Crystal structure of NucB, a biofilm-degrading endonuclease.

Bacterial biofilms are a complex architecture of cells that grow on moist interfaces, and are held together by a molecular glue of extracellular proteins, sugars and nucleic acids. Biofilms are particularly problematic in human healthcare as they can coat medical implants and are thus a potential source of disease. The enzymatic dispersal of biofilms is increasingly being developed as a new strategy to treat this problem. Here, we have characterized NucB, a biofilm-dispersing nuclease from a marine strain of Bacillus licheniformis, and present its crystal structure together with the biochemistry and a mutational analysis required to confirm its active site. Taken together, these data support the categorization of NucB into a unique subfamily of the ββα metal-dependent non-specific endonucleases. Understanding the structure and function of NucB will facilitate its future development into an anti-biofilm therapeutic agent.

INTRODUCTION

Free-living, motile bacteria can develop into a stationary, multicellular community of cells known as a biofilm, a colony of sessile cells that forms on natural or artificial moist surfaces (1,2). Medical implants and devices can be contaminated by biofilms (3), and dental caries (4), upper respiratory tract (5), ear infections (6) and chronic lung infections in cystic fibrosis patients are caused by biofilms (7,8). Soil-dwelling bacteria are associated with the biofilms of plants in both symbiotic and pathogenic relationships (9). Finally, biofilms are 101–104 times more resistant to antibiotics than free-living bacteria (10,11), and biofilms thus represent a significant world-wide challenge in society, healthcare, agriculture and industry.

A molecular glue, called the extracellular matrix (ECM), holds the biofilm together. The ECM is a relatively impermeable barrier of proteins, carbohydrates and extracellular DNA (eDNA) (1), and common themes and species-dependent differences in ECM formation in bacteria are now beginning to emerge (2). For instance, a subpopulation of Bacillus subtilis cells that are destined to sporulate (the ultimate survival mechanism for this species) are located at the periphery of the biofilm (2,12,13) akin to the fruiting bodies of fungi and myxobacteria. Degradation of the ECM results in biofilm dispersal. Nutrient levels, environmental factors and bacterially-derived small molecule effectors can all trigger signal transduction pathways that result in biofilm dispersal (14,15). Furthermore, secreted proteases (16) and glycoside hydrolases (17) degrade the protein and carbohydrate components of the ECM, respectively (15). eDNA is a critical component of the ECM (18–20) and is required for the initial adhesion phase (21). It was first shown several decades ago that biofilms treated with bovine DNase I had reduced viscosity (22) leading to biofilm dispersal (21). Indeed, cystic fibrosis can be treated with a nebulizer that contains recombinant human DNase I (Dornase Alfa) to reduce the viscosity of the patient's sputum to promote its clearance. There is increasing evidence that secreted nucleases play important roles in biofilm formation, dispersal and remodelling in many bacterial phyla, including major pathogens of humans such as Pseudomonas aeruginosa (7,8), Vibrio cholera (23) and Staphylococcus aureus (19).

An unidentified protein purified from B. subtilis cell lysates had been shown previously to degrade DNA in a divalent cation-dependent manner (24). This enzyme was found to be expressed in late stage II of sporulation (25) and was subsequently designated NucB (26). When a biofilm dispersing supernatant from a marine isolate of Bacillus licheniformis was analysed, one of the active compounds was found to be a NucB orthologue (BlNucB) (27). This enzyme was able to disperse a broad range of mono- and mixed-species biofilms by degrading eDNA and may be useful in combatting a number of biofilm-related problems (3–5,27,28). However, NucB orthologues share no sequence homology to any other protein family and thus a molecular understanding of their biochemical properties is completely lacking. To shed light on the biofilm-dispersing properties of NucB, we present here an analysis of the mode of action of this novel nuclease, its crystal structure and a mutational analysis that confirms the active site. Together, these results classify NucB as the founding member of a novel subgroup of the non-specific His-Me finger endonuclease superfamily.

MATERIALS AND METHODS

Strains and plasmids

Unless otherwise stated all chemicals and reagents were obtained from Sigma-Aldrich. Nutrient agar (NA; Oxoid) and Luria-Bertani (LB) medium was used for routine selection and maintenance of B. subtilis and Escherichia coli strains (Supplementary Table S1). Standard techniques were used for strain construction (29). Transformation of competent B. subtilis cells with plasmid DNA (Supplementary Table S2) was performed using an optimized two-step starvation procedure as described previously (30,31). Transformation of chemically competent E. coli cells was performed as described (32). General manipulation of DNA was performed using standard procedures (33). To induce NucB expression, B. subtilis was grown in Schaeffer's medium (Nutrient broth (Difco), 1 mM MgSO4, 1 mg/ml KCl, 1 mM CaCl2, 130 μM MnSO4). Supplements were added as required: 20 μg/ml tryptophan, 5 μg/ml chloramphenicol, 2 μg/ml kanamycin, 10 μg/ml zeocin, 100 μg/ml ampicillin. Site-directed mutagenesis of B. subtilis nucB (BsnucB) was performed using primers listed in Supplementary Table S3. All plasmids and strains were verified by sequencing.

Production and purification of B. licheniformis NucB

Recombinant B. licheniformis NucB (BlNucB) proteins were prepared by expression in B. subtilis NZ8900 (27) and purified as described previously (4).

Differential scanning calorimetry (DSC)

BlNucB at a concentration of 34.3 μM in a buffer of 50 mM Tris–HCl (pH 8.0), 1 mM DTT was subjected to thermal unfolding in a MicroCal VP-DSC instrument. The scan range was 25–80°C, with a scan rate of 90°C/h. The raw data were deconvoluted using the non-two state model within MicroCal Origin (www.originlab.com).

Crystallization, structure determination and refinement

Purified BlNucB was concentrated by ultrafiltration to a concentration of 50 mg/ml for crystallization at 20°C by sitting-drop vapour diffusion using a Mosquito (TTP Labtech) liquid-handling robot and a series of commercially-available crystallization screens. The initial BlNucB crystals, which were used for sulphur SAD phasing, crystallized in 200 mM sodium nitrate, 100 mM propionic acid/cacodylate/bis-tris propane buffer (pH 7.5), 20% (w/v) polyethylene glycol (PEG) 3350. BlNucB crystals subsequently used for high resolution data collection were grown from 100 mM MES–NaOH (pH 6.5), 12% (w/v) PEG 20000. All samples were cryo-protected by supplementing the crystallization mother liquor with 20% PEG 400. The sulphur SAD diffraction data were collected on beamline I24 of the Diamond Light Source synchrotron at a wavelength of 1.907 Å to a maximum resolution of 2.26 Å; 999.9° from a single crystal rotated around φ were collected. Higher resolution data were collected on beamline I04 of the Diamond Light Source synchrotron at a wavelength of 0.9795 Å to a maximum resolution of 1.35 Å.

The data were integrated and scaled in XDS (34). Space group determination was confirmed with POINTLESS (35). The crystallographic phases for BlNucB were determined by anomalous scattering from the sulphur atoms present in the protein chain with HKL2MAP (36) and the SHELXC/D/E suite (37), which correctly positioned four sulphurs, corresponding to the S atoms found in the two cysteines and two methionines in the sequence of mature BlNucB. The structure was built automatically in ARP_wARP (38) and Buccaneer (39). Five percent of the observations were randomly selected for the Rfree set, and model building and refinement cycles in Coot (40) and REFMAC5 (41) were interspersed until refinement reached convergence. The model was validated using Coot (40) and MolProbity (42). The data collection and model refinement statistics are summarized in Supplementary Table S4.

Nuclease activity assays ex vivo

B. subtilis NucB (BsNucB) proteins were natively expressed during sporulation. Cell cultures were inoculated into 2 ml of Schaeffer's media for 6 h at 37°C and these starter cultures were then diluted (1:100) into fresh Schaeffer's media and allowed to grow for 40 h at 37°C. Cells were pelleted and the culture supernatants were passed through a sterile 0.2 μm filter (Millipore) before analysis.

To determine the total nuclease activity present within each supernatant, B. subtilis genomic DNA (6 ng/μl; Qiagen DNeasy Kit) was combined with 10 μl of supernatants and incubated for 3 h at 37°C. The DNA samples were mixed with glycerol (5% final) and stained with 2X SYBR Gold (Thermo Fisher Scientific). The genomic DNA was separated using a 1% agarose gel run in TBE buffer (45 mM Tris–borate, 1 mM EDTA), and nucleic acid was visualized with a UV transilluminator.

Nuclease activity assays in vitro

Purified BlNucB was used to assess its nuclease activity in vitro. First, high molecular weight calf thymus (or salmon sperm) DNA was treated with BlNucB; high molecular weight DNA is not soluble in 2% perchloric acid whereas products of the nuclease reaction <500 bp in length are acid-soluble, the absorption of which can be measured at a wavelength of 260 nm in an end-point assay (43). 5 ng of BlNucB (final concentration 1.68 nM) was incubated at 37°C with 125 μg of calf thymus or salmon sperm DNA in a reaction buffer of 50 mM Tris–HCl (pH 8.0), 5 mM MnSO4 in a total reaction volume of 250 μl. The reaction was stopped after 60 min by mixing with an equal volume of ice-cold 4% (v/v) perchloric acid. The mixture was left to stand on ice for 40 min before the insoluble material (protein, high molecular weight DNA) was pelleted by centrifugation in a benchtop microfuge at 4°C. The supernatant was diluted 4-fold with 50 mM Tris–HCl (pH 8.0), 5 mM MnSO4 before the amount of low molecular weight DNA generated by BlNucB was measured in a spectrophotometer at 260 nm using 1 cm pathlength quartz cuvettes. All reactions were performed in triplicate from single- to triple-preparations of proteins.

Second, supercoiled, relaxed and linearized pBR322 DNA were used as nuclease substrates. Linearized pBR322, exploiting the sole BamHI restriction site in this plasmid, was generated by restriction using BamHI (ThermoFisher) according to the manufacturer's instructions. pBR322 samples at final concentrations of 200 ng/μl were mixed with 0.33 ng of BlNucB (final concentration 2.5 nM) in a reaction buffer of 10 mM Tris–HCl (pH 7.5), 5 mM MgCl2, 100 mM NaCl in a total reaction volume of 10 μl. Though it has been reported previously that maximal nuclease activity of the purified nuclease, presumed to be BsNucB, was obtained in the presence of Mn2+ ions (24), we chose to use Mg2+ ions, dilute enzyme concentrations and to conduct the reactions at room temperature to potentially observe transient intermediates in the reaction trajectory. Samples were taken at time intervals and the nuclease reaction was stopped by the addition of EDTA to 10 mM and by heating to 95°C before the reaction products were separated by 0.8% agarose gel electrophoresis and visualized by GelRed (Biotium, USA) in-gel staining and UV transillumination.

Finally, BlNucB was used against fluorescently labelled 30mer oligodeoxynucleotides (ATDbio, Southampton, UK) that incorporated two successive phosphorothioate linkages at either the 5΄-, 3΄- or both termini, with deoxythymidylyl-fluorescein isothiocyanate (dT-FITC) and/or deoxythymidylyl-tetramethyrhodamine (dT-TAMRA) incorporated at position 3 and/or 28, respectively. Stock solutions of single stranded oligodeoxynucleotides at 250 nM were prepared by re-solublizing the lyophilized DNA pellet in a hybridization buffer of 25 mM HEPES–KOH (pH 7.5), 100 mM KCl, 1 mM EDTA. Oligodeoxynucleotide stock solutions (100 μl) were mixed with an equal volume of the unlabelled complementary strand at 350 nM, heated to 85°C and then left to cool slowly overnight to yield double stranded oligodeoxynucleotides.

Single- and double-stranded fluorescent oligodeoxynucleotides were digested in a 25 μl reaction volume comprising 25 nM DNA and 100 ng of BlNucB in a buffer of 10 mM Tris–HCl (pH 7.5), 5 mM MgCl2, 100 mM NaCl. At defined timepoints samples were taken and the reaction stopped with an equal volume of a quench solution at pH 7.5 comprising 95% formamide, 5% H2O, 10 mM EDTA, and 1 μl of 10 μM unlabelled competitor strand DNA (which has the same sequence as the fluorescent oligodeoxynucleotide) to prevent re-hybridization of the fluorescent products (44). The reaction products were heated to 90°C prior to resolution on a 17% denaturing agarose gel containing 8 M urea; the gel was electrophoresed for 2.5 h at 4 W. Reaction products were visualized using a Typhoon FLA9500 scanner using excitation and emission wavelengths of 542 and 568 nm (TAMRA) and 494 and 520 nm (FITC), respectively.

RESULTS

Nuclease properties of BlNucB

To investigate the activity of BlNucB, its gene was cloned and the recombinant protein was expressed and purified using B. subtilis as host. In nuclease assays, 5 ng of BlNucB was capable of digesting 125 μg of calf thymus DNA almost completely in 60 min at 37°C (Figure 1A). Over a 15–60 min time frame, the length of the reaction products was reduced to a minimum, shorter than the 500 bp marker in the HindIII-digested' phage λ DNA ladder. A semi-quantitative analysis of the degradation of calf thymus DNA by three independent preparations of BlNucB revealed that 0.25 (±0.02; n = 9) OD units of acid soluble product was obtained per ng of BlNucB per hour in buffer containing 5 mM Mn2+ ions. Given that double stranded DNA at a concentration of 50 μg/ml has an A260 of 1 OD units, 1 ng of BlNucB can therefore produce 12.5 μg of low molecular weight DNA per hour.

Figure 1.

The nuclease activity of BlNucB against high molecular weight DNA. (A) BlNucB was incubated with (+) and without (–) calf thymus DNA for 15, 30 and 60 min at 37°C for 1 h. Samples of the digestion products were separated by agarose (0.8% w/v) gel electrophoresis and the DNA made visible by staining with ethidium bromide. HindIII-digested λ DNA (λ) is included as a marker. (B) BlNucB was incubated with a mixture of nicked (N) and supercoiled (S) pBR322 DNA at room temperature for 15 s, 30 s, 1 min, 2 min, 5 min, 10 min, 20 min and 30 min (labeled ¼ to 30 min). 10 mM EDTA (E) was added to the reaction mix prior to the addition of BlNucB for 30 min. Undigested (–) and BamHI-linearized (L) pBR322, along with HindIII-digested λ DNA (λ) are included as markers. The digestion products were separated by agarose (0.8% w/v) gel electrophoresis and the DNA made visible by staining with GelRed.

However, the calf thymus DNA used as the substrate in this experiment is a heterogeneous mix of highly polymerized single and double stranded DNA, and it is thus not possible to determine whether BlNucB acts as an endo- or an exonuclease. To address this question, we determined the effect of BlNucB on supercoiled pBR322 DNA in the presence of Mg2+ ions to potentially observe transient intermediates in the reaction trajectory. Since supercoiled DNA does not have free 5′ or 3′ termini, if BlNucB degrades this substrate it must have endonuclease activity. In the first few time points, the bands corresponding to the nicked and supercoiled topoisomers of pBR322 began to disappear such that by 1 min almost all of the supercoiled DNA had either been converted to the nicked form, or had been digested completely (Figure 1B). The substrates were degraded readily in the presence of magnesium, and degradation was inhibited completely in the presence of 10 mM EDTA (Figure 1B). By 30 min, almost all the DNA had been completely degraded. Furthermore, the ∼4.4 kb linear form of the plasmid was not observed in any of the time points, indicating that the DNA was cut one strand at a time. A relatively nuclease-resistant, ∼900 bp product that persisted beyond 20 min, but which was also completely degraded by 30 min, was observed. The nature of this product and its relative stability are unknown. Restriction linearized pBR322 DNA was also degraded completely by BlNucB (data not shown). There was no evidence of the formation of a DNA ladder in any experiment, implying that BlNucB does not have exonuclease activity. The reaction products observed are entirely in keeping with those of a monomeric, metal-dependent, non-specific endonuclease.

To confirm that BlNucB has solely an endonuclease function, a series of fluorescent oligodeoxynucleotides were synthesized in which either the 5′ or 3′ terminus, or both, was blocked from hydrolysis by the incorporation of an adjacent pair of non-hydrolysable phosphorothioates. If BlNucB had 5′ exonuclease activity, it would not degrade oligodeoxynucleotides substrates blocked at the 5′ end by phosphorothioates, whereas it would remove one base at a time and produce a ladder in the absence of phosphorothioates at the 5′ end. By similar logic, oligodeoxynucleotide substrates blocked at the 3′ end by phosphorothioates would not be a substrate for a 3′ exonuclease. A substrate blocked at both ends could not be hydrolysed by any exonuclease, but it would be hydrolysed as efficiently as a non-blocked substrate by an endonuclease. The sequences of the 30mer oligodeoxynucleotides used in these experiments can be found in Figure 2A.

Figure 2.

The nuclease activity of BlNucB against oligodeoxynucleotide substrates. (A) Sequence of the oligodeoxynucleotides used, with the position of the phosphorothioate linkages indicated with the blue asterisks. (B) BlNucB was incubated with fluorescent double stranded oligodeoxynucleotide DNA at room temperature for 1 and 15 min, before separation by 17% (polyacrylamide) denaturing gel electrophoresis and visualization on a Typhoon scanner (GE Healthcare). The non-fluorescent complementary oligodeoxynucleotide used in all reactions is sequence 7. Undigested TAMRA-labelled 30mer and FITC-labelled 18mer oligodeoxynucleotides are used as size controls, and the schematic above the gel indicates which oligodeoxynucleotide was used; phosphorothioates are indicated by the blue circles.

Representative scans of the in-gel fluorescence from the 3′ TAMRA and/or 5′ FITC fluorescent labels are shown in Figure 2B. There were two timepoints per reaction (1 and 15 min) and a FITC-labelled 18mer was included as a marker. Irrespective of the presence or absence of phosphorothioate linkages at the 5′ terminus, there was little to distinguish between the TAMRA-labelled band patterns in any experiment. Similarly, the fluorescence banding patterns from the FITC label was unaffected by the presence of phosphorothioates at the 3′ terminus. No matter how the experiment was conducted, very few products were observed that were larger than 18 bp, and there were no ladders consistent with the removal of a single base at a time from substrates with free termini. Therefore, the DNA was cleaved endonucleolytically, in approximately the centre of the double-stranded 30mer substrate, to produce products in the ∼10mer to ∼20mer range. Close inspection of the resolved fragments from the dual-labelled probe indicates that the oligodeoxynucleotide substrate was cleaved asymmetrically to produce distinct fragments containing either TAMRA- or FITC-labelled (green and red) products. Control experiments with well-characterized exonucleases confirmed that the phosphorothioates protected the termini from degradation (data not shown), and Micrococcal endonuclease also cleaved the dual-labelled probe asymmetrically (Supplementary Figure S1A). In keeping with plasmid DNA as the substrate, the presence of EDTA abrogated endonucleolysis completely (data not shown). When these experiments were repeated with single-stranded DNA substrates, BlNucB cleaved these oligodeoxynucleotides poorly; about half the substrate was cleaved in the first minute and much less cleavage was seen between 1 and 25 min, even when >7-fold more enzyme was included in this reaction (Supplementary Figure S1B). BlNucB cleaved single stranded 30 mer deoxyoligonuclotides at approximately position 20. When the secondary structure of the single stranded substrates were calculated, position 20 tended to occur at the boundary between predicted single and double stranded regions of the DNA. It has already been shown that BlNucB had no detectable activity against total RNA purified from the periodontitis-associated oral bacterium Veillonella parvula in comparison to the positive control, RNase A (4). Therefore, BlNucB is a Mn2+/Mg2+-dependent, non-specific endonuclease that can cleave both single- and double-stranded substrates, but with a preference for double stranded DNA.

Crystal structure of BlNucB

To understand the molecular basis of its nuclease activity, the structure of BlNucB was solved by single wavelength anomalous dispersion (SAD) from the intrinsic sulphur atoms in the protein and refined to a resolution of 1.35 Å (Supplementary Figure S2). Of the 8594 PDB entries to date that are described under ‘diffraction protocol’ as single wavelength experiments, fewer than 1% have been solved by S-SAD. Careful analysis of highly redundant (58-fold), moderate resolution (2.3 Å), low energy (6.3 keV) diffraction data, collected by rotating a single crystal around one axis, resulted in a straightforward structure solution of a protein with one sulphur atom per 27 residues and a Bijvoet diffraction ratio of ∼1%. Overall, BlNucB is a single, compact domain comprising five β-strands and three α-helices (Figure 3A) that forms a rough triangular pyramid with an approximate diameter of 25 Å. The core of the protein is formed by the five β-strands in a bipartite β-sheet, with β-strands with topological order 1, 5 and 2 separated from β-strands 3 and 4 by the β3-β4 loop that breaks the continuity of the β-sheet (Figure 3B). The β3-β4 loop is stabilized in part by the formation of a disulphide bond between β2 residue Cys66 and the β3-β4 loop residue Cys100 (Supplementary Figure S3). Whilst this disulphide would appear to stabilize the local structure, its reduction with 10 mM DTT had no significant effect on the degradation of calf thymus DNA by BlNucB (data not shown). The base of the pyramid is formed by α-helices 2 and 3 as well as the loop connecting α-helix 2 to β-strand 3, while the peak of the pyramid is formed by the C-terminus of α-helix 1. In keeping with the observed endonuclease activity of BlNucB, the solvent-accessible surface of BlNucB does not reveal a molecular ‘wall’ or ‘tape-measure’ (45) to confer exonuclease activity. Instead, the flat base of the pyramid contains a 15 Å deep, 10 Å wide, 20 Å long concave depression that is formed mostly by conserved amino acids (Figure 3C). The base of the depression is predominantly negatively-charged, whereas the lips of the cavity are mostly positively-charged (Figure 3D). The surface depression is of sufficient size to accommodate a single strand of DNA, presenting the scissile phosphodiester bond to the catalytic apparatus, whilst the lips of the cavity can interact with the DNA phosphate backbone.

Figure 3.

The crystal structure of BlNucB. (A) Overview of the structure of BlNucB as a cartoon, colour-ramped from blue to red, from N- to C-terminus. The secondary structure elements are labelled. (B) Topology diagram of BlNucB; α-helices are shown as red cylinders and β-strands as pink arrows. The ββα motif that defines the nuclease superfamily to which BlNucB belongs is highlighted by the grey box. (C) The surface of BlNucB coloured by sequence conservation (68), drawn in the same view as panel A. Note the concave depression in approximately the middle of the structure the mouth of which is mostly coloured deep purple, which indicates a high degree of sequence conservation. Non-conserved regions are coloured blue, with a purple-blue gradient between the two extremes. (D) Electrostatic potential mapped to the surface of BlNucB shown in the same view as above; positively-charged regions are coloured blue, negatively-charged red and uncharged regions are white. Note how the mouth of the depression is positively-charged whereas the bottom of the cavity is negatively-charged.

Though NucB orthologues appear to form a unique protein family by sequence, searches of the PDB with the atomic co-ordinates of BlNucB revealed that it does have some structural homology to the Serratia marcescens endonuclease (Smendo; PDBid: 1SMN; 46), with an RMSD of 3.6 Å on 77 superimposed Cαs (Figure 4A). At first glance, the homology between Smendo and BlNucB is not readily apparent: Smendo is approximately twice the size of BlNucB (245 residues vs 109), but the protein core and functionally essential structural elements are maintained (see below; Figure 4A). Smendo is a member of the His-Me finger family of endonucleases, where Me stands for a divalent cation and where His refers to the general base in the reaction (47). The magnesium ion in Smendo is co-ordinated directly by only one protein atom, the sidechain amide oxygen of Asn119, and the rest of the magnesium's hydration shell is filled by water molecules, one of which is co-ordinated by the mainchain amide and sidechain Nδ1 nitrogen from the His89 general base (Figure 4B). This arrangement is mostly, but not completely, maintained in other His-Me finger endonucleases; in some instances, including the Vibrio vulnificus endonuclease (Vvnendo), there is a direct contact to the cation from an acidic residue, Glu79, located immediately prior to the general base, His80 (47,48). Smendo has glycine (Gly88) adjacent to the His89 general base, explaining why the sole protein contact to the cation in this protein comes from Asn119 (46). There is no metal ion bound in the apo structure of BlNucB reported here and, despite repeated attempts, we were unable to obtain diffracting crystals containing either bound divalent cations or oligodeoxynucleotides. As with other members of the His-Me finger family, BlNucB retains the metal-binding pocket (Figure 4B), including the metal-chelator Asn117 and, as in Vvnendo (48), BlNucB encodes an acid (Asp93) immediately before Glu94 and located so that with minimal changes to the sidechain chi angles it can be positioned to co-ordinate directly to a bound divalent cation. The His-Me finger family is also known as the ββα family, which relates to the maintenance of centrally-positioned secondary structural elements that harbour the essential catalytic residues (47). The ββα nomenclature is adopted here in relation to BlNucB because it lacks the conserved histidine general base of the His-Me finger family, which has been replaced in BlNucB by Glu94, and thus NucB is the founding member of a novel subgroup of the His-Me finger family of non-specific endonucleases.

Figure 4.

BlNucB belongs to the ββα family of non-specific endonucleases. (A) The structure of BlNucB (cyan) indicates that it belongs to a novel subgroup of the ββα (or His-Me finger) family of non-specific endonucleases, as exemplified by Smendo (magenta; PDBid: 1SMN; 45). The ββα motif that defines the active site is labelled in each structure by the corresponding secondary structure element (top row). Other than this motif, the structural homology between BlNucB and Smendo is not extensive, but the surface representation in the same view (bottom row) indicates that they share the same surface depression that houses the active site. (B) A superposition of the active sites of Smendo (magenta) vs BlNucB (cyan), showing the conservation of the metal binding site in the latter. Asn119 in Smendo forms the only direct proteinaceous contact to the bound Mg2+ ion (pale pink sphere), and the water shell around the metal ion is stabilized by interactions to Gln114 and the general base, His89. The structural equivalents in BlNucB are Asn117, Asp93 and Glu94, respectively. Residues are coloured by structure, cyan for BlNucB and magenta for Smendo.

The NucB active site

In order to confirm the importance of likely active centre residues of BlNucB, the nuclease activity of several variants (Figure 5A, B) was analyzed ex vivo using B. subtilis, in which we have an established genetic system. The amino acid sequence homology between mature NucB orthologues from B. subtilis and B. licheniformis is 77% (Supplementary Figure S3), and all the metal-binding and proposed catalytic residues are conserved, indicating that functional data from one system will be directly relevant to structural and biochemical data in the other, and vice versa. Wild-type BsnucB was replaced by mutant alleles whilst retaining the native expression system and secretion signal. Cultures were grown into stationary phase to induce BsnucB expression and supernatants containing the secreted BsNucB proteins were collected. To assess overall nuclease activity the supernatants were incubated with purified chromosomal DNA. Subsequently the genomic DNA fragments were stained with a fluorescent dye, resolved using agarose gel electrophoresis, and visualized by UV transillumination (Figure 5A). Whilst the supernatant from the wild-type strain led to the digestion of the DNA substrate, the supernatants from strains expressing BsNucBH47A, (equivalent to BlNucBH53), BsNucBD87A and BsNucBD87N (BlNucBD93), BsNucBE88A (BlNucBE94), and BsNucBN111A (BlNucBN117), in addition to the ΔBsnucB mutant, showed no significant nuclease activity. In BlNucB, Asp93 is likely to stabilize the bound magnesium ion that is essential for nuclease activity most likely by co-ordinating directly to it. Alternatively, Asp93 might bind indirectly to the magnesium via a bridging water molecule, but this is less likely because of steric clashes that would result between Asp93 and His53 by introducing such a water molecule. Replacement of this aspartate with alanine will clearly disrupt the magnesium co-ordination potential of NucB, most likely a direct contact given the complete loss of nuclease activity associated with mutations at this allele and the steric clashes that would likely occur by the introduction of a bridging water molecule. The importance of the correct positioning of the ion for catalysis is underlined by the loss of activity associated with the BsNucBH47A allele, since the BlNucBH53 equivalent maintains the position of Asp93 by a hydrogen bond network in a manner that is reminiscent of a catalytic triad (Figure 5B). The third member of this triad, BlNucBS108, is positioned on the other side of His53 and mutation of the non-conserved equivalent, BsNucBD102 also has a negative impact on BsNucB activity. To confirm the importance of Asp93 for activity, the asparagine variant at this position was introduced to BlNucB by site directed mutagenesis and the protein purified by the same protocol as for the wildtype protein. In keeping with these data and the importance of Asp93 for co-ordinating the catalytically-essential metal ion, the BlNucBD93N variant had less than 1% of the wild type activity in semi-quantitative assays based on the degradation of calf thymus DNA (Supplementary Figure S4A). Purified BlNucB preparations showed a single protein band when increasing loadings were analysed by SDS PAGE and spectroscopic analysis confirmed the absence of nucleic acid. (4). Peptide mass fingerprinting of a trypsin digest of the single protein band excised from the SDS PAGE gel showed that the only protein present was BlNucB (4). Furthermore, purified BlNucB preparations had no detectable RNase activity against total bacterial RNA in comparison to RNase A and no detectable protease activity against BSA in comparison to proteinase K (4). These observations, combined with the quantitative loss of over 99% of the nuclease activity of the BlNucBD93N variant compared to the wildtype BlNucB when purified under the same conditions (Supplementary Figure S4A), argues strongly that the nuclease activity seen in Figures 1 and 2 is due to the wildtype BlNucB present and not a co-purifying contaminant.

Figure 5.

NucB nuclease activity ex vivo. (A) BsNucB mutants are defective in nuclease activity. Cells were grown in Schaeffers media to induce sporulation at 37°C. Supernatant were then mixed with chromosomal DNA in the ratio 1:3 for 3 and 6 h at 37°C before DNA was visualized using 1% agarose gel stained with SYBR Gold. Wild-type (HM715), ΔnucB (AK361), nucBH47A (AK453), nucBD87A (AK447), nucBD87N (AK449), nucBD88A (HM1766) nucBD102A (AK451), nucBN111A (HM1767). (B) The active site of BlNucB is shown after rotation of the view in Figure 4A around a horizontal axis of ∼90°. The equivalent residues in BlNucB that were mutated in BsNucB (Figure 5A) are labelled in cyan; their counterparts in BsNucB are additionally labelled in black. Though BsNucB His47 is removed from the active site, mutation of it and immediate neighbours presumably affects nuclease activity because of the loss of a stabilizing network of hydrogen bonds involving Asp87 and Asp102; the equivalents in BlNucB are His53, Asp93 and Ser108.

Thermal stability of BlNucB

The NucB family is exemplified by its small size; the mature enzyme is only 110 amino acids in length, approximately half the size of the crystallized secreted endonucleases from Anabaena (e.g. PDBid: 1ZM8), Streptococcus agalactiae (e.g. PDBid: 4QGO) and Serratia marcescens (e.g. PDBid: 1SMN), and half the size of the recombinant human DNase I used to treat cystic fibrosis. NucB is slightly bigger than the catalytic domain (∼95 residues) of the GIY-YIG endonuclease family, but unlike NucB, GIY-YIG family members are modular, with separate DNA-binding and nuclease domains though the isolated nuclease domain retains the ability to bind to and degrade DNA non-specifically (47). In the context of the potential biotechnological exploitation of NucB, the thermal stability of BlNucB was investigated by DSC. In the presence of 1 mM DTT, BlNucB had a single unfolding event with a Tm of 57.4 (±0.1)°C (Supplementary Figure S4B). When fully unfolded protein was allowed to cool passively to 25°C, and then subjected to a second round of thermal unfolding, a Tm of 58.7 (±0.1)°C was obtained (Supplementary Figure S4B). The calculated enthalpy of the first unfolding event was 71 (±10.4) kcal mol−1, whereas for the second unfolding event the enthalpy was 37 (±5.2) kcal mol−1, indicating that 52% of the BlNucB sample refolded spontaneously when cooled passively to 25°C. Semi-quantitative nuclease assays using BlNucB that had been subjected to thermal unfolding and then refolded passively showed that the refolded protein had regained approximately 36% of its activity in comparison to unheated samples (Supplementary Figure S4C). Hence, native BlNucB can be thermally unfolded in the presence of reducing agents, refold spontaneously and regain significant nuclease activity, suggesting that forced evolution of BlNucB could improve these biotechnologically desirable characteristics even further.

DISCUSSION

The BlNucB protein sequence does not identify other nuclease families when used to query the BLAST non-redundant sequence database. Although some de novo structure predictions could recapitulate the overall fold of BlNucB, the models were insufficiently accurate to identify the active site or the nuclease family to which BlNucB belongs (49), and molecular replacement using these models was unsuccessful. The structure of BlNucB was instead solved by S-SAD without recourse to sophisticated data collection strategies. Perhaps, as advocated elsewhere (50), S-SAD has greater latent potential for solving the crystallographic phase problem than appears to be generally appreciated.

Consequent structure-based searches of the PDB revealed that BlNucB belongs to the divergent His-Me finger family of endonucleases (47), despite sharing less than 12% sequence identity with its closest structural neighbour, the endonuclease from Serratia marcescens, Smendo. Structures of several representatives of the His-Me finger nucleases have been solved (48,51–54), which are involved in diverse cellular functions (47) including, for example, secreted toxins (51,52), protection from transformation by extracellular DNA (48), Holliday junction resolution (53), and prophage induction (54). NucB, on the other hand, is necessary for the degradation of eDNA for the dispersal of bacterial biofilms (27). Most His-Me finger nucleases cleave DNA non-specifically (47), and some have been shown to also work on single-stranded DNA and RNA (47). BlNucB is a non-specific endonuclease, which can hydrolyse both single- and double-stranded DNA substrates, as determined by its ability to completely hydrolyse supercoiled plasmid DNA substrates that have no free 3′ or 5′ termini (Figure 1B), and to degrade short oligonucleotides whose 5′ and 3′ termini are blocked because they were synthesized with non-hydrolysable phosphorothioates (Figure 2A, B). Since both single- and double-stranded DNA play roles in biofilm formation (55) it is entirely appropriate that NucB has evolved to disperse biofilms with precisely these properties.

His-Me finger nucleases generally utilize a single metal ion as a Lewis acid and a histidine general base for hydrolysis of the phosphodiester DNA backbone, and are characterized by a common ββα motif (47). BlNucB diverges from this general description because glutamate replaces the histidine, and this glutamate is essential for biological activity (Figure 5A); therefore BlNucB is the founding member of a novel subfamily of the His-Me finger nuclease superfamily. The histidine general base in colicin E7 (His545) and I-PpoI (His98) has each been mutated to glutamate, but the resultant enzymes lose biological activity almost entirely (56,57), indicating that other facets of the NucB active site must compensate for the absence of the histidine, but it is not immediately apparent from the BlNucB structure what this facet could be. The activity of BsNucB has already been measured as a function of pH and appears to show a bell-shaped pH dependency with a pH optimum of 7.5 (24). The apparent bell-shaped curve could suggest the involvement of two amino acids in catalysis with pKas above and below neutral. Presumably one of these amino acids is Glu94, but it is not readily apparent from the structure of BlNucB what the other ionizable group could be. Alternatively, the rapid drop-off in activity seen at pH 8 and above might reflect the precipitation of Mn(OH)2 in mildly alkaline conditions with a concomitant reduction in enzyme activity—as observed previously for Anabaena NucA (58). The drop-off in enzyme activity with increase in pH may also reflect the partial destruction of the disulphide by base-catalysed β-elimination, though this seems unlikely given the retention of enzyme activity in 10 mM DTT, sufficient to reduce the disulphide. The ββα motif contains a central pair of anti-parallel β-strands flanked by a conserved distal α-helix that sits in the major groove of DNA; the loop connecting the two β-strands is variable in length and conformation. The histidine general base (46,47,59) is located in the middle of the first of the two β-strands of the ββα motif, and does not interact directly with the bound nearby catalytic cation. Instead, the cation is co-ordinated directly usually by an asparagine found towards the N-terminus of the conserved distal α-helix, and this is the only direct contact between protein and cation in Smendo (46) because a glycine is found at the structurally equivalent position to the glutamate and aspartate found in Vvnendo (48) and NucB, respectively. Histidine is required to co-ordinate the cation in zinc-dependent His-Me finger nucleases like ColE7 (51) and ColE9 (52), whereas the magnesium- and manganese-dependent enzymes require asparagine or aspartate (e.g. 43,48,53,60). In the absence of a structure of BlNucB with a bound cation, and based on the structural similarity of BlNucB to magnesium-dependent His-Me finger nucleases, the co-ordination sphere of the bound metal ion is likely to be fulfilled by the sidechain amide oxygen of Asn117, the sidechain carboxylate of Asp93 and four water molecules, some which could be stabilized by their contacts to the sidechains oxygens of Asp93 and Glu94, and mainchain atoms from Asp93, Glu94 and Asn117 (Figure 4B).

The similarity of the catalytic apparatus of BlNucB to His-Me finger-type endonucleases is underlined by the structural superposition of the core ββα motif of BlNucB to that found in the structure of, for example, the Vvn endonuclease (48). The rmsd of the matched motifs is 0.7 Å, and this superposition places all the catalytic machinery of BlNucB in equivalent positions to those found in Vvn, and the metal ion and its co-ordinating waters that is present in the structure of Vvn, but which are absent from the structure of BlNucB, into positions that are consistent with roles in catalysis in BlNucB. Furthermore, the key amino acids (underlined) for direct magnesium co-ordination are highly conserved in NucB sequences in DRDE and DNRG motifs (Supplementary Figure S5) on the second β-strand and the α-helix, respectively, of the ββα sub-structure. The importance of these sequence motifs to metal-binding and catalysis is emphasized by the observation that mutation of individual amino acids in these motifs in BsNucB (Asp93 and Glu94 from motif 1, and Asn117 from motif 2) abrogates enzyme activity completely (Figure 5A).

In common with most other non-specific nucleases, including human DNase I (61), it is likely that NucB interacts with its double stranded substrates in the DNA minor groove (Figure 6). Base-specific interactions tend to require access to the major groove, which is not necessary for non-specific DNA interactions (62). For instance, there are only two base specific contacts in the structure of the non-specific Vvn endonuclease in the presence of DNA (48). Many of the amino acids used by Vvn to interact with DNA are conserved spatially in BlNucB and superimpose closely (Figure 6). For instance, the phosphate at –1 in the Vvn:DNA complex (48) is contacted by the Nζ atom of Lys28, the equivalent residue in BlNucB is Arg92. The sidechain of Asn127 of Vvn interacts with the scissile phosphate and its O5′ oxygen; Asn117 is the equivalent residue in BlNucB. The scissile phosphate in Vvn is also contacted by the guanidinium group of Arg99, which overlaps with that of BlNucB Arg77 despite the 8 Å separation of their respective Cα atoms. These arginine residues probably play critical roles in their respective proteins by stabilizing the developing negative charge on the phosphate in the transition state.

Figure 6.

A model of the interaction of BlNucB with DNA. A model for the interaction of BlNucB (cyan) with post-cleavage DNA (orange worm) based on the shared structural feature of the ββα motif in Vvn endonuclease (silver; PDBid: 1OUP; 52). Key amino acids for Vvn DNA recognition are shown as sticks and are labelled in silver K28, E79, R99 and N127, with cyan labelled BlNucB structural equivalents R92, D93, R77 and N117. Vvn K100 does not seem to have a structural equivalent in BlNucB and the interaction of Vvn residues W85 and W94 with the DNA backbone remote from the active site is likely performed by BlNucB R70. The scissile phosphate is labelled with a red ‘0’ and the –1 phosphate is also indicated.

B. subtilis encodes another glutamate-dependent variant of the ββα endonuclease family with 78% sequence identity to NucB; the gene for this nuclease, NucA, is found adjacent on the chromosome to that of the presumed nuclease inhibitor, Nin. When lysates were mixed from B. subtilis strains that lacked either nin, nucA or nucB, Nin was capable of inhibiting both NucA and NucB in nuclease assays (63). Does the structure of BlNucB provide insight into nuclease inhibition by Nin? A crystal structure of NucA in complex with its cognate inhibitor, NuiA (Supplementary Figure S6A), has been solved from the cyanobacterium, Anabaena (61). Though Anabaena NucA and BlNucB share no significant sequence homology, Anabaena NucA is also a member of the His-Me finger family of nucleases. Anabaena NuiA is a competitive inhibitor of Anabaena NucA and binds to the nuclease in a site that overlaps with that of the substrate DNA (60). However, Nin and NuiA share no meaningful sequence homology and inspection of their structures (the unpublished structure of Nin has been determined by a structural genomics consortium, PDBid: 4MQD) reveals that they also share little structural homology (Supplementary Figure S6B). It thus remains to be seen how Nin exerts its inhibitory effect on NucA and/or NucB.

Finally, various properties of BlNucB are pertinent to its biotechnological potential in biofilm dispersal (28). BlNucB is a potent, non-specific endonuclease that can degrade double- and single-stranded DNA and topologically complex molecules, such as supercoiled DNA. BlNucB is a thermally robust enzyme that regains nuclease activity after a heat-cool cycle (Supplementary Figure S4B, C). By contrast, the recombinant human DNase I used in treatments for cystic fibrosis requires two N-linked glycosylation sites for full enzyme activity, thermal stability and protease resistance (64); moreover, human DNase I does not refold spontaneously after thermal denaturation (65). Possible methods to further enhance the in vivo allergenicity, activity and stability of BlNucB to address the health and societal challenges raised by biofilms include gene site saturation mutagenesis and/or the introduction of novel disulphides; these methods have found utility in improving xylanases for paper pulping (66) and viruses for vaccine production (67). Future work in our laboratories will focus on the precise roles played by the various elements of the catalytic machinery, determining how NucA and NucB utilize glutamate in catalysis, and how Nin exerts its inhibitory effects on NucA and NucB.

AVAILABILITY

The structure of NucB has been deposited in the Protein Data Bank (http://www.rcsb.org/pdb/home/home.do) under accession number 5OMT.

Click here for additional data file.