The impact of sampling, PCR, and sequencing replication on discerning changes in drinking water bacterial community over diurnal time-scales

Highlights

• PCR and sequencing variability masks differences between replicate water samples.

• Sequencing replication allows for reliable detection of low abundance taxa.

• Bacterial community richness, structure, and membership changes diurnally.

• Relative abundance of dominant taxa can change in excess of 10% within one day.

• PCR/sequencing replication must be included in drinking water microbiome studies.

Abstract

High-throughput and deep DNA sequencing, particularly amplicon sequencing, is being increasingly utilized to reveal spatial and temporal dynamics of bacterial communities in drinking water systems. Whilst the sampling and methodological biases associated with PCR and sequencing have been studied in other environments, they have not been quantified for drinking water. These biases are likely to have the greatest effect on the ability to characterize subtle spatio-temporal patterns influenced by process/environmental conditions. In such cases, intra-sample variability may swamp any underlying small, systematic variation. To evaluate this, we undertook a study with replication at multiple levels including sampling sites, sample collection, PCR amplification, and high throughput sequencing of 16S rRNA amplicons. The variability inherent to the PCR amplification and sequencing steps is significant enough to mask differences between bacterial communities from replicate samples. This was largely driven by greater variability in detection of rare bacteria (relative abundance <0.01%) across PCR/sequencing replicates as compared to replicate samples. Despite this, we captured significant changes in bacterial community over diurnal time-scales and find that the extent and pattern of diurnal changes is specific to each sampling location. Further, we find diurnal changes in bacterial community arise due to differences in the presence/absence of the low abundance bacteria and changes in the relative abundance of dominant bacteria. Finally, we show that bacterial community composition is significantly different across sampling sites for time-periods during which there are typically rapid changes in water use. This suggests hydraulic changes (driven by changes in water demand) contribute to shaping the bacterial community in bulk drinking water over diurnal time-scales.

Introduction

Deep nucleic acid sequencing techniques continue to reveal the complexity of the drinking water (DW) microbiome, unveiling the presence of diverse and abundant microbial communities in full-scale drinking water distribution systems (DWDS) (McCoy and VanBriesen, 2012, Pinto et al., 2014). Treated bulk DW may contain from thousands to millions of cells per liter (Lipphaus et al., 2013, Lautenschlager et al., 2013), comprising hundreds to thousands of bacterial operational taxonomic units (OTU's) (Pinto et al., 2014). Given this microbial abundance and diversity, previous studies have suggested that DWDSs can no longer be considered as simple water conveyance systems, but should be considered as “reactors” (Lautenschlager et al., 2013, Liu et al., 2013) where a range of chemical, physical, and biological forces influence the quality of water at the consumer's tap. Studies designed to characterize the effects of these forces have assessed changes in microbial communities (primarily bacterial). For instance, the impact of processes at drinking water treatment plants (DWTPs) (Roeselers et al., 2015, Pinto et al., 2012), the structure of the DWDS (Pinto et al., 2014), water main flushing (Douterelo et al., 2013, Douterelo et al., 2014), disinfectant type (Williams et al., 2005), water stagnation in the DWDS and premise plumbing (Lipphaus et al., 2013, Lautenschlager et al., 2010) have been previously studied.

The majority of DW studies focus on the changes in structure and membership of microbial communities over varying temporal and spatial scales with the aim of deducing which parameter(s) underlies the observed changes. Studies thus far have investigated a range of temporal scales: annual (Pinto et al., 2014), monthly (McCoy and VanBriesen, 2012), weekly (Sekar et al., 2012) and daily (El-Chakhtoura et al., 2015). Similarly, studies targeting spatial variability in DW bacterial communities tend to span large regions of the complex distribution system (McCoy and VanBriesen, 2012, Pinto et al., 2014), while smaller spatial scales have been minimally explored. Despite the increasing application of nucleic acid sequencing technologies to characterize DW microbial communities, there has been little understanding of how the observed patterns and their magnitudes are influenced by the variability associated with sample collection and processing. From a methodological perspective, studies have investigated the impact of different types of filters for sample concentration (Revetta et al., 2011, Poitelon et al., 2009), DNA extraction methods (Hwang et al., 2012, Feinstein et al., 2009), PCR amplification protocols, and the primers used to target a particular region of the 16S rRNA gene (Chakravorty et al., 2007, Vasileiadis et al., 2012). Though it may be difficult to correct for these variabilities, it is imperative that sufficient replication efforts at all/appropriate steps are undertaken. This is particularly critical at small spatial and temporal scales, where the magnitude of changes under investigation may be significantly affected by methodological biases, leading to inaccurate and/or incomplete conclusions.

In this study, we test the ability of DNA sequencing-based approaches to capture changes in DW bacterial communities over small temporal scales when informed by replication at multiple levels; namely sampling sites, sample collection, PCR amplification, and amplicon sequencing. Additionally, we assessed the bacterial community variability within and between each of these nested levels over diurnal temporal scales. We selected the diurnal time scale to address the gap in current literature. Further, we explore the diurnal trends to determine time scales within a day over which changes in DW bacterial community (structure and membership) are significant enough to not be masked by methodological biases. And finally, informed by the outcomes of this study, we provide recommendations to improve and optimize future studies aimed at studying the DW microbiome in full-scale systems and discuss the implications of diurnal variability in bacterial community for the design of future studies.