Correspondence: Renee E. Sockett, Institute of Genetics, School of Biology, Medical School, University of Nottingham, Derby Road, QMC, Nottingham, NG7 2UH, UK. Tel: +44 115 823 0325; fax: +44 115 823 0338; e-mail: email@example.com
Bdellovibrio bacteriovorus are small, vibroid, predatory bacteria that grow within the periplasmic space of a host Gram-negative bacterium. The intermediate-filament (IF)-like protein crescentin is a member of a broad class of IF-like, coiled-coil-repeat-proteins (CCRPs), discovered in Caulobacter crescentus, where it contributes to the vibroid cell shape. The B. bacteriovorus genome has a single ccrp gene encoding a protein with an unusually long, stutter-free, coiled-coil prediction; the inactivation of this did not alter the vibriod cell shape, but caused cell deformations, visualized as chiselled insets or dents, near the cell poles and a general ‘creased’ appearance, under the negative staining preparation used for electron microscopy, but not in unstained, frozen, hydrated cells. Bdellovibrio bacteriovorus expressing ‘teal’ fluorescent protein (mTFP), as a C-terminal tag on the wild-type Ccrp protein, did not deform under negative staining, suggesting that the function was not impaired. Localization of fluorescent Ccrp–mTFP showed some bias to the cell poles, independent of the cytoskeleton, as demonstrated by the addition of the MreB-specific inhibitor A22. We suggest that the Ccrp protein in B. bacteriovorus contributes as an underlying scaffold, similar to that described for the CCRP protein FilP in Streptomyces coelicolor, preventing cellular indentation, but not contributing to the vibroid shape of the B. bacteriovorus cells.
Bdellovibrio bacteriovorus are predatory bacteria that prey upon a wide range of Gram-negative bacteria (Lambert, 2006). To achieve this highly motile, vibroid, attack-phase B. bacteriovorus seek out, attach to and then squeeze through a small pore in the outer membrane of the prey, entering the prey periplasm (Abram et al., 1974). Previous work has shown that prey entry occurs by the action of type IV pili, and striking electron microscopic observations show that B. bacteriovorus cells locally contract around the site of prey entry. This contraction travels over the entire length of the cell (Abram et al., 1974; Evans et al., 2007; Mahmoud & Koval 2010). Once inside, the B. bacteriovorus round up the prey cell wall, forming a bdelloplast, growing within, using molecules acquired from ordered hydrolytic breakdown of the prey, elongating into either a long-vibroid or a coil-shaped growth-phase cell (Lambert, 2006). Once prey resources have been depleted, the growth-phase cell septates, forming multiple motile progeny that lyse the bdelloplast (Lambert, 2006).
Our previous work showed that the B. bacteriovorus cytoskeleton has adapted to the challenges presented by the unique predatory lifestyle of this bacterium, and showed that the two MreB homologues played differing roles in cell elongation within the bdelloplast (Fenton et al., 2010). Protein secondary-structure prediction software has identified a large family of cytoskeletal elements in bacteria, which are structurally similar to eukaryotic intermediate-filament (IF) proteins (Lupas et al., 1991; Lupas, 1996; Ausmees et al., 2003; Bagchi, 2008). Since their discovery in 2003, they have been shown to play varying roles in the bacterial cell architecture such as crescentin (CreS) in Caulobacter crescentus, which establishes and maintains its vibroid/coiled-cell shape; FilP in Streptomyces coelicolor plays a role in cell rigidity; and finally, in Helicobacter pylori, two IF-like proteins (Ccrp59 and Ccrp1143) play roles in maintaining cell morphology (Ausmees et al., 2003; Bagchi, 2008; Waidner et al., 2009). Here, we show that the B. bacteriovorus genome contains one predicted IF-like protein (CCRP) and we investigate its role in prey cell entry and in B. bacteriovorus cell morphology.
Materials and methods
Bacterial strains and growth media
A full list of the strains used in this study can be found in Table 1. Genome-sequenced strain B. bacteriovorus HD100 (Stolp & Starr, 1963; Rendulic, 2004) was used throughout this study, and was grown by predation on Escherichia coli S17-1 (Simon et al., 1983) in Ca/HEPES buffer using standard culturing methods described in Lambert et al. (2003). Ca/HEPES buffer supplemented with 50 μg mL−1 kanamycin (Kn) and kanamycin-resistant E. coli S17-1:pZMR100 prey were used to maintain B. bacteriovorus strains with genome-integrated kanamycin resistance cartridges (Rogers, 1986).
Table 1. Strains used in this study
Sources or references
E. coli S17-1
thi, pro, hsdR−, hsdM+, recA; integrated plasmid RP4-Tc∷Mu-Kn∷Tn7 used as a donor for the conjugation of plasmids into B. bacteriovorus
Donor E. coli strain containing the pSET151 suicide plasmid with inserted kanamycin-interrupted HD100 ccrp (Bd2697) ORF with 1 kb 5′- and 3′-flanking genomic DNA, used for gene interruption in B. bacteriovorus
HD100 strain with a kanamycin-interrupted ccrp gene (Bd2697)
B. bacteriovorus HD100 Bd2345∷Knr
HD100 strain with a kanamycin-interrupted ABC gene Bd2345. Knr cassette placed in the equivalent genome position of the B. bacteriovorus 109JK, which has no predatorily defective phenotype (Lambert et al., 2003)
B. bacteriovorus HD100 ccrp-mtfp
HD100 strain carrying integrated plasmid pAKF41a at the ccrp (Bd2697) locus
Inactivation of ccrp and Bd2345 genes in B. bacteriovorus
Gene interruptions by kanamycin cassette insertion into B. bacteriovorus HD100 were carried out as described previously (Lambert et al., 2003; Evans et al., 2007). Briefly, constructs were prepared by the amplification of a region of the HD100 genome containing either ccrp (Bd2697) or Bd2345 and 1 kb flanking genomic DNA, and were inserted into the pGEM7 vector (Promega); subsequent gene inactivation was achieved using kanamycin cassette insertion into the unique NruI site of the ccrp ORF and the EcoRV site of the Bd2345 ORF, and transferred into the mobilizable pSET151 plasmid (Bierman, 1992), forming the pAKF22 and pLH008 deletion constructs, respectively. These were then introduced into B. bacteriovorus cells by conjugation using the S17-1 donor strain described fully in Evans et al. (2007); candidate mutants were screened and gene knockout candidates were confirmed by Southern blot. We were able to isolate the ccrp mutant directly from a predatory host-dependent culture, without the need to go through host–prey-independent growth for selection.
Cryoelectron microscopic imaging
Sample preparations were carried out using the methods described in Borgnia et al. (2008). Images were taken on a Tecnai T12 transmission electron microscope (TEM).
Five microlitre droplets of bacterial cells were applied to holey carbon grids (Quantifoil MultiA; Micro Tools GmbH, Germany), previously glow discharged for about 30 s and coated, for scale, with 15 nm protein A–gold conjugates (BB International, Cardiff, UK). The grids were manually blotted and quenched in liquid ethane using a manual gravity plunger. Vitrified specimens were then transferred into an FEI Tecnai 12 TEM or a Tecnai Polara TEM (FEI Company, Hillsboro, OR). The images were recorded on a 2 × 2 k CCD camera, under low-dose conditions, at 120 kV (Tecnai 12) or 200 kV (Tecnai Polara). The nominal magnifications were in the range 6000–18 000 and 4–6 μm underfocus values.
Bdellovibrio bacteriovorus attack-phase cells were negatively stained using 0.5% uranyl acetate (URA) (Sigma), pH 4.0, for 30–45 s using the methods described elsewhere (Evans et al., 2007). Cells were observed at 100 kV using a JEOL JEM 1010 TEM.
Ccrp C-terminal mTFP fusion in B. bacteriovorus
C-terminal tagging of B. bacteriovorus proteins with a bright monomeric fluorescent protein mTFP was carried out as described previously in Fenton et al. 2010. In brief, C-terminal tagging of B. bacteriovorus Ccrp protein with mTFP was achieved by the amplification of a 927-bp fragment of the ccrp ORF from the HD100 genome, representing 76% of the entire ORF. Primer designs removed the stop codon of ccrp and introduced both EcoRI site and KpnI sites used to ligate the fragment in frame with mtfp; this construct was transferred into the mobilizable pK18mobsacB vector (Schafer et al., 1994), forming pAKF42a, and conjugated into B. bacteriovorus HD100 using the methods described previously (Evans et al., 2007). Single genome integration of pAKF42a into the HD100 genome, producing a fluorescent, in-frame fusion, was confirmed by Southern blotting and by direct sequencing of DNA from the genomes of the resultant fluorescent strains. When translated, the Ccrp–mTFP fusion protein has five linker amino acids bridging the two proteins with the sequence VPRSS.
Fluorescence microscopy and FM4-64 staining
Bdellovibrio bacteriovorus attack-phase cells were stained with the FM4-64 membrane stain (Invitrogen) at a final concentration of 10 μg mL−1 and incubated in the dark for 5 min before detection. FM4-64 stains the membranes of B. bacteriovorus, including the membranous flagellar sheath (Ai, 2006). Fluorescence and bright-field images were visualized on a Nikon Eclipse E600 epifluorescence microscope using a × 100 lens (NA: 1.25), with either CFP (excitation, 420–454 nm; emission, 458–500 nm) or hcRed (excitation, 550–600 nm; emission, 610–665 nm) filter blocks for the detection of mTFP and FM4-64 fluorescence, respectively. Images were acquired using a Hamamatsu Orca ER camera and analysed using iplab software, version 3.64. mTFP fluorescence images were background corrected using the 3D filter tool and normalized within the iplab software; FM4-64 images are displayed raw in Fig. 1d. MreB inhibitor S-(3,4-dichlorobenzyl)isothiourea (A22) was dissolved as a concentrated stock of 10 mg mL−1 in methanol and added to cells at concentrations from 1 to 100 μg mL−1 in comparison with methanol-only controls.
Results and discussion
Identification of an IF-like protein in B. bacteriovorus
IF-like proteins in bacteria have been identified using a protein secondary-structure prediction program coils (http://www.ch.embnet.org/software/COILS_form.html), which successfully predicts the characteristic coiled-coil domains found within these proteins (Lupas et al., 1991; Lupas, 1996). However, multiple proteins are known to contain regions of a coiled-coil structure without being IF elements (Bagchi, 2008; Graumann, 2009). A true IF protein homologue must have both a good coiled-coil prediction, and critically, no other predicted domains; it has been suggested that proteins fulfilling these criteria be named coiled-coil-rich-proteins (CCRP) (Bagchi, 2008; Graumann, 2009; Waidner et al., 2009). An exhaustive search of the B. bacteriovorus genome revealed one predicted CCRP protein encoded by the Bd2697 ORF. Therefore, we conclude that Bd2697 is the only structural IF-like gene in the B. bacteriovorus genome, hereafter called ccrp. Unusually for an IF protein, the coiled-coil prediction of this gene product did not have any recognizable ‘stutter’ regions, where coiled-coil prediction breaks down (Fig. 1a) (Lupas et al., 1991; Lupas, 1996; Bagchi, 2008).
Ccrp of B. bacteriovorus has limited homology, by wublast2 (http://blast.jcvi.org/cmr-blast/), to the CreS protein of Caulobacter (21% identity, 43% similarity, 1.5e-07) or to the FilP protein of Streptomyces (24% identity, 42% similarity, 7.2e-09). This low level of primary sequence homology is expected for CCRP-type proteins (and very poor sequence conservation is seen between the documented CCRP proteins crescentin and FilP) (Bagchi, 2008). In both cases, repeating E, A and R residues can be seen along the homologies to B. bacteriovorus Ccrp, probably as part of the coiled-coil motifs. Interestingly, homology was not significant with either protein at the N-terminus of Ccrp, indicating that the nature of attachment of the Ccrp at the N-terminus might differ, as the first 27 amino acids of CreS are required for membrane attachment (Cabeen, 2009). This is further discussed later.
In order to study the role of the ccrp gene in the B. bacteriovorus life cycle, a strain carrying a deletion of ccrp by kanamycin cassette insertion was constructed using the methods described previously (Fenton et al., 2010; Lambert et al., 2003). Deletion strains were examined by cryoelectron microscopy to determine whether their vibroid morphology had been altered by the mutation. Surprisingly, all cells of the ccrp∷Kn strain were vibroid in shape, as was the kanamycin-resistant Bd2345∷Kn control (Fig. 1b). In contrast to what has been concluded regarding the role of the CreS, CCRP protein in determining the shape of C. crescentus, we conclude that Ccrp does not maintain vibroid cell shape in B. bacteriovorus (Ausmees et al., 2003).
A larger number of ccrp∷Kn B. bacteriovorus cells were visualized for any morphological differences, in comparison with cells without a ccrp deletion, by negative staining of whole attack-phase cells with 0.5% URA, pH 4.0, for TEM (Fig. 1c). Interestingly, this revealed that, in contrast to the usual wild-type smooth appearance of all the Bd2345∷Kn control cells, all cells of the ccrp∷Kn strain had a dented and creased appearance, not seen previously (Fig. 1b, c). Negative staining of B. bacteriovorus cells with URA does not cause membrane damage observed with other stains used for electron microscopy, and positively stains the cell wall, causing slight cell folding in wild-type and Bd2345∷Kn control cells (Fig. 1c) (Abram & Davis, 1970). In contrast to control strains, the surfaces of the ccrp∷Kn strain are severely creased and turned inwards, creating deep indentations at both poles in 29% of the cells (n=191), a feature not seen either by light microscopy or by cryoelectron microscopy (Fig. 1c).
That this denting and deformation did not have an effect on cell viability was shown by the wild-type predatory rates of the ccrp∷Kn strain (measured by microscopic observation of the rates of E. coli prey bdelloplast formation and lysis and by the rate of OD600 nm decline of prey E. coli cells), its long-term survival at levels comparable to the wild type in buffer alone and its short-term survival during treatment with up to 0.1% glycerol, which was used to try to provide an osmotic challenge to the cells in case their response was altered (data not shown). The cell deformations described here are consistent with the work published on the IF-like protein FilP in S. coelicolor, which shows that CCRP proteins can act as an underlying protein scaffold contributing to cell rigidity, previously thought to be a function of the cell wall and turgor pressure (Bagchi, 2008).
Interestingly, the homology between Ccrp and FilP, mentioned in Identification of an IF-like protein in B. bacteriovorus, although weak, does include a conserved AQVD motif seen in FilP at amino acids 19–22 and in B. bacteriovorus Ccrp at amino acids 33–36. This motif, along with other extra amino acids, is shared by FilP family proteins, but not crescentin (Bagchi, 2008). Thus, Ccrp from B. bacteriovorus may have a more FilP-like nature than a crescentin-like nature.
Localization of a Ccrp–mTFP fusion protein in B. bacteriovorus
We showed previously that tagging of cellular proteins with a bright, monomeric, fluorescent protein, mTFP, in B. bacteriovorus could be used to determine cellular address and function (Fenton et al., 2010; Ai, 2006). A C-terminal ccrp–mtfp fusion was cloned and recombined, on several separate occasions, into the B. bacteriovorus genome using the methods described previously (Fenton et al., 2010). In contrast to reports on crescentin in C. crescentus, the Ccrp–mTFP fusion protein appeared to be fully functional, as the crushing and denting phenotypes revealed under negative staining of ccrp-deletion strains were never observed (data not shown) (Ausmees et al., 2003). The fluorescent Ccrp–mTFP signal in attack-phase B. bacteriovorus cells was generally evenly distributed, but showed a bias towards the cell poles (Fig. 1d). In only some cells could fainter more peripherally located thread-like, fluorescent regions be observed (Fig. 1d, A and B). Partitioning of the signal could be observed in some cells where there was a clear fluorescent signal bias to either pole (Fig. 1d, C). Using the FM4-64 membrane stain to visualize the sheathed flagellum, and thus determine cell polarity, showed that single Ccrp–mTFP foci could be located to either pole of the attack-phase cell (Fig. 1d, A and B) (Iida, 2009).
The helical filament of the actin-like protein MreB spatially organizes many proteins within the cytoplasm of diverse bacterial cells and helps anchor CreS fibres to the cell poles in C. crescentus (Charbon et al., 2009; Graumann, 2009). The presence of the MreB-specific inhibitor A22 (at a final concentration of 10 μg mL−1) did not significantly alter the Ccrp–mTFP distribution patterns in B. bacteriovorus attack-phase cells (data not shown). Previous work has shown that A22 concentrations of 10 μg mL−1 modify MreB activities in B. bacteriovorus without affecting the long-term viability (Fenton et al., 2010); in contrast to work carried out on CreS, our results suggest that the Ccrp–mTFP localization in B. bacteriovorus is independent of the MreB cytoskeleton (Charbon et al., 2009).
We conclude that the Ccrp–mTFP fusion protein in attack-phase B. bacteriovorus was predominantly evenly located throughout the cell and that the absence of Ccrp in cells caused a creased appearance by TEM. In some cells, the Ccrp–mTFP fusion protein showed a positional bias towards either B. bacteriovorus cell pole at frequencies that were similar to the cell denting bias observed for the ccrp-deletion strain under negatively stained TEM (Fig. 1c, d). The similarity of these two frequencies may suggest that the absence of the Ccrp in a portion of the mutant B. bacteriovorus population (where Ccrp would have been positioned near the poles in that fraction of wild-type cells) makes that portion of the population more susceptible to the insults of negative staining near the cell poles, producing the subpolar dents seen. In the case of crescentin, this scaffold protein has a submembranous peripheral location (Charbon et al., 2009), but we have not been able to confirm this for Ccrp, although the fluorescence tagging has provided some evidence (Fig. 1d). We are aware that the addition of a fluorescent tag to Ccrp may have affected its assembly at wild-type positions, and we note that in some of our fluorescent cells, fainter filamentous fluorescence can be visualized closer to the cell periphery (Fig. 1d, A and B). It is too early to say for sure whether these faint filamentous structures provide evidence of membrane-associated attachment of B. bacteriovorus Ccrp; this needs more detailed localization studies. Also, the lack of conservation of the approximately 30 N-terminal amino acids of Ccrp with either that of FilP or CreS makes predictions about localization impossible, as in CreS, at least the N-terminus is responsible for membrane association (Cabeen, 2009).
It was perhaps surprising that a protein that could localize to discrete foci in B. bacteriovorus cells, and whose absence produced large-scale denting of the cell surface, when visualized by negative staining, did not affect the entry of B. bacteriovorus into prey cells. As B. bacteriovorus entry to prey was previously shown to be dependent on type IV pili, it is clear that the structures that anchor the pilus retraction machinery at the cell pole do not require this Ccrp protein to be firmly positioned in the cell (Evans et al., 2007; Mahmoud & Koval, 2010). This is also the case for the cytoskeletal MreB proteins, whose alteration does have a marked effect on the subsequent progression of the predatory cycle, but not upon the initial invasion of prey (Fenton et al., 2010). The question as to whether cytoskeletal proteins or peptidoglycan interactions are key to allowing B. bacteriovorus cells to be dragged into prey by pilus retraction remains open. Our results suggest that while Ccrp in B. bacteriovorus does not contribute to the vibroid cell shape, it significantly contributes to the smoothness of the B. bacteriovorus cell shape by acting as an internal protein scaffold.
We thank C.J. Wagner for her initial identification of a coiled-coil containing protein in Bdellovibrio, Cezar Khusugaria for advice and assistance with the cryoelectron microscopy on the Tecnai Polara TEM and Marilyn Whitworth for technical assistance. This study was funded by a BBSRC PhD Studentship for A.K.F. to R.E.S., HFSP grant RGP57/2005 to R.E.S. for L.H. and NIH core funding to S.S. for C.B. A.K.F. carried out the majority of the experiments, designed parts of the experimental programme including the mTFP fusions and coauthored the paper. L.H. constructed the Bd2345∷Kn B. bacteriovorus control strain, assisted with TEM analysis and critically read the manuscript. C.B. and A.K.F. carried out cryoelectron microscopic analysis under the supervision of S.S., and R.E.S. designed the experimental programme, supervised the research, coauthored and revised the paper.