Into the deep: Evaluation of SourceTracker for assessment of faecal contamination of coastal waters

Highlights

• Microbial source tracking tool SourceTracker was optimised by artificial community analysis.

• Application of SourceTracker to coastal waters found treated wastewater was a major contaminant.

• Detection of contaminated wastewater at the sampling locations correlated with rainfall events.

• Contamination of coastal waters by the Yarra River estuary was shown to occur after rainfall.

• Sand resuspension was found to correlate with increases in treated effluent detections at beaches.

Abstract

Faecal contamination of recreational waters is an increasing global health concern. Tracing the source of the contaminant is a vital step towards mitigation and disease prevention. Total 16S rRNA amplicon data for a specific environment (faeces, water, soil) and computational tools such as the Markov-Chain Monte Carlo based SourceTracker can be applied to microbial source tracking (MST) and attribution studies. The current study applied artificial and in-laboratory derived bacterial communities to define the potential and limitations associated with the use of SourceTracker, prior to its application for faecal source tracking at three recreational beaches near Port Phillip Bay (Victoria, Australia). The results demonstrated that at minimum multiple model runs of the SourceTracker modelling tool (i.e. technical replicates) were required to identify potential false positive predictions. The calculation of relative standard deviations (RSDs) for each attributed source improved overall predictive confidence in the results. In general, default parameter settings provided high sensitivity, specificity, accuracy and precision. Application of SourceTracker to recreational beach samples identified treated effluent as major source of human-derived faecal contamination, present in 69% of samples. Site-specific sources, such as raw sewage, stormwater and bacterial populations associated with the Yarra River estuary were also identified. Rainfall and associated sand resuspension at each location correlated with observed human faecal indicators. The results of the optimised SourceTracker analysis suggests that local sources of contamination have the greatest effect on recreational coastal water quality.

Introduction

Coastal watersheds provide the public with recreational amenities for both primary and secondary contact activities. However, stresses from increased urbanisation and population growth can negatively impact these ecosystems resulting in degradation of water quality. Faecal contamination, particularly of human origin, is recognised, worldwide, as one of the leading causes of waterway pollution and represents a major on-going, public health risk (Mallin et al., 2000). Water quality guidelines have been established based on faecal indicator organism (FIO) levels. But failure to directly identify the source of contamination and implement targeted mitigation strategies often results in faecal contamination levels exceeding regulatory standards (Handler et al., 2006, Huang et al., 2015, Shahidul Islam and Tanaka, 2004).

Historically, identification and assessment of faecal contamination in recreational coastal waters has primarily relied on the detection of FIOs such as enterococci or Escherichia coli (Neave et al., 2014). A fundamental limitation of these methods is that FIOs can originate from a variety of sources. Thus, the detection of E. coli, for example, does not indicate origin; information essential to the implementation of targeted mitigation strategies by regulatory bodies. Further complicating these analyses is the presence of naturalised environmental FIO populations which can lead to over-estimation of true faecal contamination levels (Ishii et al., 2006).

The advent of high-throughput sequencing (HTS) technologies has significantly advanced microbial source tracking (MST) studies. In particular, for the identification of sources of faecal contamination within complex systems (Neave et al., 2014, Casanovas-Massana et al., 2015, Ervin et al., 2014, Korajkic et al., 2015, Newton et al., 2013). Thus, it is now feasible, and becoming increasingly commonplace, to characterize the entire microbial community of faecal and environmental sources of interest (Korajkic et al., 2015, Knights et al., 2011a). In response, computational MST tools, such as Ichnaea and SourceTracker, have been developed to utilise HTS data to accurately estimate the proportion of contamination within an environment (Casanovas-Massana et al., 2015, Knights et al., 2011b). The approaches assume that diversity between faecal-derived and endogenous communities will enable discrimination between the populations based on their microbial signatures (Korajkic et al., 2015, Knights et al., 2011b). These methods, often referred to as signature-based MST, have been previously applied to identify faecal contamination in coastal waters, recreational beaches, lakes and to define changes in bacterial communities associated with sewage decomposition (Neave et al., 2014, Korajkic et al., 2015, Newton et al., 2013, Knights et al., 2011b, Flores et al., 2011). However, it is noted that most previous studies have applied small local source datasets, or integrated alternative publically-available HTS data sources. The consequences of which can bias results due to over representation of data for specific sources; i.e. human faecal datasets are present at higher proportions in public databases in comparison to other animal faeces (such as flying foxes and possums). Consequently, the requirement for large regionally-specific source datasets is an increasingly recognised need within the field (Knights et al., 2011b, Harwood et al., 2014).

Previous water-based investigations have also not defined the limitations of the computational tools. Thus, the conditions required to ensure confidence in the source attribution results is unknown. Limited research has been conducted for equivalent studies within built and human environments (Lax et al., 2014, Tito et al., 2012). To increase confidence and support the results of computational MST, studies, such as those of Neave et al. (2014), have alternatively applied culture-based FIO results. As previously stated, the use of FIOs and culture based methods do have inherent limitations associated with the techniques and targeted organism.

The current study applied HTS and SourceTracker software to identify signatures of human and animal derived waste in coastal waters and associated recreational beaches. SourceTracker applies a Bayesian framework to estimate the proportion of each source contributing to a designated sink sample (Knights et al., 2011b) (a brief overview figure of the model is provided in Supplementary Material 1). The tool forms part of the QIIME open access bioinformatics platform for analysis of metagenome and 16S rRNA amplicon data (Caporaso et al., 2010). To ensure confidence and define the limitations of the SourceTracker, this study investigated artificial community analysis under default and altered parameter conditions. The results of which were then applied to laboratory scale MST investigations to ensure the selected condition performed as expected. The selected conditions were then applied to data from 436 local and regionally specific water and faecal sources to track contamination across summer 2014-15 at three recreational beaches near Port Phillip Bay (Victoria, Australia). The data was examined to identify the main sources of contamination. The results from this study will assist water managers with achieving better understanding of the sources of contamination in these beach catchments.