Impact of drinking water conditions and copper materials on downstream biofilm microbial communities and Legionella pneumophila colonization

Authors


Abstract

Aims

This study examined the impact of pipe materials and introduced Legionella pneumophila on downstream Leg. pneumophila colonization and microbial community structures under conditions of low flow and low chlorine residual.

Methods and Results

CDC biofilm reactors containing either unplasticized polyvinylchloride (uPVC) or copper (Cu) coupons were used to develop mature biofilms on Norprene tubing effluent lines to simulate possible in-premise biofilm conditions. The microbial communities were characterized through 16S and 18S rRNA gene clone libraries and Leg. pneumophila colonization was determined via specific qPCR assays. The Cu significantly decreased downstream microbial diversity, approximately halved bacterial and eukaryotic abundance, with some groups only detected in uPVC-reactor tubing biofilms. However, some probable amoeba-resisting bacteria (ARB) like Mycobacterium spp. and Rhodobacteraceae were significantly more abundant in the Cu than uPVC-reactor tubing biofilms. In particular, Leg. pneumophila only persisted (postinoculation) within the Cu-reactor tubing biofilms, and the controlled low chlorine residue and water flow conditions led to a general high abundance of possible free-living protozoa in all tubing biofilms. The higher relative abundance of ARB-like sequences from Cu-coupons vs uPVC may have been promoted by amoebal selection and subsequent ARB protection from Cu inhibitory effects.

Conclusions

Copper pipe and low flow conditions had significant impact on downstream biofilm microbial structures (on plastic pipe) and the ability for Leg. pneumophila colonization post an introduction event.

Significance and Impact of the Study

This is the first report that compares the effects of copper and uPVC materials on downstream biofilm communities grown on a third (Norprene) surface material. The downstream biofilms contained a high abundance of free-living amoebae and ARB, which may have been driven by a lack of residual disinfectant and periodic stagnant conditions. Given the prevalence of Cu-piping in buildings, there may be increased risk from drinking water exposures to ARB following growth on pipe/fixture biofilms within premise drinking water systems.

Introduction

Legionella spp. are frequently identified in drinking water biofilms following water treatment. One species, Legionella pneumophila, accounts for some 30% of U.S. drinking water-related outbreaks and about 80% of drinking water-related deaths in the United States (Craun et al. 2010). Complex, but poorly characterized microbial communities that develop in drinking water biofilms appear to contribute to the survival and protection of these environmental, amoeba-resisting bacterial pathogens (Greub and Raoult 2004). Specifically, Legionella and Mycobacterium spp. are considered intracellular parasites of free-living amoebae (FLA) (Thomas et al. 2010), which support their survival and proliferation in various biofilm environments (Kilvington and Price, 1990). Recent studies by Corsaro et al. (2010) and Thomas et al. (2006) have illustrated the significant association between the presence of Legionella and Mycobacterium spp. and FLA in various drinking water systems, and their results supported the role of FLA as a reservoir/amplification host for these opportunistic pathogens in water systems. FLA, such as Acanthamoeba polyphaga and Vermamoeba (Hartmannella) vermiformis, are ubiquitous in drinking water systems and are resistant to commonly used disinfection methods (Thomas et al. 2008; Thomas and Ashbolt 2011). Some FLA are also pathogens in their own right (e.g. various Acanthamoeba spp. and Naegleria fowleri), as well as being able to host intracellular bacterial pathogens (Thomas and Ashbolt 2011).

There are many factors that shape biofilm development within premise plumbing, such as residual disinfectant concentrations, flow conditions, and pipe materials/corrosion products. For example, Ciesielski et al. (1984) reported that Legionella counts in a stagnant household header water tank were 10 times higher (at 100 CFU ml−1) than other tanks in use (<10 CFU ml−1). Further, distribution system and premise-plumbing designs that may result in more growth and lower chlorine residuals include oversized lines, dead ends, warming of the drinking water, closed valves and shower heads, largely due to longer residence (stagnation) times (Crozes and Cushing 2000). Bacteria may also grow on pipe joints, valves, elbows, tees and other fittings due to the changing water movement, materials used and stagnant areas of the premise plumbing (Berry et al. 2006). Such biofilm growth is likely to be greater on certain plastic than metal materials (van der Kooij et al. 2005; Yu et al. 2010; Bucheli-Witschel et al. 2012). Early work also indicated that high bacterial counts were seen in sampled drinking water from taps after overnight stagnation (LeChevallier et al. 1987).

Copper pipes have long been considered to suppress microbial growth in drinking water (van der Kooij et al. 2005), which may influence the growth of environmental pathogens. Focusing on copper vs plastic pipes, Morvay et al. (2011) reported that copper coupons were colonized by lower numbers of cells than PVC surfaces, and Lehtola et al. (2005) demonstrated faster biofilm growth on polyethylene than on copper pipes. Nevertheless, various other metal pipe materials appear to support more biofilm biomass and diversity than what develops on copper (Doğruöz et al. 2009; Jang et al. 2011). Overall, the least amount of biofilm formation potential and microbial diversity appears to occur on copper coupons (Yu et al. 2010). Most recently, Buse et al. (2014a,b) reported preferential colonization and release of Leg. pneumophila on Cu-associated biofilms to those from uPVC, as well as significantly different microbial communities. However, what appears unreported is the possible downstream microbial impact from different upstream pipe materials, which may impact the development of amoeba-resisting bacteria (ARB) and the biofilm community's ability to support or suppress introduced Leg. pneumophila (Wang et al. 2013).

The aim of this study was to explore the impact of upstream pipe materials (uPVC vs Cu) and introduced Leg. pneumophila with or without a host amoeba (A. polyphaga) in CDC biofilm reactors developed under conditions of low chlorine residual and periodic stagnant conditions. In particular, the ability for Leg. pneumophila colonization within downstream microbial communities developed on a third material, Norprene (Cole-Parmer, Vernon Hills, IL) tubing lines use to simulate plastic fixtures and potential increased growth conditions for environmental pathogens.

Materials and methods

Bacterial and amoebal culture preparation and enumeration

The preparation, culture and enumeration of Leg. pneumophila strain Lp02, A. polyphaga (ATCC 30461) and V. vermiformis (ATCC 50237) was as previously described (Buse and Ashbolt 2011). Briefly, cells were harvested on the day of inoculation and were washed twice with 5 ml filtered, autoclaved tap water (fat H2O). Densities of A. polyphaga cells were determined using a hemacytometer and then diluted to 6·8 × 104 cells ml−1 in fat H2O for inoculation into CDC biofilm reactors (BioSurface Technologies Corp., Bozeman, MT).

CDC biofilm reactors and sample collection

Six CDC biofilm reactors (http://www.biofilms.biz), half containing either 24 copper (Cu) or unplasticized polyvinylchloride (uPVC) coupons (surface area of 1·3 cm2.coupon−1), were fed tapwater at ambient temperature (20·8 ± 1·6°C) and with minimal chlorine residual (Free Cl2: 0·169 ± 0·175) (Buse et al. 2014a). The reactor ‘feed’ water was sampled daily 1 year prior to and for 4 months postmicrobial inoculation. Free and total chlorine (Cl2), pH, temperature, conductivity and per cent dissolved oxygen (DO) were previously reported (Buse et al. 2014a). The microbial community compositions of the Cincinnati drinking water have also been described (Gomez-Alvarez et al. 2012 and Buse et al. 2013). Each reactor consisted of a 1 l jacketed glass vessel containing approx. 400 ml of water and 24 coupons and an outlet overflow port connected to 3/8″ inner diameter Norprene food grade tubing (Cole-Parmer, Vernon Hills, IL). Drinking water was delivered via a 23-L holding tank to allow for dechlorination at a continuous flow rate of 40 ml h−1 using a peristaltic pump (L/S® Variable-Speed Digital Drive; Cole-Parmer), providing a hydraulic resident time of 10 h. The reactors stood on a magnetic stir plate set to turn on every 2 h for 30 min at approx. 100 rev min−1 to simulate periods of stagnation then shear flow. Drinking water biofilms were allowed to establish on the coupon surfaces for 1 year before Leg. pneumophila and A. polyphaga inoculation. Three days before the inoculation, the existing Norprene tubing (connected to the effluent port) was disconnected from the reactor and 60–90 cm of the tubing were cut for biofilm sample collection. A piece of tubing (approx. 13 cm−2, designated pretubing) was opened up latitudinally and biofilm (0·3–0·7 g wet weight biomass, 23–53 mg cm−2) was collected with a sterilized stainless steel scraper and placed into a preweighted bead beating tube with 0·1 mm silica beads (MP Biomedicals LLC, Solon, OH) and 300 μl T&C lysis buffer (Epicenter Biotechnologies, Madison, WI).

The six CDC reactors were disconnected from the water supply and inoculated as pairs with sterile fat H2O (designated post-uPVC-Con and post-Cu-Con: Negative controls for uPVC- and Cu-containing reactors), 106 CFU ml−1 of Leg. pneumophila (designated uPVC-Lp and Cu-Lp), or 106 CFU ml−1 of Leg. pneumophila and 500 cells ml−1 of A. polyphaga (designated uPVC-Lp/Ap and Cu-Lp/Ap). After the 24-h inoculation period, each reactor was connected back to the peristaltic pump (approx. 2·4 volume changes per day) with new sterile Norprene tubing effluent lines. After 140-day postinoculation, the Norprene biofilm samples (designated post-tubing) were collected, as described above. Hence, the pretubing and post-tubing samples were designated uPVC- (Cu-) Lp (Lp/Ap) according to its connection to either coupon material within the differently inoculated CDC biofilm reactors. For the detection of Leg. pneumophila and A. polyphaga, both the pretubing and post-tubing controls were the negative controls.

Biofilm DNA isolation, PCR and clone libraries

DNA was extracted in four tubes per biofilm tubing sample using MasterPure™ DNA Purification Kit following the manufacturer's instructions (Epicenter Biotechnologies). DNA extracts were further purified using the Genomic DNA Clean & Concentrator kit as recommended (Zymo Research Corp., Irvine, CA). Final DNA extracts were eluted in 100 μl molecular grade water and DNA concentrations estimated with a Nanodrop ND-1000 Spectrophotometer (NanoDrop Technologies, Inc., Wilmington, Delaware). This protocol was described, validated for various biofilm samples and demonstrated to remove PCR inhibition (Lu et al. 2013). To obtain genus level identification from approx. 800 bp sequences, Sanger method of DNA sequencing was used following the procedures briefly described as follows. Conventional PCR was performed using general bacterial and eukaryote-targeted primes (Table S1) to develop clone libraries of bacteria and eukaryotes. The PCR master mix reaction (20 μl final volume) contained 2 μl of DNA, 0·5 units of TaKaRa Ex Taq polymerase (TaKaRa Bio Inc, Otsu, Shiga, Japan), 1× Ex Taq buffer, 1·25 mmol l−1 each dNTP mixture and 2 μmol l−1 of each universal 16S rRNA gene PCR primers (27F and 786R) (Amann et al. 1995) and 18S rRNA gene PCR primers (Ami6F1 and Ami9R) (Thomas et al. 2006) (Table S1), which target all bacteria and eukaryota, respectively. Four PCR per CDC reactor tubing were performed as described in previous studies (Turner et al. 1999; Thomas et al. 2006) and the quality of PCR products were confirmed via agarose gel electrophoresis. The four replicate amplified products from each sample were pooled and cloned using a TOPO TA Cloning kit (Invitrogen, Carlsbad, CA) according to the manufacturer's recommendation. Two clone libraries, for 16S and 18S rRNA genes respectively, were produced for each duplicate reactor tubing sample by randomly picking approx. 200 colonies and sequencing each in both directions using ABI Big Dye sequencing chemistry (Applied Biosystems, Foster City, CA) on an ABI 3730 XL DNA analyzer (Cincinnati Children's Hospital, Cincinnati, OH).

Sequence analysis

Raw sequences were edited using Sequencher (Gene Codes Corp., Ann Arbor, MI). Chimeric sequences were identified using UCHIME (Edgar et al. 2011) and DECIPHER (Wright et al. 2011), and removed from further analyses. Rarefaction curves were used to determine whether the sampling effort was sufficient to represent the possible richness in the sample and represent the rate at which new sets of Operational Taxonomic Units (OTUs) were detected as the sample size increased (Morris et al. 2002). Rarefaction curves were calculated with the software mothur (Schloss et al. 2009). Prior to analysis, clone libraries were normalized by randomization to the smallest dataset (i.e. 99 and 98 clones for 16S and 18S rRNA gene sequences, respectively). Sequences were aligned and clustered with 97% sequence identity as the cut-off point for each Operational Taxonomic Unit (OTU) using the software mothur v1.25.1 (Schloss et al. 2009; http://www.mothur.org). Taxonomic identification of 16S rRNA gene sequences was performed using the Classifier tool (Ribosomal Database Project II release 10.28) (Cole et al. 2009) and NCBI/BLASTn. Phylogenetic trees were constructed from the alignments of 16S or 18S rRNA gene sequences based on the maximum likelihood method. The software mega v5.03 (Tamura et al. 2011) was used to build trees using 500 replicates to develop bootstrap confidence values. Normalized libraries were used to calculate the species richness (S), species richness estimators (ChaoI), and the Shannon's (H) diversity index, using the software mothur (Schloss et al. 2009). Nonmetric multidimensional scaling (nMDS) analysis based on the Bray-Curtis similarity coefficient of the transformed data (log[x + 1]) was used to describe the relationships among microbial communities based on the relative distribution of OTU groups. The nMDS ordination plots were generated using the software past v2.14 (Hammer et al. 2001). A two-way crossed analysis of similarities (anosim) was used to identify significant differences (P < 0·05) between biofilm community assemblages pre- and postinoculation and between upstream reactor coupon material (Cu vs uPVC) using R values (R test statistic), where near zero indicates a true null hypothesis of no difference between groups, whereas those greater than 0 (up to 1) indicate dissimilarities between groups (Clarke 1993). Sequences were compared with publicly available databases (BLASTn, NCBI) to identify the closest matching sequences.

Nucleotide sequence accession numbers

Representative tubing biofilm 16S and 18S rRNA gene sequences from cloning experiments were deposited in GenBank with accession numbers KF680665 - KF680742 and KF680743 - KF680771, respectively.

qPCR

For qPCR assays targeting Leg. pneumophila (Lp02), sidF, dotA and rtxA virulence genes were used (Faucher et al. 2011; and Lu et al. 2013), while assays targeting A. polyphaga and V. vermiformis were JDP1/JDP2 (Schroeder et al. 2001) and hv1227F/1728R (Kuiper et al. 2006), respectively. Standard curves were generated with the DNA isolated using the same method described above from serial dilutions of cultures in triplicate (102–108 cells ml−1). The four duplicate biofilm DNA (20–100 ng μl−1) and their 10-fold dilutions per tubing sample were set up on a 96-well plate in duplicate, and qPCR performed on an Applied Biosystems 7900 HT Fast Real-time PCR System (Applied Biosystems, Foster City, CA). Each reaction mixture (20 μl final volume) contained 1× Power SYBR Green PCR Master Mix, which is Built-In Hot Start, (Applied Biosystems) qPCR conditions consisted of an initial denaturation step of 10 min at 95°C, followed by 40 cycles of 15 s at 95°C, for all three PCR according to the manufacturer's instruction. The annealing step varied for each PCR with 30s at 63, 56, or 58°C for Lp02, V. vermiformis, and A. polyphaga assays, respectively. Each assay was completed with extensions cycles for 30 s at 72°C with a final hold at 72°C for 5 min. The dissociation curves were analysed to confirm the amplification of a single product. Reactions showing multiple amplification products were not used in further analysis. The cycle thresholds (Ct) were calculated using the Sequence Detection Systems software v 2.3 (Life Technologies). All qPCR assays were performed in 96-well plates containing DNA standards and no-template controls. The qPCR efficiency was calculated using the following equation: Efficiency % = 100 × (10(−1/slope) – 1) (Applied Biosystems). To evaluate DNA inhibition, each DNA extract was tested using TaqMan Exogenous Internal Positive Control (IPC) Reagents (a VIC-labelled probe) manufactured by ABI (Applied Biosystems), in addition of the qPCR against10-fold dilution of DNA. The data were expressed as cell equivalent (CE), based on densities of the culture (colony forming units), which were used to generate standard curves, against the wet weight of tubing biofilm (CE g−1 cm−2).

Statistical analysis for qPCR data

The qPCR data were compared between pre- and postinoculation for uPVC- or Cu-reactor tubing biofilms either uninoculated or inoculated with Lp or Lp/Ap. Two-way comparisons were undertaken with a General Linear Model (GLM, sas Systems ver. 9.2, SAS, Cary, NC) with overall F-test and T-test assuming samples had equal variance, and the response variable in the analysis was the qPCR Ct value.

Results

qPCR of Legionella pneumophila

The three qPCR assays for Leg. pneumophila (targeting three different virulence genes, sidF/dotA/rtxA) had similar amplification efficiencies, were within an effective range (91/94/93% respectively) and showed minimal PCR inhibition, that is absence of change for 10-fold dilutions of DNA when tested using IPC (PF-test = 0·429) for postinoculation tubing biofilm DNA. Each qPCR assay yielded the same result: that is Leg. pneumophila was only detected in postinoculated, rather than in pre-inoculated and the negative controls, tubing biofilm samples. Based on Ct, the quantity of Leg. pneumophila in the uPVC-Lp/Ap reactor tubing biofilms was significantly lower than in the Cu-Lp and Cu-Lp/Ap reactor tubing biofilms. However, there was no significant difference between the Cu-Lp and Cu-Lp/Ap reactor tubing biofilms (Table 1).

Table 1. Quantification of Legionella pneumophila and Vermamoeba vermiformis (cell equivalents [CE] g−1 cm−2 by qPCR) in PVC- and Cu-CDC reactor tubing biofilmsa
 PVC-controlPVC-LpPVC-Lp/ApCu-controlCu-LpCu-Lp/ApP (T-test)
TargetMean (StDev)Mean (StDev)Mean (StDev)Mean (StDev)Mean (StDev)Mean (StDev)PVC vs Cu
  1. a

    From 4-mo old Norprene tubing biofilms downstream of CDC reactors spiked with Leg. pneumophila (Lp), with Leg. pneumophila and Acanthamoeba polyphaga (Lp/Ap) or not spiked (control); no Leg. pneumophila was detected in the pretubing biofilms.

  2. b

    Below detection limits, which were approx. 100 CE g−1 cm−2 for the three assays, but commonly detection limits were below 100, typically down to 10 CE g−1 cm−2.

  3. c

    Quantity was obtained from quantifiable qPCR, which was consistent in three replicates for each tubing, although the quantities were below detection limits.

Leg. pneumophila
sidF BDLbBDL45 (18)cBDL235 (58)147 (32)<0·0001
dotA BDLBDL55 (32)cBDL233 (137)207 (221)<0·0001
rtxA BDLBDL12 (5)cBDL108 (4)98 (7)c<0·0001
V. vermiformis 1496 (330)863 (50)795 (61)243 (39)23 (1)c437 (217)<0·0001

Bacterial tubing biofilm community

Rarefaction curves for Bacteria domain communities did not plateau, suggesting that further sampling would have revealed additional OTUs (Figure S1). However, when rare members (i.e. singletons) were excluded from the rarefaction analysis, the rarefaction curves did plateau, indicating that the most predominant bacterial groups were likely identified (Figure S1). Rarefaction curves for eukaryotic communities did not rise as the sample size increased (Figure S1).

There were 16 (of 125) major bacterial OTU groups (abundance > 1% at 97% similarity) identified in the reactors' tubing biofilms from twelve 16S rRNA gene libraries (N = 2 replicate × 6 reactor tubing biofilms) representing the total biofilm communities (Table 2). The genus Mycobacterium represented 38% of the total abundance and dominated in all biofilm samples (Table 2). Of the mycobacterial OTU's, some exhibits high similarity to sequences obtained from Mycobacterium bolletii, Mycobacterium chelonae and Mycobacterium abscessus (99% sequence identity), while the others were similar to Myco. abscessus (identity: 84–96%), Mycobacterium immunogenum (identity: 91%), Mycobacterium sp. (identity: 85–96%) (Fig. 1). The α-Proteobacteria OTU groups accounted for 10% and were mostly associated with known amoeba-resisting bacteria (ARB) (Pagnier et al. 2008) (Table 2, Fig. 1). The other major group (>8·5% of total sequences) was related to Bacteroidetes (Table 2 and Fig. 1). There was only one clone sequence similar to Legionella sp. in pretubing, but no Leg. pneumophila was identified, which was consistent with the qPCR assays targeting the three virulence genes. Nonetheless, there were significant differences (two-way anosim, P = 0·01, Table S2) in community structures between pre- and postinoculation periods (global R of dissimilarity: 0·759) and between uPVC- and Cu-reactor tubing biofilms (global R: 0.889). There was no dissimilarity between tubing biofilms downstream of inoculation treatments (control, Lp and Lp/Ap, Fig. 2). Compared with post-tubing biofilms, the sequences similar to unidentified bacteria (OTU003), Sphingopyxis and Aquabacterium were more abundant, while Myxococcales, Phenylobacterium, Flavobacterium and Rhdobacteraceae were less abundant in pretubing biofilms (Fig. 2 in coordinate 2, Table 2, Table S3). Compared with Cu-reactor tubing biofilms, uPVC-reactor tubing biofilms displayed more diversity (Shannon indices: 3 vs 2), and higher abundance. More than half of the dominant bacterial OTU (abundance > %1) groups were more abundant in the PVC- than Cu-reactor tubing biofilms (Table 2). However, some OTUs such as Mycobacterium spp. (21% vs 54%) and Rhodobacteraceae (0·3% vs 6·1%) were significantly less abundant (P ≤ 0·016) in the PVC than Cu-reactor tubing biofilms (Fig. 2 in coordinate 1, Table 2, Table S3).

Table 2. Bacterial distribution (%) and classification of OTU (at 97% cut-off level) for those with abundance >1% in the total community
OTULowest common ancestorPVCCopperTotal (%)P (t-test: PVC vs Cu)
PrePostPrePost
TaxonomyLevelControlLpLp/ApControlLpLp/ApControlLpLp/ApControlLpLp/Ap
OTU001 Mycobacterium Genus19·212·137·410·15·142·453·573·758·638·439·461·6 37·6 0·003
OTU002α-proteobacteriaClass1·02·01·07·11·02·022·26·19·1001·0 4·4 0·291
OTU003BacteriaDomain4·019·24·01·0006·19·111·1000 4·5 0·932
OTU004BacteroidetesPhylum9·113·116·29·17·10000000 4·5 0·002
OTU005 Bradyrhizobium Genus03·04·09·11·09·12·0001·016·210·1 4·6 0·874
OTU006 Sphingomonas Genus18·29·18·13·03·0001·001·01·00 3·7 0·037
OTU007 Phenylobacterium Genus01·01·01·014·115·2000000 2·7 0·087
OTU008RhodobacteraceaeFamily0002·00003·03·014·112·14·0 3·2 0·035
OTU009 Sediminibacterium Genus1·0003·05·103·02·01·0010·10 2·1 0·527
OTU010 Ohtaekwangia Genus5·13·006·13·08·1000002·0 2·3 0·009
OTU011 Aquabacterium Genus8·14·08·12·000002·01·01·00 2·2 0·049
OTU012BacteriaDomain02·01·012·14·0001·00002·0 1·9 0·192
OTU013BacteroidetesPhylum6·11·01·03·09·11·0000000 1·8 0·028
OTU014 Flavobacterium Genus02·00001·000018·24·01·0 2·2 0·282
OTU016 Aquabacterium Genus2·02·0001·06·1001·02·003·0 1·4 0·441
OTU018BurkholderialesFamily1·04·002·03·01·01·0001·000 1·1 0·040
Total (%)   74·7 77·8 81·8 70·7 56·6 85·9 87·9 96·0 85·9 76·8 83·8 84·8 80·2  
Figure 1.

Bacterial phylogenetic tree constructed from the alignments of 16S rRNA clones representing >75% of the total variation. The software mega v5.03 was used to build trees based on the maximum likelihood method using 500 replicates to develop bootstrap confidence values.

Figure 2.

Nonmetric multidimensional scaling (nMDS) ordination plot of 16S bacterial clone libraries based on Bray-Curtis dissimilarity using the software past v2.14 (Hammer et al. 2001). A two-way crossed analysis of similarities (anosim) identified significant differences (< 0·05) in the biofilm community assemblages of pre- and postinoculation and between coupon material (Cu vs PVC) (Table S2). Contribution of OTUs that explained ≈62% (SIMPER analysis) of the dissimilarity within samples is listed in Table S3. (image_n/jam12578-gra-0001.png) Pre; (▼) Post.

Micro-eukaryotic community in tubing biofilms

There were 29 OTU groups identified from twelve 18S rRNA gene clone libraries (n = 1176 sequences, 840 bp average length). Most could be identified as free-living protozoa (FLP) and fungi. Most fungal sequences were similar to Paecilomyces nostocoides (Identity: 95–99%). The FLP-like sequences were divided into six groups (Group 1–3 and 5–7) when using alignments of 18S rRNA clones representing >95% of the total variation (Fig. 3). Group 1 was similar to previously isolated strain eukaryote sp. CRIB-09 (GenBank accession no. DQ123626) described as an unidentified protist (Thomas et al. 2006), originally isolated via amoebal enrichment from one tap water swab. The Group 2 was similar (identity 90–95%) to the sequences present in a detritus bacteria–bacterivorous protozoan culture system (van Hannen et al. 1999) and a FLA-Legionella culture (Valster et al. 2010), and this FLA-like eukaryote (OTU21) predominated in all libraries (Table 3). Group 3 consisted of V. vermiformis-like sequences (identity: 93–99%) as the second dominant group and its overall average abundance was about 32%. The last three groups were Spumella-like flagellate sequences (ID -99%) and flagellate Neobodo designis-like sequences (ID 87- 94%) (Group 5–6 and 7, respectively). Similar to bacteria, there were differences in community structures between pre- and post-tubing biofilms and between uPVC- and Cu-reactor tubing biofilms (Fig. 3). Protist-, Spumella and Vermamoeba-like sequences appeared to be more abundant in the downstream uPVC- (39%) than Cu- (24%) reactor tubing biofilms (Table 3, PT-test = 0·012, 0·014 and 0·060, respectively), suggesting the possible upstream impact of copper. The FLA-like eukaryote domain EUK-OTU21 seemed to be a very stable population as there was no significant difference between pre- and postinoculation, nor for uPVC- and Cu-reactor tubing biofilms. However, the quantity of V. vermiformis measured using qPCR clearly displayed significantly more biomass from uPVC- than Cu-reactor tubing biofilms (Table 1), indicating that copper negatively impacted V. vermiformis abundance.

Table 3. Eukaryotic distribution (%) and classification of OTU (at 97% cut-off level) for those with abundance >1% in the total community
OTULowest common ancestorPVCCopperTotal (%)P (t-test: PVC vs Cu)
PrePostPrePost
TaxonomyLevelControlLpLp/ApControlLpLp/ApControlLpLp/ApControlLpLp/Ap
  1. ND, not determined.

OTU01UnclassifiedND7·104·12·05·13·102·00000 2·0 0·012
OTU02UnclassifiedND00000010·20024·500 2·9 0·188
OTU05 Spumella Genus21·417·321·44·1025·507·102·002·0 8·4 0·014
OTU14SordariomycetesClass000006·133·71·051·01·0012·2 8·8 0·107
OTU21UnclassifiedND13·356·119·449·058·29·221·463·328·636·786·757·1 41·6 0·304
OTU26 Hartmannella Genus51·021·454·128·625·552·033·723·520·429·612·225·5 31·5 0·060
OTU32UnclassifiedND00014·32·00000000 1·4 0·272
Total (%)   92·9 94·9 99·0 98·0 90·8 95·9 99·0 96·9 100·0 93·9 99·0 96·9 96·4  
Figure 3.

Eukaryotic phylogenetic tree constructed from the alignments of 18S rRNA clones representing >95% of the total variation. The software mega v5.03 was used to build trees based on the maximum likelihood method using 500 replicates to develop bootstrap confidence values.

Discussion

Downstream biofilm seeding and communities

The effluent tubing was used to simulate downstream drinking water premise-plumbing connections. While biofilms grown on Norprene tubing have not previously been described, it is widely used in association with drinking waters from bioreactors, water tanks and fixtures according to the manufacturer. The microbial biofilm community compositions from this study are similar to those previously described via different methods (Buse et al. 2014a) and have also been largely reported for other drinking water biofilms (Yu et al. 2010; Gomez-Alvarez et al. 2012; Buse et al. 2013; Wang et al. 2013), although the community structures varied. For example, Acetobacteraceae, Bacteroidetes, Burkholderiales, Mycobacterium, Proteobacteria, Rhodobacter, Sediminibacterium, Sphingopyxis, Chrysophyceae and Vermamoeba-like sequences presented as the major groups (abundance > 1%) both in bioreactor coupon biofilms (Buse et al. 2014b) and in current study tubing biofilms. This suggested that the downstream biofilms were seeded from upstream shedding or drinking water, rather than being specifically associated to the material of the tubing.

Unlike the upstream coupon biofilm, the downstream tubing biofilm microbial communities were dominated by ARB and free-living protozoa, especially, FLA. The upstream minimal residual chlorine residual and periodically stagnant and slow water flow would likely encourage thicker biofilms, although tubing material (food grade plastic with smooth surface) could have played some role. According to Koh et al. (2013), thicker biofilm supported more multiplication of eukaryotic cells. Moreover, those biofilms contained both prokaryotic and eukaryotic cells often formed extensive dense and thick mature biofilms as reported by Singleton et al. (1997) and Koh et al. (2013). In addition to the synergy/cooperation or antagonism interactions in the tubing biofilm (Simões et al. 2007), the bacteria-protozoan interactions (prey-bacteria, predator, antipredatory bacteria) among microbial populations could be an important factor for such microbial community structures. Some of the bacterial groups adhered to the tubing surface may become prey, although some might have developed mechanisms to resist predation. Of the predators identified in the current study, FLA (Vermamoeba and an unclassified eukaryotic clone OTU21) and flagellates (Spumella-like flagellate) dominated all biofilm communities examined, and were similar to bacteriovorous protozoa previously reported (van Hannen et al. 1999; Boenigk et al. 2005). For example, Spumella elongata-like flagellates (Chrysophyceae) were reported to be among the dominant bacterivorous flagellates (Boenigk et al. 2005). Of particular interest to Leg. pneumophila and other ARB growth in tubing biofilms were FLP that were not predominant in the upstream coupons with less biofilm growth (Buse et al. 2014b) than the Norprene tubing. Possible ARB identified in the current study were those closely related to Mycobacterium and α-Proteobacteria, as members of both groups have been recovered from bacteria-amoeba co-culture of drinking water samples (Bradyrhizobium, Burkholderiales, Rhodobacter and Sphingomonas (Corsaro et al. 2010; Pagnier et al. 2008; Thomas et al. 2006, 2008)). Various nontuberculous mycobacteria (NTM) are known to effectively adhere to different pipe materials (Mullis and Falkinham 2013), so NTM continuously released from the upstream CDC reactor biofilms (Buse et al. 2014b) may have preferentially attached onto downstream tubing, so accounting for their dominance (about one third of bacteria identified).

Upstream impact on downstream

Unlike previous studies that focused on direct effects on biofilm from different pipe materials (Lehtola et al. 2005; Doğruöz et al. 2009; Yu et al. 2010; Jang et al. 2011 and Morvay et al. 2011), the different OTU microbial communities observed appear to have resulted from indirect and/or direct effects of free copper. Indirect copper effects may have resulted from Cu-coupon reactor biofilms and/or members detaching and establishing on the downstream tubing. Direct effects may have resulted from Cu ions/oxides directly impacting downstream biofilm development on the Norprene tubing (e.g. Lu et al. 2013). Given the general presence of Mycobacterium spp. and α-Proteobacteria with or without upstream Cu material, these members may be more resistant to Cu-impacts than other bacteria (discussed further below) or may be associated with FLP's resistance to copper, an effect warranting further investigation.

We are unaware of previous work on the impact of upstream pipe material on the downstream microbial structure of drinking water biofilms, although the potential for upstream members to influence downstream biofilms has been suggested for drinking water distribution systems (Pinto et al. 2012). In contrast, there are numerous reports on the direct effect of copper repressing biofilm growth, and even microbial diversity. For example, Yu et al. (2010) and Morvay et al. (2011) noted fewer cells within drinking water biofilms on copper coupons than with uPVC or other materials, and Lehtola et al. (2005) demonstrated faster biofilm growth on polyethylene than on copper pipes. Our previous work (Buse et al. 2014b) and others (Yu et al. 2010; Jang et al. 2011; Wang et al. 2014) described greater microbial diversity on plastic materials than from biofilms formed on metals. Lastly, FLA have also been recovered from copper and galvanized steel coupons exposed to drinking water, but not culturable Leg. pneumophila (Doğruöz et al. 2009) unless known to have been recently introduced. Most intriguing is the suggestion of Buse et al. (2013) that copper pipe may enhance the development of viable but nonculturable (VBNC) yet infectious Leg. pneumophila cells. In particular, upstream-spiked Leg. pneumophila was detected from the downstream tubing biofilm 4 months after spiking, and it was only detected in the uPVC-Lp/Ap, but not uPVC-Lp reactor tubing biofilms, suggesting A. polyphaga might have promoted its persistence in the bioreactors (Buse et al. 2013) and colonization of the down-stream tubing biofilms (current study). This result was based on qPCR assays, which have been demonstrated to be more sensitive than direct-culturing approaches (Lee et al. 2011; Buse et al. 2013). The level of approx. 103 equivalent cells ml−1 or g−1 was not detectible by our culture of Leg. pneumophila. The unexpected persistence of Leg. pneumophila in biofilms developing downstream of Cu-coupon reactors postinoculation (Table 1) is particularly noteworthy, given the ubiquitous use of copper piping for premise drinking water distribution (hot and cold water). The relevant eukaryotic hosts for Legionella and similar pathogens within drinking water is unclear, and may not necessarily be directly related to the most predominant cell types present, but influenced by them. Hence, insights from upstream processes that may select from hosts/pathogen pairs and seed downstream sites in distribution systems could be as important as sites for downstream amplification of environmental pathogens (Wang et al. 2013). There is stronger evidence of intracellular growth of Leg. pneumophila than the very much broader group of mycobacteria, for which some members may grow intracellularly, but not freely in the environment. There is actually no clear data describing the proportion of either group that may grow freely in the biofilm vs intracellularly in that niche.

Risk of the upstream impact on downstream

The major finding of the current work is the dominance of FLA, especially V. vermiformis, Mycobacterium and the colonization of event-introduced Leg. pneumophila in downstream biofilms, and their possible impact on public health. The latter clearly has potential public health implications. The human health risks from FLA are described by Kinnear (2003), who provided evidence that V. vermiformis and Acanthamoeba castellanii may produce a cytopathic effect on keratocytes in vitro, and harbour water-based environmental pathogens within the genera Mycobacterium and Legionella (Berry et al. 2006; Valster et al. 2009; Garcia et al. 2013). ARB such as members of Legionella, Mycobacterium and other pathogenic bacterial genera would appear to be more hazardous in low chlorine residue (Wang et al. 2014), slow-flow conditions that benefit biofilm growth and more FLA due to the greater bacterial productivity. Hence, while there were significant changes in relative abundance for the major members, FLA that are known to serve as hosts for Leg. pneumophila and Mycobacterium spp. in aquatic samples (Rowbotham 1980, 1986) appeared to have been selected. Although many different FLA genera may (under laboratory conditions) support intracellular growth of drinking water pathogens such as Legionella (Lau and Ashbolt 2009), V. vermiformis is frequently isolated from engineered water systems (Valster et al. 2009), and documented as a host for numerous facultative bacterial pathogens (Greub and Raoult 2003, 2004). Further, of the various ARB pathogens described (Pagnier et al. 2008), we identified very similar sequence representatives of Mycobacterium and α-Proteobacteria from the tubing microbial community that may also have amplified in the tubing FLA. It is well known that various NTM can cause severe lung diseases (Falkinham 2011) and that they result in the major hospitalization health burden costs associated with drinking water (Collier et al. 2012). Some of the NTM-like OTUs identified in the current study may contain human pathogenic strains, including, Myco. bolletii (Koh et al. 2009) and Myco. chelonae (Simmon et al. 2011). The other identified NTM-like OTUs require further study, because Mycobacterium chelonae-abscessus complex alone or accompanied with Myco. immunogenum, Mycobacterium massiliense, Myco. bolletii, and Mycobacterium salmoniphilum can cause invasive infections in immunocompetent and immunocompromised hosts, as seen in the North-eastern United States (Simmon et al. 2011). Of particular importance to Leg. pneumophila occurrence, our results indicate increased potential health risk was associated with low chlorine residual and fluctuating stagnant conditions for the biofilm communities that developed downstream of Cu-containing reactors. This would infer that the upstream microbiome and/or minerals released from copper must have a strong influence on downstream biofilm development on non-Cu, non-uPVC materials that supported Leg. pneumophila. We are unaware of any specific association between copper pipe materials and Leg. pneumophila outbreaks associated with drinking water (Craun et al. 2010). On a more positive note, this upstream ability to manipulate downstream microbiomes may hold promise for potential novel ARB control approaches as generally promoted by Wang et al. (2013).

Acknowledgements & disclaimer

This research was supported by the Pathfinder Innovation Project – Funded by the Office of Research and Development, United States Environmental Protection Agency (U.S. EPA). The views expressed in this article are those of the author(s) and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.

Conflict of interest

No conflict of interest declared.

Ancillary