Article pubs.acs.org/est
Correlations between Environmental Variables and Bacterial Community Structures Suggest Fe(III) and Vinyl Chloride Reduction As Antagonistic Terminal Electron-Accepting Processes Noam Shani,†,§ Pierre Rossi,‡ and Christof Holliger†,* †
Laboratory for Environmental Biotechnology, School of Architecture, Civil and Environmental Engineering, Ecole Polytechnique Fédérale de Lausanne, EPFL ENAC IIE LBE, Station 6, 1015, Lausanne, Switzerland ‡ Central Environmental Laboratory, School of Architecture, Civil and Environmental Engineering, Ecole Polytechnique Fédérale de Lausanne, EPFL ENAC IIE CEL, Station 2, 1015, Lausanne, Switzerland S Supporting Information *
ABSTRACT: Natural attenuation of anaerobic aquifers contaminated with tetrachloroethene (PCE) often results in the accumulation of the intermediates cis-dichloroethene and vinyl chloride (VC) which are even more toxic than the parent compound. Reasons for this accumulation were investigated in a PCE-contaminated aquifer in which VC accumulation has previously been shown to occur using stable isotope techniques. Multifactorial analysis of bacterial community structure data and environmental variables showed that in general terminal electron-accepting processes were shaping the bacterial community structures. Both VC and Fe(III) reduction were key but antagonistic terminal electronaccepting processes. The phylogenetic affiliation of terminal restriction fragments (T-RFs), together with correlation analyses, showed that T-RFs having significant correlation with VC reduction were closely affiliated to the genus Dehalococcoides and to uncultured bacteria belonging to the “Lahn Cluster” within the class Dehalococcoidetes. A T-RF that negatively correlated with a “Lahn Cluster” T-RF was affiliated to the genus Rhodoferax that contains members identified as iron-reducing bacteria. The higher affinity of Fe(III)-reducing bacteria for hydrogen compared with VC-reducing bacteria might explain why VC accumulated locally at the studied site. In conclusion, the combination of molecular and numerical ecology approaches was helpful to identify reasons for the accumulation of toxic dechlorination intermediates and could become a useful tool for characterizing contaminated sites in general.
■
INTRODUCTION Monitored natural attenuation is an alternative strategy for the bioremediation of aquifers contaminated by chlorinated ethenes (CEs), such as tetrachloroethene (PCE) or trichloroethene (TCE).1 The main drawback of this technique is the frequent undesired accumulation of lower CEs, such as the toxic intermediates cis-dichloroethene (cDCE) and vinyl chloride (VC).2−5 An incomplete sequence of dechlorination or a reduced efficiency of the dechlorination process of the lower CEs were both identified as potential reasons for their accumulation.6 Incomplete or impeded reductive dechlorination has been formerly explained by (i) the absence of bacterial strains able to completely dechlorinate to ethene,7−9 (ii) the depletion of organic substrates used as electron donors by H2producing secondary fermenting guilds (syntrophs),10,11 and (iii) the competition for electron donors with other terminal electron-accepting processes (TEAPs).12−14 Reasons for the accumulation of lower CEs have been investigated in microcosm studies or enrichment cultures.15−18 Hoelen & Reinhard15 showed that sulfate reduction affected the rate of © 2013 American Chemical Society
dechlorination of lower CEs in microcosm experiments. Addition of toluene stimulated H2 production via fermentation and increased the dechlorination rate. Similar conclusions were obtained by Heimann et al. in mixed anaerobic cultures.16 Among experiments carried out in situ, Aeppli et al.19 investigated a PCE-contaminated site (Zuchwil Quaternary aquifer, Switzerland) using stable isotope analysis to determine the extent and the rates of the in situ reductive dechlorination process. Qualitative evidence of the organohalide respiration (OHR) of PCE was provided by the detection of significant concentrations of daughter compounds, such as TCE, cDCE, and VC, as well as by the observation of a gradual dechlorination pattern along a transect from the source zone to the fringe of the contaminant plume. Carbon isotopic Special Issue: Rene Schwarzenbach Tribute Received: Revised: Accepted: Published: 6836
October 5, 2012 March 3, 2013 March 13, 2013 March 13, 2013 dx.doi.org/10.1021/es304017s | Environ. Sci. Technol. 2013, 47, 6836−6845
Environmental Science & Technology
Article
Figure 1. Map of the Zuchwil testsite showing monitoring wells with colored circles (left). The source of the PCE contamination is shown with a star. The general groundwater flow direction is shown with a blue arrow. Colors of the circles indicate the local oxidation−reduction status at the monitoring location. Green: oxic/nitrate-reducing; orange: nitrate to Fe(III)-reducing; red: Fe(III)-reducing and brown: methanogenic. Concentrations of CEs measured for the two sampling campaigns (A: April and B: October 2008 respectively) are shown on the left.
low hydraulic conductivity ((1−8) × 10−5 m/s) and an average hydraulic gradient of 0.3%, which lead to Darcy groundwater velocities of 1−8 m/yr. Thirty-one groundwater samples were collected during two sampling campaigns (April and October 2008) from 17 monitoring wells offering a vast panel of contamination conditions from the highly contaminated source zone (RB1/99) to the plume fringe (KB6, P20) (Figure 1). Groundwater samples for chemical analyses were collected following the procedure described in Aeppli et al. 19 Concentrations of major anions and cations, Fe2+ and Mn2+, TOC, CEs, ethene and methane were measured and used to define the aquifer oxidation−reduction state using a dedicated Excel spreadsheet program.20,21 Additional groundwater samples were collected for the analysis of microbial communities. Detailed information on groundwater sampling and chemical analyses is provided in Supporting Information (SI) S3.1 and S3.2. Analysis of Bacterial Communities. Water samples were filtered through 0.2 μm autoclaved polycarbonate membranes (Isopore Membrane Filters, Millipore) with a mobile filtration system (Filter Funnel Manifolds, Pall Corporation). DNA was extracted using the PowerSoil DNA Extraction Kit (MoBio Laboratories) following the manufacturer’s instructions, except that the samples were processed in a bead-beater (Fastprep FP120, Bio101) at 4.5 m/s for 30 s after the addition of solution C1. Presence of OHR-related genes (16S rRNA and reductive dehalogenase genes) was assessed by PCR amplification with specific primers (see the detailed procedure in SI S3.3 and Table S1). BCS were analyzed using terminalrestriction fragment length polymorphism (T-RFLP). Both analytical process and downstream numerical treatment of TRFLP profiles were based on the protocol described in Rossi et al.22 with modifications (see SI S3.4).
signatures of CEs indicated that PCE was efficiently transformed at most sampling points, with degrees of transformation of more than 60%. Furthermore, isotopic mass balance based on the concentration-weighted 13C isotopic signatures of CEs allowed the assessment of the potential of VC accumulation. The conclusion of this study has been that VC was accumulating overall, while complete OHR process leading to ethene was occurring locally. The reasons for the VC accumulation have not been investigated further but it has been postulated to be the result of a competition between sulfate-reducing bacteria and OHR bacteria able to reduce VC.19 Based on the results obtained by Aeppli et al.,19 the investigations carried out in this study aimed at identifying the mechanisms leading to in situ accumulation of VC at the Zuchwil site. To this end, we developed a sequential investigation procedure involving a combination of molecular and numerical statistical analyses: (i) PCR detection of both OHR bacteria and reductive dehalogenase genes to evaluate the presence of the biochemical potential for complete CE dechlorination, (ii) determination of the correlations between the bacterial community structure (BCS) and environmental variables using multivariate statistics to elucidate the mechanisms responsible for the incomplete dechlorination of CEs, and (iii) identification of terminal restriction fragments (TRFs) and their corresponding phylotypes that significantly correlated with TEAP variables to validate the mechanisms proposed in the previous step of the procedure.
■
MATERIAL AND METHODS Groundwater Sampling and Chemical Analyses. The Zuchwil test site and the occurring natural attenuation process have been previously described in detail.19 The site has a very 6837
dx.doi.org/10.1021/es304017s | Environ. Sci. Technol. 2013, 47, 6836−6845
Environmental Science & Technology
Article
Figure 2. Heatmap of pairwise correlations between environmental variables. The color key of the correlations is shown on the left. Statistically significant correlations are indicated with a color surrounding the squares. Dark green: p < 0.001; light green: p < 0.01; orange: p < 0.05.
Identification of T-RFs. Cloning and Sequencing. Selected DNA samples were used to construct clone libraries when the presence of T-RFs showing significant correlations with selected environmental variables could be detected. Clone libraries were screened first by analyzing each clone separately with T-RFLP analysis. Clones producing a T-RF showing significant correlations with environmental variables were sequenced. As a control, clone sequences were digested in silico with HaeIII and compared to the experimental length measured by T-RFLP. Sequences showing a discrepancy higher than 5 bp between the in silico and the experimental T-RF were discarded. The remaining sequences were compared with those available in GenBank30 by BLAST31 searches to find the nearest relatives and percentage of gene sequence identity. Neighborjoining phylogenetic trees were constructed with ClustalX 2.132 and MEGA533 (SI Figures S2−S5). The detailed procedure is available in SI S3.5. Pyrosequencing and PyroTRF-ID. Six aquifer samples (AB2, AKB2, AKB6b, BKB1, BKB5b, and BRB1/99) were selected based on the clustering of all T-RFLP profiles (n = 30) using the Hellinger distance and the Ward algorithm, and analyzed by pyrosequencing (GATC, Germany) using a 454/Roche GSFLX Titanium Instrument (Roche, NJ). Amplicons of the V1− V3 region of the 16S rRNA genes were produced and purified as described above. PyroTRF-ID34 running on the Vital-IT platform (Vital-IT, Swiss Institute of Bioinformatics, Switzerland) was used for data analysis and for the unambiguous taxonomic affiliation of the T-RFs. Digital T-RFLP profiles were produced on the basis of 5234 reads per sample on average, with a mean length of 402 bp. The detailed procedure is available in SI S3.6.
Multivariate Statistical Analyses Based on T-RFLP Profiles. All multivariate statistical analyses were carried out with R and the package Vegan.23,24 Ruzicka dissimilarities were calculated between at least three replicate T-RFLP profiles per sample. The bacterial T-RFLP profile at the centroid (and thus showing the lowest dissimilarity with all related replicates) was selected for further analysis. Resulting bacterial profiles were Hellinger transformed,25 whereas environmental data sets were standardized.26 Relationships between environmental variables were assessed by calculating pairwise Pearson product-moment correlation coefficients as well as their p-values using a permutation test with 1000 permutations. Bacterial community and environmental data sets were jointly analyzed using multifactorial analysis (MFA).27,28 For specific analysis, the environmental data set was subdivided in three main groups of variables: (i) the proportion of each CE among the total amount of all CEs (variable group “%CEs”), (ii) environmental variables related to TEAPs (variable group “TEA”), and (iii) all remaining variables (variable group “OTHERS”) (see SI Tables S2 to S5 for details). Numerical correlations and p-values between the environmental and bacterial community data sets were obtained by computing a multivariate generalization of the squared Pearson correlation coefficient (so-called RV multivariate correlation coefficients)29 between the BCS data set and (i) the whole environmental data set, (ii) each defined group of environmental variables (“%CEs”, “TEA” and “OTHERS”), and (iii) single environmental variables. Spearman pairwise correlations were computed between single T-RFs and environmental variables so as to reveal significant correlations with ongoing biogeochemical processes. 6838
dx.doi.org/10.1021/es304017s | Environ. Sci. Technol. 2013, 47, 6836−6845
Environmental Science & Technology
■
Article
RESULTS
Table 1. RV Correlation Coefficients Calculated between Bacterial Community Data Set and Whole Environmental Data Set (ENV), Each Defined Group of Environmental Variables (%CEs, TEA, OTHERS), and Each Environmental Variable Separately
Presence of Biochemical Dechlorination Potential. Known OHR bacteria and vinyl chloride reductive dehalogenase genes were detected in almost all samples by end point PCR indicating that the biochemical potential for complete reductive dechlorination of CEs was present at the site. Desulf itobacterium sp. was detected in all sampling wells whereas Dehalococcoides sp. was detected in all but two wells (RB1/93 and KB7b). Dehalobacter sp. was detected in all but six wells (RB3/93, B2, KB4a, KB5b, KB6a, and KB6b), Desulfuromonas sp. in all but five wells (RB3/93, KB1, KB4b, KB6a, and KB7a), and Sulf urospirillum sp. in all but three wells (RB3/93, KB5a, and KB5b). Vinyl chloride reductive dehalogenase-encoding genes vcrA and bvcA were detected in almost all wells, except in RB3/93, KB7a, and KB7b for vcrA and RB3/93, KB3a, and KB7a for bvcA. Microcosms inoculated with aquifer material from different monitoring wells showed that the biochemical potential for complete dechlorination was expressed under optimal conditions resulting in production of ethene (SI Figure S6). Pairwise Correlations between Environmental Variables. The oxidation−reduction state analysis based on major anion and cation concentrations indicated that the main part of the contamination plume was under Fe(III)-reducing conditions (Figure 1). Pearson pairwise correlations were calculated to explore the relationships between environmental variables (Figure 2). “NO3−” and “SO42‑” were correlated negatively with all other terminal electron acceptor (TEA)related variables. Both were present in relatively high concentration in the oxic water of reference well RB3/93 and in much lower concentrations in the other monitoring wells of the Zuchwil aquifer. Sulfate reduction was apparently more active in sections of the aquifer with high organic carbon content, as indicated by the negative correlation between “SO42‑” and “TOC”. “Fe2+” was positively correlated with bivalent cations such as “Sr2+”, “Ca2+”, and “Mg2+”, but also with “%VC”, a variable which indicates whether VC accumulated or not at a specific sampling well. Overall, “Fe2+” and “%VC” were following an almost identical pattern of correlations. The correlation between “Fe2+” and “%cDCE” was not significant. “PCE” was positively correlated with “Eh”, and all CEs were correlated with each other, the strongest correlations being displayed by adjacent compounds in the sequence of dechlorination. More detailed information on measured environmental variables is provided in SI Tables S2 to S5. Relationships between Bacterial Community Structures and Environmental Variables. In total, 30 bacterial community profiles were obtained from samples taken in April (n = 17) and October (n = 13) and were used to determine correlations with the environmental variables presented above. RV correlation coefficients and p-values calculated between the bacterial community data set and environmental variables data sets are summarized in Table 1. An MFA ordination plot with environmental variables displayed in three groups, namely “%CEs”, “TEA” and “OTHERS”, is depicted in Figure 3. All three groups of variables were significantly correlated with BCS, “TEA” showing the highest correlation. The two “TEA” variables with highest RV correlation coefficients of 0.34 and 0.32 were “VC” and “Fe2+”, respectively. However, no significant correlation was obtained between these two variables (Spearman’s rank correlation ρ = 0.19, p = 0.31; Figure 2).
variable or group of variables
RV coefficient
p-valuea
significanceb
ENV %CEs TEA OTHERS PCE [nM] PCE [%] TCE [nM] TCE [%] cDCE [nM] cDCE [%] VC [nM] VC [%] NO3− Fe2+ Mn2+ SO42‑ TOC Mg2+ Ca2+ Sr2+ Na+ NH4+ K+ Cl− Ehc Temperature ECd pH
0.62 0.34 0.54 0.49 0.24 0.19 0.29 0.18 0.22 0.19 0.34 0.25 0.20 0.32 0.18 0.26 0.24 0.25 0.28 0.28 0.17 0.18 0.18 0.23 0.23 0.26 0.21 0.16
