A Predator Unmasked: Life Cycle of Bdellovibrio bacteriovorus from a Genomic Perspective

See allHide authors and affiliations

Science  30 Jan 2004:
Vol. 303, Issue 5658, pp. 689-692
DOI: 10.1126/science.1093027


Predatory bacteria remain molecularly enigmatic, despite their presence in many microbial communities. Here we report the complete genome of Bdellovibrio bacteriovorus HD100, a predatory Gram-negative bacterium that invades and consumes other Gram-negative bacteria. Its surprisingly large genome shows no evidence of recent gene transfer from its prey. A plethora of paralogous gene families coding for enzymes, such as hydrolases and transporters, are used throughout the life cycle of B. bacteriovorus for prey entry, prey killing, and the uptake of complex molecules.

Bdellovibrio bacteriovorus is a highly motile delta-proteobacterium that preys on other Gram-negative bacteria (1). Bdellovibrio are ubiquitous in nature, having been discovered in a wide variety of environments that include both aquatic and terrestrial habitats as well as mammalian intestines (2). Their prey includes plant, animal, and human pathogens (3). Thus, studying the molecular mechanisms of Bdellovibrio's large complement of degradative enzymes gives insight into those targets in prey cells that have proven to be evolutionarily successful points of attack and offers cues for the design of antimicrobial agents.

Here we report the complete genome sequence of B. bacteriovorus strain HD100, which can only grow in the presence of prey. Despite the small dimensions of Bdellovibrio cells (0.2 to 0.5 μm wide and 0.5 to 2.5 μm long), its genome consists of 3,782,950 base pairs (bp) on a single circular chromosome and is predicted to code for 3584 proteins (fig. S1). No evidence was found for additional episomes, with the exception of a single copy of the novel IS element ISBba777 and a previously unknown Bdellovibrio prophage (Table 1). Despite having ready access to the prey's genetic information, no recent prey-derived horizontal gene transfer can be observed. Such a transfer from predator to prey should have resulted in regions in the B. bacteriovorus genome with GC content higher than the average, because the range of soil prey bacteria that Bdellovibrio was originally isolated from all have a higher than average GC content. The four regions of the genome that do deviate from the average GC content are high in AT and code for ribosomal, lipopolysaccharide (LPS) synthesis, prophage, and restriction modification genes, which in general are higher in AT in any bacterial genome (Table 1).

Table 1.

General features of the B. bacteriovorus HD100 genome (European Molecular Biology Laboratory accession no. BX842601).

Species B. bacteriovorus
    Strain HD100 (Deutsche Sammlung Mikroorganismen DSM50701)
Size 3,782,950 bp
    GC content 50.7%
    GC in coding areas 50.4%
Predicted number of ORFs 3584
    Coding sequences (CDS) similar to known proteins 1995
    Conserved hypothetical proteins 382
    Hypothetical proteins 1207
    Coding potential 93%
    Average CDS length 982 bp
rRNA operons 2
tRNA genes 36
Mobile genetic elements 1 IS element
4 transposases
Four regions of deviating GC content LPS synthesis
Prophage insertion in tRNAmet
Ribosomal gene cluster
Restriction modification system
CDS coding for hydrolytic enzymes 150 proteases/peptidases
20 DNases
9 RNases
10 glycanases
15 lipases
89 other

Of the predicted coding sequences, 55% can be assigned to a putative function by their homology to entries in the public database without suggesting a particular phylogenetic relationship to any of the completed microbial genomes. Eleven percent of the predicted open reading frames (ORFs) have homology to proteins of unknown function, whereas 34% cannot be assigned any function (Table 1, fig. S2, and table S1). In addition to the protein-coding genes, 36 transfer RNA (tRNA) genes were found, together with two ribosomal RNA (rRNA) clusters, each consisting of one copy of 5S, 16S, and 23S genes (Table 1).

Published experimental evidence [(3) and references therein] and the predicted gene content model the predatory life cycle of Bdellovibrio in eight stages (designated I through VIII) (fig. S3).

In attack phase (stage I), B. bacteriovorus swims at high speed using a single sheathed polar flagellum with a characteristic dampened filament waveform (4). Six clusters of motility and flagellar synthesis genes were found together with six copies of flagellin genes at four independent loci (supporting text). The detection of prey does not seem to depend on genes for the reception of homoserine-lactones generated by dense prey populations, nor does Bdellovibrio seem to be capable of producing known quorum-sensing compounds by itself.

Once B. bacteriovorus has collided with a prey cell, it remains reversibly attached to it for a short “recognition” period (5), after which it becomes irreversibly anchored via the pole opposite the flagellum (stage II). Multiple candidate genes for adhesion are found in the genome. Besides passive protein-protein and LPS-LPS interactions between outer membrane components, we suggest that active adhesion also occurs, based on the presence of multiple pili genes in Bdellovibrio. This mechanism combines adhesion and prey invasion. At least four clusters of pil genes were found on the chromosome (table S2), as were numerous dispersed pil genes coding for type IV pili (TFPs); this is the fourth highest number reported for any bacterial genome to date. TFPs or fimbriae perform multiple functions in many bacterial cells, such as protein secretion, adherence, DNA uptake, and twitching motility, depending on the composition of their central pilin fiber (6). Pilus tips are known to hold diverse sets of adhesive biopolymers that are specific to different surfaces (6). Bdellovibrio pili can be discerned at the prey interaction pole in electron micrographs (7). We think that a set of twitching-type pili pulls B. bacteriovorus through a previously generated penetration pore (see below) in the outer membrane, while attached to the inner side of the prey peptidoglycan. Squeezing of bacteria through a pore narrower than a bacterial cell has recently been shown to occur with TFPs in an in vitro assay (8).

A previously identified prey interaction (hit) locus (9) lies in a section of a pil and adherence gene cluster (6) (Fig. 1). Mutations in this locus were shown to be characteristic of prey independence; however, no function could be assigned to any ORF at this locus. Our analysis shows that three of the four formerly predicted hit locus ORFs (9) are inaccurate, with only one of the ORFs corresponding to a predicted gene from this analysis (Bd0108, Fig. 1). No putative function can be assigned to the previously reported hypothetical protein encoded by hit; however, hit is clearly part of a transcriptional unit together with a gene coding for a cell wall–associated protein with a cellulose-binding domain (wapA, Bd0109), the flagellar pilus assembly genes tadA (Bd0111) and tadB (Bd0110), and additional pil genes that encode structural elements of a type IV pilus. Tad and pil genes have been shown to be essential for tight adherence and colonization (10).

Fig. 1.

The host interaction locus (hit) of B. bacteriovorus. The previously described hit locus (host interaction locus, orange bar) (9), consisting of a 950-bp sequence, is congruent with a section of the B. bacteriovorus genome that is predicted to code for pilus and adherence genes. The genes in this cluster are likely to functionally interact as structural elements of a pilus and are predicted to be transcribed as one transcriptional unit. The previously predicted gene hit corresponds to the ORFBd0108 in the B. bacteriovorus HD100 genome and is part of the pilus and adherence transcriptional unit, together with a gene wapA coding for a cell wall–associated protein with a cellulose-binding domain (Bd0109), the flagellar pilus assembly genes tadA (Bd0111) and tadB (Bd0110), and additional pil genes of the type IV pilus. The hit locus containing the tad and pil gene cluster seems to have been inserted in between two chemotaxis genes (cheY, Bd0102; and mcp, Bd0121), because they are flanked on both sides by large potentially noncoding regions (red arrows) for which only short artificial or hypothetical ORFs could be found. The predicted ORFs of this region were assigned the following putative functions: chp, (conserved) hypothetical protein; WapA, cell wall–associated protein with a cellulose-binding domain; tad, tight adherence Flp pilus-assembly proteins TadB and TadA; pil, pilus assembly; cpaB, Flp pilus-assembly protein CpaB; MCP, methyl-accepting chemotaxis protein; hprT, hypoxanthine-guanine phosphoribosyltransferase; *, tetratrico peptide repeat domain–containing protein; comL, competence lipoprotein; argD, acetylornithine/succinyl-diaminopimelate aminotransferase; dapE, succinyl-diaminopimelate desuccinylase.

Before invasion of the prey cell can be achieved, B. bacteriovorus generates a small opening in the prey cell's outer membrane and peptidoglycan layer (stage III), which is ultimately resealed. A mixture of hydrolytic enzymes is applied in a locally targeted manner that prevents excessive damage to the prey and counters diffusion. Candidate genes for this function include serine (74), cysteine (8), and aspartate (1), as well as metal-dependent (53), proteases (table S3). There is also a burst of glycanase activity [previously reported (11)], which solubilizes prey peptidoglycan early in invasion and is sustained throughout the predatory cycle.

Upon entry into the prey periplasm, B. bacteriovorus sheds its flagellum (stage III). Soon afterward it starts cellular events such as DNA replication and the synthesis of cellular biopolymers (stage IV) (12).

Once B. bacteriovorus has entered the periplasm, it alters the cell shape of its prey, forming a joint structure termed the bdelloplast (stage V). The previously rod-shaped prey becomes rounded by a process believed to involve the hydrolysis of peptide crosslinks and glycanase activity (13). Nevertheless, it has been shown that the bdelloplast is osmotically stable (11), which suggests that the structural integrity of the prey's peptidoglycan is maintained. Several genes for glycan-modifying enzymes, such as soluble and membrane-bound lytic murein transglycosylases (8 and 1) and a membrane-bound effector of murein hydrolases, are found in the genome (table S3).

At this stage, B. bacteriovorus begins to extract and ingest solutes from the prey's cytosol via a broad range of transport systems from both the adenosine triphosphate (ATP)–binding cassette (ABC, 147) and major facilitator superfamily (MFS, 97) classes of transporters (figs. S4 and S5 and table S4). Only 7 out of 127 bacterial genomes have been reported so far to code for even higher numbers of ABC-type transporter genes. B. bacteriovorus has to transport charged organic substances not only across the prey's cytoplasmic membrane but also across its own outer and inner membranes. The genomic data illuminate the potential diversity of substances that could be taken up, including transporters for multidrug resistance, organic solvent resistance, amino acids and peptides, and phosphate and nitrate transporters (figs. S4 and S5 and table S4). It remains to be seen whether the prey's membrane transporter systems are used in reverse to translocate cytoplasmic solutes to Bdellovibrio [as suggested in (14)] or whether the predator actively inserts its own gene products for this purpose into the prey's cytoplasmic membrane (14). In this regard, at least 15 of the hydrolytic enzymes are predicted to be extracellularly localized, suggesting a B. bacteriovorus–driven process (table S3).

It is of particular interest to determine which prey components are used as nutrients and which are used as direct source material for the de novo synthesis of biopolymers by Bdellovibrio. Apparently, B. bacteriovorus HD100 is capable of synthesizing only 11 of the amino acids needed for protein synthesis from intermediates of energy metabolism. Moreover, 10 amino acid degradation pathways are missing (table S5). However, all the enzymes are present that are required for the formation of the full range of activated tRNAs (fig. S1 and Table 1). These findings suggest that B. bacteriovorus is capable of protein biosynthesis only while it has access to amino acids from its prey. The two- to threefold increase in cell size and the flagellum elongation observed after B. bacteriovorus's release from prey (2) presumably depend on prey-derived amino acid pools that were stored during the bdelloplast stage.

Previous biochemical experiments (15) indicate that ATP is directly obtained from the prey's cytosol; however, no homologs of known ATP transport proteins were seen in the Bdellovibrio genome like those described for the endoparasitic bacteria Chlamydia and Rickettsia (16). The mechanism for ATP transport remains unknown. B. bacteriovorus does, however, have a full complement of metabolic enzymes needed to generate ATP via glycolysis, the tricarboxylic acid cycle, and fatty acid degradation, as well as the ability to carry out oxidative respiration. This observation and the large number of predicted hydrolytic enzymes indicate that Bdellovibrio degrade prey molecules and resynthesize their own from the constituent bases, sugars, or acids rather than directly transporting and using complex metabolic intermediates of the prey.

The filamentous B. bacteriovorus cell that has grown to several multiples of its normal size begins to septate (stage VI). Although the gene products for chromosome partitioning and septation are homologous to those encoded by known genes (mreB, mbl, ftsZ, and smc), the mechanism itself is likely to be different, because one long filamentous cell divides into many progeny cells of equal size, and odd numbers of progeny are common (7). Because the amount of replicated chromosomal DNA of the Bdellovibrio cells adds up to a multiple of the available nucleotides generated from the prey's hydrolyzed chromosome, it is evident that Bdellovibrio has to synthesize large amounts of nucleotides from scratch. The organism has been found to contain a complete set of genes for purine and pyrimidine metabolism. There also is a complete LPS biosynthesis pathway, which gives rise to a unique lipid A molecule that has been described to be less antigenic than those of enteric bacteria (17).

Within the exhausted prey protoplast, the B. bacteriovorus progeny develop into flagellated cells that are ready for further attack (stage VII). At this stage, B. bacteriovorus produces hydrolytic enzymes (18) to dissolve the remaining peptidoglycan layer and the outer membrane of the prey cell (stage VIII) to release the progeny.

Bdellovibrio's hydrolytic arsenal is used in at least three distinct stages of its life cycle: prey entry (III), degradation of biopolymers (V), and the exit from the bdelloplast (VIII) (fig. S3). The largest group of paralogous hydrolytic enzymes, by far, is found in the family of proteases and peptidases (150), glycanases (10), deoxyribonucleases (DNases) (20), ribonucleases (RNases) (9), and lipases (15) (Table 1 and table S3). The significance of the proteases and peptidases becomes apparent when considering their density in the genome, which is the highest reported so far, with the exception of the minimalist genome of Buchnera aphidicola. The enzymatic activities of these gene products must be contained and organized in a timely fashion. The transcriptional control of these genes is likely to be carried out by numerous sigma factors and interacting proteins (21), as well as numerous transcription factors regulated by sensor histidine kinases (40) (table S1).

The ability of B. bacteriovorus to conduct a highly orchestrated and focused hydrolytic assault on bacterial prey depends on its type I and type II secretion systems, as well as on the twin-arginine translocation system (19). The absence of type III and type IV secretion systems, together with the fact that there are no reports of Bdellovibrio invading mammalian cells (20), is encouraging for their possible development as living antibiotics in future pharmacological applications. In addition, Bdellovibrio's large complement of proteases and other hydrolases, revealed by the genome sequence, may provide a valuable reservoir of enzyme-based antimicrobial substances. Understanding their modes of operation and targets, as well as the mechanisms of predator resistance to these compounds, will aid the hunt for new antibacterial strategies.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S5

Tables S1 to S5


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article