The sequence encoding Ala31-Gln274 of EndA(H160A) from S. pneumoniae was cloned into the pET30M expression vector as described previously (20). EndA(H160A) was expressed as a His-GST fusion protein in Rosetta2(DE3) cells, overnight at 18°C. The cells were pelleted by centrifugation, resuspended in 25 mM Tris pH 7.5, 500–750 mM NaCl and lysed by sonication. The soluble fraction was bound in-batch to glutathione sepharose 4B resin and EndA(H160A) was cleaved from the resin by TEV protease, overnight at 4°C. Uncleaved fusion protein contaminants were removed by a brief incubation with Ni-NTA agarose. Unbound protein was concentrated and loaded onto a Superdex 200 16/60 column equilibrated with 25 mM Tris pH 7.5, 75 mM NaCl. Peak fractions were collected and concentrated to ∼50 mg/ml.

EndA(H160A) was crystallized by mixing 1 μl of protein (51 mg/ml) with 1 μl of 0.1 M Tris pH 7, 17% PEG8K, 0.2 M MgCl₂. Crystals were grown by sitting drop vapor diffusion (21) at room temperature. The highest resolution data set was obtained from a glutaraldehyde cross-linked crystal. This crystal was transferred from the original drop into a 3.5 μl hanging drop containing 0.1 M Tris pH 7, 75 mM NaCl, 0.2 M MgCl₂ and 20% PEG8K (mother liquor). The crystal was placed in a drop over a vapor diffusion chamber containing a sitting drop bridge with 3.5 μl of 25% glutaraldehyde and 500 μl mother liquor for 8 min at room temperature. The crystal was then soaked in a solution containing 0.1 M Tris pH 7, 75 mM NaCl, 0.2 M MgCl₂, 20% PEG8K, 11.6 mM trimethyl lead acetate and 9.63 mM triethyl lead acetate for 43 h at room temperature for potential phasing opportunities. The soaked crystal was transferred to a cryoprotectant (0.1 M Tris pH 7, 75 mM NaCl, 0.2 M MgCl₂, 20% PEG8K, 15% ethylene glycol, 11.9 mM trimethyl lead acetate and 13.75 mM triethyl lead acetate) in four steps, then flash frozen in liquid nitrogen and placed in a stream of nitrogen gas cooled to −180°C. The data were collected at Southeast Regional Collaborative Access Team (SER-CAT) BM-22 beamline at the Advanced Photon Source (Argonne National Laboratory), at a wavelength of 0.94 Å, with the MAR225 area detector. The data were processed at 1.75 Å, in space group P2₁2₁2₁ using HKL2000 (22) (Table 1). All attempts at solving the structure of EndA(H160A) using molecular replacement with other DRGH nucleases as search models, heavy metal soaking or multi-wavelength anomalous dispersion with engineered selenomethionine-substituted protein failed to provide useful phase information. The phase problem was solved by crystallizing EndA as a MBP(wt)-EndA(H160A) fusion protein (23) and using a partial model of EndA(H160A) from this data to solve the 1.75 Å structure by molecular replacement (Supplementary Data). The quality of the electron density maps was improved by density modification and non-crystallographic averaging. An initial round of refinement was performed in crystallographic refinement package (CNS) (24), and the resulting structure was used as a starting model for iterative building using AutoBuild (25) in PHENIX (26). About 420 amino acids for two molecules of EndA(H160A) (86.1%) were built, with an R_free of 23%. The final model for EndA(H160A) was generated by iterative cycles of manual model building and refinement in O (27) and CNS (24) (R_free 21.6%). The refined model for EndA(H160A) contains amino acid residues Ser46-Tyr124 and Ser136-Gln274 in molecule A and residues Lys43-Lys125 and Ser134-Gln274 in molecule B (Table 1). No glutaraldehyde cross-linkages or lead derivative molecules were found in the density. Ramachandran statistics were determined by MolProbity (28), with 98.7% of residues in the favored regions, and 100% in allowed regions. Table 1.

Data collection and refinement statistics

The H160G mutation in EndA was obtained by site-directed mutagenesis, using pET30M-EndA(H160A) as a template for the reaction. Additional mutants were generated on the H160G background using the same method. All EndA proteins were expressed and purified as for EndA(H160A).

EndA proteins were diluted to 20 nM in activity buffer (20 mM Tris, 10 mM sodium acetate, 10 mM MES, 25 mM NaCl, 5 mM MgCl₂, pH 8) and incubated with 15 ng/μl of supercoiled pBluescript SK(+) plasmid for 30–35 min, in the presence or absence of 0.2 M imidazole. This concentration of imidazole was chosen for the reactions, as all phases (supercoiled, nicked open circle, linearized and degraded) of digestion are visible.

The reactions were quenched by addition of loading dye containing EDTA. Samples were run on a 0.8% (w/v) agarose gel, dissolved in 1× Tris–Acetate–EDTA buffer. DNA species in the gel were visualized by ethidum bromide staining, scanned with a Typhoon fluorescence imager and analyzed using ImageQuant. Nuclease activity assays were performed in triplicate.

Agarose (1% w/v) was dissolved in activity buffer (20 mM Tris, 10 mM sodium acetate, 10 mM MES, 25 mM NaCl, 5 mM MgCl₂, ±0.2 M imidazole, pH 8) and high molecular weight salmon sperm DNA (30 μg/ml) and ethidium bromide (1 μg/ml) were added immediately prior to casting. EndA proteins were diluted to 5 mg/mL in activity buffer (20 mM Tris, 10 mM sodium acetate, 10 mM MES, 25 mM NaCl, 5 mM MgCl₂, ±0.2 M imidazole, pH 8) and 1 μl of each protein solution was placed in a small well on the SRED plates. About 0.5 μl of DNase I (1 unit/μl) was added to each plate as a positive control. The plates were incubated overnight at 37°C. Nuclease activity is identified by visualization of a dark ‘halo’ around the well site, as seen under UV_254nm illumination. The relative extent of nuclease activity for each protein was determined by measurement of the halo radius. This experiment was performed in triplicate.

A 24mer oligonucleotide (5′-GCAAGGTGCGTGCAGCCTTGGACC-3′) was annealed to a complementary oligonucleotide by denaturation and slow cooling. The annealed DNA was mixed with 10 μM EndA protein in a 4:1 molar ratio (11.5 mM Tris pH 8, 10 mM MgCl₂, 7.5 mM NaCl) and incubated on ice at 4°C for 2 h. About 2 μl of High Density Tris Borate-EDTA (TBE) Loading Buffer was added to each sample, and 5μl of this mixture was loaded onto a 1% agarose/1% acrylamide gel in 0.25× TBE. DNA and protein species were serially stained on the same gel with ethidium bromide and SYPRO Orange, respectively and scanned using a Typhoon fluorescence imager. About 610 nm band-pass and 555 nm band-pass emission filters were used to detect the ethidium bromide and SYPRO Orange signals, respectively.

The nuclease domain (Ala31-Gln274) of EndA from S. pneumoniae, that contained the inactivating H160A mutation and also lacking the N-terminal membrane localization signal sequence, was crystallized and refined as summarized in Table 1 (PDB ID code 3OWV). A model for molecular replacement of EndA was obtained by crystallizing an N-terminal MBP fixed-arm fusion with the EndA nuclease domain and building a partial model to the EndA density which contained the core structural elements (see ‘Materials and Methods’ section, and Supplementary Data). The final structure for the EndA nuclease domain was refined through a combination of automated and manual model building. The protein core is comprised of a central β-sheet consisting of five antiparallel β-strands (β-strands 3,4,7,8 and 9, Figure 1 and Supplementary Figure S1). The ‘front’ face of the molecule contains the active site and putative DNA-binding surface. β-strands 5/6 and α-helix D form the ββα-metal finger motif that creates the catalytic center (Figures 1B and 2A). The ‘back’ face of the molecule contains little secondary structure. Aside from a single ordered loop between β-strands 2 and 3 traversing this face, most of the sheet is exposed to solvent (Supplementary Figure S2). Analysis of the electrostatic surface of EndA underscores the small size of the active site (Figure 1C and D), and reveals a fairly even distribution of positively and negatively charged residues on both faces of the protein. Residues Asn126-Gly135 connecting β-strands 4 and 5 are disordered in this structure, as in other crystal forms we were able to obtain, and were not included in the calculation of the electrostatic surface potential. Mass spectrometric analysis of dissolved crystals yielded a molecular weight consistent with the expected protein construct size rather than proteolyzed fragments (data not shown). Due to the exclusion of the putative transmembrane domain (Leu9–Leu18), and the lack of overt hydrophobic patches on the protein surface, the current structural information reveals little information on the mechanism by which EndA associates with the bacterial outer membrane. Figure 1.

Sequence and structure of EndA. (A) Sequence and secondary structural elements of EndA from S. pneumoniae. α-Helices are indicated as blue rectangles and β-strands as green arrows. The DRGH motif is boxed in gray. The position of the general base (His160) is shown in red. Mutated residues involved in catalysis are shown in pink. Residues likely involved in DNA substrate binding and/or catalysis are shown in purple. The ‘finger-loop’ is shown in cyan. Disordered regions are highlighted in yellow. (B) Ribbon diagram of EndA, showing the ‘front face’ of the molecule that contains the active site. Secondary structural elements are color-coded as in (A) and the magnesium ion is colored purple. The positions of the termini and the ends of the disordered loop are labeled. (C) Electrostatic surface potential of the ‘front face’ of EndA, in the same orientation as in (B). The surface was calculated using the Adaptive Poisson–Boltzmann Solver tool in PyMOL (47). The potential ranges from −8 kTe⁻¹ (red) to 8 kTe⁻¹ (blue). The location of the active site is highlighted by a yellow circle. (D) Electrostatic surface potential of the ‘back face’ of EndA [Disordered side chains lacking density were modeled onto the EndA surface for panels (C) and (D).].

Historically, active wild-type EndA is purified directly from S. pneumoniae or expressed in cell-free systems (15,20). In contrast, bacterial expression of the inactive EndA(H160A) variant yielded large amounts of extremely soluble (>75 mg/ml) and stable protein. Nuclease activity for EndA(H160A) could be rescued by addition of imidazole (38,39) using a plasmid conversion nuclease activity assay (20). This assay measures conversion of supercoiled substrate to nicked open circle, linear duplex or degraded fragments. Nuclease activity was recovered in the presence of imidazole, however, the extent of rescue was relatively low (Figure 3). One goal of our crystallographic studies was to use structural information from EndA(H160A) to develop a platform from which more robust imidazole rescue experiments could be performed, using easily expressed, catalytically inactive nuclease variants. Figure 3.

Plasmid conversion nuclease activity assay for EndA(H160A) and EndA(H160G). The assay measures conversion of the supercoiled plasmid substrate to nicked open circle or linear (double strand breaks). The assay was performed in the presence of 200 mM imidazole and 10 mM MgCl₂).

Many ββα-metal finger nucleases exhibit an interesting evolutionary active site design, which is optimized to bind a hydrated divalent cation (32,34–36). The structure of EndA contains a bound magnesium ion coordinated by Asn191, and five water molecules (Figure 4). This ion was refined as magnesium because of the high concentration of MgCl₂ (200 mM) present in the crystallization buffer, the octahedral coordination geometry and because the bond distances (2.07–2.21 Å) between the ion and the water molecules were consistent with those expected from magnesium. In order to determine whether EndA displays a preference for certain divalent metal cations, the plasmid conversion assay was performed in the presence of 200 mM imidazole and 10 mM of various cations. Imidazole-rescued EndA(H160G) activity appears to have little preference between Mg²⁺ and Mn²⁺, but shows little detectable activity in the presence of Ca²⁺ or Zn²⁺ (Supplementary Figure S3B). Figure 4.

Mutagenesis of catalytic and DNA-binding residues in the active site of EndA. (A) Surface rendering of EndA, with the modeled DNA substrate from the VVN nuclease [PDB ID code 1OUP (42)] superimposed in the active site. (B) Stereo ribbon diagram of the active site of EndA, displaying the ‘finger-loop’ (cyan), catalytic residues (pink) and DNA-binding residues (purple) targeted for mutagenesis. The disordered loop is drawn in dashed yellow. Putative positions of Arg127 and Lys128 on the disordered loop are shown as purple triangles. (C) Ribbon diagram of the ‘finger-loop’ position (cyan), relative to the putative position of a bound DNA substrate (tan).

Since the EndA(H160G) exhibited greatly increased nuclease activity in imidazole rescue experiments, relative to EndA(H160A), the greater dynamic range available with the glycine substitution allowed for a more precise measurement of small changes in catalytic activity. Therefore, we used this mutant as a ‘background’ for evaluating the effects of additional mutations. These mutants were then tested for levels of nuclease activity in the plasmid conversion assay and the single-radial enzyme diffusion (SRED) assay (42).

To date, attempts to crystallize EndA in the presence of various DNA substrates have been unsuccessful. To determine the residues involved in catalysis and DNA binding outside of the canonical DRGH motif, DNA from another ββα-Metal finger nuclease—VVN from Vibrio vulnificus [PDB ID code 1OUP (43)]—was superimposed into the active site of EndA (Figure 4) based on the ββα-Me motif. The superimposed DNA fits snugly into a shallow cleft on the surface of EndA, immediately outside the active site (Figure 4A). It should be noted that the disordered loop between β-strands 4 and 5 would likely lie in close proximity to the DNA, over the small cavity containing the catalytic center (Figure 4A and B). Upon detailed examination of the superimposed DNA into the EndA(H160A) structure, we mutated several residues in the immediate vicinity of the active site (Arg127/Lys128, His154, Gln186, Gln192, Asn202, Glu205 and Arg209/Lys210) to alanine and evaluated their effects on DNA substrate binding and/or catalysis. We included N191A in our study as a negative control, since that mutation was nearly as deactivating as alanine substitution for His160 (20), which is proposed to act as the general base (44). To elucidate the role of the ‘finger-loop’ in α-helix D (Figure 4C), we created alanine substitution mutants for Glu196/Tyr197 and a deletion mutant where residues Gln194-Thr199 were removed from the construct (ΔLoop). Given the ‘finger-loop’ position proximal to the putative DNA binding site, it is tempting to speculate that this feature could play a role in DNA substrate binding or recognition.

All alanine substitution mutants appeared indistinguishable from the original EndA(H160G) with regards to solubility, stability and behavior during size exclusion chromatography. Plasmid conversion nuclease activity assays with imidazole rescue were performed for each mutant, and evaluated with respect to EndA(H160G). These assays showed that, as expected, the H160G/N191A mutant displayed no detectable nuclease activity assay (Figure 5A and B). Therefore, mutations made on the H160G background, with the aid of imidazole rescue, behave similarly to the same mutations on wildtype background. The H160G/E205A and H160G/R127A/K128A mutations also had no detectable activity. H160G/H154A and H160G/R209A/K210A had minimal activity levels, while H160G/Q186A and H160G/Q192A were less affected. H160G/N202A showed activity levels nearly comparable to the original H160G alone. Surprisingly, the ‘finger-loop’ mutants showed no significant differences from the H160G mutant, though the ΔLoop variant appeared slightly more active. Figure 5.

Nuclease activity assays for EndA variants generated on the background of H160G. (A) Plasmid conversion nuclease activity assay for EndA(H160G) variants generated on the background of H160G. Supercoiled plasmid substrate was incubated with EndA(H160G) variants in the presence or absence of 200 mM imidazole. The supercoiled substrate (sc) is first nicked to create the open circle (oc) species, further nicking activity eventually creates double-strand breaks, linearizing the plasmid. A high level of nuclease activity degrades the linearized substrate to small oligonucleotide fragments, which can be visualized as smears. (B) Bar graph illustrating the relative levels of nuclease activity for the EndA(H160G) variants. The error bars reflect the standard deviation for n = 3 replicates of this experiment. (C) Single radial enzyme diffusion assay for EndA(H160G) variants. A SRED agarose plate containing high molecular weight salmon sperm DNA, MgCl₂ and ethidium bromide. EndA(H160G) variants were spotted onto this plate in equal amounts. Nuclease digestion is visualized by a dark ‘halo’ around each injection site. (D) Halo radii for each EndA(H160G) variant were measured, and are displayed on the bar graph as a ratio of the H160G background. The error bars reflect the standard deviation for n = 3 replications of this experiment.

In order to discover whether loss of nuclease activity could be correlated with DNA substrate binding, an electrophoretic mobility shift assay was performed using all EndA(H160G) variants (Figure 6). Due to its high predicted pI (9.51), native gel electrophoresis of EndA(H160G) alone results in negligible migration of the protein (Figure 6A, lane 2). A complex formed by non-specific binding of the 24 base pair duplex DNA substrate to the protein (Figure 6B) alters the pI, enabling its migration toward the cathode. The DNA/protein complexes formed in this assay appear to be relatively unstable and of low affinity, presenting as a low-molecular weight smear migrating above the unbound DNA (Figure 6B and C). This behavior is consistent with the non-sequence specific nature of EndA interactions with DNA. Mutations which are thought to undermine the catalytic mechanism (H160G/N191A) would be expected to exhibit ‘wildtype’ substrate binding, but no nuclease activity. Interestingly, all variants except for the H160G/R127A/K128A and H160G/R209A/K210A mutants were capable of binding the DNA substrate to an extent similar to that of the ‘background’ H160G mutant (Figure 6C). These results suggest that residues Arg127/Lys128 and Arg209/Lys210 are likely involved in substrate binding. Conversely, the E205A mutation displays no alteration in substrate binding, yet has no detectable nuclease activity. Therefore, Glu205 appears to be crucial for catalysis. Figure 6.

DNA-binding assays for EndA(H160G) variants. Equal amounts of EndA(H160G) proteins were mixed with a 24 base pair duplex DNA substrate under low salt conditions, then separated using an electrophoretic mobility shift assay. (A) Protein-containing species were stained with SYPRO Orange (blue, Sigma-Aldrich). (B) Molecular species containing DNA were stained with ethidium bromide (red) on the same gel. (C) Images showing DNA (red) and protein (blue) were merged using ImageJ (48), highlighting the DNA/protein complexes in purple.