Bacterial diversity and community structure of a sub-surface aquifer exposed to realistic low herbicide concentrations

Abstract

An increasing number of herbicides are found in our groundwater environments. This underlines the need for examining the effects of herbicide exposure on the indigenous groundwater microbial communities, as microbial degradation is the major process responsible for the complete removal of most contaminants. We examined the effect of in situ exposure to realistic low concentrations of herbicides on the microbial diversity and community structure of sub-surface sediments from a shallow aquifer near Vejen (Denmark). Three different community analyses were performed: colony morphology typing, sole-carbon source utilisation in Biolog^®EcoPlates, and denaturing gradient gel electrophoresis. Cluster analysis demonstrated that the microbial communities of those aquifer sediments that acclimated to the herbicide exposure also had similar community structure. This observation was concurrent for all three community analyses. In contrast, no significant effect was found on the bacterial diversity, except for the culturable fraction where a significantly increased richness and Shannon index was found in the herbicide acclimated sediments. The results of this study show that in situ exposure of sub-surface aquifers to realistic low concentrations of herbicides may alter the overall structure of a natural bacterial community, although significant effects on the genetic diversity and carbon substrate usage cannot be detected. The observed impact was probably due to indirect effects. In future investigations, the inclusion of methods that specifically detect relevant microbial sub-populations and functional genes is therefore recommended.

Microbial diversity, Microbial community structure, Herbicide, In situ exposure, Sub-surface aquifer

1 Introduction

During the past decade an increasing number of herbicides has been detected in groundwater aquifers [1]. These include, e.g., phenoxy acid, atrazine and phenyl urea compounds. Groundwater constitutes a major drinking water resource, and studies exploring the impact of herbicides on sub-surface microbial communities are important since microbial degradation is the primary process involved in the complete transformation of most contaminants [2].

Numerous studies have examined the effect of herbicides on surface soil microbial communities (for reviews see, e.g. [3,4]). The traditional methods that have been used are measures of microbial biomass as well as carbon and nitrogen metabolism [4]. These measures do not necessarily indicate if changes in the microbial diversity or community structure occurred and such measures should also be included [4] due to the possible correlation between microbial diversity and ecosystem function [5]. Only few studies have directly examined the effect of herbicide application on microbial diversity and community structure in soil environments [6–9], and no studies have evaluated such effects in groundwater environments. Both the diversity and the structure describe the composition of a microbial community. However, whereas the diversity is indicated by a numerical value, the community structure is a figurative measure and usually illustrated by multivariate analysis. Besides, the diversity also describes the heterogeneity of a microbial sample.

Only a small percentage of the indigenous soil bacteria is culturable [10], emphasising the importance of applying cultivation-independent techniques in the analysis of natural microbial communities. Bakken [11] found, however, that the culturable bacteria most likely play an important role in many microbial processes and therefore it is reasonable also to include cultivation-dependent techniques in microbial community analyses. During the past few years the approach of using both cultivation-dependent and cultivation-independent techniques in the analysis of microbial community composition has become increasingly widespread [7,8,12–14]. Several experimental methods have been used, e.g., colony morphology typing, time of appearance of bacterial colonies, sole-carbon-source utilisation in Biolog microtiter plates, denaturing (DGGE) and temperature (TGGE) gradient gel electrophoresis, terminal-restriction fragment length polymorphism (T-RFLP), DNA reassociation, 16S rRNA hybridisation, sequencing of 16S rDNA clone libraries (for review see, e.g. [4,15]). In the present study, we used colony morphology typing, sole-carbon-source utilisation in Biolog EcoPlates and DGGE analysis in order to describe the composition of microbial communities from a sub-surface aquifer. Colony morphology typing is a classical approach [14,16,17] and relies on the grouping of bacterial colonies according to phenotypic characters [18]. With the Biolog^® microtitre plates (Biolog, Hayward, CA), the oxidation and turnover of single carbon sources in microtiter wells is measured spectrophotometrically by reduction of a tetrazolium dye to formazan [19]. Biolog plates have been used in several studies to evaluate changes of microbial community structure due to environmental stress, e.g., mercury [14,20] and herbicides [7,8]. With DGGE analysis, DNA fragments of equal length but with differences in GC-content are separated [21]. The DGGE approach is widely used and has especially been applied to describe the phylogenetic (16S rDNA) composition of natural bacterial communities [8,12,14,20,22,23].

In agricultural fields herbicides are typically applied in concentrations in the mg kg⁻¹ soil range. In contrast, herbicide concentrations measured in groundwater below such fields are often below 1 μg l⁻¹ (equivalent to 0.2 μg kg⁻¹ sediment) [1]. Higher groundwater concentrations are measured below landfills (i.e., 10–250 μg l⁻¹) [24,25], although these concentrations are still much lower than observed in surface soils. Therefore, to get a realistic picture of how herbicides affect sub-surface microbial communities, studies need to be carried out at low herbicide concentrations.

We studied the bacterial diversity and community structure in sub-surface soils from Vejen aquifer (Denmark) that had been continuously exposed in situ to a mixture of herbicides in concentrations <40 μg l⁻¹ (equivalent to <7 μg kg⁻¹ sediment) [26]. After an initial lag phase of 80–100 days, the phenoxy acids mecoprop (2-[2-methyl-4-chlorophenoxy]propionic acid) and dichlorprop (2-[2,4-dichlorophenoxy]propionic acid) were degraded rapidly in situ, and herbicide concentrations decreased to below detection within the first meter from the injection wells [26]. The highest exposed sediments had an increased potential for phenoxy acid degradation [27] and prevalence of specific herbicide degraders as well as catabolic genes (tfd) [28]. This effect was most pronounced in the aquifer sediments A1, A2 and B2 sampled close to the injection (Fig. 1). The aim of the present study was to evaluate if the in situ exposure to realistic low herbicide concentrations in the Vejen aquifer had also affected the overall structure and diversity of the indigenous sub-surface bacterial communities. This was investigated by use of colony morphology typing, sole-carbon-source utilisation in Biolog^®EcoPlates, and DGGE analysis. Principal component and cluster analysis was used to group sediments with similar bacterial community structure. Two measures of diversity were evaluated: the richness – determined as the number of colony morphotypes, carbon sources used in EcoPlates, and DNA bands in DGGE gels – and the Shannon diversity index [29].

1

Field plot of Vejen aquifer. Plan view showing the bromide tracer plume and the location of sediment sampling points inside (A1, A2, B1, B2, C1 and C2) and outside (NX1–NX4) the herbicide-exposed area. At sampling (40 days after the field injection had been terminated), bromide as well as herbicides were no longer present within the monitoring network [26].

2 Materials and methods

2.1 Field site

Sediment was collected from an unconfined oligotrophic shallow aquifer located at Vejen (Denmark). The aquifer consists of a 10-m thick medium to coarse-grained sand deposit (glacio-fluvial) with occurrences of gravel, fine sand and silt [30]. The hydraulic conductivity was 2 × 10⁻⁴ m s⁻¹ and the groundwater velocity approximately 0.5 m day⁻¹[26]. The investigated part of the aquifer was aerobic with oxygen contents of 7.0–8.7 mg l⁻¹ and the pH ranged from 4.99 to 5.55 [27]. The temperature of the aquifer is approximately 10 °C [27]. Part of the aquifer was previously exposed to herbicides during a natural gradient field injection experiment [26]. A mixture of the herbicides mecoprop, dichlorprop, DNOC (2-methyl-4,6-dinitrophenol), bentazone (3-[1-methylethyl]-1H–2,1,3-benzoth iadiazin-4[3H]-one 2,2-dioxide), isoproturon (N, N-dimethyl-N′-[4-(1-methylethyl)phenyl]urea) and 2,4-dichlorbenzamide (BAM, a degradation product of dichlobenil) was continuously injected into the aquifer for a period of 216 days, creating a contaminant plume. Bromide was included as non-reactive tracer. The concentration of each herbicide was approximately 40 μg l⁻¹ immediately down-gradient of the injection wells. Migration of bromide and the herbicides was monitored using a dense network of multilevel samplers [26].

2.2 Sampling

In August 1999, 10 sub-surface sediment samples (each of approximately 6 kg) were collected from the herbicide-exposed area (samples A1, A2, B1, B2, C1 and C2) and from a non-exposed area outside the contaminant plume (samples NX1–NX4) of the Vejen aquifer (Fig. 1) [27,28]. Sampling of the sub-surface sediments was performed 40 days after termination of the in situ herbicide field injection [26]. Herbicide-exposed sub-surface sediment was collected as three intact cores at different distances from the injection wells (0.5 m for A sites, 1.5 m for B sites and 4.5 m for C sites). Two samples from each core were obtained: an upper sample (number 1) and a lower sample (number 2). The distance between the upper and lower sediment samples was approximately 0.5 m. The non-exposed sub-surface sediment was collected similarly at upper (NX1 and NX3) and lower (NX2 and NX4) depths. All sediment cores were collected with a stainless steel piston sampler from the saturated zone, 5.00–6.25 m below surface. The individual sediment samples were homogenised thoroughly and stored at 4 °C. Sediments for DNA analysis were stored at −20 °C. Laboratory experiments were set up a maximum of 9 days from field sampling.

2.3 Colony morphology typing

One gram of sediment was suspended in 9.0 ml sterile NaCl (0.9% w/v), vortexed for 1 min and 10-fold dilutions were prepared. Sediment suspensions were spread on nystatin-amended (50 μg ml⁻¹) R2A agar plates (Difco, Detroit, MI) and incubated at 20 °C for 4 days. A total of three 1-g sub-samples of each sediment sample were analysed. Furthermore, three replicate analyses of each sediment sub-sample were performed. Bacterial colonies were grouped into morphotypes on the basis of visual differences using the following characteristics as described by Smibert and Krieg [18]: colour, diameter, edge, smoothness, brightness, flatness and special structures. R2A plates containing approximately 100 colonies were analysed and a total of 50 colonies from each replicate sub-sample were described (i.e., 150 colonies from each sediment sub-sample or 450 colonies from each sediment sample on a total basis). The relative abundance of a specific morphotype, p_i, was expressed as the ratio between the number of colony forming units (CFUs) of that group and the total number of CFUs described. Three bacterial colonies of each morphotype were analysed by repetitive extragenic palindromic PCR (REP-PCR) analysis as described by Johnsen et al. [31] in order to evaluate if colonies belonging to different morphotypes were also genetically unrelated.

2.4 Sole-carbon-source utilisation

The utilisation of 31 single carbon sources by the Vejen sub-surface bacterial communities was measured in Biolog^®EcoPlates (Biolog, Hayward, CA). The EcoPlates were inoculated with NaCl-extracts (see Section 2.3) originating from three individual 1-g sub-samples of each sediment to a density of 1.0 × 10⁶ bacterial cells well⁻¹. Bacterial densities were determined from the same NaCl-extracts by direct counting of three replicates of each sub-sample by epifluorescence microscopy (Olympus BX50, Olympus, Hamburg, Germany) after staining with 4′,6-diamidino-2-phenyl-indoldihydrochlorid (DAPI) (2 μg ml⁻¹) for 10 min in darkness [28]. The plates were incubated at 20 °C in darkness in plastic bags containing water-soaked paper towel in order to minimise evaporation from the wells. The turnover of each carbon source was measured spectrophotometrically (A₆₅₀) by reduction of tetrazolium violet to formazan. Readings were performed every 12 or 24 h for a 7-day period by use of a microplate reader (ThermoMax, Molecular Devices, Sunnyvale, CA) and subtracted the absorbance of the control well (without carbon source). The relative use of a specific carbon source, p_i, was expressed as the ratio between the A₆₅₀ of the specific EcoPlate well and the total A₆₅₀ of all 31 wells.

2.5 DGGE analysis

Community DNA was extracted from three 0.5-g sub-samples of each sediment combining bead beating and freeze–thaw procedures as described previously [28]. The two bacterial primer pairs 338f and 518r [22], and 518f and 785r [32] were used to amplify different regions of the 16S rDNA gene in order to comply with the preferential amplification of some DNA sequences compared to others [33]. A 40-bp GC-clamp was added to the 338f and 785r primers in order to increase the separation of DNA bands in DGGE analysis [22]. Five μl of DNA extracts were added to 50 μl of PCR mixtures containing: 1× GeneAmp^® PCR Buffer (Perkin–Elmer, Applied Biosystem, Foster City, CA), 100 μM of each dNTP (Perkin–Elmer), 0.5 μM of each primer (Gibco-BRL Custom Primers, Life Technologies, Paisley, Scotland) and 1.25 U AmpliTaq Gold^TM DNA polymerase (Perkin–Elmer). PCR was performed using the program: 10 min at 95 °C; 35 cycles of 30 s at 95 °C, 30 s at 55 °C and 1 min at 72 °C; 6 min at 72 °C. The presence of 16S rDNA segments was evaluated by agarose (1.5%) gel electrophoresis and the amount of amplified bacterial DNA was estimated by comparing band intensities to a standard curve derived from the intensities of the bands of a 100-bp molecular weight size marker (Promega, Madison, MI). Quantitative measurements were performed on a Gel Doc 2000 system (Bio-Rad, BioRad Laboratories, Inc., Hercules, CA) using the image analysing software Multi-Analyst^®/PC (Bio-Rad). DGGE was performed with the Dcode^TM Universal Mutation Detection System (Bio-Rad). Equal amounts of PCR product were loaded onto 8% (w/v) polyacrylamide gels (40% acrylamide/bis solution, 37.5:1, Bio-Rad) in 1× TAE (40 mM Tris, 20 mM acetate and 1.0 mM Na₂-EDTA). Polyacrylamide gels with a denaturant-gradient ranging from 25% to 60% (100% denaturant contains 7 M urea 40% (w/v) deionised formamide) were run at 60 °C for 17 h. After electrophoresis, gels were stained with SYBR^®Gold (1:10,000 dilution, Molecular Probes, Eugene, USA) for 1 h and briefly rinsed with Milli-Q water. Gels were analysed using the Multi-Analyst^®/PC software (Bio-Rad). The number of DNA bands in each lane was recorded. The relative abundance of a specific DNA band, p_i, was expressed as the ratio between the intensity of that band and the total intensity of all DNA bands in the lane.

2.6 Bacterial diversity and community structure analyses

Shannon diversity indices, H′, of colony morphology typing, sole-carbon-source utilisation and DGGE analysis were calculated using the equation H′=−∑p_ilog_ep_i[29], where p_i indicate the relative abundance of a biotype (see also the description of specific methods). The richness was determined as the number of colony morphotypes appearing on R2A, sole-carbon-sources used in EcoPlates and DNA bands in DGGE gels. Sediments were divided into groups of higher and lower H′ and richness based on statistically significant differences using Student's t-test (0.05 significance level). Principal component analysis (PCA) was performed on relative data (p_i-values) of all community structure analyses using the software MATLAB with the 5.3 PLS toolbox from Eigenvector Research (Eigenvector Research, Inc., Manson, WA). To evaluate the bacterial community structure and possible relation between sediments, cluster analysis was performed applying the PCA data. All principal components that could explain a variation of >5% between samples were included in the analysis. A distance of 0.15 to K-Means Nearest Group in cluster analysis was set as limit to discriminate between groupings. Cluster analysis was performed on data from all three sub-samples for colony morphology typing and sole-carbon-source utilisation. For DGGE, however, the analysis was only performed on one replicate set of data as only one sub-sample of each sediment could be run on the same gel. The equitability, J′=H′/H′_max, was calculated in order to compare Shannon indices.

3 Results

3.1 Bacterial diversity and community structure based on colony morphology typing

The number of bacterial heterotrophs appearing on R2A was significantly higher for the herbicide-exposed sediments A1, A2 and B2 (3.55–6.98 × 10⁷ CFU g⁻¹) compared to all other samples (P<0.01). Sediments B1, C2 and NX2 showed the lowest population sizes (7.50 × 10³–1.56 × 10⁴ CFU g⁻¹) (P<0.05). For sediments C1, NX1, NX3 and NX4, the number of culturable bacteria was between 1.64 × 10⁵ and 1.55 × 10⁶ CFU g⁻¹. Sediments A1, A2 and B2 also showed the highest number of colony morphotypes (P<0.05) (Fig. 2 and Table 1), however, generally there was no correlation between the number of CFUs and the number of bacterial morphotypes. Based on colony morphology typing, the aquifer sediments were distributed into two superior groups of richness with sediments A1, A2 and B2 harboured in one group and all the other sediments in another group. A total of 36 colony morphotypes were described, but none were represented by all sediment samples. The morphotypes 3 and 7 were observed in all the herbicide-exposed sediments, except for B1, and were not represented among the non-exposed samples (Fig. 2). In contrast, no morphotype was exclusively represented among the non-exposed sediments. Cluster analysis based on colony morphology typing (Fig. 3A) revealed that the aquifer sediments were distributed into seven structural groups: (1) A1, A2 and B2, (2) B1, (3) C1, (4) C2, (5) NX1, NX3, (6) NX2 and (7) NX4. For the non-exposed sediments NX1–NX4, single morphotypes constituted 65–91% of the heterotrophic populations (Fig. 2). In contrast, the herbicide-exposed sediments (except for B1) had a more even distribution, with the largest proportion constituted by a single morphotype being 26–44% (Fig. 2). These data indicate that the diversity of colony morphotypes within the herbicide-exposed sediments was higher compared to the non-exposed samples. This was also seen from the Shannon indices (H′) showing that the diversity of sediments A1, A2, B2, C1 and C2 was significantly higher compared to all other samples (P<0.05) (Table 1), thus distributing the sediments into two superior groups.

2

Relative incidence, p_i, of bacterial colony morphotypes in Vejen aquifer sediments. Data represent means ± SD of three replicate measurements.

1

Shannon indices (H′) and richness values of Vejen aquifer microbial communities based on colony morphology, EcoPlate and DGGE analyses^a

Sediment	Colony morphology		EcoPlate (72 h)		DGGE (338–518)
	H′	Morphotypes	H′	Carbon sources	H′	DNA bands
Herbicide-exposed
A1	2.05 ± 0.19	11.7 ± 1.15	2.66 ± 0.02	16.0 ± 0.00	2.93 ± 0.27	18.7 ± 0.58
A2	1.75 ± 0.16	11.7 ± 0.58	2.72 ± 0.04	17.7 ± 0.58	2.46 ± 0.23	12.0 ± 0.00
B1	0.84 ± 0.11	4.00 ± 1.00	1.08 ± 0.01	3.00 ± 0.00	2.17 ± 0.17	9.00 ± 0.00
B2	1.95 ± 0.06	11.3 ± 0.58	2.69 ± 0.14	16.3 ± 2.52	2.69 ± 0.20	14.7 ± 0.58
C1	1.64 ± 0.32	8.00 ± 1.00	2.39 ± 0.07	11.3 ± 0.58	2.58 ± 0.11	13.3 ± 0.58
C2	1.40 ± 0.36	7.33 ± 2.08	2.39 ± 0.08	11.7 ± 0.58	2.42 ± 0.13	12.0 ± 0.00
Non-exposed
NX1	0.95 ± 0.17	6.33 ± 1.53	2.57 ± 0.02	15.0 ± 0.00	2.81 ± 0.21	18.0 ± 1.00
NX2	0.81 ± 0.06	3.33 ± 0.58	0.59 ± 0.55	2.00 ± 1.00	2.60 ± 0.19	13.3 ± 0.58
NX3	0.45 ± 0.10	4.67 ± 0.58	2.46 ± 0.02	13.0 ± 0.00	2.25 ± 0.15	11.0 ± 0.00
NX4	0.39 ± 0.07	4.33 ± 0.58	2.04 ± 0.00	9.00 ± 0.00	2.22 ± 0.17	10.0 ± 0.00

^aThe Shannon index (H′) was calculated from the equation H′=−∑p_ilog_ep_i[29], where p_i represent the relative incidence of a colony morphotype, use of carbon sources in EcoPlates and DNA bands in DGGE gels. The richness was calculated as the numbers of bacterial colony morphotypes, carbon sources used in EcoPlates and DNA bands in DGGE gels. Data are represented as mean values±SD of three replicate measurements.

3

Relatedness of the microbial communities in Vejen aquifer shown by cluster analysis derived from data of: (A) colony morphology typing, (B) sole-carbon-source utilisation in EcoPlates after 72 h of incubation and (C) DGGE analysis using the primer pair 338f and 518r. The cluster analyses were performed on PCA analysed data (see Section 2).

REP-PCR was performed on three representatives of each of the 36 colony morphotypes (data not shown). REP sequences were successfully amplified only from 25 of the morphotypes. Among those, 19 gave rise to identical REP-PCR patterns for colonies belonging to the same morphotype, while six bacterial morphotypes differed between replicates. None of the bacterial morphotypes had REP-PCR patterns that were identical to other morphotypes, though. This indicates that bacterial colonies belonging to different morphotypes were also different at the genetic level.

3.2 Bacterial diversity and community structure based on sole-carbon-source utilisation

In order to discriminate between immediate and potential capabilities of the aquifer communities, the sole-carbon-source utilisation pattern of Vejen aquifer samples was followed in EcoPlates during a 7-day incubation period. Two time points were chosen for detailed analyses of the bacterial diversity and community structure: one at early (72 h) (Fig. 4) and one at late (168 h) (data not shown) stages of carbon substrate use. The early carbon source utilisation response was similar for the majority of the sediments but B1, NX2 and NX4 each showed unique community patterns (Fig. 4). This was further illustrated by cluster analysis (Fig. 3B). PCA analysis was likewise performed on all the other data collected during the 7-day incubation of EcoPlates and similar community structures were found for all time points with sediments A1, A2, B2, C1, NX1 and NX3 constantly grouping together, while B1, NX2, NX4 and occasionally C2 each grouped separately (data not shown). Four significantly different groups could be separated for the richness ((1) A1, A2, B2, NX1 and NX3, (2) C1 and C2, (3) NX4, and (4) B1 and NX2) and the Shannon indices ((1) A1, A2, B2 and NX1, (2) C1, C2 and NX3 (3) NX4, and (4) B1, NX2) (Table 1). The late carbon substrate utilisation responses resulted in similar groupings of the sub-surface sediments as the early response with respect to community structure, richness as well as Shannon diversity (data not shown). Looking at the specific carbon sources, no general difference was found between the herbicide-exposed and non-exposed samples. The only discrimination was the transformation of phenylethylamin (carbon source 30) after 72 h in the herbicide-exposed sediments A1, A2 and B2. d-Mannitol (carbon source 11) was the only carbon source transformed by all sediment communities following 72 h of incubation. After 168 h, also d-xylose (carbon source 9), N-acetyl-d-glucosamine (carbon source 12) and l-asparagine (carbon source 25) were transformed in all samples. There was no carbon source that was not transformed by at least one of the sediments following the 7-day incubation of the EcoPlates.

4

Relative utilisation of carbon sources, p_i, by the microbial communities of Vejen aquifer sediments after 72 h incubation in EcoPlates. Data represent means ± SD of three replicate measurements. Carbon source: 1, pyruvic acid methyl ester; 2, Tween 40; 3, Tween 80; 4, α-cyclodextrin; 5, glycogen; 6, d-cellobiose; 7, α-d-lactose; 8, β-methyl-d-glucoside; 9, d-xylose; 10, I-erythritol; 11, d-mannitol; 12, N-acetyl-d-glucosamine; 13, d-glucosaminic acid; 14, glucose-1-phosphate; 15, d, l-α-glycerol phosphate; 16, d-galactonic acid-γ-lactone; 17, d-galacturonic acid; 18, 2-hydroxy benzoic acid; 19, 4-hydroxy benzoic acid; 20, γ-hydroxybutyric acid; 21, itaconic acid; 22, α-keto butyric acid; 23, d-malic acid; 24, l-arginine; 25, l-asparagine; 26, l-phenylalanine; 27, l-serine; 28, l-threonine; 29, glycyl-l-glutamic acid; 30, phenylethylamine; 31, putrecine.

3.3 Bacterial diversity and community structure based on DGGE analysis

The genetic diversity and community structure was evaluated by DGGE analysis by applying the two primer pairs 338f and 518r (Fig. 5), and 518f and 785r (data not shown). With each primer pair, all sediment samples had unique DGGE patterns although some DNA bands were represented in every sample. No DNA bands were unique for the herbicide-exposed sediments. The DGGE patterns of the three replicate sub-samples of each sediment sample were very alike (data not shown), which is also illustrated by the low standard deviations obtained for both the Shannon indices and richness values (Table 1). Cluster analysis was performed on single replicate data solely (Fig. 3C) due to the inability of running all triplicate samples on the same gel (see Section 2). However, each triplicate sample can be considered to group tight in cluster analysis due to the high similarity within replicates. For the primer pair 338f and 518r, the sediment communities could be divided into six structural groups: (1) A1, A2 and B2, (2) B1, (3) C1, NX1 and NX3 (4) C2, (5) NX2, and (6) NX4. This grouping was very different compared to the one obtained with the primer pair 518f and 785r: (1) A1, B2 and NX1, (2) A2, B1 and NX4, (3) C1, C2 and NX2, and (4) NX3. Furthermore, DGGE analysis performed with the primers 338f and 518r (Table 1) resulted in higher values of richness (mean 13.2, maximum 19) than obtained with 518f and 785r (mean 10.0, maximum 13). This also gave rise to differences in the richness-based grouping with seven significantly different groups obtained with the primers 338f and 518r (Table 1): (1) A1 and NX1, (2) B2, (3) C1 and NX2, (4) A2 and C2, (5) NX3, (6) NX4, and (7) B1, but only three with the primers 518f and 785r (data not shown). Similarly, a higher Shannon index was obtained with the DGGE (338–518) analysis (mean 2.51, maximum 2.93) (Table 1) compared to the DGGE (518–785) analysis (mean 1.96, maximum 2.48) (data not shown). The DGGE (518–785) analysis thus had a lower resolution and resulted in lowered diversity indices (J′= 0.85) compared to the application of the primers 338f and 518r (J′= 0.97). With both DGGE analyses no significant sub-grouping of the sediments could be done (P>0.05) (Table 1, and data not shown).

5

DGGE analysis of Vejen aquifer community DNA using the primer pair 338f and 518r. NC, negative control without DNA template.

4 Discussion

Generally for all the 10 sub-surface sediments, and for each of the three community analyses, it was found that the analysis of three individual sub-samples gave rise to similar results. This means that the different structure and diversity found between the sediment samples was due to true differences in the bacterial community composition. The 10 sediments were collected from the field within a small area and a maximal horizontal and vertical distance of 5 and 0.5 m, respectively. This illustrates that the microbial community structure was very heterogeneous, even within a small distance. Similar observations were done by Fredrickson et al. [34] who found that the heterotrophic bacterial community of deep cretaceous sediments differed between segments originating from the same core. Many parameters may affect the microbial community composition in natural environments. For the Vejen aquifer, the pH and redox potential was very constant between the 10 sampling sites, however, some differences were found in concentrations of various inorganic ions and physical parameters of the sediments [27]. The high microbial heterogeneity found in sub-surface environments emphasises the importance of collecting several samples from each site when evaluating the impact of in situ contamination on natural microbial communities.

The three methods that were applied in this study describe different biological parameters and furthermore their analysis is performed on diverse sub-populations of the bacterial community. A stringent comparison between the methods is therefore not possible. Despite hereof, convergent results were obtained when analysing the bacterial community structure of the aquifer sediments with the three methods. Thus, with the cluster analysis the sediments A1, A2 and B2 always grouped together, as did NX1 and NX3. In contrast, the sediments B1, NX2 and NX4 each revealed unique community structures different from all the other sediments.

This convergence, however, was not found with the richness and Shannon diversity analyses. We find it relevant, though, to look at the data in combination. For example it could be speculated that a higher number of genotypes found in the DGGE analysis would result in a higher number of phenotypes measured by colony morphology typing since both methods attempt to describe bacterial taxons. However, such correlation was not found since the correlation coefficient obtained when comparing the richness of colony morphology typing with the two DGGE analysis was 0.26 (338–518) and 0.21 (518–785). This lack of correlation was possibly due to the divergent bacterial populations that are assayed with the two methods. However, there may also be experimental limitations with the ability of the methods to describe bacterial taxons. Concerning the DGGE analysis, each DNA band theoretically corresponds to a bacterial taxon. This view is too simplified, though, as not all 16S rDNA gene segments of different sequence can be separated during electrophoresis, and furthermore one bacterial isolate can harbour a variable number of heterogenous rRNA genes [23]. With the colony morphology typing we found that bacterial colonies belonging to different morphotype were also different at the genetic level. This finding is in agreement with observations done by Haldeman and Amy [17]. Although the colony morphology typing is a very subjective approach it may thus seem fairly applicable to discriminate between bacterial taxons. However, with this approach only approximately 1% of the total bacterial population is encountered. With the DGGE analysis, on the other hand, the dominating taxons are detected, and species that constitute as low as 1% of a population can be detected [22].

It could also be speculated that an increasing number of bacterial biotypes would result in a likewise increase in the number of carbon sources transformed in the EcoPlates. A moderate correlation was found when comparing the richness of the EcoPlate analysis with that of the colony morphology typing (R²= 0.67), however, the same was not obtained when comparing with the richness in the DGGE analyses (R²= 0.23 for both DGGEs). This discrepancy may be due to the fact that the colony morphology typing is a culture-dependent method, while the DGGE analysis is not. The Biolog analysis is not a truly culture-dependent assay, but it is generally accepted that it selects for the minority of the bacterial population (mainly γ-proteobacteria) that are fast-growing and best suited for laboratory conditions [35]. Therefore, data of the Biolog analysis may be better correlated with the colony morphology typing than with the DGGE analysis.

When comparing the Shannon indices between the individual methods, the correlation coefficients were low and ranged from 0.11 to 0.46. Constantly, the DGGE (338–518) analysis showed the best correlation to the other methods. Nübel et al. [13] similarly compared the Shannon indices obtained from different methods, when describing the microbial diversity of oxygenic phototrophs in microbial mats. The higher correlation coefficients (R²= 0.46–0.69) found in their study may partly be due to the fact that they solely applied cultivation-independent techniques.

In situ exposure by low herbicide concentrations in Vejen aquifer showed the highest impact on the acclimation of the microbial communities of sediments A1, A2 and B2, when evaluated by measurements of increased phenoxy acid degradation potentials [27] and prevalence of specific herbicide degraders and catabolic genes (tfd) [28]. The present study showed that these same sediments constantly grouped together in the various cluster analyses. This indicates that the in situ herbicide-exposure also had an impact on the overall bacterial community structure of the Vejen aquifer. In contrast, no effect was seen with respect to richness and Shannon indices, except for data obtained with the colony morphology typing. The microbial communities of sediments B1 and C1 were less acclimated to the phenoxy acid herbicides compared to those of A1, A2, B2 and C2 [27,28]. This may in part be due to a lower in situ phenoxy acid exposure in these sediments [27] and a lack of tfd genes involved in the transformation of phenoxy acid herbicides [28]. In the present study all community analyses demonstrated that sediment B1 had a unique community structure and consistently grouped distinct to the other aquifer sediments. Furthermore, B1 always belonged to the groups of lowest richness and Shannon indices. It might thus be that the low heterotrophic density and generally low diversity at B1 explain the lack of acclimation to herbicide degradation. Sediment C2, in which tfd genes were detected [28], also had a low heterotrophic density but the richness and Shannon indices were higher compared to B1, and the microbial community thus might contain a higher taxonomic and catabolic versatility.

In surface soils, herbicide-exposure can change the microbial community structure and decrease the microbial biomass [6–9]. In the Vejen aquifer, the total bacterial counts (approx. 10⁷ cells g⁻¹) were similar for the herbicide-exposed and non-exposed sediments [28]. This is in agreement with the fact that the mass of field-injected phenoxy acid herbicides at a maximum could increase the bacterial population size by 2 × 10⁵ cells g⁻¹[27,28]. On the other hand, an increase in the fraction of culturable bacteria was recorded for the herbicide-exposed sediments A1, A2 and B2. We suggest that the observed increases in culturable bacterial density in these sediments could be due to a physiological response imposed by the activity of indigenous phenoxy acid degraders of Vejen aquifer. During in situ herbicide degradation, soluble carbon is probably released (i.e., waste products from carbon metabolism) and the presence of these small substrate concentrations might have caused a general change in the physiological state of the microbial community resulting in an impact on the culturability of the microorganisms. The majority of indigenous microorganisms in natural environments can be considered non-culturable, which is partly due to the physiological state of the cells, i.e., they can have entered a starvation or a viable but non-culturable state [36]. However, recovery from starvation (i.e., resuscitation) can be fast and can result from the presence of even subgrowth levels of exogenous substrates [36–38]. Bakken [11] also suggests that there may be a correlation between the metabolically active cells and the culturable cells of a microbial community. In laboratory batch experiments, phenoxy acid turnover was significantly higher in sediments A1, A2 and B2 than in all other sediments [27], thus indicating that the general activity of these sediments was possibly also the highest.

Similar to our observations for the Vejen aquifer, Yang et al. [9] found that the genetic diversity in herbicide-exposed surface soils was kept at high levels. In contrast, Atlas et al. [6] registered a decrease in the genetic diversity after treatment of soils with 2,4,5-trichlorophenoxyacetic acid. It must be noted, however, that the concentrations of herbicides applied to these surface soils were approximately 1000 times higher (mg kg⁻¹) compared to the concentrations measured in Vejen aquifer. This must be taken into account when evaluating the effect of contaminants on microbial diversity and stresses the need of using relevant concentrations in order to make realistic considerations.

The results of this study indicate that in situ exposure of sub-surface aquifers to realistically low concentrations of herbicides may alter the composition of the overall bacterial community. This effect, however, was only found when looking at the structural composition in terms of cluster analysis. Except for the culturable fraction, no significant effect was found on the bacterial diversity. The exposure of sub-surface microbial communities to natural herbicide concentrations thus may not alter their general catabolic versatility – and in this respect does no harm. As the density of microbial phenoxy acid degraders only constituted approximately 0.1% of the bacterial population [28], the observed impacts were most likely due to indirect effects. This emphasises the importance of including methods that more specifically describe the presence and activity of relevant microbial sub-populations and functional genes, as the effect of contaminant exposure may very likely be seen only on a minority of subgroups and thus will vanish in the description of the total microbial community.

Acknowledgements

This work was supported by the Danish Environmental Research Programme, Pesticides and Groundwater. We gratefully acknowledge the assistance of Spire Maja Kiersgaard and Rasmus Rune Hansen with the field sampling.