The wild-type B. bacteriovorus strain HD100 (Rendulic et al., 2004) and the axenic B. bacteriovorus strain HI (Roschanski et al., 2011) were used in this study. B. bacteriovorus HD100 was grown at 30°C by predation on Escherichia coli S-17 in HEPES buffer (25 mM HEPES, 2 mM CaCl₂, 3 mM MgCl₂, pH 7.8) and 200 rpm, or on double-layer plates [bottom layer – YPSC medium (0.1% Yeast Extract, 0.1% Pepton, 0.05% Sodium Acetate, 0.025% Magnesium Sulfate; pH 7.6] with 1% agar, top layer – YPSC with 0.6% agar and supplemented with E. coli S-17 liquid culture, both layers were supplemented with 0.025% CaCl₂ after autoclaving). E. coli was grown in LB medium (liquid or agar) at 37°C. B. bacteriovorus HI was grown in PYE medium (1% Bacto Peptone, 0.3% yeast extract, 2 mM CaCl₂, 3 mM MgCl₂, pH 7.6) at 30°C and 200 rpm.

The oriC-type replication origins in the genomes of B. bacteriovorus HD100 [GenBank entry BX842601.2], B. bacteriovorus str. Tiberius [GenBank entry CP002930.1], Bdellovibrio exovorus JSS [GenBank entry CP003537.1], and Halobacteriovorax marinus SJ [GenBank entry FQ312005.1] were predicted using the following stepwise procedure: (1) The annotation of the dnaA gene in the genome was validated by TBLASTN (version 2.2.30) (Shiryev et al., 2007) using the DnaA sequence of E. coli K-12 MG1655 [GenBank entry AAC76725.1] as a query. (2) The approximate genomic location of oriC was roughly determined based on the inflection point (minimum) of the genome’s cumulative GC-skew, which was obtained from the Comparative Genometrics website (Roten et al., 2002) or the GenSkew webserver¹ with the following parameter settings: Nucleotide1: G, Nucleotide2: C, Windowsize: 500, Stepsize: 100. (3) WebSIDD (Bi and Benham, 2004)² was used under default settings (37°C, 0.1 M salt, circular DNA, copolymeric) and negative superhelicity values in the range of σ = -0.04 (low) to σ = -0.06 (high) in increments of 0.005, in order to identify putative DUE(s) (Kowalski and Eddy, 1989) in intergenic regions near (±10 kb) of the GC-skew inflection point (minimum). (4) DnaA boxes were assigned manually using the E. coli consensus, 5′-TTWTNCACA (Schaper and Messer, 1995), and allowing for three mismatches. (5) A prediction was considered significant if a DnaA box could be assigned to a position of approximately two helical turns distant from the border of a strong DUE. Prediction output data were obtained as raw text files and further processed with Microsoft Excel v97SR-1 and Corel Draw v.11.

The B. bacteriovorus dnaA gene was PCR amplified from chromosomal DNA with primers P-1 and P-2 (Supplementary Table S1), cut with BamHI and XhoI and then cloned into the pET28a(+) expression vector linearized with the same restriction enzymes. The 6HisBdDnaA protein was produced in E. coli BL21 containing pET28a(+)BddnaA. When the culture reached an OD₆₀₀ = 1.9, fusion protein synthesis was induced by addition of 1 mM IPTG, after which cells were incubated for 3 h at 37°C. The bacteria were harvested by centrifugation (10 min, 5000 g, 4°C) and the bacterial pellets were stored at -20°C. The purification of 6HisBdDnaA was performed as described previously (Zawilak-Pawlik et al., 2006), except that LG₁₀₀ buffer (45 mM HEPES/KOH, pH 7.6, 100 mM potassium glutamate, 10 mM magnesium acetate, 1 mM DTT and 20% sucrose) was used in place of the LG₂₀₀ buffer.

The interactions of the DnaA protein with DNA were analyzed as previously described (Zawilak et al., 2001; Donczew et al., 2015) with minor modifications. The IRD-700-labeled BdoriC fragment (12 fmol, 623 bp, PCR amplified using primers P-5/P-4 and the pOCBdoriC plasmid as the template) and an IRD-700-labeled control DNA fragment (12 fmol, 620 bp, PCR-amplified using primers P-5/P-14 and pOCBd2045 as the template) were incubated with recombinant DnaA proteins of B. bacteriovorus (BdDnaA), E. coli (EcDnaA), and Pseudomonas putida (PpDnaA) at 30°C for 20 min in Marians’ binding buffer [20 mM HEPES/KOH, pH 8.0, 5 mM magnesium acetate, 1 mM EDTA, 4 mM DTT, 0.2% Triton X-100, 100 μM ATP, and 100 μg ml^-1 bovine serum albumin (BSA)]. The reaction was carried out in the presence of a non-specific competitor [poly(dA-dC)•poly(dG-dT), 50 ng; Sigma, P0307]. The formed complexes were chilled on ice for 2 min and separated by electrophoresis (5 V/cm) on 4% polyacrylamide gels in 0.5× TBE (89 mM Tris, 89 mM borate, 1 mM EDTA) at 20°C. The gels were analyzed using an Odyssey CLx Infrared Imaging System and the Image Studio software (Li-Core Biosciences).

For surface plasmon resonance (SPR) analysis, a 652-bp BdoriC fragment was PCR amplified with biotinylated primer P-6 and non-biotinylated primer P-3, and immobilized on the chip surface (Sensor Chip SA) in a BIAcore T200 apparatus. Approximately, 100 response units (RUs) of DNA were immobilized. A non-DnaA-box DNA fragment (649 bp, PCR amplified using primers P-6 and P-13) was used as a negative control. Measurements were performed in HKM buffer (25 mM HEPES, pH 7.6, 100 mM potassium acetate, 1 mM magnesium acetate, 0.005% Tween 20) (Pei et al., 2007) in the presence of the DNA competitor, poly(dA-dC)•poly(dG-dT) (final concentration, 50 μg/ml) at a continuous flow rate of 15 μl min^-1. At the end of each cycle (180 s association followed by 90 s dissociation), the bound proteins were removed by washing with 0.05% (w/v) SDS for 20 s, and the flow channels were equilibrated with HKM buffer until the baseline was stable. The data were analyzed using the BIA evaluation 3.0 software program.

The P1 nuclease assay was performed as previously described (Donczew et al., 2012). The pOCBdoriC plasmid (112 pmol) was incubated with DnaA proteins (0, 17.5, 35, 70, and 140 pmol from B. bacteriovorus, E. coli, or P. putida), and the presence of unwound DNA was examined by P1 treatment followed by digestion with SspI. The digestion products were visualized on a 1% agarose gel using a Molecular Imager^® Gel Doc^TM XR+ System and the Image Lab Software (Bio-Rad).

The E. coli strains, WM1785 and its polA derivative, WM1838 (polA^-, fadA::Tnl0), were used as host strains in the ori assay (Woelker and Messer, 1993). Chemically competent WM1785 and WM1838 cells were heat-shock-transformed using 50 ng of the appropriate plasmid (pBR322, pOC170, pBR322BdoriC, or pOCBdoriCΔori). The transformed cells were then cultivated on agar plates with tetracycline (12.5 μg/ml; for pBR322 and pBR322BdoriC) or ampicillin (100 μg/ml; for pOC170 and pOCBdoriCΔori) overnight at 30°C (for WM1838) or 37°C (for WM1785).

DNA modification with dimethyl sulfate (DMS) was performed as previously described (Sasse-Dwight and Gralla, 1991; Donczew et al., 2015). The reaction mixtures (50 μl) contained 25 mM HEPES/KOH, pH 7.6, 12% (v/v) glycerol, 1 mM CaCl₂, 0.2 mM EDTA, 5 mM ATP, 0.1 mg/ml BSA, 15 nM pOCBdoriC, and 6HisBdDnaA protein (0, 200, 400, 800, or 1600 nM). After the mixtures were incubated at 30°C for 10 min, 3.6 μl of 150 mM DMS (Sigma) was added to a final concentration of 10 mM, and the incubation was continued for 5 min. The reaction was quenched by the addition of 100 μl of cold Stop Buffer (3 M ammonium acetate, 1 M 2-mercaptoethanol, 20 mM EDTA). The samples were precipitated with cold ethanol, dried, dissolved in 100 μl of 1 M piperidine, and incubated at 90°C for 30 min. DNA was purified by gel filtration on Sephacryl S500 (Sigma) spin columns equilibrated in molecular-grade water. The DMS modification pattern was monitored by PE [primer extension (PE)] analysis using primers P-7, P-8, P-9, and P-10. For each PE reaction, 0.3 units of Taq DNA polymerase (Thermo Scientific), 20 fmol of DNA template, and 350 fmol of ³²P-labeled primer were used. PE was performed using 30 cycles of 30 s at 95°C, 30 s at 55°C, and 60 s at 72°C. The samples were then separated on a 8% polyacrylamide gel under denaturing conditions and scanned with a Typhoon 8600 Variable Mode Imager (GE Healthcare).

Replication initiation point (RIP) mapping was performed essentially as previously described (Gerbi and Bielinsky, 1997; Bielinsky and Gerbi, 1999; Matsunaga et al., 2003; Donczew et al., 2012). B. bacteriovorus cells were grown in 1000 ml HEPES buffer supplemented with P. putida cells (OD₆₀₀ = 1.0), and P. putida cells were grown in LB medium (OD₆₀₀ = 1.0). B. bacteriovorus were grown until the solution became viscous and slightly clear, whereupon the medium was passed through a 0.45-μm filter and then pelleted. The bacterial pellets were resuspended in 30 ml of TEN buffer (50 mM Tris-HCl, pH 8.0, 50 mM EDTA, 100 mM NaCl) and disrupted by the addition of sodium dodecyl sulfate (SDS) and sodium sarcosyl (final concentration, 1% each). The mixture was subjected to three-step extraction with phenol:chloroform:isoamyl alcohol (25:24:1, v/v) and after that 1.1 g/ml CsCl and 6 μl Midori Green Advanced DNA Stain (Nippon) were added to the aqueous phases. The genomic DNA was purified by CsCl gradient ultracentrifugation. To enrich the replication intermediates, the total isolated DNAs (75 μg for B. bacteriovorus and 426 μg for P. putida) were passed through BND-cellulose columns (Sigma-Aldrich) pre-equilibrated with NET buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA and 1 M NaCl). The columns were washed with five volumes of NET buffer, and DNA was eluted at 50°C with NET buffer containing 1.8% caffeine. To remove nicked DNA, the recovered DNAs (48 μg for B. bacteriovorus and 58 μg for P. putida) were subjected to phosphorylation by T4 kinase (Thermo Scientific) followed by λ-exonuclease (Thermo Scientific) digestion. The PE reactions contained 1 unit of vent (exo-) DNA polymerase (Thermo Scientific), 0.6 μg of prepared DNA, and 350 fmol of ³²P-labeled primer P-19. After 35 cycles of reaction (30 s at 95°C, 30 s at 55°C, and 60 s at 72°C), the amplified products were separated on an 8% polyacrylamide gel under denaturing conditions and analyzed with a Typhoon FLA 9500 Biomolecular Imager (GE Healthcare).

Immunoprecipitation assays were performed as described elsewhere (Jakimowicz et al., 2002). Briefly, B. bacteriovorus HI cells were grown to OD = 1.0 in 40 ml of PYE medium, and then formaldehyde was add to final concentration 1% (v/v) and the samples were incubated for 30 min. Anti-6HisBdDnaA polyclonal antibodies (ProteoGenix) were used to precipitate BdDnaA-DNA nucleoprotein complexes, and PCR was used to amplify regions of interest (primers P-3/P-4 for oriC and primers P-11/P-12 for non-box DNA). The PCR fragments were resolved on 1.5% agarose gels and analyzed using a Gel Doc^TM XR+ Imaging System (Bio Rad).

To identify the oriC of B. bacteriovorus, we employed different in silico tools, including analysis of gene arrangement, GC-skew analysis, identification of DnaA boxes, and the WebSIDD tool (SIDD, stress-induced DNA duplex destabilization) for localizing the DNA-unwinding element (DUE) (for details see Materials and Methods). Similar to the previous findings of Gao et al. (2013), we obtained predictions for oriC-type replication origins within the ∼250-bp-long dnaA-dnaN intergenic regions of the genomes of B. bacteriovorus (HD100 and str. Tiberius strains) and other species belonging to this genus, including Halobacteriovorax marinus SJ and B. exovorus JSS (B. exovorus JSS is not yet included in the DoriC data set created by Gao et al., 2013). The predicted Bdellovibrio oriC region contains eight putative DnaA boxes (see Supplementary Figure S1; Figure 6). Interestingly, two of the putative DnaA boxes (boxes 7 and 8; Supplementary Figure S1; Figure 6) are unusually situated within the dnaA gene, at its 3′-end. The position of the DUE could be readily derived from the SIDD plots generated for all of the analyzed organisms, with the exception of B. exovorus JSS. As seen for other bacterial origins, the first in the cluster of DnaA boxes could be assigned to a position approximately two helical turns distant from the border of a strong SIDD site (see Supplementary Figure S1; Figure 6). In the case of B. exovorus JSS, we were able to assign a “predicted oriC” based on the high similarity of DnaA box distances and orientations in this organism and the two B. bacteriovorus strains for which we were able to predict oriCs. In all four genomes, the predicted oriC is flanked upstream by the dnaA, rpmH, rnpA, and yidC (oxaA) genes, and downstream by the dnaN, recF, gyrB, and gyrA genes. This particular oriC gene context is also found in many genomes from the Actinobacteria and Firmicutes (Ogasawara et al., 1985).

In sum, the in silico-predicted B. bacteriovorus oriC region contains origin-characteristic elements (DnaA boxes and a DUE) and is located between the dnaA and dnaN genes within the gene cluster of rnpA-rpmH-dnaA-dnaN-recF-gyrB-gyrA, which is conserved in some bacterial species.

To determine whether the in silico-predicted B. bacteriovorus oriC region is bound in vivo by the initiator BdDnaA protein, we performed immunoprecipitation assays using antibodies against the purified 6HisBdDnaA protein (Supplementary Figure S2). The formaldehyde cross-linked BdDnaA-BdoriC complexes formed in B. bacteriovorus grown under host-independent conditions were enriched by affinity chromatography, and the released DNA fragments were identified by PCR (Figure 1A). We obtained strong PCR signals using primers for the BdoriC fragment but not a non-box DnaA fragment (Figure 1A), indicating that BdDnaA-BdoriC complexes were successfully detected.

To examine whether the BdDnaA protein interacts with BdoriC in vitro, we applied electrophoretic mobility shift assays (EMSAs) and SPR. Our EMSAs demonstrated that the BdoriC region, but not the non-DnaA box fragment, was bound specifically by the BdDnaA protein (Figure 1B). Interestingly, although the BdoriC contains eight putative DnaA boxes, only one nucleoprotein complex was observed. Moreover, increasing the protein concentration did not lead to the formation of additional higher-molecular-weight complexes. The interaction between BdDnaA and BdoriC was also confirmed by SPR analysis (Figure 1C), which showed that the RU values were proportional to the BdDnaA concentration.

Together, these findings indicate that the in silico-predicted B. bacteriovorus oriC is bound specifically by the initiator protein, BdDnaA, in vitro and in vivo.

To further define the in silico-predicted DnaA boxes and gain additional insight into their abilities to bind the BdDnaA protein, we applied footprinting experiments using DMS. This agent primarily methylates deoxyguanosine residues, making the proximate phosphodiester bond susceptible to piperidine cleavage (Maxam and Gilbert, 1977). The pOCBdoriC plasmid, which contained the entire oriC region, was incubated with increasing concentrations of BdDnaA protein, and then subjected to DMS modification and subsequent PE of piperidine-cleaved DNA. We identified 16 nucleotides that exhibited BdDnaA-dependent protection from DMS modification: eight Gs on the upper strand and eight Gs on the lower strand (Figure 2). Fourteen of them are located within the seven of eight in silico-predicted DnaA boxes, while the remaining two lie between in silico-predicted DnaA boxes (box; Figure 2 and see also Figure 6). Thus, the DMS footprinting confirmed the results of our in silico analysis and allowed us to identify additional DnaA-binding sites. Moreover, this analysis showed that the two unusually located DnaA boxes, boxes 7 and 8 (Figure 6A), are bound by the BdDnaA protein (Figure 2D).

Based on the assumption that, as in other bacteria, the DnaA box sequence of B. bacteriovorus consists of nine nucleotides, we aligned the DNA sequences in the vicinity of the protected nucleotides and obtained a proposed consensus sequence for the BdDnaA-binding motif, 5′-NN(A/T)TCCACA-3′, which we designated the DnaA box (Figure 6B; Supplementary Table S2).

Collectively, these analyses show that BdDnaA specifically binds to eight sites (Figures 2A–D) within the identified BdoriC region of B. bacteriovorus.

Interestingly, the binding mode of the E. coli DnaA protein to the oriC of E. coli (EcoriC) (Weigel et al., 1997) appears to differ from the interaction between B. bacteriovorus BdDnaA and the BdoriC. In contrast to B. bacteriovorus, in which only a single nucleoprotein complex was formed (Figure 1B), the interaction of E. coli and P. putida DnaAs with the BdoriC region yielded multiple discrete nucleoprotein complexes that formed a ladder of retarded bands on the gel indicating that the DnaA boxes were sequentially bound by the DnaA proteins (Figure 3). This difference in orisome formation prompted us to question how the arrangement of DnaA boxes and/or the properties of DnaA influence the formation of nucleoprotein complexes, and whether the formation of a single BdDnaA-BdoriC complex is specific to B. bacteriovorus. To answer these interesting questions, we analyzed orisome formation in heterologous systems. The BdoriC region from B. bacteriovorus was incubated with DnaA proteins from prey organisms (E. coli or P. putida) and the formed nucleoprotein complexes were analyzed using EMSAs. In these in vitro heterologous systems (as in the homologous system; Figure 1B), the observed nucleoprotein complexes formed in a protein-concentration-dependent manner (Figure 3). Surprisingly, the DnaA proteins of the prey organisms exhibited higher affinities toward the B. bacteriovorus oriC than the BdDnaA from B. bacteriovorus toward its own BdoriC region. The BdoriC region from B. bacteriovorus was almost completely bound at the lowest tested concentration of DnaA from E. coli and P. putida (1 nM; Figure 3), whereas in the homologous system, the nucleoprotein complex was detectable only at B. bacteriovorus DnaA concentrations >3 nM (Figure 1B). As the concentrations of E. coli or P. putida DnaA proteins increased, the complexity of the band pattern increased until a critical point was reached, whereupon the ladder pattern was replaced by diffuse, highly retarded bands indicative of large complexes (Figure 3). Thus, whereas the B. bacteriovorus DnaA appears to exhibit a unique binding mode to the B. bacteriovorus oriC, the DnaA proteins of prey organisms bind their own and B. bacteriovorus oriC regions in a similar manner. We also observed that the oriC region from B. bacteriovorus was not specifically bound by DnaA proteins from non-prey organisms (e.g., Streptomyces coelicolor; data not shown).

Together, the results of in vitro analysis indicate that DnaA proteins from prey bacteria bind the BdoriC region specifically and with a high affinity, forming multiple nucleoprotein complexes.

To experimentally verify the in silico-predicted B. bacteriovorus DUE within BdoriC, we used a P1 nuclease assay. Supercoiled pOCBdoriC plasmids containing all of the predicted DnaA boxes (Supplementary Table S1) were incubated with increasing amounts of BdDnaA and subsequently treated with P1 nuclease, which hydrolyzes single-stranded DNA at the opened helix and hence linearizes the unwound plasmid (Donczew et al., 2012). Subsequent digestion with SspI allowed us to approximate the region unwound by BdDnaA. We detected faint bands providing evidence for DnaA-dependent DNA unwinding; the observed DNA fragments were ∼1.2 and 1.5 kb, indicating that the P1 hydrolysis site corresponded to the in silico-predicted DUE of B. bacteriovorus (Figure 4B1). We also observed an additional DNA fragment of about 2.3 kb (Figure 4A), likely corresponding to a DnaA-independent P1-sensitive site located within the plasmid origin of pBR322 (pOCBdoriC). It was previously suggested that the ori of pBR322 contains a helically unstable region (Kowalski et al., 1988; Donczew et al., 2012).

Since EcDnaA and PpDnaA strongly bound BdoriC, we examined whether these prey proteins could unwind BdoriC. Surprisingly, P1- and SspI-mediated digestion generated the same patterns obtained using BdDnaA (Figure 4B2,3), indicating that both proteins could in vitro unwind DNA within the BdoriC region. The efficiency of DNA unwinding was much higher for the prey proteins (particularly that of E. coli) than for BdDnaA (Figure 4B2,3). Moreover, when we performed P1 assays with an equimolar mixture of DnaA proteins from B. bacteriovorus and E. coli, strong bands were observed at a minimal protein concentration (35 nM; Figure 4B4) that was even lower than that found to yield a similar result with the E. coli protein alone (70 nM; Figure 4B2).

To precisely map the unwound region, we performed PE of P1-cleaved pOCBdoriC plasmids using Taq polymerase and ³²P-labeled primers flanking the in silico-predicted DUE (for details see Materials and Methods and Supplementary Table S1). Since BdDnaA alone yielded only faint signals in the P1 assay, we used an equimolar mixture of B. bacteriovorus and E. coli proteins (Figure 4C). The observed extension products confirmed that DNA unwinding occurs within the in silico-predicted DUE sequence at the 5′-end of the oriC region, and allowed us to estimate the unwound region as spanning ∼55 bp (Figures 4C and 6A). Moreover, RIP mapping showed that BdoriC is the replication initiation site in vivo (Figure 5).

Together, these results show that BdoriC is unwound at the 5′-end by its own BdDnaA, as well as by DnaA proteins from the prey species, E. coli and P. putida.

Since BdoriC was specifically bound and unwound by the DnaA from E. coli, we next tested whether the predator’s oriC could initiate replication in its prey. To investigate ability of BdoriC to promote replication in E. coli, we performed a set of heterologous transformations in which the pBR322 plasmid (a negative control) and its derivatives carrying BdoriC or EcoriC (a positive control) regions (Supplementary Table S1) were assayed for oriC-dependent initiation of replication in the E. coli polA^- strain (Langer et al., 1996; Zawilak-Pawlik et al., 2005). ColE1-type plasmids (such as pBR322) require DNA polymerase I for their replication; thus, only a construct containing a functional oriC region (and conferring pBR322-encoded ampicillin resistance, Amp^R) can replicate in the absence of DNA polymerase I (such as found in E. coli polA^-). Among the analyzed constructs, only the pOC170 plasmid carrying the EcoriC was replicative (Table 1), suggesting that the BdoriC region does not promote the initiation of replication in E. coli. Indeed, no plasmid containing the BdoriC region yielded ampicillin-resistant transformants in the polA-deficient strain, even after prolonged incubation.

Together, these results show that B. bacteriovorus oriC is not a substrate for replication in its prey organisms.