Multiple lines of evidence to identify sewage as the cause of water quality impairment in an urbanized tropical watershed.

Indicator bacteria, which are conventionally used to evaluate recreational water quality, can originate from various non-human enteric and extra-enteric sources, hence they may not be indicative of human health risk nor do they provide information on the sources of contamination. In this study we utilized traditional (enterococci and Escherichia coli) and alternative (Clostridium perfringens) indicator bacteria, F-specific coliphage, molecular markers for microorganisms associated with human sewage (human-associated Bacteroides and polyomaviruses), and microbial community analysis tools (16S rRNA gene fragment amplicon sequencing), to identify and evaluate human sewage-related impact in the Manoa watershed in Honolulu, Hawaii. Elevated concentrations of enterococci (geometric mean ranging from 1604 to 2575 CFU 100 mL) and C. perfringens (45-77 CFU 100 mL) indicated impairment of the urbanized section of the stream, while indicator bacteria concentrations decreased downstream in the tidally influenced Ala Wai Canal. The threshold values triggering water quality violation notifications in Hawaii were exceeded in 33.3-75.0% of samples collected at sites in the urbanized section of Manoa Stream, but were not exceeded in any of the samples collected at an upstream site located in a forested area. Correlation between indicator bacteria concentrations and rainfall amounts was weak to moderate but significant (E. coli R = 0.251, P = 0.009; enterococci R = 0.369, P < 0.001; C. perfringens R = 0.343, P < 0.001), while concentrations of human fecal-associated molecular markers were not significantly correlated with rainfall (human-associated Bacteroides, R = 0.131, P = 0.256; human-associated polyomaviruses, R = 0.213, P = 0.464). Presence of human sewage was confirmed by detection of human-associated Bacteroides and human polyomavirus in the urbanized section of Manoa Stream (83.3-100% and 41.7-66.7% positive samples respectively). It was further confirmed by microbial community analyses which suggested that an average 2.4-3.4% of the total bacterial population in this section was associated with sewage. Microbial community profiles were significantly influenced by rainfall (R = 0.4390, P < 0.001), pH (R = 0.3077, P = 0.006), salinity (R = 0.2614, P = 0.038), and conductivity (R = 0.2676, P = 0.031). Although microbial diversity fluctuated throughout the watershed, it was lower in the impaired section. Leaking sewer systems and illegal cross-connections are implicated in the impairment of the watershed, hence both the sewer and the storm water lines should be routinely inspected. Collectively, our data suggest that information derived from the analysis of microbial communities complements current marker-based microbial source tracking techniques and environmental monitoring programs.