AWWA WQTC63992

For library-dependent bacterial source tracking (BST) methods, bacteria (e.g. E. coli) isolated from water samples are compared to reference libraries of genetic and phenotypic fingerprints of bacterial isolates from known human and animal sources to identify the sources of fecal pollution. BST technology continues to be developed and applied to new studies despite controversy over the usefulness of BST, especially the library-dependent methods. This paper presents a recently completed E. coli BST project that included comparison of one phenotypic and three genetic library-dependent BST methods. E. coli isolates were characterized by automated ribotyping using the Qualicon RiboPrinter, repetitive sequence polymerase chain reaction (ERIC-PCR), pulsed field gel electrophoresis (PFGE) and Kirby-Bauer antibiotic resistance analysis (KB-ARA). These methods span the spectrum in their ability to resolve differences in related bacterial strains, training and labor required, and cost. Source samples were collected from municipal wastewater treatment plant influent/effluent and septage samples (collectively referred to as "domestic sewage"), livestock, wildlife, and pet fecal samples. In total, 2,275 E. coli isolated from 813 known source samples were screened using ERIC-PCR to exclude clonal (identical) isolates and maximize diversity of the library. A total of 883 E. coli isolates from 745 different source samples were selected for the library and analyzed by all four BST methods. Fingerprints of isolates were compared based on the Pearson-product similarity coefficient with a "best match" approach to a single isolate at greater than or equal to 85% similarity for ERIC-PCR, RiboPrinting and KB-ARA, and greater than or equal to 70% for PFGE. Data from the different BST methods were also combined in composite data sets, and a 70% similarity cutoff was used to allow for variation of the individual methods. PFGE had the highest rate of correct classification (RCC) of any individual method in jackknife analysis of the library isolates, but could only identify 20% of the 555 E. coli isolated from 412 water samples. A four-method (ERIC-RP-PFGE-ARA) composite data set was found to have the next highest RCC values ranging from 22% to 83% based on a seven-way split into domestic sewage, pet, cattle, other livestock, avian, non-avian, and wildlife sources. These RCCs were 3- to 7-fold higher than random probability. Congruence measurements of individual methods and composite data sets revealed that the ERIC-PCR and RiboPrinting (ERIC-RP) and ERIC-PCR and KB-ARA (ERIC-ARA) composite data sets were 90.7% and 87.2% similar, respectively, to the four-method composite data set. Further, these two-method composite data sets identified the same leading source of fecal pollution as the four-method composite data set. In future studies, a "toolbox" approach using ERIC-RP or ERIC-ARA composite data, along with library-dependent methods to aid with sampling design, may help provide cost, labor, and time savings while not compromising the integrity of the BST results. Includes tables, figure.