ARTICLE pubs.acs.org/est
Impact of Seawater-Quality and Water Treatment Procedures on the Active Bacterial Assemblages at Two Desalination Sites C-L de O. Manes,†,‡,* C. Barbe,^ N. J. West,§,|| S. Rapenne,^ and P. Lebaron†,‡,§,|| †
UPMC Univ Paris 06, UMR 7621, LOMIC, Observatoire Oceanologique, F-66651, Banyuls/mer, France CNRS, UMR 7621, LOMIC, Observatoire Oceanologique, F-66651, Banyuls/mer, France § UPMC Univ Paris 06, UMS 2348, Observatoire Oceanologique, F-66650 Banyuls/Mer, France CNRS, UMS 2348, Observatoire Oceanologique, F-66650 Banyuls/Mer, France ^ Veolia Environnement—Centre de recherche sur l'eau, Chemin de la Digue 78600 Maisons Laffitte, France
)
‡
bS Supporting Information ABSTRACT: Inorganic and organic compounds, particles and microorganisms in intake waters are mainly responsible for fouling of reverse osmosis membranes, which reduces the efficiency of the desalination process. The characterization of seawater quality to better predict its fouling potential remains a challenge for the desalination field and little is known about the seasonal variability of water quality parameters in the coastal waters used to supply desalination plants. In this study, standard water quality methods were combined with flow cytometry and molecular methods (16S rRNA sequencing and fingerprinting) to assess in parallel, the physicochemical properties, the microbial abundance and the active microbial community composition of the intake waters and their associated pretreated waters at two desalination sites from July 2007 to July 2008. The overall assessment of quality parameters revealed that microfiltration followed by slow sand filtration were the most efficient in removing microorganisms than the conventional dual media filtration routinely used in full-scale desalination plants, and that all treatments were inefficient for organic matter reduction. Temporal variation of the environmental parameters such as temperature, turbidity and silt density index only moderately affected the bacterial community structure in raw waters, but that interestingly, water treatment compartments changed the composition and diversity of the metabolically active bacterial populations and thus create distinct ecological post-treatment niches.
’ INTRODUCTION The actual bottleneck in the reverse osmosis (RO) based desalination process is RO membrane fouling, which hinders plant performance. Membrane fouling can be caused by inorganic, organic compounds and particles, but also by the attachment of microorganisms to membrane surfaces and subsequent biofilm formation.13 Recent studies point out that organic and biofouling are the major foulants of seawater RO membranes.4,5 Nutrient-rich waters feed desalination plants as they are mostly located in coastal areas, therefore a comprehensive characterization of seawater quality is needed to better predict their fouling potential which remains a challenge for the desalination research field. Seawater quality can be interpreted differently according to its use. For the desalination field reliable indicators to predict the fouling potential of RO feedwater are important for preventing and diagnosing fouling at the design stage of RO plants, and to access pretreatments performance during plant operation.6 The silt density index (SDI) is widely used to monitor water quality and to roughly predict the fouling potential of the intake and treated waters in terms of colloids and particles removal in r 2011 American Chemical Society
full-scale desalination plants. However, fouling problems have been reported even with very low SDI values, that is, SDI < 1.7 Complementary analyses characterizing potential organic and microbial foulants in seawater have therefore been recently incorporated as water-quality parameters for evaluating pretreatment efficacies upstream the RO units in some desalination plants.8,9 The study of organic matter (OM) removal through the different types of pretreatments allowed an evaluation of their efficiency whereas pigment content determination assessed the raw water quality in terms of autotrophic microorganism abundances.8 The seasonal monitoring of microbial abundance would permit the identification of algal bloom periods, which could lead to higher organic loads into the plant, and may be a very useful parameter to assess pretreatment efficiencies. Similarly, bacterial community input should be investigated in more detail as they play a key role Received: March 11, 2011 Accepted: June 16, 2011 Revised: June 10, 2011 Published: June 16, 2011 5943
dx.doi.org/10.1021/es200799t | Environ. Sci. Technol. 2011, 45, 5943–5951
Environmental Science & Technology in downstream biofilm establishment on SWRO membranes.1,5 Seasonal bacterial community shifts have been associated with variations in environmental parameters and to differential natural OM supply and composition.1012 Therefore it is not only important to estimate the microbial content in the RO feed waters but also the shifts in bacterial communities associated with seasonal changes of water-quality parameters that may reveal active RO membrane colonizers. Recent analysis on bacterial community diversity coupled to biogeochemical parameters demonstrated that cell abundances were unrelated to activity since some bacterial groups present in lower abundance may exhibit high metabolic activity.13 Therefore describing the active bacterial community in waters feeding desalination plants could help to (i) determine the influence of water quality parameters on bacterial community structure on site and (ii) assess the impact of pretreatments in removing and/or favoring certain bacterial groups present in raw water. The goal of this study was to apply new methodologies for water quality control in water treatment lines for a comprehensive analysis of quality parameters not only in terms of known physicochemical/chemical characterization but also for microbiological content. The bacterial community structure as well as water quality parameters of raw and treated waters from two desalination sites were monitored monthly over a one-year period (July 2007 to July 2008). Furthermore we assessed the impact of conventional, dual media filtration and nonconventional, microfiltration and beach-well, water treatments on the metabolically active bacterial populations by investigating in detail the taxonomic composition of bacterial communities at a specific time point coinciding with an elevated organic load from feed waters.
’ MATERIALS AND METHODS Processes and Samples Description. Water sampling began
in July 2007 and ended in July 2008 at two desalination sites, Site 1 and Site 2 both located on the NW Mediterranean coast. Water quality from these two sites differs according to the raw water origin and the pretreatment processes applied. Site 1 is composed of two parallel treatment lines. The first treatment line is a conventional pretreatment, which consists of an addition of sulphuric acid, ferric chloride and polymer for the coagulationflocculation step followed by a dual media filtration (anthracite and sand, dual media filtration-DMF) at a filtration rate of 7.5 m/h. The second parallel treatment train is a membranebased pretreatment, a microfiltration (0.1 μm MF). In Site 2, water pretreatment consists of a beach-well with a sand depth of around 0.81 m at a slow filtration rate 0.4 m/h (slow sand filtration-SSF). The sampling points were located after each step of the pretreatment in order to evaluate separately their efficiency for the different parameters monitored. Sampling frequency and period varied according to the availability and the proximity of the studied site. For Site 1 a monthly survey was undertaken (12 time points from July 2007 to July 2008, excepting November 2007), whereas for Site 2, fewer samples were taken but still covering all seasons except winter (six time points: monthly from July 2007 to October 2007 covering summer and autumn and then on April 2008 (spring) and July 2008 (summer)). Colloidal and Particulate Fouling Indicators. Turbidity as well as Silt Density Index measurements were performed to assess the colloidal and particulate fouling potential of studied waters.
ARTICLE
Turbidity was measured using a portable nephelometer or an online sensor to assess roughly the amount of suspended solid within studied waters. The limit of quantification is around 0.3 NTU (nephelometric turbidity unit). SDI measurements were performed according to the ASTM Method. Water was passed through acetate cellulose 0.45 μm membrane filter (47 mm) at a constant pressure of 30 psi. The initial time (t0) to collect 500 mL of permeate was measured and the same measurement was performed after i min (3 for raw water or 15 for pretreated water) of filtration (ti). The SDI was calculated using the following formula: SDI ¼
ti t0 100 x 15 ti
Natural Organic Matter (NOM) Characterization. Total organic carbon (TOC), total nitrogen (TN), and UV254 analyses were determined using a Shimadzu TOC-V analyzer.14 The limit of quantification was 0.5 and 0.07 ppm for TOC and TN concentration respectively. The UV absorption value was measured by spectrometry at 254 nm and represents a surrogate parameter for aromaticity measurement of the natural organic matter. Using these parameters, two indexes were calculated, SUVA and the molar ratio Corg/Norg. Both indexes give indications regarding the nature and origin of the organic matter. Accordingly, SUVA values below the threshold value (670 nm) fluorescence due to phycoerythrin and chlorophyll pigments, respectively. Bacterial cells were prestained with the fluorescent dye SyberGreen I (Molecular Probes, Eugene, OR) for 10minutes before analysis. Acquisition data were performed using CellQuest software (BDBiosciences). Fluorescent 1.002 μm beads (Polysciences Inc., Europe) used as an internal standard were systematically added to each analyzed sample. For each cell population discriminated, mean fluorescence and mean light scatter were normalized, dividing them by homologue signals from beads, making results intercomparables. 3. Nucleic Acids Extraction and cDNA synthesis. Water samples were obtained from two desalination sites. Water samples collected in different compartments, before and after pre treatments (described above), were filtered onto 0.2 μm pore size Sterivex filter units (46 L) (Millipore) after prefiltration through 3 μm pore size, 47 mm Nucleopore polycarbonate filters (Millipore). Sterivex units were immediately frozen and shipped. The DNA and RNA coextraction method from water samples was performed as described previously using a modified protocol1 and the Qiagen All Prep kit according to the manufacturers protocol. Total RNA extracts were reverse-transcribed using the SuperScript VILO cDNA synthesis kit (Invitrogen) according to the manufacturer’s protocol and by adding the 16S rRNA gene reverse specific primer 786RMOD (50 -ACTACCGGGGTATCTAAKCC-30 ), specially designed to target the great majority of bacterial groups using the ARB software.18 4. PCR-CE-SSCP Fingerprinting. PCR amplification of short bacterial fragments (200 bp) of the 16S rRNA gene V3 region was performed using the specific primers w49-F (50 -ACGGTCCAGACTCCTACGGG-30 )19 and w34-R-TET (50 -TTACCGCGGCTGCTGGCAC-30 )20 which was 50 labeled with phosphoramidite (TET, Applied Biosystems, Applera Corp. Norwalk, Connecticut). PCR amplification was performed as described previously.1 PCR product (410 ng) was mixed with deionizedformamide and the internal size standard Gene-Scan-400 HD labeled with ROX (Hi-Diformamide Applied Biosystems, Applera Corp. Norwalk, Connecticut). Samples were denatured at 94 C for 5 min and placed immediately in a water/ice bath for 10 min before separation by capillary electrophoresis single-stranded conformational polymorphism (CE-SSCP) using a ABI310 Genetic Analyzer (Applied Biosystems, Applera Corp. Norwalk, Connecticut). Electrophoresis was carried out at 12 kV and 30 C for 30 min per sample, according to Delbes et al. (2000). The electropherograms were analyzed by the Genescan software (Applied Biosystems, Applera Corp. Norwalk, Connecticut) using the second-order least-squares size calling method to normalize mobility between different runs as described previously.21 SAFUM software22 was used to normalize the total area of the SSCP profiles and the mobilities between different runs using the ROX internal standard. The similarity of the total (DNA) bacterial community compositions of the samples by means of SSCP profiles was examined by clustering analysis and nMDS plots from distance matrices comprising 750 data points calculated using the BrayCurtis similarity algorithm taking into account both peak presence and relative abundance. The differences between groups were tested by a one-way ANOSIM analysis using the PRIMER software (PRIMER-E Ltd., Plymouth, UK).
ARTICLE
5. 16S rRNA Library Construction. Bacteria-specific primers 27FMOD (50 -AGRGTTTGATCMTGGCTCAG-30 )23 and 786RMOD (50 -ACTACCGGGGTATCTAAKCC-30 ) were used to amplify a partial 16S rRNA cDNA fragment based on the method described previously.24,1 For each DNA sample, PCR reaction mixture contained approximately 2 ng DNA template, 0.1 mM primers, 0.8 mM dNTPs, 2 U of Super-Taq Polymerase and SuperTaq buffer (HT Biotechnology, Cambridge, UK). The cycling conditions were: an initial denaturation of 3 min at 94 C followed by 25 cycles of 1 min at 94 C, 1 min at 50 C and 2 min at 72 C with a final extension of 10 min at 72 C. Bands of the correct size were verified by agarose gel electrophoresis. The PCR products were gel purified (Gel Extraction Kit, Qiagen) and approximately 5 ng DNA was cloned the same day using the TOPO TA cloning kit (Invitrogen) according to the manufacturer’s instructions. 6. Sequencing and Phylogenetic Analysis. Plasmid DNA (192 clones per library) was sequenced using the BigDyeTM terminator kit and the 3730xl Automatic Sequencer by Macrogen (Seoul, Korea) using primer 27F (50 -AGAGTTTGATCMTGGCTCAG-30 ). Electropherograms were treated for base calling and trimming using PHRED software.25 Chimera checking was performed using the CHECK_CHIMERA program.26 Sequences (918) were analyzed with the Ribosomal Database Project RDP - Release 10 (http://rdp.cme.msu.edu/) for alignment, taxonomic classification and library comparison.27 Hierarchical taxonomic classification of the libraries was performed using the ‘Naive Bayesian rRNA Classifier’ version 2.028 and the “Taxonomic Outline of the Bacteria and Archaea”, release 7.8 as RDP default’s engines. Clustering analysis into operational taxonomic units (OTUs) at a cutoff level 97% identity was performed with the Clusterer software using single linkage analysis,29 which generated 104 clusters (Supporting Information (SI) Table 2). OTUs were identified to the lowest taxonomic level possible by using the RDP taxonomic classifier only considering the classifications with a bootstrap value of > = 80% and by using the SILVA100 ARB database.30 Clone libraries were analyzed statistically with the PAST software.31 The similarities of the RNA libraries were analyzed using the BrayCurtis algorithm in a dendrogram taking into account OTU abundances. The sequence data from this study, one OTU representative, have been submitted to the GenBank database under the accession numbers: HQ327004-HQ327109.
’ RESULTS AND DISCUSSION Water Quality Survey of Intake and Treated Waters at Two Desalination Sites. Water quality indicators were investigated to
characterize raw seawater and to assess distinct water treatments performances in removing particulate, organic matter and microbial content in two desalination sites (pilot plants) using a polyphasic approach. The particulate and colloidal fouling potential of raw and pretreated waters were characterized by measurements of turbidity, which can be linked to the suspended solid content and the silt density index (SDI) together with other physicochemical parameters: temperature, pH and conductivity. These parameters were monitored and recorded throughout a one-year period (July 2007 until July 2008) to assess seasonal variations, and to have an overview of raw water quality feeding at both RO desalination sites. Water samples from the two sites (1 and 2) were collected at different locations corresponding to the raw water (RW) and treated waters: after dual media filtration 5945
dx.doi.org/10.1021/es200799t |Environ. Sci. Technol. 2011, 45, 5943–5951
Environmental Science & Technology
ARTICLE
Table 1. Annual Variation of Water Quality Parameters Analyzed in Different Water Types (Raw and Treated) from Two Desalination Sites: Site 1 and Site 2 from July 2007 until July 2008a RW1 (n = 12)
DMF (n = 12)
MF (n = 12)
RW2 (n = 6)
SSF (n = 6)
Temperature (C)
17.3 ( 6.50
17.2 ( 6.25
17.2 ( 6.26
20.2 ( 4.70
20.1 ( 5.40
pH Conductivity (mS/cm)
8.2 ( 0.1 56.8 ( 2.75
6.7 ( 0.3 57.5 ( 1.4
6.7 ( 0.4 57.5 ( 1.5
8.1 ( 0.1 57.7 ( 0.5
8.1 ( 0.2 57.4 ( 0.9
0.33 ( 0.11
0.04 ( 0.00
0.03 ( 0.00
0.94 ( 0.99
0.48 ( 0.25
18.23 ( 1.60
3.35 ( 0.28
2.36 ( 0.42
20.28 ( 4.41
2.92 ( 0.74
UV254(m1)
1.72 ( 0.41
1.25 ( 0.28
1.38 ( 0.39
1.31 ( 0.14
1.39 ( 0.20
SUVA
1.56 ( 0.29
1.42 ( 0.34
1.43 ( 0.24
1.62 ( 0.45
1.91 ( 0.39
70.0 ( 12.5
76.7 ( 12.5
71.7 ( 17.5
61.7 ( 9.2
Turbidity (NTU) SDIb
TOC (μM)
93.3 ( 19.2
TN (μM)
37.1 ( 1.1
TOC/TN pigmentsc (μg/L)
2.60 ( 0.69 0.57 ( 0.39
0.05 ( 0.03
picophytoplanktond (Cells/mL)
1.2 104 ( 1.0 104
3.2 103 ( 3.1 103
total bacteria (Cells/mL)
3.3 10 ( 1.3 10
1.0 10 ( 7.9 10
5
na
5
5
na 0.03 ( 0.02 bd 1.4 10 ( 1.8 10
5
4
4
5.7 ( 1.4
7.1 ( 1.4
12.67 ( 3.81 0.21 ( 0.25
8.97 ( 2.36 0.04 ( 0.06
2.9 104 ( 3.1 104
1.1 104 ( 2.7 104
5.2 10 ( 2.3 10
5.8 104 ( 3.1 104
5
5
na, not analysed; bd, below detection limit; errors are (1δ. RW1, raw water from Site 1; DMF, water after a dual media filtration treatment in Site 1; MF, water after a microfiltration treatment in Site 1; RW2 raw water from Site 2; SSF,water after a slow sand filtration treatment in Site 2. b Silt index measured for 3 min in raw waters and 15 mm in pretreated waters. c Pigments analyzed = chlorophyll a and phaeophytin a. d Picophytoplancton detected= Synechococcus and Prochlorococcus a
(DMF) and a microfiltration (MF) at Site 1, and after a slow sand filtration (SSF) at Site 2. The frequencies of the samplings at each facility are detailed in the Materials and Methods section. At Site 1 intake seawater (RW1) temperature varied from 13.2 C on March 2008 to 21.2 C on September 2007, while other parameters such as turbidity, conductivity and SDI were less variable throughout the sampling period (Table 1, Figure 1A). The lower mean SDI values (2.4) in MF compared to DMF treatment (3.4), and the absence of major fluctuations in MF treated waters, indicates that MF is a more suitable pretreatment for removing particles in this Site (Table 1 and SI Figure SI1). In Site 2, turbidity, SDI and temperature showed greater seasonal variation in raw water: respectively, 105.3% (from 0.17 to 0.53 NTU), 21.7% (from 15.4 to 20.5) and 23.3% (from 14.7 C in Apr/08 to 23.8 C in Aug/08) (Figure 1C). The comparison of raw and effluent waters of the beach-well (SSF) demonstrated that there was no significant difference in temperature or conductivity (Table 1). The turbidity of the raw water was around 1 NTU against 0.4 NTU for the treated water. Similarly SDI values were lower for the treated water than for the raw water (Table 1) indicating that SSF treatment efficiently removed particles. The determination of natural organic matter (OM) content in seawater is essential to design an adapted pretreatment process upstream of the RO units, as OM is known to be involved in the early stages of biofilm establishment.32 Furthermore a study of organic matter removal through the different types of pretreatments could help to determine their efficiency. Broad OM composition (TOC, TN, and UV) analysis in samples from Site 1 revealed that these parameters varied more on a yearly basis for the RW1 sample than for treated waters (DMF and MF) as revealed by the standard deviation (Table 1). An increase of UV values in RW1 was observed in the summer period reaching 2.5 m1, whereas for the rest of the year, they barely reached 1.6 m1 (data not shown). The high UV values in summer were concomitant with elevated pigment concentrations in the same period (see below, Figure 2B). The calculated SUVA index was on average
