Auxiliary ATP binding sites support DNA unwinding by RecBCD

The RecBCD helicase initiates double-stranded break repair in bacteria by processively unwinding DNA with a rate approaching ∼1,600 bp·s−1, but the mechanism enabling such a fast rate is unknown. Employing a wide range of methodologies — including equilibrium and time-resolved binding experiments, ensemble and single-molecule unwinding assays, and crosslinking followed by mass spectrometry — we reveal the existence of auxiliary binding sites in the RecC subunit, where ATP binds with lower affinity and distinct chemical interactions as compared to the known catalytic sites. The essentiality and functionality of these sites are demonstrated by their impact on the survival of E.coli after exposure to damage-inducing radiation. We propose a model by which RecBCD achieves its optimized unwinding rate, even when ATP is scarce, by using the auxiliary binding sites to increase the flux of ATP to its catalytic sites.


SUPPLEMENTARY NOTE 1 Nucleotide binding to RecBCD cannot be accounted by the two catalytic sites
Binding of mant-nucleotides to RecBCD and RecBCD×DNA measured by FRET exhibited a biphasic binding pattern. Since the RecBCD complex possesses two well defined catalytic nucleotide-binding sites, in RecB and RecD, one might postulate that these two binding phases reflect binding of nucleotides to each one of these catalytic sites, or that a cooperative model of two binding sites can give rise to such pattern. However, here we show that the biphasic nature of the binding isotherm cannot be accounted for by two binding sites. In what follows we analyze the possible cases involving two binding sites and show that none of them can give rise to a biphasic curve.
Model 1: Two independent binding sites. In this case, each site represents binding of a nucleotide to a catalytic site and the sites are non-cooperative in their binding. It is well known that binding of a single ligand to a single site gives rise to a hyperbolic pattern. Hence, the overall occupancy, , can be expressed as a weighted sum of two hyperbolas, i.e.
where is the ligand concentration and ! and " the dissociation constants of the binding sites. The derivative of this curve, is always positive and decreases with increasing concentrations. Therefore, will always have a hyperbolic pattern regardless of the values of ! and " . Simply put, the sum of two concave hyperbolas is also a concave function (Supplementary Fig. 2A). Since for a biphasic pattern, the derivative of the curve should decrease in the first phase and then increase in the second, this model cannot give rise to a biphasic pattern. is the most common way to describe cooperativity. Two cooperative sites on a macromolecule will result in a Hill coefficient that is non-unity. For positive cooperativity, the Hill coefficient is greater than unity resulting in a sigmoidal pattern which cannot explain the data ( Supplementary Fig. 2B). For negative cooperativity, the Hill coefficient is lower than unity resulting in a derivative, that is always positive and decreasing with increasing ligand concentration, resulting in a hyperbolic-like curve ( Supplementary Fig. 2C). Hence, this model cannot explain our data.
Thus, two sites are insufficient to give rise to a biphasic binding pattern. Following this argument, one can decompose the binding curve into two phases, one hyperbolic with high affinity and one sigmoidal with weak binding. The decomposition is shown in Supplementary Fig. 2D. The first hyperbolic phase could be due to a single site, or multiple loosely coupled binding sites. The second binding phase represents a sigmoidal pattern, necessitating additional sites (at least two, and cooperative) that are distinct from the site/s giving rise to the first phase. Assuming that the affinities of the weak and strong sites are well separated, we phenomenologically describe the total occupancy by the weighted sum of two Hill equations: where ( % , & ) are the microscopic dissociation constants, and ( ' , ( ) the Hill coefficients of the strong and weak phases, respectively. As shown in Supplementary Fig. 2D, this model can result in a biphasic pattern. Figs. 1A, B and D show that it fits our data well.

SUPPLEMENTARY NOTE 2 mant-Nucleotides binding kinetics detailed analysis
Time courses of mantADP (Fig. 3A) and mantATP ( Supplementary Fig. 6) binding to RecBCD displayed a double-exponential pattern. The fast phase observed rate -)*% +,%-/ of mantATP and mantADP binding to RecBCD and RecBCD·DNA complexes exhibited a hyperbolic concentration dependence on mant-nucleotide ( ) concentration ( Fig. 3B and Supplementary  Fig. 6). Hence, it is modeled by a reaction mechanism where a rapid equilibrium step is followed by an isomerization step to a high fluorescence complex × * , as follows: This two-step binding reaction predicts that the observed rate constants for binding should obey a rectangular hyperbola according to: where ! is a measure of the strength of the collision complex between RecBCD and , and ±" are the on (+) and off (-) rates of isomerization. Supplementary Table 6 summarizes the rate constants of mant-nucleotide binding to RecBCD and RecBCD·DNA complexes. The kinetics of both mantATP and mantADP binding to RecBCD are extremely fast with a relatively tight collision complex: ! is in the ranges of tenths of µM in all cases. Additionally, the overall apparent second order binding constants, 2" / ! , are an order of magnitude faster than many other molecular motors studied 1,2 . For comparison, it is ~30 fold faster than Rep; a SF1 DNA helicase 3 . Remarkably, the maximum rate of the observed isomerization is faster than 1000 s -1 at 6° C supporting rapid kinetics of nucleotide binding to RecBCD and RecBCD·DNA complexes.
Binding of mantATP to RecBCD or RecBCD·DNA is irreversible, with an intercept being indistinguishable from the origin ( $" ≈ 0).
The binding kinetics for mantATP and mantADP, in the absence or presence of a DNA substrate, is overall very similar in its mechanism and rates (Supplementary Fig. 9 & 10 and Supplementary Table 6). This may suggest that RecBCD has an overall weak selectivity for the type of nucleotide. The structural implication may be that RecBCD nucleotide binding sites are highly accommodating and translating almost every collision complex to a productive, high affinity state which may promote rapid catalysis and high turnover.

Supplementary
178,000 ± 1,600     Figure 1A. B. Transient kinetics of mantADP binding at normal conditions. Data shown as in Figure 3B. C. Equilibrium binding of mantADP in the presence of 2 mM adenosine. Data shown as in Figure  1B. D. Binding kinetics of mant-ADP in the presence of 2 mM adenosine. Data shown as in Figure  3D. E. Unwinding rates in the presence of 2 mM adenosine. Data shown as in Figure 4D. F. Equilibrium binding of mantADP at varying salt concentrations (75, 200, 300 mM NaCl in blue green and purple, respectively). Data shown as in Figure 1B. G. Unwinding rates as a function of salt for low (lower) and high ATP (higher). Data shown as in Figure 4B. RecBCD was purified following the same protocol as described in the 'Online Materials', immediately subjected to an additional MonoQ purification step and eluted with a salt gradient. The shaded area shows the fraction collected. B. The MonoQ-purified protein was tested for nucleotide binding and compared with the titration in Fig. 1B, showing nearly identical biphasic binding. Lines show the best fit to Eq. 1 (Methods). Fig. 17: Calibration of the equilibrium dialysis experiments indicates minimal loss across the membrane. Absorbance spectra of ADP: 20x diluted, before (yellow) and 10x diluted, after (dashed blue, side 1 and dashed red, side 2) equilibrium dialysis. The initial half absorbance at 259 nm was determined as 0.8332 (corresponding to 0.5411 M of ADP), the post-dialysis absorbance was determined as 0.8307 (0.5394 M of ADP) on the left side and 0.8267 (0.5368 M ADP) on the right side, indicating a total loss of 0.006 M of ADP and a 0.54% error. Inset: Zoom-in to the peaks at 259 nm.