INTRODUCTION

Top Abstract Introduction Materials and Methods Results and Discussion References

Bovine brucellosis is a disease that affects livestock with a high incidence in South America. The etiological agent is the pathogenic bacterium Brucella abortus, a gram-negative nonmotile coccobacilus that produces abortions, sterility, and orchitis. It also affects humans, producing undulant fever, migraine, nausea, arthritic pain, and partial or total motion incapability (20). Although Brucella affects humans, it does not spread among them.

Classically the characterization and identification of virulence genes was carried out by the generation of random mutants searching for avirulent phenotypes (7). Although useful, this approach has led to the identification of a limited number of genes. Some modifications, such as targeted mutagenesis, have been developed in order to improve the isolation of avirulent mutants (9). One of the drawbacks for the identification of virulence genes is that in most cases these genes are induced within the host and are not expressed under normal laboratory culture conditions (18).

Cloning, characterization, and identification of B. abortus virulence genes is important for understanding the molecular pathogenesis of this intracellular microorganism. It might also help for the identification of antigens useful for the development of diagnosis tests and new vaccines.

Today it is possible to obtain the complete genomic sequence of a microorganism and, by comparison with known sequences deposited in protein or nucleotide sequence databases, to assign functions to genes and identify regulatory sequences. As of October 2000, the complete sequences of the genomes of 28 bacterial species have been released (see http: //www.ncbi.nlm.nih.gov/PMGifs/Genomes/micr.html), and many others are in different stages of completion. With multiple genomes sequenced, it is possible to delineate highly conserved protein families (17). Such information may be critically important, for example, to assign virulence functions to those protein families that are conserved among different bacteria.

In this work we have begun a large-scale sequencing of random B. abortus genomic fragments in order to obtain the sequence of a representative number of genes and facilitate the identification of genes involved in virulence and other host interaction functions. We describe 1,899 B. abortus genomic sequence surveys (GSSs) which represent an interesting start that may be used in combination with targeted mutagenesis in functional studies and identification of virulence genes.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results and Discussion References

Genomic libraries. DNA used for construction of the libraries was isolated by CsCl-ethidium bromide equilibrium centrifugation. Three different libraries from B. abortus strain 2308 were constructed, with two in the plasmid vector pBluescript SK() (Stratagene). The first library was made by partial digestion of total DNA with Sau3AI. Restriction fragments were size fractionated by agarose gel electrophoresis, and fragments between 0.8 and 2.2 kbp were recovered with Geneclean (Bio 101, Inc.) and cloned into the dephosphorylated BamHI site of the vector. The second library was constructed with random DNA fragments generated by using a nebulizer. After treatment with mung bean nuclease, phenol extraction, and ethanol precipitation, the DNA was blunt ended with T4 DNA polymerase and Klenow fragment. Fragments were size fractionated by agarose gel electrophoresis, and those in the range between 1.5 and 3 kbp were recovered and cloned into the dephosphorylated EcoRV site of the vector. The third library was constructed with DNA sheared by nebulization to an average size of 2 kb. The random fragments were cloned into a modified M13 vector using the double adaptor method (3).

Template preparation. Transformed bacteria were plated on LB agar containing ampicillin (100 µg/ml), X-Gal (5-bromo-4-chloro-3-indolyl--D-galactopyranoside (40 µg/ml), and IPTG (isopropyl--D-thiogalactopyranoside (100 µg/ml). White colonies were grown overnight at 37°C in 3 ml of 2× TY containing ampicillin (100 µg/ml) (15). Plasmid DNA templates for sequencing reaction were prepared from 1.5 ml of culture by an alkaline lysis method with minor modifications (15), followed by a polyethylene glycol 8000 precipitation. The amount of isolated DNA template was estimated on 1.0% agarose gel by comparison to serial dilutions of pBluescript II KS(+) (Stratagene). M13 phage DNA templates were prepared by using a glass fiber-filtration method (2).

DNA sequencing. Sequencing reactions at the San Martín University were performed in a Genius thermal cycler (Techne) using a Dye Terminator Cycle Sequencing Ready Reaction Kit with AmpliTaqr DNA polymerase (FS enzyme) (Applied Biosystems), following the protocols supplied by the manufacturer, and analyzed in an ABI prism 377 sequencer (Applied Biosystems). Single-pass sequencing was performed on each template using T7 primer. Sequencing reactions at the INTA Institute were carried out with primers 691 (GCCGCTCTAGACTAGTGGA) or 721 (GTCGACGGTATCGATAAGC) in a Perkin-Elmer 9600 thermocycler by using Dye Terminator Cycle Sequencing kits (Applied Biosystems). The fluorescent-labeled fragments were purified from the unincorporated terminators with Centri-Sep minicolumns (Princeton Separation, Adelphia, N.J.). The samples were resuspended in formamide and subjected to electrophoresis in an ABI 373 automatic sequencer. Sequencing reactions at the Department of Molecular Evolution, Uppsala University, Uppsala, Sweden, were carried out with the -21 M13 Forward primer on phage DNA templates and analyzed in an ABI prism 377 sequencer (Applied Biosystems).

Sequence analysis. Local homology searches were performed in a PC computer running Linux, using the BLAST suite of programs (1). BLAST searches against the national Center for Biotechnology Information (NCBI) nonredundant protein database were performed remotely using the Netblast client. The BLAST programs and the Netblast client are distributed by the NCBI (ftp://ncbi.nlm.nih.gov).

5

5

Nucleotide sequence accession numbers. Sequence data have been deposited in the dbGSS division of GenBank under the following accession numbers: AQ752928 to AQ752940, AZ048471 to AZ49844, and AZ302564 to AZ303170.

	RESULTS AND DISCUSSION

Top Abstract Introduction Materials and Methods Results and Discussion References

Three libraries were constructed as described in Materials and Methods, with inserts with an average size of 2 kbp. A total of 1,899 clones were successfully sequenced. After deleting vector sequences and unreliable data, an average length of 421 bases per clone was obtained and used for database searches. About 805,000 bp of genomic sequences were generated. In order to identify overlapping sequences, all sequences were subjected to contig assembly by using the Phred/Phrap/Consed system (courtesy of B. Ewing, P. Green, and D. Gordon, University of Washington, Seattle). This analysis generated 362 contigs and 978 isolated singlets (sequences having no nonvector match to any other read), representing 633,500 bp of unique genomic sequence. This represents ~20% of the estimated 3.2-Mb B. abortus genome.

Sequence similarities identified by the BLASTx program were considered statistically significant, with an E of <10⁵. Among the 1,899 GSSs obtained, 1,199 matched sequences deposited in the GenBank databases, and 700 either did not have a significant match or had no match at all (Table 1). To detect putative coding sequences within the latter group we used the testcode algorithm. About 37% of the sequences were found to be potentially coding; hence, they might represent novel or B. abortus-specific protein coding genes. About 3.9% of the GSSs matched Brucella sp. sequences, while 14.0 and 45.2% matched sequences from other members of the -2 subgroup of the division Proteobacteria and from other organisms, respectively.

Taking into account that the main aim of this project was gene discovery, we grouped 1,125 GSSs with significant homology to non-Brucella sequences according to matches to the same database entry. After deleting redundant GSSs, 925 out of 1,125 remained as nonredundant GSSs, thus representing 925 different genes that have not been previously described in the Brucella genus (see http://www.iib.unsam.edu.ar/genomelab/brucella/gss.html).

To get an insight into the functional diversity of our random sequences we compared our GSSs to the sequences present in the COG database developed by Tatusov et al. (17) and derived the functional classification that is associated with each COG in the COG database. Our results show that 470 nonredundant GSS could be related to 316 COGs and could be classified into 15 broad functional categories. Some GSSs were classified into more than one category and thus were included in the "mixed function" group, the remaining sequences were grouped under the "no related COG" category (see above-mentioned website). The distribution of putative genes with assigned COG category is shown in Fig 1. The largest number (18%) was related to the mixed-function group. Other categories include sequences related to general functions (15%); amino acid transport and metabolism (14%); translation, ribosomal structure, and biogenesis (11%); and energy production and conversion (8%).

View larger version (54K): [in this window] [in a new window]   FIG. 1.   Functional classification of B. abortus GSSs, showing the distribution of predicted genes according to their putative biological functions. A total of 470 nonredundant GSSs out of 925 with an E value of <105 were classified into 15 categories. GSSs that were classified into more than one category in the COG database were included in the mixed-function group (see website mentioned in Results and Discussion).

A detailed analysis of the putative genes identified is not within the scope of this work and will certainly be done by interested researchers in the field. It is, however, worthwhile mentioning the finding of an important number of interesting matches to sequences from other bacteria such as those of the genera Rickettsia, Agrobacterium, and Rhizobium, which, like Brucella, belong to the -2 subgroup of proteobacteria. In our survey, 220 out of 925 putative new genes identified showed similarities to bacterial sequences belonging to this group of bacteria (see above-mentioned website). This evidence reinforces the close phylogenetic relationship among these bacteria, a fact already pointed out by using 16S RNA sequence analysis (13).

Obtaining the complete chromosomal sequence of B. abortus will enable the identification of most of the potential virulence genes by comparison with other pathogens. Nevertheless, in the short run, a reasonable accomplishment might be the identification of a large proportion of its gene content by approaches like random genomic sequencing. In this study we have identified 925 genes, most of them representing novel genes for the Brucella spp. Among them, we have identified several B. abortus homologs of virulence and pathogenesis genes from other pathogens, including a homolog to both the Shda gene from Salmonella enterica serovar Typhimurium and the AidA-1 gene from Escherichia coli (GSSBru133). The AidA-1 gene encodes a cell adhesion molecule in enteropathogenic strains of E. coli which facilitates cell colonization. The same GSS is also homologous to the virulence gene virG of Shigella spp. which encodes a product involved in motility of the bacteria into the cell cytoplasm and spreading to the neighboring cells (11). No adhesion molecules have been described so far in Brucella. Thus, it will be interesting to obtain null mutants for the Brucella homolog of AidA-1 and study its possible role in cell adhesion and invasion. On the other hand, the Brucella homolog of virG might be involved in actin polymerization during phagocytosis.

Recently, it has been shown that some proteins implicated in the export of flagellar components are similar to components of the type III secretion machinery (8). Furthermore, it has been reported that the flagellar apparatus itself can secrete virulence factors (21). In our dataset we found several GSSs with high similarity to flagellar basal body genes like flip (GSSBru021), fliM (GSSBru838), and flgG (GSSBru1354) and to fliI (GSSBru1959), an ATPase associated to the flagellum biosynthesis (5). These findings suggest that, although Brucella is a nonmotile bacterium, it might have an active type III secretion machinery and/or a flagellar type secretion system not related to the biogenesis of the flagellum.

Other GSSs displayed significant homologies to genes encoding components of the type IV secretion machinery; among these, GSSBru0998 exhibited sequence similarity to traL of E. coli, and GSSBru162 and GSSBru1401 showed high similarity to virB9 and virB10 from Bordetella pertussis and Rhizobium etli, respectively. Tra genes are homologous to and colinear with genes found in the virB operon of Agrobacterium tumefaciens, which transfer the Ti plasmid from the bacteria to the nucleus of the plant cell (4), whereas in B. pertussis this system exports the pertussis toxin into the host cell (19). We have recently knocked out the gene homolog to virB10 in a virulent strain of B. abortus and showed that this gene is essential for intracellular survival and virulence (16). In addition, it has been recently shown that a region of the Brucella suis genome which is highly homologous to the A. tumefaciens virB operon is also required for intracellular multiplication (14).

Other GSSs displayed a significant homology with several Rhizobium meliloti genes (see above-mentioned website) related to nodule nitrogen fixation (GSSBru219 and -1400), synthesis of nod factors (GSSBru355), nodulation protein symbiotic plasmid (GSSBru223), and nodule bacteroid differentiation (GSSBru550). Among them, GSSBru550 was highly similar to the bacA gene, which encodes a putative cytoplasmic membrane transport protein that is essential for the symbiosis between Rhizobium meliloti and alfalfa (10). Recently, it has been shown that a mutation of the bacA homolog in B. abortus decreased intracellular survival (12). Thus, it will be interesting to know the functions of these proteins in Brucella. Also, it will be interesting to know the function of the R. meliloti homolog genes in Brucella, a mammalian intracellular pathogen.

Interestingly, we present here evidence showing a number of GSSs that are highly similar to both Agrobacterium virulence and Rhizobium nodulation genes. Why genes that seem to be very specific for the interaction of these soil bacteria with their hosts are conserved in Brucella is a question that remains to be answered but suggests that the ancestors of these organisms might have shared a common environment.

Particularly in bacteria, where gene density is high, it is increasingly apparent that random sequencing of genomic DNA is an efficient way to identify protein homologs. To date (October 2000), there are 333 protein sequences reported in the GenBank database for the Brucella genus; 198 of them correspond specifically to B. abortus. (htpp://www.ncbi.nlm.nih.gov/Taxonomy/). In the present study, random sequences totalling only 19.8% of the genome resulted in the identification of 925 putative new Brucella genes, which represents a large improvement over the known genes reported to date. These data, already available to researchers interested in the field, might provide new insights into the biology of Brucella.