Comparison of the microbiomes of two drinking water distribution systems—with and without residual chloramine disinfection

Background Residual disinfection is often used to suppress biological growth in drinking water distribution systems (DWDSs), but not without undesirable side effects. In this study, water-main biofilms, drinking water, and bacteria under corrosion tubercles were analyzed from a chloraminated DWDS (USA) and a no-residual DWDS (Norway). Using quantitative real-time PCR, we quantified bacterial 16S rRNA genes and ammonia monooxygenase genes (amoA) of Nitrosomonas oligotropha and ammonia-oxidizing archaea—organisms that may contribute to chloramine loss. PCR-amplified 16S rRNA genes were sequenced to assess community taxa and diversity. Results The chloraminated DWDS had lower biofilm biomass (P=1×10−6) but higher N. oligotropha-like amoA genes (P=2×10−7) than the no-residual DWDS (medians =4.7×104 and 1.1×103amoA copies cm−2, chloraminated and no residual, respectively); archaeal amoA genes were only detected in the no-residual DWDS (median =2.8×104 copies cm−2). Unlike the no-residual DWDS, biofilms in the chloraminated DWDS had lower within-sample diversity than the corresponding drinking water (P<1×10−4). Chloramine was also associated with biofilms dominated by the genera, Mycobacterium and Nitrosomonas (≤91.7% and ≤39.6% of sequences, respectively). Under-tubercle communities from both systems contained corrosion-associated taxa, especially Desulfovibrio spp. (≤98.4% of sequences). Conclusions Although residual chloramine appeared to decrease biofilm biomass and alpha diversity as intended, it selected for environmental mycobacteria and Nitrosomonas oligotropha—taxa that may pose water quality challenges. Drinking water contained common freshwater plankton and did not resemble corresponding biofilm communities in either DWDS; monitoring of tap water alone may therefore miss significant constituents of the DWDS microbiome. Corrosion-associated Desulfovibrio spp. were observed under tubercles in both systems but were particularly dominant in the chloraminated DWDS, possibly due to the addition of sulfate from the coagulant alum. Electronic supplementary material The online version of this article (10.1186/s40168-019-0707-5) contains supplementary material, which is available to authorized users.

Supplemental Text S1 Water mains from winter-shutoff sites Though collected during normal operation, the water mains at these sites are shutoff annually during the cold-weather months due to freezing concerns; though technically still operational and full of drinking water, these mains lie adjacent to gate-valves that are closed to completely halt flow. Water in the main on the opposite side of each gate-valve is drained to prevent ice formation and expansion damage. These dead ends are likely stagnant for months-long periods every year.

S2 Library-size normalization for beta metrics
Beta diversity was assessed using the generalized UniFrac distances and checked using the unweighted UniFrac distances and Bray-Curtis dissimilarity as alternatives. Depending on the metric, read counts were normalized to account for the uneven library sizes [S1]. No normalization was performed for generalized UniFrac (i.e. original counts were used) [S2]. For unweighted UniFrac, counts were equally subsampled to the lowest count of any sample [S3]. Bray-Curtis dissimilarity was calculated using read counts normalized with cumulative sum scaling [S4].

S3 Synthesis of qPCR standards
Standards were created using either plasmid DNA (Nitrosomonas oligotropha-like amoA genes) or custom gBlocks gene fragments (archaeal amoA genes). Plasmid DNA was prepared from PCR amplification using positive controls, followed by ligation with pGEM-T Easy cloning vectors (Promega, Madison, WI, USA) and transformation into Escherichia coli JM109. After purification with the QIAprep Spin Miniprep Kit (QIAGEN, Hilden, Germany), plasmid DNA was stained with Hoechst 33258 dye and quantified on a TD-700 fluorometer (Turner Designs, Sunnyvale, CA, USA) using calf thymus DNA as a standard. For archaeal amoA gene standards, custom gBlocks gene fragments were synthesized by Integrated DNA Technologies using a reference fragment from Nitrosopumilus maritimus (GenBank accession HM345610; https://www.ncbi.nlm.nih.gov/genbank).

S4 qPCR of under-tubercle samples
The 16S rRNA gene copy numbers for under-tubercle samples are not shown due to the lack of an appropriate normalization parameter. Surface area was deemed inappropriate as there appeared to be substantial differences in the masses of corrosion solids recovered from under different tubercles. Dry mass of corrosion solids may have been a suitable parameter, but unfortunately, masses were not quantified. Furthermore, AOB and AOA were not considered relevant for the under-tubercle communities, which were ostensibly oxygen-deficient, so qPCR was not performed targeting either amoA gene target.

C1
Winter shutoff sites Figure S1. Locations of sample collection sites, with linear geographic distances relative to the drinking water treatment plants in the (a) chloraminated and (b) no-residual drinking water distribution systems in the United States and Norway, respectively. Total chlorine and assimilable organic carbon concentrations in drinking water are labelled near the respective measurement sites. Geographic scale between a and b are 1:1; cardinal directions and geographic features have been masked to retain anonymity of the two participating municipalities.   Indices represent mean average of 10 random subsamples (without replacement). Subsamples were performed at 10 intervals within the range of 10 to 74 880 sequences per sample (i.e., the median library size). Dashed vertical lines indicate the minimum library size among all samples (17 041 sequences per sample); some samples terminate before this threshold due to the interval size of the rarefaction analysis.  AOB, log 10 (16S rRNA genes) NOB, log 10 (16S rRNA genes)

Sample type
Water-main biofilms Drinking water

Sample type
Water-main biofilms Drinking water Under tubercle Figure S8. Principal coordinates analysis of generalized UniFrac for all samples collected from the chloraminated and no-residual drinking water distributions systems. This represents the between-sample (i.e., beta) diversity for water-main biofilms, drinking water, and under tubercle. Percentages = variance explained.