Introduction

It has been argued that because of their small size and high rates of dispersal, microorganisms have a global distribution and therefore microbial taxa do not exhibit biogeographical patterns and are ubiquitous (Baas-Becking, 1934). The extent, to which this is true, is a matter of considerable debate (Bell et al., 2005b; Fenchel and Finlay, 2005; Whitfield, 2005) and has profound implications for the extent of global microbial diversity, its conservation and its exploitation. The consequence of ubiquitous dispersal and colonization would be the mixing of all species across all continents and none of the regional differences that are familiar in the ecology of larger organisms, for example, the occurrence of marsupials only in Australia and South America. There is evidence to support the notion of ubiquity of some microbial eukaryotes (Finlay, 2002) but equally, there is compelling evidence of restricted distribution, species area relationships and endemism, in microbial communities and populations (Cho and Tiedje, 2000; Oda et al., 2003; Papke et al., 2003; Whitaker et al., 2003; Green et al., 2004; Horner-Devine et al., 2004; Bell et al., 2005a; Reche et al., 2005; van der Gast et al., 2005). To date, claims for bacterial endemism have been based on the observed geographical distribution of particular bacterial taxa correlated with their genetic divergence and measured environmental variables (Hughes Martiny et al., 2006). These observational case studies are a powerful means of developing theoretical constructs to explain the apparently endemic distribution of some microbial taxa. However, in addition to true endemism, there are alternative ecological explanations for patterns that suggest the occurrence of endemic bacterial communities (Rauch and Bar-Yam, 2004) and there is a need for experimental testing of theories of biogeography. This was highlighted recently as an important but neglected area of ecological research (Thorpe, 2005). One of the major limitations to assessing apparently endemic distribution patterns in bacterial communities is our poor knowledge of habitat heterogeneity and the selection pressure this imposes on bacteria (Baas-Becking, 1934; Horner-Devine et al., 2004).

Achromatium are large uncultured sulfur-oxidizing bacteria found in freshwater sediments (Gray and Head, 1999; Head et al., 2000). Although Achromatium are relatively large compared to most bacteria, their size is fairly typical of many eukaryotic microorganisms and well below the size limitation proposed to be permissive for effective dispersal by agents such as wind, animals and birds (Finlay, 2002). As such, these organisms represent a useful model group for studying the ecological mechanisms dictating the distribution of microbial diversity. Achromatium species form a coherent phylogenetic group within the Gammaproteobacteria and share distinct characteristics (cell size and shape, presence of calcite and sulfur inclusions; Gray et al., 1999a, 1999b) which indicate that members of the genus are functionally similar. Most communities of Achromatium spp studied in any detail occur in shallow, organic carbon and iron-rich freshwater sediments at circumneutral pH, where concentrations of dissolved sulfide are below detection limits. These low-sulfide conditions are unlike those typically encountered in environments that harbour other giant sulfur bacteria, where dissolved sulfide is abundant. As a result, Achromatium cannot form conspicuous mats and cannot rely on the diffusive flux of free sulfide, as do other giant sulfur bacteria. Instead, they are distributed throughout the zone of sulfate reduction and may even use other reduced sulfur resources (Gray et al., 1997; Gray and Head, 1999; Head et al., 2000). Achromatium communities comprise coexisting species, some of which exploit different redox-defined niches as a basis for coexistence (Gray et al., 2004). A survey of the geographical distribution of Achromatium in lake sediments in the North of England and Germany has indicated that individual lakes are dominated by different genetically distinct Achromatium spp (Gray et al., 1999a). Of these freshwater lakes, two (Rydal Water (RY) and Hell Kettles (HK)) are located in the North of England and are separated by approximately 60 miles.

Purely observational data on the geographical distribution of Achromatium species in RY and HK (Gray et al., 1999a) suggesting that each community comprised endemic Achromatium species, can be interpreted in three ways, illustrating the difficulty in interpreting the ecological significance of purely observational data. Firstly, the lakes may have been colonized by a single ancestral Achromatium sp, which, following geographical separation, diverged by adaptive radiation (Rainey and Travisano, 1998) to produce the distinct communities of Achromatium spp observed in each lake today (that is, true endemism). Alternatively, the sites are geographically remote with respect to bacterial migration (that is, island-like) (MacArthur and Wilson, 1967), and their community composition is a result of restricted dispersal of Achromatium spp and undersaturation with respect to global Achromatium diversity. Here it would follow that regardless of the existence of truly endemic Achromatium species, not all possible or best adapted Achromatium species have dispersed to every lake, a well-known phenomenon in the ecology of island communities, which leads to species impoverishment in insular communities (Begon et al., 1996). Finally, the lakes have been repeatedly seeded with a wide variety of Achromatium spp and the predominant Achromatium spp observed are a consequence of selection of those most fit for the prevailing environmental conditions (Baas-Becking, 1934; Begon et al., 1996). We have used Achromatium-bearing sediments from RY and HK to experimentally test which of these mechanisms most likely explain the observed geographical distribution of Achromatium spp. The experiment was designed to simulate the dispersal of new organisms to a previously uncolonized location.

Materials and methods

Sampling and experimental setup

Grab samples of sediment containing Achromatium cells were obtained from a wetland area on the margins of Rydal Water, Cumbria, UK (54°219N, 2°519W) and Hell Kettles, located south of Darlington, County Durham, UK (54°299N, 1°339W) (Gray et al., 1999a). Sediments and samples of overlying water (5 cm3 sediment with 5 cm3 water) from RY and HK were mixed in sterile test tubes (30 ml capacity) and closed with sterile cotton wool. Sediments were mixed at three different ratios (RY volume %:HK volume %) namely 90:10, 75:25 and 50:50. In addition, unmixed sediment microcosms, that is, those containing pure sediment from one site or the other, were prepared for comparison. A set of microcosms was prepared and immediately killed as time zero controls. In these microcosms, each of the mixed sediments initially contained the accumulated diversity of the parent sediments. This experimental strategy increased the diversity of the Achromatium community but did not increase the population size above the notion carrying capacity of the artificially constructed environments. A duplicate set of sediment microcosms was incubated at 10 °C for 60 days.

Achromatium cell counts and DGGE analysis of Achromatium community structure

Prior to cell extraction for DGGE analysis of Achromatium community composition, an aliquot of well-mixed sediment (100 μl) was removed and made up to 1 ml for the determination of total Achromatium cell counts using a Sedgwick Rafter counting chamber (Gray et al., 1997). Changes in Achromatium community composition over time in the experimental microcosms were assessed by comparison of DGGE profiles of PCR-amplified Achromatium 16S rRNA gene fragments. Sediments from experimental microcosms were sequentially filtered through 100 and 64 μm nylon meshes to remove larger sediment particles. Crude cell suspensions obtained in this way were purified as described previously (de Boer et al., 1971; Head et al., 1996) and were used for DNA extraction. Many Achromatium spp precipitate intracellular calcium carbonate, a unique property that makes them very dense and amenable to the gravity-based separation procedure. All of the experimental microcosms yielded a visible cell suspension. DNA for PCR was obtained by boiling suspensions of purified Achromatium cells in Tris-EDTA buffer. 16S rRNA gene fragments were amplified using primers pA (5′-AGAGTTTGATCCTGGCTCAG-3′) (Edwards et al., 1989) and AO703r (5′-CTGGTATTCCTCCTGATATC-3′). AO703r was designed using the probe and PCR primer design software tool Primrose (Ashelford et al., 2002) and checked for specificity using the Ribosomal Database Project (RDP II) probe match analysis tool (Cole et al., 2007). Comparison of the primer sequences with RDP II Release 9.47 (1 February 2007) identified a number of matches with non-target sequences. However, AO703r is selective for Achromatium 16S rRNA gene sequences, and the use of purified Achromatium cell suspensions ensured that only 16S rRNA gene fragments from Achromatium spp were amplified using pA and AO703r. This was tested empirically by excising and sequencing bands from DGGE gels and by analyzing 16S rRNA gene clone libraries obtained with pA and AO703r.

PCR amplification was carried out according to a previously described method (Rowan et al., 2003) with an annealing temperature of 62 °C. PCR products were obtained from all 30 experimental microcosms except for a single replicate of the 90% RY:10% HK mix. PCR products generated using pA and AO703r were amplified using primer 3 (5′-CCTACGGGAGGCAGCAG-3′) containing a GC clamp (5′-CGCCCGCCGCGCGCGGCGGGCGGGGCGGGGGCACGGGGGG-3′) and primer 2 (5′-ATTACCGCGGCTGCTGG-3′) (Muyzer et al., 1993). PCR products were purified using a Qiagen PCR clean up kit (Qiagen, Crawley, UK). DGGE analysis was conducted using the D-Gene denaturing gradient gel electrophoresis system (Bio-Rad, Hercules, CA, USA). Polyacrylamide gels (10% polyacrylamide; 0.75 mm thick; 16 × 16 cm) were run in 1 × TAE buffer (40 mM Tris-acetate, 1 mM EDTA (pH 8.3)). A denaturant gradient ranging from 30% to 52% denaturant (100% denaturant is 7 M urea plus 40% (v/v) formamide in 1 × TAE) was used. Gels were run at 60 °C for 4 h at a constant 200 V and stained for 30 min in SYBR green I (1/10 000 in 1 × TAE; Sigma, Poole, UK). Stained gels were viewed using a Fluor-S MultiImager (Bio-Rad).

Validation of the PCR-DGGE approach

To validate the PCR-DGGE approach used to study Achromatium communities, bands present in DGGE profiles were confirmed as being derived from Achromatium spp by comparative analysis of partial 16S rRNA gene sequences. Sequences were obtained either directly from bands excised from DGGE gels and/or on the basis of co-migration of bands in DGGE profiles with cloned Achromatium 16S rRNA gene fragments. Bands were excised from the DGGE gel using a sterile plastic pipette tip. DNA was eluted from the gel fragments by incubation in 50 μl of Tris-EDTA buffer (20 min, 20 °C). An aliquot of the eluted liquid (1 μl) was used as template for amplification with primers 2 and 3 (Muyzer et al., 1993). The 16S rRNA fragments (each approximately 190 bp in length) were sequenced using primer 2. A clone library of longer 16S rRNA gene fragments was generated from purified Achromatium cells obtained from RY sediment using primers pA and AO703r. PCR-amplified 16S rRNA gene fragments were purified using a Qiagen PCR clean up kit (Qiagen) and cloned using a Qiagen PCR cloning kit (Qiagen) with the pDrive cloning vector and Qiagen EZ competent Escherichia coli. All procedures were carried out according to the supplier's instructions. White colonies containing inserts were selected at random, and DNA was extracted by boiling cells picked from the colony in Tris-EDTA buffer for 3 min. Amplification of DNA inserts was performed using primers pUCr (5′-CAGGAAACAGCTATGAC-3′) and pUCf (5′-GTTTTCCCAGTCACGAC-3′) according to a previously described method (Rowan et al., 2003). PCR products of the correct size from 50 randomly selected clones were screened by DGGE, after amplification with primers 2 and 3 (see above) to identify groups of different sequences from the clone library.

Eight groups of clones were identified by their migration in DGGE gels. Representative clones from each of these groups were sequenced. Initially, 23 of the cloned 16S rRNA genes were sequenced. This demonstrated that each of the groups identified on the basis of their migration in DGGE gels was homogeneous. On this basis the sequence of the entire amplified 16S rRNA gene fragment (approximately 700 bp.) was determined for a representative clone from each of the groups, using the primer pC (5′-CTACGGGAGGCAGCAGTGGG-3′) (Edwards et al., 1989). These eight 16S rRNA sequences have been deposited in the GenBank database with accession nos. DQ288131–DQ288138. All sequencing was conducted using the DyeDeoxy chain termination method using a 3730 xl DNA analyzer (Applied Biosystems, Warrington, UK).

Phylogenetic distance analysis was conducted using the Jukes and Cantor (1969) correction for multiple substitutions at a single site and the neighbour-joining method (Saitou and Nei, 1987) as implemented in the TREECON package (van De Peer and De Wachter, 1994). Bootstrap resampling was conducted with 100 replicates. The analysis included 34 sequences and covered positions 100–202 and 221–722 (E. coli numbering). Comparison of the short sequences obtained from DGGE bands with the GenBank database using BLAST (Altschul et al., 1990), and with sequences obtained in this study, confirmed that all sequences represented Achromatium spp.

For analysis of Achromatium community DGGE profiles from the microcosms, the Bionumerics software package (Applied Maths, Austin, TX, USA) was used to produce a normalized composite gel with reference to markers included on the DGGE gels (van Verseveld and Röling, 2004). Band-matching data were exported to a Microsoft Excel spreadsheet for subsequent analysis. Primer 6 for Windows (Version 6.1.5, Primer-E Ltd, Plymouth, UK) was used to construct Bray-Curtis similarity matrices for non-metric multidimensional scaling (MDS) followed by analysis of similarities (ANOSIM) using band designation as variables of presence and absence (Clarke and Warwick, 2001).

The MDS algorithm uses distance information from a Bray-Curtis similarity matrix to place each data point in low-dimensional space while maintaining distance relationships between each point. The resulting MDS plots are representations of how different the treatments are from each other based on clustering of like samples. To determine if the observed clusters are statistically significant, a pairwise ANOSIM was conducted on the similarity matrix. ANOSIM generates an R statistic, which is calculated by comparing the mean distances within user-defined groups and between groups (the latter of which should be greater if the groups are different), and ranges from 0 (completely random pattern) to 1 (treatments are completely separated). A P-value is then computed by randomizing the samples to determine if the R-value generated from the user-defined groups is greater than any randomly generated R-value.

Results

Diversity of Achromatium in the Rydal Water and Hell Kettles sites and validation of the PCR-DGGE approach employed

A total of six bands were excised from a DGGE profile (Figure 1) from a microcosm which comprised 75% RY:25% HK sediments (time zero, replicate 2). Four band-derived fragments (designated C1, C2, C3 and C6) that were also present in the parent RY community were sequenced. In addition, two further sequences (designated as C4 and C5) were obtained, which corresponded with the two principle bands present in the HK parent community. Comparative sequence analysis demonstrated that all of the six sequences from the excised DGGE bands were related to Achromatium sequences previously recovered from the RY and HK sites (Gray et al., 1999a).

Figure 1
figure 1

Band assignments for Achromatium 16S rRNA sequences obtained from bands excised from the DGGE gel (C1–C6) or by cloning and sequencing of a larger fragment of PCR-amplified 16S rRNA gene (RYA, RYB, RYE, RYG and RYH).

As an additional confirmation of the specificity of the PCR-DGGE approach, comparative analysis of the 50 cloned 16S rRNA gene fragments selectively amplified from purified RY Achromatium cells demonstrated that all eight sequence types identified (designated RYA to RYH) clustered with the Achromatium assemblage within the Gammaproteobacteria (Figure 2). Of these, RYA, RYB, RYE, RYG and RYH all co-migrated with bands identified in DGGE profiles from the microcosms (Figure 1).

Figure 2
figure 2

Phylogenetic distance tree based on the comparative analysis of partial 16S rRNA sequences (RYA to RYH) recovered from Achromatium cells purified from Rydal Water sediment. Fourteen Achromatium 16S rRNA sequences recovered from three sites in Northern England (Rydal Water (RY), Hell Kettles (HK) and Jenny Dam (JD)) and two sites in Germany (Lake Fuchskuhle and Lake Stechlin) (Glöckner et al., 1999) were included in the analysis. GenBank accession nos. for each sequence are provided in parenthesis. The tree was rooted with respect to the Desulfobacter postgatei DSM 2034T (M26633) 16S rRNA sequence. The scale bar denotes 2% sequence divergence and the values at the nodes indicate the percentage of bootstrap trees that contained the cluster to the right of the node. Bootstrap values less than 50 are not shown.

Achromatium community size in the experimental microcosms

Differences in total Achromatium community size in the experimentally mixed microcosms (Figure 3) were as a consequence of differences in the Achromatium community size in the parent sediments from which they were constructed. HK sediment sustained approximately 10 times more Achromatium cells than RY sediment. Accordingly, the microcosms represented a gradient of increasing Achromatium community size correlated with increasing contributions from the HK sediment. On this basis, although the ratio of the two sediments (RY:HK) by volume in the microcosms were 100:0, 90:10, 75:25, 50:50 and 0:100, the estimated ratios of cells and hence specific Achromatium sustaining resources contributed by the parent sediments (RY%±s.e.:HK%±s.e.) were 100:0, 41±25:59±25, 21±15:79±15, 8±6:92±6 and 0:100. In the sediment microcosms incubated for 60 days, all replicate microcosms still contained Achromatium cells (Figure 3). For all but one of the experiments (100% HK microcosms), no significant difference (P>0.05) in the total abundance of the Achromatium cells was found between day 0 and day 60. Replicating cells were observed both at time zero and after 60 days in all of the microcosms; however, despite the relatively steady-state population sizes, cell turnover rates could not be calculated because Achromatium spp have not been cultured in the laboratory and no realistic estimate of division times are available.

Figure 3
figure 3

Abundanace of Achromatium cells in sediment microcosms. Microcosms at time zero (grey bars), microcosms after 60 days of incubation (white bars). Error bars represent s.e. (n=3). The sediment ratios (Rydal Water (RY):Hell Kettles (HK)) by volume were 100:0, 90:10, 75:25, 50:50 and 0:100. The HK sediment sustained approximately 10 times the number of Achromatium cells per unit volume than RY sediment. Accordingly, the estimated ratios of cells and resources contributed by the parent sediments (RY %±s.e.:HK %±s.e.) were 100:0, 41±25:59±25, 21±15:79±15, 8±6:92±6 and 0:100.

Achromatium community composition in the experimental microcosms

A composite gel including DGGE profiles from all of the microcosms (Figure 4a) was subjected to numerical and statistical analysis (Figure 4b). It should be noted that the profiles are derived from purified preparations of Achromatium cells. It is unlikely, therefore, that any bands were derived from extracellular DNA released from dead cells.

Figure 4
figure 4

(a) Composite DGGE gel of Achromatium 16S rRNA gene fragments derived from the experimental microcosms. Lane 1–3 and 4–6 represent, respectively, Achromatium community profiles from microcosms containing 100% RY sediment at time zero ( × 3) and after 60 days ( × 3). Lanes 7 and 8 and 9–11 represent, respectively, Achromatium community profiles from microcosms containing 90% RY:10% HK sediments at time zero ( × 2) and after 60 days ( × 3). Lanes 12–14 and 15–17 represent, respectively, Achromatium community profiles from microcosms containing 75% RY:25% HK sediments at time zero ( × 3) and 60 day ( × 3). Lanes 18–20 and 21–23 represent, respectively, Achromatium community profiles from microcosms containing 50% RY:50% HK sediments at time zero ( × 3) and 60 day ( × 3). Lanes 24–26 and 27–29 represent, respectively, Achromatium community profiles from microcosms containing 100% HK sediment at time zero ( × 3) and 60 day ( × 3). (b) MDS analysis of DGGE profiles of Achromatium 16S rRNA gene fragments derived from the experimental microcosms. Coloured symbols indicate profile identity based on the ratio of sediment from the different lakes. RY:HK: 100:0 (red diamonds), 90:10 (magenta circles), 75:25 (green squares), 50:50 (light blue triangles) and 0:100 (dark blue inverted triangles). The position of symbols reflects how different DGGE profiles are from each other based on their distance in a two-dimensional plot. Distance is derived from Bray-Curtis similarity coefficients calculated from the DGGE profiles. Filled symbols indicate replicate analyses from microcosms at time zero, open symbols indicate replicate analyses from microcosms following 60 days of incubation. HK, Hell Kettles; MDS, multidimensional scaling; RY, Rydal Water.

Comparison of Achromatium community structure by MDS analysis (Figure 4b) showed that at time zero, Achromatium communities from the parent sediments 100% RY (filled red diamonds) and 100% HK (dark blue filled inverted triangles) fell into two mutually exclusive clusters, which were separated along the abscissa. This indicated that the Achromatium community structure in the two Achromatium parent communities was different at the outset of the experiment, a finding which was consistent with previous studies of RY and HK, which demonstrated that the different sediments were dominated by distinct Achromatium lineages (Gray et al., 1999a). After 60 days, the DGGE profiles from the unmixed sediments, 100% RY (open red diamonds) and 100% HK (open dark blue triangles), were still separated on the abscissa to the same extent. A comparison of the two groups by ANOSIM confirmed the separation of these communities with the maximum R statistic of 1 (P<0.01). Thus, during the incubation period, both the Achromatium population size (Figure 3) and community composition (Figure 4) were broadly maintained in the parent sediments.

Analysis of the DGGE profiles from microcosms containing mixtures of sediments, at the start of the experiment, indicated that a gradient of diversity had been established based on the initial sediment ratios. At time zero (filled symbols in Figure 4b), the 75% RY:25% HK microcosms (filled green squares) grouped midway along the abscissa but were not significantly separated from the time zero microcosms containing 90% RY:10% HK sediment (filled magenta circles) or microcosms comprising 100% RY sediment (filled red diamonds) (R=−0.108, P=0.29; ANOSIM). They were, however, separated from the communities in time zero microcosms containing 100% HK sediments (filled dark blue inverted triangles), and with one exception the communities from microcosms containing 50% RY:50% HK sediment (filled light blue triangles) (R=0.723, P=0.036; ANOSIM). This gradient of diversity, graphically represented by variation along the abscissa of the MDS plot, is consistent with the ratio of sediment from each source in the microcosms and the observed differences in population size (Figure 3). For instance, in the time zero microcosms containing 90% RY:10% HK sediment, the estimated contribution of RY cells determined by cell counts was 41%. This correlates well with the relative intensity of bands attributable to RY Achromatium (39.7±15.2) in the corresponding DGGE profiles. In the 75:25 sediment mix, the estimated contribution of Rydal cells determined by cell counts was 21% and the relative intensity of bands attributable to RY Achromatium was 26.7%±7.3%. Importantly, these results indicate that there is little bias in the PCR amplification of 16S rRNA genes from the different Achromatium communities when mixed together.

After 60 days of incubation (open symbols in Figure 4b), there was no longer any significant difference (R=−0.105, P=0.64; ANOSIM) between the Achromatium communities in the microcosms containing 75% RY:25% HK sediments (open green squares) and the Achromatium communities in the sediment containing 100% HK sediment (open blue inverted triangles) and the communities in microcosms containing 50% RY:50% HK sediments (open light blue triangles) profiles. However, there was now a significant difference between the communities in microcosms with 75% RY:25% HK sediments (open green squares), the communities in microcosms with 90% RY:10% HK sediments (open magenta circles) and the communities present in 100% RY sediment (open red diamonds) (R=0.728, P=0.012; ANOSIM). Taken together, these data indicate a shift in the communities from mixed sediments, towards the composition of the Hell Kettle community and away from the Rydal community. After 60 days, the relative intensity of bands attributable to RY Achromatium was reduced to 6.7%±3.1%. In contrast, after 60 days, the Achromatium communities in the microcosms with 90% RY:10% HK sediments (open magenta circles) were still distinct from the Achromatium communities in microcosms containing 100% HK sediment (open inverted blue triangles) and those containing 50% RY:50% HK sediments (open light blue triangles) (R=0.743, P=0.005; ANOSIM). In the MDS plots, the Achromatium communities in some of the microcosms were also clearly separated along the ordinate based on incubation time (R=1, P<0.001; ANOSIM).

Discussion

Validation of the experimental approach

It has been shown in experimental microcosms similar to those used here that redox gradients and the small-scale distribution of Achromatium species are rapidly re-established when microcosms are re-constituted from bulk sediments (Head et al., 2000). Furthermore, Achromatium community composition is maintained when re-constituted sediments are incubated (as in this study) with oxic overlying water (Gray et al., 2004). The microcosms used in this study, therefore, are suitable vehicles for the experimental manipulation and perturbation of Achromatium communities.

Nevertheless, two important objectives of our experimental design were to validate the PCR-DGGE approach used for community analysis and to discriminate general experimental effects from putative environmental selection. To this end, ‘time zero’ mixed-sediment microcosms were killed immediately to show which RY and HK Achromatium species were detectable by the PCR-DGGE approach and, therefore, provide a reference pattern to identify experimental changes in the mixed sediments over time. The subsequent disappearance of RY bands after 60 days under certain environmental conditions therefore cannot be attributed to PCR bias since it is known that when RY cells are present alongside HK cells, they were detectable in the DGGE profiles. In addition, control microcosms containing sediment from a single source were prepared using the same procedures as microcosms containing mixtures of sediments from two locations. During the 60-day incubation, the Achromatium communities in these control microcosms did not separate along the abscissa in MDS plots as was observed in some of the mixed microcosms. In contrast, significant separation of Achromatium communities along the ordinate of the MDS plots was observed in all control microcosms and in all microcosms containing sediment mixtures. On this basis, we conclude that distributions along the ordinate are attributable to general time-related effects of incubation in microcosms, whereas those along the abscissa can be attributed to environmental and community mixing and environmental selection (see below).

If environmental selection plays a pivotal role in determining the geographical distribution of related but genetically distinct species, it was reasoned that the selective pressure on species from different locations would correlate with the sediment they occupy. If sediments from two environments containing different Achromatium species are mixed, there will be a specific ratio of the two sediments where the species from one geographical location will be favoured over the other, due to the sediment conditions more closely resembling their home environment. This environmental selection and phenotypic differentiation would be manifest by the growth or maintenance over time, of particular Achromatium spp accompanied by antagonistic effects on other Achromatium spp.

The role of environmental selection in the geographical distribution of Achromatium spp

In the mixed-microcosm experiments, population size was not increased above the notional carrying capacity of the sediments even though Achromatium diversity was artificially increased. On this basis we did not anticipate, or observe (Figure 3), a major decline in total Achromatium numbers in the mixed sediments during the course of 60-day incubation. This precludes the exclusion of Achromatium spp which were present and detectable in low abundance at time zero, simply on the basis of a general decline in the Achromatium population. On this basis, the pattern of community change observed in microcosms containing 75% RY:25% HK sediments, that is, selection against the RY community and persistence of HK species even when the sediment mix contained a substantial proportion of RY sediment demonstrates that the HK Achromatium species have a selective advantage in the mixed sediments. This selective advantage was not found to be significant as the sediment mixture became dominated by RY sediment (90% RY:10% HK, Figure 4b). It should be noted (see above) that the notional resources contributed to the microcosms based on the numbers of cells sustained in the unmixed sediments meant that even the most Rydal-like sediment mix (90% RY:10% HK) had a dominance of HK-derived resources, at least in terms of those specific components of the resource pool that sustain Achromatium spp if not the overall physicochemical conditions.

Although we have not directly measured dispersal, our experiments represent an extreme manifestation of dispersal of Achromatium spp from one location to another. Our results show that even when there is large-scale immigration, the Achromatium community composition correlates with the nature of the host sediment. This indicates that environmental selection determines the composition of the Achromatium community in the two lakes. This experimental finding is consistent with observed environmental differences between the sites (Head et al., 2000) and some limited information on functional differences between the geographically separated Achromatium populations. RY has low sulfate (15–168 nmol cm−3) and high organic carbon (14.8–15.9 wt%) concentrations, Measured sulfate reduction rates are in the range of 10–300 nmol cm−3 per day and reduced sulfur species concentrations range from 3.4 to 9.5 μmol cm−3. HK sediments typically contain higher concentrations of sulfate (9870–11160 nmol cm−3) and organic carbon (21.3–22.2 wt%), which result in higher rates of sulfate reduction (1824±73 nmol cm−3 per day) and higher concentrations of reduced sulfur species (16.01–95.32 μmol cm−3). It has also been shown (by an inability to amplify specific gene targets) that unlike RY populations, HK Achromatium may lack the enzymes adenosine phosphosulfate (APS) reductase and ribulose bisphosphate carboxylase/oxygenase (RuBisCO), suggesting that the metabolic potential of the populations is different. Regardless of these functional differences, it has now been experimentally shown that the realizable ecological niches of the Achromatium species from RY are not present in the HK sediment. Accordingly, regardless of the frequency of dispersal, the RY species are unlikely to become predominant members of the HK community. We conclude that at a regional level, dispersal and environmental selection are important factors in determining the distribution of specific bacterial taxa and if this translates to larger geographical scales, then the global inventory of Achromatium will be principally dictated and restricted by habitat heterogeneity.

The initial driver for experimentally testing factors that lead to observed biogeographical patterns in bacteria was the ambiguity inherent in interpretation of purely observational environmental, functional and phylogenetic data. The default interpretation of biogeographical patterns is often endemism. However, it is apparent that these observed patterns can be interpreted in a number of ways as discussed above. Suggestions of microbial endemism (Cho and Tiedje, 2000; Oda et al., 2003; Papke et al., 2003; Whitaker et al., 2003; Green et al., 2004; Bell et al., 2005a, 2005b; Reche et al., 2005; van der Gast et al., 2005) arising from allopatric speciation are thought to be the consequence of segregation of populations when rates of evolution at a single location exceed rates of dispersal even in the absence of obvious physical barriers. Accordingly, the particular genetic marker (and its rate of evolution) used to identify putative endemic populations will affect the conclusions drawn. In this case, we have used a relatively conserved genetic marker (16S rRNA sequences) and have demonstrated that environmental selection is significant in dictating the biogeographical patterns observed in Achromatium. We have also provided an experimental approach that can be used to test the generality of this phenomenon in other bacterial taxa. Hitherto, this has been an important barrier to rigorous interpretation of biogeographical patterns in bacterial communities (Fenchel and Finlay, 2005). Nevertheless, it is likely that in different environments and for different bacteria, the relative importance of environmental selection and geographic dispersal in dictating microbial community structure will vary.