Microbial source tracking in highly vulnerable karst drinking water resources

Water resources situated in areas with underlying karst geology are particularly vulnerable to fecal pollution. In such vulnerable systems, microbial source tracking (MST) methods are useful tools to elucidate the pathways of both animal and human fecal pollution, leading to more accurate water use risk assessments. Here, we describe the application of a MST toolbox using both culturedependent bacteriophage and molecular-dependent 16S rRNA assays at spring and well sites in the karstic St Imier Valley, Switzerland. Culture-dependent and molecular-dependent marker performance varied significantly, with the 16S rRNA assays displaying greater sensitivity than their phage counterpart; HF183 was the best performing human wastewater-associated marker while Rum2Bac was the best performing ruminant marker. Differences were observed in pollution regimes between the well and spring sampling sites, with the spring water being more degraded than the well site. Our results inform the choice of marker selection for MST studies and highlight differences in microbial water quality between well and spring karst sites. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/wh.2017.215 om https://iwaponline.com/jwh/article-pdf/16/1/138/240149/jwh0160138.pdf 020 D. Diston (corresponding author) A. Baumgartner R. Felleisen Federal Food Safety and Veterinary Office FSVO, Sector Laboratories, Schwarzenburgstrasse 165, 3003 Bern, Switzerland E-mail: d_diston@yahoo.co.uk R. Robbi Cantonal Laboratory, Bern, Switzerland


INTRODUCTION
Karst areas are an important source of drinking water throughout the world, with Switzerland drawing 80% of its drinking water from groundwater aquifers (Tripet ).
Fractures, sinkholes, and other conduits which are characteristic of karst systems, facilitate the rapid movement of water from the terrestrial surface to receiving subterranean waterbodies, with rainfall-driven influx of microbial pollution a particular problem (Pronk et al. ; Sinreich et al. ; Diston et al. ). This rainfall-driven overland flow may transport a range of human and animal pollutants to water resources in karst systems and as karst systems have short water retention times, in the order of hours as opposed to days (Pronk et al. ), little reduction of this terrestrialderived microbial load occurs between system input and output.
Contamination of important drinking water resources by fecal pollution may lead to outbreaks of illness and economic degradation (Beaudeau et al. ), with human fecal pollution representing a greater risk to water users than animal fecal pollution (Soller et al. ). Traditional indicators of non-specific fecal pollution such as indicator bacteria, are a useful first step in identifying incidences of aquatic fecal contamination. However, as their discriminatory ability is limited and they offer no indication of pollutant source, opportunities to rectify pollution problems and undertake water use risk assessments are limited.
Microbial source tracking (MST) techniques are powerful tools, able to identify the sources of pollution in catchments using either library-dependent or library-independent methods. Application of MST techniques allows targeted remediation strategies to be employed, saving time and money (see review by Ahmed et al. () for a discussion of current MST tools).
Here, we describe the application of a MST toolbox employing both culture-and molecular-dependent parameters within two highly vulnerable karst water resources. Previous work has shown that karst systems within Switzerland are vulnerable to rainfall-driven microbial contamination (Diston et al. ) and therefore we undertook more intense sampling than has been previously conducted in previous Swiss MST studies. The aims of this paper are: (i) to compare the sensitivity of culture-and molecular-dependent MST indicators in two highly vulnerable karst catchments (one spring and one well subject to similar land use); and (ii) to trial 16S rRNA Bacteroidales indicators not previously used in Switzerland.

Sample locations and collection of water samples
Historic data provided by the Cantonal Laboratory, Bern allowed the identification of 14 sites within the St. Imier valley (Bern, Switzerland) where thermotolerant Escherichia coli were frequently recorded in groundwater samples (data not shown). To assess the magnitude of fecal pollution, preliminary sampling was undertaken at all 14 sites on two occasions using fecal coliforms (FC), intestinal enterococci (ENT), and the human wastewater-associated marker HF183 (Seurinck et al. ). Two sites, Sonceboz-Sombeval and Tavannes, were found to consistently have higher concentrations of all three parameters and were selected for detailed study ( Figure 1).
The sites at Sonceboz and Tavannes were sampled 13 times between June and December 2014 at approximately 2-week intervals. Both sites are related to karst aquifers and are presumed to respond rapidly to rainfall. The sampling point at Sonceboz was in a pumping house, which extracts water from the aquifer via a well, whereas the sampling point at Tavannes was a small reservoir receiving water directly from the aquifer. Although the sampling sites are located only 4 km apart, they belong to different drainage basins. Land use at both sites is similar, with a mixture of extensive cattle grazing, meadows, forest, and small towns of less than 5,000 residents (2015 GIS data provided by Canton Bern Department of Economics).
At each site, 2 × 1 gallon (4.55 L) grab samples of untreated drinking water were collected in sterile polypropylene carboys (Nalgene, USA), with sample containers being pre-rinsed in situ three times before sample collection.
Water samples were kept on ice blocks in a polystyrene cool box and transported back to the Federal Food Safety and Veterinary Office laboratory (Köniz, Bern, Switzerland) within 2 hours. All filtrations were conducted within 6 hours of sample collection.

MST indicators and fecal indicator bacteria
A range of culture-and molecular-dependent, sourcespecific and general MST parameters were assessed in this study. Three bacteriophage groups were assayed: bacteriophages infecting human-specific host strains Bacteroides  Table 1.
500 mL of water sample was filtered on a 47 mm 0.2 μm cyclopore track etched membrane (Whatman, USA). The filter was folded and transferred into a 15 mL DNA-free, screw-top test tube containing 0.5 mL of GITC buffer (5 M guanidine isothiocyanate, 100 mM EDTA (pH 8.0), 0.5% Sarkosyl), inverted and stored at À20 W C until DNA extraction.
QIAmp DNA mini kit (Qiagen, USA) was used to extract DNA from thawed filters. 3 μL DNA aliquots, and dilutions thereof, were used in qPCR assays and were run on an ABI  quantified using a Nanodrop spectrophotometer. Standard curves using extracted plasmid DNA from 1.0 × 10 5 to 1 copy/reaction were created and run in the qPCR assays. The concentrations of the primers and probes in the qPCR systems remained unchanged from the original papers. Blank filtrations using sterile Elga Purelab Ultra water were conducted in parallel. Where samples returned a negative result, they were re-tested using further dilutions to investigate possible inhibition.

Human enteric viruses
Norovirus genogroups I and II, human enterovirus, and human rotavirus group A were assayed using qPCR.
Samples were concentrated using the method of Katayama et al. () and although this method uses only 1 L of water, it has been used successfully in many environmental High Capacity cDNA Archive Kit (Applied Biosystems, USA) was used to synthesize cDNA, which was then stored at À20 W C until required.
The viral qPCR assays remained unchanged from the original publication, with the exception of substituting TAMRA for BHQ-1 in all systems (Table 1)

Rainfall data
Rainfall data were sourced from the Swiss Rainfall Bulletin (Federal Office of Meteorology and Climatology MeteoSwiss) and 1-5 day cumulative totals were calculated.
The nearest weather station to the sampled sites was Courtelary, approximately 9 and 12 km from Sonceboz and Tavannes, respectively.

Statistical methods
As the data were non-normally distributed (Shapiro-Wilk test) and log transformation did not produce normally dis-

Water characteristics
Temperature of the samples varied throughout the monitoring period (typically between 5 W C and 12 W C at both sites) and there were no significant relationships between temperature and parameter concentrations observed at either site.

General indicators of fecal pollution
Across both sites, FIB were detected in 100% of samples (n ¼ 26), with FC always present at a higher concentration that ENT (median ¼ 182 and 50 CFU/100 mL, respectively; both a lower detection rate (54%, n ¼ 14) and median concentration value of 3.69 × 10 5 copies/100 mL than AllBac.
These data suggest that AllBac is more suitable for identifying general fecal pollution in Switzerland than GenBac. The data also highlight the usefulness of cost-effective FIB when compared to 16S rRNA assays for indicating the presence of non-specific fecal pollution in karst catchments.

Human MST indicators
While FIB and the general indicators discussed above high-   and sampling campaign of our study, shows that HuBac has poor specificity for human wastewater, 64.6%, whereas HF183 has a sensitivity of 94.6%. These data explain why HuBac performed poorly in our study. It is advisable to assess the specificity of HF183 within Switzerland if further studies using these markers are to be conducted.

Ruminant MST indicators
Previous work found that BacR performed poorly in Switzerland (Diston et al. ) and it was for this reason that we included an alternative ruminant indicator in the study. Rum2Bac was detected more frequently than BacR, 54% and 35%, respectively, although the median BacR concentration was higher, 7.05 × 10 4 cn/100 mL compared to 5.93 × 10 3 cn/100 mL for Rum2Bac. Specificities of between 99 and 100% have been reported for BacR

Viral pathogens
Viral pathogens were largely absent from water samples, with norovirus GI and GII, group A rotavirus not being detected while enterovirus was found in only one sample (two other samples gave positive results that were on or below the limit of detection (LOD) of 40 virions per liter).
In this positive sample, FIB, somatic coliphage, HF183, and HuBac were co-occurring, whereas human-specific phage indicators were absent in this sample.

Comparison of sites
It is clear from both the FIB and the MST data that Tavannes  Although FIB were detected in 100% of samples from both sites, indicating ubiquitous fecal contamination, AllBac was only detected in 54% of samples at Sonceboz (this situation was mirrored to a lesser extent by GenBac). Moreover, both ruminant and human wastewater-associated 16S rRNA markers also followed this pattern; HuBac, BacR, and Rum2Bac were detected on a single occasion at Sonceboz whereas at Tavannes they were present at 85%, 62%, and 100%, respectively. As indicated by the GIS data, predominant land use patterns are very similar, suggesting something This may explain why, despite having similar land use, the microbial contamination problem at Tavannes is worse than that at Sonceboz.
Strong linear correlations were found between GB-124 and    It is of note that there were more significantly statistical/ stronger positive linear correlations at Sonceboz than Tavannes. As the water at Sonceboz is likely to be sourced from a small hydrological area, few pollutant sources are likely to contribute to the microbial load, leading to increased parameter correlation. The hydrological situation at Tavannes is more complex, with the spring draining a larger area than Sonceboz. Therefore, more pollutant sources with varying degrees of hydraulic conductivity contribute to the microbial fingerprint of the site, causing poor correlation of the parameters.

CONCLUSIONS
We analyzed two karst water sources for a range of microbiological parameters, and for the first time trialled new molecular-dependent MST indicators in Switzerland. This study has added to the understanding of how both cultureand molecular-dependent MST indicators relate to each other in karst systems, and the geographical limitation of indicators. The primary conclusions from this study may be summarized as follows: • Molecular indicators appear to offer a more sensitive assessment of human wastewater-associated pollution in karst systems.
• Despite being geographically close and having similar land use, drinking water sourced from springs (Tavannes) may be of worse microbial quality than that of wells (Sonceboz).
• Disparities in group-specific indicators were observed; Rum2Bac appears to be more sensitive than BacR, HF183 is more sensitive than HuBac and AllBac is more sensitive than GenBac.
In order to increase understanding of the geographical limitations in terms of sensitivity and specificity, the proposed MST indicators should be evaluated in both karst and nonkarst catchments in Switzerland. Further work could include studies using greater volumes of groundwater, thereby enhancing viral detection, and in conjunction with viral infectivity assays, could better inform microbial risk assessments by improving selection of the most appropriate MST toolbox.