Article pubs.acs.org/est
MS2 Bacteriophage Reduction and Microbial Communities in Biosand Filters Hanting Wang,† Takashi Narihiro,†,‡ Anthony P. Straub,§ Charles R. Pugh,† Hideyuki Tamaki,†,‡ Johnathan F. Moor,† Ian M. Bradley,† Yoichi Kamagata,∥ Wen-Tso Liu,† and Thanh H. Nguyen*,† †
Department of Civil and Environmental Engineering, University of Illinois at Urbana−Champaign, 205 N. Mathews, 3230 Newmark Lab, Urbana, Illinois 61801, United States ‡ Bioproduction Research Institute, National Institute of Advanced Industrial Science and Technology (AIST), Central 6, Higashi 1-1-1, Tsukuba, Ibaraki 305-8566, Japan § Department of Chemical and Environmental Engineering, Yale University, 9 Hillhouse Avenue, Mason Lab, New Haven, Connecticut 06520-8286, United States ∥ Bioproduction Research Institute, National Institute of Advanced Industrial Science and Technology (AIST), 2-17-2-1, Tsukisamu-Higashi Toyohira, Sapporo, Hokkaido 062-8517, Japan S Supporting Information *
ABSTRACT: This study evaluated the role of physical and biological filter characteristics on the reduction of MS2 bacteriophage in biosand filters (BSFs). Three full-scale concrete Version 10 BSFs, each with a 55 cm sand media depth and a 12 L charge volume, reached 4 log10 reduction of MS2 within 43 days of operation. A consistently high reduction of MS2 between 4 log10 and 7 log10 was demonstrated for up to 294 days. Further examining one of the filters revealed that an average of 2.8 log10 reduction of MS2 was achieved within the first 5 cm of the filter, and cumulative virus reduction reached an average of 5.6 log10 after 240 days. Core sand samples from this filter were taken for protein, carbohydrate, and genomic extraction. Higher reduction of MS2 in the top 5 cm of the sand media (0.56 log10 reduction per cm vs 0.06 log10 reduction per cm for the rest of the filter depth) coincided with greater diversity of microbial communities and increased concentrations of carbohydrates. In the upper layers, “Candidatus Nitrosopumilus maritimus” and “Ca. Nitrospira defluvii” were found as dominant populations, while significant amounts of Thiobacillus-related OTUs were detected in the lower layers. Proteolytic bacterial populations such as the classes Sphingobacteria and Clostridia were observed over the entire filter depth. Thus, this study provides the first insight into microbial community structures that may play a role in MS2 reduction in BSF ecosystems. Overall, besides media ripening and physical reduction mechanisms such as filter depth and long residence time (45 min vs 24 ± 8.5 h), the establishment of chemolithotrophs and proteolytic bacteria could greatly enhance the reduction of MS2.
■
well-received, and continually used.3,5 An estimated 430 000 BSFs have been implemented by diverse communities in over 60 countries as a sustainable drinking water treatment solution.6 A BSF consists of a concrete or plastic hollow chamber packed with a filtration sand layer. Two layers of gravel underlie the filtration sand to allow drainage of water throughout the length of the filter. A diffuser plate sits 5 cm above the filtration sand layer and is used to prevent influent water from disturbing the sand. Typically, every 16−32 h, water is poured into the inlet basin and is filtered through the sand
INTRODUCTION The World Health Organization estimates that 760 million people around the world lack access to improved sources of drinking water.1 Many of these people live in impoverished communities, where children are among those most affected by diarrheal diseases.2 In these areas, a centralized water treatment system is often infeasible.3 However, recent studies have shown that point-of-use (POU) systems, in which water is treated at the location of consumption, are able to reduce diarrheal diseases by 30−40%.3,4 In particular, biosand filters (BSFs), or intermittently run slow sand filters, have been identified as one of most promising POU systems.3 BSFs effectively improve several water quality parameters (e.g., turbidity and concentrations of microorganisms), producing 20−40 L of safe drinking water per day. These systems are easy to operate, © 2014 American Chemical Society
Received: Revised: Accepted: Published: 6702
January 28, 2014 May 20, 2014 May 23, 2014 May 23, 2014 dx.doi.org/10.1021/es500494s | Environ. Sci. Technol. 2014, 48, 6702−6709
Environmental Science & Technology
■
media by gravity, displacing water that had been occupying the filter from the previous charge out of the outlet tube. The filter design prevents the water level from falling below the filtration sand layer, which is critical for the maturation or ripening of a biologically active zone (Schmutzdecke or biolayer) comprising the top 5 cm of the filter media. The biological composition and the role of this biolayer on the BSF function have not been determined. The two most recent versions of the BSF designed and implemented by the Center for Affordable Water and Sanitation Technology (CAWST) are Version 9 and Version 10.6 The Version 9 filter has a 45 cm thick filtration sand layer and a 20 L recommended charge volume, whereas Version 10 has a 55 cm filtration sand layer and a 12 L recommended charge volume to allow a longer contact time between water and the filter media.6 BSFs should be able to reduce both pathogenic bacteria and virus concentrations to ensure public health protection for people who use the filtered water. This requirement has motivated a number of laboratory BSF studies. While most studies agree that BSFs can effectively reduce bacteria (>4 log10) and remove turbidity to under 2 NTU after 6−8 weeks, reported virus reduction efficacy has varied widely in studies lasting 6−10 weeks.7−9 The high removal of suspended solids and large-size pathogens is linked to physical and biological removal by the biolayer, enabling enhanced sieving, adsorption, and predation.10 However, within the first 60 days of operation, this biological zone in filters is not as effective in reducing virus concentrations.7,8 Only one study has shown reduction of viruses in BSFs with an efficiency close to the 4 log10 drinking water standard set by USEPA, and this high reduction occurred only after >200 days of operation.10 Clearly, additional longterm studies are needed not only to evaluate virus reduction efficiency but also the factors influencing virus reduction. To date, most studies have focused on physical factors influencing virus reduction such as ripening.7,8,10,11 Only one study indirectly suggested that microbial activities influence MS2 reduction by a BSF operated for 50 days.11 The addition of sodium azide, which can suppress growth of aerobic bacteria, but is not likely to change the surface properties of the sand media, caused a decrease in the rate of MS2 reduction by BSF.11 These results suggest the role of active aerobic bacteria on MS2 reduction.11 The logical questions to ask are what aerobic bacteria are present in the BSF and how does their growth influence MS2 reduction. Because sodium azide does not influence the growth of anaerobic bacteria, their presence and roles have not been investigated. The individual and synergistic contributions from physical/operational characteristics and microbial communities developed inside the filter on virus reduction in both unripened and ripened BSFs remain unclear. To further understand this, we (1) evaluated MS2 reduction in three full-scale concrete Version 10 BSFs, (2) investigated the effect of three characteristics associated with a filter (i.e., depth, media ripening, and residence time) on virus reduction efficacy in a BSF, and (3) used genomic techniques to characterize microbial communities present in the BSF and analyze their role in virus reduction. This represents the first study in which genomic sequencing techniques have been used to determine the microbial communities in a BSF. Improved knowledge of the mechanism and magnitude of virus reduction can be used to enhance the design and efficacy of future BSFs, ultimately leading to improved performance and reduced disease incidence.
Article
MATERIALS AND METHODS
Virus Selection and Infectivity Assay. Male specific type 2 bacteriophage (MS2) (ATCC 15597-B1) was chosen as a surrogate virus due to its similarities in size and morphology to human enteric viruses.12 MS2 was replicated and purified as described previously.13,14 The purified MS2 stock, concentrated to ∼1011 plaque forming units (PFU) per mL, was stored in 10 mM NaCl at 4 °C. Enumeration of MS2 concentrations was performed using the double agar layer procedure.15 Dilutions obtaining plaque counts between 30 and 300 plaques were used to calculate the PFU per mL. Biosand Filter and PVC Column Experiments. An outline of the experiments conducted in this study is provided in the Supporting Information (Figure S1). Three concrete BSFs were constructed using specifications provided by CAWST for Version 10 of their filter design.6 Six sampling ports were installed in one of the filters, called the concrete port filter, at depths (cm) of 5.4, 10.9, 16.3, 21.7, 32.6, and 54.3 prior to packing (TOC art). Gravel and sand were manually sieved using mesh sizes recommended by CAWST.6 The upper and lower gravel layers consisted of gravel sieved to sizes of 1− 6 mm and 6−12 mm, respectively. The sand was sieved to a size of 0.7 mm or smaller. Plug flow conditions were verified for the concrete port filter using 4 pore volumes of a 0.1 M NaCl tracer solution. The conductivity of the effluent water was recorded every minute and used to construct the breakthrough curve shown in Figure S2a. As determined through the tracer breakthrough curve, the pore volume was 13 L for the Version 10 filter; i.e., all of the 12 L charge poured in this filter remains in the filter for at least one residence time. Feed water was collected from a natural aquifer underneath Newmark Civil Engineering Laboratory (205 N. Matthews, Urbana, Illinois, 61801) allowing a long-term study over various seasons. Prior to use, water was passed through a greensand filter to remove iron and manganese, which could precipitate and change the surface characteristics of the BSF. Newmark groundwater, characterized in previous studies, was found to contain 1−2 mg/L TOC16 and turbidity between 0.25 and 0.7 NTU.10 Newmark groundwater has also been shown to form biofilms in previous studies.16,17 In addition, the sand media and groundwater were analyzed for total dissolved Al, Ca, Fe, Mg, and Zn for the sand media and Fe, S, and Cl for the groundwater using inductively coupled plasma-optical emission spectrometry (ICP-OES). Daily influent and effluent samples from the filter were analyzed for pH, dissolved oxygen concentration, and alkalinity. This well-characterized groundwater was chosen so that the biofilm was developed under a relatively steady state condition over four years from 2010 to 2013. For MS2 reduction experiments, the filters were charged daily with 12 L of feedwater containing MS2 at approximately 107 PFU/mL. Flow rates, which ranged between 0.18 and 0.4 L/min throughout the duration of the experiments, were measured daily and used as an indicator for filter ripening. Note that the concrete port filter was run with Newmark groundwater for 5 weeks before MS2 reduction experiments were started. In addition to the concrete port filter with biofilm developed on the filter media, a 10 cm diameter PVC column, called the PVC port column, with seven ports inserted along the length of the column was constructed to model a full-scale concrete Version 10 BSF with unripened sand media. The PVC port column had ports installed at the same depths as the 6703
dx.doi.org/10.1021/es500494s | Environ. Sci. Technol. 2014, 48, 6702−6709
Environmental Science & Technology
Article
concrete port filter, but with an extra port inserted at a depth of 43.4 cm. A tracer test for the PVC port column (Figure S2b) demonstrated a nearly identical breakthrough curve to the concrete port filter. Prior to starting MS2 reduction experiments, a 9 day equilibration period removed air trapped between sand grains and stabilized the flow rate. Influent, effluent, and port samples for short (∼10 min) and long (24 ± 2 h) residence times were taken for four runs between 9 and 28 days of operation. MS2 Bacteriophage Reduction Sampling and Sand Sampling. Effluent samples for virus assay were collected after approximately 1 and 2 L of water had passed through the PVC port column and the concrete port filter, respectively. Port samples were collected after added water flowing through the outlet tube ceased, and these samples are referred to as short residence time (averages of ∼45 min and ∼10 min taken throughout the durations of the experiments for the concrete port filter and the PVC port column, respectively) samples. Long residence time port samples were collected after 24 ± 8.5 and 24 ± 2 h of residence time for the concrete port filter and the PVC port column, respectively, prior to the following daily charge. These two residence times were selected to study whether allowing water to stay inside the filter increased MS2 reduction. All virus samples were collected in 1.5 mL centrifuge tubes, stored at 4 °C, and analyzed within 24 h of sampling. MS2 reduction results for different conditions were compared using statistical two-tailed t-test. After 240 days of operation, three core sand samples used for pyrosequencing were taken from the concrete port filter using a 1 in. diameter PVC tube that was marked at each port depth. Due to the weight of the sand suspension, the core samples for Port 6 were incomplete, and only samples for Ports 1−5 were analyzed. Sand samples were also taken at a depth right above Port 1 to represent samples from the biolayer, giving a total of 18 sand samples from the three core samples. The tubes were cut at each port depth, and 4 g of sand were extracted at each port. Two grams of sand, combined with 20 mL of nanopure water in a 50 mL centrifuge tube, were used for quantifying extracted carbohydrate and protein. The remaining 2 g of sand, combined with 20 mL of phosphate buffer solution in a 50 mL centrifuge tube, were used for pyrosequencing. All samples were stored at −80 °C until analysis. Microbial Diversity by 16S rRNA Gene Pyrosequencing. DNA extraction from the sand samples (ca. 1 g wet weight each) was performed by FastDNA SPIN Kit for Soil (MP Biomedicals, Carlsbad, CA, USA) according to the manufacturer’s instructions. PCR, 16S rRNA gene pyrosequencing, and data analyses were performed by a protocol described previously18 and sequenced at the W.M. Keck Center, part of the Roy J. Carver Biotechnology Center at the University of Illinois at Urbana−Champaign. Detailed descriptions of the methods for biosand filter experiments, MS2 bacteriophage and sand sampling, extractable carbohydrate and protein analysis, and pyrosequencing can be found in the SI.
Figure 1. (A) Comparison of the log10 reduction of MS2 for three Version 10 biosand filters. (B) Effect of media ripening on cumulative log10 reduction of MS2 (unaveraged, part B; averaged, part 3) with short residence time of 10 min for days 9−27 and 45 min for days 43− 240. For parts B and C, data obtained for 9−27 days were from the PVC port column. Data for longer times were from the concrete biosand port filter. Detection limits for days 138−152 and for days 223−240 were 6.4 log10 and 5.3 log10, respectively. Different MS2 stock concentrations used for MS2 reduction experiments led to different detection limits.
significantly higher than the reduction for days 9−27 (p = 1.2 × 10−6), indicating a difference in media ripening between the two time periods, where the filter was unripened before 27 days and was ripened after 43 days. Between days 9 and 27, the unripened filter showed an exponential dependence of reduction on depth (i.e., a straight line on semilog plot with R2 = 0.75 for short residence time shown in Figure 2A). The average cumulative MS2 reduction throughout the entire filter depth for days 9−27 was 2.2 log10 (Figure 1C). This observed reduction in MS2 for unripened filters is significantly higher than data obtained with surface water sources reported in previous studies7,11 probably because the groundwater source16 used in this study contains 1.5 mM Ca2+ and 1.0 mM Mg2+. The presence of divalent cations has been shown to substantially increase adhesion of MS2 to organic matter coated sand surfaces due to inner-sphere complexation with carboxylate groups on MS2 surfaces and the organic matter structure.20,21 Results for the ICP-OES analysis, which used concentrated hydrofluoric acid solution to digest the sand media (0 ppm for Al and Fe, 28.66 ppm for Mg, 64.1 ppm for
■
RESULTS Impact of Media Ripening on MS2 Reduction throughout Filter Depth. The MS2 reduction in three concrete Version 10 biosand filters is shown in Figure 1A. All filters reached the EPA standard19 of 4 log10 reduction by 43 days. A comparison of the log10 removal of MS2 for unripened vs ripened filters as a function of filter depth is shown in Figure 1B. The cumulative MS2 reduction for days 43−204 is 6704
dx.doi.org/10.1021/es500494s | Environ. Sci. Technol. 2014, 48, 6702−6709
Environmental Science & Technology
Article
Impact of Residence Time on MS2 Reduction in the BSF. The average cumulative MS2 reductions in the unripened PVC port column for short and long residence times are shown in Figure 2. MS2 reduction fit a linear correlation of R2 = 0.75 and R2 = 0.87 for short and long residence times, respectively. MS2 reduction reached 2.0 log10 and 4.8 log10 for short and long residence times, respectively. A 4 log10 MS2 reduction could not be reached until Port 6 under a long residence time. The average cumulative MS2 reduction as a function of depth and residence time for the ripened port filter is shown in Figure 2B. MS2 reduction reached an average of 5.2 log10 and 5.6 log10 cumulatively for short and long residence times, respectively. In both cases, the highest portion of the MS2 reduction was observed for the biolayer, where the water had traveled through 5 cm of sand. Specifically, 1.6 log10 and 2.8 log10 reductions of MS2 were observed for the biolayer for the short and long residence times, respectively. The samples collected from the ports after the biolayer for both residence time conditions showed further contribution to the overall increase of MS2 reduction but at a decreasing rate (Figure 2). This correlation between MS2 reduction and filter depth suggests that the reduction in the biolayer is significantly higher than in the rest of the filter. Microbial Diversity in the Filter. A total of 122 962 16S rRNA gene pyrosequencing reads were obtained from the 18 sand samples (triplicates for biolayer and Ports 1−5), and 3026 operational taxonomic units (OTUs) were classified using a sequencing similarity of 97% (Table S2). The calculated Good’s coverage values were 90.27−96.88%, indicating that the OTUs detected were sufficient to estimate the microbial diversity in the BSF ecosystems. According to the Chao1 nonparametric estimators, the sand samples contained approximately 1.36− 2.25-fold more OTUs than that detected using 16S pyrotag. Based on different α-diversity indexes (i.e., OTU numbers, Chao1, Shannon, and Simpson), the microbial diversities of BSF ecosystems decreased with filter depth (Figure S3). The similarity of microbial compositions among the sand samples was compared using the principle coordinate analysis (PCoA) based on weighted UniFrac, and a clear shift in the microbial community structures was observed among the sand samples taken at different filter depths (Figure 3A). The OTUs assigned to the phylum Proteobacteria (especially the classes Betaproteobacteria and Gammaproteobacteria) were the dominant populations in the BSF columns (>16% of the total populations in all samples) (Figure 3B, Table S4). Within the family Enterobacteriaceae of the class Gammaproteobacteria, an OTU (OTU2983) was closely related to E. coli strain DH1 (99.7% identity; AP012030) (Table S4) but at a low abundance ( 0.19 for all six ports). This may indicate that the filter was fully ripened, and the reduction had reached its maximum of 5.3 log10 and dropped below the detection limit of the experiment. During this period, the plateau observed between Ports 5 and 6 likely represented the detection limit for MS2 reduction (Figure 1C). 6705
dx.doi.org/10.1021/es500494s | Environ. Sci. Technol. 2014, 48, 6702−6709
Environmental Science & Technology
Article
Figure 3. (A) Principal coordinate analysis based on the abundances of 16S rRNA gene OTUs (weighted UniFrac) and separating the depths of the core sand columns. For this analysis, 16S rRNA gene OTUs observed here were normalized to 4000 reads per sample. (B) 16S rRNA gene pyrotagbased microbial compositions across depths in BSF column. The relative abundances of each phylogenetic group were calculated on the average of the three core sand columns.
Acidiferrobacter,24−26 were often observed over the entire filter depth (Figure 4D and E). For microorganisms involved in N cycling, 7 and 30 OTUs were classified into the genera “Ca. Nitrosopumilus” and Nitrospira, which are well-known ammonia-oxidizing archaea (AOA) and nitrite-oxidizing bacteria (NOB), respectively (Table S4). In particular, the “Ca. Nitrosopumilus”-related OTU3410 reached a maximum level in the biolayer (8.8 ± 5.3% of the total populations) (Figure 4F). Likewise, the abundance of Nitrospira-related OTU300 was also the highest in the biolayer (6.3 ± 2.6% of the total populations) (Figure 4G). In addition to nitrifying microbes, microbial populations relevant to denitrification, nitrogen fixation, and anaerobic ammonia oxidation (anammox) were detected. OTUs related to Sulf uricella (OTU1971) and Azoarcus (OTU3007) were detected and could act as denitrifying bacteria along with some Thiobacillus and Thiohalomonas populations. An OTU
(OTU454) related to the nitrogen-fixer Azospirillum was observed throughout the filter depth at low population densities (1.0−2.0% of total populations). Previously known anammox members of the class Kueneniae in the phylum Planctomycetes were detected at low abundances (a total of 7 OTUs). These N-cycling microbes may contribute to remove N-related contaminants (e.g., ammonia and nitrite) in groundwater, which have been shown to contain 388 mg S/L, 0.91 mg NH3-N/L, 0.054 mg NO2-N/L, and
