Microbial source tracking by DNA sequence analysis of the Escherichia coli malate dehydrogenase gene

Criteria for sub-typing of microbial organisms by DNA sequencing proposed by Olive and Bean were applied to several genes in Escherichia coli to identify targets for the development of microbial source tracking assays. Based on the aforementioned criteria, the icd (isocitrate dehydrogenase), and putP (proline permease) genes were excluded as potential targets due to their high rates of horizontal gene transfer; the rrs (16S rRNA) gene was excluded as a target due to the presence of multiple gene copies, with different sequences in a single genome. Based on the above criteria, the mdh (malate dehydrogenase) gene was selected as a target for development of a microbial source tracking assay. The mdh assay was optimized to analyze a 150bp fragment corresponding to residues G191 to R240 (helices H10 and H11) of the Mdh catalytic domain. 295 fecal isolates (52 horse, 50 deer, 72 dog, 52 seagull and 69 human isolates) were sequenced and analyzed. Target DNA sequences for isolates from horse, dog plus deer, and seagull formed identifiable groupings. Sequences from human isolates, aside from a low level (ca. 15%) human specific sequence, did not group; nevertheless, other hosts could be distinguished from human. Positive and negative predictive values for two- and three-way host comparisons ranged from 60% to 90% depending on the focus host. False positive rates were below 10%. Multiple E. coli isolates from individual fecal samples exhibited high levels of sequence homogeneity, i.e. typically only one to two mdh sequences were observed per up to five E. coli isolates from a single fecal sample. Among all isolates sequenced from fecal samples from each host, sequence homogeneity decreased in the following order: horse>dog>deer>human and gull. For in-library isolates, blind analysis of fecal isolates (n=12) from four hosts known to contain host specific target sequences was 100% accurate and 100% reproducible for both DNA sequence and host identification. For blind analysis of non-library isolates, 18/19 isolates (94.7%) matched one or more library sequences for the corresponding host. Ten of eleven geographical outlier fecal isolates from Florida had mdh sequences that were identical to in-library sequences for the corresponding host from California. The mdh assay was successfully applied to environmental isolates from an underground telephone vault in California, with 4 of 5 isolates matching sequences in the mdh library. 146 sequences of the 645bp mdh fragment from five host sources were translated into protein sequence and aligned. Seven unique Mdh protein sequences, which contained eight polymorphic sites, were identified. Six of the polymorphic sites were in the NAD+ binding domain and two were in the catalytic domain. All of the polymorphic sites were located in surface exposed regions of the protein. None of the non-silent mutations of the Mdh protein were in the 150bp mdh target. The advantages and disadvantages of the assay compared to established source tracking methods are discussed.