High-density PhyloChip profiling of stimulated aquifer microbial communities reveals a complex response to acetate amendment


Correspondence: Jillian F. Banfield, Department of Earth and Planetary Science, University of California, Berkeley, CA 94720, USA. Tel.: 1 510 642 9488; fax: 1 510 643 9980; e-mail: jbanfield@berkeley.edu


There is increasing interest in harnessing the functional capacities of indigenous microbial communities to transform and remediate a wide range of environmental contaminants. Information about which community members respond to stimulation can guide the interpretation and development of remediation approaches. To comprehensively determine community membership and abundance patterns among a suite of samples associated with uranium bioremediation experiments, we employed a high-density microarray (PhyloChip). Samples were unstimulated, naturally reducing, or collected during Fe(III) (early) and sulfate reduction (late biostimulation) from an acetate re-amended/amended aquifer in Rifle, Colorado, and from laboratory experiments using field-collected materials. Deep community sampling with PhyloChip identified hundreds-to-thousands of operational taxonomic units (OTUs) present during amendment, and revealed close similarity among highly enriched taxa from drill core and groundwater well-deployed column sediment. Overall, phylogenetic data suggested that stimulated community membership was most affected by a carryover effect between annual stimulation events. Nevertheless, OTUs within the Fe(III)- and sulfate-reducing lineages, Desulfuromonadales and Desulfobacterales, were repeatedly stimulated. Less consistent, co-enriched taxa represented additional lineages associated with Fe(III) and sulfate reduction (e.g. Desulfovibrionales;Syntrophobacterales;Peptococcaceae) and autotrophic sulfur oxidation (Sulfurovum;Campylobacterales). Data implies complex membership among highly stimulated taxa and, by inference, biogeochemical responses to acetate, a nonfermentable substrate.


Research at the Rifle Integrated Field Research Challenge (IFRC) site, Colorado, USA, tests the efficacy of using indigenous microbial communities for the remediation of low-level uranium contamination. Experiments consistently demonstrated reductive immobilization of uranium from groundwater during organic carbon (acetate) stimulation and Fe(III) and sulfate reduction (Anderson et al., 2003; Williams et al., 2011), as illustrated in Fig. 1. Similar results have been reported elsewhere (e.g. Finneran et al., 2002ab; Akob et al., 2008). In a number of field-scale uranium bioremediation studies, including at Rifle, Geobacteraceae were identified as the dominant Fe(III)-reducing bacteria (IRB) (Holmes et al., 2002; Anderson et al., 2003; North et al., 2004) and were implicated in uranium reduction, as several Geobacter species are known to enzymatically reduce U(VI) (e.g. Lovley et al., 1991; Shelobolina et al., 2008). Other phylogenetically diverse bacteria, including certain sulfate-reducing bacteria (SRB), also reduce U(VI) (Lovley & Phillips, 1992; Suzuki et al., 2004; Wall & Krumholz, 2006) and likely contribute to U(VI) reduction in contaminated environments (e.g. Nevin et al., 2003; North et al., 2004). It follows that many organisms, some of which may be present at low abundance levels, could have the potential to reduce U(VI) in the Rifle aquifer. However, relatively little is known about the extent and identity of these and other stimulated taxa, or the range of biogeochemical processes that may impact upon bioremediation at Rifle.

Figure 1.

Characteristic geochemical profiles of Rifle groundwater during acetate amendment illustrating microbial Fe(III) (IR) and sulfate (SR) reduction, forming aqueous Fe(II) and sulfide, and reductive immobilization of aqueous U(VI). As sulfate reduction becomes increasingly dominant, Fe(II) is partly removed from solution by sulfide and precipitated as FeS. Curves represent geochemical data collected over 120 days during first-time amendment. Data collected from eight wells (D01-8, see Williams et al., 2011; Fig. 1 for a plot map) were averaged, and time series were fitted with locally weighted scatterplot smoothing (LOESS) curves. Note, the rate of transition between the major TEAPs is greatly enhanced where the system has been stimulated in the previous year.

The PhyloChip microarray is a cost-effective, 16S rRNA gene–based method for documenting the presence of organisms across a wide range of abundance levels and can provide information about between-sample relative taxa abundances at the family level. The coverage level provided by the PhyloChip is comparable to that obtainable with 454 pyrosequencing of amplified 16S rRNA genes (e.g. DeAngelis et al., 2011), and phylotypes identified using the traditional clone library method have been shown to almost exclusively represent a subset of those identified by PhyloChip (DeSantis et al., 2007). The G2 PhyloChip microarray has been used to investigate complex community responses to external stimuli (e.g. Tsiamis et al., 2008; La Duc et al., 2009; Godoy-Vitorino et al., 2010), including trends in key bacterial family abundances, such as those of Geobacteraceae, during stimulation of uranium-contaminated soil (Brodie et al., 2006). A more recent version of the PhyloChip, G3, expands upon the range of operational taxonomic units (OTUs) identifiable from just fewer than 9000 to almost 60 000 (Hazen et al., 2010).

Prior to this research, stimulated Rifle groundwater and sediment bacterial community composition has been characterized by clone library analysis (using up to 100 clones) and denaturing gradient gel electrophoresis (DGGE) (Anderson et al., 2003; Chang et al., 2005; Vrionis et al., 2005; Holmes et al., 2007), which while identifying the most dominant organisms may have failed to capture the underlying community structure or flanking (less-abundant) community composition. To further our understanding of bacterial populations important to bioremediation, we used the G3 PhyloChip microarray to assess trends in microbial community diversity across a collection of Rifle sediment and groundwater samples that had been biostimulated to differing extents and lengths of time, and to establish the efficacy of bench-top (ex situ) and field-based (in situ, down-well) incubation experiments as a proxy for biostimulated subsurface sediment. Specifically, we sought (1) to define the range of taxa enriched during remediation efforts, in particular those that may impact on the biogeochemical cycling of Fe, S and U, and (2) to characterize the similarity and stability of microbial community composition and structure across experimental conditions.

Materials and methods

Sample descriptions

The samples used in this study are listed in Table 1 and represent a collection from various in situ or ex situ experiments conducted in 2008 and 2009 within, or using material from, the Rifle aquifer (see site description in Anderson et al., 2003). Samples consist of unamended ‘background’ sediments (BGS08/09); naturally reduced (NR) and acetate-amended drill-core sediments (LAS, MAS, HAS); laboratory- and field-based flow-through column sediments (ECS/Q, ICS/Q); enrichment cultures of SRB derived from sediment (ECA/L); and groundwater samples collected during acetate amendment (GW). Each sample type is described below. All samples were flash-frozen immediately upon collection and stored at −80 °C. The well gallery used for the 2008 and 2009 stimulation experiments was first amended during the summer in 2007, and as such constitutes a prestimulated gallery.

Table 1. Description of samples analyzed
Sample IDSample locationSample typeMajor TEAPTEAP ProgressYAmendmentAmount of e donorDate
  1. Background (control) sediments were collected up-gradient, UG, of amendment.

  2. T, time point; IR, Fe(III) reduction; SR, sulfate reduction; NOM, naturally occurring organic matter; L, low; M, moderate; H, high. The year (Y) of stimulation denotes whether experiments were conducted in portions of the aquifer, or using aquifer materials, that were as follows: 1, pristine (first-year amendment); 2, stimulated during the previous summer (second-year amendment); 3, stimulated during the last two summers (third-year amendment).

BGS08Background (UG)Subsurface sedimentUnamended2008
BGS09Background (UG)Subsurface sedimentUnamended2009
GW T1-3Well D04GroundwaterIREarly25–10 days acetateExcess, H2008
GW T4-6Well D04GroundwaterIRLate213–23 days acetateExcess, H2008
NRCore LQ112 16-18′Subsurface sedimentSRProlongedOngoing NOM2009
LASCore P106 10′Subsurface sedimentSRProlonged24 months acetateLimited, L2008
MASCore P105 16′Subsurface sedimentSRProlonged24 months acetateLimited, M2008
HASCore P104 19′Subsurface sedimentSRProlonged24 months acetateExcess, H2008
ICSWell P104In situ sediment columnSRProlonged31 month acetateExcess, H2009
ICQWell P104In situ quartz columnSRProlonged31 month acetateExcess, H2009
ECS T1-3Bench topEx situ sediment columnSREarly prolonged11–3 months acetateExcess, H2009
ECQ T1-3Bench topEx situ quartz columnSREarly prolonged11–3 months acetateExcess, H2009
ECA 1-6Bench topEnrichment cultureSR1AcetateExcess, H2009
ECL 1-2Bench topEnrichment cultureSR1LactateExcess, H2009

Background sediments were collected in 2008 (BGS08) and 2009 (BGS09) up-gradient of injection wells from saturated portions of the aquifer using a backhoe. Acetate was employed in bench-top and subsurface experiments to stimulate microbial growth. Concentrations were typically in excess of microbial rates of consumption throughout the course of field and laboratory stimulation experiments (see Table 1).

Acetate-stimulated groundwater samples (GW T1-6) were obtained along a 6-point time course spanning from early Fe(III) reduction to the onset of sulfate reduction in 2008. Specifically, samples were collected 5, 7, 10, 13, 20, and 23 days after the addition of acetate to the subsurface was commenced in July 2008. Acetate was injected continuously from a 50-mM stock solution, which underwent a dilution of approximately 1 : 10 in groundwater. The injection process is described in detail by Williams et al. (2011). Samples were pumped from well D04 and filtered as described by Wilkins et al. (2009).

Following 110 days of near-continuous acetate injection during the 2008 experiment (Williams et al., 2011), sediment samples dominated by sulfate reduction were collected from drill cores (L-HAS). The injected concentration of acetate was increased to 150 mM during sulfate reduction from days 38 to 110. The subsurface sample P104-19′ (well ID P104, sample depth 19′) was collected 0.5 m down-gradient; P105-16′, 5 m down-gradient; and P106-10′, 9 m down-gradient of injection wells. The amount of acetate received by these samples decreased with increasing distance from injection wells.

Further acetate-stimulated sediment samples were obtained from the aquifer after incubating flow-through columns in situ (ICS-Q) within well P104, for 1 month (July–August 2009). During this period, acetate was injected into the aquifer up-gradient of P104 from a 150-mM stock solution, attaining approximately 1 : 10 dilution in groundwater. Cylindrical columns (either 2.5 cm wide × 20.3 cm long or 5.1 wide × 10.2 cm long) were packed with either fine-grained (< 2 mm) freshly collected background Rifle sediment (BGS09) or dithionite–citrate–bicarbonate washed (Chang & Jackson, 1957) quartz sand underlain by a 2-cm-thick layer of fresh Rife background sediment for microbial inoculation. Flow was achieved by pumping (acetate-amended) groundwater up through the column using peristaltic pumps located at the ground surface. Flow rates corresponded to a pore water velocity of approximately 1 m day−1, approximately twice the rate of groundwater within the aquifer. Sediment and quartz were collected from columns upon sacrifice.

Drilling in 2009 also recovered subsurface sediment (well ID LQ112, sample depth 16-18′) from a NR sulfidic region of the aquifer that was putatively stimulated by autochthonous organic matter. The core was obtained outside areas previously impacted by acetate amendment.

Laboratory experiments, emulating field studies, utilized ex situ ‘bench-top’ flow-through cylindrical columns (ECS-Q) (2.5 cm wide × 15 cm long) that were packed with clean quartz sand, underlain by an inoculum of background (BGS08) Rifle sediment (~ 10% column volume, < 2 mm fraction) mixed 50 : 50 with quartz sand. Columns were flushed continuously with anoxic (99 : 1 N2/CO2) Rifle groundwater that contained ~ 10 mM of naturally occurring sulfate and 15 mM of added acetate. Flow rates corresponded to a pore water velocity of 0.5 m day−1. Columns were sacrificed at the onset of sulfate reduction, after 31 days of incubation, and during ongoing sulfate reduction, after 56 and 78 days. Both sediment and quartz were collected from each column.

Fast-growing SRB, as might be expected to occur in the field in the presence of excess electron donor, were enriched using a modified minimal (no yeast extract) Postgate B medium (Postgate, 1984) or Rifle groundwater, both with added vitamins and with minerals (see Handley et al., 2009 and references therein) and 10 mM acetate (ECA) or 20 mM lactate (ECL). Media were inoculated with 10% w/v Rifle background sediment, incubated at 30 °C, and subcultured 4–6 times.

DNA extraction and amplification

Genomic DNA was extracted in duplicate from sediments using the PowerSoil™ DNA Isolation kit (Mo Bio Laboratories, Inc, Carlsbad, CA) and from groundwater and cultures using the FastDNA® SPIN Kit for Soil (MP Biomedicals, Solon, OH). Near full 16S rRNA gene amplification was undertaken using 8-gradient (annealing temperatures, 48–58 °C) 25-cycle PCR in order to minimize PCR bias, as described by DeSantis et al. (2007). Reactions were performed using the general bacterial primers 27f (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492r (5′-GGTTACCTTGTTACGACTT-3′) (Lane, 1991) and Clonetech Titanium taq (Mountain View, CA), and uracil was incorporated during fragment synthesis (2 : 1 dTTP/dUTP). The primers used span the 16S rRNA gene region used for probe creation (DeSantis et al., 2007). Pooled products were concentrated to 20 μL by ethanol precipitation with glycogen, and fragment size and quality were checked by gel electrophoresis.

Probe fragmentation and hybridization

PCR products were fragmented to 50–200 bp using uracil-DNA glycosylase and apurinic/apyrimidinic endonuclease 1 (Affymetrix, Santa Clara, CA). Reaction mixes were spiked with amplicons from prokaryotic metabolic genes, yeast genes, and Arabidopsis genes of known concentration (final concentration range: 4.62–651.9 pM) to serve as internal controls, and contained 500 ng of sample amplicons (or between 225 and 372 ng for samples NR and ECQ T2 r1). Reactions were carried out at 37 °C for 60 min and inactivated at 93 °C for 2 min. Fragmentation products were terminally biotinylated using the Affymetrix GeneChip DNA Labeling Reagent kit, following the manufacturer's instructions. Labeled products were hybridized to 25-mer 16S rRNA gene probes on custom-made GeneChip chips (G3 PhyloChip) using the Hybridization Module (Affymetrix). Reactions were performed according to the manufacturer's protocol, with DNA denaturation at 99 °C for 5 min and hybridization overnight at 48 °C while chips rotated at 60 r.p.m. Hybridized chips were washed and stained in the automated Affymetrix fluidics station, and fluorescence was measured using the Affymetrix GeneChip Scanner 3000 7G (see DeSantis et al., 2007). Chip design is described by DeSantis et al. (2007) and Hazen et al. (2010).

Data treatment

Chip OTUs comprised 37 ± 9.6 SD (or between 2–50) probe pairs targeting closely related organisms with ≥ 97.5% 16S rRNA gene similarity (Hazen et al., 2010). Following amplicon hybridization to chip, the mean intensities of all spiked in DNA controls were averaged for each chip and scaled to 10 000 in the initial treatment of the data. Background hybridization intensities were subtracted from probe intensities as follows. Probe pairs comprised one perfectly matching (PM) probe and one mismatching (MM) probe with a substitution at central base number 13 along the 25-mer probe length (Hazen et al., 2010). Threshold criteria for scoring OTUs as positive are described in detail by Hazen et al. (2010) and are outlined here. The difference between PM and MM hybridization intensities, the d score, was determined. The likelihood that the difference in intensities (d score) of a given OTU came from a distribution more similar to positive rather than negative controls was then calculated, yielding the r score. An OTU was considered present if (1) at least 18 probe pair signals were counted and (2) the quartiles of ranked r scores were at least 0.7, 0.95, and 0.98 for the first, second, and third quartiles, respectively.

OTUs were then assessed against a stringent secondary criteria that penalized any OTUs likely to be positive because of probe cross-hybridization. Potentially, cross-hybridizing probes were deemed to be those sharing perfect identity among their central 17-mer. Adjusted r scores (rx) were generated by dividing the number of external subfamilies that passed the first threshold by the number of external subfamilies that potentially cross-hybridize with specific probes. Subfamilies with a third quartile rx value of ≥ 0.48 were accepted. Subfamilies were demarcated as OTU groups with 72% similar heptamers. The means of PM probe hybridization scores, minus the highest and lowest scores, were used for analyses. Corrected hybridization scores and presence–absence data are given in the Supporting Information, Data S1.

In analyses comparing the relative abundances of taxa, OTUs were discounted if probe scores fell below the secondary threshold across all samples. Hybridization scores were then normalized to total array intensity to adjust for differences in sample intensities, attributed to differences in taxa diversity owing to the loss of available hybridization sites for highly abundant taxa in samples with less even, and typically less rich, communities (a negative linear trend is observed between summed chip intensities and maximum intensity values, R2 = 0.92, prior to normalization to total array intensity). For diversity measures, including presence–absence data, the hybridization scores of OTUs within individual samples were set to zero if below the threshold. As zero is a nonstandard hybridization score value, when extraction replicate data were averaged, if a value was present for only one replicate, the nonzeroed value was used.

Hybridization scores in PhyloChip have been shown to directly relate to gene abundance through a positive linear relationship (DeSantis et al., 2005; Brodie et al., 2007). However, scores are not quantitative within a sample, owing to differences in the hybridization efficiencies of probes. Even so, the relative abundances of individual taxa are comparable across samples.

Community analyses

Differences in community diversity among samples were examined with rank abundance curves (rank of OTU vs. hybridization scores), which graphically portray observed sample community richness (S = total number of OTUs) and evenness (slope of line as a representation of the distribution of OTU abundance). Key information from the curves was summarized as richness and maximum hybridization score data and portrayed relative to average background scores (score - average background sediment score).

Unamended background sediment is used as a baseline throughout the study to gauge changes in diversity and the enrichment of taxa-stimulated communities. While unamended sediment cannot be considered an exact proxy for unamended groundwater, by applying the same baseline uniformly, we are still able to compare differences and similarities across stimulated sample types and determine whether similar taxa are highly abundant in stimulated sediment and groundwater communities.

Hierarchical clustering and nonmetric multidimensional scaling (NMDS), in primer 6 (PRIMER-E Ltd, Plymouth, UK), were used to assess clustering patterns among samples. Both methods utilized Bray–Curtis dissimilarity matrices. Hierarchical clustering was performed using the group average method. Stress values of ≤ 0.1 indicate that NMDS plots are a good representation of sample relationships in two-dimensional space. Stress is determined in primer 6 by the goodness of fit of a regression line to samples plotted according to pairwise distance in the ordination vs. percent similarity in the Bray–Curtis matrix.

Analysis of similarity (anosim), in primer 6, was used to test the alternate hypothesis that there are differences between sample groups. R statistics were calculated using 999 permutations to determine the difference in average rank dissimilarities between and within groups, based on Bray–Curtis similarity matrices. Communities were considered different if the R statistic (on a 0–1 scale) approached 1, moderately overlapping for R statistics near 0.5 and highly overlapping for R statistics near 0. Permutational anova (permanova) was used to determine the similarity of extraction replicates by comparing the average dissimilarities between extraction replicates and samples (http://www.stat.auckland.ac.nz/~mja/Programs.htm#Mine; Anderson, 2001; McArdle & Anderson, 2001). Bray–Curtis dissimilarities were used for the analysis.

The similarity percentage method, SIMPER, in primer 6 was used to determine the contribution of each OTU to differences between groups based on average Bray–Curtis dissimilarities. For comparison, the standard deviations (SD) of OTU hybridization scores between sample groups or relative to background were also calculated to gauge the variation in abundances (as employed by Brodie et al., 2006). Key taxa were taken as those occurring in the top 10% of SIMPER contributions and top 100 SDs.

The relative abundances of all taxa, and key taxa determined by SIMPER and SDs, were rendered visually as heatmaps using the made4 v1.22 package (Culhane et al., 2005) in R (http://www.r-project.org/; Ihaka & Gentleman, 1996). Hybridization scores were scaled by row (OTU) across all samples, according to the default settings. Specifically, scores for each OTU were adjusted to values spanning the range of ±3. Samples were grouped using a correlation metric distance and average linkage cluster analysis. This method enables comparisons among samples (columns) for each OTU, but not comparisons among taxa (rows). The OTU accession numbers for probe-targeted sequences are given in Table S1.

Given the large number of probes on the PhyloChip, the phylogenetic assignment of OTUs examined in the ‘key taxa’ heatmap was re-verified by blast matches (Altschul et al., 1990), and by the construction of neighbor-joining and maximum-likelihood phylogenetic trees with 1000 boot-strap replicates using mega v5.0 software (http://www.megasoftware.net; Tamura et al., 2011). Trees were created using OTU probe-targeted sequences from GenBank. Neighbor-joining trees employed evolutionary distances estimated with the maximum composite likelihood method (Tamura et al., 2004). The highest log-likelihood (maximum-likelihood) tree was obtained using the maximum parsimony method when the number of common sites was < 100 or < 1/4 of the total number of sites, or using the BIONJ method with MCL distance matrix when the number of common sites was greater than this (Tamura & Nei, 1993).


Similarities in bacterial composition and abundance patterns

OTUs were detected across a range of different phyla in both stimulated and unstimulated samples, including the seven major chip targeted phyla (Acidobacteria, Actinobacteria, Bacteroidetes, Chloroflexi, Firmicutes, Planctomycetes, and Proteobacteria) and Candidate Divisions (heatmap in Fig. S1). The heatmap shows that background sediment communities were diverse, with relatively abundant members widely distributed across these phylogenetic groups, especially in the 2009 collected sediment community (BGS09). Some heterogeneity in the community compositions of these two samples, which were collected from different locations within the aquifer, is evident. Drill-core sediments collected after acetate amendment (L-HAS), and NR sediment communities, grouped closely with the background communities, and displayed a similar broad distribution of enrichment across the phylogenetic groups. Communities in enrichment cultures were also broadly enriched and clustered with the HAS and NR. Both acetate and lactate resulted in enrichment communities with very similar profiles, although relative abundances were generally much higher with lactate. In contrast, community abundance patterns appear to have been much less evenly distributed across phylogenetic groups in acetate-amended column sediments/quartz (IC and ECS/Q) and groundwater samples. Even so, removing the enrichment cultures from the heatmap (data not shown) caused HAS to group with the in situ column communities owing to discernable similarities in composition.

Overall, total community compositions differed primarily by sample type and, to a lesser extent, by time (or degree of stimulation). Hierarchical clustering (Fig. 2a) revealed that, at similarities > 96%, communities were separated into four distinct clusters according to the sample type, similar to the groups observed in the full community heatmap (Fig. S1). ANOSIM results demark significantly different clusters on the NMDS plot (Fig. 2b). The fourth cluster is not shown in Fig. 2 to better resolve clusters 1–3. Cluster 4 represents a dominant, but poorly constrained group of enrichment culture communities and is shown in Fig. S2. Removing this cluster does not alter the character of the remaining clusters. anosim global and pairwise R statistics were between 0.841 and 0.959 (P-values < 0.001) differentiating all four clusters.

Figure 2.

Community cluster analysis. (a) Group average hierarchical clustering of OTU data into distinct groups (1–4). Abbreviations: Earlier (E) or later (L) time points; lower (LA) or higher (HA) acetate; number of years (Y) of amendment the aquifer or aquifer-derived material has undergone (see Table 1). (b) NMDS of OTU hybridization score data from groups 1–3, depicting two-dimensional separation of samples based on overall community composition. Shaded areas demark clusters in (a) that differ significantly from one another based on anosim results. Extraction replicate data are averaged in the NMDS plot, but display identical clustering patterns to unaveraged data.

Specifically, cluster 1 groups well-sourced samples, namely GW and field-deployed column (ICQ-S) communities collected during dominant phases of Fe(III) reduction and sulfate reduction, respectively. Communities in groundwater differ according to earlier and later time points (GW T1-3 vs. T4-6), while the composition of in situ column communities differs based on quartz or sediment column matrix. Within cluster 2, ex situ (bench-top) column communities group with the inoculating background sediment (BGS08) and exhibit a temporal and column matrix cluster effect. However, the duration of sulfate reduction appears to be a stronger driver for clustering than matrix (quartz vs. sediment). In cluster 3, background (BGS09) and low and moderately stimulated subsurface sediments (L-MAS) group closely, while NR and more highly stimulated subsurface communities are less similar. Cluster 4 comprises sulfate-reducing enrichment culture communities grown with acetate or lactate, but does not cluster expressly based on electron donor.

To dampen the effect of highly abundant taxa, cluster analysis using square-root-transformed hybridization scores was also performed, but results differed only by slightly higher overall similarity values (data not shown). Agreement between DNA extraction replicates evident from hierarchical clustering (Fig. 2a) was confirmed by average dissimilarity scores between these replicates and samples, generated by permanova. Dissimilarity scores were, on average, 0.48 (±0.1 SD) between replicates and 2.41 (±0.7 SD) between samples.

Community richness and diversity

Overall community diversity was lower in stimulated communities relative to background sediment, reflected by decreases in taxa number (richness) and decreases in the evenness of taxa abundance (Fig. 3; for rank abundance curves, see Fig. S3). Fe(III)-reducing groundwater and sulfate-reducing sediment communities shared a similar range of richness values, with over twofold lower richness values relative to background in some samples. Interestingly, diversity increased somewhat in groundwater communities collected in late Fe(III) reduction (T5-6), as conditions approached the transition to sulfate reduction. Quartz-colonized communities exhibited lower overall richness than sediment-associated column communities; however, the in situ quartz community exhibited greater diversity than that of an analogous drill-core sediment (HAS). As expected, the lowest community richness and evenness occurred in SRB enrichment cultures.

Figure 3.

Bar chart, with staggered bars showing average richness (S) data and maximum hybridization scores (MHS) relative to (minus) average background (BGS08/09) values (S = 3497 ± 335; MHS = 14563 ± 47). S and MHS values summarize rank abundance curves. Negative S and MHS values are indicative of values lower than those in the background sediment communities. The table gives the range of original S values.

Richness values for the NR sediment community were low (319 and 346). This may be partly due to the loading of less than an optimal amount of DNA onto the chips. As such, these data were excluded from comparisons in Fig. 3. Nevertheless, total array hybridization scores were within the range of other samples, suggesting high abundance of detected OTUs. Owing to lower DNA, total array-normalized values may overestimate NR OTU relative abundances, but this would not affect the community structure, nor the identification of dominant taxa (see below).

Identity of key bacterial taxa enriched during biostimulation

In order to ascertain the bacteria most responsive to acetate amendment or natural attenuation, changes in OTU abundances were compared to background communities and between-sample groups. Results were approximately comparable using SIMPER and SD methods (Table 2), and no difference was noted between SIMPER analyses using untransformed or square-root-transformed data. The relative abundances of OTUs from enriched lineages are portrayed within a cluster heatmap (Fig. 4). Results indicated a notable increase over background and low-acetate sediment levels in OTUs representing or very closely related to genera known to reduce Fe(III) and/or be involved in sulfur redox cycling (Fig. 5).

Figure 4.

Heatmap depicting the relative abundances of highly stimulated lineages known to reduce/oxidize sulfur species and/or reduce Fe(III). Ep, Epsilonproteobacteria; Dp, Deltaproteobacteria; F, Firmicutes; N, Nitrospira; Gp, Gammaproteobacteria. Colors in the bar at the top of the heatmap distinguish sample groups (NR, yellow; BGS, black; L-HAS, brown; GW, dark blue; IC, pale blue; EC, red; ECS, pale gray; ECQ, dark gray). Extraction replicate and culture data are averaged. No difference was evident between heatmap results for replicates, and few differences were apparent among cultures.

Figure 5.

16S rRNA gene neighbor-joining phylogenetic tree of key highly enriched taxa identified by SIMPER and SD analyses (boldface with black circles). One representative (probe-targeted) sequence per OTU was used. The tree does not represent actual sequences from Rifle aquifer material, but is an close approximation based on probe matches. All sequences were obtained from GenBank. Reference taxa are shown in regular font. Bootstrap values ≥ 50 are shown.

Table 2. Elevated taxa in sample groups relative to background sediment communitiesThumbnail image of

In particular, OTUs, representing or closely related to Epsilonproteobacteria known to oxidize (Sulfurovum and Helicobacteraceae) and/or reduce (Sulfurospirillum) sulfur, were highly increased in acetate-amended sediment and groundwater communities over background (Fig. 4). These bacteria were most enriched in the in situ columns, which were incubated in a well gallery subject to a third summer of acetate amendment. The heatmap also shows fairly abundant Sulfurovum-like OTUs in the 2009, but not 2008, background sediment.

Deltaproteobacteria lineages known to reduce sulfate were increased in all stimulated samples (Fig. 4). Of these, Desulfobacterales were consistently enriched across all amended samples, excluding enrichment cultures (Fig. 4 and Table 2). A high degree of enrichment in Desulfobulbaceae, Desulfomicrobiaceae, and Syntrophobacterales appeared to be less consistent among samples. The Firmicute family, Peptococcaceae, were common only to NR sediment and communities stimulated under pristine conditions (first-year amendment), in ex situ columns and enrichment cultures. Peptococcaceae are highly enriched in cultures using either acetate or lactate as electron donor.

OTUs in the Desulfuromonadales, an order well known for dissimilatory Fe(III) reduction, were highly increased in groundwater and column sediments (ICS and ECS T2 and 3), but decreased in groundwater collected during late-stage Fe(III) reduction (Fig. 4). Curiously, Desulfuromonadales became increasingly enriched alongside that of SRB lineages (i.e. Desulfobacterales and Peptococaceae) in ex situ columns sediments, which were harvested during early-stage to late-stage sulfate reduction. SIMPER results suggest that within the Desulfuromonadales, Geobacteraceae and Desulfuromonadaceae were differently enriched, being each more highly enriched in Fe(III)-reducing groundwater or sulfate-reducing sediments, respectively (Table 2). In general, Desulfuromonadales appeared to be poor colonizers of the quartz matrix in ex situ, but not in situ columns.

SIMPER and SD analyses (data not shown) indicate that the 2008 background sediment had higher abundance of Pseudomonaceae than the 2009 sediment, likely explaining why these communities do not cluster together in full community dendrogram and NMDS plots (Fig. 2).

Clustering of samples based on highly enriched taxa (Fig. 4) differs in certain respects to that based on full community data (Fig. 2 and Fig. S1). In terms of enriched taxa, moderate- and high-acetate subsurface drill-core sediments more closely resemble in situ columns and groundwater, owing to overall similarities among highly enriched S-cycling and Fe(III)-reducing lineages. Moreover, ex situ columns cluster with enrichment cultures, apparently owing to shared high abundances of Peptococcaceae.


Observed richness, diversity, and co-enrichment of IRB and SRB

Use of high-density PhyloChip expanded our estimation of community richness within the Rifle aquifer over 10-fold, with large numbers of observed bacterial OTUs (795–3132) still detected after the addition of excess acetate to subsurface Rifle sediment and groundwater. The observed richness for the Rifle background sediment was also up to four times greater than that determined by Brodie et al. (2006) for unamended U-contaminated sediment from Oak Ridge, TN, using an earlier version of the PhyloChip with fewer probes (1000 OTUs identified). However, to what extent the observed richness values in this study approximate actual richness values, in terms of reaching saturation, is not known.

Consistent with previous clone library and DGGE studies conducted at the site, PhyloChip data indicate that Desulfuromonadales (dissimilatory IRB) and Desulfobacterales (SRB) were enriched during acetate-stimulated Fe(III) and sulfate reduction (Anderson et al., 2003; Chang et al., 2005; Vrionis et al., 2005; Holmes et al., 2007). Deeper community sampling with PhyloChip data clearly shows that the enrichment of these lineages was not mutually exclusive, even in the very early stages of acetate-induced Fe(III) reduction (Fig. 4). Desulfobacterales enrichments occur early in the Fe(III)-reducing stage of biostimulation, and Desulfuromonadales continue to be enriched throughout sulfate reduction.

The co-enrichment of Fe(III)- and sulfate-reducing lineages, and evidence of their concurrent activity in Rifle-related laboratory and field studies (Komlos et al., 2008a; Miletto et al., 2011; Williams et al., 2011), may be explained by the mixture of easily reducible and recalcitrant iron minerals within the aquifer (Postma & Jakobsen, 1996; Komlos et al., 2008b). It seems likely that once easily reducible, poorly crystalline Fe(III) sources are expended within the aquifer, Desulfuromonadales colonize and slowly reduce more recalcitrant forms of Fe(III), such as goethite. The enrichment of Desulfuromonadales in Fe(III)-bearing Rifle sediment, but not in quartz sand matrices in ex situ column experiments, suggests a need for attachment to consumable Fe(III) oxides. Attachment to quartz in the in situ columns may have been due to the presence of suspended Fe(III) oxides in the aquifer.

Miletto et al. (2011) recently demonstrated that dissimilatory sulfite reductase (dsrB) genes used in sulfate reduction were expressed predominantly by Desulfobacteraceae and, to a lesser extent, by Desulfobulbaceae and Syntrophaceae, in a second-year Rifle biostimulation experiment. Although samples analyzed in this study represent a diverse collection of nonreplicated time course or single time point experiments, and different amendment time points, PhyloChip data indicate that enrichment of these SRB lineages, particularly Desulfobacteraceae, was typical during acetate amendment experiments and largely independent of sample type. Desulfobacteraceae and/or other Desulfobacterales became enriched in samples regardless of (1) sample incubation and collection procedure, that is, whether in situ or ex situ column, drill core, or groundwater; (2) whether samples were acetate-stimulated or from a NR part of the aquifer; or (3) the progress of stimulation, that is, dominant terminal electron–accepting process (TEAP), or with or without prestimulation (first- to third-year amendments).

In addition, other enriched OTUs were detected using PhyloChip that belong to phyla known to include sulfate reducers (Desulfovibrionales and Peptococcaceae) and Fe(III) reducers (Shewanellaceae, Aeromonadaceae, Peptococcaceae, and Sulfurospirillum). Among likely SRB candidates, Peptococcaceae were the most variable, and their detection in past studies has appeared transitory (Anderson et al., 2003) or atypical (Vrionis et al., 2005). In Anderson et al.'s study (2003), this study, and further experiments within the Rifle aquifer (K.M. Handley, K.W. Wrighton, C.S. Miller, J.F. Banfield., unpublished data), Peptococcaceae were abundant exclusively in experiments using pristine (not previously stimulated) aquifer material or in NR sediment, suggesting that they may represent an early-stimulation SRB group when exposed to high levels of added acetate.

Mixed sulfate-reducing communities

Although several SRB groups are co-enriched, the prevalence of Desulfobacterales in the acetate-amended field sediments is not surprising. Many genera from this family, most notably Desulfobacter, can grow using acetate as a sole electron donor and carbon source while reducing sulfate (e.g. Bak & Widdel, 1986; Platen et al., 1990; Lien & Beeder, 1997; Purdy et al., 1997; Kuever et al., 2001). Certain species within other detected SRB lineages, namely Syntrophobacterales and Desulfotomaculum (in the Peptococcaceae), can also couple sole acetate oxidation to sulfate reduction (e.g. Widdel & Pfennig, 1977; Oude Elferink et al., 1999). How competitive these bacteria are during bioremediation likely depends upon their concentration-specific affinity for acetate and sulfate (Laanbroek et al., 1984), in addition to the availability of other limiting nutrients. The ability of Peptococcaceae to form spores (Campbell & Postgate, 1965; Stackebrandt et al., 1997) may also account for the apparent dominance of this group in first-time stimulated sediments, although alternative possibilities cannot be excluded.

Aside from Desulfotomaculum, other Peptococcaceae genera, Desulfovibrionales, or Desulfobulbaceae species are not known to couple acetate oxidation with sulfate reduction, although some species can reduce sulfate mixotrophically when using small amounts of acetate as a carbon source and H2 as an electron donor (e.g. Lien et al., 1998; Dias et al., 2008). Alternatively, their presence may be result from cryptic growth supported by the acetate-fed biomass. Competition for acetate by acetoclastic methanogenesis is unlikely to be important at the high sulfate concentration present in the Rifle groundwater (Dar et al., 2008), until very low sulfate/acetate ratios are attained during peak sulfate reduction (Williams et al., 2011).

Interestingly, diverse SRB appear to coexist in the NR sediment, where sulfate reducers from all of the identified taxonomic groups (Desulfobacterales, Desulfovibrionales, Syntrophobacterales, and Peptococcaceae) were abundant relative to nonreduced background sediment. Most of these lineages (Desulfobacterales, Desulfovibrionaceae, and Peptococcaceae) also include species capable of Fe(III) reduction (Lovley et al., 1993; Ramamoorthy et al., 2006; Haouari et al., 2008). Both groups may be implicated in the formation of uranium-enriched framboidal pyrite associated with these NR zones in the Rifle aquifer (Qafoku et al., 2009).

Phylogenetic evidence for sulfur cycling

Among the most highly enriched OTUs in acetate-stimulated samples, in addition to SRB, were those associated with lineages of chemolithoautotrophic Epsilonproteobacteria. These bacteria – Sulfurovum, Helicobacteraceae (Sulfuricurvum or Sulfurimonas), and Campylobacteraceae (Sulfurospirillum deleyianum) – are known to couple nitrate or oxygen reduction to the oxidation of sulfide to S0, or S0 and thiosulfate to sulfate (Hoor, 1975; Eisenmann et al., 1995; Inagaki et al., 2004; Kodama & Watanabe, 2004). While 16S rRNA gene phylogeny is not necessarily congruent with physiology, the well-characterized nature of these organisms lends strong evidence toward their potential activity in the Rifle aquifer.

Indeed, the aquifer contains approximately 6–10 mM of sulfate in the groundwater, which is largely consumed during periods of high acetate loading (e.g. 10 mM). Equimolar concentrations of H2S produced during sulfate reduction are capable of generating a large supply of reduced sulfur, such as FeS and S0 (Williams et al., 2011), that may fuel cyclic oxidation and reduction. Microbial oxidation of elemental or reduced sulfur species could be supported by small amounts of dissolved oxygen or nitrate present in micromolar concentrations in the groundwater (Williams et al., 2011). In fact, high ammonium-to-nitrate ratios detected during a previous stimulation experiment in the aquifer (Mouser et al., 2009) may indicate the occurrence of denitrification.

Other notable OTUs enriched during amendment that may also have the potential to contribute to sulfur redox cycling are Sulfurospirillum, Desulfuromonadales (Geobacteraceae, Desulfuromonadaceae, Pelobacteraceae), and Shewanellaceae – all of which include species capable of reducing thiosulfate, S0, nitrate, or nitrite (Lovley et al., 2004). The nature of abundant Nitrospiraceae OTUs, however, is largely enigmatic, although sequence identity indicates that the OTUs are most closely related to sulfate-reducing Thermodesulfovibrio (Henry et al., 1994) and putatively sulfur-oxidizing Magnetobacterium (Jogler et al., 2010) species.

Phylogenetic diversity of bacteria potentially capable of uranium reduction

Direct microbial enzymatic reduction of U(VI) yields similar energy as that for the reduction of crystalline Fe(III) minerals (Finneran et al., 2002ab) and occurs in the presence of Fe(II) and sulfide without interference by abiotic reduction (Lovley et al., 1991; Finneran et al., 2002ab). Geobacter is an obvious candidate for U(VI) reduction during acetate stimulation in the Rifle aquifer owing to its ability to grow using uranium as an electron acceptor (Lovley et al., 1991), and its high abundance, even during sulfate reduction (Williams et al., 2011). Nevertheless, we also observed an increase in the abundance of other lineages that, in addition to Fe(III) and sulfate reduction, include species that are able to reduce U(VI). It is possible that these lineages (Shewanellaceae, Desulfovibrionaceae, and Peptococcaceae) may also account for some portion of the U(VI) reduced during biostimulation (Caccavo et al., 1992; Lovley & Phillips, 1992; Tebo & Obraztsova, 1998; Suzuki et al., 2004). With the presence of multiple candidates for U(VI) reduction with different electron donor and nutrient affinities, it is difficult not to speculate that stimulated biological uranium reduction at the site may be robust against shifts in community composition.

Similarities and differences among stimulated communities

In acetate amendment experiments, it made little difference to the enrichment of Desulfuromonadales and Desulfobacterales whether samples were collected during dominant Fe(III)- or sulfate-reducing phases, from groundwater, column, or drill-core sediments, or from first-, second-, or third-year stimulation experiments. Only the degree of enrichment altered, evidently owing to either the length of incubation or whether previous stimulations had occurred. Re-amendment of the Rifle aquifer from 1 year to another is known to cause a legacy effect, inducing a more rapid community response and transition through dominant TEAPs (Callister et al., 2010). Results here also suggest that community composition is affected and that while experiments involving a section of nonpristine, previously stimulated aquifer may fail to capture the enrichment of Peptococcaceae, the use of pristine sediments could underestimate the importance of Desulfobacterales, particularly with respect to long-term amendment. Enrichment cultures, in particular, missed the progression from Peptococcaceae to Desulfobacterales.

Experimental approaches clearly yielded differences in OTU abundance across diverse phylogenetic groups. Communities from amended columns and groundwater samples appeared depleted in OTUs from a number of phyla relative to amended drill-core sediment communities. However, considering the similarity in the composition of highly enriched taxa from column, groundwater, and drill-core samples (all from second- or third-year aquifer stimulations), it is uncertain whether observed differences in total community diversity would have any implications for remediation processes or major community function.

It is not within the scope of this study to determine whether or how groundwater communities collected during experiments at Rifle might differ from coupled sediment communities; however, it is notable that although groundwater and sediment samples from field-based experiments represent different time points during stimulation or successive stimulation events (second-vs. third-year stimulations), PhyloChip results show that groundwater communities can be at least as diverse as those in subsurface sediments (HAS and ICS in Fig. 3), and share strong similarities in highly stimulated taxa composition (Fig. 4). Groundwater sampling is commonly used to study subsurface communities (e.g. Anderson et al., 2003; Fields et al., 2005; Holmes et al., 2007; Miletto et al., 2011), even though planktonic and sediment-attached aquifer communities are expected to differ somewhat in composition (Alfreider et al., 1997; Lehman et al., 2001; Flynn et al., 2008; Anneser et al., 2010). Regardless of potential biases in groundwater sampling, data also suggest that down-well, flow-through column incubations may provide a convenient alternative, permitting sampling of the whole (planktonic and attached) wet sediment communities.


PhyloChip-based analysis yields an in-depth view of community phylogenetic affiliation and richness, providing evidence for both compositional complexity among acetate-stimulated taxa and reproducibility in key taxa among experiments. We consistently detected the same highly stimulated lineages (Desulfuromonadales and Desulfobacterales) throughout acetate-promoted Fe(III) and sulfate reduction, accompanied by a less consistently enriched contingent of other lineages. Of the differences observed among experimental approaches, year of aquifer stimulation (i.e. first, second, or third) appears to be the largest factor affecting the composition of highly stimulated taxa. Compositional similarities among drill-core and in situ flow-through column communities, particularly among highly stimulated taxa, indicate that the latter may serve as a suitable proxy and tractable method for studying subsurface stimulation. The phylogenetic affiliations of taxa, enriched during amendment, suggest that many may share the ability to use the same electron donors and acceptors, such as Fe(III), sulfate, and potentially also U(VI). Phylogenetic data also provide evidence for other biogeochemical processes, in particular re-oxidation of sulfur or reduced sulfur species. We speculate therefore that the presence of multiple lineages that able to compete for resources may increase the functional efficiency of the system and the variety of niches exploited.


Funding was provided by the Environmental and Remediation Sciences Program, Office of Science, Biological and Environmental Research, US Department of Energy. We thank Kate Campbell (US Geological Survey, Menlo Park) for her help with column design and Shuk Chan (University of California, Los Angeles) for help with field implementation. We also thank our anonymous reviewers for their helpful comments.