Pyrosequencing Demonstrated Complex Microbial Communities in a Membrane

M&E Papers in Press. Published online on March 18, 2011 doi:10.1264/jsme2.ME10205

2 3 Revised ME10205 4

6 Pyrosequencing Demonstrated Complex Microbial Communities in a Membrane

7 Filtration System for a Drinking Water Treatment Plant

1 1 1 1 9 SOONDONG KWON , EUNJEONG MOON , TAEK-SEUNG KIM , SEUNGKWAN HONG , and HEE-

1* 10 DEUNG PARK

11 12 1School of Civil, Environmental and Architectural Engineering,Proofs Korea University, Anam- 13 Dong, Seongbuk-Gu, Seoul 136-713, South Korea

15 (Received December 1, 2010 － Accepted ViewFebruary 17, 2011)

17 * Corresponding author. 18 E-mail: [email protected];Advance Tel: +82-2-3290-4861; Fax: +82-2-928-7656. 19

20 Running headline: Microbial Communities in a Filtration System

2 Microbial community composition in a pilot-scale microfiltration plant for drinking water

3 treatment was investigated using high-throughput pyrosequencing technology. Sequences of

4 16S rRNA gene fragments were recovered from raw water, membrane tank particulate matter,

5 and membrane biofilm, and used for taxonomic assignments, estimations of diversity, and the

6 identification of potential pathogens. Greater bacterial diversity was observed in each sample

7 (1,133 – 1,731 operational taxonomic units) than studies using conventional methods,

8 primarily due to the large number (8,164 – 22,275) of sequences available for analysis and

9 the identification of rare species. Betaproteobacteria predominated in the raw water (61.1%),

10 while Alphaproteobacteria were predominant in the membrane tank particulate matter 11 (42.4%) and membrane biofilm (32.8%). The bacterial communityProofs structure clearly differed 12 for each sample at both the genus and species levels, suggesting that different environmental 13 and growth conditions were generated duringView membrane filtration. Moreover, signatures of 14 potential pathogens including Legionella, Pseudomonas, Aeromonas, and Chromobacterium

15 were identified, and the proportions of Legionella and Chromobacterium were elevated in the

16 membrane tank particulate matter, suggesting a potential threat to drinking water treated by 17 membraneAdvance filtration. 18

19 Key words: membrane, drinking water, pathogen, pyrosequencing, microbial community

2 Drinking water contaminated by pathogenic protozoa, bacteria, and viruses can cause

3 diseases (14, 15, 25, 26). Statistics indicate that 126 drinking water-related disease outbreaks,

4 429,000 cases of illness, 653 hospitalizations, and 58 deaths occurred in the United States

5 during the years 1991 - 1998 (4).

7 Sand filtration and disinfection are commonly used to purify drinking water. Sand filtration

8 removes particulate matter including microorganisms at the surface or in the middle of the

9 sand bed. Direct collisions, van der Waals force, surface charge attraction, and diffusion are

10 known to be involved in the capture of particulate matter by sand filters (12). Generally, the 11 filtration process is affected by several operational parameters (e.g.Proofs, linear velocity, backwash 12 rate, etc.) (12) and design conditions (e.g., grains size, depth of sand bed, etc.) (7). Current 13 increased water quality requirements make Viewit more difficult to design and utilize sand filters 14 for drinking water treatment (26).

16 Microfiltration (MF) or ultrafiltration (UF) with membranes is an attractive alternative to 17 sand filtrationAdvance for drinking water production mainly due to an excellent ability to remove 18 microorganisms as well as suspended solids and colloids, without the need for high

19 concentrations of disinfectants. Because the membranes used to purify drinking water have

20 pores that are smaller (typically 0.04 – 0.2 m) than microorganisms (typically 0.5 – 5.0 m),

21 microorganisms are effectively rejected through a sieving mechanism (30), although some

22 microorganisms (e.g., ultramicrobacteria (28)) can pass through the membranes. However,

23 defects on a membrane’s surface can decrease sieving efficiency, allow pathogens to pass

24 through the membrane, and affect public health, and it is important to test the integrity of

2 conducted by counting particles in filtered water and/or checking pressure-induced decay by

3 applying high pressure to the membranes. In addition, it is also important to know which

4 pathogens can persist in membrane systems to prepare for a possible entering of pathogens

5 into the public water supply. Pathogens are usually detected by culture and colony counting

6 methods (13), microscopic observation (35), and PCR (13).

8 In submersed membrane filtration operated in a dead-end mode for drinking water treatment,

9 membrane modules are installed in a tank (the membrane tank), and particulate matter is

10 concentrated in the membrane tank by filtration through the membrane. In addition, aeration 11 is frequently applied from the bottom of the membrane tank to minimizeProofs the accumulation of 12 foulants on the surface of the membrane during backwash and/or filtration periods (5, 29). 13 This operation can result in the concentrationView of particulate organic matter while oxygen is 14 dissolved in the tank, resulting in an environment suitable for the growth of aerobic

15 microorganisms, although biocidal treatment affects the growth of microorganisms in the

16 membrane system. The membrane tank therefore can behave like a bioreactor that facilitates 17 the growthAdvance of diverse microorganisms, including some pathogens. If the membrane tank 18 promotes the growth of pathogenic microorganisms and the membrane surface has some

19 defects, it would result in a potential threat to people who consume the water.

21 Although the potential exists for the growth of microorganisms in a membrane tank, such

22 growth has not been well reported or characterized. Thus, the objectives of this study were 1)

23 to investigate bacterial community composition and diversity and 2) to identify potential

24 pathogenic bacteria in membrane tanks. To this end, we collected biomass samples from a

2 system, and characterized bacterial 16S rRNA gene sequences using a high-throughput

3 pyrosequencing technique, and then analyzed the sequences using bioinformatics tools.

5 Materials and Methods

6 The pilot plant and its operating conditions

7 A pilot-scale drinking water treatment plant was set up at the Kuei municipal drinking water

3 -1 8 treatment plant (Seoul, South Korea) which produces 650,000 m d and provides treated

9 water to residents of northern Seoul. The Kuei treatment plant takes water from the Paldang

10 reservoir located in the upper parts of the Han River and treats the water by alum coagulation, 11 flocculation, sedimentation, sand filtration, and chlorination. AsProofs shown in Fig. 1, the pilot 3 12 plant primarily consisted of a membrane tank (working volume = 1.2 m ) and a produced

3 3 13 water tank (0.1 m ). Raw water was pumpedView to the membrane tank at a rate of 5.1 m /h. The 14 water was filtered through submersed hollow-fiber membranes by a suction pump at a rate of

3 -1 15 4.8 m h , and the filtered water was delivered to the produced water tank. At the bottom of

3 -1 16 the membrane tank, the concentrate was withdrawn at a rate of 0.3 m h . The pilot plant was

-2 -1 17 operated Advanceat a flux of 60 LMH (L m h ). A daily total of 91 cycles of suction (15 min), 18 backwashing (0.5 min), and relaxation (0.17 min) were processed, followed by one cycle of

19 maintenance cleaning (2 min) and relaxation (15 min). During backwash, treated water from

20 the produced water tank was back-flowed across the membrane to the membrane tank at a

3 -1 21 rate of 7.2 m h , without addition of a chemical. Maintenance cleaning was practiced similar

-1 22 to the backwash operation except for providing 15 mg L of NaOCl. A total of four

® 23 horizontal-type membrane modules were installed in the membrane tank (Cleanfil -S20H,

24 KOLON Industry, South Korea), and air was continuously supplied at the bottom of the

2 module consisted of polyvinylidene fluoride hollow-fiber membranes with a nominal pore

2 3 size of 0.07 m. The overall membrane surface area of the four modules was 80 m . The pilot

4 plant had been operated since April, 2006 and had consistently produced water with low

5 levels of turbidity (0.04  0.01 Nephelometric Turbidity Units (NTU)) irrespective of highly

6 variable raw water quality (turbidity = 10.4  15.9 NTU).

8 Sampling, DNA extraction, and PCR amplification

9 After the pilot plant was operated for 30 months, during August 2009, particulate matter was

10 collected from the membrane tank using a bucket-type sampler. An attached biofilm sample 11 was also harvested by scraping the biofilm formed on the middleProofs of hollow-fiber membranes, 12 using a sterilized spatula after lifting the membrane modules. For a comparison of bacterial 13 communities, raw water was also sampled Viewfrom the source water equalization basin using a 14 bucket-type sampler. Samples were immediately stored in an ice box before being transported

15 to the laboratory. A total of 20 L of raw water and 2 L of the membrane tank water were

16 passed through a membrane filter (0.2 m) to collect particulate matter. It took several hours 17 to filter theAdvance samples. Immediately after the filtration, the particulate matter was used for the 18 extraction of total DNA. Total DNA was extracted in duplicate using the UltraClean soil

19 extraction kit (Mobio Laboratory, Solana Beach, USA) following the manufacturer’s protocol.

20 For membrane biofilm samples (~ 0.2 g), the same kit was used for extraction of total DNA.

21 For the amplification of bacterial 16S rRNA gene fragments, the PCR primers 27F (5'-

22 AGAGTTTGATCMTGGCTCAG-3’) (8) and 518R (5'-ATTACCGCGGCTGCTGG-3’) (18)

23 were used. RDP’s Probe Match utility (http://rdp.cme.msu.edu) demonstrated that 65.1% and

24 87.9% of sequences within domain Bacteria matched with the 27F and 518R primers,

2 South Korea), 2.5 units of EF-Taq polymerase (Solgent), 0.2 mM dNTP mix, 0.1 M of each

3 primer and 100 ng of template DNA. The PCR profile was as follows: 95°C for 10 min; 35

4 cycles at 94°C for 45 s, 55°C for 1 min and 72°C for 1 min, with a final extension at 72°C for

5 10 min. The duplicate PCR products were pooled and purified using the QIAquick gel

6 extraction kit (Qiagen, Hilden, Germany). The purified products were used for

7 pyrosequencing.

9 Pyrosequencing

10 The ends of the purified PCR products (~ 1 g) were blunted, and short adaptors (14-bp long) 11 were ligated onto both ends for sorting sequences by key as wellProofs as for providing a priming 12 region. The modified products to be sequenced were attached to DNA capture beads, one 13 fragment per bead, and amplified using emulsiViewon-based clonal amplification. The beads were 14 set into the wells of a PicoTiterPlate device (1 of 8 lanes), with appropriate chemicals and

15 enzymes, and inserted into the Genome Sequencer FLX Titanium Series (454 Life Science,

16 Branford, USA) for pyrosequencing. All of the procedures followed the manufacturer’s 17 directionsAdvance (454 Life Science) and were conducted at Macrogen (Seoul, South Korea). 18

19 Data analysis

20 Initially, trimBarcode.pl Perl script (Macrogen) was used to sort the raw 16S rRNA gene

21 sequences obtained from pyrosequencing by key (i.e., sequences from the raw water,

22 membrane tank particulate matter, and membrane biofilm), to discard low quality and short

23 (< 250-bp long) sequences, and to trim the primer sequences. In addition, the sequences were

24 further refined by removing potential chimeric sequences with the Mothur utility (24). The

2 done using the Classifier (32) provided by the RDP. At a 3% cutoff, richness and diversity

3 indices (i.e., observed OTUs, Chao1 estimator, Shannon index, and ACE), rarefaction curves,

4 rank abundance curves, and Venn diagrams were obtained using the Mothur utility (24). The

5 input file of the Mothur utility was generated by aligning the processed sequences and

6 constructing a distance matrix using the RDP’s Pyrosequencing Aligner and RDP’s Column

7 Formatted Distance Matrix, respectively, according to the developer’s introduction.

8 Differences in bacterial community composition were evaluated by a proportional test (31)

9 using Minitab software (http://www.mintab.com).

10 11 Phylogenetic and statistical analyses Proofs 12 Legionella and Chromobacterium 16S rRNA gene sequences were retrieved from the total 13 sequences obtained in this study using RDP’sView Classifier, while their related sequences were 14 retrieved from the NCBI database (http://www.ncbi.nlm.nih.gov/) using the BLAST utility.

15 Then, the retrieved sequences were multiply aligned using ClustalX version 1.83 (27). A

16 neighbor-joining tree (22) was generated and a bootstrap analysis was performed using the 17 same softwareAdvance with 1000 resampling trials. The calculated tree was drawn using TreeView 18 version 1.6.6 (20).

20 Sequence deposition

21 The 16S rRNA gene sequences determined in this study have been deposited in GenBank

22 under accession numbers GU738023 to GU784790.

24 Results and Discussion

2 In order to investigate bacterial community composition, 16S rRNA gene sequences were

3 retrieved using pyrosequencing from three samples: water in the raw water equalization basin

4 (raw water), particulate matter in the membrane tank (membrane tank particulate matter), and

5 biomass on the surface of membranes in the membrane tank (membrane biofilm). A total of

6 22,275 (average length = 418.9 bp), 13,702 (428.6 bp), and 8,164 bacterial 16S rRNA gene

7 sequences (average length = 426.5 bp) were obtained from the raw water, membrane tank

8 particulate matter, and membrane biofilm, respectively, and used for the community analyses.

10 Sequences were initially assigned to corresponding bacterial taxa using the RDP classifier 11 with an 80% cutoff value, which identified 8, 12, and 12 phyla Proofsfor the raw water, membrane 12 tank particulate matter, and membrane biofilm, respectively. Community composition by 13 phylum is summarized for each sample in Fig.View 2 (A). All three samples had a similar bacterial 14 community composition at the phylum level, with a preponderance of Proteobacteria (62.9 –

15 83.3% of each sample’s total sequences) and Bacteroidetes (9.0 – 19.4%). Members within

16 Nitrospira, Acidobacteria, OD1, Actinobacteria, Verrucomicrobia, Cyanobacteria, 17 Chloroflexi,Advance TM7, Firmicutes, Plactomycetes, and Gemmatimonadetes were also found, but 18 the percentage of these sequences were relatively low (< 2.9 %). It should be noted that

19 significant fractions of sequences (0.2 – 19.0 %) were assigned to the unclassified phylum.

20 The sequences were unlikely to be generated from non-specific PCRs or new organisms due

21 to the high identity with sequences in the NCBI database (data not shown). Rather, the

22 accumulation of 16S rRNA gene sequences in the RDP database is outpacing our ability to

23 classify them as Sanapareddy et al. inferred (23).

2 (P < 0.001). Fig. 2 (B) shows a classification by class within Proteobacteria. The raw water

3 was predominated by Betaproteobacteria (61.1% of total sequences), while both the

4 membrane tank particulate matter (42.4%) and membrane biofilm were predominated by

5 Alphaproteobacteria (32.8%). The differences among samples were more evident at the

6 genus level (P < 0.001). Table 1 shows the 20 genera found frequently in each sample. In the

7 raw water, 48.0% of sequences were attributed to Curvibacter, which was not significant in

8 the membrane tank particulate matter (0.14%) or membrane biofilm (0.12%). Flavobacterium,

9 Pseudomonas, Porphyrobacter, and other minor members followed Curvibacter in

10 prevalence. No sequences constituting more than 10% of all sequences were found in either 11 the membrane tank particulate matter or membrane biofilm samples.Proofs In the membrane tank 12 particulate matter, Sphingomonas, Undibacterium, Nitrospira, Sphingopyxis, Porphyrobacter, 13 and GP4 were mainly identified, whileView in the membrane biofilm, Undibacterium, 14 Sphingopyxis, and Sphingomonas were the principal members.

16 The differences in microbial community composition were made obvious by analyzing OTUs 17 shared amongAdvance the three samples. Mothur’s Venn diagram analysis was used to identify the 18 numbers of shared OTUs among samples (3% cutoff) as shown in Fig. 3. Only 85 OTUs were

19 shared among the three samples, which constituted 6.9%, 4.9%, and 7.5% of the raw water,

20 the membrane tank particulate matter, and the membrane biofilm, respectively. This suggests

21 that each sample had a distinct bacteria community composition possibly due to different

22 environmental and growth conditions. The shared OTUs predominantly belonged to the

23 genera Pseudomonas, Porphyrobacter, Undibacterium, Sphingopyxis, Sphingomonas, and

24 Curvibacter. It is also worth noting that 638 OTUs were shared by the membrane tank

2 membrane tank particulate matter and the raw water, and 111 OTUs were shared by the

3 membrane biofilm and the raw water. This indicates that the bacterial community

4 composition of the membrane tank particulate matter was more similar to that of the

5 membrane biofilm than that of the raw water.

7 Freshwater habitats comprise a specific bacterial community in which Proteobacteria (mostly

8 Alpha-, Beta-, and Gammaproteobacteria) predominate, although Actinobacteria,

9 Bacteroidetes, Nitrospira, Chloroflexi, Cyanobacteria, and Verrucomicrobia are also

10 frequently present (6, 36). The bacterial communities found in drinking water treatment 11 plants, using freshwater as a source, appear not to differ muchProofs from those in freshwater. 12 Several researchers demonstrated the dominance of Proteobacteria in drinking water 13 treatment systems (2, 6, 33), however, it is Viewlikely that particular unit operations employed by 14 drinking water treatment systems affect bacterial composition. Eichler and coworkers (6)

15 studied bacterial communities from source water, water treatment plants, and tap water. They

16 demonstrated that flocculation and sand filtration did not influence the major bacterial species 17 of sourceAdvance water, but chlorination changed the composition, as evidenced by an RNA 18 fingerprint analysis. The results of our study also demonstrated distinctive bacterial

19 communities for different water treatment stages. Proteobacteria constituted 83.3% of all

20 bacterial sequences of the raw water (Fig. 2 (A)), while Beta-, Alpha-, and

21 Gammaproteobacteria constituted 61.1%, 11.3%, and 10.2%, respectively (Fig. 2(B)).

22 However, the proportions of Proteobacteria were 62.9% and 65.7% in the membrane tank

23 particulate matter and membrane biofilm, respectively. Within the phylum Proteobacteria,

24 the bacterial community difference was more significant (P < 0.001). The portions of

2 the membrane biofilm, respectively, while the portions of Alphaproteobacteria for each

3 sample were 42.4% and 32.8%, respectively. Although this study demonstrated distinct

4 bacterial communities for the three samples, the differences were not necessarily caused by

5 the environment at each stage due to the long period of the pilot operation (i.e., 30 months)

6 and variations of bacterial community composition in the raw water with time.

8 Several researchers have reported an abundance of Alphaproteobacteria in biofilms from

9 membrane filtration systems. Chen and coworkers used isolation and cloning experiments to

10 demonstrate the dominance of Alphaproteobacteria in biofilms formed on MF membranes 11 used to treat wastewater effluent (2). In addition, Chon and coworkerProofss (3) retrieved 68 clones 12 of Alphaproteobacteria from a total of 120 bacterial clones from foulants attached on a 13 membrane used to treat drinking wateViewr. Our study corroborated the dominance of 14 Alphaproteobacteria in the membrane biofilm, and revealed the dominance of

15 Alphaproteobacteria in the membrane tank particulate matter. The increase in

16 Alphaproteobacteria demonstrated the likelihood that the membrane filtration produced an 17 environmentAdvance conducive for enriching members within Alphaproteobacteria. During 18 membrane filtration, particulate matter tends to be concentrated in the membrane casing for

19 pressurized types and in the membrane tank for submersible types of filtration, and could

20 serve as a growth substrate for some microorganisms. In addition, air is supplied during

21 backwash or air sparging periods. Thus, aerobic microorganisms that can easily utilize the

22 particulate matter would have a selective advantage over other types of microorganisms,

23 though further research is needed to verify whether members within Alphaproteobacteria

24 have such an advantage.

2 Bacterial diversity

3 A total of 1,228, 1,731, and 1,133 operational taxonomic units (OTUs) were identified for the

4 raw water, membrane tank particulate matter, and membrane biofilm, respectively (Table 2).

5 For a comparison of species richness among the three samples, rarefaction curves were

6 generated using a 3% cutoff as shown in Fig. 4. None of the curves became flatter,

7 demonstrating that more sampling events (i.e., more sequences) are required to explain the

8 large fraction of OTUs in the three samples. The particulate matter sample curve was the

9 steepest, reflecting the highest species richness among the samples. The Chao1 index

10 estimated 1,665, 2,657, and 1,943 OTUs at a 3% cutoff for the raw water, membrane tank 11 particulate matter, and membrane biofilm samples, respectively,Proofs also demonstrating the 12 highest bacterial diversity for the membrane tank particulate matter. Other nonparametric 13 diversity indices such as Shannon index andView ACE gave same result (Table 2). 14

15 The numbers of 16S rRNA gene sequences analyzed (8,164 – 22,275 sequences) were

16 significantly larger than those from conventional cloning and sequencing methods, which 17 generallyAdvance employed less than 200 sequences. The increased numbers made it possible to 18 detect more microorganisms (i.e., 1,133 – 1,731 OTUs at a 3% cutoff). These numbers are

19 much greater than those in previous studies based on clone libraries in which generally less

20 than 100 OTUs were identified in samples from freshwater (6, 36) and drinking water

21 treatment systems (6, 17, 21, 33). The numbers of OTUs observed in this study were even

22 higher than those from biofilm in two water meters investigated by pyrosequencing (133 and

23 208 OTUs) (11). The increased numbers of sequences also revealed the presence of rare

24 species including pathogenic bacteria. Rare species contributed a significant portion of the

2 of the sequences (i.e., relative OTU abundance = 0.01) constituted more than 75% of the

3 bacterial diversity (i.e., observed OTUs) for each sample.

5 Signatures of pathogens

6 Occasionally, poorly treated water gives rise to outbreaks of waterborne diseases, which

7 emphasizes the importance of controlling pathogens during drinking water treatment (19). In

8 this study, signatures of potential bacterial waterborne pathogens were identified in all

9 samples (i.e., the raw water, the membrane tank particulate matter, and the membrane biofilm)

10 by comparing the retrieved sequences with known sequences of pathogens obtained from the 11 NCBI database. Potential pathogens are listed in Table 3. The pathogensProofs were mostly found 12 in the genera Legionella, Pseudomonas, Aeromonas, and Chromobacterium, although there 13 were differences in numbers for each sample.View These bacteria are reported to be responsible 14 for Legionellosis, Pneumonia, Enteritis, and Sepsis, respectively, and are frequently found in

15 contaminated water distribution systems (25). The raw water had 1,443 Pseudomonas

16 sequences, 13 Legionella sequences, and one Aeromonas sequence. The membrane tank 17 particulateAdvance matter had 17 Legionella , 20 Pseudomonas, one Aeromonas, and 11 18 Chromobacterium sequences, while the membrane biofilm had 5 Pseudomonas sequences

19 and one Chromobacterium sequence. The percentage of potential pathogens was 6.54% in the

20 raw water, but decreased to 0.36% and 0.07% in the membrane tank particulate matter and

21 membrane biofilm, respectively. The decreased percentage of potential pathogens in the

22 membrane system might be caused by residual chlorine supplied during the cleaning period

23 when the produced water was back-flowed across the membrane into the membrane tank with

-1 24 15 mg L of NaOCl. Interestingly, Pseudomonas sequences decreased significantly in the

2 Chromobacterium (0.00% → 0.08%) sequences increased. These results demonstrated the

3 likelihood that chlorine was effective in reducing the number of Pseudomonas in the

4 membrane tank, but much less effective in reducing the numbers of Legionella and

5 Chromobacterium. The phylogenetic association of the sequences of Legionella and

6 Chromobacterium observed in the membrane tank particulate matter with other pure-culture

7 sequences is presented in Fig. 6. Fifteen out of 17 Legionella sequences (88%) showed > 99%

8 sequence identity and were clustered with L. feeleii, L. maceachernii, and L. oakridgensis,

9 while eight out of 11 Chromobacterium sequences (73%) demonstrated >99% sequence

10 identity and were clustered with C. aquaticum. The three Legionella species are commonly 11 found in water and soil and are known to cause lung infectionsProofs and pneumonia (16). The 12 pathogenesis of C. aquaticum is not clear, but phylogenetically similar to C. violaceum (96.8% 13 identity) which can cause serious blood Viewpoisoning in human (34). Pathogens frequently 14 detected in freshwater such as Escherichia coli, Shigella, Salmonella, Vibrio, Helicobacter,

15 and Mycobacterium were not found in this study.

16 17 AlthoughAdvance this study demonstrated complex microbial communities in a membrane filtration 18 system, it has not clarified how the bacterial community diversity of a drinking water

19 treatment system influences the quality of water produced or how the presence of pathogen

20 signatures threatens the safety of water. In addition, it is not clear why some potential

21 pathogens persisted, or possibly concentrated in the membrane tank; further research is

22 required to resolve these issues. Nevertheless, this study clearly demonstrated that there were

23 numerous bacterial species in the membrane tank of a submersed membrane system for

24 producing drinking water, and some potential pathogenic bacteria persisted even in the

2 through, the treated water may pose an increased chance of contamination. Although

3 membrane filtration is an emerging technology that can replace traditional sand filtration,

4 caution should be paid to monitoring the integrity of the membrane, and water disinfection

5 might be practiced even after membrane filtration, to reduce the potential risk of pathogens in

6 the water produced.

8 Acknowledgements

9 We would like to thank to the engineers of KOLON Industry for providing samples and

10 operational data. This study was funded by grants from the National Research Foundation of 11 Korea to H.-D. Park (K20901001306-09B1200-10310) and Proofsby the Korea Science and 12 Engineering Foundation to S. Hong (R01-2006-000-10946-0). 13 View 14 References

15 1. Adham, S.S., J.G. Jacangelo, and J.M. Laine. 1995. Low-pressure membranes: Assessing

16 integrity. J. AWWA 87: 62-75. 17 2. Chen, AdvanceC.L., W.T. Liu, M.L. Chong, M.T. Wong, S.L. Ong, H. Seah, and W.J. Ng. 2004. 18 Community structure of microbial biofilms associated with membrane-based water

19 purification processes as revealed using a polyphasic approach Appl. Microbiol.

20 Biotechnol. 63: 466-473.

21 3. Chon, K., J.S. Chang, H. Oh, E. Lee, and J. Cho. 2009. Community of bacteria attached on

22 the PVDF MF membrane surface fouled from drinking water treatment, in Seoul, Korea.

23 Drinking Water Eng. Sci. 2: 35-39.

24 4. Craun, G.F., N. Nwachuku, R.L. Calderon, and M.F. Craun. 2002. Outbreaks in drinking-

2 5. Cui, Z.F., S. Chang, and A.G. Fane. 2003. The use of gas bubbling to enhance membrane

3 processes. J. Membr. Sci. 221: 1-35.

4 6. Eichler, S., R. Christen, C. Holtje, P. Westphal, J. Botel, I. Brettar, A. Mehling, and M.G.

5 Hofle. 2006. Composition and dynamics of bacterial communities of a drinking water

6 supply system as assessed by RNA- and DNA-based 16S rRNA gene fingerprinting. Appl.

7 Microbiol. Biotechnol. 72: 1858-1872.

8 7.Ellis, K.V., and W.E. Wood. 1985. Slow sand filtration. Crit. Rev. Environ. Sci. Technol. 15:

9 315-354.

10 8. Frank, J.A., C.I. Reich, S. Sharma, J.S. Weisbaum, B.A. Wilson, and G.J. Olsen. 2008. 11 Critical evaluation of two primers commonly used for amplificationProofs of bacterial 16S rRNA 12 genes. Appl. Environ. Microbiol. 74: 2461-2470. 13 9. Guo, H., Y. Wyart, J. Perot, F. Nauleau,View and P. Moulin. 2010. Low-pressure membrane 14 integrity tests for drinking water treatment: A review. Water Res. 44: 41-57.

15 10. Hong, S., F. Miller, and J. Taylor. 2001. Assessing pathogen removal efficiency of

16 microfiltration by monitoring membrane Integrity. Water Sci. Technol.; Water Suppl. 1: 43- 17 48. Advance 18 11. Hong, P.Y., C.C. Hwang, F.Q. Ling, G.L. Andersen, M.W. LeChevallier, and W.T. Liu.

19 2010. Pyrosequencing analysis of bacterial biofilm communities in water meters of a

20 drinking water distribution system. Appl. Environ. Microbiol. 76: 5631-5635.

21 12. Huisman, L., and W.E. Wood. 1974. Slow sand filtration. World Health Organization,

22 Geneva, Switzerland.

23 13. Lazcka, O., F.J. Del Campo, and F.X. Munoz. 2007. Pathogen detection: A perspective of

24 traditional methods and biosensors. Biosens. Bioelectron. 22: 1205-1217.

2 efficiency in achieving safe drinking water. World Health Organization and IWA

3 Publishing, London, UK.

4 15. Lovins Ⅲ, W.A., J.S. Taylor, and S.K. Hong. 2002. Micro-organism rejection by

5 membrane systems. Environ. Eng. Sci. 19: 453-465.

6 16. Madigan, M.T., and J.M. Martinko. 2006. Brock Biology of Microorganisms. Prentice

7 Hall, Upper Saddle River, USA.

8 17. Martiny, A.C., H.-J. Albrechtsen, E. Arvin, and S. Molin. 2005. Identification of bacteria

9 in biofilm and bulk water samples from a nonchlorinated model drinking water distribution

10 system: detection of a large nitrite-oxidizing population associated with Nitrospira spp.

11 Appl. Environ. Microbiol. 71: 8611-8617. Proofs

12 18. Muyzer, G., E.C. Dewaal, and A.G. Uitterlinden. 1993. Profiling of complex microbial 13 populations by denaturing gradient gelView electrophoresis analysis of polymerase chain 14 reaction-amplified genes coding for 16S rRNA. Appl. Environ. Microbiol. 59: 695-700.

15 19. Okabe, S., M. Oshiki, Y. Kamagataand , et al. 2010. A great leap forward in microbial

16 ecology. Microbes Environ 25: 230-240. 17 20. Page,Advance R.D.M. 1996. TreeView: An applicati on to display phylogenetic trees on personal 18 computers. Comput. Appl. Biosci. 12: 357-358.

19 21. Poitelon, J.-B., M. Joyeux, B. Welté, J.-P. Duguet, E. Prestel, O. Lespinet, and M.S.

20 DuBow. 2009. Assessment of phylogenetic diversity of bacterial microflora in drinking

21 water using serial analysis of ribosomal sequence tags. Water Res. 43: 4197-4206.

22 22. Saitou, N., and M. Nei. 1987. The neighbor-joining method - a new method for

23 reconstructing phylogenetic trees. Mol. Biol. Evol. 4: 406-425.

24 23. Sanapareddy, N., T.J. Hamp, L.C. Gonzalez, H.A. Hilger, A.A. Fodor, and S.M. Clinton.

2 pyrosequencing. Appl. Environ. Microbiol. 75: 1688-1696.

3 24. Schloss, P.D., S.L. Westcott, T. Ryabinand , et al. 2009. Introducing mothur: Open-source,

4 platform-independent, community-supported software for describing and comparing

5 microbial communities. Appl. Environ. Microbiol. 75: 7537-7541.

6 25. Sharma, S., P. Sachdeva, and J.S. Virdi. 2003. Emerging water-borne pathogens. Appl.

7 Microbiol. Biotechnol. 61: 424-428.

8 26. Taylor, J.S., and S.K. Hong. 2000. Potable water quality and membrane technology. Lab

9 Medicine 31: 563-568.

10 27. Thompson, J.D., T.J. Gibson, F. Plewniak, F. Jeanmougin, and D.G. Higgins. 1997. The 11 CLUSTAL_X windows interface: flexible strategies for multipleProofs sequence alignment aided 12 by quality analysis tools. Nucleic Acids Res. 25: 4876-4882. 13 28. Torrella, F., and R.Y. Morita. 1981. MicrocuViewltural study of bacterial size changes and 14 microcolony and ultramicrocolony formation by heterotrophic bacteria in seawater. Appl.

15 Environ. Microbiol. 41: 518-527.

16 29. Ueda, T., K. Hata, Y. Kikuoka, and O. Seino. 1997. Effects of aeration on suction pressure 17 in a submergedAdvance membrane bioreactor. Water Res. 31: 489-494. 18 30. Van der Bruggen, B., C. Vandecasteele, T. Van Gestel, W. Doyen, and R. Leysen. 2004. A

19 review of pressure-driven membrane processes in wastewater treatment and drinking water

20 production. Environmen. Prog. 22: 46-56.

21 31. Walpole, R.E., R.H. Myers, S.L. Myers, and K. Ye. 2007. Probability & statistics for

22 engineers & scientists. Prentice Hall, Upper Saddle River, USA.

23 32. Wang, Q., G.M. Garrity, J.M. Tiedje, and J.R. Cole. 2007. Naive bayesian classifier for

24 rapid assignment of rRNA sequences into the new bacterial taxonomy. Appl. Environ.

2 33. Williams, M.M., J.W.S. Domingo, M.C. Meckes, C.A. Kelty, and H.S. Rochon. 2004.

3 Phylogenetic diversity of drinking water bacteria in a distribution system simulator. J. Appl.

4 Microbiol. 96: 954-964.

5 34. Young, C.-C., A.B. Arun, W.-A. Lai, W.-M. Chen, J.-H. Chao, F.-T. Shen, P.D. Rekha, and

6 P. Kämpfer. 2008. Chromobacterium aquaticum sp. nov., isolated from spring water

7 samples. Int. J. Syst. Evol. Microbiol. 58: 977-880.

8 35. Yousef, A.E. 2008. Detection of bacterial pathogens in different matrices: Current

9 practices and challenges, p. 31-48 In Zourob, M., S. Elwary, and A. Turner (ed.), Principles

10 of Bacterial Detection: Biosensors, Recognition Receptors and Microsystems, Springer, 11 New York, USA. Proofs 12 36. Zwart, G., B.C. Crump, M.P. Kamst-van Agterveld, F. Hagen, and S.K. Han. 2002. 13 Typical freshwater bacteria: An analysisView of available 16S rRNA gene sequences from 14 plankton of lakes and river. Aquatic Microb. Ecol. 28: 141-155.

16 Advance

Copyright 2011 by the Japanese Society of Microbial Ecology / the Japanese Society of Soil Microbiology Table 1. The 20 genera frequently identified in the raw water, membrane tank particulate matter, and membrane biofilm

Percentage of sequences (%) Genus Membrane tank Membrane Raw water particulate matter biofilm Rheinheimera 1.44 0.00 0.00 Polaromonas 1.23 0.06 0.00 Gemmatimonas 0.00 0.27 0.72 Bdellovibrio 0.00 0.55 0.94 Cellvibrio 0.01 0.59 1.32 Methylibium 0.01 0.25 0.81 Duganella 0.03 0.26 1.26 Massilia 0.01 0.46 0.85 OD1 0.00 1.40 0.71 Novosphingobium 0.68 0.76 0.16 Rhodobacter 0.12 1.07Proofs 0.16 Nitrospira 0.00 2.90 1.62 Flavobacterium 14.72 0.24 0.16 GP4 0.00View 2.44 1.08 Sphingopyxis 0.28 2.94 5.22 Porphyrobacter 3.34 2.55 1.91 Sphingomonas 0.51 5.78 3.75 Undibacterium 0.33 3.77 8.70 Curvibacter 48.04 0.14 0.12 Pseudomonas 6.48 0.15 0.06 AdvanceTotal 77.23 26.58 29.55

Copyright 2011 by the Japanese Society of Microbial Ecology / the Japanese Society of Soil Microbiology Table 2. Number of 16S rRNA gene sequences analyzed, observed OTUs, Chao 1, Shannon index, and ACE for the raw water, membrane tank particulate matter, and membrane biofilm samples

No. of Observed OTUs Chao 1 Shannon index ACE Sample sequences analyzed. 3% 5% 3% 5% 3% 5% 3% 5% Raw water 22,275 1,228 755 1,665 986 5.30 4.60 1,660 999 Membrane tank 13,702 1,731 1,281 2,657 1,902 5.89 5.48 2,707 1,877 particulate matter Membrane biomass 8,164 1,133 872 1,943 1,395 Proofs5.58 5.15 2,458 1,636 View

Advance

Copyright 2011 by the Japanese Society of Microbial Ecology / the Japanese Society of Soil Microbiology Table 3. Bacterial pathogenic signatures from the raw water, membrane tank particulate matter, and membrane biofilm samples

Number of sequences1 membrane Genus Species tank Membrane Related disease Raw water particulate biofilm matter Legionella L. pneumophila 0 0 0 Legionellosis L. rubrilucens 0 0 0 Legionellosis L. jordansis 0 0 0 Legionellosis Others 13 17 0 Unknown Pseudomonas P. aeruginosa 593 3 1 Pneumonia P. alcaligenes 14 3 1 Pneumonia P. pseudoalcaligenes 1 2 0 Pneumonia Others 835 12 3 Unknown Aeromonas A. hydrophila 0 0 0 Enteritis A. caviae 0 Proofs0 0 Enteritis A. veronii 0 0 0 Enteritis A. sorbia 0 1 0 Enteritis Others 1 0 0 Unknown Chromobacterium C. violaceum View0 0 0 Sepsis Others 0 11 1 Unknown Total 1,457 49 6 Percentage 6.54 0.36 0.07 1Classification was based on an identity higher than 97% using the 16S rRNA gene sequences of the designated species.

Advance

Fig. 1. A schematic representation of the pilot-scale drinking water treatment plant using membrane filtration.

Fig. 2. Taxonomic assignment of 16S rRNA gene sequences retrieved from the raw water, membrane tank particulate matter, and membrane biofilm classified by (A) phylum and (B) class within the phylum Proteobacteria.

Fig. 3. Venn diagram of shared OTUs among the raw water, membrane tank particulate matter, and membrane biofilm.

Proofs

Fig. 4. Rarefaction curves of OTUs in the raw water, membrane tank particulate matter, and membrane biofilm evaluated by 3% sequenceView variation.

Fig. 5. Rank abundance curves of OTUs in the raw water, membrane tank particulate matter, and membrane biofilm evaluated by 3% sequence variation. The values corresponding to 0.01 of relativeAdvance OTU abundance were 120, 125, and 229 abundance ranks for the membrane tank particulate matter, membrane biofilm, and raw water, respectively. The abundance ranks constituted the top 6.1, 9.7, and 16.7% of the total observed OTUs for each sample.

Fig. 6. Neighbor-joining phylogenetic tree based on Legionella and Chromobacterium 16S rRNA gene sequences recovered from the membrane tank particulate matter and pure cultures of these bacteria obtained from Genbank (http://www.ncbi.nlm.nih.gov/). Bacillus subtilis was used as the outgroup. The scale bar represents 0.1 substitutions per site. Clone sequences

Copyright 2011 by the Japanese Society of Microbial Ecology / the Japanese Society of Soil Microbiology exhibiting more than 99% identity are indicated by the number of other sequences in parentheses following the sequence name. Genbank accession numbers for pure culture strains are indicated in parentheses following the strain name. Bootstrap values providing 50% support are indicated at the nodes.

Proofs View

Advance

From raw water Suction Raw water equalization basin pump pump

Produced water

Produced water Air blower Chemical NaOCl tank pump tank Membrane module Proofs Backwash Membrane tank pump

View Disposal Concentrate pump

Advance

Unclassified Phylums Raw water (A) Membrane tank particulate matter Minor Phylums Membrane biofilm

OD1

Acidobacteria

Nitrospira

Bacteroidetes Proteobacteria Proofs 0 20 40 60 80 100 Percentage View Unclassified Proteobacteria Raw water (B) Membrane tank particulate matter Membrane biofilm Epsilonproteobacteria

Deltaproteobacteria

AdvanceGammaproteobacteria

Betaproteobacteria

Alphaproteobacteria

0 20406080100 Percentage

Membrane tank particulate matter 1731 OTUs Membrane biofilm 1133 OTUs

1029 OTUs 553 OTUs Proofs469 OTUs 85 OTUs

64 OTUs View26 OTUs

1053 OTUs Advance

Raw water 1228 OTUs

2000

1500

1000 OTUs

500 Membrane tank particulate matter Membrane biofilmProofs Raw water

0 0 5000 10000View 15000 20000 25000 Number of sequences sampled

Advance

Membrane tank particulate matter Membrane biofilm Raw water

e 0.1 c

0.01 tive OTU abundan OTU tive a

Rel 0.001 Proofs 195 (15.9%) 275 (24.3%) 353 (20.4%) 0.0001 View 0 500 1000 1500 2000

Abundance rank order Advance

Bacillus subt ili s (GU991 853) Membrane tank sequence F1Q32TO03C6FU7 Chromobacterium Membrane tank sequence F1Q32TO03C26AZ 50 100 Chromobacterium piscinae (AJ871127) Chromobacterium haemolyticum (DQ785104) 60 Membrane tank sequence F1Q32TO03C1JGI 60 62 75 Membrane tank sequence F1Q32TO03DINVS (+7 membrane tank sequences) Chromobacterium aquaticum (EU109734)

93 Chromobacterium violaceum (M22510) Chromobacterium subtsugae (()AY344056) 90 Chromobacterium pseudoviolaceu (AJ871128) Legionella israelensis (X73408) Legionella londiniensis (AB5148) 100 Legionella spiritensis (AB21117) Legionella brunensis (Z32636) Membrane tank sequence F1Q32TO03DN104 49 Legionella jordanis (X73396) Legionella jamestowniensis (X73409) 46 62 Legionella hackeliae (M36028)

60 Legionella oakridgensis (Z32642) Legionella maceachernii (X60081) Proofs Legionella feeleii (NR026117) Membrane tank sequence F1Q32TO03DA2G9 (+12 membrane tank sequences) 100 Legionella 89 Membrane tank sequence F1Q32TO03C7IJ9 Membrane tank sequence F1Q32TO03DTRMN 95 Legionella rubrilucens (Z32643)View Legionella erythra (Z32638) Membrane tank sequence F1Q32TO03DUBE4 80 Legionella cincinnatiensis (X73407) 90 Legionella sainthelensi (AB185331) 50 55 Legionella longbeachae (AY44474)

100 Legionella moravica (Z49729) Legionella quateirensis (Z49732) Legionella cherrii (X73404) Legionella pneumophila (FN66782) Fluoribacter dumoffii (Z32637) Advance49 Legionella tucsonensis (Z32644) 96 Legionella steigerwaltii (X7340) Fluoribacter bozemanae (M36031)

0.1 substitution/site