Open Access

Research Article

• Download Article XML
• Download Article PDF
• Download Citation
• E-mail this Article
• Order Reprints
• Print this Article

Introduction

The epsilon subdivision of the Gram-negative Proteobacteria comprises multiple genera contained within three major families: Campylobacteraceae, Helicobacteraceae and Nautiliaceae. The majority of well-characterized species in this subdivision are members of genera within the first two families, including Campylobacter, Arcobacter and Sulfurospirillum in the Campylobacteraceae, and Helicobacter and Wolinella in the Helicobacteraceae. Many of these species are pathogenic, e.g. Campylobacter jejuni [1] and Helicobacter pylori [2], and/or are associated with a particular host or hosts, e.g. Campylobacter upsaliensis [3] and Helicobacter mustelae [4]; however, several species are free-living, e.g. Sulfurospirillum spp. [5], and they are not considered to be pathogenic.

The genus Arcobacter is an unusual taxon within the epsilon subdivision in that it contains both pathogenic and free-living species found in a wide range of environments. Currently, Arcobacter contains four recognized species: A. butzleri [6], A. cryaerophilus [7], A. skirrowii [8] and A. nitrofigilis [9]. A. butzleri, A. cryaerophilus and A. skirrowii have been isolated from animals and humans [10], while A. nitrofigilis is a nitrogen-fixing bacterium isolated originally from Spartina aterniflora roots in an estuarine marsh [9]. In addition to these established Arcobacter species, three new species have been described recently: 1) the obligate halophile A. halophilus sp. nov., isolated from a Hawaiian hypersaline lagoon [11], 2) A. cibarius sp. nov., isolated from broiler carcasses [12] and 3) Candidatus A. sulfidicus, a sulfide-oxidizing marine organism that produces filamentous sulfur [13]. In addition, several potential, novel Arcobacter species, based so far on only 16S rDNA sequence data, have been identified in: the flora of deep-sea hydrothermal vents [14], hydrocarbon-contaminated seawater [15], a low-salinity petroleum reservoir [16], infected or dead coral surfaces [17], deep-sea sediments [18], tube worms [19], anaerobic sludge [20], and a circulated dairy wastewater lagoon [21]. These studies demonstrate clearly that the genus is associated strongly with fresh-water and marine environments. In fact, although A. butzleri, A. cryaerophilus and A. skirrowii have been isolated often from animals or food sources, they have been isolated frequently also from water or water systems [22]–[30].

Arcobacter butzleri is the best characterized of all Arcobacters. A. butzleri cells are small, spiral and motile [10], similar morphologically to the taxonomically–related Campylobacter. Nonetheless, notable differences exist between A. butzleri and Campylobacter spp. Classified initially as an “aerotolerant Campylobacter”, along with A. cryaerophilus [10], A. butzleri is able to grow aerobically, at variance with most Campylobacters which are microaerophilic. However, A. butzleri grows also under microaerobic and anaerobic conditions [10]; thus, this bacterium can grow at all oxygen concentrations. Additionally, Campylobacter spp. grow generally between 37°C and 42°C [31], whereas A. butzleri is more psychrophilic with a temperature range between 15°C and 37°C, although some strains can grow at 42°C [10]. Furthermore, A. butzleri is more halotolerant than most Campylobacter spp., with some strains able to grow at 3.5% NaCl [10].

While A. butzleri is isolated often from aqueous environments, it is isolated also from multiple animals and food sources. It has been found in pigs [32] and ground pork [33], [34], chicken carcasses [35], [36] and other poultry [37], as well as in beef [38], [39], lamb [34] and the feces of other animals [40]. A. butzleri has also been isolated increasingly from human diarrheal stool samples [41]–[47]. The clinical symptomatology described for A. butzleri typically includes diarrhea and recurrent abdominal cramps [10], although A. butzleri-related bacteremia has also been reported [48], [49]. Prouzet-Mauléon et al. [46] reported an isolation frequency of 1% from human clinical stool samples. Additionally, Vandenberg et al. [47] reported an isolation frequency of 3.5% from diarrheic stool samples. Although co-infection with other enteric pathogens was reported by Prouzet-Mauléon and Vandenberg, in the majority of samples (14/15 clinical stool samples and 55/67 patients, respectively) no other enteric pathogen was detected. Houf et al. reported that Arcobacters were isolated from 1.4% (7/500) of asymptomatic human stool samples [50]. However, all seven isolates were typed as A. cryaerophilus; A. butzleri was not isolated. Similarly, Vandenberg et al. [47] reported also that A. butzleri was isolated more frequently from diarrheic stool samples than from non-diarrheic stool samples. Thus, the isolation of A. butzleri from diarrheic stool samples is likely to be relevant clinically and is probably not due to the organism being merely a human commensal. Therefore, these data suggest strongly that A. butzleri is an emerging pathogen [10], where transmission, as with C. jejuni, occurs probably through consumption of contaminated food or water. The low level of incidence reported in human clinical samples is most likely an underestimate, due to sub-optimal isolation and/or detection methods [46].

Relatively little is known about A. butzleri, compared to other members of the epsilon subdivision, but the wealth of genomic information from other epsilonproteobacterial taxa provides a solid foundation to compare and contrast A. butzleri to its taxonomic relatives. The genomes of multiple species of Epsilonproteobacteria have been sequenced; these include: four strains of C. jejuni subsp. jejuni ([51]–[54]: strains NCTC 11168, RM1221, 81-176 and 81116, respectively); C. jejuni subsp. doylei strain 269.97 (CP000768.1); Campylobacter coli strain RM2228, Campylobacter lari strain RM2100 and C. upsaliensis strain RM3195 [52]; Campylobacter fetus subsp. fetus strain 82-40 (CP000487.1); Campylobacter curvus strain 525.92 (CP000767.1); Campylobacter concisus strain 13826 (CP000792.1); Campylobacter hominis strain ATCC BAA-381 (CP000776.1); Sulfuromonas denitrificans strain ATCC 33889 (formerly Thiomicrospira denitrificans [55]: CP000153); Wolinella succinogenes strain DSM 1740 [56]; Helicobacter hepaticus strain ATCC 51449 [57]; three strains of H. pylori ([58]^-[60]: 26695, J99 and HPAG1, respectively); Helicobacter acinonychis strain Sheeba [61]; and the deep-sea vent taxa Nitratiruptor sp. and Sulfurovum sp. (strains SB155-2 and NBC37-1, respectively [62]). This study presents the genomic sequence of a human clinical isolate, A. butzleri strain RM4018, a derivative of the type strain ATCC 49616. The genomic data revealed multiple differences between A. butzleri and other members of the Campylobacteraceae, as well as pathways and systems vital for its survival in diverse environments.

Results and Discussion

General features

The genome of Arcobacter butzleri strain RM4018 contains 2,341,251 bp; as such it is the second largest characterized epsilonproteobacterial genome to date, smaller than the genome of Sulfurovum strain NBC37-1 (2,562,277 bp) but larger than both the genomes of S. denitrificans strain ATCC 33889 (2,201,561 bp) and W. succinogenes strain DSM 1740 (2,110,355 bp). The G+C content of the RM4018 genome (27%) is remarkably low. A summary of the features of the strain RM4018 genome is provided in Table 1. A diagrammatic representation of the RM4018 genome is presented in Figure S1.

Consistent with its size, the RM4018 genome is predicted to encode 2259 coding sequences (CDSs). Based on pairwise BLASTP comparisons of proteins predicted to be encoded by these CDSs against proteins in the NCBI non-redundant (nr) database (release 10/13/2007), and on the presence of various Pfam and PROSITE motifs, 1011 (45%) of the predicted proteins were assigned a specific function, 505 (22%) were attributed only a general function, and 743 (33%) were considered proteins of unknown function (Table 1). A complete list of the CDSs predicted to be present within the genome of strain RM4018 and their annotation is presented in the supplementary Table S1. A breakdown of the CDSs by function is presented in supplementary Table S2.

Relationship of A. butzleri to other taxa

Arcobacter butzleri is a member of the family Campylobacteraceae which includes also the genera Campylobacter and Sulfurospirillum. Given the close taxonomic relationship between Arcobacter and Campylobacter, it is noteworthy that 17.2% and 12.4% of the RM4018 proteins have their best match in proteins encoded by S. denitrificans and W. succinogenes, respectively, both members of the Helicobacteraceae (Table 2). Moreover, approximately 25% of the RM4018 proteins have their best match in proteins encoded by Sulfurovum or Nitratiruptor, deep-sea vent Epsilonproteobacteria isolated from a sulfide mound off the coast of Japan [62].The percentage of S. denitrificans homologs with best matches is greater than that of the eight sequenced Campylobacteraceae species combined (13.1%). These differences are reduced somewhat by comparing the top five matches and not just the best match; however, even using these parameters, S. denitrificans (9.6%) remains the closest related organism (Table 2). Among the Campylobacters, C. fetus (3.1% best matches) is the closest to A. butzleri, although it is possible that other Campylobacters, whose genomes are as yet un-sequenced, may have higher degrees of similarity to A. butzleri than C. fetus.

In the entire A. butzleri strain RM4018 proteome (unique and non-unique), 61.5% (1390/2259) of the proteins have their best matches in proteins encoded by the epsilonproteobacterial taxa, and 79.0% (1785/2259) are similar to Proteobacterial proteins. Most of the matches to non-Epsilonproteobacteria are found within the gamma subdivision (Table 2), in genera such as Marinobacter (Alteromonadales), Oceanospirillum (Oceanospirillales) and Pseudomonas (Pseudomonadales). Other phyla with a moderate number of protein matches include Firmicutes (2.5%), especially Clostridiales (1.5%), and Bacteroidetes/Chlorobi (2.8%). Among the 2259 predicted CDSs of strain RM4018, 315 (13.9%) had no homolog within the nr database using a minimum expect (E) value of 1×10⁻⁵, a minimum identity of 25% and a minimum alignment length of 75% (Figure 1, Table 2, Table S3).

Although approximately 30% of the A. butzleri RM4018 proteins with homologs within the nr database have their best matches to non-epsilonproteobacterial proteins, the approximately 550 genes encoding these proteins are not distributed randomly through the RM4018 genome, but rather clustered with respect to position and protein function (Table S2, Table S3). Interestingly, many of these clusters contain proteins involved in transport, and several contain discrete loci associated with a single function, such as the urease and quinohemoprotein loci described below. The genes contained in these loci are found often in the same order as in other taxa, suggesting lateral transfer. However, the G+C content of these gene clusters is not significantly different from the genome as a whole, suggesting that if these clusters were acquired from other non-Arcobacter taxa the acquisition was not recent.

Generally, strain RM4018 proteins involved in “housekeeping” functions (e.g. amino acid biosynthesis, fatty acid biosynthesis and protein synthesis) have homologs well-conserved among the other Epsilonproteobacteria (Table 3). Nevertheless, a number of other major functional categories are more divergent. For example, 19% (13/70) of the strain RM4018 proteins involved in DNA replication and repair have either no homologs or homologs of low similarity (≤ 35% identity) among the other epsilonproteobacterial taxa; this proportion increases to 37% (26/70) if proteins with homologs only within the Campylobacteraceae are included (Table 3). Other divergent functional categories include: sulfur/nitrogen metabolism (34%), transcriptional regulators/σ-factors (47%), signal transduction (49%), cell envelope (37%), chemotaxis (70%) and antibiotic resistance (36%) (Table 3). These divergent functional categories will be discussed in further detail.

Methyl-directed mismatch repair and the absence of polynucleotide G:C tracts

Members of the Epsilonproteobacteria lack multiple genes in the methyl-directed mismatch repair system (MMR). The MMR system depends upon the presence of three main functions: 1) the MutSLH endonuclease complex, 2) a methylation system to identify parental vs. daughter strands and 3) multiple 5′→3′ and 3′→5′ single-strand DNA exonucleases, e.g. RecJ, ExoI and ExoVII [63]. The A. butzleri RM4018 genome is not predicted to encode MutL or MutH. However, it is predicted to encode a MutS2 family MutS protein, which is distinguished from MutS1 by the presence of a C-terminal Smr domain. It has been proposed that the Smr domain has a MutH-like nicking endonuclease function [64]. Thus, the A. butzleri MutS2 protein may contain both MutS and MutH domains, obviating the need for a MutL scaffold protein. A. butzleri strain RM4018 is not predicted to encode a Dam DNA-adenine methyltransferase (data not shown) used commonly in MMR systems, but the presence of a Dcm DNA-cytosine methyltransferase was demonstrated experimentally in strain RM4018 (data not shown); this Dcm function may be used by the A. butzleri MMR system to distinguish between parental and daughter strands. Although the annotated epsilonproteobacterial genomes contain the 5′→3′ exonuclease RecJ, they do not contain the 3′→5′ exonucleases ExoI or ExoVII. A 32-fold increase in +1 frameshifts and an 11-fold increase in -1 frameshifts were observed in an E. coli ExoI⁻ ExoVII⁻ mutant [65]; therefore, one possible outcome of such a defect in MMR systems would be the formation and extension of hypervariable G:C tracts, such as those identified in all characterized Campylobacter and Helicobacter genomes. A major distinguishing characteristic of the A. butzleri RM4018 genome is the lack of such G:C tracts, a feature shared by the S. denitrificans and W. succinogenes genomes. A comparison of these three genomes revealed the presence of two to four members of the DnaQ superfamily, to which the 3′→5′ exonuclease ExoX belongs [66]. Thus, one or more of these DnaQ homologs may provide the missing 3′→5′ exonuclease function, and it is conceivable that the absence of hypervariable G:C tracts in A. butzleri strain RM4018 and the presence of these tracts in Campylobacter and Helicobacter may be due to the presence or absence of a functional MMR system, respectively.

Sulfur assimilation, oxidation and the biosynthesis of sulfur-containing amino acids

Arcobacter butzleri strain RM4018 contains a number of genes required for sulfur uptake and assimilation (Figure 2). These genes include those encoding the sulfate ABC transporter CysATW, the sulfate binding protein Sbp, the ATP sulfhydrylase CysDN, the adenosine phosphosulfate (APS) reductase CysH, the sulfite reductase proteins CysI and CysJ and the siroheme synthase CysG. CysD and CysN have been identified also in C. coli, but the position of these genes, along with the APS kinase-encoding gene cysC and the 3′(2′),5′-bisphosphate nucleotidase-encoding gene cysQ, is in the capsular locus, suggesting that these genes in C. coli are involved in the formation or modification of the capsule and not sulfur assimilation per se.

Intracellular sulfate is reduced to sulfite in bacteria by one of two pathways. Both pathways first convert sulfate to APS via the ATP sulfhydrylase CysDN [67]. The first pathway converts APS to PAPS (phosphoadenosine phosphosulfate) using the kinase CysC, and then reduces PAPS to sulfite using the PAPS reductase CysH [67]; PAPS toxicity in some taxa is decreased through the conversion of PAPS to APS via CysQ [68], [69]. The second pathway, identified originally in plants and subsequently in taxa such as P. aeruginosa [70] and M. tuberculosis [71], reduces APS to sulfite directly using the APS reductase CysH. The presence of conserved two-cysteine motifs in strain RM4018 CysH, characteristic of bacterial APS reductases [70], [71], and the absence of both CysC and CysQ homologs, suggests that sulfate reduction to sulfite in this strain does not use a PAPS intermediate.

The sulfite generated by strain RM4018 CysH has two potential fates: reduction to sulfide and assimilation into the sulfur-containing amino acids L-cysteine and L-methionine (Figure 2), or oxidation to sulfate, facilitated by a cytochrome c multienzyme complex encoded by the sox genes (ab0563-ab0570). Initial data indicated that sulfite is oxidized by strain RM4018 (data not shown), and also that it can grow in minimal media without added cysteine or methionine (data not shown). Hence, both metabolic fates are possible, and the genetic switch modulating oxidation or reduction of sulfite remains to be identified. Homologs of the sox genes have been identified in multiple taxa, including Paracoccus pantotrophus [72], [73], Chlorobium tepidum [74] and Rhodovulum sulfidophilum [75]. The Sox clusters of some organisms can contain as many as 15 genes, but seven genes (soxXYZABCD) are essential for sulfur oxidation [76]. These seven genes encode four proteins: the heterodimeric c-type cytochrome SoxXA, the heterodimeric sulfur-binding protein SoxYZ, the heterotetrameric SoxCD sulfur dehydrogenase and the thiol sulfate esterase SoxB [76]. Sulfite, sulfide, sulfur and thiosulfate are possible substrates for the Sox complex; thiosulfate and sulfide oxidation requires the entire complex, while sulfite oxidation does not require SoxCD [76]. The presence of SoxCD in strain RM4018 would suggest that thiosulfate is oxidized by this organism. However, other sox genes important for thiosulfate oxidation, e.g. soxV [77], [78], are absent in this strain, as well as the Sbp-related thiosulfate-binding protein CysP. Thus, it is possible that only sulfite and sulfide are oxidized in strain RM4018. Sox proteins are present also in the related epsilonproteobacterial taxa Sulfurovum, Nitratiruptor and S. denitrificans (Table S2). However, there are multiple differences between the sulfur oxidation systems of these related organisms and those of strain RM4018: 1) the sox genes in strain RM4018 form a single cluster instead of two clusters, 2) the strain RM4018 sox cluster contains only one copy of soxY and soxZ and 3) no significant similarity exists between the SoxXA protein of strain RM4018 and the SoxXA proteins of Sulfurovum, Nitratiruptor and S. denitrificans. The sox cluster of strain RM4018 was not detected in 12 additional A. butzleri strains (see below). However, a cluster of sox genes similar to that of strain RM4018 has been identified in A. halophilus (data not shown). Thus, it appears that strain RM4018 may have acquired the sox cluster through lateral gene transfer from another Arcobacter species and that the sulfur oxidation machinery of Arcobacter and, e.g., Sulfurovum are likely to be distinct evolutionarily.

The genome of strain RM4018 contains all of the genes necessary for the biosynthesis of L-cysteine, L-methionine and iron-sulfur clusters (Figure 2). Consistent with the proposed absence of thiosulfate metabolism, the bacterium has only a gene encoding the CysK cysteine synthase A and not the CysM cysteine synthase B that utilizes thiosulfate instead of sulfide to produce S-sulfocysteine [67]. Strain RM4018 also contains both the cobalamin-dependent and cobalamin-independent homocysteine transmethylases MetH and MetE.

The central metabolism of A. butzleri

The general function of the citric acid/tricarboxylic acid (TCA) cycle is to oxidize organic tri- and di-carboxylic acids to provide energy and biosynthetic precursors for metabolism. Arcobacter butzleri strain RM4018 encodes several proteins homologous to other epsilonproteobacterial TCA cycle enzymes (Figure 3), e.g. isocitrate dehydrogenase (AB1321), 2-oxoglutarate dehydrogenase (AB0852-AB0855), malate dehydrogenase (AB1322) and citrate synthase (AB0307). However, A. butzleri strain RM4018 is predicted putatively to encode two aconitate hydratases and two fumarate dehydratases. Additionally, enzymes that catalyze two TCA cycle steps are absent apparently in strain RM4018.

The genes ab0275 and ab1447 encode proteins homologous to the E. coli aconitate hydratases AcnA and AcnB, respectively. Arcobacter butzleri AcnB shares between 62% and 76% identity with other epsilonproteobacterial aconitases. Arcobacter butzleri AcnD, encoded by ab0275, does not have significant similarity with any of the known epsilonproteobacterial proteins, but has an 82% similarity with the PrpD 2-methylisocitrate dehydratase of Alkaliliminicola ehrlichei as well as 68% similarity with E. coli AcnA. In addition, the proximity of ab0275 to genes that encode the methylcitrate pathway enzymes PrpB and PrpC suggests that in strain RM4018 AcnD is the aconitase of the methylcitrate pathway and AcnB is the aconitase of the TCA cycle. Also, the absence of the 2-methylcitrate dehydratase PrpD and the presence of acnD next to ab0276, which encodes the AcnD-associated protein PrpF, suggests that A. butzleri strain RM4018 contains the alternate AcnD/PrpF methylcitrate pathway [79].

Escherichia coli can express three fumarases: A, B and C. The first two contain iron-sulfur clusters and are unstable aerobically; fumarase C is stable in the presence of oxygen. The genome of strain RM4018 is predicted to encode two fumarases: AB0722 and AB1921. AB0722 has 74% similarity to E. coli fumarase C and would be predicted to be active under aerobic conditions. AB1921 has 48% similarity to E. coli fumarases A and B and would require microaerobic or anaerobic conditions. AB1921 also has 55–64% similarities with two proteins encoded by the obligate microaerophile W. succinogenes and two encoded by H. hepaticus. Interestingly, the sequence similarities with the W. succinogenes proteins WS1766 and WS1767, and the H. hepaticus proteins HH1702 and HH1793 correspond to the first 282 and the last 185 residues of AB1921, respectively, suggesting that the two polypeptides in W. succinogenes or H. hepaticus are fused into one protein in A. butzleri.

No genes encoding the SucCD succinyl-CoA synthetase or the SdhABCD succinate dehydrogenase have been identified in the A. butzleri genome, but the gene cluster ab0296-ab0298 encodes proteins with high similarities (71–85%) to the fumarate reductase FrdABC of C. jejuni, H. hepaticus, H. pylori, and W. succinogenes, which catalyzes the reaction in the reductive direction converting fumarate to succinate. Some bacteria such as E. coli and C. jejuni have genes encoding both Sdh and Frd, and others, e.g. H. pylori and W. succinogenes, encode only Frd. The similar structures of Sdh and Frd preclude predicting, solely from sequence analyses, whether an enzyme is one or the other. Regulation of transcriptional levels by the oxygen content in the atmosphere permits differentiation between both enzymes; obligate aerobes encode Sdh and anaerobes encode Frd. Initial experiments with A. butzleri indicated the presence of fumarate reduction and no succinate oxidation, and this activity increased several-fold in bacteria grown under anaerobic conditions relative to bacteria grown under aerobic conditions (data not shown). These results provided evidence supporting the identification of A. butzleri FrdABC (ab0296-ab0298) as a fumarate reductase (Figure 3).

The genome of strain RM4018 may contain also a pathway, encoded by the gene cluster ab1917-ab1921, which resembles segments of the 3-hydroxypropionate cycle [80], [81] and the citramalate cycle [82], [83], and interconverts glyoxylate and propionyl-CoA with pyruvate and acetyl-CoA (Figure 3). AB1917 and AB1920 are similar to the CitE citrate lyase. In addition, AB1918 is an acetyl-CoA synthetase with a CaiC domain in residues 50-550, and a CitE domain in residues 550-825. CitE domains have strong similarity to malyl-CoA lyases and have lesser homology to malate synthetase domains. Thus, any of these three proteins could function as a malyl-CoA lyase. AB1919 has a ~73% similarity to enol-CoA hydratase which catalyzes reversible reactions interconverting 2-enoyl-CoA compounds, such as mesaconyl-CoA, and 3-hydroxyacyl-CoA compounds, such as β-methylmalate-CoA. If the acetyl-CoA synthetase encoded by ab1918 and the anaerobic fumarate hydratase encoded by ab1921 were broad specificity enzymes, the former could catalyze the synthesis of mesaconyl-CoA from mesaconate, and the latter the interconversion of mesaconate and citramalate. Finally, citramalate could be synthesized from pyruvate and acetyl-CoA by one of the CitE enzymes. The gene ab1069 encodes an acyl-CoA thioester hydrolase which can catalyze the synthesis of malyl-CoA from malate or the reverse reaction. One of the three citrate lyases mentioned could then convert malyl-CoA to glyoxylate and acetyl-CoA and glyoxylate (Figure 3).

Arcobacter butzleri does not grow on acetate, citrate, propionate or acetate with propionate, and it grows on fumarate, lactate, malate and pyruvate (data not shown). These data suggested that the methylcitrate pathway would function to produce oxaloacetate for the TCA cycle and propionyl-CoA. The latter metabolite together with glyoxylate synthesized from malate via malyl-CoA would be converted to acetyl-CoA and pyruvate. Acetyl-CoA could be synthesized also from pyruvate via the pyruvate dehydrogenase complex encoded by ab1480-ab1482. Alternatively, the methylcitrate pathway could have a regulatory role converting any excess propionyl-CoA to succinate and pyruvate. Through the activity of a malic enzyme encoded by ab1083, pyruvate can be carboxylated to malate and used to replenish the TCA cycle.

Anaerobic and aerobic respiration

A. butzleri has a full complement of genes for aerobic/microaerobic respiration including those encoding NADH:quinone oxidoreductase, ubiquinol cytochrome c oxidase, ferredoxin, cytochrome bd oxidase, cytochrome c oxidase (cbb3-type), and F1/F0 ATPase (Figure 4), but it also has limited ability for anaerobic respiration. Potential electron donors, in addition to NADH, are hydrogen, malate and formate. One large gene cluster encodes three FeNi hydrogenases with one uptake hydrogenase (hupSL) and two membrane associated proteins encoded by hydABCDF and hyaABCD, the latter two predicted to be anchored to the membrane by the b-type cytochromes HydC and HyaC, respectively. There are two formate dehydrogenases, one selenocysteine homolog and one cysteine homolog, suggesting that selenium may be important in their regulation. Also present is a malate:quinone oxidoreductase (Mqo) and a putative lactate dehydrogenase (AB0728), suggesting that lactate may be a potential electron donor. Electrons may be transferred also to the menaquinone pool through the Sox system described above, the number of electrons depending on the substrate oxidized, 2 e⁻ for sulfite and 8 e⁻ for sulfide.

Fumarate, nitrate and nitrite are used by strain RM4018 as electron acceptors. Fumarate can be reduced to succinate by the fumarate reductase FrdABC. Nitrate can be reduced to ammonia via a periplasmic nitrate reductase NapAB and a pentaheme nitrite reductase NrfAH. The nap operon has the same gene number and order as that seen in Campylobacter species. It is possible also that trimethylamine–oxide (TMAO) and/or dimethylsulfoxide (DMSO) may serve as alternative electron acceptors. Analysis of the strain RM4018 genome indicates the presence of the pentaheme c-type cytochrome TorC encoded by ab1150. The CDS ab1151 has been annotated as bisC, but the BisC family includes also other anaerobic dehydrogenases, such as the TMAO and DMSO reductases TorA and DmsA, respectively. The genes torC and torA are co-transcribed usually with torD, but no TorD homolog was detected in the strain RM4018 genome. Other electron acceptors may be present: CDS ab1360 encodes a putative cytochrome b-type nitric oxide reductase, and ab1987 encodes a putative nitrite/nitric oxide reductase. Additional investigations will be necessary to determine the function of these putative electron acceptors.

In addition to the respiratory proteins described, A. butzleri strain RM4018 is predicted to encode a quinohemoprotein amine dehydrogenase (QHAmDH), the presence of which is novel in the epsilon subdivision. QHAmDH is a heterotrimeric (αβγ) protein that deaminates oxidatively a variety of aliphatic and aromatic amines, e.g. n-butylamine and benzylamine [84], and is unusual in that the small (γ) subunit contains an intrinsic quinone cofactor [in A. butzleri, cysteine tryptophylquinone (CTQ)], formed by the covalent linkage of a cysteine residue to an oxidized tryptophan residue [85]. The presence of this intrinsic quinone cofactor permits the transfer of electrons directly to the cbb3 oxidase through either a cytochrome c [86] or blue-copper protein (e.g. azurin [87]) intermediate (Figure 4). The QHAmDH locus of strain RM4018 encodes also a radical SAM (S-Ado-Met) family protein (AB0466). A radical SAM protein (ORF2) is found also in the QHAmDH locus of Paracoccus denitrificans and has been proposed to play an important role in the formation of the CTQ cofactor [88]. Amino acid motifs conserved in ORF2 are 100% identical to those in AB0466, as well as to the relevant amino acids in the γ subunit AB0467, suggesting that the strain RM4018 QHAmDH locus encodes a functional dehydrogenase. M9 minimal media amended with 0.5% n-butylamine-HCl (v/v), benzylamine-HCl (w/v) or methylamine-HCl (w/v) as a sole carbon source does not support the growth of strain RM4018 (data not shown). However, substrate specificity has been demonstrated among the amine-utilizing taxa [89]; therefore, it is likely that RM4018 utilizes an as yet unidentified aliphatic or aromatic amine. Finally, two aldehyde dehydrogenases, encoded putatively by ab0376 and ab2135, also novel in the Epsilonproteobacteria, were identified in strain RM4018 (Figure 4). Thus, it is possible that the aldehydes generated from the QHAmDH are oxidized further.

Urease

Arcobacter butzleri strain RM4018 contains six genes (ab0808-ab0813) involved in the degradation of urea. In bacteria this catabolism involves generally three sets of genes: a nickel-containing urease (composed of α, β and γ subunits), urease accessory proteins which deliver the nickel to the urease and a nickel uptake system. The urease of strain RM4018 is functional (data not shown), as determined by a phenol red/urea assay [90], although the level of activity is not as high as that found in other urease-producing taxa (e.g. UPTC C. lari). As in Helicobacter, the urease α and β subunits are fused, and the strain RM4018 urease subunits show high homology to Helicobacter urease subunits. However, differences exist with the urease loci of both Helicobacter and UPTC C. lari. First, the gene order of the locus itself, ureD(AB)CEFG, is similar to the gene order of the urease loci in Klebsiella, Proteus, E. coli O157:H7 and Vibrio, but not the Helicobacter locus ure(AB)IEFGH. Second, although the urease enzyme itself is similar to the Helicobacter urease, the accessory proteins UreD, UreE and UreF, had greater similarity to those identified in Bacillus, Lactobacillus and Psychromonas (Table S3); additionally, the nickel-binding protein UreE has a histidine-rich C terminus, found in multiple UreE proteins, but not in those from Helicobacter or C. lari. Also, unlike Helicobacter, no obvious nickel uptake system, such as the H. hepaticus nikABDE or H. pylori nixA, was found in strain RM4018. A putative nickel transporter, AB1752, was identified, but it is unclear whether it is specific for nickel, or is a heavy-metal-ion transporter. Finally, although the A. butzleri urease may serve to degrade exogenous urea, it may also degrade endogenous urea, formed during putrescine biosynthesis, specifically during the conversion of agmatine to putrescine by SpeB (AB1578).

Surface structures

SDS-PAGE analysis suggested that A. butzleri strain RM4018 can express lipooligosaccharide (LOS). To date, little is known about the roles of these molecules in A. butzleri. They generate great attention in bacterial pathogens since LOS/lipopolysaccharide (LPS) are major inducers of proinflammatory responses, are immunodominant antigens, and play a role in host cell interactions. The LOS biosynthesis locus of strain RM4018 (ab1805-ab1833) showed a similar organization to those of Campylobacter [51], [52], [91], [92], and is thus dissimilar from the loci of Helicobacter and Wolinella. At both ends of the locus are genes involved in the addition of heptose to the oligosaccharides and the encoded proteins are similar to those of other Epsilonproteobacteria (Table S2). Within the locus, genes are found whose products have functions related to LOS/LPS biosynthesis, including several glycosyltransferases, but the proteins encoded have greater similarity to proteins outside of the Epsilonproteobacteria (Table S3). This LOS/LPS biosynthesis region is conserved among 12 unrelated A. butzleri strains, based on comparative genomic indexing (described in detail later), which likely distinguishes A. butzleri from C. jejuni, where the LOS biosynthesis region is an intraspecies hypervariability region [93]. The gene conservation of this region in A. butzleri resembles more closely the conservation of the LPS core biosynthesis region occurring among many of the Salmonella enterica serovars [94].

Many bacteria synthesize structurally diverse polysaccharide polymers, O-antigen and capsule that are major antigenic determinants. It is possible that A. butzleri strain RM4018 produces O-antigen, since there is a locus (ab0661-ab0697) that encodes several additional glycosyltransferases. This region has two copies of wbpG, hisH, and hisF, found also in the Pseudomonas aeruginosa B-band O-antigen locus [95], and many of the other encoded proteins have greater similarity to proteins from bacteria outside of the Epsilonproteobacteria (Table S2). Although this region could be involved putatively in capsular formation, the absence of conserved kps capsular genes in this region, combined with the presence of O-antigen-related genes, suggests that A. butzleri strain RM4018 produces O-antigen and not capsule; however, further investigations will be necessary to determine the nature of the A. butzleri cell-surface structures. Like many O-antigen biosynthesis regions, the A. butzleri region appears to represent an intraspecies hypervariability region with the RM4018 region, present in only 1 of the 12 A. butzleri strains examined using comparative genomic indexing.

Arcobacter butzleri strain RM4018 is a motile bacterium that synthesizes a polar flagellum. Many of the flagellar apparatus proteins encoded by strain RM4018 have homologs in other epsilonproteobacterial taxa. However, phylogenetic analysis of selected flagellar proteins suggests that the flagellar apparatus of strain RM4018 has an evolutionary history distinct from those of Campylobacter and Helicobacter (Figure 5A). This distinct history is supported also by predicted differences between strain RM4018 and Campylobacter/Helicobacter in flagellar gene regulation (see below). Additionally, the flagellar genes of strain RM4018 are highly clustered, compared to the flagellar genes of Campylobacter jejuni (Figure 5B). The primary cluster in strain RM4018 contains 20 flagellar genes (ab1931-ab1961) with the other two flagellar clusters containing eight (ab0197-ab0208) and three (ab2238-ab2244) genes. The flagellar genes of the related organism Nitratiruptor are also highly clustered with the primary cluster containing 36 flagellar and chemotaxis genes. Significantly, the Nitratiruptor flagellar proteins also appear to be distinct phylogenetically from those of both A. butzleri and Campylobacter/Helicobacter. Moreover, the primary flagellar cluster of Nitratiruptor has an atypical G+C content, suggesting acquisition through horizontal gene transfer [62]. Although the G+C content of the strain RM4018 flagellar genes is not atypical, it is possible that the flagellar genes of this organism were acquired via a similar mechanism.

Prophage and genomic islands

The genome of the A. butzleri strain RM4018 is predicted to contain a prophage. The size of this prophage is approximately 38 kb and spans genes ab1655-ab1706. BLASTP comparison of the predicted phage proteins to proteins from other bacteriophage indicates that this prophage is a member of the mutator (Mu) bacteriophage family. The size of the prophage is similar to other Mu-like bacteriophage, and it contains proteins similar to the Mu transposition proteins A and B in addition to coat, baseplate, and tail proteins. Mu-like bacteriophage have been found in multiple bacterial taxa, including E. coli, Neisseria meningitidis, Deinococcus radiodurans, Haemophilus influenzae, Burkholderia cenocepacia [96], [97] and notably C. jejuni (CMLP1: strain RM1221[52], [93]). Indeed, 26 of the 50 predicted RM4018 Mu-like phage proteins are similar to those encoded by CMLP1 (Table S2). Hence, it is proposed to name this bacteriophage AMLP1 (Arcobacter Mu-like phage). Mu-like bacteriophage have been identified in other epsilonproteobacterial taxa, including Campylobacter (11 species), Arcobacter (3 species), and Helicobacter bilis ([93]; Miller and Mendoza, unpublished data). Some of these Mu-like phage are similar to CMLP1, but many show marked variation in gene content and gene sequence, indicating that CMLP1 and AMLP1 are members of a diverse Mu bacteriophage family common to Proteobacteria.

The genome of strain RM4018 also contains three small genomic islands, termed ABGI1, ABGI2 and ABGI3 (for Arcobacter butzleri genomic island). ABGI1 is 26,918 bp (bp 1,324,568-1,351,485) and contains 29 genes (ab1330-ab1358). ABGI2 is 15,973 bp (bp 1,703,320-1,719,292) and contains 13 genes (ab1721-ab1733). ABGI3 is 4,907 bp (bp 2,103,976-2,108,882) and contains eight genes (ab2090-ab2097). Additionally, ABGI1-3 are bordered by direct repeats of 21, 21 and 25 bp, respectively. The presence of genomic islands within the epsilon subdivision is not unusual. They have been identified in C. jejuni (CJIE2 and CJIE3; [52]), H. hepaticus (HHGI; [57]) and H. pylori (cagPAI; [58]). CJIE2 and CJIE3 are bounded by direct repeats, contain an integrase gene at one end, and have integrated into the 3′ end of a tRNA [52], [93]. Consistent with the C. jejuni islands, all three RM4018 genomic islands contain terminal integrase genes and ABGI1 has inserted into the 3′ end of a leucinyl-tRNA; however, ABGI2 and ABGI3 did not integrate into tRNAs, although ABGI3 is in close proximity to a cysteinyl-tRNA. CJIE3 has been proposed to be an integrated plasmid, based on the similarity of this element to the C. coli strain RM2228 megaplasmid [52], [93]; in contrast, none of the protein functions encoded by elements ABGI1-3 suggested a plasmid origin. Interestingly, ABGI2 does encode proteins similar to the Type I restriction enzymes HsdS and HsdM. Thus, the only Type I, II, or III restriction/modification enzymes present in strain RM4018 are encoded by a genomic island. The role of ABGI3 is unknown.

Antibiotic resistance

Arcobacter butzleri strain RM4018 was resistant to 42 of the 65 antibiotics tested (see supplementary Table S4). This level of resistance is remarkably high, considering that the multi-drug-resistant C. coli strain RM2228 was resistant to only 33 of the same 65 antibiotics ([52] and Table S4). Strain RM4018 was resistant to all macrolides and sulfonamides tested and to all of the β-lactam antibiotics, with the exception of the β-lactam cephalosporin ceftazidime. Strain RM4018 had resistance to some quinolones, i.e. nalidixic acid and oxolinic acid, and also to chloramphenicol and 5-fluorouracil (5FU). The pattern of antibiotic resistance in strain RM4018 is consistent, in part, with the resistances of 39 A. butzleri strains tested against a smaller set of 23 antibiotics by Atabay and Aydin [98], with the exception that strain RM4018 was resistant to chloramphenicol and nalidixic acid. No plasmids were detected in strain RM4018; therefore all resistance mechanisms would be chromosomal in nature.

In many cases, the antibiotic resistance observed for strain RM4018 was due to the presence or absence of genes characterized previously in terms of antibiotic resistance. For example, chloramphenicol resistance is due most likely to the presence of a cat gene (ab0785), which encodes a chloramphenicol O-acetyltransferase. β-lactam resistance is due probably to the three putative β-lactamases (AB0578, AB1306 and AB1486) present in the RM4018 genome; β-lactam resistance may be enhanced also by the presence of the lrgAB operon (ab0179, ab0180) which modulates penicillin tolerance in Staphylococcus [99], [100]. Uracil phosphoribosyltransferase, encoded by the upp gene, catalyzes the first step in the pathway that leads to the production of the toxic analog 5-fluorodeoxyuridine monophosphate; mutations in the upp gene have been shown to lead to increased 5FU resistance [101], [102]. Thus, the absence of upp in A. butzleri strain RM4018 results presumably in high 5FU resistance.

Although strain RM4018 is resistant to some quinolones, mutations implicated previously in Campylobacter [103] and Arcobacter [104] quinolone resistance at Thr-86, Asp-90 and Ala-70 in the DNA gyrase subunit GyrA are not present in strain RM4018. It is probably not a coincidence that strain RM4018 is susceptible to hydrophilic quinolones (e.g. ciprofloxacin and norfloxacin) and resistant to hydrophobic ones (e.g. nalidixic acid and oxolinic acid). These data suggest that the mechanism of quinolone resistance in strain RM4018 is not at the level of the gyrase, but rather of uptake. Decreased quinolone uptake is associated with either increased impermeability or the activity of efflux pumps [105]. Hydrophobic quinolones alone can transit across the phospholipid bilayer but all quinolones can enter the cell through porins [105]. It is possible that the phospholipid bilayer of strain RM4018 has reduced permeability towards hydrophobic quinolones in conjunction with modifications in the porins to permit passage of only hydrophilic quinolones. A more likely scenario is the presence of a hydrophobic quinolone-specific efflux pump. Examples of pumps with specificity towards one class of quinolone are known; for example, the NorA protein of Staphylococcus aureus has been shown to be involved in the specific efflux of hydrophilic quinolones [106].

Virulence determinants

Putative virulence determinants have been identified in Campylobacter, but little is known about potential virulence factors in A. butzleri. Since campylobacterioses and reported A. butzleri-related illnesses have similar clinical outcomes [10], it might be expected that some C. jejuni virulence factors would be found in Arcobacter. In fact, some virulence determinants identified in C. jejuni, have homologs within A. butzleri. For example, the fibronectin binding proteins CadF and Cj1349 have homologs in strain RM4018 (AB0483 and AB0070, respectively). Moreover, homologs of the invasin protein CiaB, the virulence factor MviN, the phospholipase PldA and the TlyA hemolysin are present in strain RM4018 (AB1555, AB0876, AB0859 and AB1846, respectively). However, it has not been determined if these putative virulence determinants are functional or if their function and role in A. butzleri biology is similar to the function of their Campylobacter homologs.

On the other hand, several Campylobacter virulence-associated genes are not present in the RM4018 genome. Most notably, the genes encoding the cytolethal distending toxin CDT (cdtABC) are absent from strain RM4018. CDT is an exotoxin which irreversibly blocks eukaryotic cells in the G₁ or G₂ phase [107]; cdtABC genes have been identified in Helicobacter hepaticus and several characterized Campylobacter genomes. The absence of cdtABC in strain RM4018 correlated well with a study by Johnson and Murano [108] which was unable to detect cdt genes in Arcobacter by PCR. A. butzleri strain RM4018 also contains no PEB1 or JlpA adhesin homologs [109], [110].

Analysis of the strain RM4018 genome identified two additional putative virulence determinants: irgA (ab0729) and hecAB (ab0941-ab0940). The irgA gene in V. cholerae encodes an iron-regulated outer membrane protein [111], and an IrgA homolog has been demonstrated to play a role in the pathogenesis of urinary tract infections by uropathogenic E. coli [112]. Adjacent to irgA is ab0730, which encodes a putative IroE homolog. IroE is a siderophore esterase found also in uropathogenic E. coli [113]. The other novel virulence determinant, HecA, is a member of the filamentous hemagglutinin (FHA) family, and hecB encodes a related hemolysin activation protein. FHA proteins are distributed widely among both plant and animal pathogens, e.g. HecA in Erwinia crysthanthemi contributes to both attachment and aggregation and is involved in epidermal cell killing [114]. Consistent with this distribution, RM4018 HecA homologs occur in both plant- (Pseudomonas syringae, Ralstonia solanacearum) and animal-pathogens (Burkholderia cepacia, Acinetobacter spp. and uropathogenic E. coli).

Interacting with the environment: chemotaxis and signal transduction

Microbial life is characterized by continuous interactions between bacteria and their environment. The ability of microorganisms to monitor environmental parameters is a prerequisite for survival. Hence, bacteria have evolved different mechanisms such as sensory histidine kinases, methyl-accepting chemotaxis proteins, sigma(σ)/anti-sigma factor pairs, and adenylate and diguanylate cyclases to monitor and rapidly adapt to changes in their environment.

Arcobacter butzleri is well equipped with a large number of these systems, allowing it to survive in diverse ecological niches. Prime bacterial mechanisms of environme