Article pubs.acs.org/est
Microbial Community Dynamics during Acetate Biostimulation of RDX-Contaminated Groundwater Joshua A. Livermore,† Yang Oh Jin,† Richard W. Arnseth,‡ Michael LePuil,‡ and Timothy E. Mattes†,* †
Department of Civil and Environmental Engineering, 4105 Seamans Center, University of Iowa, Iowa City, Iowa 52242, United States ‡ TetraTech, Inc., Oak Ridge, Tennessee 37830, United States S Supporting Information *
ABSTRACT: Biostimulation of groundwater microbial communities (e.g., with carbon sources) is a common approach to achieving in situ bioremediation of organic pollutants (e.g., explosives). We monitored a field-scale approach to remediate the explosive RDX (hexahydro-1,3,5trinitro-1,3,5-triazine) in an aquifer near the Iowa Army Ammunition Plant in Middletown, IA. The purpose of the study was to gain insight into the effect of biostimulation on the microbial community. Biostimulation with acetate led to the onset of RDX reduction at the site, which was most apparent in monitoring well MW309. Based on previous laboratory experiments, we hypothesized that RDX degradation and metabolite production would correspond to enrichment of one or more Fe(III)reducing bacterial species. Community DNA from MW309 was analyzed with 454 pyrosequencing and terminal restriction fragment length polymorphism. Production of RDX metabolites corresponded to a microbial community shift from primarily Fe(III)-reducing Betaproteobacteria to a community dominated by Fe(III)-reducing Deltaproteobacteria (Geobacteraceae in particular) and Bacteroidetes taxa. This data provides a firsthand field-scale microbial ecology context to in situ RDX bioremediation using modern sequencing techniques that will inform future biostimulation applications.
■
INTRODUCTION
A promising carbon source for RDX bioremediation is acetate.3 A wealth of field and laboratory studies has shown that acetate amendment is likely to enrich for Fe(III)-reducing organisms1,9,15−18 (e.g., Geobacter spp.) and competitively exclude sulfur reducers.19,20 Furthermore, multiple independent laboratory experiments have shown that Fe(III)-reducing bacteria transform RDX either by direct reduction, or indirectly by reaction between RDX and the Fe(II) that is produced during Fe(III) respiration or electron shuttles.7,21−24 Based on these two observations, it is reasonable to expect RDX degradation in aquifers during acetate amendment would correspond to enrichment of Fe(III)-reducing bacteria. However, no field investigation has yet documented a link between an Fe(III)-reducing taxon and in situ RDX transformation. To contribute a field-based perspective, we investigated a bioremediation application at the Iowa Army Ammunition Plant (IAAAP) in Middletown, Iowa which was implemented to remove a dilute aquifer RDX plume that migrated off-site. We collected groundwater samples from the RDX-contaminated aquifer before, during, and after acetate injections from
Injecting carbon substrates into groundwater is a strategy often used to stimulate microbial activity in the subsurface, alter terminal electron accepting conditions, and promote anaerobic reduction of groundwater pollutants such as radionuclides1,2 and explosives.3 One particular priority contaminant being treated with this bioremediation strategy is the explosive RDX (hexahydro-1,3,5-trinitro-1,3,5-triazine). The practice of carbon injection is based on observed biologically mediated RDX decomposition,3,4 which could proceed by a variety of pathways, including sequential denitration,5 direct enzymatic cleavage,6 reduction by biogenic Fe(II), 7 and others.3,8 Previous studies aimed at identifying RDX-degraders in enrichment culture have implicated Geobacter, Acetobacteria, and Clostridia species;9 Gordonia and Williamsia species;10 and Desulfovibrio species.11 Additionally, stable isotope probing (SIP) studies have implicated Sphingobacteria,12 Enterobacter, and Pseudomonas,13 and various Proteobacteria, Bacteroidia, Clostridia, and Spirochetes.14 A single field-based study examined the prevailing in situ microbial community during RDX bioremediation by cheese whey addition in one aquifer and a mulch biowall in another.4 In contrast to potential RDXdegraders implicated in enrichment cultures and SIP experiments, a diverse array of opportunist heterotrophs was observed. © 2013 American Chemical Society
Received: Revised: Accepted: Published: 7672
March 22, 2013 June 4, 2013 June 19, 2013 June 19, 2013 dx.doi.org/10.1021/es4012788 | Environ. Sci. Technol. 2013, 47, 7672−7678
Environmental Science & Technology
Article
extracted from the filter pieces using the MoBio UltraClean soil DNA kit with the cell lysis step performed on a Biospec minibeadbeater 8 for 2 minutes. For high-throughput sequencing, the forward primers 8fm and 1492r were used to amplify the entire 16S rRNA gene then 0.5 μL of this reaction was used as the template for a second reaction with primers targeting the V1−V3 region 26 modified with adapters for the 454 amplicon pyrosequencing reaction and barcodes for sample identification (SI Table S3). We sent gel purified reamplifications to the University of Iowa DNA facility for quality control testing on an Agilent Bioanalyzer (Santa Clara, CA) and subsequent sequencing on the 454 GS FLX instrument (Branford, CT). Quality filtered sequence reads were processed using Quantitative insights into microbial ecology (Qiime).27 Samples were filtered to retain sequences between 200 and 1000 bases with quality scores greater than 25 units and no more than three primer base mismatches. Operational taxonomic units (OTUs) were selected using Uclust 28 at a 97% sequence similarity threshold and thereafter were represented by the most abundant sequence in each cluster. Chimeric OTUs were identified using ChimeraSlayer 29 and only nonchimeric OTUs were retained. For TRFLP we used 16S primers targeting the V1−V3 region with forward primer tagged with a 5′ 6-FAM fluorophore. PCR product concentrations were measured using the dsDNA broad range kit for the Qubit fluorometer (Invitrogen, Carlsbad, CA). Primer sequences and thermocycling conditions are provided in SI Table S3. Following amplification, purified PCR product (1 μg) was digested with MspI (New England Biolabs, Ipswich, MA) at 37 °C for four hours then desalted by glycogen precipitation. Digested PCR product (70 ng) was sent to the University of Iowa DNA facility for electrophoresis on an Applied Biosystems 3730 DNA analyzer (Life Technologies Corporation, Carlsbad, CA) with the GeneScan 500 LIZ size standard. Output was analyzed using Peak Scanner 1.0 (Applied Biosystems) followed by TREX 30 yielding a table with each terminal restriction fragment (TRF) treated as an operational taxonomic unit (OTU) with abundance defined by measured peak area. TRFLP peak abundances were averaged across replicates where available. Phylogenetic Identification of Sequences and Terminal Restriction Fragments. Representative pyrosequencing OTU sequences were assigned to the Phylum taxonomic level using the RDP classifier (implemented in Qiime) with the Greengenes core set as the source database.31 Precise sequence identification was achieved using phylum specific maximum likelihood phylogenies. OTU sequences and environmentally relevant reference sequences were aligned using the SINA aligner on the SILVA Web site.32 Alignments were manually inspected in Seaview and phylogeny computed using PhyML33 with optimal parameters determined by the ModelGenerator software and assessed confidence using 500 bootstrap replicates. To phylogenetically identify terminal restriction fragments (TRFs), we predicted fragment sizes of pyrosequencing OTUs using TRiFLE.34 Additionally, we produced clone libraries from two samples (MW309 (4/22/2009) and EMW02 (5/5/2009)) using the Topo TA 2.1 cloning kit (Invitrogen, Carlsbad, CA). Clones were sequenced at the University of Iowa DNA facility, and classified using the RDP classifier.35 Clones were also digested and fragment analyzed as above, and TRF lengths predicted using TRiFLE. We used logistic regression to evaluate whether Fe(III)-reducing Betaproteobacterial or
injection sites and wells up-gradient and down-gradient from injections. Using 454 pyrosequencing, terminal restriction fragment length polymorphism (TRFLP) analysis and clone libraries, we tracked variations in microbial community structure that resulted from aquifer acetate biostimulation and compared these with concurrent changes in geochemical parameters. This approach provides new insights into the relationship between microbial community dynamics and RDX reduction at the field scale in a biostimulated aquifer.
■
MATERIALS AND METHODS Site Characterization. The Iowa Army Ammunition Plant located near Burlington, IA (40° 48′ N, 91° 15′ W) is the origination point of RDX groundwater contamination. Located within the Mississippi River valley, the contaminated aquifer is of alluvial origin and composed of medium to coarse grain sized sands. Groundwater flow velocities range approximately between 25 and 40 m/yr based on slug tests, measurements of hydraulic gradient, and porosity.25 Our investigation focused on wells in this aquifer within a one mile radius (Supporting Information (SI) Figure S1). Acetate was injected into wells SMW04-SMW08 and SMW10, whereas MW309, EMW02, and EMW04-EMW07 are monitoring wells located down-gradient of the injection site and are thus subject to impact by acetate. MW505 and MW502 are background wells located up-gradient of the injection wells. As of the final five year site review (March 2011), the acetate amendment strategy was performing adequately to attain remediation goals. Additional information can be found in the EPA region 7 cleanup record (www.epa. gov/region7/cleanup) or the Iowa Army Ammunition plant administrative record (iaaap.adminrecord.com). Sampling Methodology. Groundwater samples were taken before, during, and after liquid sodium acetate injections which occurred in October 2007 (1000 kg), January 2008 (1500 kg), March 2008 (545 kg), August 2008 (1180 kg), and April 2009 (1315 kg). Groundwater wells were low-flow purged (99% Deltaproteobacteria), RDX and Fe(II) were detectable, DO was low, and RDX metabolites were at their highest measured concentrations (Figure 3, SI Table S5). At the other extreme (sample EMW06 on 9/28/2007) Betaproteobacteria Fe(III)-reducers were more abundant (100%), RDX and Fe(II) were detectable, and DO was very low, but only two of the RDX metabolites were detectable and at very low concentrations (Figure 3, SI Table S5). Logistic regressions based on TRF abundance and RDX metabolite detections in 34 samples from 14 wells indicate this relationship is statistically significant (Table 1). The proportion of Fe(III)reducers that were Deltaproteobacteria (Geobacter spp.) was the most significant predictor variable (p = 0.009, 0.019, and 0.034 for MNX, DNX, and TNX, respectively). While this statistical relationship establishes a link between Fe(III)-reducing Deltaproteobacteria and RDX metabolites at this site, additional studies are necessary to establish this relationship at other sites. We observed predicted TRFs associated with Fe(III)reducing bacteria in the background wells MW502 and MW505. Specifically, we identified a 92 bp TRF as Albidiferax ferrireducens in MW505 on 7/21/2009, and a 121 bp TRF as an unclassified Rhodocyclales in MW505 on 5/5/2009), and a 162 bp TRF as Geobacter sp. in well MW502 on 10/10/2008 and 12/5/2008. Each of these putative Fe(III)-reducing species were rare members of the microbial community in these samples (70%) of Bacteroidetes sequences that were enriched were represented by OTUs in unique Flavobacteriales groups (SI Figure S5). Flavobacteria are known to reach high abundance in aquatic habitats following addition of organic materials and are thought to be opportunistic heterotrophs.39 Other Bacteroidetes OTUs were placed in Chitinophagaceae and Saprospiraceae (both class Sphingobacteriales), clade WchB1-32, and sequences in a unified clade containing sequences from clades WchB1-69 and KD1-131 (SI Figure S4). Members of WchB1-69 are found in other Fe(II) reducing environments40,41 but are not known to be Fe(III)-reducers. The level of Bacteroidetes stimulation observed here is contrary to 7675
dx.doi.org/10.1021/es4012788 | Environ. Sci. Technol. 2013, 47, 7672−7678
Environmental Science & Technology
Article
to RDX degradation and accumulation of the metabolites MNX, DNX, and TNX. While our data indicates that the level of acetate addition permitted Deltaproteobacterial Fe(III)reducers to outcompete Betaproteobacterial Fe(III)-reducers at this site, by itself this observation is insufficient to establish a link to RDX biodegradation. We established this link by showing a statistically significant association between RDX reduction product detection and the shift between dominant Fe(III)-reducing taxa (Figure 3, SI Table S5). Given that certain Betaproteobacterial Fe(III)-reducers can transform RDX (e.g., Albidiferax (Rhodoferax) ferrireducens14,17) it is possible they also catalyzed RDX degradation at this site, perhaps producing a set of metabolites we did not measure. However, we could not determine such a relationship from our data as fluctuating influent concentrations of RDX preclude inferences based on covariance between RDX concentrations and microorganisms. Our microbial community analysis was based on DNA extracted from groundwater samples. These can be considered composite samples of the microbial community from within the radius of influence of the wells. Sediment samples (which are grab samples of the microbial community at discrete points) were not analyzed for comparison, so it is possible that certain sediment-associated microbial community members were not representatively sampled in this study. However, because Geobacteraceae were observed, and are known to exhibit attached growth45,46 this suggests that the aquifer sedimentassociated community was sampled to some extent. Because the RDX concentrations at this site were low (