VRR-Nuc Domains Are Structurally Related to Holliday-Junction-Resolving Enzymes
Despite considerable effort, we have thus far failed to crystallize FAN1 or isolate soluble VRR-Nuc domains from any of the FAN1 orthologs. However, we were able to solve the first VRR-Nuc domain structures using single-domain proteins derived from three bacteria and bacteriophage as surrogates: Psychrobacter (psNUC); Streptococcus equi bacteriophage P9 (stNUC); and Salmonella enteritidis typing phage 3 (saNUC) (Figure 2A). Crystallographic statistics are presented in Figure S2A.
stNUC was solved by single-wavelength anomalous diffraction (SAD) methods on crystals of selenomethione-substituted protein and refined against native 1.3 Å resolution data (Figures 2A and S2B). One molecule is present per asymmetric unit (AU) and comprises a central four-stranded β sheet surrounded by three α helices in a αβββαβα topology (Figure 2B). The protein crystallized as a dimer with monomers related by a crystallographic 2-fold axis and interacting through a substantial hydrophobic interface formed by the central β sheets of opposing molecules and bounded on either side by the α2 helix from each partner. Electrostatic surface representations highlight the position of the negatively charged active sites on each monomer (Figure 2A), which bind a single magnesium ion in a geometry that is distorted from a classical octahedral arrangement toward a square pyramidal configuration (Figure 2C). K47 is presumably involved in positioning the catalytic water molecule as would be expected from homology with other members of the restriction endonuclease-like superfamily (Knizewski et al., 2007). Q57, a conserved residue within VRR-Nuc domains (http://pfam.xfam.org/family/vrr_nuc), makes a hydrogen bond to K47 that is absent in archaeal Holliday junction resolvase (HJR) structures. The active sites are surrounded by positively charged residues in a figure-of-eight pattern, as seen in Sulfolobus solfataricus Hjc (Protein Data Bank [PDB] ID 1HH1; Bond et al., 2001), consistent with a role in substrate DNA recognition and orientation.
saNUC was solved by molecular replacement using the stNUC structure as a search model (Figure 2A), revealing a dimeric arrangement of two molecules in the AU essentially identical to that seen in the stNUC crystals. Despite only having 34% sequence identity with stNUC, both have core structures that are remarkably similar (root-mean-square deviation = 1.3 Å over 86 Cα) and the orientation of active site residues is essentially identical to stNUC.
psNUC was solved by SAD and refined against 2.0 Å resolution data. Here, the crystals contain six molecules per AU, comprising three dimers that, like those of saNUC and stNUC, form around a central hydrophobic β sheet interface. There are, nonetheless, several elaborations (Figure 2A). In particular, the linker between β1-β2 forms a partially structured extension over the active site surface. Such structures are common in Holliday-junction-resolving enzymes, including RecU (PDB ID 1ZP7), and likely play a role in the binding and orientation of substrate (Cañas et al., 2011). Hexacoordination of a magnesium ion is seen at each active site, involving the E-PD-EXK motif and two water molecules in a configuration stabilized by the conserved Q102 (Figure 2C). The active site surface is bounded by a ring of positively charged residues to allow substrate recognition (Figure 2A).
DALI searches (Holm and Sander, 1995) show that the structures with greatest similarity to the VRR-Nuc domains are all Holliday-junction-resolving enzymes exemplified by Hjc (Figure 2B). This observation raised the possibility that dimeric VRR-Nuc domain proteins selectively process four-way DNA junctions, and we therefore investigated the nuclease activities of these three proteins.