16S rRNA-based assays for quantitative detection of universal, human-, cow-, and dog-specific fecal Bacteroidales: A Bayesian approach

We report the design and validation of new TaqMan® assays for microbial source tracking based on the amplification of fecal 16S rRNA marker sequences from uncultured cells of the order Bacteroidales. The assays were developed for the detection and enumeration of non-point source input of fecal pollution to watersheds. The quantitative “universal” Bacteroidales assay BacUni-UCD detected all tested stool samples from human volunteers (18 out of 18), cat (7 out of 7), dog (8 out of 8), seagull (10/10), cow (8/8), horse (8/8), and wastewater effluent (14/14). The human assay BacHum-UCD discriminated fully between human and cow stool samples but did not detect all stool samples from human volunteers (12/18). In addition, there was 12.5% detection of dog stool (1/8), but no cross-reactivity with cat, horse, or seagull fecal samples. In contrast, all wastewater samples were positive for the BacHum-UCD marker, supporting its designation as 100% sensitive for mixed-human source identification. The cow-specific assay BacCow-UCD fully discriminated between cow and human stool samples. There was 38% detection of horse stool (3/8), but no cross-specificity with any of the other animal stool samples tested. The dog assay BacCan-UCD discriminated fully between dog and cow stool or seagull guano samples and detected 62.5% stool samples from dogs (5/8). There was some cross-reactivity with 22.2% detection of human stool (4/18), 14.3% detection of cat stool (1/7), and 28.6% detection of wastewater samples (4/14). After validation using stool samples, single-blind tests were used to further demonstrate the efficacy of the developed markers; all assays were sensitive, reproducible, and accurate in the quantification of mixed fecal sources present in aqueous samples. Finally, the new assays were compared with previously published sequences, which showed the new methodologies to be more specific and sensitive. Using Bayes’ Theorem, we calculated the conditional probability that the four assays would correctly identify general and host-specific fecal pollution in a specific watershed in California for which 73 water samples had been analyzed. Such an approach allows for a direct comparison of the efficacy of different MST methods, including those based on library-dependent methodologies. For the universal marker BacUni-UCD, the probability that fecal pollution is present when the marker is detected was 1.00; the probability that host-specific pollution is present was 0.98, 0.84, and 0.89 for the human assay HF160F, the cow assay BacCow-UCD, and the dog assay BacCan-UCD, respectively. The application of these markers should provide meaningful information to assist with efforts to identify and control sources of fecal pollution to impaired watersheds.