The Brucella abortus strain S19, a spontaneously attenuated strain, has been used as a vaccine strain in vaccination of cattle against brucellosis for six decades. Despite many studies, the physiological and molecular mechanisms causing the attenuation are not known. We have applied pyrosequencing technology together with conventional sequencing to rapidly and comprehensively determine the complete genome sequence of the attenuated Brucella abortus vaccine strain S19. The main goal of this study is to identify candidate virulence genes by systematic comparative analysis of the attenuated strain with the published genome sequences of two virulent and closely related strains of B. abortus, 9–941 and 2308. The two S19 chromosomes are 2,122,487 and 1,161,449 bp in length. A total of 3062 genes were identified and annotated. Pairwise and reciprocal genome comparisons resulted in a total of 263 genes that were non-identical between the S19 genome and any of the two virulent strains. Amongst these, 45 genes were consistently different between the attenuated strain and the two virulent strains but were identical amongst the virulent strains, which included only two of the 236 genes that have been implicated as virulence factors in literature. The functional analyses of the differences have revealed a total of 24 genes that may be associated with the loss of virulence in S19. Of particular relevance are four genes with more than 60bp consistent difference in S19 compared to both the virulent strains, which, in the virulent strains, encode an outer membrane protein and three proteins involved in erythritol uptake or metabolism.
Citation: Crasta OR, Folkerts O, Fei Z, Mane SP, Evans C, et al. (2008) Genome Sequence of Brucella abortus Vaccine Strain S19 Compared to Virulent Strains Yields Candidate Virulence Genes. PLoS ONE 3(5): e2193. doi:10.1371/journal.pone.0002193
Editor: Jean-Nicolas Volff, Ecole Normale Supérieure de Lyon, France
Received: December 6, 2007; Accepted: March 13, 2008; Published: May 14, 2008
Copyright: © 2008 Crasta et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The project was funded by the Virginia Bioinformatics Institute to Bruno Sobral. The data analysis was also funded through the National Science Foundation (NSF grant # OCI-0537461 to O. Crasta). Preparation and submission of genomic sequences to GenBank and annotation were funded by NIAID Contract HHSN26620040035C to Bruno Sobral and are available at PATRIC's site (http://patric.vbi.vt.edu/).
Competing interests: The authors have declared that no competing interests exist.
Brucella spp. are gram-negative, facultative, intracellular coccobacilli that may cause brucellosis in humans and livestock. The main economic impact of infection in animals is reproductive failure , whereas in humans it is undulant fever and if untreated, a debilitating chronic disease . Brucella are also identified as potential agricultural, civilian, and military bioterrorism (category B) agents. They are particularly hard to treat as they infect and replicate within host macrophages. Brucellosis in livestock animals is controlled by vaccination . Human brucellosis is treatable with antibiotics, though the course of antibiotic treatment must be prolonged due to the intracellular nature of Brucella.
B. abortus “strain 19” or S19 (here after, S19) is a spontaneously attenuated strain discovered by Dr. John Buck in 1923 , . The underlying molecular or physiological mechanisms causing the loss of virulence are not well understood. Live, attenuated strain S19 had been used worldwide since the early 1930s as an effective vaccine to prevent brucellosis in cattle, until it was replaced by strain RB51 during the 1990s. S19 maintains its smooth appearance derived from the presence of the extracellular lipopolysaccharide (LPS) while the other vaccine strain, RB51, with rough characteristics devoid of O-chain, does not elicit antibodies against the O-side polysaccharide , . Further attenuation or optimization of S19 will be necessary to develop a human vaccine strain, which could be approved through the “Animal Rule” regulatory mechanism . Such modification could result from the expression of additional vaccine candidate proteins to enhance vaccine efficacy , or from the inactivation of genes encoding additional virulence factors to reduce residual virulence found in humans.
There are four papers describing the genomes from the following different strains/species of Brucella: B. melitensis , B. suis, B abortus strain 2308 (here after 2308), and B. abortus strain 9–941 (here after 9–941) –; in addition, two more genome sequences of Brucella sp. have been sequenced and have been made available by the Pathosystems Resource Integration Center (PATRIC, http://patric.vbi.vt.edu/). All sequenced strains to date are virulent. Therefore, we set out to sequence the first genome of an attenuated, live vaccine strain, with the main objective of identifying the genes associated with the virulence or lack thereof, through comparison of the newly sequenced genome with that of the virulent counterparts.
We used pyrosequencing (454 Life Sciences Corporation)  together with conventional sequencing to determine the complete genome sequence of S19. Here we have described the newly sequenced genome and identification of unique genes by its comparison to genomes of virulent strains of B. abortus, 2308 and 9–941 , . Comparative genomic analysis identified a number of candidate genes that can be mutated with the aim of further attenuating wild-type or other vaccine strains.
Results and Discussion
The combination of pyrosequencing and Sanger sequencing allowed for the rapid (one day of sequencing) and comprehensive (more than 99.5% of the genome) closure and assembly of the genome of S19. The average length of the sequence reads was 110 bp. Using the Roche GS-FLX™ we were able to improve read lengths to an average of 230 bp (data not shown).
Genome Sequence Properties
The 3.2 Mb S19 genome is comprised of two circular chromosomes (Table 1): one 2,122,487 bp long and the other 1,161,449 bp long. The average GC content of the two chromosomes is 57%. Not surprisingly, S19's genome showed remarkable similarity in size and structure to those of its virulent relatives, B. abortus 9–941 and 2308. The size of the S19 genome is within 5 kb of 9–941 (3.283Mbp) and 2308 (3.278 Mbp) genomes. The S19 genome sequence shows over 99.5% similarity compared to the genomes of 9–941 and 2308.
Table 1. Genome properties of the newly sequenced genome of B. abortus strain S19 in comparison with the known genome sequence of two virulent strains.doi:10.1371/journal.pone.0002193.t001
The S19 chromosomes and their comparison to the 9–941 and 2308 chromosomes are shown in Figure 1. The online version of the Figure 1 is interactive (and is available at http://patric.vbi.vt.edu/). A total of 2,047 and 1,086 open reading frames (ORFs) were identified on the first and second chromosomes, respectively. More than 96% of these ORFs are identical to the ORFs of 9–941 and 2308. Functional assignment of the ORFs was carried out by BLASTP searches against four Brucella genomes. S19 exhibits very high genome-wide collinearity with 9–941 and 2308 in Chromosome 1. Chromosome 2 has perfect collinearity with both 9–941 and 2308 (Figure 2.). Of the 3,062 predicted ORFs, more than 79% had BLASTP hits to the cluster of orthologous groups (COG) database with an e-value less than 1e–4. A total of 571 ORFs (18%) are hypothetical proteins (Table 2 and Supplemental Table S1).
Figure 1. Complete DNA sequence of Brucella abortus strain S19.
The concentric circles show, reading outwards: GC skew, GC content, AT skew, AT content, COG classification of proteins, CDS on reverse strand, ORFs on three frames in reverse strand, ORFs on three frames in forward strand, CDS on forward strand and COG classification of proteins on forward strand. The genes that differ from both 2308 and 9–941 strains are labeled.doi:10.1371/journal.pone.0002193.g001
Figure 2. Comparative genomic analysis of the Brucella abortus strain S19 with the virulent strains 9–941 and 2308 using the MUMer program.doi:10.1371/journal.pone.0002193.g002
Table 2. COG-based functional categories of B. abortus S19 coding sequences.doi:10.1371/journal.pone.0002193.t002
Genome wide comparisons were between S19 and the virulent 9–941 genomes to identify all the single nucleotide polymorphisms (SNPs) between the two genomes. A more stringent criteria was used to identify SNPs using all the sequenced reads, then to identify the differences in ORFs (sections below). A total of 201 SNPs were identified in S19 when compared to 9–941, and are listed in Table 3. The exact position and alignment of all the SNPs at the nucleotide level and protein level (if the SNP is within an ORF) is given in the Supplemental Table S2. Forty-four of the 201 SNPs (22%) were located in intergenic regions. Forty-nine SNPs were synonymous substitutions, encoding the same amino acid (aa). Sixty-five SNPs were conservative, non-synonymous substitutions, encoding a different aa with similar properties. Radical non-synonymous substitutions were found in 36 of the SNPs, resulting in the incorporation of an aa with a net change in charge or polarity. These SNPs between S19 and 9–41 were also compared for consistency by comparing the S19 genome to 2308 genome. A total of 39 single nucleotide differences in ORFs that were consistently different between S19 and its two virulent counterparts and their relevance to virulence are described in the section below.
Table 3. Single nucleotide polymorphisms detected in B. abortus strain S19 as compared to the strain 9–941.doi:10.1371/journal.pone.0002193.t003
Identification of Virulence Associated Differences between the Attenuated and Virulent Strains
The main focus of the work was to identify all the ORFs that are different in S19 when compared to its virulent relatives, 9–941 and 2308, which provide a complete basis for the attenuation of S19. Pairwise and reciprocal comparisons were made between S19 and the published genome sequences of 9–941 and 2308, to identify genes that are 100% identical and consequently, those non-identical ORFs with any differences (Table 4). In each pairwise comparison, predicted genes from one genome were aligned to the whole genome sequence of the second strain and vice versa, to allow for differences in gene annotation among the genomes. In each of the genes-to-genome pairwise comparisons, more than 95% of the genes from strain-1 were identical to a corresponding sequence from the whole genome of strain-2. The bulk of the rest of the genes (an average of 4%) had only one nucleotide (nt) difference, while less than 1% of the genes showed differences of more than 1 nucleotide. The number of genes that were non-identical was greater in the comparison between S19 and 9–941 than in the comparison between S19 and 2308. The results of the pairwise and reciprocal comparisons between S19, 9–941 and 2308 are given in the Supplemental Table S3.
Table 4. Reciprocal gene to genome pairwise comparison between B.abortus strains S19, 9–941, and 2308.doi:10.1371/journal.pone.0002193.t004
The pairwise and reciprocal genes-to-genome comparisons allowed us to account for the genes that were predicted in one genome but not predicted in the other. A total of 260 and 214 ORFs identified in 9–941 and 2308, respectively, were not predicted in S19. More than 95% of these sequences were identical to S19 and more than 91% of these were annotated as “hypothetical”. Similarly, of the ORFs identified in S19 a total of 321 and 311 were not predicted in 2308 and 9–941, respectively, and more than 90% of these S19 sequences were identical to those in the 2308 and 9–941 genomes. As expected, the differences in number of ORFs between the three strains were much smaller when the same methods were applied for ORF prediction (data not shown). The differences in the gene annotations are corrected through the curation efforts by the PATRIC project and the new annotations are available at http://patric.vbi.vt.edu/.
Comprehensive pairwise and reciprocal genes-to-genome comparisons of all the predicted ORFs in the three genomes of S19, 2308 and 9–941 revealed only 263 ORFs that were non-identical (<100% homology) between S19 and any of the two virulent genomes, 9–941 and 2308. The data are summarized in Table 5 and the details of the genes and their differences are given in the Supplemental Table S4. Of the 263 ORFs identified as non-identical between S19 and any of the two virulent strains (Table 5), 70 ORFs showed nucleotide changes but not aa changes, therefore they were not further pursued from the perspective of explaining the differences in virulence. A total of 148 ORFs showed differences between S19 and only one of the two virulent strains, while there was no difference compared to the other. This included some of the ORFs that have been implicated in virulence (.e.g., AroA, 3-phosphoshikimate 1-carboxyvinyltransferase), but are not discussed because of their indifference in one of the two virulent strains. The remaining 45 ORFs showed consistent sequence deviation in S19 compared to both the virulent strains, while both virulent strains maintained identical sequences. These 45 ORFs that were consistently different between the attenuated S19 and the two virulent strains were evaluated to identify candidate virulence associated ORFs. The clusters of orthologous groups of proteins ( COG) functional classification of the 45 ORFS, with consistent differences between S19 and both the virulent strains (OCDs), and all of the 263 non-identical differences, is shown in Table 6. While the non-identical differences were distributed in a total of 19 COG classes, the OCDs were clustered in 11 COG classes (excluding no hits and unknown).
Table 5. Number of genes identified to be different between attenuated strain S19 genome and the virulent strains 9–941 and 2308.doi:10.1371/journal.pone.0002193.t005
Table 6. COG-based functional categories of ORFs identified as differences between S19 and the virulent B. abortus strains 9–941, and 2308.doi:10.1371/journal.pone.0002193.t006
The details on the differences between S19 and the two virulent strains of all the 45 OCDs are shown in Table 7. Four of the 45 OCDs (ORFs BruAb1_0072, BruAb2_0365, BruAb2_0366, and BruAb2_0372 in 9–941) showed more than 60 bp differences between the attenuated and the virulent strains and were considered as “Major Virulence Associated Differences” (“priority 0” in Table 7). The remaining 41 OCDs had only less than 10 bp difference and hence were considered as “Minor OCDs”. All the major virulence associated differences and some of the OCDs with possible association to virulence are described in the sections below. Further follow up experiments are being designed to test the functions of some of the virulence associated ORFs by mutation studies of S19 and the virulent strains and their responses during infection, and hence they are not part of this manuscript.
Table 7. List of ORFs with consistent differences (OCDs) between the attenuated strain S19 and the two virulent strains, 2308 and 9–941, but not different within the virulent strains.doi:10.1371/journal.pone.0002193.t007
Major Virulence Associated Differences
Rearrangement in an outer membrane protein: The most striking and consistent difference between S19 and the two virulent strains was in the region of the ORF, BAB1_0069 of 2308, encoding a putative 1,333 aa outer membrane protein. Compared to both 2308 and 9–941, this locus in S19 suffers a 1,695 nt. deletion, corresponding to nucleotides 805–2499 of BAB1_0069. The deletion removes amino acids 269–833 and therefore the predicted ORF in S19 is only 768 aa long (Figure 3).
Figure 3. Diagramatic representation of the alignment of the outer membrane protein loci (BAB1_0069) across sequenced Brucella genomes.doi:10.1371/journal.pone.0002193.g003
A structural similarity search against the Protein Data Bank revealed two significant hits to Yersinia enterocolitica YadA and the Haemophilus influenzae Hia adhesins, respectively. Both YadA and Hia are important virulence factors involved in adhesion, invasion, and serum resistance. The structural similarity between YadA and BAB1_0069 resides in amino acids 62–265 of YadA and 1064–1251 of BAB1_0069. This region in YadA was recently shown to consist of a globular head region consisting of nine-coiled left-handed parallel β-rolls (LPBR) . Phagocytic uptake by host macrophages is clearly essential for a robust immune response, and evasion of this uptake is a protective mechanism for extra-cellular pathogens such as the Yersiniae. Brucella on the other hand are intracellular pathogens and infect initially and mainly macrophages, and therefore depend on an efficient mechanism for adhesion and cellular uptake.
Although the 1,695 nt deletion in the outer membrane protein is consistent between S19 and both the virulent strains, further studies are needed to associate this protein to the lack of virulence in S19, as further examination of the same ORF in other virulent species of Brucella reveals the presence of similar deletions (Figure 3) in B. melitensis and B. suis. The B. melitensis locus contains 3 deletions. The first deletion results in a 512 aa deletion. The locus also has two single nucleotide deletions, causing a frame shift, such that the ortholog in B. melitensis contains only the terminal 365 aa (NP_540789.1). The B. suis genome has two deletions, resulting in an ORF predicted to encode 740 aa. The deletions in S19 and B. suis are identical in length and position and further examination of the nucleotide and protein sequences and alignments revealed the presence of a tandem repeat sequence of 339 nucleotides (113 aa). Six copies of the repeated sequence occur in the full-size ORFs of strains 2308 and 9–941 (Figure 3). The variant ORFs could have arisen by deleting portions of the coding regions/proteins through ‘recombination’ between the 1st and 6th repeat sequence in strain S19 and B. suis, and the 1st and 5th repeat in B. melitensis. Although the presence of similar deletions in other virulent species eliminates the possibility of association of the deletion to lack of general virulence, it is possible that the deletion may be associated with species specific (B. abortus) host preference, which needs to be tested in mutation experiments.
Rearrangements in the erythritol catabolic operon (eryC, eryD) and related transporter (eryF): The second largest gene rearrangement in S19 compared to both the virulent strains, 2308 and 9–941, occurs in the erythritol (ery) operon. The erythritol operon contains 4 ORFs for eryA, eryB, eryC and eryD respectively. Compared to the virulent strains, S19 has a 703 nucleotide deletion which interrupts both the coding regions of eryC (BAB2_0370) and eryD (BAB2_0369). The deletion affects the C terminal part of eryC and the N-terminal part of eryD proteins from B. abortus strains 2308 (BAB2_0369), 9–941 (BruAb2_0365) and B. suis (BRA0867). Figure 4A shows the alignment of the predicted protein with the C-terminal part of eryD. The deletion in eryC and eryD ORFs of S19 has been previously shown , . The importance of this deletion in the attenuation of S19 has also been studied using Tn5 insertions and complementation analysis, revealing that it is not sufficient or required for virulence in a mouse model .
Figure 4. Protein alignment of the ribose transporter protein, S19 BAbS19_II03540; A. with the homologs from strain 9–941 (BruAb2_0372), 2308 (BAB2_0376) and B. suis (BRA0859); B. Protein alignment of the ribose transporter protein from strain 2308 (BAB2_0376) with the putative Rhizobium leguminosarum eryF protein.
The sequence deleted in S19 BAbS19_II03540 is indicated by the strike-through mark-up of the 2308 sequence.doi:10.1371/journal.pone.0002193.g004
Eryrthritol metabolism by Brucella has been identified as a trait associated with the capability of the pathogen to cause abortions in livestock. The preferential growth of Brucella in the foetal tissues of cattle, sheep, goats and pigs was also shown to be due to the high concentration of erythritol . Although Brucella infects and cause brucellosis in other organisms such as human, rat, rabbit and guinea pig, overwhelming infection of the placental and foetal tissues is not observed, which is also associated with low concentrations of erythritol . According to Garcia-Lobo and Sangari , the cultures of “strain 19” provided by the USDA before 1956 showed differences in growth in the presence of erythritol and in erythritol oxidation rates. During that time, erythritol sensitive cultures were selected and used to substitute the previous batches of vaccine, which was then renamed “US19” or just “S19”. Jay F. Sperry and Donald C. Robertson  elucidated the pathway of erythritol catabolism in Brucella using radiolabelling experiments. Later, Tn5 mutagenesis of virulent 2308 revealed four ery genes (eryABCD) proposed to exist as an operon . The ery operon in S19 was also analyzed and shown to contain a deletion of 702 bp affecting two genes, eryC and eryD .
In a murine model, when attenuated S19 and virulent 2308 strains were compared to genetically engineered strains, including: (1) a knock-out mutant of 2308 (ΔeryCD), (2) a naturally reverted S19 strain (ery resistant), and (3) S19 strain transformed with the wild type ery operon, there was no direct correlation between erythritol metabolism and in vivo colonization . However, the experiments that have been performed in mice do not address the question of whether or not an eryCD mutation would attenuate B. abortus 2308 in a pregnant ruminant as the B. abortus dhbC mutant is shown to be extremely attenuated in pregnant cattle . These results suggest that S19 has lost some additional and essential, yet unknown, mechanism of virulence in mice.
S19 also contains a 68 nucleotide deletion in the ORF corresponding to BAB2_0376 of B. abortus 2308, which results in a 114 aa N terminal truncation (Figure 4 AB). BAB2_0376 encodes a putative inner-membrane translocator sodium dicarboxylate symporter, which is highly conserved among many gram negative bacteria, proteobacteria, enterics and gram positive bacteria. Christopher K. Yost et al.  recently characterized a region of the Rhizobium leguminosarum 3841 responsible for erythritol uptake and utilization, which is highly conserved within the corresponding genome region of Brucella spp. In R. leguminosarum 3841 the eryABCD erythritol catabolic operon is flanked by a putative operon containing a hypothetical protein, a putative nucleotide binding protein (EryE), permease (EryF) and a periplasmic binding protein (EryG). Transposon mutation of eryF abolished erythritol uptake. Because BAB2_0376 has 85% sequence similarity with the predicted eryF ORF (pRL120201) and the conservation of gene order with the eryEFG operon, it is very likely that BAB2-0376 also encodes a putative erythritol transporter. Hence, we hypothesize that mutation of the gene in S19, in addition to the large deletion in the eryABCD operon, further contributes to the inability of S19 to metabolize erythritol. Further experiments are designed to test the combined impact of the eryCDF on the attenuation of the virulent strains.
Minor Virulence Associated Differences
As shown in Table 7, a total of 41 ORFs were identified as minor OCDs, which showed a consistent difference of less than 10bp between the attenuated S19 strain and both the virulent strains, 9–941 and 2308. These ORFs were grouped into 11 classes based on the COG classification. Only two minor OCDs, carboxyl transferase (BruAb1_0019) and Enoyl-acyl-carrier proteins and (BruAb1_0443) showed more than 1 bp difference (“priority 1” in Table 7) resulting in frameshifts, while the remaining 39 minor OCDs showed only single nucleotide differences (“priority 2 to 4” in Table 7). Among the 39 minor OCDs that showed single nt differences, 12 ORFs showed either variation among reads within S19 or were transposases (“priority 5 to 4” in Table 7) and hence were omitted from discussion. Six OCDs had frameshifts between S19 and the two virulent strains (“priority 2” in Table 7), which included one protein involved in Lipid transport and metabolism (BAB1_0967), an IclR family transcriptional regulator (BruAb1_0229), aUDP-glucose 4-epimerase (BruAb2_0379), a Glycerophosphodiester phosphodiesterase domain-containing protein (BruAb1_1172), an aldehyde dehydrogenase family protein (BruAb2_1016), and an intimin/invasin family protein (BAbS19_I18830). Some of the 12 minor OCDs (“priority 3” in Table 7) that resulted in the net charge change included excinuclease ABC subunit B (BruAb1_1504), ABC transporter, periplasmic substrate-binding protein (BruAb1_0208), IolC myo-catabolism protein (BruAb2_0517), outer membrane autotransporter (BAbS19_I18870), CTP synthetase (BruAb1_1140), Omp2b, porin (BruAb1_0657), and three hypothetical proteins. We also identified 9 OCDs that showed amino acid changes without change in the net charge (“priority 4” in Table 7).
Some of the minor OCDs and their relevance to attenuation (besides the four major differences) are described below. Although we have discussed the association of some of the OCDs to the virulence (in the virulent strains) or lack there of (in S19), further mutation experiments are needed and are underway to verify their functional relevance. As mentioned before, the main focus of this project is to catalog all the consistent differences between the attenuated strain S19 and the two virulent strains (Table 7).
We identified a total of four ORFs under the classification of lipid transport and metabolism. They are: carboxyl transferase family protein (BruAb1_0019), enoyl-[acyl-carrier-protein] reductase [NADH] (BruAb1_0443), membrane protein involved in aromatic hydrocarbon degradation (BAB1_0967), and acetyl-coenzyme a synthetase (BAbS19_II07300). The differences in the first four OCDs caused frameshift or premature stop codons in the S19 ORF, while the difference in the last caused change in aa (V to I) without any change in the net charge. The impact of the difference in carboxyl transferase family protein needs a follow up to assess its impact on the virulence, as B. abortus encodes two additional carboxyl transferases, BAB1_1210 (510 aa) and BAB1_2109 (AccD, 301 aa). It is not clear if the mutation in BAB1_0019 may be compensated for by the presence of two additional ACCase beta subunit genes. Mycobacterium tuberculosis encodes six AccDase proteins, and at least three of these have a putative role in virulence –. The locus encoding enoyl-[acyl-carrier-protein] reductase [NADH] (BruAb1_0443) showed 2 bp difference between S19 and the two virulent strain, while the other two ORFs (BAB1_0967, and BAbS19_II07300) showed single bp difference.
The locus encoding transcriptional regulator, IclR family (BruAb1_0229) contains a one base deletion in S19, causing a frameshift in the C-terminus and premature termination of the ORF, changing the C-terminal 28 aa of BruAb1_0229. BruAb1_0229 encodes a 264 aa protein with homology to the IclR family of regulatory proteins . The C-terminal half of the protein contains the Helix-turn-helix DNA binding motif . The change of the C-terminal 28 aa in the S19 protein removes or changes Helix 9 of the TM-IclR structure.
Comparison of the differences between S19 and the virulent strains to the Brucella virulence factors
As mentioned before, the main focus of this project was to identify candidate ORFs that are associated with the virulence (in 9–941 and 2308) or lack there of (in S19). A comprehensive analysis of the genomes revealed 263 non-identical differences. As described above, we have identified 45 ORFs that were consistently different between S19 and the two virulent strains. As an independent evaluation, we also compared all the 263 non-identical differences (as listed in the Supplemental Table S4) to the virulence factors that are described in literature as experimentally characterized in Brucella, which is described below.
Large-scale screens and testing in model systems have been performed in Brucella to identify virulence factors associated with the pathogenesis and virulence. Rose-May Delrue et al.,  studied the literature and identified a total of 192 virulence factors that have been characterized using 184 attenuated mutants, as well as an additional 44 genes, which have been characterized for virulence in Brucella by different groups –. The sequences of all these 236 genes were compared to the 263 ORFs that we have identified as non-identical differences between S19 and any of the two virulent strains of B. abortus. The comparison yielded a total of 26 ORFs that were non-identical between S19 and any of the two virulent strains (<100% homology) as shown in Table 8. However, out of the 22, only three ORFS (BRA0866:BAB2_0370, BRA1168:BAB2_1127, and BR1296:BruAb1_1350) were identified as showing consistent differences between S19 and the two virulent strains (9–941 and 2308). Amongst these only two (BAB2_0370, BAB2_1127) were included as virulence associated ORFS, as the third one did not show difference at the amino acid level. The ORF, BAB2_0370 (EryC), was identified as a major virulence associated difference and is described in detail in the above section. The other ORF, BAB2_1127, which encodes a hypothetical protein associated with UPF0261 protein CTC_01794 (COG5441S) in B melitensis (BMEII0128) was screened in a murine infection model through signature-tagged mutagenesis (STM), and was used to identify genes required for the in vivo pathogenesis of Brucella  with a combined attenuation score of 4 . None of the other ORFs listed in Table 8 were consistently different between S19 and the two virulent strains.
Table 8. Comparison of the list experimentally characterized Brucella virulence factors to the non-identical differences between S19 and the virulent strains.doi:10.1371/journal.pone.0002193.t008
Besides the comparisons of the sequences of the ORFs of the 236 virulence factors, the intergenic regions upstream of these genes (and all other ORFs in the genome) were also used to compare the attenuated and virulent strains. None of the differences at the intergenic regions were found to be consistent between S19 and both the virulent strains (data not shown).
The Brucella abortus strain S19 is a spontaneously attenuated strain, which has been used as a vaccine strain in vaccination of cattle against brucellosis for six decades , . Although it has been studied extensively, the physiological and molecular mechanisms causing the attenuation are not known . The classical studies that evaluated the attenuation of S19 were the discovery of partial deletion and loss of function of two proteins ery operon (eryCD), and their characterization and mutational analysis –. At least in murine models, the deletion of these two genes was not enough to cause attenuation, suggesting that S19 has lost some additional and essential, yet unknown, mechanism of virulence in mice.
We have determined the complete genome sequence of S19 and conducted a comprehensive comparative analysis using the whole genome sequence of two virulent strains, 9–941 and 2308 and the newly sequenced attenuated strain, S19. Our comparative analyses agreed with previous studies to reveal >99% homology among the genomes sequences . The differences in the method of gene prediction used in three different genomes has been corrected and shown on the PATRIC website (http://patric.vbi.vt.edu/). We conducted pairwise and reciprocal gene-to-genome comparisons to identify all of the 263 non-identical differences between S19 and the two virulent genomes, out of which only 45 ORFs or “OCDs” were consistently different between the attenuated S19 strain and both the virulent strains, 9–941 and 2308.
Among the 45 OCDs, only four ORFs had more than 60 nt difference between S19 and the virulent strains with no difference observed within the virulent strains (priority “0” in Table 7). The results revealed one additional ORF encoding protein involved in erythritol uptake (eryF), while confirming the previous findings on the 703 nt deletion in eryCD. We also identified an outer membrane protein with 768 aa deletion in S19. Although this difference is consistent when compared to the virulent strains of B. abortus, the virulent strain of B. suis also contain this deletion, making it a less probable candidate for attenuation. However, its role within B. abortus needs testing.
Besides the four major differences, we identified a prioritized list of 24 OCDs with minor but consistent differences between S19 and the two virulent strains, which included eight OCDs with frameshifts (priority 1 and 2), 12 OCDs with aa changes that cause a change in the net charge of the protein (priority 3) and nine OCDs with aa changes without change in the net charge (priority 4). Some of the intriguing differences with possible relevance to attenuation include four proteins involved in lipid transport and metabolism, two proteins involved in transcription, two transporter proteins, two outer membrane proteins, and several hypothetical proteins. We believe the characterization of these few proteins using mutation and host response studies will yield an account of the attenuation of S19.
Materials and Methods
Strain Information, Extraction and Characterization of The Genomic DNA
B. abortus S19 was obtained from the National Animal Disease Center collection. It was originally isolated from the milk of American Jersey Cattle by Dr. John Buck in 1923 , . Total genomic DNA was extracted and purified by the modification of a previously described method . An aliquot of the DNA was subjected for analysis using the Bioanalyzer (Agilent technologies) and was confirmed for no degradation of the DNA. An aliquot of 10ug of DNA was used for the sequencing via pyrosequencing (see below), and the remaining stock was maintained for further sequencing and completion of the gaps.
Genome Sequencing, Whole Genome Assembly
The first round of high-throughput sequencing was performed via pyrosequencing . A total of two, four-hour runs were performed to generate a total of ~800 thousand sequences with an average length of about 100 bases, resulting in more than 20X coverage of the whole genome of the strain. The quality filtered reads were then assembled into contigs using the Newbler assembler (http://www.454.com/). A total of 701 contigs with at least two contributing fragments were formed, of which 172 contigs had sequence lengths ranging from 0.5 to 123 kb, with an average of 18.8kb.
The 172 contigs were aligned to the whole genome sequence of the B. abortus 9–941  to identify the putative gaps to be sequenced in the whole genome of the B. abortus strain S19. Primers were designed and the genomic DNA from the B. abortus strain S19 was used as a template in PCR to amplify the segments that needed to be sequenced. The purified PCR amplicons were used as templates in sequencing. The newly generated sequences, together with the contigs, were used to determine the whole genome sequence.
Gene Prediction and Annotation
Putative protein-encoding genes of the S19 genome were identified with Glimmer . Genes consisting of fewer than 33 aa were eliminated, and those containing overlaps were manually evaluated. Start sites of each predicted gene were tuned by TiCO . Sequences from the intergenic regions were compared to non-redundant protein databases to identify genes missed by the Glimmer prediction. tRNAs were identified with tRNAscan-SE , while ribosomal RNAs were identified by comparing the genome sequence to the rRNA database . The whole genome sequence of B. abortus S19 is deposited in GenBank (accession # CP000887 and CP000888). The whole genome sequence and existing annotations were also submitted for additional curation (gene prediction and protein annotation) using the genome annotation pipeline of the PathoSystems Resource Integration Center (PATRIC)  and are made available at http://patric.vbi.vt.edu/.
Whole genome sequences and the predicted gene sets of B. abortus 2308 and 9–941 were downloaded from NCBI RefSeq . SNPs between S19 and 9–941 genomes, as well as between S19 and 2308 genomes, were identified by mapping all of the 800,000 reads generated by the 454 machine to the reference genomes (9–941 and 2308, respectively) using the 454 whole genome mapping software which includes a high-confidence SNP identification module–MutationDetector (http://www.454.com).
To identify potential genes that differ between the attenuated strain S19 and other two B. abortus virulent strains, 2308 and 9–941, pair-wise and reciprocal comparisons were performed by aligning the predicted genes of one strain to the whole genome sequence of the second strain and vice versa. The genomes were automatically compared for missed gene calls, indels, frameshifts, and other sequence variants by a program called GenVar . The improved annotations of the ORFs were performed by PATRIC and are made available at http://patric.vbi.vt.edu/.
Functional assignment of the ORFs of Brucella abortus Strain S19 using BLASTP searches against the protein datasets
(0.54 MB XLS)
Single Nucleotide Polymorphisms (SNPs) identified between S19 genome and 9–941 genome
(0.20 MB XLS)
Pair-wise and reciprocal comparisons between the genes and the genomes of the strains S19, 9–941 and 2308
(1.29 MB XLS)
List of genes identified to be different between attenuated strain S19 genome and the virulent strains, 9–941 and 2308
(0.09 MB XLS)
The authors thank Drs. Stephen Boyle and Shirley Halling for their valuable suggestions and comments in preparation of this manuscript, Drs. Gerard Irzyk and Tom Jarvie for their help in generating the first draft sequence, Drs. Yan Zhang, Saroj Mohapatra, and Mr. Thero Modise for curation of the data, and Mrs. Emily Berisford and Mrs. Carol Volker for editing the manuscript. The authors are also grateful to Dr. M. Roop and Reviewer#2 for their valuable feedback and comments.
Conceived and designed the experiments: OC CE BS SM. Performed the experiments: CE SM BB. Analyzed the data: OC OF ZF SM GY LD. Contributed reagents/materials/analysis tools: BB. Wrote the paper: OC CE OF SM BB.
- 1. Corbel MJ (1989) Brucellosis: Epidemiology and prevalence worldwide. In: Young EJ, Corbel MJ, editors. Brucellosis: clinical and Laboratory Aspects. Boca Raton, FL: CRC Press. pp. 25–40.
- 2. Young EJ (2000) Brucella species. In: Mandell GL, Bennet JE, Dolin R, editors. Principles and Practice of Infectious Diseases. Philadelphia: Churchill Livingstone. pp. 2386–2393.
- 3. Cutler SJ, Whatmore AM, Commander NJ (2005) Brucellosis–new aspects of an old disease. J Appl Microbiol 98 6: 1270–1281.
- 4. Graves RR (1943) The Story of John M. Buck's and Matilda's Contribution to the Cattle Industry. Journal of American Veterinary Medical Association 102: 193–195.
- 5. Nicoletti P Nielsen K, Duncan JR, editors. (1990) Vaccination. Animal Brucellosis Boca Raton: CRC Press. 284–299.
- 6. Diaz R, Jones LM, Leong D, Wilson JB (1968) Surface antigens of smooth brucellae. J Bacteriol 96 4: 893–901.
- 7. Poester FP, Goncalves VS, Paixao TA, Santos RL, Olsen SC, et al. (2006) Efficacy of strain RB51 vaccine in heifers against experimental brucellosis. Vaccine 24 25: 5327–5334.
- 8. Langford MJ, Myers RC (2002) Difficulties associated with the development and licensing of vaccines for protection against bio-warfare and bio-terrorism. Dev Biol (Basel) 110: 107–112.
- 9. Vemulapalli R, He Y, Cravero S, Sriranganathan N, Boyle SM, et al. (2000) Overexpression of protective antigen as a novel approach to enhance vaccine efficacy of Brucella abortus strain RB51. Infect Immun 68 6: 3286–3289.
- 10. Chain PS, Comerci DJ, Tolmasky ME, Larimer FW, Malfatti SA, et al. (2005) Whole-genome analyses of speciation events in pathogenic Brucellae. Infect Immun 73 12: 8353–8361.
- 11. DelVecchio VG, Kapatral V, Redkar RJ, Patra G, Mujer C, et al. (2002) The genome sequence of the facultative intracellular pathogen Brucella melitensis. Proc Natl Acad Sci U S A 99 1: 443–448.
- 12. Halling SM, Peterson-Burch BD, Bricker BJ, Zuerner RL, Qing Z, et al. (2005) Completion of the genome sequence of Brucella abortus and comparison to the highly similar genomes of Brucella melitensis and Brucella suis. J Bacteriol 187 8: 2715–2726.
- 13. Paulsen IT, Seshadri R, Nelson KE, Eisen JA, Heidelberg JF, et al. (2002) The Brucella suis genome reveals fundamental similarities between animal and plant pathogens and symbionts. Proc Natl Acad Sci U S A 99 20: 13148–13153.
- 14. Margulies M, Egholm M, Altman WE, Attiya S, Bader JS, et al. (2005) Genome sequencing in microfabricated high-density picolitre reactors. Nature 437 7057: 376–380.
- 15. Nummelin H, Merckel MC, Leo JC, Lankinen H, Skurnik M, et al. (2004) The Yersinia adhesin YadA collagen-binding domain structure is a novel left-handed parallel beta-roll. Embo J 23 4: 701–711.
- 16. Sangari FJ, Aguero J, Garcia-Lobo JM (2000) The genes for erythritol catabolism are organized as an inducible operon in Brucella abortus. Microbiology 146 ( Pt 2): 487–495.
- 17. Sangari FJ, Garcia-Lobo JM, Aguero J (1994) The Brucella abortus vaccine strain B19 carries a deletion in the erythritol catabolic genes. FEMS Microbiol Lett 121 3: 337–342.
- 18. Sangari FJ, Grillo MJ, Jimenez De Bagues MP, Gonzalez-Carrero MI, Garcia-Lobo JM, et al. (1998) The defect in the metabolism of erythritol of the Brucella abortus B19 vaccine strain is unrelated with its attenuated virulence in mice. Vaccine 16 17: 1640–1645.
- 19. Smith H, Williams AE, Pearce JH, Keppie J, Harris-Smith PW, et al. (1962) Foetal erythritol: a cause of the localization of Brucella abortus in bovine contagious abortion. Nature 193: 47–49.
- 20. Garcia-Lobo JM, Sangari FJ (2004) Erithritol Metabolism and Virulence in Brucella. In: Lopez-Goni I, Moriyon I, editors. Brucella- Molecular and Cellular Biology. Norfolk, NR18 OJA, England: Horizon Bioscience.
- 21. Sperry JF, Robertson DC (1975) Erythritol catabolism by Brucella abortus. J Bacteriol 121 2: 619–630.
- 22. Bellaire BH, Elzer PH, Hagius S, Walker J, Baldwin CL, et al. (2003) Genetic organization and iron-responsive regulation of the Brucella abortus 2,3-dihydroxybenzoic acid biosynthesis operon, a cluster of genes required for wild-type virulence in pregnant cattle. Infect Immun 71 4: 1794–1803.
- 23. Yost CK, Rath AM, Noel TC, Hynes MF (2006) Characterization of genes involved in erythritol catabolism in Rhizobium leguminosarum bv. viciae. Microbiology 152 Pt 7: 2061–2074.
- 24. Gago G, Kurth D, Diacovich L, Tsai SC, Gramajo H (2006) Biochemical and structural characterization of an essential acyl coenzyme A carboxylase from Mycobacterium tuberculosis. J Bacteriol 188 2: 477–486.
- 25. Gande R, Gibson KJ, Brown AK, Krumbach K, Dover LG, et al. (2004) Acyl-CoA carboxylases (accD2 and accD3), together with a unique polyketide synthase (Cg-pks), are key to mycolic acid biosynthesis in Corynebacterianeae such as Corynebacterium glutamicum and Mycobacterium tuberculosis. J Biol Chem 279 43: 44847–44857.
- 26. Lin TW, Melgar MM, Kurth D, Swamidass SJ, Purdon J, et al. (2006) Structure-based inhibitor design of AccD5, an essential acyl-CoA carboxylase carboxyltransferase domain of Mycobacterium tuberculosis. Proc Natl Acad Sci U S A 103 9: 3072–3077.
- 27. Portevin D, de Sousa-D'Auria C, Montrozier H, Houssin C, Stella A, et al. (2005) The acyl-AMP ligase FadD32 and AccD4-containing acyl-CoA carboxylase are required for the synthesis of mycolic acids and essential for mycobacterial growth: identification of the carboxylation product and determination of the acyl-CoA carboxylase components. J Biol Chem 280 10: 8862–8874.
- 28. Molina-Henares AJ, Krell T, Eugenia Guazzaroni M, Segura A, Ramos JL (2006) Members of the IclR family of bacterial transcriptional regulators function as activators and/or repressors. FEMS Microbiol Rev 30 2: 157–186.
- 29. Zhang RG, Kim Y, Skarina T, Beasley S, Laskowski R, et al. (2002) Crystal structure of Thermotoga maritima 0065, a member of the IclR transcriptional factor family. J Biol Chem 277 21: 19183–19190.
- 30. Delrue RM, Lestrate P, Tibor A, Letesson JJ, De Bolle X (2004) Brucella pathogenesis, genes identified from random large-scale screens. FEMS Microbiol Lett 231 1: 1–12.
- 31. Seleem MN, Boyle SM, Sriranganathan N (2008) Brucella: A pathogen without classic virulence genes. Veterinary Microbiology. doi: 10.1016/j.vetmic.2007.11.023.
- 32. Hornback ML, Roop RM 2nd (2006) The Brucella abortus xthA-1 gene product participates in base excision repair and resistance to oxidative killing but is not required for wild-type virulence in the mouse model. J Bacteriol 188 4: 1295–1300.
- 33. Lavigne JP, Patey G, Sangari FJ, Bourg G, Ramuz M, et al. (2005) Identification of a new virulence factor, BvfA, in Brucella suis. Infect Immun 73 9: 5524–5529.
- 34. Gee JM, Kovach ME, Grippe VK, Hagius S, Walker JV, et al. (2004) Role of catalase in the virulence of Brucella melitensis in pregnant goats. Vet Microbiol 102 1–2: 111–115.
- 35. Arellano-Reynoso B, Lapaque N, Salcedo S, Briones G, Ciocchini AE, et al. (2005) Cyclic beta-1,2-glucan is a Brucella virulence factor required for intracellular survival. Nat Immunol 6 6: 618–625.
- 36. Kim S, Watarai M, Kondo Y, Erdenebaatar J, Makino S, et al. (2003) Isolation and characterization of mini-Tn5Km2 insertion mutants of Brucella abortus deficient in internalization and intracellular growth in HeLa cells. Infect Immun 71 6: 3020–3027.
- 37. Endley S, McMurray D, Ficht TA (2001) Interruption of the cydB locus in Brucella abortus attenuates intracellular survival and virulence in the mouse model of infection. J Bacteriol 183 8: 2454–2462.
- 38. Forestier C, Deleuil F, Lapaque N, Moreno E, Gorvel JP (2000) Brucella abortus lipopolysaccharide in murine peritoneal macrophages acts as a down-regulator of T cell activation. J Immunol 165 9: 5202–5210.
- 39. Gee JM, Valderas MW, Kovach ME, Grippe VK, Robertson GT, et al. (2005) The Brucella abortus Cu,Zn superoxide dismutase is required for optimal resistance to oxidative killing by murine macrophages and wild-type virulence in experimentally infected mice. Infect Immun 73 5: 2873–2880.
- 40. Celli J, Gorvel JP (2004) Organelle robbery: Brucella interactions with the endoplasmic reticulum. Curr Opin Microbiol 7 1: 93–97.
- 41. Bandara AB, Contreras A, Contreras-Rodriguez A, Martins AM, Dobrean V, et al. (2007) Brucella suis urease encoded by ure1 but not ure2 is necessary for intestinal infection of BALB/c mice. BMC Microbiol 7: 57.
- 42. Lestrate P, Delrue RM, Danese I, Didembourg C, Taminiau B, et al. (2000) Identification and characterization of in vivo attenuated mutants of Brucella melitensis. Mol Microbiol 38 3: 543–551.
- 43. Delcher AL, Harmon D, Kasif S, White O, Salzberg SL (1999) Improved microbial gene identification with GLIMMER. Nucleic Acids Res 27 23: 4636–4641.
- 44. Tech M, Pfeifer N, Morgenstern B, Meinicke P (2005) TICO: a tool for improving predictions of prokaryotic translation initiation sites. Bioinformatics 21 17: 3568–3569.
- 45. Computer ProgramLowe T (1997) tRNAScan-SE. St. Louis, MO: Washington University School of Medicine.
- 46. Wuyts J, Perriere G, Van De Peer Y (2004) The European ribosomal RNA database. Nucleic Acids Res 32 Database issue: D101–103.
- 47. Snyder EE, Kampanya N, Lu J, Nordberg EK, Karur HR, et al. (2007) PATRIC: The VBI PathoSystems Resource Integration Center. Nucleic Acids Res 35 Database issue: D401–406.
- 48. Pruitt KD, Tatusova T, Maglott DR (2005) NCBI Reference Sequence (RefSeq): a curated non-redundant sequence database of genomes, transcripts and proteins. Nucleic Acids Res 33 Database issue D501–504.
- 49. Yu GX, Snyder EE, Boyle SM, Crasta OR, Czar M, et al. (2007) A Versatile Computational Pipeline for Bacterial Genome Annotation Improvement and Comparative Analysis, with Brucella as a Use Case. Nucleic Acids Research. doi: 10.1093/nar/gkm377.