|Home | About | Journals | Submit | Contact Us | Français|
The E. coli RecBCD enzyme facilitates the loading of RecA onto single-stranded DNA produced by the combined helicase/nuclease activity of RecBCD. The nuclease domain of RecB protein, RecBnuc, has been previously shown to bind RecA. Surprisingly, RecBnuc also binds to phage and eukaryotic homologs of RecA, leading to the suggestion that RecBnuc interacts with the polymerization motif that is present in all three proteins. This mode of interaction could only be with monomeric RecA, as this motif would be buried in filaments. We show that RecBnuc binds extensively to the outside of RecA-DNA filaments. Three-dimensional reconstructions suggest that RecBnuc binds to the ATP-binding core of RecA, with a displacement of the C-terminal domain of RecA. Solution experiments confirm that the interaction of RecBnuc is only with the RecA core. Since the RecA C-terminal domain has been shown to be regulatory, the interaction observed may be part of the loading mechanism where RecB displaces the RecA C-terminal domain and activates a RecA monomer for polymerization.
The RecBCD complex plays a key role in homologous genetic recombination in E. coli. It serves to act as both a helicase and nuclease that together generate single-stranded DNA (ssDNA) tails that become covered with the RecA protein and which then act in strand invasion reactions1. It has been shown that RecBCD also plays a specific role in the loading of RecA onto these ssDNA tails2-4, and the RecA-binding portion of RecBCD has been shown to reside in the RecB nuclease (RecBnuc) domain3,5. Remarkably, it has been shown that RecBnuc also interacts with the phage and eukaryotic RecA homologs, UvsX and Rad51, respectively5. RecA and UvsX have a C-terminal domain6 that is absent in Rad51, while Rad51 has an N-terminal domain7 that is absent in RecA and UvsX. All three of these proteins, however, share a structurally conserved ATP- and DNA-binding core that is also present in such disparate proteins as the F1-ATPase8 and many helicases9.
Since RecA, UvsX and Rad51 appear to have a common “polymerization” motif that forms a part of the subunit-subunit interface in the filaments that are formed by these proteins7,10,11, it was suggested that RecBnuc binds to this motif5. Such a mode of interaction might be similar to that put forward for the Rad51-binding BRC motifs within BRCA2, since it was suggested10 that the BRC motifs mimic the polymerization motif in Rad51, bind to Rad51 monomers, and block polymerization of Rad5112. These BRC-Rad51 interactions appear to be more complicated, because it has now been shown that some BRC motifs can also bind to Rad51-DNA filaments13, an interaction that must involve a region of the Rad51 subunit that is not part of a buried polymerization motif within the filament.
We have examined the interaction of RecBnuc with RecA-DNA filaments using electron microscopy (EM). Naked RecA-DNA filaments (Fig. 1a) display the characteristic striations arising from the ~ 90−95 Å pitch RecA helix. After incubation with RecBnuc (with a 1:1 stoichiometry of RecA:RecBnuc) the filaments become less smooth (Fig. 1b), suggesting a binding of RecBnuc along the entire length of these filaments. The extensive binding was confirmed using three-dimensional reconstruction. We used the Iterative Helical Real Space Reconstruction (IHRSR) method14,15 to reconstruct the RecA-DNA-RecBnuc filaments from these EM images of negatively stained samples. We started the procedure with 48,633 overlapping segments (each 70 pixels or 332 Å in length). The IHRSR approach failed to converge to the same solution from different starting points, which is an indication of heterogeneity in structure16,17. The heterogeneity might come from partial binding of RecBnuc to RecA-DNA filaments, with incomplete occupancy resulting in structural heterogeneity. The heterogeneity might also arise from different binding modes, similar to what is seen for certain actin-binding proteins that can interact with F-actin polymorphically18-21. We therefore used three reference volumes to sort these segments based both on occupancy and binding modes: a naked RecADNA filament, a RecA-DNA filament with additional mass on the outside of the filament, and a RecA-DNA filament with additional mass in the groove of the filament. A multi-reference sorting against projections of these reference volumes was then used to classify image segments into three different groups. Slightly more than a third of the segments were classified as naked RecA-DNA filaments, ~ 40% of the segments were characterized as having additional mass on the outside of the RecA-DNA filaments, and the remaining quarter of the segments were classified as having additional mass in the groove. Such classification does not necessarily mean that the sorting is valid. One independent test of this sorting was using the same reference volumes to sort images of naked RecA-DNA filaments (Fig. 1a). Surprisingly, a significant amount (~ 25%) of the naked RecA-DNA segments were classified as having additional mass in the groove, comparable to the fraction of RecA-DNA-RecBnuc segments sorted into this category. This suggested to us that this category may simply be an artifact of incomplete stain penetration into the groove. However, no significant fraction of naked RecA-DNA segments was classified as having additional mass on the outside of the filament, the largest category in the RecA-DNA-RecBnuc filaments.
The classification method just described is model-dependent, so we also used a model-independent approach to sort the RecA-DNA-RecBnuc filaments based upon the projected density at different radii. By extracting segments showing the highest density on the outside of the RecA-DNA filament, a more homogeneous sample was obtained that showed extensive binding by RecBnuc (Fig. 2b) when compared to a control reconstruction of naked RecA-DNA filaments (Fig. 2a). The helical parameters of the RecA-DNA filaments bound by RecBnuc (an axial rise of 14.8 and a twist of 59.0° per subunit) were very similar to those found for naked RecA-DNA filaments, showing that the binding of RecBnuc introduces no major change in the helical geometry. A high degree of confidence exists in the interpretation of the domain structure within a low-resolution RecA-DNA reconstruction (Fig. 1a), since our model for this filament22 has been confirmed by a recent high-resolution crystal structure of an active RecA-DNA filament23. We have found no way to fit both RecA (PDB 3CMW) and the RecBnuc structure24 (PDB 1W36) into the reconstructed volume (Fig. 2b). The simplest explanation is that RecBnuc binds to the core of RecA, with the C-terminal domain of RecA shifted extensively by RecBnuc into different positions so that it is not seen in the helically averaged reconstruction. Such motions of the RecA C-terminal domain would be consistent with what has previously been observed by both EM22 and x-ray crystallography25. A molecular model showing RecBnuc bound to the core of RecA (Fig. 2b) suggests that the full molecular volume of RecBnuc cannot be accommodated. We suggest that the reconstruction is an average of naked RecA subunits (where the RecA C-terminal domain is normally positioned) and RecA subunits where RecBnuc is bound and the C-terminal domain is displaced.
This model for RecBnuc binding to a RecA-DNA filament suggests that RecBnuc should bind to the core of RecA in solution but not to the C-terminal domain. We have tested this by examining the interaction between the RecBnuc domain and the N-or C-terminal fragments of RecA. A construct was produced that contained the N-terminal region of RecA, comprising the first 268 residues which encompasses the N-terminal and the ATPase domains (ΔCRecA), fused with SUMO to both facilitate solubility and provide a hexahistidine tag. Pull-down assays were performed and an excess of untagged RecBnuc was mixed with ΔCRecA and Ni-agarose beads. After extensive washing, proteins retained on the beads were eluted and a complex between RecBnuc and ΔCRecA was detected (Fig. 3a). We also undertook pull-down assays with an alternative construct, in which the C-terminal domain of RecA (δNRecA, residues 269−352) was fused to SUMO. In this case, no interactions were detected (Fig. 3b), supporting the proposal that it is the core of RecA that interacts with the nuclease domain of RecB.
It was shown5 that RecBnuc also interacts with human Rad51 (hRad51), so we have used similar procedures to look at the binding of RecBnuc to hRad51-DNA filaments. When naked hRad51-DNA filaments (Fig. 1c) are compared with filaments after incubation with RecBnuc (Fig. 1d), filaments are seen to be decorated by RecBnuc in a similar manner to the decoration of RecADNA filaments (Fig. 1b). We used the IHRSR method to reconstruct the hRad51-DNA-RecBnuc filaments from EM images of negative stained samples. As with the RecA-DNA-RecBnuc filaments, the hRad51-DNA-RecBnuc filaments were judged to be heterogeneous by the failure of the IHRSR approach to converge to the same solution from different starting points. Similar sorting methods were used as for the RecA-DNA-RecBnuc filaments, and we were able to generate a three-dimensional reconstruction (Fig. 2d) from a relatively homogeneous subgroup. As with the case for RecA, our interpretation of the low-resolution domain structure (Fig. 2c) of the naked Rad51-DNA filaments26 has been confirmed by high-resolution x-ray crystal structures of these filaments7,11. What is also similar to the situation with the RecA-DNA-RecBnuc filaments is that the reconstruction of the decorated volume (Fig. 2d) is unable to accommodate both a molecule of hRad51 and a molecule of RecBnuc. However, the reconstructed density can be explained by RecBnuc bound to the core of Rad51 (Fig. 2d), with the Rad51 N-terminal domain completely missing. In this model, the interaction of RecBnuc with the Rad51 core would be very similar to the interaction between RecBnuc and the RecA core (Fig. 2b).
Is it reasonable to imagine that the Rad51 N-terminal domain would be invisible in Fig. 2d, presumably due to disorder? The RecBnuc bound to a Rad51 subunit one turn above would be in steric conflict with the N-terminal domain of another Rad51 subunit, and this would generate large displacements of the N-terminal domains. There are many observations showing that this domain is highly mobile. In a crystal structure of a related protein, archaeal RadA, only one N-terminal domain is seen in seven subunits due to large disorder27. In a crystal structure of the related Dmc1 protein, no N-terminal domains are seen in an octamer of subunits28. Electron microscopic reconstructions of yeast Rad51 filaments have shown that in the presence of a point mutation (G103E) within the N-terminal domain, the N-terminal domain is never seen after helical averaging29. Thus, the large displacements of the N-terminal domain that would be generated by the binding of RecBnuc to hRad51-DNA filaments appear to be consistent with many previous observations.
Although the complex of RecBnuc bound to hRad51-DNA filaments is not of biological significance, such binding must arise from RecBnuc recognizing a conserved structural surface that is also present in RecA, as suggested previously5. In contrast to the original prediction, however, this surface is not part of the subunit-subunit interface in the filament. On the other hand, the complex of RecBnuc with a RecA-DNA filament has great biological relevance, and provides insights into how RecBCD loads RecA onto ssDNA4. Although the stoichiometric binding of a RecB domain to a RecA polymer is not in itself expected to recapitulate what happens in a cell, stoichiometric binding of actin-binding proteins to F-actin and microtubule-binding proteins to microtubules have been extremely useful in understanding the structural basis of the interactions that will occur biologically at much lower stoichiometries. Many observations suggest that the RecA C-terminal domain plays a regulatory role in RecA filament formation and homologous recombination30-33. This regulatory role appears mainly to be negative, because truncations of the C-terminal domain can lead to enhanced binding to DNA and other activities in vitro, while C-terminal truncated RecA proteins can lead to constitutive induction of the SOS response in vivo34. Presumably, the C-terminal truncation of RecA leads to a protein that will more readily form activated filaments on double-stranded DNA within the cell, thereby cleaving the LexA repressor in the absence of the normal signal for DNA damage, the presence of naked ssDNA. We therefore think that it may be significant that the binding of RecBnuc to the core of RecA displaces the C-terminal domain. This may be the first step in the activation of a RecA subunit for polymerization, a process that is normally negatively regulated by the presence of the C-terminal domain.
This work was supported by NIH grant GM035269 (to E.H.E.) and CRUK (D.B.W.). The clone for overepression of the 6xHis-tagged RecBnuc was a generous gift from Dr. S. Kowalczykowski.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.