Introduction

Hypersaline ecosystems are distributed globally and represent a wide range of ecosystem types, including salt lakes, soda lakes, hypersaline springs, salt flats, playas, solar salterns, and ancient salt deposits (Terry et al., 2000; Oren, 2002; Ventosa et al., 2008). Hypersaline water, soils, and sediments harbor active and diverse microbial communities (Oren, 2002; Humayoun et al., 2003; Caton et al., 2004; Ley et al., 2006; Mesbah et al., 2007) and contain organisms representing all three domains of life (Oren, 2002). Studies of hypersaline ecosystems have improved our understanding of the biology of extreme environments, and many have resulted in the discovery of novel organisms and enzymes with enhanced potential for biotechnological applications (Caton et al., 2004; Nicholson and Fathepure, 2005; Ventosa et al., 2008). The majority of studies published on the microbiology of hypersaline ecosystems focus on aquatic communities (Oren, 2002; Humayoun et al., 2003; Demergasso et al., 2004; Mutlu et al., 2008) and/or microbial mats (Mouné et al., 2003; Ley et al., 2006), whereas far fewer have attempted to characterize hypersaline soils and sediments (David et al., 2005; Dong et al., 2006; Jiang et al., 2006; Mesbah et al., 2007; Ventosa et al., 2008).

The advent of pyrosequencing technology has provided the ability to examine microbial communities with unprecedented levels of coverage and detail (Sogin et al., 2006; Roesch et al., 2007; Sundquist et al., 2007; Acosta-Martínez et al., 2008). Pyrosequencing has enabled the production of sequence libraries that are orders of magnitude larger than conventional Sanger sequence libraries, and at a fraction of the cost (Cardenas and Tiedje, 2008). Such advances provide a major step forward in our ability to characterize microbial communities, particularly those that are highly complex, have been under-sampled historically, and/or may harbor members of the ‘rare biosphere’ (Sogin et al., 2006). Furthermore, the high-throughput, low-cost nature of pyrosequencing affords new opportunities to study the dynamics of microbial communities in time and space or under fluctuating environmental conditions.

A recent global survey of bacterial communities from natural environments found that sediment communities may be more phylogenetically diverse than those from any other environment, including soils (Lozupone and Knight, 2007), and that salinity is a major factor in determining microbial community composition (Lozupone and Knight, 2007). Given these factors and the tremendous resolution afforded by new sequencing technologies, a tag-encoded pyrosequencing study was undertaken to characterize the structure and diversity of soil and sediment microbial communities at La Sal del Ray, a naturally occurring and previously uncharacterized salt lake ecosystem in southern Texas, USA. A transect approach was used in combination with community quantitative PCR (qPCR), 16S rRNA tag-pyrosequencing, 16S rRNA gene cloning and sequencing, and multivariate statistics to characterize the lake's soil and sediment microbial communities and identify the relationships they share with their environment.

Materials and methods

Site description and sampling

La Sal del Rey is a shallow (1 m depth), 215-hectare hypersaline lake located within the Lower Rio Grande Valley National Wildlife Refuge, Hidalgo County, TX (26°31′55′′ N, 98°03′50′′ W). Mean annual temperature for the region is 23.8 °C, and annual rainfall averages 583 mm (based on National Climatic Data Center information for McAllen, TX) (NOAA, 2009). The geological history of La Sal del Rey is uncertain; however, the lake, which sits in a depression on the landscape and has no major outlets, is thought to have formed either through result wind erosion or the presence of a salt dome (Barton, 1925). Historical accounts describe its use, from pre-settlement times through the American Civil War, as a salt mine (Hawkins, 1947), and more recent descriptions of the lake report its total salinity to be as high as 177.9 g l−1, with NaCl as its dominant salt (Cole and Brown, 1967).

Soil and sediment samples were collected in June 2008 from La Sal del Rey's shoreline and lakebed. Samples were collected at eight points (T3-0, T3-65, T3-130, T3-195 T3-260, T3-325, T3-390, and T3-455), spaced 20 m from one another, along a 140 m transect. The transect originated in a vegetated, upland area (that is site T3-0), extended across exposed lakebed sediments, and continued toward the water's edge (that is sites T3-390 and T3-455). From point T3-0 to point T3-455, the sediments became increasingly water-logged, grew darker in color, and smelled increasingly of volatile sulfur compounds. Upland vegetation included mesquite (Prosopis glandulosa), cactus (Opuntia lindheimeri), and scattered grasses, including Setaria leucopila, Hilaria berlangeri, and Digitaria californica. Sparse patches of shoregrass (Monanthochloe littoralis) were also present. The lake appeared to have contracted in area over the course of the spring, exposing portions of the lakebed and producing an ice-like layer of salt over some points along the transect.

Surface soils and sediments were collected to a depth of 10 cm using a cylindrical bulb planter with a diameter of 6.5 cm. Temperatures were collected at mid-depth to characterize the ‘average’ conditions at each sampling point, recognizing that the variation in temperature between the surface and depth has the potential to be large. In cases where solid salt crusts were present, they were removed before sampling. Samples were placed into sterile plastic bags, sealed, and held on ice until return to the laboratory. One portion of each sample was dried at 45 °C and used for physical and chemical analysis, and a second portion was frozen at −80 °C and used for DNA extraction.

Physical and chemical analysis

Samples were dried to a constant weight in a forced air oven at 45 °C. They were then sieved to 2 mm and submitted to the Soil, Water, and Forage Testing Laboratory at Texas A&M University for pH (Schofield and Taylor, 1955), detailed salinity (Rhoades and Clark, 1978), electrical conductivity (Rhoades, 1982), and phosphorus, magnesium, calcium, and sodium analyses (Mehlich, 1978), as well as organic carbon and total nitrogen analysis (McGeehan and Naylor, 1988).

DNA extraction and purification

Community DNA was extracted from transect sediments using a PowerMax DNA isolation kit (Mo Bio Laboratories Inc., Carlsbad, CA, USA) and a modified version of the manufacturer's protocol. Fifteen gram aliquots of sediment and 15 ml of bead solution were added to each bead-beating tube. After 5 min of bead beating, lysozyme was added (final concentration of 1 mg ml−1), and samples were incubated at 37 °C for 1 h. Following lysozyme treatment, solution ‘C1’ was added and samples were incubated at 65 °C for 30 min. The manufacturer's protocol was followed from this point onward. Following elution, DNA samples concentrated by ethanol precipitation, resuspended in 10 mM Tris (pH 8.0), and purified using a 1.2% (w/v) low-melting point agarose gel. DNA bands were excised from the gel, extracted using a Wizard SV Gel and PCR Clean Up kit (Promega, Madison, WI, USA), and quantified with a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA).

Community quantitative PCR

Community qPCR assays, based on Fierer et al. (2005) and Lima and Sleep (2007), were used to evaluate the relative abundances of Bacteria and Archaea along the survey transect. Assays were performed in triplicate, using an Eppendorf Mastercycler ep realplex thermal cycler (Eppendorf, Hamburg, Germany). Each 10 μl reaction contained: 4.5 μl 2.5 × RealMasterMix with 20 × SYBR solution (5Prime, Inc., Gaithersburg, MD, USA), 1.0 μl BSA (10 mg ml−1), 0.5 μl of each primer (10 μM,), 2.5 μl molecular grade water, and 1.0 μl template DNA (2.5 ng μl−1). Thermocycling consisted of an initial denaturation at 95 °C for 15 min, followed by 40 cycles of 95 °C for 1 min and annealing temperature (see below) for 30 s, and 72 °C for 1 min. Primer set Eub338/518 was used with an annealing temperature of 53 °C for Bacteria (Fierer et al., 2005) and primer set ARCH85F/313R was used with an annealing temperature of 61 °C for Archaea (Lima and Sleep, 2007).

Plasmid standards for Bacteria and Archaea were generated from the genomic DNA of Escherichia coli DH10B(pUC19) (obtained from Carlos Gonzales, Texas A&M University) and Methanosarcina acetivorans C2A (obtained from William Metcalf, University of Illinois). Regions of interest were amplified from each organism using a FailSafe PCR kit (Epicentre Biotechnologies, Madison, WI, USA) and their corresponding qPCR primer sets. The PCR products were confirmed for size on a 1.2% (w/v) low-melting point agarose gel with UV light, excised, and extracted using a Wizard SV Gel and PCR Clean Up kit (Promega). Then, they were cloned into a pGEM-T Easy vector (Promega), and plasmids were isolated using a Wizard SV Miniprep kit (Promega). Plasmid DNA concentrations were quantified using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies) and serial dilutions of 5.0 × 10−3 to 5.0 × 10−7 ng μl−1 DNA were amplified to generate standard curves.

Amplification efficiency was calculated as described by Pfaffl (2001). rRNA gene copy numbers were estimated from the plasmid-based standard curves, and were calculated as a function of DNA concentration, the average molecular weight of a base pair of DNA (660 g mol−1), the size of our cloning vector, and the size of our DNA fragment of interest. Standard curves generated from our bacterial and archaeal controls displayed linear relationships between the log-values of their gene copy numbers and calculated threshold cycles (r20.95). Amplification efficiencies were 1.85 and 2.49 for Archaea and Bacteria, respectively, and were within the range of those reported by others (Fierer et al., 2005 and references therein). Gene copy numbers were adjusted to accommodate the potential for Bacteria and Archaea to carry multiple copies of the 16S rRNA gene. Average values of 4.14 copies bacterium−1 and 1.70 copies archaeon−1 were obtained from the Ribosomal RNA Database (Lee et al., 2009) (accessed 18 August 2009) and used in these calculations.

Bacterial 16S rRNA amplification, cloning, and sequencing

The 16S rRNA gene was PCR-amplified from community DNA using primers 27F and 1492R (Lane, 1991). Each 25 μl reaction contained 100 ng template DNA, FailSafe buffer E (Epicentre Biotechnologies), 1.25 U Taq polymerase, and forward and reverse primers at a final concentration of 0.1 μM each. Thermocycling was conducted in a 2720 Thermal Cycler (Applied Biosystems, Foster City, CA, USA) under the following conditions: initial denaturation at 95 °C for 1 min; 30 cycles of denaturation at 94 °C for 1 min, annealing at 55 °C for 1 min, and extension at 72 °C for 1 min, 30 s; and a final extension at 72 °C for 7 min. Five replicate amplifications were performed for each sample, and their products were combined for downstream use.

PCR products were confirmed for as described above. Product bands were excised, purified using a Wizard SV Gel and PCR Clean Up kit (Promega), and cloned into a pGEM-T Easy vector (Promega). Approximately 300 positive colonies were selected from each sample, and glycerol stocks of each were submitted to Agencourt Bioscience Corporation (Beverly, MA, USA) for single-pass Sanger sequencing with primer 515F (Lane, 1991).

Bacterial tag-encoded amplicon pyrosequencing

Purified community DNA samples were submitted to the Research and Testing Laboratory (Lubbock, TX, USA) for tag-pyrosequencing. Samples were amplified with modified versions of primers 530F and primer 1100R, and the amplicons were sequenced using Roche 454 Titanium chemistry, producing reads from the forward direction from 530F (Acosta-Martínez et al., 2008). Sequences were quality trimmed according to Acosta-Martínez et al. (2008).

Sequence analysis and community comparisons

Sanger sequence data were quality checked, aligned, trimmed to a similar length, and assigned putative identities using the Ribosomal Database Project (RDP) pipeline (Cole et al., 2007) (accessed 26 January 2009). The RDP pyrosequencing pipeline and Classifier function were used to align and assign identities to the 454 sequence data (accessed 26 January 2009). Sequences from the Sanger clone libraries were screened for chimeras using Bellerophon (Huber et al., 2004), and potential chimeras were excluded from further analysis. Those 454 sequences that did not align with the 16S rRNA region were considered to be sequencing errors and were excluded from further analysis.

Distance matrices were constructed for both the Sanger and 454 sequence libraries using the dist.seqs function in MOTHUR, version 1.4.1 (Schloss et al., 2009). MOTHUR was also used to assign sequences to operational taxonomic units (OTUs, 97% similarity) and calculate both Shannon's diversity index values (H’) and ChaoI richness estimates. Phylogenetic trees were constructed using the relaxed neighbor-joining algorithm in Clearcut (version 1.0.9) (Sheneman et al., 2006), and between-site comparisons of phylogenetic structure were conducted using the parsimony test in Treeclimber (Schloss and Handelsman, 2006). Parsimony test scores with P-values <0.05 were considered to represent significant differences, and pairwise comparisons of individual libraries were conducted only if the study-wide (that is global) test was found to be statistically significant (Schloss and Handelsman, 2006). Bonferonni correction for multiple comparisons was used among the pairwise comparisons, adjusting our significance level to P0.0017 (Neter, 1996).

Nonmetric multidimensional scaling (NMDS) of sites based on OTU composition, and Mantel tests of community correlation with environmental variables, were conducted using PC ORD, version 5.0 (MjM Software, Gleneden Beach, OR, USA) in default mode. NMDS and the Mantel tests were carried out on both the Sanger and 454 sequence libraries. OTUs occurring as singletons were removed from the data set before analysis to down-weight the effects of rare ‘species.’ Sørenson distance was used for both the NDMS ordination and the ‘species’ distance metric in our Mantel tests (McCune and Grace, 2002); Euclidian distance was used to describe the underlying structure of our environmental variables (Fierer and Jackson, 2006).

Sanger sequence reads were submitted to NCBI GenBank under accession numbers GQ893028 through GQ894720, and the 454 sequence reads were submitted to the NCBI Short Read Archive under accession number SRA009427.2.

Results

Soils and sediments chemical characteristics

The chemical composition of the transect sediments is summarized in Table 1. The sites exhibited a narrow range of pH values, and all contained sodium as their dominant cation. The sites varied widely with respect to water content and nutrient and cation concentrations. Strong positive correlations were observed with respect to: (1) water content, sodium or potassium concentration, and electrical conductivity (r2=0.71–0.77); (2) sodium, magnesium, and potassium concentration (r2=0.79–0.93); (3) sodium concentration and electrical conductivity (r2=0.98); and (4) phosphorus concentration with both organic carbon content (r2=0.77) and pH (r2=0.86). Sediment organic carbon and total nitrogen concentrations, however, were not well correlated with one another (r2=0.41).

Table 1 Soil and sediment physical and chemical properties

Community qPCR

On the basis of the relative proportion of gene copies detected in the community qPCR assays, Bacteria were numerically dominant among the La Sal del Rey sediments, accounting for 97% of the rRNA gene copies detected at each site. Adjusting these values to reflect the average number of rRNA copies that may be found in each organism type, Bacteria accounted for 92–99% of community membership. Although Archaea represented only a small proportion of the microbial community within each site, their relative abundance tended to increase along the transect. They were nearly absent at the terrestrial end of the transect (sites T3-0 and T3-65) but increased to 8% of the community at the aquatic end of the transect. Archaeal abundance correlated well with sample water content (r2=0.80) and to a lesser degree with electrical conductivity (r2=0.62) and sodium content (r2=0.53).

Community composition, diversity, and estimated richness

A total of 39 590 16S rRNA sequences were generated through 454 sequencing, with an average read length of 455±85 bp (mean±s.d.). The 454 sequence libraries ranged in size from 1403 sequences at site T3-0 to 6745 sequences at site T3-325 and contained between 744 OTUs at site T3-0 and 3 478 OTUs at site T3-65 (Table 2). An additional 1693 16S rRNA sequences, averaging 425±5 bp in length (mean±s.d.) after trimming, were generated through cloning and single-pass Sanger sequencing. The Sanger clone libraries ranged in size from 185 sequences at sites T3-0 and T3-130 to 230 sequences at site T3-390 and contained between 68 OTUs at site T3-390 and 143 OTUs at site T3-195 (Table 2). Overall, a total of 16 596 unique OTUs were identified among the 454 sequence libraries, and 658 unique OTUs were identified among the Sanger sequence libraries. No single OTU accounted for >0.32% of the entire data set, but to convey a better sense of the composition of each community, the 10 most abundant OTUs from each sampling point, along with their putative identities, have been provided in Supplementary Table 1.

Table 2 Summary of sequence library sizes, operational taxonomic units (OTUs), and diversity and richness estimates

The 454 libraries detected a greater variety of Archaea and low-abundance bacterial taxa than did the Sanger sequence libraries (Table 3); however, the most abundant taxa were detected with similar frequencies across both sequencing projects (Figure 1). Members of at least 24 bacterial and 2 archaeal phyla were detected along the study transect. Proteobacteria and Bacteroidetes were encountered most frequently, representing 24–45% and 4–26% of each library's membership, respectively (Figure 1). Unclassified Bacteria (that is those that could not be identified at the phylum level with at least 80% confidence in RDP) accounted for an additional 25–47% of each sequence library. Among the Archaea, members of the phylum Euryarchaeota (particularly members of the order Holobacteriales) were encountered with the greatest frequency and accounted for the majority of archaeal sequences identified here.

Table 3 Detection and distribution of low abundance phyla
Figure 1
figure 1

Bacterial community composition of soils and sediments located along a transect through a hypersaline lake ecosystem that included terrestrial sites (T3-0), exposed lakebed sediments (T3-65 to T3-260) and water-logged, ‘aquatic’ sites (T3-325 to T3-455). The 454 sequencing (panel a) and cloning and Sanger sequencing (panel b) were used to characterize the sediment communities. Both libraries were sequenced in the forward direction from positions 515–530 of the 16S rRNA gene.

Shannon diversity index values (H’) suggest that diversity varied along the study transect. With the exception of the relatively small T3-0 454 sequence library, H’ values tended to decrease as conditions became increasingly water-logged and salt-rich (Table 2), indicating that community richness and/or evenness declined along the transect. Chao I richness estimates suggest that approximately half of the estimated diversity contained within these communities was captured by our sequencing efforts (Table 2). They also suggest that further sequencing would be likely to yield additional unique OTUs; however, greater yields would be expected from the near-shore and intermediate-distance communities (for example sites T3-65, T3-130, and T3-195) than from those located farther away from shore (for example sites T3-260 to T3-455).

Analysis of communities via the Parsimony test

Parsimony tests conducted using the 1693 Sanger sequences found no significant differences among the transect communities with respect to their phylogenetic structure (P=0.132). In contrast, when the parsimony test was conducted using the 39 950 sequences generated through 454 sequencing, all of the transect communities were significantly different from one another. The global test was statistically significant (P<0.001), as were all of the pairwise comparisons of the individual transect locations (P0.001). Additional comparisons of sequence library similarity (that is overlap as measured by the Yue-Clayton index) can be found in Supplementary Table 2.

Microbial community correlations with physicochemical characteristics and NMDS ordination of communities based on OTUs

Mantel tests conducted with both the Sanger and 454 sequences found that La Sal del Rey's microbial communities correlated significantly with sample water, phosphorus, and total organic carbon content, as well as sample pH (Table 4). A significant correlation between sample calcium concentration and the Sanger communities was also detected.

Table 4 Mantel correlations highlight the relationships shared between environmental traits and community composition

The relationships shared between the transect microbial communities and the environmental variables that correlated well with them can be seen in the clustering of the transect sites in NMDS plots shown in Figure 2. Site water content, the best correlated variable measured here, can be seen in the segregation of the driest, most ‘terrestrial’ site T3-0, by itself, as well as the clustering of the most water-logged, ‘aquatic’ sites (T3-325, T3-390, and T3-455) with one another. The remaining four transect points formed a less cohesive ‘intermediate’ cluster. The influence of the other well-correlated variables can also be observed. For example, organic carbon concentration, phosphorus content, and pH appear to contribute to the segregation of site T3-0 from the other transect communities. Organic carbon concentration also appears to underlie the clustering of the transect into its three community types (Table 1), contributing to site T3-325's tendency to cluster with the ‘aquatic’ sites rather than the ‘intermediate’ sites despite its water content being more ‘intermediate’ in nature.

Figure 2
figure 2

NMDS of transect points based on OTU composition determined by (a) 454 sequencing and (b) cloning and Sanger sequencing. Clustering of the transect communities into three groups, roughly corresponding to sample water content can be observed. Samples T3-390, T3-455, and T3-325 represent the most aquatic sites along the study transect, and site T3-0 represents the driest, most terrestrial site. Those clustered in the middle of the figure represent a set of conditions that were intermediate relative to the aquatic and terrestrial clusters. OTUs were defined as sequences sharing 97% similarity with one another.

The distribution of OTUs across La Sal del Rey fell into one of three patterns: (1) occurring at most or all of the transect sites, but varying in abundance relative to site conditions; (2) occurring with cluster-specific distributions (that is terrestrial, intermediate, or aquatic); or (3) occurring with site-specific distributions (that is occurring exclusively in one of the eight transect sites). OTUs representing Acidobacteria, Actinobacteria, Rhodobacteraceae, and Desulfohalobiaceae were found at most or all of the sites along the study transect but varied in the abundance relative to site water content. Acidobacteria and Actinobacteria declined in abundance as conditions became more water-logged and salt-rich (Figure 1), whereas the abundances of the Rhodobacteraceae and Desulfohalobiaceae increased under the same conditions. The terrestrial-specific cluster of OTUs was composed of Bacteroidetes (particularly members of the genus Balneola), members of the GP10 division of the Acidobacteria, and Myxococcales-like organisms (Supplementary Table 1). Those OTUs occurring exclusively within the intermediate cluster (T3-65, T3-130, T3-195, and T3-260) included Cyanobacteria, other Bacteroidetes, Chromatiales, and a number of unclassified Gammaproteobacteria (Supplementary Table 1). The aquatic cluster (sites T3-325, T3-390, and T3-455) was characterized by OTUs representing Rubrimonas- and Roseicyclus-like bacteria, Desulfohalobiaceae, several unclassified Bacteroidetes-like organisms, and a variety of unclassified Bacteria (Supplementary Table 1). OTUs having site-specific distributions represented nearly 80% of the OTUs identified here, and the number (that is richness) of site-specific OTUs occurring within a given location was found to be negatively correlated with site water content (r2=0.69, data not shown).

Discussion

In any discussion of PCR-based community characterizations, it should be acknowledged that these approaches are not without potential biases, including differential DNA extraction efficiencies, the potential for species to carry multiple copies of the 16S rRNA gene, and biases in primer selectivity (von Wintzingerode et al., 1997; Lee et al., 2009). Furthermore, it is recognized that tag-pyrosequencing is also prone to error, including difficulty interpreting homopolymeric regions of sequence and producing amplicons sequences of anomalous length, both of which have the potential to over-inflate diversity estimates (Kunin et al., 2009). However, sequence library descriptions of community composition are often corroborated by other approaches to community characterization (Papineau et al., 2005 and references therein), and recent evidence suggests that thorough controls on data quality and clustering can help to limit the effects of pyrosequencing errors on diversity estimates (Kunin et al., 2009).

Although no technique is perfectly bias-free, it is clear that advances in sequencing technology are expanding our view of the microbial world in unprecedented ways. New methods are increasing dramatically the number of sequences that may feasibly be included in a single study, allowing deeper coverage and providing opportunities for new insights regarding microbial communities and their interactions with the environment (Sogin et al., 2006; Turnbaugh et al., 2006; Roesch et al., 2007; Acosta-Martínez et al., 2008). Although culture-independent studies of microbial communities have been carried out in hypersaline soils and sediments in the past, this study represents the first to have done so at La Sal del Rey, and it is the first among such studies to have used community qPCR and the combination of Sanger and tag-encoded pyrosequencing technologies.

The combination of techniques used in this study provided a detailed and powerful snapshot of the microbial communities occurring at La Sal del Rey. We found that the system's soil and sediment microbial communities harbored a wide variety of taxa and contained a large proportion of individuals that had no close match within the existing molecular databases. Despite our 454 sequence libraries being an order of magnitude larger than their corresponding Sanger libraries, and having been constructed using different primer sets, similar descriptions of community composition and similar correlations with environmental variables (Table 4) were obtained with each of the sequencing approaches. Major taxa were detected with similar relative abundances (Figure 1), corroborating the findings of other studies that have performed side-by-side comparisons of tag-encoded pyrosequencing versus Sanger clone libraries (Turnbaugh et al., 2006; Jones et al., 2009). Likewise, major taxa were detected with abundance levels comparable to those that have been described in other hypersaline ecosystems. The ratio of Bacteria to Archaea at La Sal del Rey resembles those reported by others who have attempted to quantify their abundance in other hypersaline sediments (Mouné et al., 2003; Ley et al., 2006). Bacteria appear to be numerically dominant in these systems, and among them three phyla—the Proteobacteria, Bacteroidetes, and Firmicutes—typically account for the majority of taxa found within each community (Mouné et al., 2003; Dong et al., 2006; Mesbah et al., 2007).

Although the agreement among our 454 sequence libraries, Sanger sequence libraries, and those described by others in the literature was quite good relative to the abundances of the major taxa, we found that our 454 libraries included taxa that had not been captured by our Sanger sequencing efforts or those described in other studies (Humayoun et al., 2003; Demergasso et al., 2004; Dong et al., 2006; Jiang et al., 2006; Ley et al., 2006; Mesbah et al., 2007). These additional taxa generally occurred with low abundance and included members of the Deinococcus-Thermus, Thermotogae, and candidate division TM7, as well as a number of OTUs that could be identified, at best, as unclassified Bacteria. Akin to the detection of the ‘rare biosphere’ described by Sogin et al. (2006), the enhanced sampling depth achieved with tag-encoded pyrosequencing improved our detection of low abundance taxa (Table 3) and allowed us to profile our transect microbial communities with greater detail.

In addition to offering a broader view of the composition of La Sal del Rey's microbial communities, the increased sampling resolution achieved with the 454 libraries also improved our statistical power and enhanced our ability to detect differences among the transect communities. As communities increase in their complexity, it is thought that the minimum number of sequences (that is sampling effort) needed to distinguish differences among them should increase as well (Singleton et al., 2001; Schloss and Handelsman, 2006). Parsimony tests conducted with our modestly sized Sanger libraries (average size=211 sequences) did not find significant differences among the transect communities; however, when the parsimony test was repeated using the larger 454 libraries (average size=4948 sequences), highly significant differences were detected among all transect sites.

Differences in microbial community diversity and phylogeny may result from a number of different factors, key among these being those characteristics, which define a community's physical and chemical environment. In sediment ecosystems, physical stratification and chemical gradients help to create and maintain high levels of diversity within and between bacterial communities (Torsvik et al., 2002; Lozupone and Knight, 2007). Sodium concentration and electrical conductivity displayed some of the strongest gradients along our study transect, and given the importance of salinity in shaping microbial community structure at global scales, we expected that these factors would have an influential function at La Sal del Rey, as well. Although small portions of the communities at La Sal del Rey (for example members of the archaeal class Holobacteriales) did display positive relationships with factors related to salinity, we found that overall community structure was better correlated with other environmental factors. These included sample water content, organic carbon concentrations, phosphorus content, and pH.

Perhaps the most striking among these factors, with respect to its influence on community structure, was sample water content. Despite the strong, positive correlations shared between sediment water content and factors related to salinity (for example sodium concentration and electrical conductivity), it appears that water's relationship with microbial community structure was primarily a function of its influence on sediment oxygen concentrations. Although we did not measure sediment oxygen concentrations directly, water content has been used as an indicator of sediment oxygen content in other systems (Dong et al., 2006; Jiang et al., 2006; Lozupone and Knight, 2007), and field observations of sediment color and odor suggest this to be the case at La Sal del Rey as well. Taken in whole, the Mantel correlations described above suggest that the microbial communities occurring along our study transect are shaped by their requirements for oxygen, the carbon substrates they metabolize, and their pH tolerances, rather than their tolerance or affinity for salt. Although salinity may have a major function in determining community composition at global scales (Lozupone and Knight, 2007), the results of this study suggest that its influence is less important at local scales, particularly in systems that are already salt-rich.

One of the central goals of microbial ecology is to understand the relationships that microbial communities share with their environment, and an extension of this goal is to understand how those relationships play out at local, regional, and global scales. Although this study provides a detailed snapshot and expands our understanding of microbial community variation at the local scale, it is clear that we have only just scratched the surface. We still have much to learn about these communities and the relationships that they share with one another, as well as the factors that shape these communities and others at larger scales. As sequencing technologies continue to advance, it is anticipated that progress toward these goals will improve steadily and provide greater insight into the diversity, complexity, and dynamics of microbial systems.