w a t e r r e s e a r c h 6 5 ( 2 0 1 4 ) 3 6 2 e3 7 0

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.elsevier.com/locate/watres

Effects of microbial redox cycling of iron on cast iron pipe corrosion in drinking water distribution systems Haibo Wang, Chun Hu*, Lili Zhang, Xiaoxiao Li, Yu Zhang, Min Yang Key Laboratory of Aquatic Science and Technology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 10085, China

article info

abstract

Article history:

Bacterial characteristics in corrosion products and their effect on the formation of dense

Received 28 January 2014

corrosion scales on cast iron coupons were studied in drinking water, with sterile water

Received in revised form

acting as a reference. The corrosion process and corrosion scales were characterized by

21 July 2014

electrochemical and physico-chemical measurements. The results indicated that the

Accepted 26 July 2014

corrosion was more rapidly inhibited and iron release was lower due to formation of more

Available online 6 August 2014

dense protective corrosion scales in drinking water than in sterile water. The microbial community and denitrifying functional genes were analyzed by pyrosequencing and

Keywords:

quantitative polymerase chain reactions (qPCR), respectively. Principal component anal-

Bacterial community

ysis (PCA) showed that the bacteria in corrosion products played an important role in the

Corrosion scales

corrosion process in drinking water. Nitrate-reducing bacteria (NRB) Acidovorax and

Nitrate-reducing bacteria

Hydrogenophaga enhanced iron corrosion before 6 days. After 20 days, the dominant bac-

Microbial redox cycling

teria became NRB Dechloromonas (40.08%) with the protective corrosion layer formation. The Dechloromonas exhibited the stronger corrosion inhibition by inducing the redox cycling of iron, to enhance the precipitation of iron oxides and formation of Fe3O4. Subsequently, other minor bacteria appeared in the corrosion scales, including iron-respiring bacteria and Rhizobium which captured iron by the produced siderophores, having a weaker corrosion-inhibition effect. Therefore, the microbially-driven redox cycling of iron with associated microbial capture of iron caused more compact corrosion scales formation and lower iron release. © 2014 Elsevier Ltd. All rights reserved.

1.

Introduction

Iron pipes have been used in drinking water distribution systems (DWDSs) for over five centuries (McNeill and Edwards, 2001), and they usually are covered with deposits of corrosion products. The water quality may be deteriorated during

* Corresponding author. Tel.: þ86 10 62849628; fax: þ86 10 62923541. E-mail address: [email protected] (C. Hu). http://dx.doi.org/10.1016/j.watres.2014.07.042 0043-1354/© 2014 Elsevier Ltd. All rights reserved.

transport through DWDSs to the use point due to iron corrosion. For example, corrosion can produce suspensions of iron particles that give the tap water a red, brown, or yellow color, or a dirty appearance (Sarin et al., 2004a). In the past several years, several serious red water cases occurred in some cities of the United States owing to switching of source water (Imran et al., 2005; Brodeur et al., 2006). In

w a t e r r e s e a r c h 6 5 ( 2 0 1 4 ) 3 6 2 e3 7 0

October, 2008, red water appeared in some areas in a northern city of China, soon after 80% of the local source water was replaced by source water from a neighboring province. It was noticed that the areas suffering from red water were historically supplied with local groundwater, while the areas without red water were historically supplied with local surface water. Yang et al. (2012) reported that thick corrosion scales or densely distributed corrosion tubercles with higher Fe3O4 ratio were mostly found in pipes transporting surface water, but thin corrosion scales and hollow tubercles with higher content of g-FeOOH and FeCO3 were mostly discovered in pipes transporting groundwater. These indicated that the structure and composition of corrosion scales played important roles in changes of distribution water quality from the interaction of corrosion scales with finished water. Therefore, study of the formation mechanism of dense and stable corrosion scales on the surface of iron pipes is very useful for controlling the water quality in DWDSs. Sarin et al. (2004a) indicated that corrosion scales growth was a result of continued corrosion followed by a combination of precipitation and oxidation of corrosion products. Typical iron corrosion scales may be composed of goethite (a-FeOOH), lepidocrocite (g-FeOOH), magnetite (Fe3O4), siderite (FeCO3), ferric hydroxide (Fe(OH)3), ferrous hydroxide (Fe(OH)2), and calcium carbonate (CaCO3) (Peng et al., 2010; Sarin et al., 2004a; Tang et al., 2006). Several environmental factors affect the corrosion rates and composition of the corrosion products, such as water quality, flow conditions, and microorganisms (Beech and Slunner, 2004; Gerke et al., 2008; Sarin et al., 2001). Microbiologically influenced corrosion (MIC) has been of interest due to its complicated role in corrosion processes: it can either accelerate or inhibit corrosion. A prevailing theory of MIC holds that biofilms enhance corrosion by inducing the establishment of corrosion cells with microbial aerobic respiratory activity within biofilms (Landoulsi et al., 2008). However, some findings suggested that biofilms can protect the metal from corrosion under certain conditions (Chongdar et al., 2005; Dubiel et al., 2002). For example, corrosion inhibition was caused by the reduction of ferric ions to ferrous ions and enhanced consumption of oxygen, both of which derive from iron-respiring bacteria respiration (Dubiel et al., 2002). Moreover, sulfate-reducing bacteria (SRB) can cause severe corrosion problems of metal pipes. However, nitrate-reducing bacteria (NRB) will overcome SRB, because for a given electron donor, the energy gained from nitrate reduction is greater than the energy obtained from sulfate reduction. Elimination of SRB by NRB can inhibit hydrogen sulfide production and reduce the corrosion rate (Zarasvand and Rai, 2014). Although a few possible mechanisms involving the effect of biofilms on corrosion were proposed, little research had been reported on the effect of microorganisms in a complicated community on the elemental composition and crystalline phase of corrosion scales. Also, previous researchers mainly focused on the study of corrosion behavior with electrochemical methods, but little importance was attached to the effect of the microbial phase in a complicated environment on the corrosion process and stable dense corrosion scales formation.

363

The objective of this study was to study bacterial characteristics in corrosion products and their effect on the formation of stable corrosion scales on the surface of cast iron coupons in drinking water, with sterile water acting as a reference. The corrosion process and corrosion scales were characterized by electrochemical and physico-chemical measurements. Quantitative real-time PCR (qPCR) was used to monitor changes in the microbial community of corrosion scales according to specific groups: total bacteria by 16S rRNA gene and denitrifiers by the functional genes nosZ, nirK and nirS. Most probable number enumerations of Fe(III)-reducing and nitrate-dependent Fe(II)-oxidizing microorganisms were detected under different conditions for different corrosion scales biofilms. Pyrosequencing was used to monitor changes in diversity of the microbial community, including bacteria related to iron redox cycling.

2.

Materials and methods

2.1.

The tested water and water quality

Three kinds of water samples were used in the experiments. The drinking water was collected from a drinking water treatment plant in the north of China, which was treated by coagulation, flocculation, sedimentation, sand filtration, and biologically-activated carbon filtration (prior to entering the chlorine contact tanks). The sterile water was obtained by sterilizing the above drinking water at 120  C and 1 bar for 20 min. In addition, the third tested water was obtained by the addition of a given amount of Ca(HCO3)2 to the sterile water to obtain the same water parameters as the drinking water. Water quality parameters were measured according to standard methods (EPA of China, 2002) for the three tested waters (Table S1 and Table S2). The calculation methods for Langelier saturation index (LSI) and Ryznar stability index (RSI) are given in the supporting information. Differences of water quality were measured using analysis of variance (ANOVA) with a significance threshold of a ¼ 0.05.

2.2.

Coupon preparation and experiment design

Cast iron coupons were used in this study. The elemental composition of the cast iron coupons was C 19.08%, O 6.09%, Si 2.06%, Ca 0.58%, P 0.65%, S 1.60%, Fe 65%, Cu 1.98%, Mn 0.92%, Zn 2.04%. The surface of each coupon was polished with 1200grit emery paper, and each coupon was rinsed with sterile deionized water thrice, degreased in acetone, followed by sterilizing in 70% ethanol for 8 h, and then dried aseptically in a laminar flow cabinet. The coupons were exposed to UV light for 30 min prior to use. In the experiments, ten cast iron coupons, 80  15  5 mm, were immersed in covered 1.5 L glass fiber-reinforced plastic bottles filled with each of the three waters containing chlorine disinfection, respectively. For the effect of biofilms, the collected drinking water from water utilities was used without any treatment, and the chlorine residual was controlled at 0.5 mg/L with sodium hypochlorite solution. In the sterile experiment, the drinking water sterilized at 120  C was used with chlorine residual 0.5 mg/L. In the control experiment, the

364

w a t e r r e s e a r c h 6 5 ( 2 0 1 4 ) 3 6 2 e3 7 0

sterile water with the addition of Ca(HCO3)2 was used under otherwise identical conditions. The water was displaced with fresh water at 2 day intervals to simulate the intermittent water flow environment in actual pipes according to a reported method (Tang et al., 2006). Each experiment was run in triplicate.

2.3.

Electrochemical measurements

All electrochemical measurements were performed in a threeelectrode electrochemical cell, with a platinum electrode used as the counter electrode, and a saturated calomel electrode (SCE) as the reference electrode. A cast iron coupon with exposed surface area of 1 cm2 was used as the working electrode, and the coupon was collected from the bottles with drinking and sterile water at different times. The Tafel plots were recorded at a scan rate of 2 mV/s within the range of 1000 mV to 0 mV versus the open circuit potential (OCP), and analyzed to determine the corrosion potentials (Ecorr) and corrosion current densities (icorr) using an Electrochemical Workstation CHI 660 D (CH Instrument, China). Electrochemical impedance spectroscopy (EIS) measurements were made at the open circuit potential (OCP) using a 10 mV amplitude sinusoidal signal over frequencies ranging from 102e105 Hz. EG&G ZsimpWin software was used to fit the impedance data to a modified Randles circuit. All of the above experiments were repeated three times for reproducibility to characterize the corrosion of the cast iron coupons. Moreover, the corrosion rate, measured by the weight loss method, is given in the supporting information.

2.4.

Surface analysis

The sampling process was kept under a strictly aseptic and dark environment. After sampling, the coupons were dried by freeze-drying under vacuum conditions. Then, the corrosion products were scraped from the inner wall of the coupons with a clean razor blade for surface characteristics analysis. The surface characteristics and corrosion features were examined by Field Emission Scanning Electron Microscopy (FESEM) (Hitachi, SU8020). Crystalline phase composition was analyzed using an X-ray powder diffractometer (XRD, X'Pert PRO MPD; PANalytical, The Netherlands). A TriStar II 3020 surface area and porosity analyzer (Micromeritics, Norcross, GA) was used to determine the specific surface area and pore size of powder samples by N2 gas adsorption and desorption.

2.5.

DNA extraction and quantitative real time PCR

To characterize the microbial phase, corrosion scales with biofilm were collected from the coupons using a previously described method (Teng et al., 2008). DNA extraction was conducted using a FastDNA spin kit for soil (Qbiogene, Solon, OH). The quantitative real time polymerase chain reactions (qPCR) of 16S rRNA for total bacteria was performed using the primer pairs 1369F and 1492R with the probe 1389F according to Suzuki et al. (2000). In addition, genes encoding nitrite reductase (nirS or nirK) and nitrous oxide reductase (nosZ) are the most common molecular markers for denitrifiers. Therefore, the quantification of nirS, nirK, and nosZ genes was

performed using the primer pairs 1F and 5R, R3cd and cd3aF, Z-F and 1622R, respectively (Throback et al., 2004). The qPCR was performed according to reaction mixture and thermal cycling conditions described previously (Guo et al., 2013). Standard curves were generated with serial ten-fold dilution (109e102 copies per microliter) of the plasmids. All samples, including standards and negative controls, were performed in triplicate. The R2 and slopes of standard curves and amplification efficiency values for quantification were as follows: 0.994, 3.356 and 98.6% for nirK, 0.996, 3.388 and 97.3% for nirS, 0.997, 3.352 and 98.8% for nosZ, and 0.992, 3.409 and 96.5% for 16S rRNA, respectively. The quantity of selected genes was expressed as log10 gene copies per gram of corrosion scales.

2.6.

454 pyrosequencing

For gene library construction, DNA was amplified by PCR using primer sets 341F (50 -CCTACGGGAGGCAGCAG-30 ) and 1073R (50 -A CGAGCTGACGACARCCATG-30 ) for the V3eV6 region of the 16S rRNA gene. The pyrosequencing was carried out on a Roche massively parallel 454 GS-FLX Titanium sequencer (Roche 454 Life Sciences, Branford, CT, USA) according to standard protocols (Pinto et al., 2012). After sequencing, low-quality sequences were removed by eliminating those with length shorter than 150 bp and those containing any ambiguous bases. Barcodes and primers were then trimmed from the resulting sequences. PCR chimeras were filtered out using Chimera Slayer (Ma et al., 2013). The reads flagged as chimeras were extracted out, and the non-chimera reads then formed the effective reads for each sample. Operational taxonomic unit (OTU) clustering was performed by setting a 0.03 distance limit using the MOTHUR program. BLAST taxonomic classification down to the phylum, class and genus level was then performed using MOTHUR via the SILVA106 database with a set confidence threshold of 80%. Principal component analysis (PCA) has been used to correlate both physicochemical and clone library matrices (Bouskill et al., 2012). It was used to correlate the bacteria at genus level, corrosion current densities (icorr) and total iron concentration in waters in this paper by the CANOCO program (CANOCO 4.5 for Windows).

2.7.

Most probable number enumerations

The abundance of culturable nitrate-reducing, Fe(III)-reducing and nitrate-dependent Fe(II)-oxidizing bacteria of the biofilm inside corrosion scales of cast iron coupons were analyzed using a three-tube most probable number (MPN) technique. The method is given in supporting information.

3.

Results

3.1.

Water quality changes in different waters

Fig. 1 shows the total iron concentration in drinking and sterile water at different times. In drinking water, the total iron concentration was about 12 mg/L before 4 days. However, from 6 d to 20 d, it quickly decreased to about 2 mg/L, and after

365

w a t e r r e s e a r c h 6 5 ( 2 0 1 4 ) 3 6 2 e3 7 0

the direct release of iron from dissolution of corrosion scales at the sampling time (Mutoti et al., 2007). Before 40 days, the average ratio of [Fe2þ]/[Fe3þ] in sterile water (0.07) was similar to in drinking water (0.08) in Fig. S3 (p ¼ 0.59). After 40 days, the ratio in sterile water was about 0.10, while it was 0.03 in drinking water, which was significantly different between the two waters (p < 0.0001). The results indicated that after 40 days, corrosion scales in drinking water were more stable and had lower iron release compared to those in sterile water.

3.2. Corrosion process of cast iron coupons in different waters

Fig. 1 e Total iron concentration in drinking water and sterile water at different times. Error bars represent the standard deviation from the average of three replications.

20 d, it gradually decreased to below 1 mg/L. Notably, the total iron concentration arrived at a relatively stable value of 0.40 mg/L after 40 d, although there was a little fluctuation from 40 d to 120 d. In sterile water, the total iron concentration was about 23 mg/L before 6 days. After 6 d, it decreased to 10.66 mg/L at 10 d. From 12 d to 20 d, it was gradually decreased to 4.47 mg/L. From 20 d to 120 d, it fluctuated from 2 to 4 mg/L. Both tested drinking water and sterile water exhibited differences in pH (p ¼ 0.02), alkalinity (p < 0.001), Ca2þ (p < 0.001) and hardness (p < 0.001) by ANOVA analysis (Table S1). However, ANOVA analysis showed that LSI (p ¼ 0.98) and RSI (p ¼ 0.12), which indicated the potential of scale formation and corrosion (Melidis et al., 2007; Rawajfeh and Shamaileh, 2007), were the same between the waters. Moreover, both effluents exhibited almost the same pH (p ¼ 0.32) and alkalinity (p ¼ 0.82) during 2-day stagnation periods in DWDSs (Fig. S1), and the decrease of hardness was the same (about 30 mg CaCO3/L) for both waters throughout the experiment. The sterile water with the addition of Ca(HCO3)2 had the same water parameters as the tested drinking water (Table S2, p > 0.05). The total iron concentration in its effluents was same as that in the sterile water (p ¼ 0.40, Fig. S2). These results indicated that the differences in corrosion produced by the water chemical parameters of the tested waters in the experiments could be neglected. Therefore, the effect of the biofilm was examined under the following experiments. Furthermore, the turbidity and ferrous ion of the effluents showed the same changes as the total iron concentration in drinking and sterile waters (Fig. S3). ANOVA analysis showed that in the effluents, the total iron concentration (p ¼ 0.0008), turbidity (p ¼ 0.0002) and ferrous ion (p ¼ 0.0001) were significantly different between the drinking water and sterile water, respectively. Fe(III) was the main component of the total iron in water due to the oxidization by dissolved oxygen and chlorine residual in the bulk water (Sarin et al., 2004b), while Fe(II) was mainly from

The Tafel analytical data are shown in Table 1. The corrosion current densities, icorr, of cast iron coupons in drinking water increased from 22.68 mA cm2 at 2 d to 27.47 mA cm2 at 6 d, then decreased to 18.63 mA cm2 at 20 d. However, after 20 d, it decreased quickly to 2.79 mA cm2 at 40 d. From 40 d to 120 d, it gradually decreased and arrived at a relatively stable value of about 1 mA cm2. The corrosion potential, Ecorr, underwent a positive shift from 0.708 V to 0.496 V during the whole experiment, and stabilized at 0.50 V after 40 d. These results indicated that after 40 d, corrosion reached a stable period, and a protective layer formed on the cast iron coupons in drinking water. In contrast, in the sterile water, icorr remained at a relatively high value, and it increased from 29.66 mA cm2 at 2 d to 39.98 mA cm2 at 40 d. From 40 d, it decreased and reached 2.77 mA cm2 at 120 d. Moreover, during the whole experiment, Ecorr exhibited a positive shift from 0.757 V to 0.648 V, which was more negative than that in drinking water, implying the corrosion scales of cast iron were less stable in sterile water than those in drinking water. Correspondingly, Fig. S4 illustrates the impedance spectra of the corroded coupons in different waters. Clearly, in drinking water, the diameter of the impedance loops markedly increased with time, and was higher than that in sterile water during the whole experiment. To describe the impedance response of the corrosion phenomenon in different waters, an equivalent circuit (Fig. S5) was proposed to represent the formation of a corrosion product film on the surface of the cast iron coupons. In drinking water, the resistance of the passive film Rf increased from 1519 U at 2 d to 18160 U at 120 d, and the charge transfer resistance Rct also increased from 45 U at 2 d to

Table 1 e Tafel analysis of polarization curves of the cast iron coupons in drinking water and sterile water at different times. Exposure time (day)

2 6 20 40 60 90 120

Ecorr (V)

icorr (mA cm2)

Drinking water

Sterile water

Drinking water

Sterile water

0.708 0.708 0.662 0.529 0.518 0.496 0.506

0.757 0.755 0.727 0.718 0.702 0.692 0.648

22.68 27.47 18.63 2.79 2.41 0.86 1.35

29.66 35.58 25.24 39.98 11.38 4.37 2.77

366

w a t e r r e s e a r c h 6 5 ( 2 0 1 4 ) 3 6 2 e3 7 0

3.3. Morphology and structure of corrosion scales in different conditions

Fig. 2 e XRD patterns of the corrosion scales on cast iron coupons in (A) drinking water and (B) sterile water (G e goethite, L e lepidocrocite, M e magnetite, S e siderite, Ca e calcite).

8955 U at 120 d (Table S3). In contrast, in sterile water, Rf decreased from 897 U at 2 d to 389 U at 6 d, and increased to 888 U at 20 d, then decreased to 243 U at 40 d. After this, the Rf increased gradually and tended toward 4222 U at 120 d. Similarly, Rct decreased from 100 U at 2 d to 22 U at 40 d, and then increased and tended toward 3129 U at 120 d (Table S3). Both Rf and Rct of the cast iron coupons in drinking water were much higher than those in sterile water, respectively, indicating the lower corrosion rate and higher corrosion inhibition in drinking water. The same results were obtained by weight loss method in Fig. S6. The corrosion rate obviously decreased with corrosion time in two systems, and the greater corrosion decrease was observed in drinking water. The results indicated that the corrosion scales formed on the surfaces of cast iron coupons in drinking water were more protective against corrosion than those in sterile water, which was in agreement with the lower iron release in drinking water.

The XRD patterns showed the differences between the corrosion scales of the cast iron coupons in different waters (Fig. 2). In drinking water, the crystalline compounds were goethite (a-FeOOH) and lepidocrocite (g-FeOOH) before 6 days. Their peak intensities increased till 60 d, then decreased from 60 d to 120 d. Meanwhile, at 20 d, magnetite (Fe3O4) and calcite (CaCO3) appeared, and their peak intensities increased from 20 d to 120 d, indicating their crystalline growth. The results indicated that Fe3O4 was probably formed from the transformation of a-FeOOH and g-FeOOH in drinking water. In sterile water, the main compounds were a-FeOOH, g-FeOOH, CaCO3 and siderite (FeCO3). These peak intensities increased with increasing time from 20 d to 120 d. There was no Fe3O4 formation in corrosion scales through the whole experiment period in sterile water. Fig. S7 shows FESEM micrographs of corrosion scales on the surface of cast iron coupons in different waters. In drinking water, FESEM exhibited a predominant laminar structure for the corrosion scales at 6 d. However, a number of globular tubercles formed and became the main structure at 40 d. In contrast, in sterile water, the main morphology was flower-like crystals and the scales became porous with increasing corrosion time. Fig. S8 shows that the nitrogen adsorption/desorption isotherms of corrosion scales formed on iron coupons in drinking water and sterile water exhibited type IV nitrogen isotherms with hysteresis loops (Sing et al., 1985), which indicated that the corrosion scales had a mesoporous structure. Moreover, the mean pore size and pore volume of the sample in drinking water were smaller than those in sterile water (Table S4). Their Brunauer-Emmett-Teller (BET) surface areas were about 42.01 and 50.75 m2/g in drinking water and sterile water, respectively. The results indicated that the corrosion scales in drinking water were more stable and compact than the scales formed in sterile water. Therefore, it could be deduced that the bacteria in corrosion products could change the morphology of corrosion scales and induce the corrosion products to become thicker and more compact, enhancing the inhibition of iron corrosion and release.

3.4. Bacterial community structure and denitrifying functional gene abundance Pyrosequencing of the sample in drinking water yielded 4085, 6254, 5299 and 5953 effective sequence tags at 6 d, 20 d, 40 d and 60 d, respectively. Significant Good's coverage was achieved for all samples (Table S5). There were 442, 661, 739, 747 OTUs in the samples taken at different times using cutoffs levels of 3% (Fig. S9 and Table S5). The Shannon diversity index was 3.82, 4.23, 4.82 and 4.83 at different times, respectively (Table S5), and it could be inferred that the enriched OTUs in the bacterial community were distributed more evenly after 40 d. From the 454 high-throughput sequencing, the dominant bacterial phyla within corrosion scales were found to be proteobacteria (99.9e94.1% of the total sequences), followed by Bacteroidetes, planctomycetes and Actinobacteria. The class level classification for all sequences detected in different samples is

w a t e r r e s e a r c h 6 5 ( 2 0 1 4 ) 3 6 2 e3 7 0

shown in Fig. S10. The main subclasses of proteobacteria were Alphaproteobacteria and Betaproteobacteria. The relative abundance of the bacteria related to iron redox cycling in corrosion scales grouped by genus is shown in Fig. 3 and Table S6. At 6 d, the relative abundance of nitrate-reducing bacteria (NRB) was 41.81%, and among them, 32.32% was Acidovorax, a nitratereducing Fe(II)-oxidizing bacterium (Muehe et al., 2009), and 9.26% was Hydrogenophaga (Shapleigh, 2006), a hydrogenconsuming corrosive bacterium. At 20 d, the dominant genera were still NRB (47.30%). However, among them, Acidovorax decreased to 7.54%, while a new genus Dechloromonas (37.20%), which was related to the iron redox cycling (Weber et al., 2006a), became the dominant bacteria, with other two NRB Azospira (Weber et al., 2006a) (2.11%) and Hyphomicrobium (Stein et al., 2001) (0.25%) associated with the iron redox cycling. In addition, 3.18% iron-respiring bacteria appeared, including iron-reducing bacteria (IRB) Rhodobacter (Dobbin et al., 1996) (1.83%) and iron-oxidizing bacteria (IOB) Sediminibacterium (Wang et al., 2012) (1.35%). At 40 d, Dechloromonas increased to 46.08% and was still the dominant genus, with two minor NRB Azospira (2.87%) and Hyphomicrobium (3.38%). Meanwhile, the iron-respiring bacteria were at 3.64%, predominantly including Rhodobacter (1.54%), Rhodomicrobium (Kappler and Straub, 2005) (0.22%) and Sediminibacterium (1.88%). Rhizobia (Arif et al., 2012) including Bradyrhizobium, Mesorhizobium and Rhizobium, which could produce siderophores, increased to 1.12%. At 60 d, Dechloromonas was still the dominant genus, although it decreased to 30.76%. Azospira and Hyphomicrobium were 5.16% and 1.71%, respectively. The iron-respiring bacteria and Rhizobia increased to 7.73% and 5.28%, respectively. From the analysis above, the main NRB were related to microbial redox cycling of iron. Therefore, the denitrifying functional genes were analyzed quantitatively by qPCR. Fig. 4 showed the abundance of nirS, nirK, nosZ and 16S rRNA genes copies at different times. The 16S rRNA gene copy numbers stabilized at about 2.0  1011 gene copies/g, indicating the stability of the microbial biomass. However, the three denitrifying functional genes underwent different changes. The

Fig. 4 e qPCR of the 16S rRNA and the functional gene nirK, nirS and nosZ of the biofilm bacteria in corrosion scales on cast iron coupons at different times. Error bars represent the standard deviation from the average of three replications.

nirK gene stabilized at about 1.0  1010 gene copies/g. nosZ decreased gradually from 7.3  109 gene copies/g at 6 d to 1.0  109 gene copies/g at 60 d. nirS increased from 7.4  109 gene copies/g at 6 d to 2.7  1010 gene copies/g at 20 d, and then it stabilized at about 2.0  1010 gene copies/g. The total denitrifying gene numbers were 5.20%, 11.09%, 17.55% and 12.38% of 16S rRNA gene copies at 6 d, 20 d, 40 d and 60 d, respectively (Fig. S11). This trend was consistent with the abundance change of NRB Dechloromonas at different times. Furthermore, MPN showed that the abundance of culturable Fe(III)-reducing bacteria was approximately 9.3  106 cells/g, nitratedependent Fe(II)-oxidizing bacteria were approximately 5.1  1011 cells/g, and nitrate-reducing bacteria were approximately 7.0  109 cells/g in the corrosion scales biofilm (Table S7). This indicated that NRB could induce Fe redox cycling inside corrosion scales.

4.

Fig. 3 e Relative abundance of biofilm bacterial genera related to iron redox cycling in corrosion scales on cast iron coupons at different times.

367

Discussion

Notably, in sterile water, the resistance of the passive film Rf exhibited an increase and decrease, moreover, the charge transfer resistance Rct gradually decreased before 40 d. After that, the two parameters increased and tended toward stable values at 120 d. Correspondingly, the total iron concentration tended toward about 4 mg/L. The total iron concentration in waters of DWDSs was mainly from iron pipe corrosion and the corrosion scales solubility (Mutoti et al., 2007). Although the corrosion rate in the sterile water decreased with corrosion time, the main corrosion product, including FeCO3, a-FeOOH, g-FeOOH, was always unstable and less protective (Tang et al., 2006; Yang et al., 2012), moreover, FeCO3 is favorable to the release of iron ions (Tang et al., 2006). Thus, iron release from the corrosion scales was higher and fluctuated during the whole corrosion time in sterile water, so it did not correlate with the changes of icorr very well. These results indicated that the corrosion of cast iron coupons was proceeding throughout the experiment in sterile water.

368

w a t e r r e s e a r c h 6 5 ( 2 0 1 4 ) 3 6 2 e3 7 0

In contrast, in drinking water, the resistance of the passive film Rf decreased before 6 d, and then greatly increased, while the charge transfer resistance Rct exhibited a great increase before 90 d, and the two tended toward stable values which were much greater than those in sterile water at 120 d, and the iron release tended toward 0.4 mg/L. The formation of the stable corrosion product Fe3O4 decreased the iron release (Tang et al., 2006). Therefore, at this stage, the iron release mainly came from iron corrosion in drinking water, showing good correlation with the changes of icorr. The results indicated that microorganisms played an important role in the prompt corrosion inhibition and corrosion product formation. Different corrosion-related bacteria appeared in the corrosion scales at different corrosion times. At 6 d, both Acidovorax and Hydrogenophaga were dominant. Acidovorax was a nitratedependent Fe(II)-oxidizing bacterium, and its respiration resulted in promotion of corrosion and the precipitation of Fe(III) oxides (Muehe et al., 2009; Weber et al., 2006b), while Hydrogenophaga could induce corrosion by hydrogen consumption (Videla and Herrera, 2009). Therefore, the two bacteria increased corrosion and precipitation of iron oxides in this period, leading to lower iron release compared with the sterile water. PCA analysis by correlation of the bacteria, the corrosion current density (icorr) and total iron concentration in drinking water also verified that Acidovorax and Hydrogenophaga were positively associated with icorr and total iron concentration (Fig. 5). This enhanced corrosion of cast iron coupons and the iron release were higher in drinking water before 6 days. However, after 6 d, the main corrosion-related bacteria in corrosion products became Dechloromonas (37.20%), and it exhibited more negative correlation with icorr and total iron concentration by PCA analysis (Fig. 5). The result confirmed that this bacteria played a stronger inhibition role in iron corrosion and release. MPN results indicated that the NRB of biofilm inside corrosion scales could induce Fe redox cycling in corrosion

scales. Furthermore, the transformation of NO3  eN and NH3eN in both waters also illustrated this point (Table S8). In sterile water, the decrease of NO3  eN varied from 0.52 mg/L to 0.71 mg/L, while the increase of NH3eN varied from 0.48 mg/L to 0.68 mg/L. The result indicated that 94.4% of the NO3  eN decrease was transformed to NH3eN by iron corrosion in sterile water (Kielemoes et al., 2000). However, the same phenomena did not occur with iron corrosion in drinking water. Few changes of NO3  eN and NH3eN were observed in the effluents, and less NO3  eN was transformed to NH3eN, indicating a bacterial denitrification role, because NO3  eN were denitrified into N2 (Kielemoes et al., 2000). It has been reported that in NO3  eN concentrations less than 7 mg/L, the oxidation and reduction of iron occurred simultaneously by the respiration of NRB Dechloromonas, resulting in Fe3O4 formation (Coby et al., 2011; Weber et al., 2006a,b). Under the experimental conditions, most Dechloromonas had the denitrifying function, and the concentration of NO3  eN was less than 2 mg/L in the drinking water. The cycling of Fe(II)/Fe(III) was driven by Dechloromonas respiration, leading to lower reduction of NO3  eN. The process enhanced the consumption of oxygen, precipitation of iron oxide and the formation of Fe3O4 in corrosion scales, causing more compact corrosion scale formation and higher inhibition of corrosion and iron release. At 40 d, the proportion of NRB Dechloromonas reached its highest value (40.08%). A number of globular tubercles formed in the corrosion scales and the iron release decreased to a stable value. At 60 d, although the abundance of NRB Dechloromonas decreased, it was still the dominant bacteria (30.76%). With the corrosion scales becoming more compact, the iron-respiring bacteria including Rhodobacter (IRB), Rhodomicrobium (IOB) and Sediminibacterium (IOB), and Rhizobia including Bradyrhizobium, Mesorhizobium and Rhizobium increased to 7.73% and 5.28%, respectively. The iron-respiring bacteria and Rhizobia exhibited less negative correlation with icorr and total iron concentration in water than Dechloromonas by PCA analysis. However, iron-respiring bacteria respiration could induce corrosion and iron release inhibition by the reduction of ferric ions to ferrous ions and consumption of oxygen, which also contributed to iron redox cycling (Dubiel et al., 2002). Moreover, Rhizobia could produce and import siderophores (e.g. ferrichrome) that capture iron to inhibit the dissolution of iron and iron corrosion (Grateron et al., 2007; Little et al., 2007). The combined action of NRB Dechloromonas, iron-respiring bacteria and Rhizobia induced the dense stable corrosion scale formation and higher inhibition of corrosion and iron release in drinking water, compared with that in sterile water. Therefore, with the increasing stability of the protective layer, the diversity of bacteria related to iron redox cycling increased, and on the other hand, the bacterial combined action accelerated the dense corrosion scales formation and maintained protective layer stability.

5. Fig. 5 e Principal component analysis (PCA) of corrosion current density (icorr), total iron concentration and the abundance of biofilm bacterial genera in corrosion scales on cast iron coupons at different times.

Conclusions

The results indicated that corrosion was more rapidly inhibited and iron release was lower due to formation of more stable and dense protective corrosion scales on cast iron coupons in

w a t e r r e s e a r c h 6 5 ( 2 0 1 4 ) 3 6 2 e3 7 0

drinking water than on those in sterile water. PCA showed that the bacteria in corrosion products played an important role in the corrosion process in drinking water. Nitratereducing bacteria (NRB) Acidovorax and Hydrogenophaga enhanced iron corrosion with the oxidation of Fe(II) before 6 days. After 20 days, the dominant bacteria became NRB Dechloromonas (40.08%) with the protective corrosion layer formation. The Dechloromonas exhibited stronger corrosion inhibition by inducing the redox cycling of iron to enhance the precipitation of iron oxides and formation of Fe3O4. Subsequently, both iron-respiring bacteria and Rhizobium which captured iron by the produced siderophores, appeared to have a weaker corrosion-inhibition effect. Therefore, the microbially driven redox cycling of iron with associated microbial capturing of iron, enhanced the formation of more compact and stable corrosion scales to inhibit iron corrosion and release in drinking water.

Acknowledgments This work was funded by the National Natural Science Foundation of China (Nos. 51308529, 21125731, 51290281), the Federal Department of Chinese Water Control and Treatment (No. 2012ZX07404002) and the project of Chinese Academy of Sciences (YSW2013A02).

Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.watres.2014.07.042.

references

Arif, K., Archana, G., Anjana, D.J., 2012. Engineering heterologous iron siderophore complex utilization in rhizobia: effect on growth of peanut and pigeon pea plants. Appl. Soil. Ecol. 53, 65e73. Beech, I.B., Slunner, J., 2004. Biocorrosion: towards understanding interactions between biofilms and metals. Curr. Opin. Biotechnol. 15, 181e186. Bouskill, N.J., Eveillard, D., Chien, D., Jayakumar, A., Ward, B.B., 2012. Environmental factors determining ammonia-oxidizing organism distribution and diversity in marine environments. Environ. Microbiol. 14, 714e729. Brodeur, T., Davis, F.S., Florence, R., Kim, M., Craig, M., Gianatasio, J., Sharp, D., Lowe, P., 2006. From red water to pump failures-corrosion control activities and related studies. Fla. Water Resour. J. 12, 42e48. Chongdar, S., Gunasekaran, G., Kumar, P., 2005. Corrosion inhibition of mild steel by aerobic biofilm. Electrochim. Acta 50, 4655e4665. Coby, A.J., Picardal, F., Shelobolina, E., Xu, H.F., Roden, E.E., 2011. Repeated anaerobic microbial redox cycling of iron. Appl. Environ. Microbiol. 77, 6036e6042. Dobbin, P.S., Warren, L.H., Cook, N.J., McEwan, A.G., Powell, A.K., Richardson, D.J., 1996. Dissimilatory iron(III) reduction by Rhodobacter capsulatus. Microbiology 142, 765e774.

369

Dubiel, M., Hsu, C.H., Chien, C.C., Mansfeld, F., Newman, D.K., 2002. Microbial iron respiration can protect steel from corrosion. Appl. Environ. Microbiol. 68, 1440e1445. EPA of China, 2002. Analysis Method for Water and Waste Water, fourth ed. Press of Chinese Environmental Science, Beijing. Gerke, T.L., Maynard, J.B., Schock, M.R., Lytle, D.L., 2008. Physiochemical characterization of five iron tubercles from a single drinking water distribution system: possible new insights on their formation and growth. Corros. Sci. 50, 2030e2039. Grateron, C., Barbosa, O., Rueda, N., Ortiz-Lopez, C., Torres, R., 2007. AS7, a bacterial siderophore valuable as a high performance corrosion inhibitor. J. Biotechnol. 131, S87eS88. Guo, G.X., Deng, H., Qiao, M., Yao, H.Y., Zhu, Y.G., 2013. Effect of long-term wasterwater irrigation on potential denitrification and denitrifying communities in soils at the watershed scale. Environ. Sci. Technol. 47, 3105e3113. Imran, S.A., Dietz, J.D., Mutoti, G., Taylor, J.S., Randall, A.A., Cooper, C.D., 2005. Red water release in drinking water distribution systems. J. Am. Water Works Assoc. 97, 93e100. Kappler, A., Straub, K.L., 2005. Geomicrobiological cycling of iron. Rev. Mineral. Geochem. 59, 85e108. Kielemoes, J., Boever, P.D., Verstraete, W., 2000. Influence of denitrification on the corrosion of iron and stainless steel powder. Environ. Sci. Technol. 34, 663e671. Landoulsi, J., Kirat, K.E., Richard, C., Feron, D., Pulvin, S., 2008. Enzymatic approach in microbial-influenced corrosion: a review based on stainless steels in natural waters. Environ. Sci. Technol. 42, 2233e2242. Little, B., Lee, J., Ray, R., 2007. A review of 'green' strategies to prevent or mitigate microbiologically influenced corrosion. Biofouling 23, 87e97. Ma, J.X., Wang, Z.W., Yang, Y., Mei, X.J., Wu, Z.C., 2013. Correlating microbial community structure and composition with aeration intensity in submerged membrane bioreactors by 454 high-throughput pyrosequencing. Water Res. 47, 859e869. McNeill, L.S., Edwards, M., 2001. Iron pipe corrosion in distribution systems. J. Am. Water Works Assoc. 93, 88e100. Melidis, P., Sanozidou, M., Mandusa, A., Ouzounis, K., 2007. Corrosion control by using indirect methods. Desalination 213, 152e158. Muehe, E.M., Gerhardt, S., Schink, B., Kappler, A., 2009. Ecophysiology and the energetic benefit of mixotrophic Fe(II) oxidation by various strains of nitrate-reducing bacteria. FEMS Microbiol. Ecol. 70, 335e343. Mutoti, G., Dietz, J.D., Imran, S.A., Taylor, J., Cooper, C.D., 2007. Development of a novel iron release flux model for distribution systems. J. Am. Water Works Assoc. 93, 102e111. Peng, C.Y., Korshin, G.V., Valentine, R.L., Hill, A.S., Friedman, M.J., Reiber, S.H., 2010. Characterization of elemental and structural composition of corrosion scales and deposits formed in drinking water distribution systems. Water Res. 44, 4570e4580. Pinto, A., Xi, C., Raskin, L., 2012. Bacterial community structure in the drinking water microbiome is governed by filtration processes. Environ. Sci. Technol. 46, 8851e8859. Rawajfeh, A.E.A., Shamaileh, E.M.A., 2007. Assessment of tap water resources quality and its potential of scale formation and corrosivity in Tafila Province, South Jordan. Desalination 206, 322e332. Sarin, P., Snoeyink, V.L., Bebee, J., Kriven, W.M., Clement, J.A., 2001. Physico-chemical characteristics of corrosion scales in old iron pipes. Water Res. 35, 2691e2697. Sarin, P., Snoeyink, V.L., Lytle, D.A., Kriven, W.M., 2004a. Iron corrosion scales: model for scale growth, iron release, and colored water formation. J. Environ. Eng. 130, 364e373.

370

w a t e r r e s e a r c h 6 5 ( 2 0 1 4 ) 3 6 2 e3 7 0

Sarin, P., Snoeyink, V.L., Bebee, J., Jim, K.K., 2004b. Iron release from corroded iron pipes in drinking water distribution systems: effect of dissolved oxygen. Water Res. 38, 1259e1269. Shapleigh, J.P., 2006. The denitrifying prokaryotes. In: Dworkin, M., Falkow, S., Rosenberg, E., Schleifer, K.-H., Stackebrandt, E. (Eds.), The Prokaryotes. Springer Press, New York, USA, pp. 769e792. Sing, K.S.W., Everett, D.H., Haul, R.A.W., Moscou, L., Pierotti, R.A., Rouquerol, J., Siemieniewska, T., 1985. Reporting physisorption data for gas/solid systems with special reference to the determination of surface area and porosity. Pure Appl. Chem. 57, 603e619. Stein, L.Y., La Duc, M.T., Grundl, T.J., Nealson, K.H., 2001. Bacterial and archaeal populations associated with freshwater ferromanganous micronodules and sediments. Environ. Microbiol. 3, 10e18. Suzuki, M.T., Taylor, L.T., DeLong, E.F., 2000. Quantitative analysis of small-subunit rRNA genes in mixed microbial populations via 50 -nuclease assays. Appl. Environ. Microbiol. 66, 4605e4614. Tang, Z.J., Hong, S.K., Xiao, W.Z., Taylor, J., 2006. Characteristics of iron corrosion scales established under blending of ground, surface, and saline waters and their impacts on iron release in the pipe distribution system. Corros. Sci. 48, 322e342. Teng, F., Guan, Y.T., Zhu, W.P., 2008. Effect of biofilm on cast iron pipe corrosion in drinking water distribution system: corrosion scales characterization and microbial community structure investigation. Corros. Sci. 50, 2816e2823.

Throback, I.N., Enwall, K., Jarvis, A., Hallin, S., 2004. Reassessing PCR primers targeting nirS, nirK and nosZ genes for community surveys of denitrifying bacteria with DGGE. FEMS Microbiol. Ecol. 49, 401e407. Videla, H.A., Herrera, L.K., 2009. Understanding microbial inhibition of corrosion. A comprehensive overview. Int. Biodeterior. Biodegrad. 63, 896e900. Wang, H.B., Hu, C., Hu, X.X., Yang, M., Qu, J.H., 2012. Effects of disinfectant and biofilm on the corrosion of cast iron pipes in a reclaimed water distribution system. Water Res. 46, 1070e1078. Weber, K.A., Urrutia, M.M., Churchill, P.F., Kukkadapu, R.K., Roden, E.E., 2006a. Anaerobic redox cycling of iron by freshwater sediment microorganisms. Environ. Microbiol. 8, 100e113. Weber, K.A., Achenbach, L.A., Coates, J.D., 2006b. Microorganisms pumping iron: anaerobic microbial iron oxidation and reduction. Nat. Rev. Microbiol. 4, 752e764. Yang, F., Shi, B.Y., Gu, J.N., Wang, D.S., Yang, M., 2012. Morphological and physicochemical characteristics of iron corrosion scales formed under different water source histories in a drinking water distribution system. Water Res. 46, 5423e5433. Zarasvand, K.A., Rai, V.R., 2014. Microorganisms: induction and inhibition of corrosion in metals. Int. Biodeterior. Biodegrad. 87, 66e74.