DAP DNA.

Graphical abstract

Introduction

Type II topoisomerases are essential ATP dependent enzymes that disentangle newly replicated chromosomes during cell division and modify the supercoiling of DNA [1]. They are heterodimers or tetramers with two-fold symmetric structures forming two chambers [2]. Type II topoisomerases change the linking number of DNA by catalyzing the passage of one DNA duplex through another (Fig. S1a). During catalysis, a DNA segment first binds to the catalytic “DNA-gate” of the enzyme composed of divalent metal-binding, topoisomerase-primase domains (TOPRIM) ^††, and winged-helix domains (WHD) where the catalytic tyrosines are located. DNA co-crystals with several type II topoisomerases revealed intercalating β hairpins from the WHDs that bend gate-segments (G-segments) sharply at two points [3–5]. Recent AFM and FRET data also indicate large overall bending angles [6]. Along opposite stands of the bent double helix, catalytic tyrosines of the enzyme form transient covalent bonds with 5′-phophoryl groups to produce temporary DNA nicks four base pairs apart. In what may be loosely coupled states in alternative pathways [7], a “transfer” DNA segment (T-segment) enters a chamber and binding of two ATP molecules latches the “N-gate” to trap the T-segment and form the critical intermediate of the enzymatic reaction [8]. After hydrolysis of ATP, opposite TOPRIM and WHD domains separate to open a gap in the nicked DNA and provide a path for the T-segment into the exit chamber. The DNA gate closes and the G-segment of DNA is re-ligated after strand passage. This conformational change may trigger the release of the transfer segment from the exit chamber [3].

Although topoisomerases are generally non-specific enzymes, kinetic and thermodynamic data, based on experiments in which re-ligation is inhibited, show that type II topoisomerases preferentially cleave certain DNA sequences [9]. Inherently bent or flexible sequences may be preferred binding sites, as suggested by electron micrographs showing type II topoisomerases bound at the apices of highly curved supercoiled plasmids [10]. Furthermore DNA sequences with high affinity identified by systematic evolution of ligands by exponential enrichment (SELEX), and kinetic analyses of the association of topoisomerase II enzymes with ten-base pair DNA fragments indicated a preference for AT-rich segments [11]. Since cleavage follows DNA binding [9], DNA bending in the initial steps of the enzymatic cycle may direct cleavage specificity. Bacterial gyrases, which can wrap about 40 bp DNA nearly 180 degrees around GyrA C-terminal domains (CTD) to catalyze negative supercoiling (Fig. S1b) [12], do prefer certain sequences [13, 14]. This preference may also depend on DNA stiffness, since gyrases operating on single DNA molecules have been observed to contract DNA molecules by 30–35 nm without creating twist just prior to and lasting throughout enzymatic activity [7].

Proteins that bend DNA are known to utilize positively charged domains to neutralize electrostatic repulsion [12] or intercalate in between base pairs to denature and introduce kinks [15]. DNA stiffness arises from base-pair stacking, electrostatic interactions between negatively charged phosphates, as well as hydrogen bonding that opposes duplex denaturation [16]. DNA segments separated by 50 nm along the contour, the persistence length of normal DNA, freely shift orientations as the molecule bends and form loops like any long narrow fiber. However, DNA segments closer than a persistence length are held apart unless ligand binding or extremely high concentrations of salt can overcome the considerable energy necessary to bend the intervening DNA [17].

Not surprisingly, histone octamers, which bend 147 base pair DNA segments into a 5 nm radius turn, have higher affinity for DNA with shorter persistence lengths [18]. Virstedt and colleagues created short DNA fragments with either inosine or diaminopurine (DAP) substitution for adenine to alter the persistence length but not the sequence of the DNA fragments. Substitutions that increased the hydrogen bonding and exocyclic amines in the minor groove of DNA significantly increased persistence length and decreased the binding affinity of histone octamers.

Although DAP is similar to adenine, it has an extra exocyclic amino group on the 2 position (Fig. 1a). This extra amino group is available for H-bonding with the electronegative oxygen atom on thymine to form a third H-bond between DAP and thymine. The melting temperature and the width of the minor groove at DAP:T base pairs are very similar to those of a G:C base pair [19], and NMR has shown that DAP-substituted, 10 bp-long DNA maintains the B-form [20]. However, the extra exocyclic amino groups of DAP:T base pairs are located in the DNA minor groove which may facilitate transitions to A- or Z-form helices [21]. The polymerase chain reaction (PCR) and many restriction enzyme digestions operate normally on DAP DNA (unpublished results) [22, 23]. Therefore, it is a convenient tool with which to investigate how the mechanical properties of DNA affect enzymatic activity.

Figure 1.

Stiffness of DAP DNA. (a) A:T and DAP:T base pairs are shown. The hollow blue arrow indicates the additional exocyclic amino group in the minor groove of a DAP:T base pair. (b) Single DAP-substituted (blue) and normal (red) DNA molecules were stretched using magnetic tweezers (inset schematic). The force vs. DNA extension data were fit using worm-like chain models revealing stiffer DAP DNA, persistence length = 70 ± 14 nm (n = 40) versus 52 ± 10 nm (n = 30), for normal DNA. (c) Single DNA molecules under 0.6 pN of tension were also twisted using magnetic tweezers (inset schematic). The slopes of linear fits to the DNA extension versus twist curves indicate larger supercoils in DAP DNA.

The aggregate evidence for DNA bending during the enzymatic cycle of type II topoisomerases suggested that increased DNA stiffness might affect the rate of strand passage. In addition there are substantial differences in the primary structures of different isoforms that may correlate with functional differences. For instance GyrB subunits from certain organisms such as Escherichia coli and Salmonella enterica serovar Typhimurium contain a 170 residue amino acid sequence that affects G-segment binding and the co-ordination of ATP hydrolysis and strand passage [24]. This domain does not have extremely positive charge that would enhance DNA binding, but rather, is thought to buttress the TOPRIM domain against the GyrA subunit to maintain the configuration of the DNA gate. This domain is not found in the human topoisomerase IIα enzyme, which might therefore have more difficulty processing stiffer DNA. Gyrases also possess C-terminal GyrA domains around which DNA wraps to create a tight positive supercoil. Strand passage reverses the overlap of the crossed strands and the DNA becomes more negatively supercoiled. In the same way that histone octamers struggle to bend stiffer DAP DNA, gyrases would not be expected to readily bend and negatively supercoil DAP DNA. To assay the effect of DNA stiffness on activities of these topoisomerases, single DNA molecules were gently stretched with magnetic tweezers to measure stiffness and monitor supercoiling.

Results

DAP substituted DNA is stiffer than normal DNA

As previously observed, DAP-subsitituted DNA may be similar to an A-form DNA double helix [18]. AFM measurements of the lengths of 1020 bp linear fragments produced by PCR measured 320 ± 30 nm (34 molecules), which indicates a axial base pair rise of only 0.31 nm. This is similar to the axial rise of A-form DNA which is slightly broader than B-form DNA, has narrower major and wider minor grooves, and has a longer helical repeat (11 bp) [25]. Normal and DAP-substituted DNA tethers were stretched using magnetic tweezers to measure persistence lengths by fitting the wormlike chain model to the force versus extension data [26], $F=\frac{k_{B}T}{P}\left[\frac{1}{4{(1-z/L)}^2}-\frac{1}{4}+\frac{z}{L}\right]$. In this equation, k_B represents Boltzmann’s constant, T represents temperature, P represents persistence length, L represents contour length, and z represents the end-to-end extension of the DNA. The model fit the data well and produced an average persistence length estimate for normal DNA of 52 ± 10 nm (n = 30), which is similar to previous measurements [27]. The model also fit the data for DAP DNA quite well, but indicated a 35% larger persistence length estimate of 70 ± 14 nm (n = 40) (Fig 1b). Long, DAP DNA was stiffer than normal DNA as expected for additional exocyclic amino groups in the minor groove, in agreement with earlier data for short, DAP DNA fragments [18].

Normal and DAP DNA also formed distinct plectonemes. A 3.2 kbp right-handed helical molecule has about 307 helical turns. As the molecule was overwound, torque initially accumulated within the molecule without changing the overall extension. Upon reaching a critical level of torque, plectonemes began to form, and the end-to-end distance decreased linearly as a function of added turns of the magnetic bead [28]. The slope (length of DNA/plectonemic wrap) of the DNA extension vs. turn curve in the plectonemic region was larger for DAP-substituted DNA (36.1 ± 2.4 nm/turn) than that of normal DNA (33.8 ± 3.7 nm/turn) (Fig. 1c). Thus, the segment of DNA required to complete a plectonemic wrap is longer for DAP than for normal DNA as would be expected for a stiffer molecule.

However, the buckling transition, the superhelical density (σ) at which the DNA molecules starts supercoiling, was similar for normal and DAP DNA, σ = 0.031 ± 0.004 and 0.032 ± 0.005 respectively. This value of σ reflects the torque that builds up in a molecule before additional twist is converted into writhe. Once writhe (plectonemes) form, the torque remains constant even as twist increases further [29]. Therefore, DAP substitution also significantly increased the torsional stiffness of the molecule as might have been expected from predictions based on molecular dynamics data [19]. Additional hydrogen bonds and stronger dipole coupling of DAP:T base pairs likely restricts the range of twist available to base pairs [22].

E. coli gyrase had higher affinity for normal DNA

Since type II topoisomerases in crystal structures appear to bend the gate segment of DNA [3–5], stiff DAP DNA was expected to be a poor substrate for binding. Therefore an electrophoretic mobility shift assay (EMSA) was used to compare the binding affinity of E. coli gyrase and human topoisomerase IIα for normal and DAP DNA. Only gyrase satisfactorily electrophoresed, and as the concentration of gyrase was increased, DNA shifted from the faster migrating, unbound (lower) band to the slower migrating, gyrase-bound position (higher band) (Fig. S3). The percentage of fluorophore-labeled DNA in the protein-complexed band was quantified by densitometry (Fig. 2). With respect to normal DNA, at gyrase concentrations higher than 184 nM the amount of gyrase-bound DAP DNA was reduced by about 20%. This corroborates a study in which the binding of IHF, which introduces a sharp bend in DNA, decreased substantially on DAP-substituted DNA [30]. It suggests that a significant conformational change associated with gyrase binding, either bending of the gate segment, or wrapping of DNA around C-terminal domains of GyrA subunits to produce the “omega” (contracted) and “alpha” (positively supercoiled) forms [7].

Figure 2.

Decreased E. coli gyrase binding to DAP DNA. Equal amounts of normal or DAP DNA were incubated with increasing E. coli gyrase concentrations and electrophoresed through non-denaturing gels (Fig. S2). Average percentages of binding to gyrase at different concentrations were determined by fluorometric densitometry.

E. coli gyrase wraps normal DNA more easily

E. coli gyrase can wrap normal DNA even in the absence of ATP at low tension [12]. Without ATP, strand passage does not proceed, DNA wrapping is reversible, and a wrapping/unwrapping equilibrium is established (Fig. 3a) [31]. For both DAP and normal DNA extended by 0.4 pN of tension in a magnetic tweezer, this equilibrium was detected as tether lengths that flickered between two values (Fig. 3b & c). In the absence of gyrase, no transitions were observed (data not shown). After introducing the gyrase into the chamber, data were recorded for several hours in sealed flow chambers at room temperature. Transitions were observed for all of the normal but only half of the DAP DNA molecules.

Figure 3.

Decreased gyrase wrapping of DAP DNA. (a) A schematic of the equilibrium between wrapped and unwrapped states of gyrase with no ATP present shows that the extension of the DNA tether switches between two levels. (b and c) Raw data (dots) and 10 second moving averages (solid lines) reveal changes in extension versus time for normal (b) and DAP (c) DNA. Histograms of the average extension (normalized to the longer unwrapped tether length) of normal (b) and DAP (c) DNA incubated with E. coli gyrase were fitted with two Gaussians representing the unwrapped (upper peak) and wrapped state (lower peak) respectively.

The histograms of DNA extension in Figure 3b and c show two distinct peaks representing unwrapped (longer) and wrapped (shorter) DNA separated by about 35 nm. This is consistent with the contraction reported for the omega and alpha conformations [7] and the length of DNA wrapped by the C-terminal domain of gyrase reported in other studies [12]. The proportion of the “wrapped” peak to the total areas under both peaks revealed that the probability for gyrase to wrap DAP-substituted DNA is distinctly lower than that of normal DNA. Employing a Boltzmann distribution, the free energy required to wrap DNA was estimated from p_wrapping = A_wrapped/A_total = e^−ΔG/kT. 5.9 kJ/mole were required for gyrase to wrap DAP-substituted DNA compared to 4.0 kJ/mol for normal DNA. This nearly 2 kJ/mole (0.8 k_BT) free energy difference is of the same order of magnitude found for histone octamer-DAP versus normal DNA binding, which also involves wrapping [18].

The lifetimes of the wrapped states for both normal and DAP DNA were tabulated and a cumulative plot of the fraction exceeding any given time were plotted (Figure 4). In some instances the wrapped state was remarkable stable, but DAP-substitution shortened the lifetime of the wrapped states, which could be either omega or alpha conformations. These distributions were best fitted with double exponential decays and the fitting parameters associated with DAP-substitution reflect relatively faster decay than for normal DNA (relatively higher fast component amplitudes and longer decay constants, supplementary Table S1).

Figure 4.

The lifetimes of the wrapped states of gyrase for DAP (red circles) and normal (blue crosses) DNA under 0.4 pN of tension were measured and the fraction greater than or equal to different time intervals was plotted. The number of measured pauses was 154 and 88 for normal and DAP DNA respectively. DAP substitution clearly shortened the average lifetime of wrapping. See Table S1 for exponential fitting data.

Gyrase-catalyzed relaxation of positively supercoiled normal and DAP DNA

Like other type II topoisomerases [3–5], gyrase is believed to bend the G-segment DNA significantly [32]. If bending were to be required in each enzymatic cycle, gyrase might catalyze supercoil relaxation more slowly on DAP-substituted versus normal DNA. Therefore, gyrase-catalyzed relaxation of positive supercoils in normal or DAP DNA under relatively high tension was measured to determine whether DNA stiffness affects the rate of strand passage (Figure 5). Note that the wrapping activity of gyrase is not required to lower the linking number of positively supercoiled DNA and is unlikely to occur under 0.6 pN of tension (see Introduction of negative supercoils in normal DNA by E. coli and Salmonella gyrases). A few minutes after introducing gyrase and ATP, 3.2 kbp DNA under 0.6 pN tension was twisted and held at 30 turns, equivalent to 10% supercoiling density. The formation of plectonemes reduced the overall extension of the DNA tether, but after random delays the DNA suddenly lengthened progressively or in multiple steps due to gyrase-catalyzed relaxation of supercoiling (Figure 6).

Figure 5.

A schematic illustration of enzyme-catalyzed supercoil relaxation monitored using magnetic tweezers. The drawing is not to scale and the following description proceeds from left to right. When magnets are rotated above a paramagnetic bead tethered to a glass slide to twist the DNA, a plectoneme forms and shortens the extension. Type II topoisomerases bind to a crossover and using ATP hydrolysis reverse it by catalyzing strand passage. Each reversal reduces the linking number of the DNA tether by two, and increases the extension which is monitored using three-dimensional tracking of the microsphere.

Figure 6.

Gyrase relaxes DAP DNA more slowly. (a) Three offset, representative records of gyrase-catalyzed relaxations of normal (left) and DAP DNA (right) under 0.6 pN tension and with 1 mM ATP. The raw data (dots) and 0.5 s moving averages (solid lines) show variable delays before processive relaxation of supercoils that restored the tether to an extended state. Occasional pauses in the processive increases of tether length are evident (circled interval). (b) By ignoring pauses and converting tether length changes to twist of DNA, gyrase was found to more slowly relax DAP (crosses) than normal DNA (open circles) at several ATP concentrations. Fits of a Michaelis-Menten model are shown for normal (solid, K_m = 0.16 mM, V_max = 2.9 turns/s) and DAP (dashed, K_m = 0.12 mM, V_max = 1.93 turns/s) DNA. (c) Average pauses during otherwise processive relaxation were similar for gyrase acting on both types of DNA at all ATP concentrations (see also Figure S3).

Three representative extension versus time curves from different normal or DAP DNA molecules at 1 mM ATP concentration are shown in Fig. 6a. These curves are characterized by bursts of strand passage events that cause length increases interrupted by pauses. Using tension-dependent calibrations established from extension vs. turns curves for each molecule (Fig. S4), length differences (nm) were converted into supercoiling changes (turns), and the rates of processive supercoil relaxation were determined for normal and DAP DNA at several ATP concentrations (Fig. 6b). E. coli gyrase relaxed supercoils in normal DNA at a maximal rates commensurate with previous measurements at 0.6 pN and 1 mM ATP concentration [33]. However, at ATP concentrations near a saturating level, the overall rate was lower for DAP-substituted DNA.

The activity vs. ATP concentration data was fitted using the Michaelis-Menten model, $V_0=\frac{V_{max}[\mathit{ATP}]}{K_M+[\mathit{ATP}]}$. For normal DNA, K_M was 0.16 mM and V_max equaled 2.9 turns/s, while the DAP substrate retarded gyrase activity by 30% (K_M = 0.12 mM, V_max = 1.93 turns/s). This mild change in the rate of relaxation of positive supercoils and the fact DAP-substitution introduces exocyclic amines into the minor groove suggest that the catalytic tyrosines of topoisomerase can effectively engage the major groove to carry out cleavage and religation. One explanation of these data might be that gyrase relaxes supercoils in DAP DNA more slowly, because a G-segment of DAP DNA is more difficult to bend. However, gyrase not only sharply bends the gate segment but also wraps adjacent segments around the C-terminal domains of GyrA subunits in the contracted “omega” intermediates which precedes strand passage [7]. The CTDs include a distinctive sequence of seven basic amino acids, the GyrA box, that orient the flanking segment for positive writhe and enable negative supercoiling by gyrase [34]. Deletion of the GyrA box or the entire CTD abolishes the capacity for negative supercoiling and slows the strand passage reaction by a little less than two-fold [35]. The fact that a similar attenuation of activity results with wild-type gyrase operating on DAP DNA suggests that decreased CTD wrapping might diminish the “grip” needed by gyrase to bend and then part the gate segment for strand passage.

Between consecutive bursts of supercoil relaxation there were delays. The distribution of pauses can reveal characteristic rate-limiting steps of the enzymatic cycle, and therefore pauses in gyrases activity were measured on normal and DAP DNA. In order to exclude pauses due to dissociation of the enzyme from the DNA substrates, the average time lag between the introduction of plectonemes and the start of supercoil relaxation by gyrase, 100 s, was considered to be an upper limit, and all longer pauses were excluded [36]. In addition, the magnetic tweezer frame rate was too slow to resolve pauses shorter than 2 s.

During the remaining pauses, gyrase was assumed to have remained associated with the DNA tether. Mean pause times were similar for both types of DNA as shown in Figure 6c, and normalized inverse cumulative pause distributions, plots of the fraction of pauses exceeding various times, for DAP DNA were very similar to those for normal DNA. Double exponential decays fit the data quite well (Fig. S5) indicating two rate-limiting pathways or paused states within the strand passage cycle. This might correspond to ATP-dependent and independent routes to reach the strand-passage competent state with a bent gate segment and a transfer segment trapped by an ATP-loaded and latched N-gate as proposed by Basu et al. [7]. Such a bifurcated pathway would be expected to shunt toward a faster, ATP-dependent pathway as the ATP concentration increases. Indeed as the ATP concentration increased, the mean pause time decreased and the amplitude of the fast decay component in the fitting of the distribution increased (Table S2, Figure S6). Unexpectedly, gyrase paused longer with 1.5 mM ATP. This may indicate that high [ATP] can prematurely close the N-gate before a T-segment has entered. Hypothetically, gyrase molecules might delay relaxation activity by shifting between relaxation modes with and without wrapping [33], but wrapping was not observed with 0.6 pN of tension, and relaxed DNA molecules under 0.6 pN of tension never shortened due to the formation of plectonemes that would have accompanied negative supercoiling as shown at lower tension (see Introduction of negative supercoils in normal DNA by E. coli and Salmonella gyrases). For gyrase, the ATP concentration affected pausing, but the stiffness of DNA did not.

Although there is high homology among isoforms of bacterial gyrases, they have distinct features that appear to correlate with the life cycle of their respective organisms [37]. For example, the CTD of GyrA in E. coli stabilizes wrapping of the DNA adjacent to the G-segment to catalyze supercoiling despite weak coupling between ATP hydrolysis and strand passage. In contrast, the M. Tuberculosis gyrase homolog has a slower rate of ATP hydrolysis linked to slower DNA supercoiling, which may reflect the comparatively long life cycle of its host bacterium. Indeed, functional differences between the gyrases of E. coli and Salmonella may even underlie different supercoiling levels observed in these bacteria during rapid growth [38]. To test this hypothesis, the strand passage by single E. coli and Salmonella gyrases were compared. These enzymes relaxed positive supercoils in single DNA molecules very similarly (Fig. S7), and although E. coli gyrase paused less frequently at a higher ATP concentration (1.0 versus 0.2 mM ATP) the basic strand passage activities of the two enzymes appear to be similar. This is consistent with the high homology between the proteins with few net charge differences that might alter activity (Fig. S8). More distinct differences in the introduction of negative supercoils by the two proteins were observed under lower tension (see Introduction of negative supercoils in normal DNA by E. coli and Salmonella gyrases).

The activity of human topoisomerase IIα on supercoiled DAP- and Normal DNA

Identical assays were performed using TOP2A, which also catalyzes a two-gate mode of strand passage. However, TOP2A cannot wrap DNA to introduce negative supercoils [39] and lacks an insert found in the GyrA subunit which may buttress the TOPRIM domain [40]. Figure 7a shows representative relaxation traces for three different DNA molecules for normal and DAP-substituted DNA respectively. Fitting the relaxation rate as a function of ATP concentration (Fig. 7b) using a Michaelis-Menten model established K_M = 0.2 ± 0.1 mM and V_max = 2.6 ± 0.3 turns/s for normal and K_M = 0.21 ± 0.09 mM and V_max = 2.4 ± 0.3 turns/s for DAP DNA which were insignificantly different.

Figure 7.

Human topoisomerase IIα relaxes DAP DNA more slowly. (a) Three offset, representative records of TOP2A-catalyzed relaxations of normal (left) and DAP DNA (right) under 0.6 pN tension and with 1 mM ATP. The raw data (dots) and 0.5 s moving averages (solid lines) show variable delays before processive relaxation of supercoils that restored the tether to an extended state. Occasional pauses in the processive increases of tether length are more evident for TOP2A acting on DAP DNA (circled trace). (b) Relaxation rates measured by converting tether length changes to twists of DNA and ignoring pauses, were similar for DAP (crosses) and normal DNA (open circles) at several ATP concentrations. Fits of a Michaelis-Menten model are shown for normal (solid, K_m = 0.2 ± 0.1 mM, V_max = 2.6 ± 0.3 turns/s) and DAP (dashed, K_m = 0.21 ± 0.09 mM, V_max = 2.4 ± 0.3 turns/s) DNA. Note that DNA tethers with a high affinity gyrase site were used for measurements at 5 mM ATP. (c) Average pauses during otherwise processive relaxation were longer for TOP2A acting on DAP DNA at all ATP concentrations (see also Figure S7).

However, TOP2A paused longer on DAP DNA (Fig. 7c) and the average pauses observed with either DAP or normal DNA decreased with increasing ATP concentration. Just as for gyrase double exponentials fit the data much better than single exponentials and produced coefficients of determination greater than or equal to 0.99 (Fig. S9). As for fittings of gyrase pauses, when [ATP] increased, the mean pause time of TOP2A decreased and the amplitude of the fast decay component in the fitting of the distribution increased (Table S2, Fig. S6).

The transition of TOP2A from a paused state was sensitive to the stiffness of the DNA molecule as well as the concentration of ATP. This may reflect the fact that TOP2A lacks a sequence of amino acids, which increases DNA affinity and decreases DNA-stimulated ATPase activity in many gyrases [40]. Without this insertion, TOP2A might not be sufficiently buttressed to readily bend a gate segment of stiffer DAP-DNA to initiate a cycle of strand passage. Since the bending of the gate segment has been shown to be critical for activity and coordination of ATP hydrolysis and strand passage [41], a larger effect on the strand passage might have been expected. However, transcription rapidly changes DNA supercoiling both ahead and behind active polymerases [42–45], so even small differences in the efficiency of gate segment bending, such as those expected on the basis of sequence variation, may be significant. Increased pausing by human topoisomerase IIα but not gyrase on DAP-substituted versus normal DNA might mean that gyrase more easily bends DNA to reach the strand-passage competent intermediate. However, other possibilities should also be considered (see discussion).

Introduction of negative supercoils in normal DNA by E. coli and Salmonella gyrases

The bends of the G-segment observed in co-crystals with DNA [3–5] are sharp kinks that do not necessarily involve an A:T base pair, the site of DAP-substitutions. In addition, gyrases wrap adjacent DNA around the CTD of GyrA creating a positive supercoil. The actual wrap of DNA encompasses about 40 bp [12], and the rate of initiation and processive introduction of negative supercoils slows in DNA under more than 0.4 pN of tension even with spermidine in the buffer [33]. In the present experiments without spermidine, the tension had to be lower before either E. coli or Salmonella gyrase introduced negative supercoils processively. Relaxation or insertion of negative supercoils depended on the tension which was toggled between 0.2 and 0.4 pN (Fig. S10). E. coli gyrase added supercoils at a rate of just 1.8 ± 0.7 turns/s while Salmonella negatively supercoiled normal DNA almost twice as fast, 3.0 ± 0.7 turns/s, and to a slightly greater extent (Fig. S11). E. coli gyrase introduced an average of 10.5 turns in each processive burst of activity while Salmonella introduced an average of 12.5 turns/burst. These single molecule measurements indicate that the higher levels of negative supercoiling observed in plasmids harvested from E. coli with respect to Salmonella in mid-log phase growth are not due to greater intrinsic supercoiling activity by the E. coli gyrase [38].

Negative supercoiling activity versus DNA stiffness

Similarly to the sensitivity of gyrase to tension, the stiffness of the DNA was expected to alter the negative supercoiling activity of gyrase. Therefore the length of normal or DAP DNA tethers under 0.3 pN of tension was monitored in the presence of E. coli gyrase. During observation totaling 174 minutes, 14 bursts of negative supercoiling ranging from 2 up to 12 supercoils were observed for normal DNA (Fig. 8a). Tethers were mechanically relaxed following each burst, and gyrase negatively supercoiled the DNA by an average of 6 turns/burst. In contrast, on stiffer DAP DNA, during a total observation of 290 minutes, gyrase only produced two supercoiling events adding one or two supercoils (Fig. 8b). Gyrase was essentially unable to introduce negative supercoils in most DAP DNA tethers. Given the efficiency with which many restriction enzymes and Taq polymerases operate on DAP DNA [22, 23], the fact that TOP2A relaxes positive supercoils in normal and DAP DNA at similar rates, and the decreased wrapping of DAP DNA by gyrase, the failure to negative supercoil DAP DNA is unlikely to result from deformation of the DNA that impedes reversible cleavage and re-ligation of the gate segment. Indeed, short DAP-substituted DNA fragments appear to be undistorted, B-form DNA[20]. Instead, due to the increased stiffness of DAP DNA, gyrase likely fails to wrap the adjacent DAP DNA segment prior to strand passage.

Figure 8.

Three representative time records (offset) showing plectonemes formed by the insertion of negative supercoils by E. coli gyrase. Normal or DAP-substituted DNA tethers were stretched with 0.3 pN of tension in the presence of E. coli gyrase and ATP. One second moving averages (black traces) of the fluctuating lengths observed for the tether (grey dots) are shown. During a total observation time of 174 minutes, 14 such bursts of processive supercoiling were observed for gyrase with normal DNA. The number of supercoils in a burst averaged 6 and ranged from 2 to 12. In contrast, with DAP-substituted DNA, during a total observation time of 290 minutes gyrase only introduced two supercoils in one tether and one in another.

Discussion

It is noteworthy that, on stiffer DAP DNA E. coli gyrase paused similarly but slowed by 30% with respect to normal DNA. On the other hand, TOP2A relaxed positive supercoils at almost equivalent rates on the two types of DNA but paused almost 2–3 times as long on DAP-substituted compared to normal DNA. The fact that processive relaxation rates for TOP2A were equivalent suggests that the DAP substitution per se does not significantly alter cleavage and re-ligation kinetics of strand passage.

For gyrase, the lower rate on stiffer DAP DNA may reflect the energy required to bend the G-segment. Lee et al. [41] has reported on mutants of the isoleucine residues in E. coli Topo IV that bent DNA to lesser degrees and found that ATPase activity correlated with bending. Crystallography of human TOP2A and yeast topoisomerase II revealed two 65–75 degree kinks in DNA which were produced by intercalating isoleucine residues into G- segment DNA [3, 4]. AFM and FRET measurements predict that the DNA bending angle is very similar for TOP2A and yeast topoisomerase II [6]. As is the case for other DNA bending proteins [46], these sharp kinks disrupt stacking but not Watson-Crick base pairing. For example, the bacterial integration host factor protein inserts prolines into the minor groove to create two sharp kinks and bend a DNA segment by 180 degrees [15]. Note that substitution of adenine with 2,6 diaminopurine (DAP) at IHF binding sites significantly reduces the binding affinity [30]. Thus it is likely that the additional hydrogen bond of DAP:T base pairs stabilizes the duplex and reduces kink formation [47]. However, gyrase and TOP2A utilize the same mechanism to relax positive supercoils in DNA under high tension, and TOP2A catalyzed strand passage at comparable rates on both types of DNA, even without a short sequence found in gyrase that is thought to consolidate the linkage between ATP hydrolysis and strand passage [24].

Instead the slower relaxation of positive supercoils observed for gyrase is likely to be related to the difficulty of bending stiffer DAP DNA around the CTDs of GyrA. As shown herein, gyrase cleaves and religates positively supercoiled DAP DNA less than a factor of two more slowly than normal DNA but essentially fails to negatively supercoil DAP DNA. Although wrapping DNA around the C-terminals of GyrA is not required for gyrase to relax positively supercoiled DNA tethers, such wraps may anchor the G-segment on the TOPRIM domains to immobilize and open it during strand passage. Given the structure of the GyrA subunits, the “omega” structure proposed by Basu et al. [7] is a likely intermediate in all strand passage events catalyzed by gyrase whether the crossover is pre-existing or generated by wrapping around a C-terminal domain. Since DAP DNA cannot be wrapped as easily as normal DNA, a subsequently weak hold on the gate segment might slow strand passage.

On the other hand, an analysis of pausing suggests that DAP DNA may fit more easily into the GyrB cavity of gyrase but not TOP2A. Corbett et al. [48] determined the structures of Topo VI bound to ATP analogs and concluded that the T- segment acts as a fulcrum upon which to lever the transducer segments apart. This seems to be a critical step for phosphate release which is rate limiting for the enzyme [49]. The shorter axial rise of DAP-DNA suggests that it likely resembles the broader A-form double helix with a diameter of about 2.3 nm. While the normal and DAP DNA might fit into the 2.0 nm cavity formed by closure of the N-gate of gyrase [50] without creating too much stress, normal and especially broader DAP DNA might obstruct closure of the N-gate of hTOP2A which has a smaller 14 × 17 angstrom cavity [51]. This could explain the extended pausing observed for TOP2A, but not gyrase, on DAP DNA. Both enzymes may have ATP-dependent and independent pathways to the intermediate state with a bent gate segment and a transfer segment trapped by an ATP-loaded and latched N-gate, but the larger GyrB chamber of gyrase should better accommodate DAP DNA.

Based on structural and electrostatic study of the C-terminal domain of gyrase, Corbett et al. proposed that gyrase introduces three 60 degree kinks to produce a total 180 degree arc in DNA [12]. Since modeling studies have revealed that the DNA double helix is most easily kinked toward the minor groove, exocyclic amino groups in the minor grove of DAP-substituted DNA may buttress the minor groove, requiring more energy to kink DAP-substituted DNA [19]. Unfortunately, reliable information regarding the energy required to kink DNA is not available [52], so it is difficult to estimate the additional energy required to wrap DAP DNA. Nonetheless gyrase struggled to negatively supercoil stiffer DAP DNA. This suggests that the C-terminal domains of the GyrA subunits have evolved just enough binding affinity to catalyze wrapping. Operating with barely enough catalytic potential would allow activity to be modulated by altered DNA stiffness due to particular sequences or small molecules like polyamines. This may have transcriptional consequences as has been shown for PrrA regulated genes in bacteria [53].

Materials and Methods

Electrophoretic Mobility Shift Assay (EMSA)

DNA fragment for EMSA

DNA fragments for EMSA experiments were produced by polymerase chain reaction (PCR) using the template pBR322 (New England Biolabs (NEB), Ipswich, MA), primers (Integrated DNA Technologies, Coralville, IA), Taq DNA polymerase (NEB) and nucleotide triphosphates (Fisher/Thermoscientific, Pittsburgh, PA) for normal DNA. 2-Amino-2′-deoxyadenosine-5′-Triphosphate (TriLink, San Diego, CA) replaced ATP in PCR of DAP-substituted DNA (Fig. 1a). 200 base pair PCR products were digested with NgoMIV and AluI to eliminate incorporated primers leaving one 5′CCGG overhang and the other end blunt. The resulting 160 bp DNA fragment included the preferential cleavage site for E. coli gyrase at position 990 of pBR322. PCR amplicons and digest products were purified using QIA quick Gel Extraction kit (QIAGEN, Valencia, CA). Klenow Fragment (3′ to 5′ exo-, NEB) was used to label the DNA by incorporating Cy5-dCTP (GE healthcare, Pittsburgh, PA) to polymerize opposite G in the 5′ overhang. The molar ratio of Cy5-dCTP to DNA was about 10:1. Cy5 is a red fluorescent dye with excitation/emission maximums at 649/670 nm. The labeled products were purified using QIAquick Nucleotide Removal kit (QIAGEN) to remove the unincorporated Cy5-dCTP.

Absorption spectra of normal and DAP DNA

To precisely determine DNA concentrations, the UV absorption at 260 nm was measured using a spectrophotometer (UV-1601 UV-visible, SHIMADZU) and the Lambert-Beer equation was solved using appropriate ε values. Since the extinction coefficients for the DAP and adenine nucleotide differ, the IDTdna “UV spectrum of DNA” tool (http://biophysics.idtdna.com/UVSpectrum.html) was used to calculate the extinction for each ssDNA containing DAP (ε₁ and ε₂). Then the excitation coefficient for dsDNA (ε_D) was calculated using ε_D = (1−h_260nm)*(ε₁+ε₂). In this formula, h_260nm= 0.287*f_AT + 0.059*f_GC, and f_AT and f_GC are the percentage of AT and GC base pairs. ε_D for the 160 bp DAP DNA fragments was calculated to be 2,411,218 l/(mole cm). ε_D for the 160 bp normal DNA was 2,564,320 l/(mole cm).

Electrophoretic Mobility shift Assay

E. coli DNA gyrase (NEB) at varying concentrations (184 – 920 nM) was incubated with 40 nM of Cy5-labeled, 160 bp double stranded DNA for 30 min. in 35 mM Tris-HCl pH 7.5, 5 mM MgCl₂, 24 mM KCl, 2mM DTT and 10% glycerol at room temperature. The total volume of the reaction was 25 μl. Reactions were loaded directly onto a 4% nativepolyacrylamide gel in 45 mM Tris-borate and 5 mM MgCl₂ [54]. After electrophoresis on ice for about 2 hours at 8V/cm, the DNA was visualized using a Typhoon Trio fluorescent scanner (GE Healthcare) with detection at 670 nm. The free and gyrase-complexed DNA bands were quantified using ImageQuant software (GE Healthcare).

Preparation of the DNA sample and flow chamber for magnetic tweezers

DNA fragment for magnetic Tweezers

Both normal and DAP-substituted DNA fragments 3220 (or 3055) bp in length were produced by PCR using plasmid DNA templates, pDL2317 [55] or pUC18-nuB104 (provided by N.P. Higgins at the University of Alabama) which contains a site from the μ phage DNA with high affinity for gyrase [13]) and forward and reverse primers containing respectively ApaI and SacII restriction sites. For DAP-substituted DNA, 2,6-diaminopurine −5′-Triphosphate (Trilink) replaced dATP in the reaction. The cycling parameters were: 3:00 min at 95 °C, 25 cycles of (1:00 min at 94 °C, 2:00 min at 50 °C, 6:00 min at 72 °C), 5:00 min at 72 °C, 3:00 min at 55 °C, and 3:00 min at 37 °C. The amplicons were purified using a QIA quick Gel Extraction kit (QIAGEN) and digested with ApaI and SacII to produce overhangs on both ends. Since pDL2317 does not contain ApaI and SacII restriction sites, only amplified DNA fragments were digested. This scheme avoided contamination of DAP-substituted DNA tethers with the normal DNA template. About 1 kbp segments of bio- and dig-labeled DNA “tails” were produced by PCR with either 1:9 biotin- 11-dUTP: dTTP or 1:9 digoxigenin-11-dUTP: dTTP in the reaction mixture using template pBluKSP. This substitution resulted in approximately 5% labeling with biotin and digoxigenin in the DNA tails. The bio- and dig-tails were digested using ApaI and SacII respectively, purified, and ligated to opposite overhangs of the 3.2 kbp normal or DAP fragments using T4 ligase (NEB).

Flow chamber assembly and preparation

The flow chambers were assembled between either two coverglasses or a coverglass and a slide with two fluid input/output holes. When low forces were employed, the re-sealable slide was convenient to prevent evaporation in longer experiments by covering the holes with small tabs of silicone gasket material. Instead the coverglass sandwich, which allowed minimal separation from the magnet, was necessary for high tension measurements. The glass surfaces were rinsed with ethanol and air-dried before assembling the approximately 40 μl chamber. To assemble the chamber, vacuum grease was applied to form a channel between two strips of double-sided tape on a slide or a 0.17 mm-thick coverslip, and a top coverglass was secured on the two strips of double-sided tape. In order to immobilize magnetic beads as references for the microscopy, 0.2 μl of 2.8 or 1 μm diameter paramagnetic, streptavidin-coated beads (DYNAL M280 or MyOne, Invitrogen, Carlsbad, CA) were diluted in 50 μl phosphate buffered saline (PBS) and drawn into the flow chamber. After at least 15 minutes incubation, the chambers were flushed with 200 μl PBS buffer to eliminate beads floating in the chamber. 60 μl of 20 μg/ml anti-digoxigenin (Roche) was introduced into the chamber and incubated overnight at 4 degrees °C. Then the flow chamber was flushed with 800 μl of PBS, and 80 μl BSA (100 μg/ml) was introduced and incubated for one hour. In parallel, 1 μl of paramagnetic beads were washed twice: first in 200 μl PBS and then in 100 μl PBS. These beads were diluted in 10 μl PBS and mixed with 2 μl of the DNA ligation product. This bead-DNA mixture was diluted in 300 μl of 10 mM Tis-HCl (pH 7.4), 200 mM KCl, 5% dimethyl sulfoxide (DMSO), 0.1 mM ethylenediaminetetraacetic acid (EDTA), 0.2 mM dithiothreitol (DTT) (lambda buffer). The flow chamber was then rinsed with 800 μl lambda buffer, and the diluted beads and DNA mixtures were introduced and incubated for one hour before beginning an experiment. During experiments, fresh solution was used to flush out unbound beads as necessary. The maximum tension ranged from 1 pN to 20 pN for MyOne beads and 20 pN to 70 pN for M280 beads.

Persistence length of DAP-substituted DNA

Persistence lengths of DNA were measured in lambda buffer in coverglass flow chambers by fitting the tension versus extension curve with a worm-like chain model. DNA extension (end-to-end distance) and tension and were determined from x, y, z excursions of the tethered bead [26] tracked in real-time using custom Matlab routines.

Gyrase wrapping assay

The gyrase wrapping experiments were performed in 35 mM Tris-HCl, 24 mM KCl, 4 mM MgCl₂, 2 mM DTT, pH=7.6 (gyrase reaction buffer) at room temperature on tethers stretched by 0.4 pN with 1 μm paramagnetic beads (DYNAL). Re-sealable flow chambers were sealed with two small pieces of silicon gasket material to permit hours of measurement without evaporation. All recordings of tether lengths were time averaged (10 s) to suppress noise due to thermal fluctuations of the bead. The lengths of time intervals during which either normal or DAP DNA tether remained contracted in the presence of gyrase were plotted as the fraction of all pauses exceeding times along the abscissa and fitted with double exponential decays.

Gyrase and human topoIIα supercoil relaxation assay

Recombinant human topoisomerase II α (USB) was used at a final concentration of about 0.6 nM. E. coli DNA gyrase (NEB) was used at a final concentration of about 9.2 nM (20 units/ml). The human topoisomerase II α (TOP2A) and gyrase relaxation experiments were performed in 50 mM Tris-HCl, 50 mM KCl, 8 mM MgCl₂, 1 mM EDTA, 0.5 mM DTT, pH=7.9 (Topo II buffer) at room temperature [36] with DNA under 0.6 pN of tension. The flow chamber was flushed with Topo II buffer before adding TOP2A or gyrase. ATP (USB) was added into the reaction mixture as indicated. Calibration of the tether length versus the number of turns of applied to the DNA tether before adding TOP2A or gyrase permitted measurements of the rates of relaxation (see Figure S4).

Gyrase negative supercoiling assay

E. coli DNA (NEB) or Salmonella (N.P. Higgins, Univ. of Alabama) gyrase were used at a final concentration of 20 nM. The negative supercoiling assays were performed in Topo II buffer at room temperature [36] at 0.3 pN. The flow chamber was flushed with Topo II buffer before adding gyrase. ATP (USB) was added into the reaction mixture at the indicated concentrations.

Highlights.

• Stiffened DNA might interfere with type II topoisomerases

• Diaminopurine-substituted (DAP) DNA is 35% stiffer than normal DNA

• E. coli gyrase slows and human topoisomerase IIα pauses when relaxing (+) supercoils in DAP DNA

• Salmonella gyrase negatively supercoils DNA slightly faster and further than E. coli gyrase

• E. coli gyrase fails to negatively supercoil DAP DNA