Bacterial diversity and ecosystem function of filamentous microbial mats from aphotic (cave) sulfidic springs dominated by chemolithoautotrophic “Epsilonproteobacteria”

Abstract

Filamentous microbial mats from three aphotic sulfidic springs in Lower Kane Cave, Wyoming, were assessed with regard to bacterial diversity, community structure, and ecosystem function using a 16S rDNA-based phylogenetic approach combined with elemental content and stable carbon isotope ratio analyses. The most prevalent mat morphotype consisted of white filament bundles, with low C:N ratios (3.5–5.4) and high sulfur content (16.1–51.2%). White filament bundles and two other mat morphotypes had organic carbon isotope values (mean δ¹³C =−34.7‰, 1σ= 3.6) consistent with chemolithoautotrophic carbon fixation from a dissolved inorganic carbon reservoir (cave water, mean δ¹³C =−7.4‰ for two springs, n = 8). Bacterial diversity was low overall in the clone libraries, and the most abundant taxonomic group was affiliated with the “Epsilonproteobacteria” (68%), with other bacterial sequences affiliated with Gammaproteobacteria (12.2%), Betaproteobacteria (11.7%), Deltaproteobacteria (0.8%), and the Acidobacterium (5.6%) and Bacteriodetes/Chlorobi (1.7%) divisions. Six distinct epsilonproteobacterial taxonomic groups were identified from the microbial mats. Epsilonproteobacterial and bacterial group abundances and community structure shifted from the spring orifices downstream, corresponding to changes in dissolved sulfide and oxygen concentrations and metabolic requirements of certain bacterial groups. Most of the clone sequences for epsilonproteobacterial groups were retrieved from areas with high sulfide and low oxygen concentrations, whereas Thiothrix spp. and Thiobacillus spp. had higher retrieved clone abundances where conditions of low sulfide and high oxygen concentrations were measured. Genetic and metabolic diversity among the “Epsilonproteobacteria” maximizes overall cave ecosystem function, and these organisms play a significant role in providing chemolithoautotrophic energy to the otherwise nutrient-poor cave habitat. Our results demonstrate that sulfur cycling supports subsurface ecosystems through chemolithoautotrophy and expand the evolutionary and ecological views of “Epsilonproteobacteria” in terrestrial habitats.

Introduction

Microbial processes occurring in the absence of light have generally been considered insufficient to support ecosystem-level processes, and until recently the dogma has been that life processes in the subsurface are dominated by heterotrophic consumption of surface-derived carbon [1–4]. But the absence of light does not preclude life, as reactive mineral surfaces and solute-rich groundwater provide energy sources sufficient for chemolithoautotrophic growth in the subsurface [5,6]. Chemolithoautotrophy is now recognized as an important ecosystem-level process in aphotic terrestrial environments, including deep aquifers [2,6] and caves [7–9]. Because subsurface habitats are relatively difficult to access, however, less is known about the biodiversity and community structure, or ecosystem functioning and carbon cycling of terrestrial chemolithoautotrophically-based microbial ecosystems.

Caves represent distinctive habitats with complete darkness, relatively constant air and water temperatures, and a poor supply of easily degradable organic matter. Consequently, most cave ecosystems depend on allochthonous organic material for energy [10,11]. Previous investigations describing microorganisms from caves and karst settings [1,12,13], including from both moist sediments and aquatic habitats, have suggested that most cave microbes originate from surface environments and are only active under optimal growth conditions[14]. However, groundwater bearing dissolved hydrogen sulfide and negligible allochthonous carbon discharges as springs into the passages of some caves [7,15–17]; hydrogen sulfide is an energy-yielding substrate for some microorganisms, and areas of these sulfidic cave springs are colonized by thick subaqueous microbial mats.

As photosynthesis is not possible in a cave, aphotic chemolithoautotrophic primary productivity can be directly investigated. Stable carbon isotope measurements and ¹⁴C-radiolabeled substrate experiments from bulk microbial mats in several sulfidic caves suggest that chemolithoautotrophy is the base of these ecosystems [7,18,19]. Molecular phylogenetic studies based on 16S rRNA gene sequences have expanded our understanding of the microbial diversity in caves[14], including those with sulfidic groundwater [8,16,20,21] and those without [9,22–25]. In sulfidic caves, the dominant bacterial groups from some subaqueous microbial mat communities belong to the “Epsilonproteobacteria”[16,21,26], while culture-based methods identified chemolithoautotrophic sulfur-oxidizing bacterial groups, including the genera Thiothrix and Thiobacillus [20,21,27]. There has been little done, however, to examine the ecology of dominant microbial groups involved with energy and nutrient transfers in cave settings, or of the physical or chemical controls that govern community structure and dynamics.

This study is part of an ongoing project to describe microbial ecosystems and nutrient cycling in sulfidic cave habitats, as proxies for deeper subsurface environments such as carbonate aquifers. We have been studying microbial mats associated with sulfidic springs in Lower Kane Cave, a small system located in the Bighorn Basin, Wyoming. The objectives of this investigation are to describe the genetic and functional diversity of microbial groups, as well as to define how community structure is controlled by habitat geochemistry. We hypothesized that community composition and structure would reflect substrate availability, and specifically that community composition would shift with changes in dissolved oxygen and hydrogen sulfide concentrations. As it is often difficult to ascertain the metabolism of certain organisms based on 16S rDNA-based phylogenies[28], elemental composition (carbon to nitrogen ratios and sulfur content) and stable carbon isotope ratio analysis of specific mat morphotypes were combined with 16S rDNA sequence phylogenies to link hypotheses of ecosystem functionality with genetic identity for the as yet uncultured microorganisms [3,29,30]. The current study complements previous investigations in which we quantified dominant epsilonproteobacterial populations in filamentous microbial mat morphotypes from Lower Kane Cave based on preliminary clone library construction and the development of two 16S rRNA-specific “Epsilonproteobacteria” fluorescence in situ hybridization (FISH) probes[26]. Our characterization of this cave ecosystem expands the ecological understanding of “Epsilonproteobacteria” and demonstrates that sulfur cycling supports this subsurface ecosystem through chemolithoautotrophy.

Materials and methods

The study site and sample collection

Lower Kane Cave (LKC) is located in the north-central portion of the Bighorn Basin and is forming within the Madison Limestone (Mississippian age); the basic hydrogeological setting is described in Egemeier[15]. There are four hydrogen sulfide-bearing springs that discharge into the cave along a fracture trace (Fig. 1(a)). The cave is actively undergoing sulfuric acid speleogenesis, a biogeochemical process by which hydrogen sulfide oxidizes to sulfuric acid in the subaqueous environment by microorganisms or subaerially on cave-wall surfaces due to sulfide gas volatilization[31]. The acid reacts with and replaces the limestone hostrock with gypsum, which is readily dissolved by groundwater undersaturated with respect to gypsum [15,31]. The net results are the removal of mass and an increase in void volume. At each of the three largest springs, dissolution of the host carbonate rock has resulted in fracture enlargement and each spring orifice area has a pool and outflow stream channel. Sparse filaments are found in all the spring orifice pools, and thick carpets of filamentous microbial mats occur along the outflow streams discharging from each spring. The Fissure and Upper Spring mats extend for ∼20m, while the Lower Spring mats are 1 m in length.

Samples of each microbial mat morphotype were aseptically collected from three spring sites and aliquots were used for bulk biomass, elemental analysis, carbon isotope analysis, and DNA extraction and clone library construction. To preserve the integrity of this sensitive ecological system, conservative quantities of microbiological materials were collected. Microbial mat morphotypes were collected separately and distinguished as white filament bundles (denoted as ‘f'), white webs (denoted as ‘w'), yellow-white patches (denoted as ‘y'), and gray filaments (denoted as ‘g'). White filament bundles in the water column or filaments from the surface of the mats were targeted for clone library construction; however, one small library was constructed with gray filaments ∼2 cm below the top of the mat for comparison. Sampling sites were numbered according to their location in meters from the back of the cave, with flow always toward the cave entrance (i.e. longer distances) (Fig. 1(a)): Fissure Spring (124-, and 127-m), Upper Spring (190-, 195-, 198-, and 203-m), and Lower Spring (one orifice and one mat sample from 248-m).

Geochemical analysis

Geochemical data were acquired at the major microbiological sample locations, as well as throughout the cave, over three years of ongoing research. Unstable parameters (pH, E_H, and dissolved oxygen) were measured using electrode methods[32]. Dissolved hydrogen sulfide, ferrous iron (Fe²⁺), and trace level dissolved oxygen were measured in the field using the Methylene Blue, Ferrozine, and Rhodazine D colorimetric methods, respectively, using CHEMetrics® chemistries (Calverton, VA) with a MiniSpec 20 field spectrophotometer[32]. Vertical profiles of dissolved oxygen through the mats were determined by fluorescence-quenching optical methods (Ocean Optics, Inc., Dunedin, FL). Unstable and reactive parameters (pH, oxygen, hydrogen sulfide, etc.) were also measured at several transects along and across the cave stream channels. Alkalinity (as total titratable bases, here dominated by bicarbonate) was determined in the field by titration to pH 4.5, and verified in the laboratory by end-point seeking autotitration[32]. Anions and acid-preserved metals were determined by ion chromatography (Environmental Protection Agency (EPA) method 9056; Manual SW-846, Test Methods for Evaluating Solid Waste, Physical/Chemical Methods; http://www.epa.gov/epaoswer/hazwaste/test/main.htm) and inductively coupled plasma mass spectrometry (EPA method 6020; Manual SW-846), respectively. Dissolved organic and inorganic carbon (DOC and DIC, respectively) were determined by Dorhman DC-180 wet-oxidation carbon analyzer (EPA method 9060; Manual SW-846). Dissolved gas species (e.g. methane, aromatic hydrocarbons, hydrogen sulfide, organosulfur gases) from the spring and stream water were analyzed by headspace gas chromatography (EPA method 5021; Manual SW-846).

Mat carbon, nitrogen, and sulfur content

Each mat sample was individually homogenized, acidified with dilute HCl, rinsed with dH₂O, repeated at least twice to ensure dissolution of carbonate mineral phases, and freeze-dried. Total organic carbon and nitrogen contents were determined by elemental analyzer interfaced with a mass spectrometer, simultaneously with carbon isotope ratio analysis (see below). Total sulfur content, as inorganic and organic sulfur compounds, was determined on a EuroEA3000 elemental analyzer (EuroVector, Milan, Italy).

Minimum bulk mat biomass was determined from dry weight analysis of the mats followed by comparison of the percent carbon in each 1 ml aliquot, using methods described in and modified from Bratbak and Dundas[33]. Briefly, replicate samples were individually homogenized, acidified with dilute HCl, weighed, freeze-dried, and re-weighed to obtain the dry weight. The percentage of carbon in each dried aliquot was determined by elemental analyzer. Cell carbon content was estimated from the standard conversion factor of 350 fg C cell⁻¹ (assuming an average cell size of 1 μm³;[33]) to determine the approximate number of cells per ml.

Carbon isotope methods

For carbon isotope ratio (¹³C/¹²C) analysis, organic carbon of 1–2 ml mat was prepared by acidifying the sample in dilute HCl to ensure removal of carbonate mineral phases. Most measurements were made by elemental analyzer interfaced with a continuous flow FinniganMAT Delta Plus mass spectrometer, but some measurements were also made by sealed tube combustion, vacuum purification, and dual-inlet VG Prism II mass spectrometer. Microbial mat carbon isotope values were compared to the values obtained from dissolved inorganic carbon (DIC), a composite of CO_2(aq), HCO⁻₃ from the cave water. DIC was extracted for ¹³C analysis by acidifying under vacuum with 100% phosphoric acid followed by cryogenic purification of the resulting CO₂, using the method modified from Hassan[34]. At the pH and temperature of the cave water (pH ∼ 7.3 at 21.5 °C), the dominant DIC species was HCO⁻₃ (∼90%) based on dissociation constants for H₂CO₃, HCO⁻₃, and CO²⁻₃ species. Carbon isotope values for the limestone were also measured by reaction with 100% phosphoric acid at 90 °C[35]. Carbon isotope values are expressed in delta (δ) notation with respect to the international standard V-PDB.

DNA extraction and PCR amplification of 16S rRNA gene sequences

Approximately 0.2–0.5 ml mat material were aseptically collected and placed into tubes containing sterilized DNA extraction buffer, identical to methods described in Engel et al.[26]. DNA purity and concentration for each extraction were determined on a GeneQuantII spectrophotometer (Amersdam Biosciences, Piscataway, NJ). Nearly full-length 16S rRNA gene sequences were PCR-amplified using the eubacterial primer pair 27f (forward, 5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492r (reverse, 5′-GGTTACCTTGTTACGACTT-3′)[36]. Amplification was performed with a Perkin Elmer 9700 thermal cycler and AmpliTaq Gold (Applied Biosystems, Branchburg, New Jersey), under the following conditions repeated for 35 cycles: denaturation at 94 °C for 1 min, primer annealing at 42 °C for 1 min, chain extension at 72 °C for 1.5 min.

16S rDNA clone library construction, clone sequencing, and phylogenetic analysis

Amplified PCR products were purified with the GeneClean II Kit (Bio101, Inc., Vista, CA), following manufacturer recommendations. Purified PCR products were cloned using the TOPO TA Cloning kit with Escherichia coli TOP10F' cells, according to manufacturer instructions (Invitrogen, Carlsbad, CA), and eleven bacterial clone libraries were constructed from different mat morphotypes. Plasmids containing 16S rDNA inserts were extracted using a standard alkaline lysis miniprep method[37]. Clone plasmids were digested simultaneously using Eco RI and Rsa I (1U each) according to manufacturer instructions (New England Biolabs) for restriction fragment length polymorphism (RFLP) analysis. RFLP patterns were visualized on 2% agarose gels stained with ethidium bromide and run in TBE (Tris–borate–EDTA)-buffer. Clones representing unique patterns from each library were selected for sequencing, and inserts from the plasmid minipreps for each clone to be analyzed were sequenced as described in Engel et al.[26].

DNA sequences were submitted to the CHECK-CHIMERA program of the Ribosomal Data Base Project (RDP) II (http://rdp.cme.msu.edu/html/)[38] to screen for and to eliminate chimeric sequences. Clone sequences were subjected to BLAST searches within the GenBank database (http://www.ncbi.nlm.nih.gov/) to determine 16S rDNA sequence similarities to culturable and not yet cultured organisms.

Nucleotide sequences were initially aligned using Clustal X[39] and then manually adjusted based on conserved primary and secondary structures. Nucleotide segments were removed that could not be unambiguously aligned, corresponding to E. coli 16s rRNA secondary structure helices 9 and 10 (bp 181–226; all alignments) (all base pair positions correspond to E. coli numbering;[40]), helix 17 (bp 452–481; all but the Betaproteobacteria alignment), helices 25 and 26 (bp 822–860; Gammaproteobacteria and Bacteroidetes/Chlorobi-Acidobacterium alignments), helix 30 (bp 1028–1032; Betaproteobacteria and Deltaproteobacteria alignments), and helix 33 (bp 995–1045; “Epsilonproteobacteria” and Bacteroidetes/Chlorobi-Acidobacterium alignments).

Phylogenetic analyses were done using minimum evolution criteria in PAUP*[41], maximum likelihood criteria using a genetic algorithm (MLga) in MetaPIGA[42], and Bayesian inference coupled with Markov chain Monte Carlo techniques (BMCMC) in MrBayes version 3.0b4[43]. For minimum evolution and BMCMC searches, a model of evolution was chosen based on likelihood ratio tests[44], as implemented in Modeltest 3.06[45]. For the MLga search, Metapiga has model choice constraints; therefore, the model was set at the most complex model allowable by the program. Minimum evolution heuristic searches were run using random addition for 500 replicates. BMCMC searches were run for 4 × 10⁶ cycles sampling every 20,000 generations at least twice to check for convergence and then combined, burning in five trees from each chain. MLga searches were run for one replicate using 16 populations of 10 individuals each. As an indication of nodal support, bootstrap analyses were performed for minimum evolution using full heuristic searches and posterior probabilities were calculated for BMCMC and MLga analyses [42,43]. Sequence similarity was calculated for the closely related clone sequences from the “Epsilonproteobacteria” using corrected distances based on the model selected by Modeltest 3.06[45].

Statistical analysis and sequence population diversity

To determine if the number of clones in each of the clone libraries was representative of the microbial diversity, rarefaction curves were produced using the approximation algorithm aRarefactWin (Analytic Rarefaction, version 1.3, S. Holland, http://www.uga.edu/strata/software/). Curves having 95% confidence levels were constructed by comparing the number of clones in each 16S rRNA gene library to the number of phylotypes from a particular library.

Clone library species richness and species dominance/evenness indices (combined to represent heterogeneity; e.g., [46,47]) were calculated based on the number of phylotypes identified from RFLP and taxonomic affiliations from BLAST searches [48–50]. The nonparametric methods Abundance-based Coverage Estimator (ACE) and Chao1, and the Shannon–Wiener biodiversity function expressed as the Shannon index (H′) were computed for each library using EstimateS (version 6.0b1, R.K. Colwell, http://viceroy.eeb.uconn.edu/estimates). The Shannon Evenness index (E) and the Simpson's Dominance index (D) were also calculated based on equations presented in Hill et al.[50].

Nucleotide sequence accession numbers

Nucleotide sequence data reported in this study are available in the GenBank database under the accession numbers AY208806 to AY208817 for LKC2-labeled clones, and AY510166 to AY510267 for LKC3-labeled clones.

Results

Geochemistry and morphologic description of microbial mats

Major ion geochemistry did not vary significantly from sample period to sample period, and waters were all dominated by Ca²⁺, HCO⁻₃, and SO²⁻₄ ions (calcium–bicarbonate–sulfate water type) (Table 1). Although the cave is forming from sulfuric acid dissolution of limestone [15,31], the spring and stream waters are buffered to circum-neutral pH by ongoing carbonate dissolution. Incoming spring water had dissolved sulfide concentrations >35 μmol l⁻¹ and non-detectable dissolved oxygen (Table 1). The concentration of dissolved sulfide and oxygen changed downstream at all the springs, such that at the end of the microbial mats sulfide decreased to non-detectable and the concentration of dissolved oxygen exceeded 40 μmol ⁻¹. The concentration of DOC in all the incoming spring waters was low at <80 μmol l⁻¹ including methane.

We observed four mat morphotypes along the spring-stream flowpaths (Fig. 2). All three spring orifice pools had gray benthic sediment and long white filament bundles were suspended in the water column. The Lower Spring (248 m) had the densest concentration of filament bundles in the orifice (Fig. 2(a)), and the microbial mat below the Lower Spring orifice was 2–5 cm thick but less than 1 m in length; overall the mat was yellowish-white in appearance (Fig. 2(b)). The Upper Spring had the longest filament bundles, at more than a meter in length in the orifice pool (Fig. 2(c) and (d)). White filament bundles coalesced on the edges of the outflow channel downstream, where the dissolved sulfide concentration decreased and dissolved oxygen concentration increased (Fig. 1(b)). Very thin, short (1 cm in length) whitish-gray filaments covered stream sediments in flowing water at the bottom of the Upper Spring outflow channel (195 m). Approximately 6 m downstream from the Upper Spring orifice, the gray filaments thickened by several centimeters and were covered by thin white webs and long white filament bundles. Some of the webs at 203 m had a bumpy or knobby texture (Fig. 2(e)). Oxygen microelectrode profiles at 203 m showed oxygen tension abruptly decreased ∼3 mm below the mat–water interface and anaerobic conditions P_o2 < Pa persisted within the 5 cm-thick mat interior, demonstrating that the mats are geochemically stratified. Although the focus of this study was on the white microbial mat morphotypes, gray filaments within the mat at 203 m (∼2–5 cm below the white mat surface) were sampled to determine if there were general differences in community structure vertically. For both the Upper and Lower Spring channel mats, dense white mats, with small (1–2 cm diameter) discontinuous yellow patches and feathery (i.e. short, thick, and branching filaments) bundles, dominated the lower reach of the outflow channels (Fig. 2(f)). Filament bundles near the orifice of the Fissure Spring (118 m) (Fig. 2(g)) were also associated with web-like structures and gas bubbles entrained within the mats (125 m) (Fig. 2(h)).

Biomass estimates, C:N ratios, and sulfur content

The biomass of the microbial mat samples was ∼10¹⁰ cells ml⁻¹ (Table 2), with gray filaments from the Lower and Upper Springs having the highest biomass. Biomass values reported here may underestimate the actual biomass because current cell conversion factors are for rod-shaped cells[33], and previous FISH analyses reveal that the mats are dominated by filamentous morphotypes[26].

The N content varied by mat morphotype, and white filament bundles and white webs had the highest N content compared to gray filaments or gray sediment (Table 2). Generally, the lower the C:N ratio, the higher the quality of the mat as a food source for the ecosystem [51,52]. The mean C:N ratios for white filament morphotypes from all the mats was 5.0 (1σ= 0.8), suggesting a high quality food source. The C:N ratios of gray filaments were higher and more variable than white morphotypes, with a mean of 15.0 (1σ= 10.5). The C:N ratios were highest for gray filaments and sediment from spring orifice sites, while the C:N ratios of gray filaments at the end of the microbial mats approached those of the white mat morphotypes (Table 2).

The sulfur content of white filament bundles was higher than the gray filaments (Table 2), presumably reflecting intracellular sulfur (as elemental S⁰) rather than organosulfur compounds. Typically, the highest sulfur content in bacterial cells, in the absence of stored sulfur, ranges up to 1% (w/w)[52]. However, the sulfur content of white filament bundles had a mean of 30.0% (1σ= 11.2%), and the white webs had consistently the highest sulfur content (Table 2). The gray filaments and sediment had significantly lower sulfur contents, with a mean of 1.9% (1σ= 0.6%), consistent with what would be predicted for bacterial biomass[52]. The sulfur content of white filaments was generally the same at the extreme upstream and downstream samples of the Upper Spring transect, but decreased by up to 10% in the middle stream reach (Table 2).

Carbon isotope systematics

The δ¹³C value for the Madison Limestone from the cave was +0.95‰, and the DIC reservoir along the Upper Spring transect had an average δ¹³C value of −7.5‰ (n = 7, 1σ= 0.1‰), and DIC from the Fissure Spring orifice water had a slightly higher δ¹³C value of −7.2‰. Microbial mat morphotypes had δ¹³C values ranging from −23‰ to −41‰ (mean −34.1‰, 1σ= 4.1) (Fig. 3). The low δ¹³C values reflect the large discrimination against ¹³C exhibited by autotrophs (e.g., ∼25‰ relative to total DIC for sulfur-oxidizing bacteria[53]).

Microbial mat morphotypes showed systematic variations in their carbon isotope compositions at most locations (Fig. 3). At all three spring locations, gray filaments consistently had among the highest δ¹³C values, whereas all coexisting white filament bundles had lower δ¹³C values. More specifically, near the distal portions of the Upper Spring mats, white feathery bundles and yellow patches (Fig. 2(f)) had some of the lowest δ¹³C values, whereas the white webs and gray filaments had the highest δ¹³C values (Fig. 3). In contrast, however, the feathery bundles from the more proximal region of the Upper Spring mats (196 m) had among the highest δ¹³C values. Moving downstream, the δ¹³C values of white filament bundles in both Upper Spring and Fissure Spring, decreased (Fig. 3).

Clone library coverage, species richness, and diversity

Eleven bacterial 16S rDNA clone libraries from four different microbial mat morphotypes were constructed and over 1000 clones were screened using RFLP. Nearly-full length 16S rRNA genes (>1300 bp) were sequenced in both directions from selected clones. Sequences from the same RFLP pattern that were 98% similar to each other were grouped as a phylotype (Table 3), and we used this classification scheme to estimate community diversity (Table 4). This level of sequence similarity takes into account micro-variations in genetic sequences due to PCR and cloning biases and variations in 16S rRNA gene copies [54,55]. Approximately 2% of the 16S rRNA gene sequences were chimera and removed from further analyses. Of the phylotypes identified, 44% had sequences that were 95% identical to GenBank sequences, corresponding to genus-level relationships[56], and 30% of the sequences were 98% identical to GenBank sequences, approximating species-level relationships[56]. The remaining phylotype sequences had 90% sequence similarity to GenBank sequences (Table 3).

Rarefaction analysis was done to determine if the libraries had saturated coverage based on the number of clones obtained per library. The rarefaction curves indicated different patterns of diversity for different morphotype libraries (Fig. 4). In the non-filament clone libraries (203g, 203w, 203y, and 248y), diversity was not fully covered compared to the saturation plateau reached for most of the white filament bundle libraries (124f, 127f, 190f, 198f) (Fig. 4). As there was an overall increase in the rate of phylotype accumulation in these unsaturated curves, major diversity within these libraries may not be well represented, although some of these libraries (e.g., 203y) do have high dominance values (Table 4).

Species heterogeneity among the clone libraries was generally low and many of the white filament libraries showed overwhelming dominance by one of two phylotypes. Species richness was higher for the non-filament morphotypes, with the white webs from 203 m and the yellow patches from 203 to 248 m showing the most diverse taxonomic representation among the eleven bacterial clone libraries (Table 3), even though observed species richness was lower than expected based on ACE and Chao1 values (Table 4). In comparison, although species richness of the white filament libraries varied, ranging from one to ten observed phylotypes, ACE and Chao1 estimates for the white filament libraries indicated that the observed phylotype numbers were close to the calculated values due to near-complete clone coverage for most of those libraries (Table 4). The diversity/dominance indices changed for the white filament clone libraries downstream for both the Upper and Lower Spring transects, such that the H′ values increased and D values decreased (Table 4).

16S rRNA gene clone libraries

The 16S rDNA clones were affiliated with several bacterial phyla (Table 3; Figs. 5 and 6). The majority of the sequences identified from the clone libraries belonged to the Proteobacteria taxonomic division, specifically the “Epsilonproteobacteria,” (68%) Gammaproteobacteria (12.2%), Betaproteobacteria (11.7%), and Deltaproteobacteria (0.8%) classes, as well as from other bacterial divisions, including the Acidobacterium (5.6%) and Bacteroides/Chlorobi (1.7%) divisions.

The “Epsilonproteobacteria” class

The highest numbers of clones from all the libraries (68%) were assigned to the “Epsilonproteobacteria” (Fig. 5; Table 3). Epsilonproteobacterial sequences from 17 phylotypes clustered into six groups based on phylogenetic position and sequence similarity, which suggests that genetic microdiversity in the microbial mats was high[57] (Fig. 5). Interestingly, regardless of morphotype location or site geochemistry, at least one epsilonproteobacterial phylotype was found in all clone libraries (Table 3). The epsilonproteobacterial groups identified from LKC have few closely related sequences from the public databases, suggesting that the diversity of these groups, and the “Epsilonproteobacteria” in general, is only now being realized.

The most abundant epsilonproteobacterial groups from all three springs formed two distinct clades, previously described as LKC group I and group II[26]. The closest relatives to the LKC groups I and II were the two environmental clones, sipK119 and sipK94, respectively (98–99% similar in nucleotide sequence), from microbial mats with a string-of-pearls morphology in sulfidic marsh springs at the Sippenauer Moor, Regensburg, Germany [58,59]. Clone sequences from LKC group I were also closely related (97–99% similar in nucleotide sequence) to environmental clones obtained from a petroleum-contaminated sulfidic groundwater storage cavity in Japan [60,61] and two clones from microbial mats from the sulfidic Cesspool Cave, Virginia[21] (Fig. 5). The closest cultured relative to LKC group I clones was Sulfuricurvum kujiense; this organism is a slightly curved rod isolated as a chemolithoautotrophic sulfur-oxidizer, capable of growth on thiosulfate, elemental sulfur, and hydrogen sulfide, and able to use molecular oxygen, nitrate, or ferric iron as electron acceptors[62]. LKC group II clones were more distantly related (90–94% similar) to miscellaneous marine, hydrothermal vent field and epibiont clones [63,64] and clones from a sulfidic cave microbial mat in Parker Cave, Kentucky[16] (Fig. 5). The phylogenetic affinities (Fig. 5) and sequence similarity of these two groups demonstrate that they are distinct from each other at more than the genus-level (85–87% similar).

LKC group III did not form a distinct phylogenetic cluster and was defined by several phylotypes from five libraries, supported by the range of sequence similarities among the sequences (91–99% similar in nucleotide sequence) and moderate boot-strap node values (Fig. 5). No LKC group III clones were found at the Lower Spring. LKC group IV, comprised of clones only from the Upper Spring, clustered closely with S. kujiense, and groundwater and cave environmental clones (Fig. 5). LKC group V had a range of sequence similarities among the group sequences (97–99% similar), indicating additional diversity that could not be resolved by RFLP. Seventeen clones from morphologically and geochemically diverse libraries, but mostly white filament bundle morphotypes, belonged to the novel sequence cluster LKC group VI (Fig. 5). The closest relatives to LKC group IV clones were environmental clones from acid mine drainage (95–96% similar).

The Gammaproteobacteria class

Twelve percent of all the clones belonged to the Gammaproteobacteria (Fig. 6(a)). Eighty-one clones formed the most abundant phylotype, closely related (99–100% similar in nucleotide sequence) to the environmental clone sipK4 from sulfidic marsh springs[58], which is also closely related to Thiothrix unzii. Several Thiothrix spp. have been identified from sulfidic caves, including Parker Cave, Kentucky[16], underwater caves and karst springs in Florida[27], and Cesspool Cave[21]. Clone library 203g was dominated by clones belonging to the Enterobacteriaceae, specifically the Pantoea and Serratia genera (Fig. 6(a)). The libraries 203g and 248y had six clones each that were closely related (99% similar) to Serratia marcescens. Nine sequences from the 124f and 203w libraries were distantly related to Beggiatoa sequences (90% similar), with one relative being the isolate Beggiatoa sp. MS-81-1c (Ahmad et al., unpublished Genbank submission) (Table 3; Fig. 6(a)). The weak sequence similarity to known Beggiatoa sequences, however, indicates that LKC clones may belong to a different, unclassified bacterial group within the Gammaproteobacteria. Beggiatoa-like filaments have been described from a marine cave in Italy using microscopy[65] and from microbial mats in Parker Cave[66], although phylogenetic investigations from Parker Cave did not support the presence of Beggiatoa [16].

The Betaproteobacteria class

Nearly twelve percent of the clones were affiliated with the Betaprotoebacteria, and were most closely related to Thiobacillus spp. (Fig. 6(b)). Three phylotypes were identified from two libraries. The closest relatives (94–95% similar in nucleotide sequence) were the environmental clone 44a-B2-21 from acid mine drainage (Labrenz and Banfield, unpublished Genbank submission) and Thiobacillus aquaesulis, a sulfur-oxidizing, facultative chemolithoautotroph[67]. Thiobacilli have been previously described from caves and mines [8,16,20,21,68], but environmental clones from those studies were not closely related to the LKC groups (Fig. 6(b)).

The Deltaproteobacteria class

Less than 1% of the clones were closely related (96–97% similar in nucleotide sequence) to Desulfocapsa thiozymogenes, the environmental clone sipK94 from the string-of-pearls mats in Germany[59], and the environmental clones SRB348 and SRB282 identified from the chemocline of the meromictic Lake Cadagno, Switzerland[69] (Table 3; Fig. 6(c)). D. thiozymogenes disproportionates thiosulfate, sulfite, or elemental sulfur to sulfate and sulfide[70].

The Acidobacterium division

One phylotype representing 5.6% of all the clones obtained from this study was closely related (96–97% similar in nucleotide sequence) to uncultured environmental clones within the Acidobacterium division. Library 203w was dominated by this clone group, and rare clones from this phylotype were found in five additional libraries (Table 3). Acidobacteria have not been identified from sulfidic cave microbial mats, but they have been identified from molecular surveys of Paleolithic cave paintings [9,25] and from submerged cave walls[23]. The closest relative was clone SJA-36 identified from an anaerobic bioreactor with trichlorobenzene contamination[71] (Fig. 6(d)). The LKC phylotype also expands the Acidobacteria-group 7 described by Liles et al.[72], which consisted of only a few environmental clones from soil, as well as the Acidobacteria-subgroup-b described by Schabereiter-Gurtner et al.[9] from La Garma Cave, Spain. Clone LKC3_156.13 had 92% sequence similarity to clone SJA-36, but also clustered as an unclassified taxonomic group within the Bacteroidetes phylum by phylogenetic analysis (Fig. 6(d)).

The Bacteroidetes/Chlorobi division

Seven phylotypes, each represented by rare ‘singleton' or ‘doubleton' clones, belonged to the Bacteroidetes/Chlorobi (BC) taxonomic group (Table 3; Fig. 6(d)). Three phylotypes (BC I–III) were closely related to environmental clones within the Bacteroides class, including environmental clones from lakes and contaminated groundwater. Four phylotypes (BC IV–VII) were related to environmental clones within the Sphingobacteria class (including the genus Cytophaga) obtained from a wide habitat range, including deep-sea hydrothermal vent metazoans, gas hydrate sediment, soil, and contaminated groundwater.

Discussion

Terrestrial subsurface environments are often inaccessible for study, limiting our understanding of ecosystem structure and dynamics, elemental cycling, and the impacts to earth and atmospheric biogeochemical processes. This investigation is part of an ongoing research program to investigate biogeochemical cycling in subterranean habitats, and we have been studying sulfidic caves as proxies for less accessible sulfidic karst aquifers. In this report our main research goals were to identify the bacterial groups comprising the cave microbial mats, to gain an understanding of how geochemistry may control microbial community diversity within the aphotic environment, and to elucidate potential ecosystem functioning and the impact of sulfur cycling and chemolithoautotrophy on the ecosystem. The results of this work demonstrate that similar microbial communities and concomitant microbially mediated biogeochemical cycles may be more widely dispersed in sulfidic groundwater habitats than previously recognized.

Geochemical controls on community structure and ecosystem function

Studies from other aquatic environments suggest that shifts in community structure could result from changes in nutrient availability, salinity, light penetration, turbidity, oxygen content, sulfide, or pH [73,74]. At present, however, there have not been any investigations that describe the controls on changing community structure from a freshwater aphotic habitat. Specifically, light penetration, turbidity, and salinity are not critical physicochemical conditions to influence these cave microbial communities, and changes in pH of the cave waters are not important because of pH buffering to circum-neutral by dissolving carbonate rock. Instead, we propose that (1) downstream variations in dissolved hydrogen sulfide concentrations, (2) increasing dissolved oxygen concentrations downstream, (3) colonization of the springs and outflow channels by “Epsilonproteobacteria”, and (4) changes in the organic carbon production and storage as a result of chemolithoautotrophy by epsilonproteobacterial groups are the most critical parameters affecting microbial community structure within the microbial mats.

The high concentrations of dissolved sulfide discharging from the springs would provide a rich energy source for sulfur-oxidizing bacteria. Although it is unlikely that abiotic autoxidation (i.e., chemical oxidation) and volatilization cause sulfide loss exclusively, there was an observed decrease in dissolved sulfide concentrations downstream (Fig. 1(b)). Abiotic autoxidation is extremely slow in poorly oxygenated water at pH ∼7.4 (the autoxidation half-life was calculated at >800 h; H₂S:HS⁻ pK 7.04) and sulfide volatilization from the water to the cave atmosphere accounts for <8% of the sulfide loss in the cave stream based on gas flux experiments[31]. With no other mechanism for sulfide loss, there would be, for example, significantly higher sulfide concentrations at the end of the Upper Spring microbial mat, as well as at the cave entrance 150 m further downstream. However, we observe a very rapid decrease in dissolved sulfide at each of the springs (Fig. 1(b)), and have demonstrated in an independent investigation that the loss is caused by microbial catalysis, even under microaerophilic conditions[31]. As the microbial mats are overwhelmingly dominated by metabolically active “Epsilonproteobacteria” based on previous investigations using FISH[26], we suggest that these organisms consume the dissolved sulfide in the cave as sulfur-oxidizers[31]. Although there is comparatively little information from culture-based studies [62,75–82], “Epsilonproteobacteria” are implicated in the oxidation of reduced sulfur compounds at low oxygen tensions in many sulfidic environments, including caves [16,21,26], deep aquifers[83], terrestrial springs and groundwater [58–62,84], oil fields[85], deep marine sediments and ocean water [86–89], hydrothermal vent sites [63,75,90–95], in association with deep-sea animal life at vent sites [64,96–100], and in engineered systems including sewage sludge and contaminated waste [101,102].

The relative abundances of epsilonproteobacterial and other taxonomic groups shifted through the microbial mats moving downstream with changing dissolved sulfide and oxygen concentrations. In general, the abundances of both epsilonproteobacterial LKC groups I and II decreased from the orifice pools downstream, and the highest abundance of LKC group I was from samples where the concentration of dissolved oxygen was very low at both the Fissure and Upper Springs. Clone libraries from the three spring orifices, which originated from habitats that are continuously replenished by sulfidic spring water, were dominated by one epsilonproteobacterial group, whereas downstream libraries had higher bacterial diversity (Tables 3 and 4). For example, at the Lower Spring all clones screened by RFLP belonged to the epsilonproteobacterial LKC group II, whereas one meter downstream in the microbial mat there were nine other bacterial groups identified, including those belonging to the Gammaproteobacteria and Betaproteobacteria (Table 3). At the Upper Spring, LKC group III was most abundant in downstream clone libraries (e.g., 203f) where the dissolved oxygen concentration was higher, suggesting that while this group may be involved with sulfur cycling, this group may prefer higher habitat oxygen content. At the three springs, there was also an increase in the abundance of Thiothrix- and/or Thiobacillus-like clones downstream, which is in accordance with the characterized metabolism of sulfide and oxygen gradient preferences for these groups [103,104] (Table 3). The pattern of occurrence for Acidobacteria, and dominance from the 203w clone library and not from upstream samples, suggests that these organisms also prefer higher habitat oxygen and lower sulfide concentrations.

Sulfur storage in the microbial mats from the three springs, as indicated by sulfur content, also changed downstream. There is no indication from cultures that “Epsilonproteobacteria” store sulfur intracellularly like Thiothrix spp.[103], although the marine epsilonproteobacterial strain “Candidatus Arcobacter sulfidicus” forms extracellular sulfur filaments [105,106] and cultures of nitrate-reducing sulfur-oxidizing “Epsilonproteobacteria” form sulfur as the metabolic end-product when nitrate is limiting or absent [77,80]. Therefore, the high sulfur content of white filaments from spring orifice samples (Table 2), which were dominated by “Epsilonproteobacteria” (Table 3), could be due to extracellular sulfur or sulfur accumulation due to nitrate-reduction. Higher sulfur content in downstream mat samples could also be due to incomplete sulfide oxidation to elemental sulfur by Thiothrix. The lower sulfur content for the 203y sample (8.6%) compared to the other morphotypes from the mat surface (Table 2) may be because the thiobacilli oxidize the sulfur within the mat due to the diminished dissolved sulfide concentration in the stream water.

While there are no known cultivated organisms closely related to LKC groups II, V, or VI clones, the closest cultured relative for LKC group I clones is strain YK-1, or S. kujiens [62]. It is possible that the organisms represented by LKC group I may also have similar metabolism to S. kujiense and grow under microaerophilic to anaerobic conditions, although nitrate and ferric iron concentrations are exceptionally low in the cave waters (Table 1). It should be noted, however, that closely related phylogenetic groups do not necessarily indicate similar ecophysiological characteristics[57], as Takai et al.[75] showed that the observed phylogenetic distribution of epsilonproteobacterial cultures isolated from deep-sea vents did not correlate with substrate or electron acceptor preferences, oxygen requirements, or geographic location. The fact that there are few sequences from the public databases that are closely related LKC epsilonproteobacterial groups suggests that the metabolic diversity of these environmental groups in the terrestrial subsurface has yet to be explored. This study expands the geographic distribution of “Epsilonproteobacteria”, significantly increases the number of sequences for “Epsilonproteobacteria” from terrestrial subsurface environments, and more importantly, characterizes the distribution of different epsilonproteobacterial groups according to physicochemical habitat and possibly ecosystem function.

Based on experiments at deep-sea vent sites where “Epsilonproteobacteria” are the first to colonize virgin surfaces, López-García et al.[100] suggest that epsilonproteobacterial groups initially and rapidly diversify metabolically within a habitat (natural or artificial), and thereby create microniches (such as anoxic regions) where other bacteria will subsequently colonize. High diversity among the specialized “Epsilonproteobacteria” would essentially maximize ecosystem functionality of other microbial groups and make the entire system more productive because of high growth rates, significantly high biomass, and quick adaptations to specific geochemical conditions of the habitat. However, Chesson et al.[30] also describe the tendency for the most productive species to also be the most dominant in a habitat, and thereby push others species to comparatively lower densities. These ecological caveats may explain why the microbial mats in Lower Kane Cave have high diversity within the “Epsilonproteobacteria”, but lower bacterial diversity overall.

The bacterial composition of the 203g clone library is one of the most telling examples of the control geochemistry has on community composition, and perhaps as an ecological consequence of “Epsilonproteobacteria” creating anoxic regions within the mat. The interior of the mat was dominated by clones closely related to two groups of Gammaproteobacteria that are characterized as facultative anaerobes with diverse metabolic capabilities [107,108]. Although rare clones closely related to D. thiozymogenes were also identified from some samples (Table 3), preliminary culture-based investigations of gray filaments and other mat samples suggest that sulfate-reducing bacterial guilds are also present in the mats, with <10⁶ cells ml⁻¹[109]. While molecular methods allow for the characterization of organisms that are difficult, if not impossible, to cultivate[110], unfortunately molecular methods can create significant biases and underestimates of particular microbial groups, especially if groups have abundances 10⁷ cells per volume [55,111]. Therefore, because this study focused on the white filament bundle morphotypes that were overwhelmingly dominated by “Epsilonproteobacteria”, it is likely that the diversity of anaerobes is underrepresented with respect to the total genetic diversity of the cave microbial ecosystem, and combined culture- and molecular-based approaches are currently being employed to better describe the diversity of the lesser abundant, anaerobic groups.

Chemolithoautotrophy in the subsurface

Most caves are energy- and nutrient-limited, commonly fed by surface streams in which photosynthetically-derived organic matter, sediments, and microorganisms are washed into the subsurface and deposited [10,11]. Previous studies have shown that microorganisms in caves associated with surface input are not chemolithoautotrophs, but instead are translocated soil heterotrophs, chemoorganotrophs, or fecal coliform bacteria from contaminated surface water[10]. Mikell et al.[12] estimate that 75% of microbial communities in most caves are heterotrophs. While we recognize that in the past the Bighorn River near the cave entrance may have had a role in inoculating the cave with microorganisms during previous flood stages, we believe that the LKC microbial communities are endemic to the cave and unaffected by surface hydrologic conditions because (1) the filamentous microbial biomass in LKC is significantly higher than the 10² to 10⁴ cells ml⁻¹ commonly found in other aquatic cave systems[112], and (2) the discharging springs contribute little to no allochthonous DOC or particulate organic carbon to the microbial community (Table 1). Because the geochemistry of the cave waters is consistent with reduced sulfur compounds being important energy sources for the microbial ecosystem and because most of the microbial groups can be associated with sulfur metabolism[31], we hypothesized that primary productivity was linked to the sulfur cycle.

We applied stable carbon isotope systematics to interpret the source of carbon to the LKC microbial mats, as well as how carbon is cycled within the mats. The overall organic carbon isotope compositions of the microbial biomass reflect significant isotopic discrimination against ¹³C relative to the inorganic carbon source, with 77% of the microbial mat samples having δ¹³C values 30‰, well below that of terrestrial biomass[113]; this demonstrates that photosynthetically-derived material is not important to the LKC ecosystem and that carbon for the ecosystem originates from chemolithoautotrophic inorganic carbon fixation. Porter[18] verified chemolithoautotrophic productivity from the white filamentous microbial mats at the Lower Spring by H¹⁴CO₃-assimilation, which suggested that there was more than six times higher autotrophic productivity than ¹⁴C-leucine-incorporation that tested for heterotrophy.

Chemolithoautotrophy in a cave ecosystem is important because it serves as the base for the cave food web, increasing both food quality and quantity [5,10]. Movile Cave, Romania, also a sulfidic cave system, has the first documented chemolithoautotrophically-based cave and groundwater ecosystem[7], and subsequently, chemolithoautotrophic microbial growth has been found in other active sulfidic cave systems, including marine caves from Cape Palinuro, Italy[65], Parker Cave[16], the Frasassi Caves, Italy [8,19], Cueva de Villa Luz, Mexico[17], Cesspool Cave[21], and the flooded Nullarbor caves, Australia[23]. The bulk of the LKC white filament bundle biomass had low C:N ratios averaging 5.0, compared to a C:N ratio of 5.7 for microbial mats from Movile Cave[5]. The C:N ratios for LKC white mat morphotypes also match previously reported ratios for bacterial cells (C:N = 3–5,[114]), but also to periphyton in surface streams (C:N = 4–8,[115]) and bacteria from a marine hydrothermal vent (C:N = 3.8–9.4,[116]). The C:N ratios are consistent with an insignificant influx and processing of allochthonous carbon, and instead suggest that carbon is provided in situ through autotrophy. In contrast, the high C:N ratios in the gray filaments samples proximal to the spring orifices, from the same locations as white filament bundles, indicate that the two microbial communities are not in communication. The high C:N ratios suggest that there is an abundant carbon supply, carbon storage due to an accumulation of processed biomass, and a reduction in nitrogen availability. Downstream the gray filament samples have C:N ratios similar to the white filament morphotypes, indicating that the white and gray microbial mat communities are in contact with each other structurally and that the gray filaments are no longer limited in nitrogen relative to the abundant supply of carbon (Table 2). The especially low C:N ratios suggest a high quality food that could be used by higher trophic levels [51,52]. Incidentally, there are large populations of endemic snails (Physa spelunca) that graze upon the microbial mats at all the LKC springs[117].

The mechanisms for inorganic carbon fixation were not evident based on carbon isotope analyses, as there are several different pathways for inorganic carbon fixation, and not all fixation pathways and their isotopic effects are known. Microorganisms that fix CO₂ by the Calvin–Benson–Bassham cycle, the predominate and most important carbon fixation pathway for photosynthetic and chemosynthetic bacteria, have isotopic values that fall into two categories based on the form of CO₂-fixing enzyme, ribulose-1,5-bisphosphate carboxylase/oxygenase (RubisCO)[118]. Nearly all of the mat samples from LKC have δ¹³C values that fit into the RubisCO form I group with δ¹³C values ranging between −27‰ and −35‰ (the ‘−30‰ group')[118]. The chemolithoautotrophic pathway using the reductive citric acid (TCA) cycle imparts a smaller (∼−10‰) carbon isotope fractionation [105,106,119]. Physicochemical conditions, such as flow velocity, water depth, temperature, pH, and CO₂ concentrations, can affect the effective isotope discrimination of autotrophs, which would result in tremendously different isotopic discrimination values [119–121]; however, the stream water in the Upper Spring transect maintains constant chemistry and turbulent flow, suggesting that these physical conditions are not important.

Variations in the carbon isotope composition among the different microbial mat morphotypes in downstream transects suggest carbon cycling between chemolithoautotrophs and heterotrophs. The carbon isotope ratios for each of the morphotypes upstream were higher than the same morphotypes downstream, especially for samples from the Upper Spring transect (Fig. 3). The systematic differences in the carbon isotope composition among the mat morphotypes at any location suggest that there may be distinct carbon isotope effects imparted by specific populations during carbon fixation. Compartmentalization of the microbial populations within a morphotype and changing abundances of bacterial populations downstream could account for the observed trend if the downstream populations express larger ¹³C discrimination. However, an alternative explanation for the downstream trend may be that mat stratification, due to redox conditions, creates an environment for nutrient spiraling[122]. The autotrophically-fixed carbon, when respired as CO₂, has a low δ¹³C value, and may be transported downstream and preferentially reassimilated by autotrophs at the mat boundary layer; the proportion and amount of autotrophic recycling of the fixed carbon derived from respiration should increase downstream[122]. Carbon isotope compositions of aerobic and anaerobic mat components reflect a complicated relationship between primary production and carbon recycling, with isotope ratios tending to increase with enhanced carbon recycling. The δ¹³C values of the anaerobic (gray filament) mat components are generally higher than coexisting white filament bundles or web and yellow patch morphotypes, and progressively converge upon those of the white morphotypes dominated by autotrophic “Epsilonproteobacteria” (Fig. 3). This reflects the assimilation and respiration of autotrophically-produced organic carbon by anaerobic heterotrophic bacteria downstream. Nutrient spiraling, as it pertains to carbon cycling, has not been previously described from chemosynthetic or subterranean ecosystems. In the future, more detailed carbon isotope ratio analyses, microautoradiography to test for specific carbon substrate uptake[123], using primer sets to amplify partial subunits of RubisCO (forms I and II) enzyme from DNA[4], and culturing of specific microbial groups will better address the type and extent of autotrophy and carbon cycling.

In this study, we combined molecular techniques with stable organic carbon isotope ratio analysis to examine the dynamics of microbial community structure and nutrient cycling in microbial mats occupying aphotic sulfidic springs. Building on our previous work describing the dominance of the cave microbial mats by “Epsilonproteobacteria”[26], we found several additional evolutionary lineages within the “Epsilonproteobacteria”, increasing the geographic diversity of this class to subsurface environments. Microbial mat bacterial diversity was low overall; certain bacterial groups were found only in one microbial mat morphotype, and most bacterial groups were rarely found or were completely absent in other morphotypes. The concentration of dissolved oxygen and dissolved sulfide controlled the distribution of sulfur-oxidizers with differing requirements for oxygen, such that those preferring higher oxygen conditions were found at the end of the microbial mats where dissolved oxygen was highest. The “Epsilonproteobacteria” provide chemolithoautotrophic energy to the ecosystem and colonize the nutrient-poor habitat and diversify genetically and metabolically, creating new habitats due to the formation of a dense mat, which increases species richness downstream. The resulting stratification of distinct microbial groups within the mats based on geochemistry and stream advection increase nutrient availability downstream, and perpetuates spiraling of carbon among multiple components of the microbial ecosystem. Future work will address the extent to which the “Epsilonproteobacteria” are distributed in other aphotic habitats, and what role these organisms may play in nutrient cycling and changing subsurface habitat conditions.

Acknowledgements

We thank the Bureau of Land Management for continuing to permit this research. We thank S. Engel, T. Dogwiler, M. Edwards, K. Mabin, R. Payn, and J. Deans for field assistance, and K. Crandall for laboratory support and critical insights. This work was supported by a National Science Foundation LExEn grant (EAR-0085576), Brigham Young University, and the Geology Foundation of the University of Texas at Austin.

References

Author notes