Molecular interplay between ComEC domains allows for selective degradation of the non-translocating strand during natural transformation

Abstract

Naturally competent bacteria can take up and incorporate environmental DNA using complex machinery in a process called natural transformation. This is a key mechanism in the spread of antibiotic resistance amongst bacteria, including many human pathogens. All competent bacteria require ComEC to transport the transforming DNA across the cytoplasmic membrane. In addition to the transmembrane domain predicted to form the DNA channel, most ComEC orthologues contain an oligonucleotide binding (OB) fold and β-lactamase-like domain. Here, we provide high-resolution structures and an in-depth characterization of the nuclease activity of the β-lactamase-like domain and the DNA-binding activity of the OB fold. We show that the in vitro nuclease activity of the β-lactamase-like domain is enhanced when the OB fold is encoded on the same polypeptide chain. Additionally, we identify a loop within the β-lactamase-like domain, positioned at the entrance of the DNA channel where the duplex DNA separates. Residues in this loop likely guide the non-translocating strand towards the nuclease domain, while a DNA channel lined with aromatic residues provides a path for the translocating strand. On the basis of our biochemical, structural, and functional characterization, we provide a model for how ComEC achieves DNA binding, degradation, and translocation.

Graphical Abstract

Graphical Abstract.

Introduction

Natural transformation is the uptake and subsequent genomic integration of exogenous DNA by naturally competent bacteria and, alongside transduction and conjugation, is a critical mechanism of horizontal gene transfer (HGT) [1, 2]. HGT is a driver of bacterial evolution, with important consequences for the spread of antibiotic resistance and other pathogenicity traits [3, 4]. Natural transformation can be distinguished from the other mechanisms of HGT, as the machinery required for DNA uptake, translocation, and integration is entirely encoded, expressed, and regulated by the competent cell [2, 5–7]. To date, transformation has been directly observed in over 80 species, including both Gram-positive and Gram-negative organisms [7, 8], yet the true prevalence is likely much higher.

Although there are differences in the cell envelopes of Gram-positive and Gram-negative bacteria, the general mechanisms governing natural transformation are remarkably conserved. In most species, type IV pili or related structures mediate DNA capture and uptake into the periplasmic space [9]. Most commonly the transforming DNA (tDNA) is linear and double-stranded DNA (dsDNA), although single-stranded DNA (ssDNA) and circular plasmid DNA can also be taken up [10–12]. Specialized pilus subunits capable of DNA binding have been identified in a number of species [13–20]. In Gram-negative bacteria, DNA binding by the pilus and subsequent pilus retraction is thought to bring the DNA into proximity with the outer membrane (OM) [16]. Pilus retraction is powered by the cytoplasmic ATPase PilT [21], yet the exact mechanisms of how the DNA is taken up across the OM and peptidoglycan layer into the periplasmic space, potentially through the secretin channel PilQ, remain to be defined. In the periplasmic space, DNA is bound by the DNA-binding protein ComEA, which ensures the unidirectional movement of DNA into the cell [22–26]. ComEA is membrane-bound in Gram-positive bacteria and capable of forming homo-oligomers through an oligomerization domain [26]. This process is thought to condense the incoming DNA in the periplasm, thereby exerting a pulling force on the DNA. In the next step, a single strand of the tDNA is translocated across the cytoplasmic membrane, which is mediated by the putative channel protein ComEC, while the non-translocating strand is degraded [27–29]. On the cytoplasmic side of the membrane, ComFC (ComF in most Gram-negative organisms) [30–32] and ComFA (a role potentially fulfilled by PriA in Gram-negative organisms) [33–37] act on the incoming ssDNA in ways that remain poorly understood. The ssDNA is protected from degradation by single-stranded DNA-binding protein (Ssb) and DNA-processing protein A (DprA) [38–40]. DprA loads the DNA onto recombinase A (RecA), which can integrate the DNA into the host genome provided there is sufficient homology between the sequences [41]. This process may further be assisted by the helicase ComM and the helicase-associated nuclease YraN [42–44].

ComEC is essential for natural transformation, with its deletion leading to total abrogation of transformation [24, 45–47]. It is present in both Gram-positive and Gram-negative competent organisms and most ComEC orthologues contain three predicted domains: the transmembrane DNA channel or competence domain (Pfam database [48], PF03772), the oligonucleotide binding (OB) fold (PF13567; domain of unknown function 4131, DUF4131), and the β-lactamase-like domain (PF00753) [8, 49–51]. Orthologues that lack either or both the OB fold and the β-lactamase-like domain have been identified in bioinformatic analysis [27], yet it remains to be shown whether all or some of the bacterial species that harbour these variants can undergo transformation. The competence domain appears to be universally conserved across competent species and is predicted to form the DNA channel critical for ComEC’s role of DNA transport across the cytoplasmic membrane [27–29]. Likewise, removal of the OB fold from the Bacillus subtilis ComEC orthologue resulted in a strain unable to undergo transformation [28].

In 1995 it was reported that a comEC deletion strain of B. subtilis was able to bind more tDNA on its cell surface (via ComEA), but showed a dramatic effect on DNA internalization [24]. A few years later in 2001, comEC null mutations were shown to prevent degradation of the non-transforming strand [52]. These observations led to the speculation that ComEC itself may harbour nuclease activity required for the degradation of the non-translocating strand. More recently, in silico predictions and subsequent confirmation of in vitro activity attributed this cryptic nuclease activity to the C-terminal β-lactamase-like domain of ComEC [29, 53]. Importantly, mutations in the B. subtilis β-lactamase-like domain that affect catalytic activity were shown to affect transformation in vivo [53, 54]. Interestingly, there are bacterial species capable of undergoing transformation that naturally lack the β-lactamase-like domain of ComEC. The Streptococcus pneumoniae ComEC orthologue likely contains an inactive β-lactamase-like domain, as it does not encode all of the predicted catalytic metal coordinating residues [29]. In this organism, the degradation of the non-translocating strand is carried out by the membrane-bound EndA nuclease [55]. The more distant Helicobacter pylori ComEC homologue neither contains the β-lactamase-like domain [45, 56], nor an EndA homologue, suggesting yet a further mechanism of strand degradation.

The β-lactamase-like domain of ComEC is a member of the metallo-β-lactamase (MBL) superfamily [57]. This superfamily of enzymes acts on a diverse set of substrates, including the β-lactam ring of the β-lactam class of antibiotics for which the name was coined. Within the MBLs, enzymes that hydrolyse the phosphoester bonds of a variety of substrates including nucleic acids and nucleotides, phospholipids, and phosphonates make up the most widespread functional group of MBLs [58]. The majority of nucleic acid processing MBLs are binuclear Zn²⁺-dependent enzymes, although other metal ions such as manganese, iron, or magnesium also occur in MBLs [58–60]. In 2021, Silale and colleagues showed that the nuclease activity of the β-lactamase-like domain of the thermophilic Gram-positive organism Moorella thermoacetica DSM 521 is dependent on Mn²⁺ coordination [53]. The general catalytic mechanism of nuclease MBLs, such as RNase J and RNase Z, relies on deprotonation of an active site water molecule by a general base. The metal ions coordinate the resulting hydroxide ion, which then acts as the nucleophile attacking the scissile phosphate of the substrate [61–65].

In this study, we provide an in-depth characterization of the nuclease activity of the β-lactamase-like domain and show that it functions as both an endo- and exonuclease, the latter with exclusive selectivity for the 5′ terminus. At the entrance of the DNA channel, an essential loop within the β-lactamase-like domain is structurally predicted to serve as a “pin loop” separating double-stranded DNA. Residues in this loop are likely also involved in guiding the 5′ strand towards the active site for degradation. Subsequently, a series of aromatic residues that line the ssDNA channel guide the uncleaved translocating DNA strand through the competence domain. While the DNA binding affinity of the β-lactamase-like domain is undetectably weak in vitro, the OB fold tightly interacts with DNA. We show that when both domains are encoded on the same polypeptide chain, the OB fold increases the local concentration of the DNA substrate in proximity with the β-lactamase-like domain’s active site, thus enhancing its nuclease activity markedly. Together with in vivo functional assays and the structural characterization of both the β-lactamase-like domain and the OB domain, we propose a mechanistic model of how ComEC functions to cleave the non-translocating strand, while the other is successfully translocated across the cytoplasmic membrane.

Materials and methods

Bacterial strains and growth conditions

Legionella pneumophila Lp02 strains (derived from L. pneumophila Philadelphia 1) were cultured in ACES [N-(2-acetamido)-2-aminoethanesulfonic acid] buffered yeast extract (AYE) liquid medium. For growth on solid medium, ACES-buffered charcoal yeast extract (CYE) was utilized. All media were supplemented with 100 μg/ml streptomycin, 100 μg/ml thymidine and 0.4 g L-cysteine and 0.135 g Fe(NO₃)₃ per litre of culture. For selection, 20 μg/ml kanamycin or 5 μg/ml chloramphenicol were added when appropriate. Supplementary Table S3 provides a list of all bacterial strains used in this study.

Plasmids

All constructs for recombinant protein expression were generated with the pOPINS vector [66]. The vector contains an N-terminal His₆-SUMO tag and inserts were cloned in frame downstream of the T7 promoter. Template DNA of the ComEC and ComEA genes from Moorella glycerini and the Sac7d gene from Sulfolobus acidocaldarius were synthesized (Twist Bioscience) prior to further cloning. Independent constructs of OB_Mg (residues 76–199) and BLACT_Mg (residues 532–797) and various fusion constructs were created. The fusion constructs encoded either OB_Mg [76–199], Sac7d_Sa [1–66] or ComEA_Mg (147–211) on the same polypeptide as BLACT_Mg (532–797), linked via an maltose-binding protein (MBP) scaffold (OB_Mg-MBP-BLACT_Mg, Sac7d_Sa-MBP-BLACT_Mg, ComEA_Mg-MBP-BLACT_Mg). GSSGSS linker sequences were introduced between the DNA-binding domain and MBP, and MBP and BLACT_Mg. Constructs for in vivo transformation assays were generated with pMMB207C by insertion of relevant constructs downstream of the Ptac promoter [67]. In-Fusion cloning and site-directed mutagenesis were carried out using the CloneAmp HiFi PCR premix (Takara) according to the manufacturer’s instructions. All plasmids used in this study are listed in Supplementary Table S4, while primer sequences can be found in Supplementary Table S5.

Protein production

All proteins were N-terminally His₆-SUMO tagged and expressed in BL21 (DE3) Escherichia coli cells. Cultures were grown in Luria–Bertani media at 37°C until an optical density at 600 nm (OD₆₀₀) of 0.6–0.8 was reached, while shaking. Cultures were then induced with 0.5 mM Isopropyl-β-D-thiogalactoside (IPTG) and further incubated for 12–18 h at 18°C, while shaking. Cells were lysed in 50 mM HEPES–NaOH (pH 7.2), 1 M NaCl, 40 mM imidazole, supplemented with 0.2 mg/ml lysozyme, 10 μg/ml DNase, and one cOmplete mini ethylenediaminetetraacetic acid (EDTA)-free protease inhibitor tablet (Roche). Cells were lysed by passing the suspension three times through an EmulsiFlex-C5 homogenizer (Avestin) at 40′000 psi. The lysate was clarified by centrifugation in a JLA-16.250 rotor (Beckman Coulter) at 30′000 × g for 60 min, filtered through a membrane with a pore size of 0.22 μm and applied to a 5 ml HisTrap HP column (Cytiva). Elution was performed either with a linear 40–500 mM imidazole gradient or by a stepwise elution with 500 mM imidazole. Protein containing fractions were pooled and dialysed against 50 mM HEPES–NaOH (pH 7.2), 50 mM NaCl, while the His₆-SUMO tag was cleaved by addition of the catalytic domain of the human SENP1 protease to the dialysate. The OB_Mg and BLACT_Mg were further purified by ion exchange chromatography using a 5 ml HiTrap Q HP column (Cytiva), collecting the protein of interest in the unbound fraction. In contrast, fusion proteins were purified using a 5 ml HiTrap SP HP column (Cytiva), eluting bound proteins using a linear salt gradient from 50 mM to 1 M NaCl. The final purification step was size exclusion chromatography of the samples using either a HiLoad 16/600 Superdex 75 pg column or a Superdex 75 Increase 10/300 GL column (Cytiva). Protein solutions were concentrated using centrifugal filter devices with a molecular weight cut-off of 10 or 30 kDa (Millipore) and the concentration was determined by measuring the specific absorption at 280 nm, using the molar extinction coefficient of 14 900 M⁻¹ cm⁻¹ for OB_Mg, 26 930 M⁻¹ cm⁻¹ for BLACT_Mg, 108 180 M⁻¹ cm⁻¹ for OB-MBP-BLACT, 101 760 M⁻¹ cm⁻¹ for Sac7d-MBP-BLACT and 94 770 M⁻¹ cm⁻¹ for ComEA-MBP-BLACT. All purification steps were performed at room temperature, except overnight tag cleavage, which occurred at 4°C.

X-ray crystallography

BLACT_Mg was crystallized using the sitting drop vapour diffusion method at 20°C at a concentration of 10 mg/ml in 30% (w/v) precipitant mix 1 (PEG 500 MME, PEG 20′000), 0.1 M buffer system 1 (1 M 2-(N-morpholino)ethanesulfonic acid (MES) and 1 M imidazole mixed in 56:44 ratio to achieve pH 6.5), 0.09 M NPS mix (0.3 M sodium phosphate dibasic dihydrate, 0.3 M ammonium sulfate, 0.3 M sodium nitrate) (well C1, Morpheus I, Molecular Dimensions). Diffraction data were collected at the Swiss Light Source (SLS) beamline X10SA (PXII) at a wavelength of 0.999989 Å. Data processing was performed within the CCP4i program suite [68, 69]. The data were indexed and scaled using iMOSFLM [70] and AIMLESS [71], respectively. There is one molecule in the asymmetric unit and the crystal belongs to the space group C 2 2 2₁. The structure was determined by molecular replacement in MOLREP [72] using an AlphaFold2-generated search model lacking any active site metal ions. The protein chain and phosphate group were iteratively built in COOT [73] and refined in REFMAC5 [74] and PHENIX [75]. The refinement strategy included positional refinement, solvent correction and individual B-factor refinement. Final statistics for the BLACT_Mg structure can be found in Supplementary Table S1.

Production of isotope-labelled OB_Mg

Uniformly ¹³C, ¹⁵N-labelled OB_Mg was produced by growing cells in M9 minimal medium containing 1 g/l ¹⁵NH₄Cl and 3 g/l ¹³C₆-glucose, supplemented with 2 mM MgSO₄, trace elements, vitamin mix, and 50 μg/ml kanamycin for selection. Protein expression and purification were performed as described above.

Data acquisition and structure determination

For nuclear magnetic resonance (NMR) resonance assignments and structure determination, samples consisting of 1.5 mM uniformly ¹³C, ¹⁵N-labelled OB_Mg in 50 mM HEPES, pH 7.2, 50 mM NaCl and 10% D₂O were used. Spectra were recorded at 25°C in 3 mm diameter NMR tubes (Bruker). 3D HNCACB [76, 77] and 3D CBCACONH [78] spectra were recorded on a 700 MHz AVNEO spectrometer equipped with a TCI cryo-probe (Bruker). The spectra consisted of 2048 × 50 × 90 complex points in the ¹H, ¹⁵N, and ¹³C dimensions with respective spectral widths of 16, 34, and 64 ppm and were recorded with 8 scans per increment resulting in 2 and 1.5 days of measurement time, respectively. Side chain assignments were hampered by the intense signals of the HEPES buffer and therefore an alternative sample with deuterated Tris (d-Tris) (Sigma, 449 105) was produced. However, with d-Tris as a buffer substance, the protein could not be concentrated to the same level as in HEPES buffer. The concentration was sufficient for a 3D HcC(aliaro)H-TOCSY spectrum [79] recorded on a 600 MHz AVIIIHD spectrometer equipped with a TCI cryo-probe (Bruker). The spectrum consisted of 1536 × 75 × 150 complex points in the ¹H, ¹H, and ¹³C dimensions with respective spectral widths of 16, 10, and 140 ppm, and was recorded with 2 scans per increment in 3 days using a recycle delay of 2 s. NOESY spectra were recorded for both types of samples: the sample in d-Tris buffer produced a clean NOESY spectrum, however, with limited sensitivity, and the sample in HEPES buffer produced a highly sensitive spectrum where the region between 2.8 and 3.9 ppm could however not be interpreted due to strong T₁ noise. In detail, time shared 3D [¹³C/¹⁵N, ¹H]-HSQC NOESYs (modified from [80]) were recorded on a 900 MHz AVNEO spectrometer equipped with a TCI cryo-probe (Bruker). The spectra consisted of 1536 × 120 × 256 complex points in the ¹H, ¹H, and ¹³C/¹⁵N dimensions with respective spectral widths of 16, 11, and 140/80 ppm, and were recorded with 2 scans per increment in 3 days.

Resonance assignments were determined with the program CARA (www.cara.nmr.ch). The signal intensities exhibited strong variations and only stretches including amino acid residues 86–155 and 182–199 could be assigned to 96% completeness (Supplementary Table S2). Automated peak picking of NOESY spectra was performed with the program Artina [81] and peak lists were manually cleaned from artefacts using the ccpnmr 2.5.1 software package [82]. Resonance assignments and peak lists from both samples were combined and were used as input for a structure calculation with Artina and Cyana (version 3.98.15 [83]). A total of 140 angle constraints were automatically generated from Cα chemical shifts, and 2098 unambiguous NOE distance restraints were used to calculate a bundle of 100 conformers, from which the 30 with the lowest Cyana target function were selected for refinement in implicit water in the program Amber20 [84] and the final 20 with the lowest energy were used to represent the structure. A total of 43 hydrogen bonds were identified in more than six structures, Ramachandran plot statistics were as follows: 92.1%, 7.7%, 0.1%, and 0.1% in favoured, allowed, generously allowed and disallowed regions, respectively, as defined by the program PROCHECK [85]. Further structural statistics can be found in Supplementary Table S2.

Nuclease activity assays

Various DNA probes were tested in nuclease activity assays (Microsynth; Supplementary Table S5). A fluorescein (FAM)-label was attached to a thymine (T) base either at the 5′ or 3′ terminus of all substrates, and some substrates contained phosphorothioate (PTO) bonds. To generate linear FAM-labelled dsDNA, a single strand of FAM-labelled DNA was annealed with the complementary unlabelled strand. Plasmid DNA, circular ssDNA and dsDNA, was obtained from ThermoFisher, Takara, and Microsynth, respectively. Reactions were performed by mixing either 10 μM linear ssDNA (50 nt), 5 μM linear dsDNA (50 bp), 67 nM circular ssDNA (M13mp18, 7429 nt) or 59 nM circular or linearized dsDNA (pBR322, 4361 bp) with 1 μM enzyme in 50 mM HEPES–NaOH (pH 7.2), 50 mM NaCl, 5 mM MnCl₂, in a total volume of 100 μl. Reactions were incubated at 50°C in a TAdvanced thermocycler (Biometra) and timepoints were taken by removing 8 μl of the reaction and quenching it with 8 μl Novex 2× Tris-borate-EDTA (TBE)–urea sample buffer (Invitrogen). Samples were resolved on 12% polyacrylamide gels containing 7 M urea and fluorescence detection was achieved using a ChemiDoc imaging system (Bio-Rad). Fluorescent band intensities were measured (GelAnalyzer V19.1) and normalized to the t = 0 time point (corresponding to either 5 μM dsDNA or 10 μM ssDNA) and the percentage of intact substrate remaining was plotted against time. All measured fluorescence intensity values were within a linear range, as confirmed by a standard curve of known DNA concentrations. The apparent initial reaction rate was determined by linear regression of the data points that lie within the initial linear part of the curve, divided by the enzyme concentration. The plasmid DNA degradation experiments were resolved on a 1% (w/v) agarose gel and visualized using UV illumination (Carestream). Each reaction was performed and analysed at least three times. Experiments for Figs 1C, 2A, and 4B and D, and Supplementary Fig. S5A were performed at the same time, therefore the data for BLACT are the same across these panels. Equally, data for OB-MBP-BLACT are the same in Fig. 4B and D. Supplementary Table S5 provides a list of all oligonucleotide substrates used for biophysical assays.

Figure 1.

The structure of BLACT_Mg reveals the catalytic mechanism of ComEC’s β-lactamase-like domain. (A) Schematic illustrating the domain organization of ComEC. Most ComEC orthologues are composed of three domains: the OB fold (orange), the competence/channel domain (grey) and the β-lactamase-like domain (blue), although versions lacking either or both the OB fold or β-lactamase-like domain also exist. (B) The crystal structure of BLACT_Mg displayed in ribbon representation with active site residues and the bound phosphate ion shown in stick representation. N- and C-termini are indicated on the structure. The inset shows a zoomed in view of the active site with ordered water molecules and the bound phosphate ion. Active site residues and the phosphate ion are shown in stick representation. The electron density around active site residues, water molecules and the phosphate ion is shown as a black mesh with the refined 2mF_O-DF_C map contoured at 1σ. (C) Time-course nuclease activity assay monitoring the degradation of ssDNA by BLACT_Mg in the presence of 5 mM of various metal ion cofactors. The image of the gel (top) was used to measure the fluorescent band intensity, from which the percentage of intact substrate remaining at each time point was plotted (bottom). Nuclease assays were conducted with 1 μM wild-type (WT) BLACT_Mg and 10 μM ssDNA substrate, 50 nt in length. The DNA substrate was fluorescently labelled with fluorescein (FAM) on the first base (thymine, T) at the 5′ terminus. Error bars represent the standard error from three technical replicates. (D) Schematic showing the proposed catalytic mechanism of BLACT_Mg. Deprotonation of a coordinated water molecule (red) by D602 functioning as a general base forms a hydroxide ion bridging the two active site metals. This ion initiates a nucleophilic attack on the scissile bond resulting in new 5′ phosphate and 3′ OH groups through nucleophilic substitution.

Figure 2.

BLACT_Mg displays endonuclease and 5′-to-3′ exonuclease in vitro. (A) Time-course nuclease activity assay comparing the degradation of 5′ OH or phosphate-modified ssDNA or dsDNA. These experiments were performed using 10 μM ssDNA (50 nt) or 5 μM dsDNA (50 bp) in order to keep the total number of cleavable phosphodiester bonds identical. The mode of FAM-labelling of ssDNA was identical to the experiments in Fig. 1, in the case of dsDNA, only one strand was fluorescently labelled. Error bars represent the standard error from three technical replicates. (B) Schematic illustrating the DNA substrates utilized in the experiment. All 50 nt ssDNA substrates are FAM-labelled (orange star), either at the first or last base (thymine, T). Each substrate carries a single phosphodiester bond marked by the letter P inside a red circle, while all other backbone sugar moieties are linked by nuclease-resistant phosphorothioate (PTO) bonds. (C) BLACT_Mg-mediated cleavage of the four substrates depicted in panel (B) monitored over a 48 h time course resulting in the appearance of distinct cleavage products. Control, fluorescein-labelled dUTP (fluorescein-12-dUTP).

Figure 4.

The OB fold serves as the DNA-binding domain for the β-lactamase-like domain. (A) Schematic showing the domain organization of WT ComEC and various fusion constructs, encoding either the OB fold or other DNA-binding domains (Sac7d from S. acidocaldarius and ComEA from M. glycerini) on the same polypeptide as the β-lactamase-like domain linked via an MBP scaffold. (B–D) Time-course nuclease activity assay comparing OB, BLACT, OB + BLACT, and OB-MBP-BLACT (B), BLACT and OB-MBP-BLACT at four different NaCl concentrations (C), and BLACT, OB-MBP-BLACT, Sac7d-MBP-BLACT, and ComEA-MBP-BLACT (D). Nuclease assays were conducted with 1 μM enzyme and 10 μM 50 nt ssDNA substrates. The mode of FAM-labelling was identical to the experiments in Fig. 1.

Fluorescence anisotropy

All DNA binding experiments were performed using either 12 bp FAM-labelled dsDNA, generated by annealing a 12-meric 5′ FAM-labelled strand and a complementary unlabelled strand or 12-meric 5′ FAM-labelled ssDNA. In some experiments, all phosphodiester bonds were replaced by nuclease-resistant PTO bonds (Microsynth; Supplementary Table S5). In contrast to the nuclease activity assays, here, the FAM moiety was attached to the 5′ phosphate of the sequence. A constant DNA concentration of 20 or 50 nM was incubated with increasing concentrations of protein in 50 mM HEPES–NaOH (pH 7.2), 50 mM NaCl for 60 min at 25°C. To test the effect of ionic strength on binding, some experiments were conducted with buffers containing different NaCl concentrations, as indicated in the relevant figures. Following incubation, 30 μl of each sample was placed in a 96-well half area black flat bottom polystyrene plate with a nonbinding surface (Corning). Fluorescence intensities, parallel and perpendicular to the excitation polarization, were measured in a Synergy2 plate reader (BioTek) (excitation: 495 nm; emission: 520 nm). The anisotropy was calculated using the following equation (Eq. 1),

(1)

where r is the anisotropy, I_|| is the fluorescence intensity in the parallel direction, and I_⊥ is the fluorescence intensity in the perpendicular direction. The calculated anisotropy for each sample was plotted against the protein concentration and the curve was fitted to a model assuming one set of binding sites in order to derive the dissociation constant (K_D) (Eq. 2).

(2)

All binding measurements were performed at least three times. As an additional control, some samples were measured twice, 60 min apart, to ensure that the binding equilibrium was fully attained at the time of measurement.

Mass spectrometry

To determine the number of coordinated manganese ions by BLACT_Mg, mass spectrometry was performed. A solution containing 34 µM BLACT_Mg was analyzed in the absence and presence of 5 mM MnCl₂. Data were collected on a Synapt G2-Si mass spectrometer using nano electrospray injection mass spectrometry.

Transformation assays

Transformation assays were performed as previously described [15]. Frozen stocks or freshly transformed Lp02 cells were streaked onto CYE plates supplemented with appropriate antibiotics, and incubated at 37°C for 3–5 days until colonies appeared. Bacteria were resuspended in a 10 ml AYE liquid starter culture and incubated overnight at 37°C while shaking. Fresh 10 ml AYE cultures were inoculated with overnight starter cultures and incubated at 30°C while shaking. At an OD₆₀₀ of 0.3, 1 ml of the culture was transferred to a new tube, mixed with 1 μg of tDNA, and further incubated at 30°C for 24 h. The tDNA fragment, containing the kanamycin resistance cassette, was described previously [15, 86]. For complementation experiments, 0.5 mM IPTG was added simultaneously with the tDNA to induce ectopic expression of genes under the control of an IPTG-inducible Ptac promoter. Serial dilutions of the culture were spread onto selective and nonselective plates and colony forming units (CFUs) were counted after 4 days of incubation at 37°C. The final transformation efficiency was calculated by dividing the number of CFUs on selective plates by the number of CFUs on nonselective plates. Plates containing fewer than 10 CFUs were not counted.

Sequence alignments and conservation analysis

A set of 2000 ComEC sequences were retrieved by BlastP (Blast v2.15.0) against the full-length M. glycerini ComEC sequence using a 95% query coverage cutoff. All searches were performed with the refseq_select database, which contains only one reference genome for each prokaryotic species to reduce redundancy in the search. The search was conducted with the Gram-positive exclusion filter followed by an identical search with the Gram-negative exclusion filter to ensure even selection of ComEC sequences from Gram-positive and Gram-negative organisms. The list of sequences was further curated manually by removing any redundant sequences. A multiple sequence alignment was performed with ClustalOmega [87]. Conservation was mapped onto the predicted structure of full-length ComEC, the BLACT_Mg crystal structure, and the OB_Mg NMR structure using the ConSurf server [88].

3D protein structure prediction and comparison

The AlphaFold3 server (https://www.alphafoldserver.com) was used to predict structural models of full-length ComEC_Mg, ComEC_Lp, ComEC_Lp in complex with manganese ions and DNA and the OB_Mg-MBP-BLACT_Mg fusion construct [89]. To identify structural homologues of BLACT_Mg and OB_Mg, their PDB files were submitted to the DALI protein structure comparison server (http://ekhidna2.biocenter.helsinki.fi/dali/) and hits with high Z-scores were chosen for further comparison [90]. Structural figures were produced in ChimeraX [91].

Results

The β-lactamase-like domain of ComEC degrades DNA through a two-metal-ion catalytic mechanism

Most ComEC orthologues are comprised of three domains: an OB fold encoded near the N-terminus, a central competence domain consisting of a bundle of transmembrane helices forming a channel and a C-terminal β-lactamase-like domain (Fig. 1A). In order to learn more about the mechanism of the β-lactamase-like domain, we sought to determine its three-dimensional structure. Previous work led to the identification of the M. thermoacetica β-lactamase-like domain orthologue as a well-expressed and soluble construct suitable for enzymatic characterization [53]. In our hands, the orthologue from M. glycerini (hereafter BLACT_Mg), a close relative of M. thermoacetica, also yielded soluble and stable preparations, and proved to be amenable to structure determination by X-ray crystallography. We determined the 1.8 Å crystal structure of BLACT_Mg, consisting of the characteristic αβ/βα fold that defines the MBL superfamily [57] (Fig. 1B and Supplementary Table S1). The active site is positioned at one end of a wide shallow groove between the sandwiched central β-sheets. Our structure contains clear density for a phosphate ion in the active site, but lacks coordinated metal ions and thus represents an inactive state. A DALI [90] search revealed that the closest structural homologues of BLACT_Mg is teichoic acid phosphorylcholine esterase (Pce) [92] from S. pneumoniae and the closest nuclease is RNase J from Streptomyces coelicolor [61] (Supplementary Fig. S1). Like most MBLs with RNA or DNA substrates, RNase J utilizes two metal ions for catalysis [58–60] and its active site is situated at the interface between the hydrolytic β-lactamase domain and an auxiliary β-CASP domain. Indeed, nucleic acid specific members of the MBL superfamily often contain additional domains involved in substrate recruitment, such as the tRNase Z exosite for tRNA binding [93], β-CASP [60], or KH [94, 95] domains for DNA or RNA binding. These domains can be inserted within the MBL fold, or occur N- or C-terminally to it [58]. In the context of full-length ComEC, the OB fold may serve as the DNA-binding domain that is lacking within the β-lactamase-like domain. While our structure of BLACT_Mg does not contain the β-CASP domain, it does display common sequence motifs found in other MBL enzymes [96], such as the characteristic H-X-H-X-D-H motif (where X is any residue), which consists of H598-X-D600-X-D602-H603 in BLACT_Mg.

In a previous study it was observed that the β-lactamase-like domain from M. thermoacetica is Mn²⁺-dependent [53]. We were unable to produce crystals in the presence of supplemented Mn²⁺, and, accordingly, there is no density that would correspond to the coordinated metal ions. In order to test whether BLACT_Mg is also a Mn²⁺-dependent nuclease, we performed a gel-based time-course nuclease activity assay using fluorescently labelled 50 nucleotide (nt) ssDNA as the substrate in the presence of various potential metal cofactors, followed by fluorescence intensity measurement of the band corresponding to the original intact substrate (Fig. 1C). The fluorescein (FAM) moiety was attached on the first base (thymine, T) at the 5′ terminus. These assays indeed confirmed that, like its close relative from M. thermoacetica, BLACT_Mg is also an Mn²⁺-dependent nuclease. The gradual decrease in band intensity without a change in its electrophoretic mobility also suggests that no or very little 3′-to-5′ exonuclease activity is taking place in this reaction. To definitively determine the coordination number of BLACT_Mg, we performed mass spectrometry on our purified BLACT_Mg sample, with and without supplementary Mn²⁺ in the solution (Supplementary Fig. S2). BLACT_Mg contains no bound Mn²⁺ without specific supplementation of the metal (as captured by our crystal structure), whereas the majority of the molecules in our sample are coordinated with two Mn²⁺ ions when BLACT_Mg was incubated with 5 mM MnCl₂. The absence of coordinated Mn²⁺ ions prevents the active site from adopting a fully catalytically competent conformation. In order to generate a general model for catalysis and to better understand the active site residue rearrangements necessary, we compared the active site of BLACT_Mg with structures of other MBL family enzymes where two metal cofactors were coordinated in the crystal structure (Fig. 1D and Supplementary Fig. S3). Based on the catalytic mechanisms of those family members and structural predictions of BLACT_Mg coordinated by Mn²⁺ ions and bound to adenosine monophosphate (Supplementary Fig. S3), we propose that D602, H603, and D600 rotate into the catalytic centre to coordinate metal ions. This would place the substrate in proximity with H725, the metal ions, and an active site water, allowing for catalysis to proceed. In this conformation, deprotonation of the active site water by the general base (D602) could proceed resulting in the formation of a hydroxide ion. This hydroxide bridges the two Mn²⁺ ions in the active site prior to nucleophilic attack on the 5′ side of the scissile phosphate. The nucleophilic substitution generates the 5′ phosphate and 3′ OH products. In general, the residues in and around the active site are highly conserved (Supplementary Fig. S4).

In order to test the effect of mutation of important residues on nuclease activity, we performed further nuclease activity assays comparing WT BLACT_Mg, a control mutation (D541A) (highly conserved but not in the active site), several residues involved in Mn²⁺-coordination (H598A, D600A, D602A, H603A, D700A) and putative DNA binding residues (H725A, S728A) (Supplementary Figs S5and S6). Mutation of most residues involved in metal coordination render BLACT_Mg virtually inactive in our nuclease activity assay, with the only exception being D600A that showed a ∼5 fold reduction in the apparent initial rate compared to WT. Similarly, substitution of the putative DNA binding residue S728 to alanine resulted in a ∼3 fold reduction of the apparent initial rate, whereas the control mutation D541A showed similar activity compared to the WT. We also tested the effect of these mutations on transformation efficiency in vivo, where the analogous residues in L. pneumophila Lp02 were targeted (Supplementary Fig. S5). Interestingly, all residues that showed reduced nuclease activity in vitro, except for S662A (corresponding to S728 in M. glycerini), showed complete abrogation of transformability. Similar mutations of catalytic β-lactamase-like domain residues in B. subtilis also resulted in reduced transformation efficiencies [53, 54], albeit to a lesser extent. Taken together, BLACT_Mg coordinates two Mn²⁺ in its catalytic centre in order to function as a DNA nuclease. When critical residues involved in the catalytic mechanism are substituted to alanine, BLACT_Mg is mostly rendered inactive both in vitro and in vivo, resulting in a severe transformation phenotype.

BLACT_Mg is an endonuclease and 5′-to-3′ exonuclease in vitro

At present it is unclear whether the β-lactamase-like domain of ComEC encounters ssDNA or dsDNA in vivo. Even when dsDNA is taken up into the periplasmic space, it is possible that the translocating and non-translocating strands are separated prior to encountering the nuclease domain, which then engages with a single stranded substrate. In order to test whether BLACT_Mg shows an intrinsic preference for ssDNA or dsDNA and whether the presence of a hydroxyl or phosphate group at the 5′ end of the DNA substrate influences the rate of DNA degradation, we performed another time-course nuclease activity assay (Fig. 2A). In order to maintain the total number of cleavable phosphodiester bonds the same between ssDNA and dsDNA substrates, we used 10 μM ssDNA and 5 μM dsDNA in this experiment. These data showed that BLACT_Mg has a clear preference for ssDNA and that the rate of degradation is slightly increased when ssDNA is modified with a 5′ phosphate group. This is consistent with the observation that the closest structural homologues of BLACT_Mg are enzymes that degrade single-stranded nucleic acid substrates and that the most likely physiological substrate is also the best substrate.

Our nuclease activity assay monitors the disappearance of the original intact substrate provided, namely FAM-labelled 50 nt ssDNA. Since the FAM moiety is attached to the first base at the 5′ end of the DNA molecule, we would observe a decrease in band intensity if BLACT_Mg degrades the substrate either with an endonucleolytic or 5′-to-3′ exonucleolytic cleavage mode. In contrast, if the substrate were degraded in a 3′-to-5′ exonucleolytic mode, we would see no change in band intensity, but rather a gradual increase in its electrophoretic mobility (gradual downward shift). As alluded to above, we do not observe such a change in electrophoretic mobility, suggesting that BLACT_Mg uses an endo- and/or 5′-to-3′ exonucleolytic mode. The slight preference for DNA substrates modified with a phosphate group at the 5′ terminus suggests that at least some exo-activity occurs, as endo cleavage events should not be affected by the terminus modification potentially some distance removed. BLACT_Mg will predominantly encounter 5′ termini with a phosphate group considering that environmental DNA was likely processed by nucleases which typically result in 5′ phosphate and 3′ hydroxyl groups [97]. However, BLACT_Mg can also degrade DNA substrates harbouring a 5′ hydroxyl group, as shown in our nuclease activity assay (Fig. 2A). In order to learn more about the cleavage mode of BLACT_Mg, we designed several distinct ssDNA substrates consisting of nuclease-resistant PTO bonds and a single cleavable phosphodiester bond and performed a nuclease assay (Fig. 2B and C). The cleavable bond was positioned either at the 5′ or 3′ terminus, or in the middle of the 50 nt ssDNA, and the FAM label was attached such that we could monitor the appearance of a single nucleotide product (Fig. 2B). These experiments showed that BLACT_Mg is able to cleave the first nucleotide from the 5′ end exonucleolytically, regardless of whether the terminus is modified by a phosphate or hydroxyl group. In contrast, there is no exonucleolytic cleavage occurring from the 3′ terminus, even after 48 h. Interestingly, we observe the occurrence of an intermediate band in our experiments (Fig. 2C), most likely due to the very slow hydrolysis of PTO bonds over time. The center of the DNA substrate seems to be more susceptible to endonucleolytic cleavage potentially due to a minimal ssDNA length requirement, or sequence preference for cleavage. Lastly, the enzyme is also clearly capable of cleaving endonucleolytically, which was further confirmed by the ability of BLACT_Mg to degrade plasmid DNA (Supplementary Fig. S7), as shown previously [53]. In summary, in vitro the BLACT_Mg functions as a 5′-to-3′ exonuclease and an endonuclease.

The OB fold of ComEC binds to DNA with high affinity through an electrostatically driven interaction

In addition to the β-lactamase-like domain and the competence/channel domain, the majority of ComEC orthologues were found to also contain an OB fold encoded near the N-terminus [27] (Fig. 1A). Structural predictions suggest that the OB fold of Gram-positive organisms like M. glycerini is located on the extracellular side of the cytoplasmic membrane [28], and in Gram-negative organisms like L. pneumophila it is located in the periplasm (Supplementary Fig. S8). In addition, the OB fold has been predicted to interact with DNA [29], making it an ideal candidate to function in DNA handover between ComEA and the β-lactamase-like domain of ComEC. Therefore, we tested whether the OB fold of M. glycerini (hereafter OB_Mg) can indeed bind to DNA. We performed a fluorescence anisotropy binding experiment where we incubated 12-meric FAM-labelled ssDNA or dsDNA with OB_Mg (Fig. 3A). This showed that OB_Mg binds to dsDNA [dissociation constant (K_D) of 0.33 μM] with approximately 2.4-fold greater affinity than to ssDNA (K_D of 0.79 μM). Next, we determined the structure of the OB_Mg using NMR spectroscopy (Fig. 3B and Supplementary Table S2). The OB_Mg structure contains a 6-stranded mixed β-sheet and a 28 residue long disordered loop between β4 and β5. OB family proteins frequently contain small α-helices between β3 and β4 or β4 and β5. However, there are no such helices in our structure, which for the β4-β5 loop could be due to the absence of contacts from the competence domain and/or the β-lactamase-like domain. Although OB fold sequences can be very divergent, both between ComEC sequences as well as between OB folds contained in other proteins, their overall fold is similar (Supplementary Fig. S9). Some OB fold containing proteins, such as the telomere end binding protein, are known to form functional dimers [98], OB_Mg behaves as a monomer in solution (Supplementary Fig. S10). Furthermore, the structure of OB_Mg does not contain any disulphide bonds, as previously described for the B. subtilis OB fold (formerly called the N-loop) [28].

Figure 3.

The interaction of OB_Mg and DNA occurs through positively charged surface residues and is dependent on ionic strength. (A) Fluorescence anisotropy binding experiments testing the interaction between 12-meric 5′ FAM-labelled ssDNA or dsDNA and OB_Mg, in a buffer containing 50 mM NaCl. (B) The solution structure of OB_Mg shown in ribbon representation (left), a superposition of all states (middle) and the corresponding topology diagram (right). The six-stranded mixed β-sheet (orange), as well as N- and C-termini are indicated on the figure. (C) Fluorescence anisotropy binding experiments to assess interaction between 12 bp 5′ FAM-labelled dsDNA and WT OB_Mg in buffers containing 50–250 mM NaCl. All fluorescence anisotropy experiments were performed in triplicate and error bars represent the standard error of the mean. (D) Dissociation constants (K_D) derived from fluorescence anisotropy experiments performed with OB_Mg mutants at 50 mM NaCl. Single and triple mutants are plotted separately. For single mutants, thresholds indicating a 2- or 3-fold increase in the mutant K_D relative to WT is marked on the graph by dashed lines. Triple mutants are indicated on the figure as group 1 (R112A/R115A/R122A), group 2 (R150A/R188A/K192A) and group 3 (R115A/R150A/R188A). (E) The fold change in K_D from panel (D) was mapped onto the surface of OB_Mg. Residues that increased the K_D by >2- or 3-fold when substituted to alanine are coloured in light pink and red respectively.

We attempted to perform chemical shift perturbation experiments to map the DNA binding interface as well as to further structurally characterize the DNA-bound state by NMR. However, at the high protein concentrations required for these DNA titration or triple resonance experiments, the complex of OB_Mg-DNA was prone to precipitation. Therefore, we again turned to fluorescence anisotropy binding experiments in order to learn more about the DNA binding mode of OB_Mg. First, we tested whether the affinity of the interaction between WT OB_Mg and dsDNA is affected by the ionic strength of the buffer (Fig. 3C). Indeed, the K_D decreased from 30 to 0.45 μM, when the NaCl concentration in the buffer was reduced from 175 to 50 mM, respectively, suggesting that this interaction is electrostatically driven. Next, we substituted several candidate DNA binding residues to alanine, focusing on positively charged residues, on the surface of OB_Mg and repeated the fluorescence anisotropy measurements (Fig. 3D and Supplementary Fig. S6). While some of these single residue substitutions had almost no effect on the affinity of the interaction (e.g. N103A, H120A, K121A), all other single amino acid substitutions increased the K_D between approximately two- and three-fold. We also created triple mutants, where we substituted three nearby residues with alanine, (group 1: R112A/R115A/R122A; group 2: R150A/R188A/K192A: group 3: R115A/R150A/R188A). These triple mutations had a greater effect on DNA binding, resulting in ∼4–19 fold increase in the measured K_D (given the WT K_D range of ∼0.33–0.8 μM). We mapped all residues that, on their own, showed greater than a two-fold increase in the K_D onto our NMR structure of OB_Mg (Fig. 3E). This revealed a binding surface on one side of the molecule that consists of R112, R115, R122, R150, R188, and K192, although the more structurally isolated R175 also seems to contribute to DNA binding. In the context of the structural prediction of full-length ComEC, this surface is located such that DNA binding to the OB fold and subsequent degradation by the β-lactamase-like domain would be conceivable (Supplementary Fig. S11). Taken together our data show that OB_Mg tightly interacts with DNA and that this interaction is electrostatically driven, occurring primarily through several positively charged residues clustered together on the surface of the molecule.

The OB fold and β-lactamase-like domain of ComEC work in concert for efficient nuclease activity

During natural transformation, competent cells are able to take up vast stretches of DNA [99], indicating that the uptake machinery must be highly efficient and processive. Yet, throughout this study, we repeatedly noted the relatively poor apparent initial rate of DNA degradation of BLACT_Mg (∼0.26 min⁻¹) and its seemingly absent affinity for its DNA substrate (K_D not determinable) (Table 1 and Supplementary Fig. S12). As mentioned previously, a close structural homologue of BLACT_Mg is RNase J, which contains an additional β-CASP domain involved in substrate binding. In the structural prediction of ComEC, the OB fold is positioned next to the β-lactamase-like domain (Supplementary Fig. S8). We hypothesized that in a manner analogous to the substrate-binding β-CASP domain of RNase J, the presence of the OB fold may affect the rate of DNA degradation by the β-lactamase-like domain. Therefore, we again performed nuclease activity assays using fluorescently labelled 50 nt ssDNA. As we are currently unable to produce sufficient amounts of full-length ComEC, we compared the nuclease activity of BLACT_Mg on its own and in the presence of OB_Mg added in solution or tethered to the BLACT_Mg domain as a fusion construct (Table 1, Fig. 4A and B, and Supplementary Fig. S6). We designed the fusion construct in a way that would mimic the overall ComEC architecture by using MBP as a scaffold, which is of similar size to the competence domain and would result in similar relative domain positioning according to its predicted structure (Supplementary Fig. S13). As controls, the nuclease activities of OB_Mg on its own and MBP-BLACT were also recorded. These data showed that adding the OB fold in solution did not change the apparent initial rate of DNA degradation significantly (0.26 versus 0.29 min⁻¹), whereas in the context of the OB-MBP-BLACT fusion construct the activity was clearly increased (approximately seven-fold increase in apparent initial rate to 1.85 min⁻¹). OB_Mg on its own did not display any nuclease activity and the MBP-BLACT construct resulted in a small increase compared to BLACT_Mg (0.30 min⁻¹ versus 0.26 min⁻¹) (Fig. 4B and Supplementary S13C). Given that DNA binding by the OB fold is heavily influenced by the ionic strength of the solution (Fig. 3C), we tested its effect on the nuclease activity of BLACT_Mg and OB-MBP-BLACT (Table 1 and Fig. 4C). Not surprisingly, the ionic strength had a profound effect on the rate of DNA degradation by the OB-MBP-BLACT fusion construct, with the apparent initial rate increasing ∼32-fold from 0.06 min⁻¹ to 1.92 min⁻¹ when the NaCl concentration was decreased from 500 to 50 mM. The effect on the BLACT_Mg alone was less pronounced (approximately two-fold increase in apparent initial rate), yet still clearly measurable. These data are in agreement with the OB fold serving as the main DNA binding platform for the β-lactamase-like domain. Next, we wondered whether the OB_Mg allows for proper substrate positioning with respect to the BLACT_Mg active site, or if the simple increase in the local concentration of DNA near the BLACT_Mg active site is sufficient to explain the increase in activity. To answer this question, we replaced the OB_Mg moiety within the fusion construct with another DNA-binding protein and performed nuclease activity assays (Fig. 4D and Supplementary Fig. S6). For this purpose, we chose the DNA-binding domain of ComEA from M. glycerini and Sac7d (an OB fold family protein) from S. acidocaldarius. Like OB_Mg, these domains are small, soluble, and thermostable DNA-binding domains that bind to DNA with similar affinities [100, 101]. This experiment showed that it does not matter whether the BLACT_Mg is fused to OB_Mg, ComEA_Mg, or Sac7d, the presence of a DNA-binding domain on the same polypeptide chain as the β-lactamase-like domain leads to a similar increase in nuclease activity. This in turn suggests that an increase in local concentration of the substrate is the main mechanism underlying this observed boost in nuclease activity.

Table 1.

Apparent initial rates of BLACT and OB-MBP-BLACT fusion constructs

Substrate	NaCl concentration (mM)	Construct	App. initial rate (min−1)
5′ FAM*-labelled 50 nt ssDNA	50	OB	No DNA degradation
		BLACT	0.26 ± 0.02
		OB + BLACT	0.29 ± 0.02
		MBP-BLACT	0.30 ± 0.01
		OB-MBP-BLACT	1.85 ± 0.02
	50	BLACT	0.26 ± 0.01
	150		0.18 ± 0.01
	250		0.16 ± 0.01
	500		0.12 ± 0.01
	50	OB-MBP-BLACT	1.92 ± 0.01
	150		0.86 ± 0.02
	250		0.20 ± 0.02
	500		0.06 ± 0.01

The fluorescence intensity of the band corresponding to the original intact ssDNA substrate was quantified in gel-based time course nuclease activity assays. Apparent initial reaction rates were determined by linear regression of the data points that lie within the initial linear part of the curve, divided by the enzyme concentration. The percentage of intact substrate is relative to the t = 0 time point in the reaction. All reactions were performed as independent triplicates.

*FAM was attached to the first base (thymine, T) at the 5′ terminus.

Key residues in ComEC guide ssDNA towards the β-lactamase-like domain active site and into the competence channel

As alluded to previously, it is not known whether the β-lactamase-like domain of ComEC encounters ssDNA that has been separated prior to the nuclease step, or dsDNA from which the enzyme is capable of selectively degrading a single strand. It is known however, that most transformation events occur with linear dsDNA as the tDNA and our results also show that the OB fold displays slightly higher affinity for dsDNA. We hypothesized that the more likely scenario is that there are structural elements within ComEC that locally destabilize the double helix of dsDNA allowing the β-lactamase-like domain to cleave a single strand, while the undegraded strand is guided towards the channel domain.

To identify putative structural elements and residues that play a role in destabilising the hydrogen bonding of the double helix and guiding of the separated strands, we carefully inspected our BLACT_Mg crystal structure and a structural prediction of full-length ComEC in complex with DNA [89] (Fig. 5A and B). We also considered that such functionally important residues should be highly conserved. We identified two highly conserved loops, spanning residues 517–527 (loop 1) and 682–694 (loop 2) in L. pneumophila ComEC, which could serve as ‘pin elements’ that would destabilize the double helix, akin to pins or wedges of helicase domains [102]. In addition, these residues may also be important to guide the 5′ strand towards the active site for degradation. Loop 1 contains a tyrosine (Y522) followed by a conserved phenylalanine (F523), while loop 2 contains a conserved arginine (R688) at its tip, followed by two conserved phenylalanines (F689 and F691). These aromatic residues could play a role in denaturing the DNA duplex through ring stacking interactions with the DNA bases and thus also guide the strand destined for degradation (5′ end) towards the active site of the nuclease domain. Moreover, we noticed that there are several other well-positioned aromatic residues, as well as some positively charged residues, that create a pathway between the OB fold and β-lactamase-like domain for the undegraded strand (3′ end) towards the entrance of the channel domain (Y108, Y140, Y154, W212, F331, Y373A). Figure 5B shows a structural prediction of ComEC in complex with DNA, where the dsDNA is beginning to show local destabilization of the double helix. We chose several residues and tested if their mutation to alanine would affect transformation in vivo in our L. pneumophila Lp02 system (Fig. 5C). These results showed that a double mutant in loop 2 (F689A/F691A) completely prevents transformation in vivo, whereas a similar double mutant in loop 1 (Y522A/F523A) did not result in decreased transformation. Mutation of F689 on its own resulted in a 2.3-fold reduction of the transformation efficiency. However, other single mutants within these loops did, in general, not produce transformation phenotypes on their own, which is consistent with observations of similar pin element single mutants in other systems (e.g. UvrD and PcrA helicases) also not being sufficient to produce a phenotype [103, 104]. It is difficult to unpick whether these phenotypes are the result of impaired DNA duplex melting, a loss of subsequent 5′ strand guiding towards the nuclease active site, or both. Nevertheless, this loop is ideally positioned to serve both these functions according to our structural prediction of ComEC in complex with DNA (Fig. 5B). Access to purified full-length ComEC in the future may open up the possibility of better deconvoluting these related functions.

Figure 5.

Key residues support DNA degradation and translocation by ComEC. (A) Structural prediction of full-length ComEC_Lp in complex with a DNA molecule composed of 12 bp complementary dsDNA and a 12 nt ssDNA 3′ overhang. Three different orientations of ComEC are displayed in ribbon representation with a transparent surface. (B) Close-up view showing candidate pin and DNA pathway lining residues chosen for mutagenesis. (C) Transformation efficiencies of parental Lp02, Lp02 ΔcomEC, and Lp02 ΔcomEC complemented by ectopic expression of WT and mutant versions of ComEC containing substitutions of residues in the regions indicated in (B). Mean transformation efficiencies of five independent biological replicates are shown with error bars representing the standard deviation. < d.l., below detection limit (d.l.) (average d.l. = 7.82 × 10⁻⁸). Statistical significance was determined on log-transformed data using an unpaired two-sided t-test with Welch’s correction [118]. The Lp02 strain complemented with WT ComEC was compared to those strains complemented with ComEC mutants. #, below d.l. in at least one replicate; n.s., not statistically significant, P >.05; *, P <.05 (P_Y522A/F523A =.01, P_F689A=.047); ***, P <.001 (P_F689A/F691A =.0003, P_W212A=.0002, P_Y373A<.0001).

We also identified two mutations of channel lining residues within the competence domain (W212A and Y373A) that reduced the transformation efficiency by 4.9-fold and 5.6-fold, respectively. However, mutation of channel lining residues within the OB fold either did not affect transformation, or reduced transformation efficiencies modestly. It is not surprising that most single mutations of channel lining residues do not produce a pronounced phenotype, given the total number of residues that contribute to this aromatically lined DNA tunnel. In summary, we believe that the conserved loop 2 is a structural element present within ComEC that orients the 5′ strand towards the active site and that might additionally act as a pin to destabilize the DNA duplex. Furthermore, a staircase of aromatic and charged residues subsequently guides the 3′ strand towards and through the channel domain.

Working model for ComEC

Our in vitro characterization of the OB_Mg and BLACT_Mg, combined with key observations made in vivo allow us to propose a working model for DNA binding, degradation and translocation by ComEC (Fig. 6). In our model, the OB fold first binds to dsDNA with high affinity. We showed that the residues involved in DNA binding are solvent-exposed and positioned, in the context of the structural prediction of full-length ComEC, in a manner that would seamlessly facilitate DNA capture and subsequent steps (Supplementary Fig. S11). OB fold binding to DNA is necessary as the β-lactamase-like domain on its own does not appear to bind efficiently. We propose that the DNA duplex may be locally destabilized through a structural pin element that allows strand separation to occur. The strand leading with its 5′ end would subsequently be selectively cleaved by the β-lactamase-like domain (Fig. 2). The inherent directionality of DNA degradation by the β-lactamase-like domain will thus establish the polarity of DNA translocation through the competence domain. This is in line with previous work that showed the strand leading with the 3′ end is transported into the cytoplasm [105]. There is no ATP consumed by ComEC for the translocation of DNA. Given that phosphodiester bond cleavage is energetically favourable and assuming that the β-lactamase-like domain will continue to cleave the 5′ strand, the growing single stranded portion of the 3′ strand is guided by the channel lining residues and threaded into the competence domain. The aromatic and charged residues that line the ssDNA channel likely guide the undegraded DNA strand via ring stacking and electrostatic interactions. Once the DNA emerges on the cytoplasmic side of the membrane, other players like ComFA (and potentially PriA in Gram-negative bacteria), may engage the emerging ssDNA and translocate along it in an ATP-dependent manner, thereby exerting a pulling force. A key feature of our model is the relative domain organization of ComEC. This defined topology ensures that only a single strand is selectively degraded, as nucleolytic attack of the other strand would require a 180 degree rotation of the β-lactamase-like domain with respect to the DNA, which is presumably held in place through interaction with the OB fold. This topological restraint, coupled with the threading and pulling of DNA through ComEC, theoretically also allows DNA degradation to proceed much more processively in vivo, explaining the observed rapid rates of DNA translocation of 80–100 nt/s in B. subtilis and S. pneumoniae [106, 107].

Figure 6.

Working model of DNA binding, degradation and translocation by ComEC. (A) Top: Possibilities of phosphodiester bond hydrolysis by BLACT_Mgin vitro shown for both ssDNA and dsDNA. Green shapes represent possible cleavage modes and configurations of BLACT_Mg, whereas red shapes depict those that cannot occur. All possible cleavage events, including endonucleolytic events, result in a 5′ phosphate and 3′ hydroxyl group as shown in circles. Bottom: The OB-MBP-BLACT fusion construct is able to cleave DNA more efficiently than BLACT alone. (B) Schematic showing key steps carried out by ComEC. 1: tDNA is bound tightly by the OB fold. 2: The 5′ strand is guided towards the nuclease active site by F689 and F691 and 5′-to-3′ exonucleolytic or endonucleolytic cleavage can occur. 3: residues that line the DNA pathway on the OB fold (Y108, Y140, and Y154) and the β-lactamase-like domain (Y522 and F523) guide the DNA towards and through the competence domain. The initial threading of the DNA towards and into the channel domain is likely driven by the activity of the β-lactamase-like domain (depicted by an arrow with a question mark). Once the 3′ strand emerges on the cytoplasmic side of the membrane, other proteins (not shown) can engage with DNA to exert a possible pulling force. The dashed box shows that possible cleavage events by the β-lactamase-like domain, depicted by blue shapes, are reduced in vivo due to the topological restraint of ComEC domains and the tDNA.

Discussion

Natural transformation has far-reaching consequences for bacterial evolution and the emergence of pathogenic strains. Central and essential to the process of DNA transport from the environment into the cell interior is the protein ComEC. Over the years, this protein has proved to be challenging to produce and detect, and as a consequence our mechanistic understanding of this critical step is severely lacking. Here, we characterized the two non-membrane domains of ComEC, the OB fold and the β-lactamase-like domain, to learn more about how ComEC binds to DNA, degrades the non-translocating strand and ultimately translocates the remaining strand. Based on our investigations, we propose a model of how ComEC transports the transforming strand across the cytoplasmic membrane of competent bacteria.

We determined the crystal structure of the β-lactamase-like domain from M. glycerini (BLACT_Mg) (Fig. 1). Like its close relative from M. thermoacetica [53], this enzyme degrades DNA substrates in a manganese-dependent manner (Fig. 1C and Supplementary Fig. S2). Our structure, in combination with careful comparison of structures of related MBL family proteins, allowed us to propose the catalytic mechanism of this nuclease (Fig. 1D). Mutation of key residues that, either prevent Mn²⁺ coordination, or interfere with substrate binding, lead to a reduction or loss of nuclease activity in vitro (Supplementary Fig. S5). In our in vivo transformation assay, all mutations that reduced nuclease activity in vitro completely abrogated transformation, except for S728 which is not involved in metal ion coordination (Supplementary Fig. S5). Such an all-or-nothing phenotype highlights the importance of efficient DNA degradation by the β-lactamase-like domain for successful DNA translocation across ComEC. It remains to be seen, whether the β-lactamase-like domains from other competent species are dependent on Mn²⁺ coordination, or whether some of these enzymes may indeed bind to other metal cofactors such as Zn²⁺, as was predicted in earlier studies [29].

Nucleases often possess both endo- and exonuclease activity [108]. A close structural homologue of BLACT_Mg is RNase J, which can also cleave its RNA substrate in both modes, and indeed switch its propensity towards one mode depending on the nature of the 5′ modification of RNA or the enzyme’s dimerization status [61, 62]. Another member of the MBL family, Artemis, can process different DNA fragments by adjusting its binding site depending on the substrate [109]. We investigated whether BLACT_Mg shows an intrinsic preference towards cleaving single- or double-stranded DNA substrates, whether or not the presence of a 5′ phosphate or hydroxyl group is preferred and the directionality of its exonuclease activity (Fig. 2). These results showed a clear preference for ssDNA and revealed that BLACT_Mg can also hydrolyse in both an endo- and exonucleolytic fashion, the latter occurring exclusively in the 5′-to-3′ direction. This would suggest that when the β-lactamase-like domain encounters DNA in vivo, that the strand leading with its 5′ end is degraded, while the 3′ strand is free to translocate. Therefore, the observed 3′-to-5′ directionality of DNA translocation through ComEC [105] can be explained mechanistically by the enzyme’s inherent directionality of DNA degradation.

Whether or not both endo- and exonuclease activities are required during DNA translocation in vivo remains to be further investigated. However, we believe it is conceivable that the β-lactamase-like domain could operate in a mixed exo- and endonucleolytic manner, and that such endonucleolytic cleavage events may occur sporadically as the enzyme skips one or several bonds once the tDNA has become engaged inside ComEC, thereby releasing units that are longer than a single nucleotide. Structurally, there is sufficient solvent space around the active site to permit endo-cleavage events that would result in the release of such longer units. Indeed, there is evidence for this from earlier work that showed that di- and tri-nucleotides are released during this process in B. subtilis and S. pneumoniae [110]. Alternatively, the endonuclease activity might serve to create ends that can be engaged by ComEC. This would either require rotation of the β-lactamase-like domain or an end-to-end rotation of the DNA. The former is impossible due to the steric constraint of the nuclease domain within the context of full-length ComEC, while the latter is highly unlikely as the exact same location on the opposite strand would need to be cleaved, following dissociation, rotation and rebinding, to create a blunt end.

Curiously, BLACT_Mg on its own does not display measurable DNA-binding activity (Supplementary Fig. S12), and our data show that the OB fold provides this critical function (Fig. 3). OB_Mg binds to DNA tightly (K_D = 0.33 μM for dsDNA) using several positively charged residues as demonstrated by the drastic effect of ionic strength on binding. Yet substitutions of residues that contribute to binding (single and triple alanine substitutions) do not reduce affinity greatly. It appears as though the OB fold of ComEC has evolved as a robust DNA binding module that cannot easily be perturbed by single amino acid substitutions.

In the absence of an auxiliary DNA-binding domain such as a β-CASP domain, the OB fold thus fulfils this function (Fig. 4). This was demonstrated by the approximately seven-fold increase in the apparent initial rate of DNA degradation when OB_Mg was fused to BLACT_Mg on the same polypeptide chain and the dependence of DNA degradation on ionic strength, mirroring the OB_Mg-DNA binding experiments. It does not appear that OB_Mg plays a role in the correct positioning of the substrate with respect to the BLACT_Mg active site in our fusion construct in vitro, since the increased DNA degradation activity can also be achieved by other DNA-binding domains. However, at present we cannot determine if the precise domain orientation within the native ComEC protein in vivo, could potentiate activity further still. Structural alignment of the MBL domains within RNase J and the predicted structure of full-length ComEC reveal the different relative positions of the OB fold and the β-CASP domain with respect to the nuclease domain (Supplementary Fig. S14), suggesting that the route taken by the RNA/DNA substrate to the nuclease domain active site differs between these two proteins. In RNase J from S. coelicolor (PDB ID: 5A0T) [61] and Deinococcus radiodurans (PDB ID: 4XWW) [62] the 5′ phosphate of the RNA substrate is coordinated by a conserved serine (S375 and S379, respectively). The S728A variant of BLACT_Mg (S662 in L. pneumophila) showed reduced nuclease activity (Supplementary Fig. S5), presumably because of its role in substrate coordination. However, in the absence of a substrate bound complex structure and due to the different positioning of the OB fold relative to the β-lactamase-like domain within ComEC compared to the domain arrangement of RNase J, it is difficult to precisely pinpoint further substrate coordinating residues. Yet it is precisely this alternative placement of the OB fold within ComEC that likely creates the DNA tunnel that will ultimately guide the nondegraded ssDNA towards the competence domain.

To more precisely understand how DNA might be bound and subsequently encountered by the active site of the β-lactamase-like domain, we performed mutagenesis and transformation experiments guided by structural predictions and our own data (Fig. 5). We identified an essential loop strategically positioned at the entrance of the DNA channel that might contribute to destabilising the DNA duplex and guiding the 5′ leading strand to the β-lactamase-like domain for degradation and the 3′ leading strand to the competence domain for translocation. We showed that substitution of two conserved phenylalanines (F689 and F691) on this loop completely abrogates transformation in L. pneumophila. Positioned in close proximity to the conserved S662 it is likely that those are further substrate coordinating residues. We propose that this loop might additionally act as a pin loop responsible for melting and separation of duplex DNA prior to downstream processing. Such a mechanism is commonly employed by helicases [102, 103], as well as other DNA-processing proteins such as T7 RNA polymerase [111], which use similar structural elements. Initially this process in ComEC would not be powered by ATP hydrolysis, but would likely be driven by the activity of the β-lactamase-like domain. Because one of the two DNA strands is immediately degraded, no torsional backpressure or overwinding problems would occur. As cleavage progresses, the ssDNA portion of the strand destined for translocation ‘grows’, which then takes the path of least resistance and is threaded into the competence domain. To this end, the relative domain organization of OB fold, β-lactamase-like domain and competence domain ensure that a clear pathway is established and the DNA is prevented from taking an alternative route. The channel lining aromatic residues help to guide the ssDNA through this pathway. Indeed, the putative DNA channel within ComEC is lined with several well-positioned and highly conserved aromatic, as well as some charged, residues (Fig. 5, 6), which is a hallmark of proteins that bind to, stabilize or translocate DNA in some manner. For example, such a helical gateway is observed in the RecJ nuclease, which is important for processive degradation of ssDNA by this enzyme [112]. Interestingly, this protein also contains an OB fold which is important for DNA binding, and its relative position with respect to the nuclease seems equally critical. In the case of natural transformation, once ssDNA emerges on the cytoplasmic side of the membrane and is engaged by an ATP-dependent DNA translocase (ComFA and potentially PriA in Gram-negative organisms), unwinding, degradation and translocation may be further accelerated.

Many proteins involved in processive reactions of RNA or DNA metabolism achieve processivity via the topological linkage model, where ring-shaped, oligomeric proteins or protein complexes encircle their linear nucleic acid substrate (examples [113–117]). Among these examples are helicases, the sliding clamp β (a processivity factor of DNA polymerase III) and lambda exonuclease. The latter is a homotrimeric enzyme with a central DNA channel that processively cleaves one strand of a dsDNA substrate by virtue of its funnel-shaped channel that is wide enough to encircle dsDNA at one end, but can only fit ssDNA at the other [117]. In our model of ComEC, the relative domain organization imparts a topological restraint to the system, which we believe is crucial for the β-lactamase-like domain to cleave DNA processively in vivo (Fig. 6). We term this strand-specific topological processivity. Since a 180° rotation of the DNA along its helical axis would not lead to the correct internal directionality and the β-lactamase-like domain cannot diffuse and rotate 180°, endonucleolytic attack of the translocating strand leading with its 3′ end cannot occur, thus sparing it from degradation. This allows the nuclease domain to engage and processively cleave the 5′ strand. Furthermore, the uncleaved strand is subsequently stabilized and guided towards the channel domain of ComEC. According to structural predictions, this channel would not be wide enough to accommodate dsDNA. Further work is required to precisely understand the energetics of DNA duplex separation, DNA degradation and ssDNA translocation.

In summary, our work allows us to propose a working model for how ComEC binds, degrades and translocates DNA through the membrane during transformation. This is an important step towards developing a complete mechanistic model of this essential protein, which has until now remained poorly characterized.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was funded by a Swiss National Science Foundation (SNSF) PRIMA grant (PR00P3_179728). Funding to pay the Open Access publication charges for this article was provided by SNSF.

Data availability

The data underlying this article are available in the article and in its online supplementary material. Coordinates and structure factors for BLACT_Mg have been deposited in the Protein Data Bank (PDB ID: 9IC4). NMR spectra and corresponding model coordinates of OB_Mg have been deposited in the BioMag Resonance Data Bank (BMRB: 34982) and Protein Data Bank (PDB ID: 9IEW), respectively.