Bacterial and Archaeal Phylogenetic Diversity of a Cold Sulfur-Rich Spring on the Shoreline of Lake Erie, Michigan

Abstract

Studies of sulfidic springs have provided new insights into microbial metabolism, groundwater biogeochemistry, and geologic processes. We investigated Great Sulphur Spring on the western shore of Lake Erie and evaluated the phylogenetic affiliations of 189 bacterial and 77 archaeal 16S rRNA gene sequences from three habitats: the spring origin (11-m depth), bacterial-algal mats on the spring pond surface, and whitish filamentous materials from the spring drain. Water from the spring origin water was cold, pH 6.3, and anoxic (H₂, 5.4 nM; CH₄, 2.70 μM) with concentrations of S²⁻ (0.03 mM), SO₄²⁻ (14.8 mM), Ca²⁺ (15.7 mM), and HCO₃⁻ (4.1 mM) similar to those in groundwater from the local aquifer. No archaeal and few bacterial sequences were >95% similar to sequences of cultivated organisms. Bacterial sequences were largely affiliated with sulfur-metabolizing or chemolithotrophic taxa in Beta-, Gamma-, Delta-, and Epsilonproteobacteria. Epsilonproteobacteria sequences similar to those obtained from other sulfidic environments and a new clade of Cyanobacteria sequences were particularly abundant (16% and 40%, respectively) in the spring origin clone library. Crenarchaeota sequences associated with archaeal-bacterial consortia in whitish filaments at a German sulfidic spring were detected only in a similar habitat at Great Sulphur Spring. This study expands the geographic distribution of many uncultured Archaea and Bacteria sequences to the Laurentian Great Lakes, indicates possible roles for epsilonproteobacteria in local aquifer chemistry and karst formation, documents new oscillatorioid Cyanobacteria lineages, and shows that uncultured, cold-adapted Crenarchaeota sequences may comprise a significant part of the microbial community of some sulfidic environments.

Cold, sulfidic springs upwelling into caves (1, 16-19) or exposed at the land surface (14, 15, 31, 39, 47, 50, 51) have recently been shown to harbor unique microbial communities, reflective of the aqueous sulfur chemistry of the upwelling groundwater or of unique cave conditions. Within these spring and cave ecosystems, new and unique Epsilonproteobacteria 16S rRNA gene sequences associated with a limited number of cultured isolates that carry out oxidation of sulfur compounds have been discovered (7). The abundance of Epsilonproteobacteria sequences in these settings and associated biogeochemical research have led to new interest in the role of microbially mediated sulfuric acid speleogenesis as an important limestone dissolution process that may contribute to the development of karst features in limestone bedrock (19). Additionally, in streamlets from sulfidic springs, unique symbioses between uncultured Euryarchaeota, Crenarchaeota, and Epsilonproteobacteria spp. that grow in whitish, macroscopically visible filaments have been described (31, 51). Sulfur cycling was identified as a major means of energy production and maintenance of microbial communities in cold, saline, perennial springs emanating from permafrost in the Arctic (47). Studies of cold, sulfidic springs have therefore provided new insights into microbial metabolism, ecology, and evolution as well as groundwater biogeochemistry and geologic processes.

All studies of sulfidic springs to date have focused on terrestrial landscapes typically associated with limestone (CaCO₃) bedrock. Limestone is one of several carbonate sedimentary rocks deposited by ancient seas, which may contain significant amounts of gypsum (CaSO₄·2H₂O) as well as pockets of hydrocarbon deposits, both a source of sulfur. Water that moves for long distances through such rocks evolves through sequential dissolution and precipitation reactions to a geochemistry that bears little resemblance to freshwater. SO₄²⁻ becomes available for microbial reduction to sulfide in aquifer zones where conditions are appropriate. Where spring waters rich in CaCO₃, CO₂, and sulfide emerge at the surface, carbonate deposition and microbially mediated sulfide oxidation occur. These processes result in tufa deposits and the whitish crusts often noted in sulfidic spring outflows (12, 13). Carbonate bedrock underlies large portions of the lower Laurentian Great Lakes. Caves in contact with lake water occur on islands in Lake Erie and along the Bruce Peninsula in Ontario, Canada. A cold, sulfidic spring is located in Ancaster, Ontario, about 5 km from the Lake Ontario shoreline (13). Recently, plumes of high-conductivity sulfidic groundwater, surrounded by whitish filamentous materials and variously colored microbial mats, were reported to occur at a 93-m depth in Lake Huron (2, 49). However, there have been few molecular surveys of Bacteria or Archaea in any Great Lakes environment, and no reports focusing on the molecular phylogenetic diversity of microorganisms associated with these Great Lakes sulfidic environments.

Along the western shoreline of Lake Erie and within Monroe County, MI, sinkholes and springs are abundant in the Silurian-Devonian carbonate bedrock, and Ca²⁺ and Mg²⁺ with SO₄²⁻ or HCO₃⁻ dominate groundwater composition (43). In some areas of Monroe County, sulfide in groundwater prohibits its use as a drinking water source. Great Sulphur Spring (GSS) was first described by Sherzer in 1900 (53) and was named for its sulfide-rich water. The spring arises from Silurian-Devonian carbonate bedrock within 0.5 km of the Lake Erie shoreline and is a convenient location for accessing sulfide-rich groundwater and for exploring potential interactions between groundwater and lake water. As part of a larger study of nearshore groundwater interactions with Lake Erie (27) and to better understand the potential role of microorganisms in sulfur chemistry of nearshore groundwater, we evaluated the chemistry and bacterial and archaeal 16S rRNA gene diversity of GSS. Our study documents a unique microbial community for the Laurentian Great Lakes, comprised in large part of new lineages and uncultivated members of the Archaea, Deltaproteobacteria, Epsilonproteobacteria, and Cyanobacteria. These sequences suggest a microbial community structure driven by (possibly H₂S-based) carbon fixation and chemolithotrophy of reduced compounds such as H₂, H₂S, or reduced nitrogen compounds, all consistent with spring geochemistry.

MATERIALS AND METHODS

Site description.

GSS (latitude 41°46′07", longitude 83°27′22") is located within the Erie Marsh preserve, managed by The Nature Conservancy and the Erie Shooting and Fishing Club (Fig. 1 and 2), and was first named and described by Sherzer in 1900 (53). The spring forms a pond 42 m in diameter and 9 to 10 m deep in the center, with a shallow shelf around the edge (Fig. 2). Mats of microbial and algal growth form on the shelf and slough off to rise and float on the pond surface (Fig. 2). The mean discharge from the major pond outlet (recorded five times between September 2003 and February 2004) was 0.14 m³/s. The spring arises from the Salina Group, a Silurian-Devonian sedimentary rock composed of dolomite [CaMg(CO₃)₂], shale, minor limestone (CaCO₃) including associated anhydrite (CaSO₄) and gypsum (CaSO₄·2H₂O), secondary calcite (CaCO₃), and celestite (SrSO₄), and some hydrocarbons (43). Tufa deposits surround the spring for some distance, and all materials in the outflow channel were covered by whitish deposits and filamentous white strands. To assess possible differences in community structure influenced by gradients of sulfur and oxygen as reported for other sulfur springs, three habitats at GSS were sampled: the spring source at approximately 11 m deep (Fig. 2), detached mats on the pond surface (Fig. 2), and the whitish filamentous materials at the outflow (drain) (not shown).

FIG. 1.

Location of GSS. (Modified from reference 27 with permission.)

FIG. 2.

Photograph of GSS showing a portion of the spring pond. Ms, mats submerged beneath shallow water; Md, detached mat material floating on pond surface; O, open area of pond over spring origin. Lake Erie is approximately 0.5 km beyond the tree line. Drain occurs to the right of the area in the image.

Sample collection.

Water from the spring origin was collected using a Grundfos submersible pump (Grundfos Pumps Corp., Olathe, KS) with Teflon-lined polyethylene tubing. The pump was lowered from the pond surface at the point of greatest depth, then raised from the bottom about 0.9 m. Water was sampled at a low flow rate to avoid entraining blackish particulates characteristic at higher flow rates. Water was delivered into and samples were processed in the field within a clean PVC-and-plastic chamber. Sample collection and processing, bottle type, filtration, and sample preservation followed the U.S. Geological Survey (USGS) National Field Manual (accessible online at http://water.usgs.gov/owq/FieldManual). Samples were shipped overnight on ice to the appropriate laboratories.

For bacterial samples from the spring origin, a higher flow rate was used so that some of the blackish particulates characteristic of water in the spring throat were also captured. An additional piece of sterile tubing permitted aseptic delivery of bacterial water samples to sterile bottles within the field-processing chamber. All spring origin water and bacterial samples were collected after 0.5 h of continuous pumping to ensure that only groundwater from the spring origin (and not pond water) was collected. Mat samples were collected by coring with a cutoff sterile large-bore syringe; the upper 1 cm of the core was cut off and transferred to sterile petri dishes. Whitish filamentous bundles at the drain were aseptically collected into wide-mouth sterile containers using sterile forceps.

Water quality.

Extensive chemical analyses allowed comparison with the previously determined chemistry of the Silurian-Devonian aquifer in Monroe County, MI (43). Water was analyzed for temperature, pH, specific conductance, and dissolved oxygen with a Hydrolab multiparameter probe (Hach Co., Loveland, CO). Alkalinity was determined in the field by titration (Hach Co.). At the USGS National Water Quality Laboratory (Denver, CO), major ions and trace metals were quantified by inductively coupled plasma-mass spectrometry (20, 24), graphite furnace atomic absorption spectrophotometry (Cr) (38), colorimetry (Br) (22), ion chromatography (Cl, SO₄), or ion-selective electrode analysis (F) (21, 22). Orthophosphate and total dissolved phosphate were determined by semiautomated colorimetry (EPA method no. 365.1), NH₄⁺ and organic nitrogen by Kjeldahl digestion (46), and NO₃⁺ and NO₂⁻ by cadmium reduction/diazotization (21). The ratios of ²H/¹H and ¹⁸O/¹⁶O were determined by mass spectrometry (29). The ratio of ¹³C/¹²C of dissolved inorganic carbon was determined by mass spectrometry and the ¹⁴C content of dissolved inorganic carbon was determined by accelerator mass spectrometry at the University of Waterloo Environmental Isotope Laboratory, Waterloo, Ontario, Canada. ¹⁴C was reported as percent modern carbon normalized to the 1950 National Bureau of Standards oxalic acid standard. Tritium levels were determined by electrolytic enrichment and gas counting at the University of Miami Tritium Laboratory, Miami, FL. Helium and hydrogen gas were analyzed using gas chromatography with thermal conductivity detection (4).

Community DNA isolation, PCR amplification, and cloning.

Prior to DNA extraction, particulates were fractionated from spring origin water samples by centrifugation at 10,000 rpm for 15 min. Quadruplicate samples (1 g) were weighed and processed separately. DNA from samples was extracted using protocols of the Ultraclean soil DNA kit (MoBio Laboratories, CA). DNA extracted from each independent sample was used as a template for the PCR amplification of bacterial and archaeal 16S rRNA genes. Primers (5′ to 3′) were BAC 8F (AGAGTTTGATCCTGGCTCAG), BAC 1492R (TACGGYTACCTTGTTACGACTT), ARCH 21F (TTCCGGTTGATCCYGCCGGA), and ARCH915R (GTGCTCCCCCGCCAATTCCT). The reaction mixture (50 μl) contained 50 to 100 ng of DNA, 1× bovine serum albumin, 1× PCR buffer, 1.5 mM MgCl₂, 250 μM of each deoxynucleoside triphosphate, 0.3 μM of each primer, and 2.5 U of Taq DNA polymerase (Invitrogen, San Diego, CA). Amplification was performed in a PerkinElmer 9700 thermal cycler (PE Applied Biosystems) with a “hot start” (94°C for 5 min), followed by 30 cycles of 94°C for 40 s, 55°C for 50 s, and 72°C for 1 min 30 s with a 10-min extension at 72°C. Amplified PCR products were purified through Qiagen PCR purification kits (Qiagen, Chatsworth, CA), and 16S rRNA gene amplicons from independent samples were pooled and cloned using a TOPO cloning kit (Invitrogen, San Diego, CA). To determine the presence of inserts of the expected size, 20 to 30 transformants were analyzed by direct PCR screening. Clones to be analyzed were lysed in 50 μl of buffer (10 mM Tris-HCl, pH 8.0) for 10 min at 95°C. Sequence inserts were PCR amplified from lysates (1 μl) with plasmid-specific primer M13F or M13R and the primer that hybridized within the insert, using reaction conditions as described above. The sizes of the inserts were checked by electrophoresis on a 1% (vol/vol) agarose gel.

Sequencing and phylogenetic analysis.

For sequencing, the PCR insert-containing subclones from Luria agar plates with kanamycin (50 μg/ml) were inoculated into 96-well plates (Qiagen, Chatsworth, CA) containing 100 μl of LB freezing medium (20.0 g Bacto tryptone, 10.0 g yeast extract, 20.0 g NaCl, 12.6 g K₂HPO₄, 3.6 g KH₂HPO₄, 1.0 g sodium citrate dihydrate, 2.0 g MgSO₄·7H₂O, 1.8 g ammonium sulfate, 88 ml glycerol, pH 7.5, 2 liters) and incubated overnight at 37°C. High-throughput sequencing of the 16S rRNA gene was performed at the Research Technology Support Facility of Michigan State University, East Lansing, MI, as per prescribed protocols (http://rtsf.msu.edu/) on an ABI 3730 genetic analyzer or an ABI Prism 3700 DNA analyzer (PE Applied Biosystems). Sequences were initially analyzed by using Sequencher software (Gene Codes, Ann Arbor, MI), and then clone sequences were phylogenetically classified using the RDP classifier from the Ribosomal Database Project II at Michigan State University (http://rdp.cme.msu.edu). Archaea and Bacteria sequences were aligned using ARB software (34). The resulting alignments were checked manually and corrected for alignment abnormalities. Neighbor-joining analysis was performed by using PAUP software (55). Bootstrap analysis was performed with 1,000 replications.

Nucleotide sequence accession numbers.

The 16S rRNA gene sequences obtained from this study have been deposited in the GenBank database under accession numbers FJ967839 to FJ968104.

RESULTS

Spring chemistry.

GSS water at the spring origin reflects the composition of the Silurian-Devonian bedrock from which it arises (Table 1). USGS well G33 is located about 1 km inland from the GSS (Fig. 1) and draws water from the Salina Group. Sinkholes, additional springs, and artesian wells are present in the area. Water from well G33 was estimated by ¹⁴C dating and geochemical composition to be greater than 15,000 years old (43). A comparison of the composition of GSS spring origin water with that of well G33 indicates an overall similarity of composition but with higher concentrations of Ca²⁺, HCO₃⁻, SO₄²⁻, and tritium and a higher percentage of modern carbon in GSS spring origin water. Each of these differences likely arises from the mixing of largely anoxic Salina Group groundwater with more aerated modern water, either within the spring pond itself or possibly via karst connections to surface water along the flow path to GSS. The lower pH of GSS spring origin water may result from oxidative reactions, such as sulfide oxidation, as anoxic groundwater mixes with oxygen along the flow path to the spring outlet. Both sulfide oxidation and release of SO₄²⁻ by interaction of acidic groundwater with sulfate-containing aquifer minerals could contribute to the relatively higher sulfate concentration in GSS water compared to that in well G33 groundwater. The presence of H₂ at 5.38 nM, and of CH₄ and S²⁻, indicates anaerobic microbial processes in the groundwater at the spring origin. The water temperature of the GSS spring origin reflects typical groundwater temperatures and is constant year round, resulting in an open pond even in winter. The chemical composition of GSS spring origin water is very different from that of typical Lake Erie water, having a much greater mineral content and relatively high concentrations of N and P (Table 1). Nevertheless, the overall elemental signature of GSS water is reflected in Lake Erie water.

TABLE 1.

Physicochemical analysis of GSS water and of adjacent waters

Parameter	GSS origin	USGS well G33, Monroe County, MIa	Lake Erie waterb
Physical parameters
Dissolved oxygen (mg/liter)	0.7	0.2	9.8
pH	6.3	7.5	8.4
Specific conductance (μS/cm)	2,590	1,730	282
Temperature (°C)	11.3	11.5	−c
Cations and anions
Ca2+ (mM)	15.65	5.49	4.83
Mg2+ (mM)	2.08	3.74	1.42
Na+ (mM)	0.46	0.78	0.39
K+ (mM)	0.06	0.07	0.03
HCO3− (mM)	4.10	2.10	1.35
Cl− (mM)	0.79	1.02	0.51
F− (mM)	0.06	0.14	0.01
SiO2 (mM)	0.22	0.18	0.02
SO42− (mM)	14.75	8.33	0.34
Nutrients
NO3−-N (mM)	0.08	<0.01	0.06
PO4-P (μM)	2.49	−	0.09
NH4-N (μM)	9.0	−	2.0
Gases
H2 (nM)	5.38	−	−
CH4 (μM)	2.70	−	−
CO2 (μM)	764	−	−
N2 (μM)	905.7	−	−
Other parameters
S2− (mM)	0.0312	0.0811	−
Sr (mM)	0.1814	0.1483	−
Fe (mM)	0.0004	0.0003	−
Li (mM)	0.0043	0.0079	−
Dissolved organic carbon (mM)	−	0.80	−
Tritium (pCi/liter)	19.04	2.32	−
14C (% modern carbon)	52.8	5.8	−

^aData obtained from reference 43.

^bData obtained from references 3, 37, and 56.

^c−, not determined.

The shallow water at the edge of the spring pool or in the drain (Fig. 2) was not analyzed for chemical composition. Specific conductance and pH of the shallow water were similar to that at depths near the spring origin; however, the temperature and dissolved oxygen concentration of the shallow water were greater (14 to 17°C and 2.8 to 4.6 mg/liter, respectively).

Community composition in the spring origin, mat, and drain clone libraries.

The number of sequences analyzed and a breakdown of the phylum affiliation of the bacterial and archaeal clone libraries from the spring origin and mat are listed in Table 2. Only 8 Bacteria sequences (1 Cyanobacteria, 1 Beta-, 2 Epsilon-, and 4 Gammaproteobacteria) and 10 Archaea sequences (all Crenarchaeota) were obtained from drain samples. These sequences are described in the appropriate sections.

TABLE 2.

Composition of phylogenetic clone libraries from the spring origin and from the mat

Classification	Spring origin		Mat
	n	% of library	n	% of library
Domain Bacteria	131		50
Chloroflexi	2	2	0	−a
Cyanobacteria (including chloroplasts)	61	47	14	28
Chloroplasts	18	14	8	16
Firmicutes	3	2	1	2
Planctomyces	2	2	0	−
Bacteroidetes	4	3	1	2
Verrucomicrobia div. nov.	1	1	0	−
Proteobacteria
Alphaproteobacteria	4	3	4	8
Betaproteobacteria	11	8	9	18
Deltaproteobacteria	13	10	3	6
Epsilonproteobacteria	21	16	3	6
Gammaproteobacteria	9	7	15	30
Domain Archaea	45		22
Crenarchaeota
Soil crenarchaeota	3	6	0	−
Miscellaneous crenarchaeota	18	40	4	18
Euryarchaeota
Marine group III	24	53	18	82

^a−, not applicable.

Phylogenetic analysis of bacterial clone libraries.

In all phylogenetic trees (Fig. 3 to 8), GSS clone designations containing an “O” indicate a spring origin source (e.g., GS1O19), those with an “M” indicate a mat source, and those with a “D” indicate a drain source. Of 14 non-Proteobacteria sequences (12 of spring origin, 2 of the mat) in five phyla (Chloroflexi, Firmicutes, Planctomyces, Bacteroidetes, and Verrucomicrobiae), no sequence was >95% similar to that of any described organism. Alphaproteobacteria sequences were not very abundant in the clone libraries, with only four sequences in the spring origin and four in the mat and most in the Rhodobacter (sulfur-oxidizing phototrophs intolerant of high sulfide concentrations) and Sphingomonas groups (Table 2). Gammaproteobacteria clones in the spring origin and mat included sequences highly similar to those of Aeromonas spp. (Fig. 3), for which sulfide production is a characteristic. Other Gammaproteobacteria sequences were moderately similar to the sequence of an uncultured gammaproteobacterium (PHOS-HE4) from deep sea sediments, and six Betaproteobacteria sequences were not similar to either environmental sequences or cultured organisms. However, the majority of Gammaproteobacteria and Betaproteobacteria clones were >95% similar to phototrophic and lithotrophic genera, such as Thiobacillus, Rhodoferax, Hydrogenophaga, or Thiothrix.

FIG. 3.

Phylogenetic analysis of bacterial 16S rRNA gene sequences obtained from the spring origin, mat and drain clone libraries and similar to the Beta- and Gammaproteobacteria sequences. The tree was inferred by neighbor-joining analysis of at least 500 homologous positions of sequence from each organism or clone. Numbers at the nodes are the bootstrap values (percentages) based on 1,000 replicates. The scale bar indicates the estimated number of base changes per nucleotide sequence position. GSS clone designations containing an “O” indicate a spring origin source (e.g., GSO-, GS1O19), those with an “M” indicate a mat source, and those with a “D” indicate a drain source.

FIG. 8.

Phylogenetic analysis of bacterial 16S rRNA gene sequences obtained from the spring origin, mat and drain clone libraries, similar to Euryarchaeota sequences. Tree construction and abbreviations are as described for Fig. 3.

Epsilonproteobacteria sequences comprised 16% of the spring origin clone library and 6% of the mat clone library and two of the seven drain clone sequences (Table 2) (Fig. 4). Epsilonproteobacterial sequences currently belong to only two valid orders. Of 17 GSS sequences in the Campylobacterales order, a group of 16, all from the spring origin, were highly similar to uncultured sequences obtained from petroleum-contaminated groundwater discharged from underground crude oil storage cavities (57). The remaining sequence, GS2M22, was similar to sequences of Dehalospirillum and “Geospirillum” spp., which oxidize reduced sulfur compounds using a variety of electron acceptors. All remaining sequences belong to a large cluster with no hierarchical taxonomic classification, provisionally named the “Thiovulgaceae fam. nov.” (7). A group of four spring origin sequences was similar to the sequence of Sulfurovum lithotrophicum, an autotroph that oxidizes elemental sulfur and thiosulfate. Sequences GS3M16 and GS1O16 were similar to sequences in groundwater groups I and II, respectively, retrieved from Lower Kane Cave, WY, in the United States (7, 17) and from sulfur springs in Germany (AJ307941, AJ307940) (50). GS1O16 was similar to the sequence of Sulfuricurvum kujiense, an autotroph isolated from oilfield waters that can oxidize the sulfur in organic compounds (32). Finally, GS2M4 was similar to environmental sequences in the “terrestrial group” of the “Thiovulgaceae” (7).

FIG. 4.

Phylogenetic analysis of bacterial 16S rRNA gene sequences obtained from the spring origin, mat, and drain clone libraries, similar to Epsilonproteobacteria sequences. Tree construction and abbreviations are as described for Fig. 3.

Deltaproteobacteria clones fell into the families Desulfobacteraceae and Syntrophobacteraceae and the class Myxococcales (Table 2) (Fig. 5). No clone was >95% similar to a characterized organism or an environmental sequence. A cluster of seven spring origin clones was moderately similar (<95%) to Syntrophus spp.; otherwise, sequences were distributed among the mat and spring origin libraries. Several environmental sequences to which Deltaproteobacteria clones were similar, including Sva0405 to Sva0679, NaphS2, salt marsh clone LCP-80, and SMP-2 (which has been isolated and named Haliangium ochraceum SMP-2) (58), were retrieved from marine environments.

FIG. 5.

Phylogenetic analysis of bacterial 16S rRNA gene sequences obtained from the spring origin, mat and drain clone libraries, similar to Deltaproteobacteria sequences. Tree construction and abbreviations are as described for Fig. 3.

Overall, 33% of GSS cyanobacterial sequences fell into families I, IV, and XIII based on 16S rRNA genes (RDP-II v9), and 77% of the sequences fell into the unclassified Cyanobacteria species (Fig. 6). Little congruity exists between cyanobacterial nomenclature based on morphology and phylogenetic relations based on 16S rRNA genes; hence, organisms assigned the same genus name based on morphology may appear in several families in Fig. 6. Three GSS sequences were highly similar to Synechocystis sp. strain PCC6803, and several other GSS cyanobacterial sequences were >95% similar to well-characterized organisms such as “Planktothrix,” Planktothricoides, Geitlerinema, and Phormidium spp. The largest group of 53 GSS sequences, in the unclassified Cyanobacteria, was most similar to the uncultured sequence DQ889938 associated with the coral Erythropodium caribaeorum and was moderately similar to “Limnothrix redekei” 165c and 007a (26). “Limnothrix redekei,” “Planktothrix agardhii,” “Planktothrix rubescens,” and Planktothricoides raciborskii were all previously in the “Oscillatoria” genus (54). Many oscillatorioid cyanobacteria have the ability to use H₂S as an electron donor for anoxygenic photosynthesis, producing S° as a product, and live in other sulfide-rich habitats (5, 25). In addition, they are typically filamentous, bloom-forming cyanobacteria among which many strains produce hepatotoxic cyclic peptides, nodularins, and microcystins. Microcystins are a water quality concern in western Lake Erie (48). In addition to cyanobacterial sequences, 26 sequences were similar to those of diatom plastids in the Chloroplast 16S rRNA gene family, genus Bacillariophyta. A previous study (44) also obtained sequences similar to those of Skeletonema and Haslea wawrikae chloroplasts, to those of Synechocystis strain PCC6803, and those within the filamentous cyanobacteria (“Oscillatoria,” Lyngbya, and Spirulina) in Lake Erie.

FIG. 6.

Phylogenetic analysis of bacterial 16S rRNA gene sequences obtained from the spring origin, mat and drain clone libraries, similar to Cyanobacteria and chloroplast sequences. Tree construction and abbreviations are as described for Fig. 3.

Phylogenetic analysis of archaeal clone library.

No archaeal sequence matched that of any cultured organism, and none was similar to sequences of methanogens or other cultured archaeal groups. With one exception, all archaeal sequences fell into clades of the phylum Crenarchaeota or Euryarchaeota for which there are no cultured representatives.

Crenarchaeota.

The Crenarchaeota sequences fell into three groups (Fig. 7) that form complexes within the ubiquitous, nonthermophilic, terrestrial, and marine uncultivated Crenarchaeota species (52). Soil Crenarchaeota spp. have been obtained from tomato roots (TRC designation Fig. 7) (9, 52) as well as soils from a variety of locations. Five sequences from the GSS drain (GSS Crenarchaeota group 1 [GSS Cren Grp 1]) were most similar to unidentified archaeon LMA137, a sequence obtained from sediment at a 101-m depth in Lake Michigan (36), indicating that some sequences detected at GSS may be more widely dispersed in the Great Lakes. Although most sequences in marine group I Crenarchaeota spp. were obtained from marine habitats (e.g., sequences U71113 and U71117 were isolated from Irish coastal waters at a 10-m depth), nonmarine sequences such as LMA137 and LMA229 have consistently clustered with this group (30). Very recently, a marine organism (“Candidatus Nitrosopumilus maritimus” [33]) was cultured from marine group I and identified as an autotrophic ammonia oxidizer. GSS Cren Grp 1 sequences were broadly similar to the sequence of this new isolate and very similar to uncultured marine crenarchaeote DQ085102, a sequence obtained in the same study. Of particular note, marine group I sequence AM055707 was obtained from a string-of-pearls-like microbial consortium in a cold sulfidic spring in Germany (31). The miscellaneous Crenarchaeota spp. comprise sequences from a variety of environments, including marine seafloor sediments, anaerobic digestors (U81774), municipal wastewater sludge (AF424768 and AF424775), lake sediment in Indiana (U59986), boreal forest lake plankton (AJ131315), and a microbial consortium within greenish-white streamers in the outlet from a cold sulfidic spring in Germany (AM055703 [31]). Sequences in the miscellaneous Crenarchaeota spp. were also obtained from a sulfide- and sulfur-rich spring in Oklahoma (15).

FIG. 7.

Phylogenetic analysis of bacterial 16S rRNA gene sequences obtained from the spring origin, mat and drain clone libraries, similar to Crenarchaeota sequences. Tree construction and abbreviations are as described for Fig. 3.

Euryarchaeota.

All Euryarchaeota sequences were similar to sequences in uncultivated Archaea in marine group III (9, 52). GSS Euryarchaeota groups 1 and 2, as well as seven additional clones, were broadly similar to uncultured archaeon sequences 2C25, 2MT8, and 2MT1, all within marine group III and isolated from sediments of a coastal salt marsh, and to unidentified archaeon CRA12 27 cm, isolated from continental shelf anoxic sediments (41). A series of clones from the spring origin (GSa2O24, GSa2O17, GSa3O3, and GSa2O15) were highly similar to sequences VAL2 to VAL68 obtained from the plankton of a boreal forest lake in Finland (28). A clone similar to these marine group III sequences (ZAR 101) was also obtained from a sulfide- and sulfur-rich spring in Oklahoma (15).

DISCUSSION

The GSS ecosystem clearly has unique attributes compared to surrounding environments. Unique aspects of spring origin water chemistry include substantially elevated concentrations of Ca²⁺, SO₄²⁻, Cl⁻, PO₄²⁻, and NH₄⁺ compared to those of Lake Erie. Moreover, the spring origin is at a constant 11°C, is anoxic with a wide array of reduced compounds (notably sulfide), and receives direct sunlight. The microbial community of GSS appears to be under strong selection by the water chemistry of the spring, resulting in a large assemblage of phototrophic bacteria, cyanobacteria, and algae, many with a notable similarity to organisms tolerant of sulfide or capable of H₂S-based phototrophy. A second community under selection by spring chemistry includes organisms capable of coupling the variety of reduced and oxidized chemical species in GSS water, as indicated by the similarity of GSS sequences to organisms in the Gamma-, Beta-, Epsilon-, and Deltaproteobacteria and Cyanobacteria taxa for which chemolithotrophy and/or sulfur metabolism is a major or the only documented means of energy generation. In virtually every case where a GSS sequence was more than 95% similar to that of a known organism, that genus is recognized for some form of sulfur metabolism or chemolithotrophy. Indicated metabolic capabilities include obligate lithotrophy using H₂S, H₂, or thiosulfate, sulfate reduction, and H₂ syntrophy. The suggested structuring of GSS bacterial sequences around sulfur metabolism is consistent with observations at other sulfidic springs. Sequences inferred to be from potential sulfur-metabolizing organisms constituted 56 to 75% of the high-Arctic saline sulfidic springs microbial community (47), 83% at Zodletone Spring (14), and >75% at Frasassi Cave (35). As many cyanobacteria also fix N₂, and reduced nitrogen is available in GSS spring origin water, the GSS bacterial community may be capable of fixing CO₂ and N₂ and completing nitrogen and sulfur cycles, based primarily or solely on chemolithotrophic metabolism. Finally, there is a heterotrophic community that is dependent on these primary producers for carbon and perhaps other essential nutrients. Such a tightly coupled community may be fostered by the relatively constant environmental conditions of the spring. Because of the spring hydrodynamics and its location, the system is essentially an autochthonous ecosystem, with few resources from outside. Under these conditions, a unique community has evolved and is maintained as a continuous culture that provides an unusual injection of chemistry and biology into the surrounding ecosystem.

The GSS microbial community may be structured in additional ways by water chemistry. GSS spring origin water has an ionic composition that includes SO₄²⁻, Ca²⁺, and HCO₃⁻ at elevated concentrations similar to those of seawater. The bedrock that now forms the Silurian-Devonian aquifer was of marine origin. Many of the sequences we detected have been previously reported only from marine environments, and the current study adds significantly to the diverse habitats and geographic locations in which uncultivated Archaea and Bacteria sequences have been detected. All the GSS Euryarchaeota clones fall into marine group III, and several GSS Crenarchaeota clones were most similar to marine sequences of marine group I. Likewise, we obtained GSS sequences similar to those of myxobacterium SMP-2 (Haliangium ochraceum), a halophile unable to grow in the absence of NaCl, and to those of deltaproteobacterium Sva0679, obtained from permanently cold, deep ocean sulfidic sediments. A new GSS clade of cyanobacterial sequences was most similar to an uncultivated sequence obtained from marine coral. Additionally, five Deltaproteobacteria sequences and several Beta- and Gammaproteobacteria sequences were most similar to sequences or isolates from marine settings. Although the abundance of marine affiliations may result in part from the extensive collections of marine sequences that have been the result of focused programs or surveys (10, 11, 45), the fact that these sequences occur both in marine settings and at GSS may also suggest the existence of similar habitats or common environmental conditions. Studies such as ours that sample habitats with unique characteristics can help to define common geochemical features associated with the distribution of these uncultured organisms and may ultimately assist in their cultivation.

Although the GSS community appeared highly structured overall, we nevertheless observed shifts in the percent composition of Epsilonproteobacteria, Deltaproteobacteria, Cyanobacteria, and Crenarchaeota spp. between the GSS spring origin and the mat (Table 2). These shifts may be due to physical or geochemical gradients, especially of oxygen and sulfide concentrations, that have been shown to influence the microbial community composition at other sulfidic springs. Higher sulfide and lower oxygen concentrations have been proposed to favor sulfide-oxidizing Epsilonproteobacteria over sulfide-oxidizing Gammaproteobacteria such as Thiothrix and to influence the relative distribution of different clades of Lower Kane Cave Epsilonproteobacteria groundwater groups I and II (18). Oxygen gradients reportedly influenced the relative distribution of Beggiatoa (preferring lower oxygen concentrations) and Thiothrix (preferring higher oxygen concentrations) species in Frasassi cave biofilms, and relatively high oxygen concentrations reportedly accounted for the relatively low percent contribution (0 to 15%) of epsilonproteobacteria to Frasassi biofilm clone libraries (35). Temperature and O₂ concentration were higher in shallow water surrounding the GSS mats than at the spring origin. Since we sampled detached mats floating on the pond surface, light exposure might also have been different than the conditions at depths near the spring origin. The more anoxic spring origin conditions could be a factor influencing the preponderance of epsilonproteobacteria and sulfate-reducing or syntrophic Deltaproteobacteria sequences in the spring origin clone library. Consistent with other sulfur springs where sulfide-rich spring water becomes oxygenated in outflows (for examples, see references 1, 6, 12, 18, and 42), Thiothrix relatives were detected only in the whitish, filamentous material from the GSS spring outflow. Cyanobacteria community composition shifted dramatically between the GSS spring origin and the mat (Table 2) (Fig. 7). The spring origin library was dominated by a group of 53 unique sequences for which the closest cultured relatives fall within the “unclassified” Cyanobacteria spp. In contrast, of the remaining 21 nonchloroplast sequences, 15 occurred only in the mat library. In other studies at sulfidic springs, cyanobacterial sulfide tolerance was correlated with specific habitat preferences, as defined by sulfide/oxygen gradients, temperature, light, and water flow rate (6, 14).

Crenarchaeota community composition at GSS also shifted from the spring origin to the mats, but the Euryarchaeota composition did not (Table 2). Unfortunately, most other studies of sulfidic springs that have also examined Archaea spp. (8, 15, 31, 47, 50, 51) do not provide good comparisons to GSS, because these focused only on mats or sediments and not source waters or because the geochemistry of these systems exhibited significant differences (salinity, hydrocarbon presence) from that of GSS. For example, both the high Arctic saline springs (47) and Zodletone Spring (15), where evaporation resulted in saline sediments, yielded euryarchaeal communities dominated by halophilic and methanogenic members not detected at GSS. Crenarchaeota sequences at GSS were more similar than Euryarchaeota sequences to those from other sulfidic springs. Soil, miscellaneous, and/or marine I groups were also represented at Zodletone Spring (15), Wind Cave (8), and the sulfidic springs in Germany and Turkey (31). The relative distribution of uncultivated Archaea sequences among different sites, and among habitats at sites such as GSS, may eventually inform cultivation methods.

One particularly interesting finding at GSS comes from a comparison to studies conducted on the whitish filamentous mats and streamers at the outflow of the sulfidic springs in Germany and Turkey (31, 39, 50, 51). Within these materials, common to many sulfidic spring outflows (12, 13) and also observed at the GSS drain, researchers observed uniquely structured consortia between Thiothrix spp. or the epsilonproteobacterium IMB1, the euryarchaeote SMI or the crenarchaeote Cre1, and an unidentified bacterial partner. The euryarchaeote SMI, the central partner in the string-of-pearls-like consortium (40), was not detected at GSS. However, of 10 Archaea sequences from the GSS drain, 5 crenarchaeote sequences (GSS Cren Grp 1) were in marine group I and broadly similar to the probable Cre1 sequence (AM055707), and of 7 Bacteria sequences at the drain, 4 were similar to those of Thiothrix spp. Additionally, one drain and three mat (GSS Cren Grp 3) sequences and GSa2O12 were broadly similar to another sequence (AM055703) in the miscellaneous Crenarchaeota species obtained from the same German spring materials. These findings suggest the possibility that these unique consortia, structured with close physical contact between the cellular partners but carrying out an unknown function, may also exist at GSS. Notably, the marine group I sequences from the German springs and from GSS are broadly similar to the only cultured nonextremophile crenarchaeon—the ubiquitous marine autotrophic ammonia-oxidizing organism “Candidatus Nitrosopumilus maritimus” (23)—and GSS Cren Grp 1 sequences are more highly similar to another sequence (DQ085102) from the same samples that yielded that isolate.

The microbial community composition and biogeochemistry of GSS may have significance on a larger scale than the spring boundaries. In previous studies of Silurian-Devonian aquifer geochemistry in Monroe County, groundwater chemistry of SO₄-dominated waters in the area could not be adequately predicted by geochemical models (43). Dissolution of carbonates by sulfuric acid produced by pyrite oxidation was proposed as a possible reaction (43), but the potential role of epsilonproteobacteria in sulfuric-acid speleogenesis had not at that time been proposed (19). If epsilonproteobacteria similar to those detected at the GSS spring origin also exist in the aquifer itself, sulfuric-acid speleogenesis may be an important microbially mediated limestone dissolution process contributing to the development of karst features in the area and influencing the variable occurrence (and hence, suitability for drinking) of H₂S in Monroe County groundwater. Additionally, plumes of high-conductivity, anoxic groundwater from the Silurian-Devonian aquifer, enriched in sulfate, sulfide, and nutrients and surrounded by whitish filamentous materials and variously colored microbial mats (referred to as biogeochemical hotspots), have been reported to occur near submerged sinkholes at 93-m depths in Lake Huron (2, 49). The present study of GSS hints at many types of biogeochemical functions for such groundwater seeps at depths in the Great Lakes.