当前位置: 首页 > 医学版 > 期刊论文 > 临床医学 > 微生物临床杂志 > 2006年 > 第6期 > 正文
编号:11259863
Identification and Characterization of Variable-Number Tandem-Repeat Markers for Typing of Brucella spp.
     Department of Statutory and Exotic Bacterial Diseases, Veterinary Laboratories Agency, Addlestone, Surrey KT15 3NB, United Kingdom

    ABSTRACT

    Members of the genus Brucella infect many domesticated and wild animals and cause serious zoonotic infection in humans. The availability of discriminatory molecular typing tools to inform and assist conventional epidemiological approaches would be invaluable in controlling these infections, but efforts have been hampered by the genetic homogeneity of the genus. We report here on a molecular subtyping system based on 21 variable-number tandem-repeat (VNTR) loci consisting of 13 previously unreported loci and 8 loci previously reported elsewhere. This approach was applied to a collection of 121 Brucella isolates obtained worldwide and representing all six classically recognized Brucella species. The size of repeats selected for inclusion varied from 5 to 40 bp giving VNTR loci with a range of diversities. The number of alleles detected ranged from 2 to 21, and Simpson's diversity index values ranged from 0.31 to 0.92. This assay divides the 121 isolates into 119 genotypes, and clustering analysis results in groups that, with minor exceptions, correspond to conventional species designations. Reflecting this, the use of six loci in isolation was shown to be sufficient to determine species designation. On the basis of the more variable loci, the assay could also discriminate isolates originating from restricted geographical sources, indicating its potential as an epidemiological tool. Stability studies carried out in vivo and in vitro showed that VNTR profiles were sufficiently stable such that recovered strains could readily be identified as the input strain. The method described here shows great potential for further development and application to both epidemiological tracing of Brucella transmissions and in determining relationships between isolates worldwide.

    INTRODUCTION

    Brucellosis is a zoonotic disease of major public health, animal welfare, and economic significance worldwide. In humans, infection with Brucella can lead to a chronic debilitating infection; in domesticated animals, the main symptom is reproductive failure. Disease in humans usually reflects occupational exposure or the consumption of unpasteurized dairy products. Brucellosis remains a major problem in many parts of the world, particularly Mediterranean regions, western Asia, and parts of Africa and Latin America (11), although in many developed countries it has been eradicated or severely curtailed by a combination of strict veterinary hygiene measures, monitoring programs, and improved food safety measures. Brucella species have also long been considered potential biological warfare agents, and in 1954 Brucella became the first biological agent to be treated as a weapon and field tested on animals under the old U.S. offensive biological weapons program. Recent history has raised awareness in this area (28), and the organism remains on the list of Centers for Disease Control and Prevention category B potential biological warfare agents (25).

    Classical Brucella taxonomists developed a classification system that recognized six species based on subtle phenotypic and antigenic differences and host specificity: B. abortus (bovine), B. melitensis (caprine and ovine), B. ovis (ovine), B. canis (canine), B. suis (porcine), and B. neotomae (desert wood rat). Some of these species are classically divided into biovars. The development of discriminatory molecular tools for identification and typing of Brucella has been problematic, reflecting the lack of genetic polymorphism in Brucella. A high degree of homology initially implied by DNA-DNA hybridization (30) has been confirmed by a variety of approaches, such as multilocus enzyme electrophoresis (MLEE) (13) and 16S rRNA sequencing. These sequences were long considered highly conserved, and a recent systematic study has confirmed 100% identity of the 16S rRNA sequences between all of the Brucella spp. (14). While the development of genus-specific and, in some cases, species-specific PCR assays for identification has been possible (3, 4), typing tools of sufficient resolution to permit epidemiological tracing of outbreaks or which might be required to try to identify the source of a bioterror attack are still lacking.

    In recent years the availability of microbial genome sequences has revolutionized DNA fingerprinting by facilitating the development of multilocus sequence-based typing approaches such as multilocus sequence typing (MLST) and multilocus VNTR analysis (MLVA). The genetic homogeneity of Brucella spp. implies that there will be insufficient sequence polymorphism in housekeeping genes to provide the resolution required to use MLST as a tool for epidemiological traceback. However, MLVA targets tandemly repeated DNA regions that are considered high-speed molecular clocks (27). The addition or deletion of repeat units reflecting either slipped strand nucleotide mispairing during replication or unequal crossover events results in a high rate of mutation at these loci. As a result, this approach has proven particularly useful as a tool for strain discrimination in bacterial species with little genomic variation, notably Bacillus anthracis (17), Yersinia pestis (17), Francisella tularensis (16), and various mycobacterial species such as Mycobacterium tuberculosis (22), Mycobacterium leprae (15), and Mycobacterium avium subsp. paratuberculosis (20). As well as the potential for high resolution, MLVA has a number of other technical advantages, most notably its relative simplicity, fast turnaround time, amenability to high-throughput approaches, applicability to nonviable and/or crude preparations, and the ability to easily compare digital results between laboratories.

    None of the existing molecular tools provide adequate resolution to confidently permit epidemiological traceback in the case of accidental import or deliberate release. However, the completion of genome sequences for a Brucella suis (21) and a Brucella melitensis (12) strain provided an opportunity to assess the presence of tandem repeats that might facilitate the development of an MLVA scheme. Initial analysis indicated the presence of many potentially useful regions of diversity in the Brucella genomes and, indeed, during the planning stages of the present study, an MLVA scheme that utilizes eight distinct copies of an octameric repeat (the "HOOF-Prints" assay) was described (5). Since our initial aim was to develop a high-resolution typing system, we have expanded on this existing scheme by assessing the performance of a number of additional highly polymorphic markers, in conjunction with those previously described, in a diverse range of isolates representing all currently recognized Brucella species. The identification of these additional markers increases the potential discriminatory capacity of the existing HOOF-Prints assay and reduces the risk of unrelated isolates having identical profiles due to homoplasy. Furthermore, during the course of this work, it became clear that, although these highly discriminatory markers should facilitate a degree of epidemiological traceback impossible in the past, the rapid mutation rate at these loci could make it difficult to assign isolates to one of the classically recognized species. We therefore assessed a number of additional tandem repeat loci with lower mutation rates for inclusion in a scheme that permits rapid speciation on the basis of apparently species-specific alleles or species-specific allelic profiles at these loci. We have thus examined VNTR markers with a range of diversity levels in 121 Brucella isolates and present an MLVA scheme that offers the potential of great discriminatory power and yet retains the capacity for isolate resolution at the taxonomic level.

    MATERIALS AND METHODS

    Isolates and template preparation. The details for all strains used in the present study are included in Table 1. The template for the VNTR PCR was either genomic DNA prepared as described previously (29) or a crude methanol extract prepared by harvesting growth into 66% (vol/vol) methanol. Where the biovar was determined in our laboratory, this was carried out according to standard procedures (1), although in some cases the biovar designations are those given by the strain providers.

    Identification of tandem repeats. Tandem repeats were identified by using the genome sequences of B. suis (21) and B. melitensis (12) and the Tandem Repeat Finder software (2). This program identified a large number of tandem repeats in addition to those described by Bricker et al. (5) and incorporated into the HOOF-Prints assay. A selection of these were assessed for their potential value in a typing scheme by designing PCR primers flanking the tandem repeats and assessing the size diversity of resulting PCR products by agarose gel electrophoresis across a panel of Brucella isolates. Thirteen novel repetitive loci were chosen for use in the assay described here to give, when used in conjunction with the existing eight HOOF-Prints loci, an assay examining variability at twenty-one loci.

    Sequence verification. In order to verify the basis of repeat copy number variation, PCR products representing at least three alternative alleles (or both alleles in the case of loci with only two allelic states) were sequenced for all 13 novel VNTR markers. PCR products were purified by passage through QiaQuick PCR purification columns (QIAGEN) and sequenced by using forward and reverse primers and an ABI BigDye terminator cycle sequencing kit (Applied Biosystems). Amplicon sizes were used in conjunction with sequence data to predict repeat copy number for each locus and isolate.

    VNTR analysis. The primer pairs used for the amplification of all loci described for the first time here are shown in Table 2. Primers used for amplification of the HOOF-Prints loci were as described previously (5), with the exception of the HOOF-Prints 7 primers that were redesigned since we found the originally reported primer set difficult to work with. PCR amplification was performed by using Fast Start Taq DNA polymerase (Roche). Each reaction mixture consisted of 2.5 μl of Fast Start 10x PCR buffer with MgCl2 (20 mM), 3.125 μl of 2 mM deoxynucleoside triphosphates, 0.05 μl of 100 μM forward and reverse primer, and 0.125 μl of Fast Start Taq polymerase in a final volume of 24.75 μl of water. As the PCR template, either 0.25 μl of purified genomic DNA or a crude methanol extract was used. Primer pairs consisted of one primer fluorescently labeled with NED, FAM, or HEX and one unlabeled primer. Initially, amplification was carried out one locus at a time but eventually amplification was multiplexed into sets of three loci using the three different dyes and reaction volumes were halved compared to the above. The PCR conditions were as follows: an initial step of 95°C for 5 min; followed by 95°C for 30 s, 55°C for 30 s, and 75°C for 1 min repeated for 30 cycles; followed by a 59-min incubation at 75°C. After PCR, the amplification products were routinely diluted 1:20 in water, and 1 μl was separated by capillary electrophoresis on an ABI377 genetic analyzer. All samples were run with GeneScan 500 (ROX) size markers as an internal standard, and the bands were sized relative to these markers by using the GeneScan software. Sizes of the HOOF-Prints loci PCR products were consistent with those described by Bricker et al. (5) (with the exception of the modified HOOF-Prints 7 primers). To ensure the comparability of datasets, we used the same numbering system and adjusted the HOOF-Prints 7 allele designations such that those presented here are equivalent to those previously described. Newly described loci were given allele designations generally based on repeat number. However, in some cases, as described in Results, alleles were found to reflect changes within regions other than the selected tandem repeat, and in these cases different alleles therefore do not necessarily reflect the numbers of tandem repeats. A full list of the sizes of PCR products obtained and the corresponding allele designations is provided in the supplemental material.

    Assessment of in vitro and in vivo stability. In order to assess the stability of VNTR profile on in vitro passage, cultures of B. abortus UK18-03-211, B. melitensis F8/01-155, and B. suis 01-5744 were maintained on serum dextrose agar slopes with the addition of 10% (vol/vol) horse serum for approximately 10 months. The cultures were transferred to fresh media at approximately three weekly intervals, at which point the population was sampled and subjected to VNTR analysis. In order to assess in vivo stability, samples obtained after experimental infection of pigs with B. suis 01-5744 were utilized. After intraconjunctival inoculation, Brucella was reisolated from the blood of six of the ten pigs initially infected by using the method of Casteneda (7). Blood culture was carried out twice weekly up to 56 days postinfection and weekly for a further 113 days. The number of occasions on which Brucella isolation was successful for an individual pig ranged from only a single bleed to up to nine bleeds. Crude methanol extracts of all Brucella strains reisolated were prepared to assess the VNTR profiles of the recovered isolates relative to the input strain.

    Computer analysis and tree construction. Genetic diversity was determined by using the Simpson's diversity index (DI) (26) via the online tool V-DICE available at the HPA website (http://www.hpa.org.uk/srmd/bioinformatics/tools/tools.htm). Values of this index can range from 0 (no diversity) to 1 (extreme diversity). Analysis of genetic relationships based on VNTR amplicon size profiles was performed by using the neighbor-joining tree algorithm in the PAUP4.0 beta version (Sinauer Associates, Inc., Sunderland, Mass.).

    Sequence accession numbers. Sequences of the VNTR loci newly described in the present study that were determined in order to verify repeat numbers have been deposited in EMBL under accession numbers AM260641 to AM260683.

    RESULTS

    Identification and characterization of loci. Since, prior to the advent of MLVA, there were essentially no molecular tools that could provide useful discrimination beyond the level of biovar, the first aim of this study was to characterize highly discriminatory loci that would facilitate the development of tools to permit epidemiological traceback of Brucella to the source of infection. During the planning stages of this work, the HOOF-Prints scheme, based on eight hypervariable octameric loci, was published by Bricker et al. (5). In order to further improve the discriminatory capacity of this scheme, to reduce the risk of match due to homoplasy, and to offer an increased choice of loci that might be applicable in a future standardized Brucella MLVA scheme, we identified further loci and compared their diversity with the HOOF-Prints loci. Tandemly repeated loci were identified by using Tandem Repeat Finder (2), and a number of loci were screened by PCR for size variation across a panel of Brucella isolates (data not shown).

    The location and characteristics of all novel loci taken forward for full assessment in the present study are described in Table 3. The initial stages of the work focused on loci that appeared variable within the classically identified Brucella species and were therefore most likely to add substantial discriminatory power to an MLVA scheme. Seven such loci were included in the present study and, as might be expected, all represented short sequence repeats of either 5 bp (VNTR 16) or 8 bp (VNTR 2, VNTR 5A, VNTR 5B, VNTR 12A, VNTR 12B, and VNTR 17). Sequencing of a number of alleles at each of these loci confirmed that the size variation in VNTR PCR products usually reflected exact changes in the number of repeats, and this information was used to predict the number of repeats present in remaining alleles. In the case of two of these repeats the situation was more complex. Sequencing of a representative of each of six different VNTR 17 alleles varying in size in 8-bp increments revealed that in addition to the tandem repeat described in Table 3, which varied from one to four copies, an additional octameric repeat varying from two to four copies is located approximately 30 bp downstream of this VNTR. Thus, in the case of VNTR 17, the combination of variation at both of these sites contributes to variation at this locus. In addition, sequencing of VNTR 16 alleles revealed that some strains harbor a 5-bp deletion immediately upstream of the pentameric repeats.

    After assessment of the shorter sequence repeats, six VNTR loci comprising of slightly larger repeats were added to the scheme. The sizes of these repeats predicted by Tandem Repeat Finder ranged from 13 bp (VNTR 21), 15 bp (VNTR27), 17 bp (VNTR 14), and 18 bp (VNTR 26 and VNTR 7) through to 40 bp (VNTR 24). Preliminary work suggested there was only very limited variation at these loci within a species, and thus it was predicted that variation at these loci between species might provide a useful taxonomic and phylogenetic framework when used in conjunction with the shorter and apparently more variable repeats. Again, representatives of a number of allelic states were sequenced. Only two alternative allelic states were detected for VNTRs 14, 21, and 27 and these corresponded to those predicted from the B. suis and B. melitensis genome sequences. For VNTR 26 the designation of alleles was simple since, despite being a rather degenerate repeat, each allele was found to reflect an exact 18-bp increment in repeat number. The situation for the two remaining loci is more complex since allele sizes were not exact multiples of the sizes predicted in Table 3, and sequencing confirmed complex patterns of variation at these two loci that do not reflect exact changes in repeats. In the case of VNTR 7, whereas most alleles did reflect 18-bp increments, one allele reflected the presence of a 3-bp deletion in one copy of the repeat. VNTR 24 was found to represent a particularly complex structure with a series of alleles that did not reflect exact changes in the 40-bp repeat originally identified. Representative sequences of all of the loci with novel variants described here have been deposited in EMBL under accession numbers given in Materials and Methods.

    Thus, in total, 13 novel loci were examined here in conjunction with the existing 8 HOOF-Prints loci to give a 21-locus MLVA scheme. With the exception of two pairs of loci (VNTR 5A and 5B, and 12A and 12B), the novel loci are widely distributed across the Brucella genome (seven on chromosome I and six on chromosome II). The two VNTR 12 loci represent octameric repeats that are separated by less than 40 bp but are highly diverse and appear to vary independently. Similarly, the two VNTR 5 loci are separated by approximately 60 bp but appear to vary at relatively high frequency independently.

    Comparative genetic diversity. The 21-locus MLVA scheme outlined above was applied to 121 Brucella isolates consisting of 44 B. abortus, 33 B. melitensis, 10 B. ovis, 25 B. suis, 3 B. neotomae, and 6 B. canis isolates (Table 1 shows the strains examined [the equivalent raw data are provided in the supplemental material]). These isolates include all species and biovar reference strains and a collection of isolates that represent a diverse range of biovars and geographical origins. In addition, a small number of isolates from more restricted geographical origins were included to assess diversity within such populations. The discriminatory power of the technique is readily demonstrated by the fact that, of the 121 genotypic profiles determined, 117 were unique. Only two pairs of strains (B. neotomae 65/196 and 65/197 and B. melitensis UK21/04 and UK26/04) gave identical profiles. With one exception, all loci could be amplified from all strains - we could not obtain a reliable amplification product from B. ovis isolates using the HOOF-Prints 5 primers and thus treated this as a null allele for all B. ovis isolates.

    The number of alleles present at each locus within the population examined here ranged from 2 in the case of VNTR 14, VNTR 21, and VNTR 27 up to 21 in the case of HOOF-Prints 7 (Fig. 1). Note that although HOOF-Prints 4 appears to have the highest number of alleles (27); this is somewhat deceptive since a deletion within the PCR product in Brucella species other than B. abortus and B. melitensis means that each repeat number has two possible allelic states (shown, for example, as 1 or 1A in the supplemental material) at this locus. In order to assess the diversity at individual loci, the DI was calculated for both the whole population and for all isolates of each Brucella species individually. Values of this index range from 0 in the case of no diversity up to 1 in the case of extreme diversity and reflect the number of alleles detected and the individual allele frequency. In general, there was a close relationship between the total number of alleles and DI with values ranging from 0.31 in the case of VNTR 14 to >0.90 in the case of HOOF-Prints 1, 4, and 7 and VNTR 12B (Fig. 1). The overall diversity of the novel short sequence repeat (5 or 8 bp) VNTR loci selected for use in the present study (range, 0.71 to 0.91; mean, 0.81) was similar to that for the HOOF-Prints loci considering the same population (range, 0.74 to 0.92; mean, 0.85), indicating similar resolving power. With the exception of VNTR 24, the six larger repeats had much lower numbers of alleles (two to four) and DI values (0.31 to 0.54). VNTR 24, although originally defined as a 40-bp repeat, had a number of complex alleles that did not reflect exact changes in repeat number and thus had a larger number of alleles (6) and a DI (0.75) within the range of the VNTR loci based on shorter sequence repeats (5 to 8 bp).

    The DI values determined within each of the individual species (Table 4) demonstrate that most of the VNTR consisting of short (5- to 8-bp) repeats are also variable within species. There are some exceptions, notably in B. ovis. This species appears to harbor less genetic diversity than the other Brucella spp., with no variation detected in 7 of 15 such VNTR loci. In contrast, all 15 of these loci were found to be variable within B. abortus and B. suis, only HOOF-Prints 3 is invariant in B. melitensis, and only two loci (HOOF-Prints 2 and VNTR 17) were found to be invariant in the small number of B. canis isolates examined. In contrast, there was limited intraspecies diversity when we considered the six VNTRs representing longer sequence repeats in this population. VNTR 21 does not vary within a species with one allele confined to B. melitensis, while a second allele is found in all other species. The only intraspecies variation for VNTR 14 is seen in the single B. suis biovar 5 isolate that shares an allele with representatives of all other species in contrast to the unique allele seen in all other B. suis isolates. VNTR 7 showed intraspecies variation in only three B. abortus isolates. VNTR 27, VNTR 24, and VNTR 26 all reveal diversity in isolates of B. abortus, B. melitensis, and B. suis, although the DI values indicate that this is less extensive than that seen using the 15 VNTR loci containing 5- to 8-bp sequence repeats.

    Genetic relationships. The genetic relationships between 105 of the 121 isolates were determined by the construction of a neighbor-joining tree (Fig. 2). Sixteen of the original isolates included in Table 1 were found to possess multiple allelic states at one or more loci and were therefore excluded from this analysis. A number of major clusters are immediately apparent that correspond to the classically recognized Brucella species. Thus, there are clusters consisting exclusively of B. abortus, B. ovis, and B. neotomae isolates. Similarly, there is a large cluster that, with the exception of a single isolate, the B. abortus biovar 3 reference strain (Tulya), is comprised entirely of B. melitensis isolates. With the exception of the B. suis biovar 5 reference strain that appears to be distantly related to all other isolates, B. suis isolates group together in a cluster that also contains B. canis. B. suis further subdivides into two clusters: one consisting exclusively of biovar 2 isolates and the other containing biovar 1, 3, and 4 isolates.

    Stability of VNTR loci. In order to assess the stability of VNTR loci over time, changes to input strains were examined after both in vivo and in vitro passage. For the in vivo study pigs were experimentally infected with B. suis 01-5744 and frequent blood culture was carried out postinfection in an attempt to reisolate Brucella. The VNTR profiles of both the inoculum and any strains recovered postinfection were determined (Table 5). Of six animals that became culture positive, isolates recovered from two of the animals at up to 63 days postinfection were identical to the input profile. Isolates from the four remaining animals showed minor changes in the VNTR profile. These changes were all either one-step increases or decreases in repeat number, often with a mixed profile such that the input allele was still visible, occurring at loci with high DI values. Thus, such changes were apparent at VNTR 12B, HOOF-Prints 5, and HOOF-Prints 7, all of which have DI values of 0.9, and a single change was noted at HOOF-Prints 8 that, while having slightly lower diversity in the overall population, has the equal highest DI value recorded for any population subset (0.93) for B. suis isolates. In order to look at the effect of in vitro passage three strains were passaged 14 times over 270 days (Table 6). The VNTR profiles of B. suis and B. melitensis isolates remained unchanged, while the profile of the B. abortus isolate showed only a one-step increase in the VNTR 12B repeat number that became apparent at passage 13 after 251 days of cultivation.

    DISCUSSION

    The aim of this study was to identify and characterize VNTR markers that, when used in combination with loci previously described, provide the robust high-resolution typing tool that Brucella epidemiologists have lacked to date to facilitate epidemiological traceback. Furthermore, by incorporation of more stable markers, it was deemed possible to develop a tool that can simultaneously address issues of Brucella taxonomy and phylogeny. We report on 13 repetitive loci not previously described in Brucella spp. that further progress toward both of these goals. The novel loci are distributed throughout the genome, with seven located on chromosome I and six located on chromosome II. Two sets of repeats (5A/B and 12A/B) were located very close to each other (ca. 40 to 60 bp apart) and were originally considered as single loci. However, eventually primers were designed such that these repeat tracts are considered separately as, for a number of reasons, this approach enhances discriminatory capacity. It enables maximum information to be gathered from the extensive diversity seen at all four of these loci rather than just considering these sites as two loci. It reduces the risk of convergence to the same allele that could result from treating both sets of repeats as a single locus. Furthermore, should there be multiple alleles at one of the repetitive tracts, only data from one locus is lost and not from both as is the case if the tracts are considered together as a single locus. As with the original HOOF-Prints loci, the 13 novel loci are located variously either intragenically, intergenically, or overlapping the predicted stop codons of putative proteins (Table 3).

    The use of VNTRs to examine phylogeny and taxonomy in bacteria is open to criticism due to concerns about rapid evolution and potential homoplasy that could result in misleading conclusions. However, in homogeneous organisms such as Brucella, where limited diversity is detectable by other approaches, the use of VNTRs offers a viable alternative to more traditional approaches. With minor exceptions, analysis based on the 21-locus MLVA scheme divided the organisms in the present study into clusters corresponding to their taxonomic designations (Fig. 2). Application of the identical analysis to the same population with only the original 8 HOOF-Prints loci fails to group all isolates of the traditionally recognized species together, while use of all 15 short-sequence repeat-containing loci (5 to 8 bp), although an improvement on the latter, also fails to completely resolve all of the species groups (data not shown). One of the minor exceptions mentioned above is the B. abortus biovar 3 reference strain (Tulya) that falls within the B. melitensis cluster. Previous studies have noted that strain Tulya is genetically distinct from field biovar 3 strains (18), and other B. abortus biovar 3 strains examined in the present study did cluster with B. abortus. Thus, the status of this reference strain and its relationship to other B. abortus biovar 3 strains requires further investigation. Except for this one strain, all isolates of B. melitensis, B. abortus, B. ovis, and B. neotomae fall into distinct clusters that reflect their traditional taxonomic status. B. canis falls on a separate branch within the B. suis cluster that is most closely related to B. suis biovar 1, 3, and 4 isolates. B. suis biovar 2 isolates form a separate branch, while biovar 5 isolates are only distantly related to other B. suis isolates. The close association of B. canis with B. suis has been noted in a number of studies using methods including amplified fragment length polymorphism (AFLP) (29), chromosomal maps (23), omp2 profiling (8, 9), MLEE (13), and insertion sequence typing (19). Similarly, studies using AFLP and MLEE have indicated that B. suis biovar 5 is distinct from other B. suis isolates (13, 29). Although the species generally separated well into distinct genetic entities, there was less evidence of many of the biovars corresponding to distinct genetic groups. Thus, in the case of B. suis, although biovar 2 and 5 isolates separate well as described above, isolates of biovars 1, 3, and 4 were not clearly separated. Similarly, there was no evidence of separation of the isolates of B. melitensis biovars 1, 2, and 3, and there were too few representatives of most of the B. abortus biovars to reach any conclusions. Genetic relationships within biovars as determined by this approach will be addressed elsewhere using much larger numbers of suitably selected isolates.

    Data gathered in the present study indicate that the use of the six novel loci based on slightly longer sequence repeats (VNTRs 7, 14, 21, 24, 26, and 27) in isolation provide an assay capable of speciation of all Brucella isolates. The potential outcomes of such an assay are illustrated in Table 7. On the basis of alleles at these loci, the 121 isolates examined here divide into 17 distinct genotypes, all of which are associated exclusively with only one of the traditionally recognized Brucella species (6 B. abortus, 4 B. melitensis, 1 B. ovis, 4 B. suis, 1 B. neotomae, and 1 B. canis). Thus, although it must first be validated by using a much larger collection of isolates, this assay would represent by far the most straightforward method currently available to speciate Brucella isolates. Existing assays that can speciate Brucella fail to distinguish all species or biovars of species (e.g., AMOS PCR [4]) and/or involve complex technical procedures and genomic DNA preparation (e.g., IS711 fingerprinting or AFLP [19, 29]). As discussed below, the use of MLVA has considerable technical advantages and could easily be run with an agarose gel-based approach, overcoming the need for expensive apparatus and specialized reagents.

    As a tool for epidemiological traceback, MLVA clearly offers a remarkable discrimination compared to any of the conventional molecular typing techniques previously applied to this organism that generally fail to provide much discrimination below the biovar level. In contrast, the 121 strains examined in the present study represent 119 distinct genotypes. Every B. abortus biovar 1 isolate examined in the present study represented a different genotype. A further indication of the power of the technique is that all isolates of B. ovis, a species long considered to harbor little diversity (24), could be separated. However, the lack of variation in 13 of 21 loci does support the suggestion that, among the Brucella species, B. ovis is particularly highly conserved. Although 13 new loci are described here, it is the 7 based on short sequence repeats of 5 to 8 bp that are most useful in developing a genotyping tool that can be used to complement conventional epidemiological investigations. Bricker et al. (6) have recently highlighted a number of inconsistencies between the B. abortus biovar and the HOOF-Prints genotype. One possible explanation for this is homoplasy resulting from convergent evolution to an identical genotype among "unrelated" strains, and Bricker et al. (6) suggested that additional polymorphic loci should be included in the scheme in order to reduce the likelihood of this. We report here on seven such polymorphic loci that, as demonstrated in Fig. 1, display levels of diversity within the same range as the original HOOF-Prints loci. Although we will describe in detail the use of this scheme in epidemiological traceback in future studies, the incorporation of small numbers of isolates from restricted geographical locations does highlight the potential of this tool in assisting conventional epidemiological approaches. Thus, there are small groups of isolates of B. abortus (isolate tree references in Fig. 2 of 28 to 36) B. melitensis biovar 1 (42 to 44), B. melitensis biovar 3 (56 to 59), and B. suis biovar 2 (97 to 99) from Portugal included in the present study. These four groups all cluster together in Fig. 2, with the exception of one of the B. melitensis biovar 3 isolates (i.e., isolate 56). However, isolates within each of these clusters still showed diversity at some of the most variable loci. Thus, the loci with the highest DI values (HOOF-Prints 1, 4, 5, and 7 and VNTR 12B) were variable within all four of these groups. A substantial number of the B. abortus biovar 1 isolates included in this study originated from Eire or Northern Ireland, and all of these isolates also cluster together, along with isolates from the United Kingdom that are known or suspected to reflect importation from these locations (isolate tree references 5, 7 to 9, 11, 14, 15, and 17 to 20). This cluster is separate from all other isolates, such as the Portuguese B. abortus biovar 1 cluster, but again all of the isolates within the cluster are distinguishable from each other, such that a broad geographical origin of an isolate might be predicted from the MLVA profile, but the technique is sufficiently powerful to further differentiate isolates within such a restricted locality. Further examples of the potential of MLVA to assist with epidemiological traceback include an isolate from a case of human brucellosis in the United Kingdom (isolate tree reference 60). In this case the isolate profile most closely matches isolates of B. melitensis from Portugal (see Fig. 2, isolate tree references 57 to 59) and, indeed, upon investigation the patient reported a history of recent travel to Portugal. There is also a well-separated cluster of B. melitensis isolates from human brucellosis in the United Kingdom (isolate tree references 66, 67, and 69 to 71). Where patient histories were available all of these isolates were associated with travel to Eritrea or Somalia. These isolates are separated from other B. melitensis isolates even on the basis of variation in the six loci that facilitate speciation (Table 7). On this basis they share a profile with only a single isolate from livestock in Tanzania R39/03-60 (this isolate was excluded from Fig. 2 due to the presence of a multiple allele at one locus). All of these examples highlight the potential power of international database of Brucella strain genotypes against which new isolates could be compared.

    The technical advantages of MLVA versus existing Brucella typing schemes are overwhelming. In addition to the huge improvement in discriminatory power, the technique is technically undemanding, and the numerical format means that data are unambiguous and readily comparable between laboratories. All Brucella isolates appear to be typeable by this approach, including the unclassified marine mammal isolates (data not shown). Only 1 of the 21 loci examined could not be amplified from all strains, and this problem was confined to B. ovis (HOOF-Prints 5). One problem was the occasional identification of multiple alleles among the most genetically variable loci. This has been reported previously (6), where the authors chose to report the major peak as the allele designation. Since in our experience the "secondary" peaks were often of identical or similar intensity, we took the approach here of reporting multiple alleles where a second peak had an intensity of >1/3 of that of the major peak. Bricker et al. (6) reported that this was a particular problem with reference isolates; we also found a particular problem with reference isolates and vaccine strains that have been extensively subcultured. Less frequently passaged recent field isolates seem less problematic. Provided there are sufficient discriminatory loci included in a scheme, the loss of data from a single locus due to the presence of multiple alleles should not be problematic. Alternatively, both possible genotypes could be considered in analysis. Further advantages of an MLVA approach include rapid turnaround time, ease of automation, and the ability to use crude material or nonviable samples; these characteristics offer huge advantages for a hazardous organism such as Brucella. As a PCR-based method, it is also theoretically possible to type directly from infected tissue or other materials without prior culture.

    A successful typing method needs to clearly differentiate unrelated isolates but to demonstrate the relationships of all organisms isolated from individuals infected through the same source. In order to formally confirm the stability of isolates obtained from a single source of infection, we examined the stability of isolates after both in vivo and in vitro passages. In both cases minor changes were seen, but they always represented one-step changes in no more than two of the most variable loci such that the strain was clearly recognizable as the input strain. After in vitro cultivation over 14 passages and some 270 days, only a single-step change in a single strain was seen toward the end of this time course. Surprisingly, this change appeared to be absolute with no intermediate mixed genotype detected. We assume that the variant was better adapted to growth on artificial media than the parent strain and rapidly outgrew the original parent strain such that this was no longer detected. However, the length and number of serial passages undertaken far exceeds the in vitro cultivation that clinical isolates would undergo prior to routine MLVA testing and indicates that in vitro cultivation does not lead to significant changes in MLVA profile. After in vivo cultivation, isolates from four of six infected animals showed some minor changes that were always single-step changes and occurred at no more than two loci. Table 5 shows that in one animal, animal 12, one-step changes apparent at one time were not always maintained. We suspect that this reflects the fact that the Brucella culture method includes an enrichment step and thus the VNTR profiles, though population based, may actually be derived from the enriched clonal descendants of a very small number of bacteria. Thus, the most obvious explanation for the results from animal 12 is that by chance on some occasions the profile resembled the input strain and by chance on other occasions we isolated organisms originating from a clone(s) that had one-step mutations at a rapidly evolving locus either alone or in conjunction with clones representing the initial input strain. Again, all output strains would be recognized as identical or very closely related to input strains particularly if ordered characters were considered in the analysis rather than considering each different allele at a locus as equally unrelated.

    The increasing movements of humans and livestock and the ongoing threat of bioterrorism highlights the needs for international epidemiological surveillance tools able to monitor pathogens such as Brucella at global levels and to track transmission routes. In order to facilitate the development of such an approach based on VNTR markers, we have characterized 13 previously unreported VNTR markers that, together with the 8 previously reported markers (5), increase the potential pool of loci that could be used in such a scheme. The loci selected clearly have different rates of evolution, giving flexibility to use all markers or use selected markers matching the nature of the population being examined. Thus, while some of the VNTRs (i.e., those based on longer sequence repeats) are not likely to be useful in establishing transmission routes, they are better informers of evolutionary scenarios and useful for establishing phylogeny across a diverse isolate collection. However, for tracking in areas of endemicity there will be little variation among these markers of relatively low diversity and, consequently, the use of more diverse markers will be more informative. The use of all 21 of the loci described here results in a scheme with potentially huge discriminatory capacity that can resolve isolates at a local level but also includes more stable markers, likely to be intrinsically less prone to homoplasy, that give a resolution at the taxonomic level that is difficult to achieve using highly variable loci alone. The most pressing requirement now is for the development of an international database of Brucella strain profiles against which new isolates can be compared. In order to make the most of the power of MLVA, it is crucial that a common set of loci are agreed upon at this early stage in the development of the technique for Brucella and that all, or a subset, of these are used internationally such that all data are comparable between laboratories. We hope that the data presented in the manuscript on both novel and previously identified loci will facilitate the rational selection of the most appropriate loci for such a scheme.

    ACKNOWLEDGMENTS

    We gratefully acknowledge the assistance of Betsy Bricker, U.S. Department of Agriculture, Ames, Iowa, in setting up the HOOF-Prints assay. We also gratefully acknowledge the role of coworkers in the Brucella field who isolated and submitted Brucella strains to the VLA over many years, providing an invaluable resource for the optimization of this technique.

    All research for this study was funded by the United Kingdom Department of Environment, Food, and Rural Affairs (DEFRA).

    Supplemental material for this article may be found at http://jcm.asm.org/.

    REFERENCES

    Alton, G. G., L. M. Jones, R. D. Angus, and J. M. Verger. 1988. Techniques for the brucellosis laboratory. INRA, Paris, France.

    Benson, G. 1999. Tandem repeats finder: a program to analyze DNA sequences. Nucleic Acids Res. 27:573-580.

    Bricker, B. J. 2002. PCR as a diagnostic tool for brucellosis. Vet. Microbiol. 90:435-446.

    Bricker, B. J., and S. M. Halling. 1994. Differentiation of Brucella abortus bv. 1, 2 and 4, Brucella melitensis, Brucella ovis, and Brucella suis bv. 1 by PCR. J. Clin. Microbiol. 32:2660-2666.

    Bricker, B. J., D. R. Ewalt, and S. M. Halling. 2003. Brucella ‘HOOF-Prints’: strain typing by multi-locus analysis of variable number tandem repeats (VNTRs). BMC Microbiol. 3:15.

    Bricker, B. J., and D. R. Ewalt. 2005. Evaluation of the HOOF-Print assay for typing Brucella abortus strains isolated from cattle in the United States: results with four performance criteria. BMC Microbiol. 5:37.

    Casteneda, M. R. 1947. A practical method for routine blood cultures in brucellosis. Proc. Soc. Exp. Biol. Med. N. Y. 64:114-115.

    Cloeckaert, A., J. M. Verger, M. Garyon, and O. Grepinet. 1995. Restriction site polymorphism of the genes encoding the major 25 and 36-kDa outer-membrane proteins of Brucella. Microbiol. 141:2111-2121.

    Cloeckaert, A., J. M. Verger, M. Garyon, and N. Vizcaíno. 1996. Molecular and immunological characterisation of the major outer membrane proteins of Brucella. FEMS Microbiol. Lett. 145:1-8.

    Corbel, M. J. 1988. International Committee on Systematic Bacteriology Subcommittee on the Taxonomy of Brucella. Report of the meeting, 5 September 1986, Manchester, England. Int. J. Syst. Bacteriol. 38:450-452.

    Corbel, M. J. 1997. Brucellosis: an overview. Emerg. Infect. Dis. 3:213-221.

    DelVecchio, V. G., V. Kapatral, R. J. Redkar, G. Patra, C. Mujer, T. Los, N. Ivanova, I. Anderson, A. Bhattacharyya, A. Lykidis, G. Reznik, L. Jablonski, N. Larsen, M. D'Souza, A. Bernal, M. Mazur, E. Goltsman, E. Selkov, P. H. Elzer, S. Hagius, D. O'Callaghan, J. J. Letesson, R. Haselkorn, N. Kyrpides, and R. Overbeek. 2002. The genome sequence of the facultative intracellular pathogen Brucella melitensis. Proc. Natl. Acad. Sci. USA 99:443-448.

    Gandara, B., A. L. Merino, M. A. Rogel, and E. Martinez-Romero. 2001. Limited genetic diversity of Brucella spp. J. Clin. Microbiol. 39:235-240.

    Gee, J. E., B. K. De, P. N. Levett, A. M. Whitney, R. T. Novak, and T. Popovic. 2004. Use of 16S rRNA gene sequencing for rapid confirmatory identification of Brucella isolates. J. Clin. Microbiol. 42:3649-3654.

    Groathouse, N. A., B. Rivoire, H. Kim, H. Lee, S. N. Cho, P. J. Brennan, and V. D. Vissa. 2004. Multiple polymorphic loci for molecular typing of strains of Mycobacterium leprae. J. Clin. Microbiol. 42:1666-1672.

    Johansson, A., J. Farlow, P. Larsson, M. Dukerich, E. Chambers, M. Bystrom, J. Fox, M. Chu, M. Forsman, A. Sjostedt, and P. Keim. 2004. Worldwide genetic relationships among Francisella tularensis isolates determined by multiple-locus variable-number tandem repeat analysis. J. Bacteriol. 186:5808-5818.

    Le Fleche, P., Y. Hauck, L. Onteniente, A. Prieur, F. Denoeud, V. Ramisse, P. Sylvestre, G. Benson, F. Ramisse, and G. Vergnaud. 2001. A tandem repeats database for bacterial genomes: application to the genotyping of Yersinia pestis and Bacillus anthracis. BMC Microbiol. 1:2.

    Ocampo-Sosa, A. A., J. Aguero-Balbin, and J. M. Garcia-Lobo. 2005. Development of a new PCR assay to identify Brucella abortus biovars 5, 6, and 9 and the new subgroup 3b of biovar 3. Vet. Microbiol. 110:41-51.

    Ouahrani, S., S. Michaux, J. Sri Widada, G. Bourg, R. Tournebize, M. Ramuz, and J. P. Liautard. 1993. Identification and sequence analysis of IS6501, an insertion sequence in Brucella spp.: relationship between genomic structure and the number of IS6501 copies. J. Gen. Microbiol. 139:3265-3273.

    Overduin, P., L. Schouls, P. Roholl, A. van der Zanden, N. Mahmmod, A. Herrewegh, and D. van Soolingen. 2004. Use of multilocus variable-number tandem-repeat analysis for typing Mycobacterium avium subsp. paratuberculosis. J. Clin. Microbiol. 42:5022-5028.

    Paulsen, I. T., R. Seshadri, K. E. Nelson, J. A. Eisen, J. F. Heidelberg, T. D. Read, R. J. Dodson, L. Umayam, L. M. Brinkac, M. J. Beanan, S. C. Daugherty, R. T. Deboy, A. S. Durkin, J. F. Kolonay, R. Madupu, W. C. Nelson, B. Ayodeji, M. Kraul, J. Shetty, J. Malek, S. E. Van Aken, S. Riedmuller, H. Tettelin, S. R. Gill, O. White, S. L. Salzberg, D. L. Hoover, L. E. Lindler, S. M. Halling, S. M. Boyle, and C. M. Fraser. 2002. The Brucella suis genome reveals fundamental similarities between animal and plant pathogens and symbionts. Proc. Natl. Acad. Sci. USA 99:13148-13153.

    Mazars, E., S. Lesjean, A. L. Banuls, M. Gilbert, V. Vincent, B. Gicquel, M. Tibayrenc, C. Locht, and P. Supply. 2001. High-resolution minisatellite-based typing as a portable approach to global analysis of Mycobacterium tuberculosis molecular epidemiology. Proc. Natl. Acad. Sci. USA 98:1901-1906.

    Michaux-Charachon, S., G. Bourg, E. Jumas-Bilak, P. Guigue-Talet, A. Allardet-Servent, D. O'Callaghan, and M. Ramuz. 1997. Genome structure and phylogeny in the genus Brucella. J. Bacteriol. 179:3244-3249.

    Ridler, A. L., M. J. Leyland, S. G. Fenwick, and D. M. West. 2005. Demonstration of polymorphism among Brucella ovis field isolates by pulsed-field gel electrophoresis. Vet. Microbiol. 108:69-74.

    Rotz, L. D., A. S. Khan, S. R. Lillibridge, S. M. Ostroff, and J. M. Hughes. 2002. Public health assessment of potential biological terrorism agents. Emerg. Infect. Dis. 8:225-230.

    Simpson, E. H. 1949. Measurement of diversity. Nature 163:688.

    van Belkum, A. 1999. The role of short sequence repeats in epidemiologic typing. Curr. Opin. Microbiol. 2:306-311.

    Whatmore, A. M., T. J. Murphy, S. J. Cutler, and A. P. MacMillan. 2004. An assessment of the potential of amplified-fragment length polymorphism (AFLP) for use in identification and typing of Brucella isolates, p. 95-99. In P. H. Elzer and K. T. Metodiev (ed.), Risk infections and possibilities for biomedical terrorism. NATO Science Series, IOS Press, Amsterdam, The Netherlands.

    Whatmore, A. M., T. J. Murphy, S. Shankster, E. Young, S. J. Cutler, and A. P. MacMillan. 2005. Use of amplified fragment length polymorphism to identify and type Brucella isolates of medical and veterinary interest. J. Clin. Microbiol. 43:761-769.

    Verger, J. M., F. Grimont, P. A. D. Grimont, and M. Grayon. 1985. Brucella, a monospecific genus as shown by deoxyribonucleic acid hybridization. Int. J. Syst. Bacteriol. 35:292-295.(Adrian M. Whatmore, Steph)