This study has demonstrated marked geographic structuring of B. pseudomallei genotypes in soil. The dramatic differences in genotype frequency over such small distances are striking, but difficult to interpret. One explanation is that the numerically dominant ST at each sampling point represents a strain with superior biological fitness compared with STs present as a minority of the population. This could relate to factors such as soil type or pH, or competition with other microbial species. This would assume that adjacent foci of soil have variable microenvironments, but it seems unlikely that nearby sampling points within a confined area of disused land would differ sufficiently to support multiple, non-overlapping niches. An alternative possibility is that of local competition between clones of B. pseudomallei. Flooding or other disturbance mechanisms would provide the means for a given clone to migrate and become established within a specific plot. Once the clone has reached a certain threshold frequency, it could repel invaders either by the production of microbicides, through phage to which they themselves are resistant, or via other killing mechanisms. The presence of a clone as a minority population could represent the ability of this strain to survive at a lower level, or could represent the boundary of a point of predominance in an adjacent point or focus.
This study also provides evidence for microevolution of B. pseudomallei
in soil. PFGE is a more sensitive marker of very rapid genetic change than MLST in this species. Alterations in banding pattern arise due to any kind of genetic event that alters the presence or absence of restriction sites anywhere in the genome, or else changes the distance between existing sites. In contrast, MLST genes are chosen specifically to code for a central housekeeping role and to be highly conserved. MLST is therefore blind to large-scale genomic rearrangements that may dramatically alter the PFGE banding pattern 
. Two of the nine STs contained strains with variable PFGE banding pattern types. We postulate that these changes represent microevolution within our sampling site rather than importation of several strains with matching ST but a different banding pattern. This is consistent with the finding that genomic islands constitute ~6% of the B. pseudomallei
K96243 genome 
. Furthermore, comparison of the whole genome sequences of B. pseudomallei
and B. mallei
indicated the capacity for genomic rearrangement and gene loss by two species that are highly related by MLST 
. Our findings are consistent with a dynamic genome that is evolving through the movement of genomic islands and rearrangements such as inversions and indels.
The co-existence in a single soil sample of a single locus variant of ST93 (ST304) can be explained by in situ
microevolution or by a chance association. ST93 and ST304 have both been isolated previously in northeast Thailand. We recovered ST93 from the environment in 1990, 1998 and 1999, each from different sampling sites situated along road 212 which runs northwest from the town of Ubon Ratchathani. The MLST database (www.mlst.net
) contains a fourth ST93 isolate that was associated with human disease in Thailand in 1998. We have also recovered ST304 from two patients with melioidosis presenting to a hospital in northeast Thailand in 1999. However, an accurate picture of the distribution and frequency of co-localization in soil of ST93 and ST304 in this region has not been defined, and it is difficult to speculate on the probability of a chance association.
A potential pitfall of this study is that the proportion of each ST was obtained after the soil sample had been prepared by mixing with water and overnight sedimentation followed by growth using rich media. Some STs may be more adapted to survival or growth after the addition of distilled water during sample preparation, or may move more efficiently into the layer of surface water that is removed for culture. It is also possible that some STs are more likely to grow on laboratory media than others, and that some STs are viable but non-culturable under the conditions used. Resolution of these issues will require the direct application of molecular tools to soil samples, and comparison of genotypes with those obtained using conventional culture and existing soil preparation methods.
Our findings have several important implications for future genotyping studies. Soil sampling at a single location will fail to identify the genotypes present at a distance of even a few meters. Furthermore, the predominance of a single ST at a given site requires that extra sampling effort is required to detect any genotypes present as a minority of the population. We estimate that the characterization of approximately 50 colonies from any single site would provide an 85% probability of detecting a genotype present at the site at a frequency of 2%. This is based on the exact 95% binomial confidence interval for ST60 at sampling point A11, which was present in the lowest proportion. The temporal stability of the genotype distribution described here is not known, and it is possible that markedly different genotypic frequencies might be recorded from the same sampling sites if the study were to be repeated at some point in the future. It is also unclear whether the degree of genetic diversity described here will be reproduced within Thailand and in other endemic countries, although a study by Pearson et al. in which genetic diversity was demonstrated by variable number tandem repeat (VNTR) analysis within a small geographic area of Australia 
suggests that this will prove to be a reproducible finding. The basis on which PCR is used in future studies to detect B. pseudomallei
in the environment also requires careful consideration. Amplification may give rise to mixed products, and DNA from strains present at low copy number may go undetected. PCR may become an appropriate technology for the detection of B. pseudomallei
, but is not an appropriate basis for subsequent genotyping unless multiple independent amplicons are evaluated.
This study has investigated B. pseudomallei in soil taken from an area of disused land. This is in contrast to many previous studies in Thailand which were conducted in rice paddies. The basis for our choice was to examine an environment free of external influences such as chemical fertilizers and pesticides, together with the effect of ploughing, planting, burning of rice stubble and the presence of rice plants. However, most disease is probably acquired in rice paddies; further studies are underway to compare and contrast the findings reported here with those from a rice paddy in the same region.