We have shown that the DNA bend angles imposed by Topo IV, hTopo IIα and yTopo II are similar and are not correlated with the extent of their topology simplification activities. This result is inconsistent with the bend angle model, in which the degree of topology simplification scales with the magnitude of the imposed bend angle (20
). The relationship between topology simplification and bend angle determined by Klenin et al.
) suggests that Topo IV should impose the largest bend angle (~310°), hTopo IIα should impose a smaller angle (~230°) and yTopo II should impose the smallest angle (~100°). (A detailed description of the derivation of the relationship between predicted bend angle and topology simplification is described in the Supplementary Data
.) However, our AFM and single-molecule FRET data indicate that the three topoisomerases impose similar bend angles. Our AFM data suggested that yTopo II imposed the largest bend angle (100
7°), followed by hTopo IIα (95
24°), and lastly Topo IV (94
13°). Our FRET data suggest larger angles of 140
16° for yTopo II, 136
17° for hTopo IIα and 126
18° for Topo IV (, ). Within each technique, all three bend angles are within 15° of one another, which we consider to be within our experimental uncertainty. Also, the angles follow the opposite order of the predicted bend angles, and there is no evidence of the ~3–fold difference in bend angles required by the bend angle model (20
). This leads us to conclude that while DNA bending is prevalent in all topoisomerases and may be an indicator of some conserved topology shifting mechanism, bending alone, within the context of the bend angle model, cannot solely explain topology simplification by type IIA topoisomerases.
Figure 8. Comparison of measured and predicted bend angles imposed by type IIA topoisomerases. Shown is a plot of topology simplification ability (R) as a function of measured and predicted bend angles for Topo IV (red squares), yTopo II (green triangles) and hTopo (more ...)
Though AFM measurements have consistently been shown to accurately measure protein–DNA interactions and conformations (35
), we verified that deposition conditions favored 2D equilibration of the DNA molecules on the mica surface and hence that the data accurately represent the conformations of both the DNA and the protein-DNA complexes (54–56
). 2D DNA equilibration is further supported by the agreement of bend angles determined by tangent measurements (both manual and automated) and EED measurements. Furthermore, the lack of correlation between the height of the DNA segments emerging from the protein and the measured bend angle is additional evidence that the protein–DNA complex equilibrated in 2D (see Supplementary Data
Although direct visualization of DNA–type IIA topoisomerase complexes and measurement of bend angles have not been previously reported, other methods, such as protein–DNA co-crystalization, single-molecule DNA manipulation and DNA cyclization have been employed to probe topoisomerase-induced DNA bending. Crystal structures of several type IIA topoisomerase–DNA complexes have been reported in the literature. These include the TOPRIM fold, which is a conserved domain required for DNA cleavage, and primary DNA-binding domain of yTopo II (25
), the breakage-reunion and TOPRIM domains of S. pneumoniae
Topo IV in the presence of the quinolones moxifloxacin and clinafloxacin (23
), and the ParE28-ParC58 fusion of A. baumannii
Topo IV in the presence of the quinolone moxifloxacin (24
). The yTopo II–DNA structure reported a DNA bend angle of ~150°, and we estimated similar bend angles from the S. pneumoniae
Topo IV–DNA crystal structure and the A. baumannii
Topo IV–DNA crystal structure. In fact, the bend angle imposed on DNA by A. baumannii
Topo IV, which has a high degree of sequence identity (61%) with E. coli
Topo IV, was slightly smaller than the bend angle imposed on DNA by yTopo II, which is consistent with our results.
Our measurement of the bend angle imposed by yTopo II from AFM images (100
7°) is smaller than the angle measured from the crystallized protein–DNA complex, though the angle measured from FRET experiments (140
16°) is much closer to this value (~150°) (25
). The discrepancy could arise from several effects, including different buffer conditions used in AFM and single-molecule FRET experiments that could have affected the bend angles. FRET experiments were performed under the exact buffer conditions as the relaxation assays, whereas AFM experiments had to be performed in a buffer optimized for AFM deposition. We observed that monovalent salt concentration, in particular, had a strong influence on bend angle in our single-molecule FRET experiments (data not shown). Regardless of the discrepancies, our AFM and FRET data both show that type IIA topoisomerases bend DNA to a similar extent. This result is consistent with published DNA–topoisomerase crystal structures, which show that all crystallized protein–DNA complexes reveal comparable DNA bending by type IIA topoisomerases from very different organisms (23–26
). Though the exact role of DNA bending in the mechanism of type IIA topoisomerases has not been determined, one study suggests that these enzymes require the DNA to be under considerable strain in order for cleavage to occur (57
). Perhaps DNA bending provides the necessary distortion of the double helix that allows the cleavage reaction and thereby the relaxation reaction to proceed.
The extent of DNA bending by Topo IV has also been estimated from single-molecule measurements of the size of plectonemic loops in supercoiled DNA with Topo IV bound, which indicate that Topo IV imposes a radius of curvature of ~6.4
nm onto DNA (27
). The radius of curvature can be related to a bend angle given assumptions about the distance over which the circular curvature approximation holds. A reasonable assumption is that the radius of curvature holds for an arc length equal to the number of DNA base pairs that interact with the protein. Estimating this length from footprinting data showing ~34
bp of protected DNA (34
), we calculated the bend angle could be as large as ~135° (assuming the protein bends the DNA over 45
bp) or as small as ~75° (assuming the protein bends the DNA over 25
bp). Our measured angle for Topo IV is consistent with this range. It should also be noted that the radius of curvature determined from the magnetic tweezers experiment is not consistent with a bend angle >180° and certainly not as large as the ~310° angle suggested by the bend angle model. DNA bending by type IIA topoisomerases has also been probed through DNA cyclization experiments (19
). Though these experiments have shown that Topo IV bends DNA and yTopo II does not, we noted in our AFM images, and others have observed (22
), that type IIA topoisomerases have a high affinity for DNA ends. This renders the cyclization data difficult to interpret since the presence of a topoisomerase on the ends would likely confound the ligation reaction necessary to achieve cyclization.
Though the bend angle model was developed to explain non-equilibrium topology simplification, the sharp DNA bending predicted by the model has other implications. Klenin et al.
) used Monte Carlo simulations to calculate the effects of sharply bent G–segment DNA on topoisomerase binding and activity. Many of these predictions have been addressed in subsequent publications, but their significance with respect to the bend angle model has not been discussed. For instance, the bend angle model suggests that a sharp bend imposed on DNA by Topo IV would result in a twenty fold higher binding affinity for positively supercoiled [(+)sc] DNA than for negatively supercoiled [(–)sc] DNA. However, the binding affinity of Topo IV for (+)sc and (–)sc DNA is the same or at most a factor of ~3 larger for (+)sc DNA (58–60
). A large bend angle imposed by Topo IV would also provide a mechanism to explain the higher efficiency of Topo IV in relaxing (+)sc DNA than (–)sc DNA (58–60
). However, recent work suggests that chiral discrimination by Topo IV results from differences in processivity rather than initial binding differences (59
). Moreover, hTopo IIα has also been shown to relax (+)sc DNA an order of magnitude more efficiently than (–)sc DNA, yet its binding affinity is slightly higher for (–)sc DNA (61
). The bend angle model also predicts that Topo IV should localize at apices of supercoils 87% of the time for (–) supercoils, but only 28% of the time for (+) supercoils (20
). Whereas this prediction has not been directly tested, data from magnetic tweezers pulling assays suggest that Topo IV has a 50% or higher affinity for (+) supercoil plectoneme apices (27
). Furthermore, recent simulation data using an improved WLC model for DNA has shown that a DNA hairpin, i.e. a sharp bend formed by a topoisomerase, is not sufficient to reproduce the experimentally observed degree of topology simplification (28
). Measurements of the DNA circle size dependence of non-equilibrium relaxation provide additional evidence suggesting that the bend angle model does not completely describe non-equilibrium topological relaxation (6
). The impact of DNA bending, and therefore the efficiency of the non-equilibrium relaxation process, would be expected to increase as the DNA circle size decreases. However, experiments with yTopo II and Topo IV show that the topology simplification activity decreases with circle size for small DNA circles and is independent of circle size for larger DNA circles. The ensemble of the available evidence from previous studies and from the measurements of the bend angle presented here suggests that the bend angle model cannot fully account for the observed non-equilibrium relaxation activity of type IIA topoisomerases. However, it is possible that non-equilibrium relaxation results from G–segment bending in combination with a second mechanism. All three topoisomerases were found to impose comparable bends onto DNA, which suggests that the bend angle model is incapable of explaining the measured differences between the topology simplification abilities of these three enzymes. However, these bend angles are somewhat consistent with the degree of bending expected from the least capable topology simplifier, yTopo II. Thus, it is conceivable that DNA bending is a conserved mechanism that is able to account for the base level of topology simplification achievable by type IIA topoisomerases. Further levels of topology simplification, as found in Topo IV and hTopo IIα, must arise from an additional enzyme specific mechanism, likely unrelated to DNA bending.
Several other models have been proposed to explain the topology simplification mechanism of type IIA topoisomerases (22
). A tracking model proposes that the enzyme binds to a DNA crossing and tracks along the DNA to trap T–segments that are catenated or knotted (15
). However, an experiment that placed tightly bound protein ‘roadblocks’ at several locations along supercoiled circular DNA did not affect non-equilibrium supercoil relaxation (16
). A three segment binding model postulates that the topoisomerase binds two potential T–segments prior to selecting one for strand-passage based on local geometry (21
). However, this model predicts an asymmetric removal of positive versus negative supercoils that would result in a skewed topoisomer distribution, which we did not observe for any of the type IIA topoisomerases studied (). Other studies have also failed to detect asymmetric supercoil removal by type IIA topoisomerases (16
), so while this model may hold in certain cases, it does not explain the more general mechanism of topology simplification. Two compelling possibilities are the hooked juxtaposition model and the kinetic proofreading model. The hooked juxtaposition model postulates that type IIA topoisomerases detect and relax specific juxtapositions of catenated, knotted and supercoiled DNA in which the G– and T–segments are bent toward one another (17
). Simulations based on lattice and WLC models indicate that strand passage at these hooked juxtapositions is sufficient to produce non-equilibrium topology simplification (28
). The kinetic proofreading model suggests that upon binding to the G–segment of DNA and encountering an initial T–segment, the topoisomerase becomes transiently activated, perhaps by binding one of the two ATP molecules. The T–segment is released and strand passage will occur if a second T–segment is captured while the enzyme remains in the active state (18
). It is possible that one of these models, perhaps coupled with the small effect arising from the bend angles imposed by the topoisomerase, could account for the non-equilibrium topology simplification activity of type IIA topoisomerases. The relatively sharp bend imposed on the G–segment DNA by type IIA topoisomerases is consistent with the hooked juxtaposition model, though it remains to be determined if bent T–segments are preferentially captured and passed. Further experiments are necessary to test these models to determine which, if either, can explain this fascinating phenomenon.