Filamentous “Epsilonproteobacteria” Dominate Microbial Mats from Sulfidic Cave Springs

Abstract

Hydrogen sulfide-rich groundwater discharges from springs into Lower Kane Cave, Wyoming, where microbial mats dominated by filamentous morphotypes are found. The full-cycle rRNA approach, including 16S rRNA gene retrieval and fluorescence in situ hybridization (FISH), was used to identify these filaments. The majority of the obtained 16S rRNA gene clones from the mats were affiliated with the “Epsilonproteobacteria” and formed two distinct clusters, designated LKC group I and LKC group II, within this class. Group I was closely related to uncultured environmental clones from petroleum-contaminated groundwater, sulfidic springs, and sulfidic caves (97 to 99% sequence similarity), while group II formed a novel clade moderately related to deep-sea hydrothermal vent symbionts (90 to 94% sequence similarity). FISH with newly designed probes for both groups specifically stained filamentous bacteria within the mats. FISH-based quantification of the two filament groups in six different microbial mat samples from Lower Kane Cave showed that LKC group II dominated five of the six mat communities. This study further expands our perceptions of the diversity and geographic distribution of “Epsilonproteobacteria” in extreme environments and demonstrates their biogeochemical importance in subterranean ecosystems.

Caves containing hydrogen sulfide-rich springs represent less than 10% of all known caves globally (42). However, these caves serve as access points into sulfidic groundwater aquifers, typically associated with geothermal regions and oil-field basins, which play an important role in global sulfur cycling. The microbial communities colonizing sulfidic cave habitats have recently received attention due to their chemolithoautotrophic metabolism that can sustain complex ecosystems in the subsurface (48) and their geomicrobiological impact due to acid production (14, 60). Filamentous bacteria dominate subaqueous cave microbial mats, and from phylogenetic analyses, stable isotope evidence, and aqueous geochemistry surveys, populations are considered to be chemolithoautotrophic, aerobic to microaerophilic sulfur-oxidizing bacteria (5, 19, 48, 59). Recently, two cultivation-independent studies based on the phylogeny of bacterial community 16S rRNA genes characterized filamentous microbial mats from the Sulphur River of Parker Cave, Kentucky, and Cesspool Cave, Virginia (5, 14). In both 16S rRNA gene libraries, most clones were affiliated with uncharacterized environmental groups within the “Epsilonproteobacteria,” but the actual abundance and the morphotype(s) of the respective organisms were not established. Additionally, cultivation of any of these filamentous bacteria from sulfidic cave mats has been unsuccessful to date. The only sulfur bacteria isolated from these caves are gram-negative, rod-shaped thiobacilli (14, 60).

The most commonly studied genera of the “Epsilonproteobacteria,” Helicobacter and Campylobacter, are often associated with the gastrointestinal tract of animals as pathogens (13, 40). Other major phylogenetic groups within the “Epsilonproteobacteria” include Arcobacter, Wolinella, Sulfurospirillum, and Thiovulum, commonly found in natural settings as living cells or in symbiotic association with animals (13, 40, 51). Recently, members of the Arcobacter were also detected in activated sludge from wastewater treatment plants (53). Members of the Thiovulum phylogenetic group have been described from many natural habitats, some of which are considered extreme environments, including caves (5, 14), springs (47), groundwater associated with oil (17, 22, 64), marine water and muds (29, 58), deep-sea hydrothermal vent sites (9, 26, 36, 45), and vent-associated metazoans (8, 27). However, relatively little is known about the ecology or physiology of most of these “Epsilonproteobacteria,” as cultured representatives for most of the detected environmental clone groups are missing. Moreover, there are few detailed studies describing the occurrence of “Epsilonproteobacteria” from terrestrial environments compared to the relatively large number of such investigations from marine habitats.

In this study we investigated bacterial communities from filamentous microbial mats associated with aphotic sulfidic springs in Lower Kane Cave, a system located in north-central Wyoming. Using the full-cycle rRNA approach, including the construction of 16S rRNA gene clone libraries and quantitative fluorescence in situ hybridization (FISH), we report on the occurrence of two distinct epsilonproteobacterial filament groups within the microbial mats in the cave. Quantitative FISH with the newly designed 16S rRNA-targeted oligonucleotide probes for both groups revealed that members of one of the groups dominated the mat communities analyzed.

MATERIALS AND METHODS

Study site and sample acquisitions.

Lower Kane Cave is located in the north-central portion of the Bighorn Basin on the western flank of the Bighorn Mountains (12). It is forming within the Madison Limestone. Two major and two minor anaerobic, hydrogen sulfide-bearing springs discharge into Lower Kane Cave along a fracture zone (Fig. 1). Each spring is associated with an orifice pool and outflow channel. Groundwater pH, temperature, dissolved oxygen, and conductivity were measured in the field with specific probes. Dissolved hydrogen sulfide and low concentrations of dissolved oxygen were measured directly in the field with the methylene blue and rhodazine D colorimetric methods, respectively, following the manufacturer's guidelines (Chemetrics Inc., Calverton, Va.).

FIG. 1.

Location of Lower Kane Cave, near Lovell, Wyo. The inset plan-view cave map, modified from Egemeier (12), shows the cave entrance and three spring sites. The Wyoming map is from http://fermi.jhuapl.edu/states/.

Samples for DNA extraction were obtained from microbial mats associated with three of the springs, and samples were taken on several occasions during 2000 and 2001. We report here the clone library results from a subset of six samples from Lower Kane Cave (Fig. 1): from August 2000, Fissure Spring orifice white filament bundles (sample 21), Upper Spring orifice white filament bundles (samples 57 and 58); from March 2001, Upper Spring orifice white filament bundles (sample 114); from August 2001, Upper Spring thin white filaments (sample 190), Lower Spring orifice white filament bundles (sample 199), and Lower Spring yellowish-white mat (sample 198). Mat samples used for FISH were collected in December 2001 from the Fissure Spring orifice, Lower Spring orifice, the white mat from the Lower Spring, Upper Spring orifice, and three white filamentous mat samples from the Upper Spring stream channel (white mat 1, 2, and 3). Samples white mat 1 and white mat 2 were separated by 4 m of stream flow; white mat 2 and white mat 3 were separated by 5 m.

DNA extraction, PCR amplification, and cloning.

Approximately 0.2 to 0.5 ml of microbial mat was aseptically collected in the cave and transferred into DNA extraction buffer containing 10 mM Tris-HCl (pH 8), 100 mM EDTA, and 2% sodium dodecyl sulfate. Total community DNA was isolated with an extraction protocol similar to the commercially available Purgene DNA extraction kits (Gentra Systems, Minneapolis, Minn.), with the following modifications: 9 μl of proteinase K (20 mg/ml) was added to each DNA extraction buffer prior to digestion; a freeze-thaw (three times at −80°C to 65°C) series was used to aid in the physical disruption of the mat structure; samples were incubated at 55°C overnight to digest cellular material; RNase was added to the digests and incubated at 37°C for up to 1 h; proteins were precipitated in 10 M ammonium acetate; and nucleic acids were precipitated in isopropanol overnight at −20°C and washed in 70% ethanol.

Nearly full-length 16S rRNA gene sequences were PCR amplified with the primer pairs 27f (forward, 5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492r (reverse, 5′-GGTTACCTTGTTACGACTT-3′), according to the protocol described by Lane et al. (23). Amplification was performed with a Perkin Elmer 9700 thermal cycler under the following conditions: denaturation at 95°C for 1 min, primer annealing at 42°C for 1 min, and chain extension at 72°C for 1.5 min. Fifty PCR cycles were used. A control tube containing sterile water instead of DNA was used as a negative control.

Amplified PCR products were purified with the GeneClean II kit (Bio101, Inc., Vista, Calif.), as recommended by the manufacturer. Purified PCR products were cloned with the Topo TA cloning kit (Invitrogen, San Diego, Calif.), following the manufacturer's instructions.

Sequencing of 16S rRNA genes and phylogenetic analysis.

Clones to be analyzed were lysed in 50 μl of buffer (10 mM Tris-HCl, 0.1 mM EDTA, pH 8.0) for 10 min at 96°C. Forty-four clones were selected for sequencing, whereby sequence inserts were PCR amplified from lysed cells with plasmid-specific primer pairs M13(−20) (5′-GTAAAACGACGGCCAGT-3′) and M13(−24) (5′-AACAGCTATGACCATG-3′) and the following PCR conditions: denaturation at 94°C for 1 min, primer annealing at 55°C for 1 min, and chain extension at 72°C for 3 min, for 35 cycles. PCR products were purified with Sephadex columns and sequenced with an ABI Big-Dye Ready Reaction kit (Perkin Elmer) with primers 27f and 1492r in conjunction with internal primers 907r (reverse, 5′-CCGTCAATTCCTTTRAGTTT-3′ [where R is G or A]) and 704f (forward, 5′-GTAGCGGTGAAATGCGTAGA-3′). Automated DNA sequencing was done on an ABI Prism 377XL sequencer (Perkin Elmer).

DNA sequences were submitted to the CHECK-CHIMERA program of the Ribosomal Database Project II at Michigan State University (http://rdp.cme.msu.edu/html/) (30). Clone sequences were subjected to BLAST searches within the GenBank database (http://www.ncbi.nlm.nih.gov/) to determine 16S rRNA gene sequence similarities to culturable and not yet cultured organisms.

The retrieved nucleotide sequences were initially aligned with Clustal X (56) and then manually adjusted based on conserved primary and secondary configuration. Phylogenetic analyses were done with maximum likelihood, minimum evolution, and maximum parsimony criteria in PAUP* (55) and Bayesian inference (20). For minimum evolution, Bayesian inference, and maximum likelihood searches, a model of evolution was chosen based on likelihood ratio tests (20), as implemented in Modeltest 3.06 (44). Heuristic searches were run for either 100 (maximum likelihood), 500 (minimum evolution), or 1,000 (maximum parsimony) replicates with random sequence addition and the tree-bisection and reconnection branch-swapping algorithm. Bayesian inference searches were run for 10⁶ cycles at least two times to check for convergence and then combined, deleting the first 1,000 trees from each search. As an indication of nodal support, bootstrap analyses were performed for maximum likelihood (100 replicates), minimum evolution (500 replicates), and maximum parsimony (1,000 replicates) criteria with full heuristic searches. For Bayesian inference analyses, posterior probabilities were used for nodal support. All trees were rooted with Desulfocapsa thiozymogenes (X95181) and Hydrogenophaga pseudoflava (AF078770) as outgroups.

16S rRNA oligonucleotide probes.

Oligonucleotide probes specific for two epsilonproteobacterial clone groups from Lower Kane Cave (group I and group II) were designed with the PROBE DESIGN tools from the ARB software package (http://www.arb-home.de) and probe designations according to Alm et al. (3). Probe LKC59 (S-*-eProt-0059-a-A-18) is specific for group I clones, and probe LKC1006 (S-*-eProt-1006-a-A-18) targets group II clones. Probe specificity was verified with the RDP PROBE-MATCH function (30) and the PROBE-MATCH tool of the ARB software package, and a 16S rRNA data set including all publicly available sequences from “Epsilonproteobacteria.” Sequences and optimal hybridization conditions for probes LKC59, LKC1006, and all other probes used are listed in Table 1, including probe EPS710, designed to target environmentally retrieved clones within the “Epsilonproteobacteria” (64) and the cultured strain YK-1 proposed as Sulfuricurvum kujiense gen. nov., sp. nov. (22; Y. Kodama and K. Watanabe, presented at the International Symposium on Subsurface Microbiology, Copenhagen, Denmark, 2002). More detailed information about the probes can be found at probeBase (http://www.microbial-ecology.net/probeBase) (28).

TABLE 1.

Probe sequences used to screen cave mat microbial populations

Probe	Target group	Probe sequence (5′ → 3′)	Target sitea	FAb (%)	Reference
EUB338	Eubacteria	GCT GCC TCC CGT AGG AGT	16S (338)	0-40	10
EUB338-II	Planctomycetes	GCA GCC ACC CGT AGG TGT	16S (338)	0-40	10
EUB338-III	Verrucomicrobia (and others)	GCT GCC ACC CGT AGG TGT	16S (338)	0-40	10
NonEUB	Negative control	ACT CCT ACG GGA GGC AGC	16S (338)	0	62
ALF968	Alphaproteobacteria	GGT AAG GTT CTG CGC GTT	16S (968)	20	38
BET42a	Betaproteobacteria	GCC TTC CCA CTT CGT TT	23S (1027)	35	32
GAM42a	Gammaproteobacteria	GCC TTC CCA CAT CGT TT	23S (1027)	35	32
HGC69a	Actinobacteria	TAT AGT TAC CAC CGC CGT	23S (1901)	25	46
LGC345A	Many Firmicutes (together with two other LGC345A probes)	TGG AAG ATT CCC TAC TGC	16S (354)	20	33
LGC354B	Same as LGC345A	CGG AAG ATT CCC TAC TGC		20	33
LGC354C	Same as LGC345A	CGG CGT CGC TGC GTC AGG		20	33
CF319a	Some members of the “Flavobacteria”	TGG TCC GTG TCT CAG TAC	16S (319)	35	31
G123Tc	Thiothrix	CCT TCC GAT CTC TAT GCA	16S (697)	40	21
EPS710	“Epsilonproteobacteria”-Thiovulum groundwater subgroup	CAG TAT CAT CCC AGC AGA	16S (710)	30d	64
LKC59	Epsilonproteobacterial group I from Lower Kane Cave	TCC TCT CAT CGT TCG ACT	16S (59)	30	This study
LKC1006	Epsilonproteobacterial group II from Lower Kane Cave	CTC CAA TGT TTC CAT CGG	16S (1006)	30	This study

^aE. coli 16S rRNA position (6).

^bFormamide (FA) percentage in the FISH hybridization buffer.

^cUsed in conjunction with a competitor probe, G123T-C (5′-CCTTCCGATCTCTACGCA-3′) (21).

^dThis formamide concentration differs from the one suggested in the original publication (64) because we applied a different hybridization temperature.

Sample fixation, FISH, microscopy, and quantification.

For FISH, microbial mat samples were collected in December 2001, shipped on dry ice, and fixed in two ways within 48 h of collection: (i) with 4% (wt/vol) paraformaldehyde for 3 h as described by Manz et al. (32), and (ii) with 50% ice-cold ethanol according to Roller et al. (46). Fixed sample material was spotted onto Teflon-coated slides and air-dried overnight before dehydration by sequential washes in 50, 80, and 100% (vol/vol) ethanol for 3 min each.

The oligonucleotide probes were synthesized and directly labeled with the monofunctional, hydrophilic, sulfoindocyanine dyes indocarbocyanine (Cy3) and indodicarbocyanine (Cy5) or with FluosPrime (5,6-carboxyfluorescein-N-hydroxysuccinimide ester), purchased from Hybaid-Interactiva (Ulm, Germany). Hybridization and washes were performed as described by Manz et al. (32). The salt concentration in the wash buffer was adjusted to the formamide concentration in the hybridization buffer according to Manz et al. (32). Washes were performed at 48°C.

Optimal hybridization stringency for probes EPS710, LKC59, and LKC1006 was determined by increasing the formamide concentration of the hybridization buffer in increments of 5 or 10% while maintaining a constant hybridization temperature of 46°C. Due to the nonavailability of suitable reference cells, the optimal hybridization stringency for the three probes evaluated was defined by the highest stringency allowing unambiguous visual detection of probe target cells in fixed samples of white filamentous microbial mats from two different sampling locations (Upper Spring white mat 1 and Upper Spring white mat 3; Fig. 1).

To determine the percentage of all cells detected with the bacterial probe set, EUB338I-III mix-hybridized mat samples were additionally stained with 10 μl of a 10,000-fold-diluted SYBR Green I (FMS Bioproducts, Rockland, Maine) working solution in the dark for 10 min at room temperature. Slides were then washed briefly with double-distilled H₂O and air-dried. Before examination, samples were covered with the antifading agent Citifluor AF1 (Chemical Laboratory, Caterbury, England). An LSM510 scanning confocal microscope (Zeiss, Oberkochen, Germany) equipped with an Ar ion laser (450 to 514 nm) and two HeNe lasers (543 and 633 nm) was used to visualize FISH results. All images were recorded with a Plan-Apochromat 63x (1.4; oil immersion) objective. Image processing was performed with the LSM510 software package (version 1.6). Quantification of probe-detected cells was achieved with the Carl Zeiss Vision KS400 software package in conjunction with the R.A.M (Relative Area Measurement) macro, as described by Schmid et al. (49).

Nucleotide sequence accession numbers.

The 16S rRNA gene sequences determined in this study have been submitted to GenBank with accession numbers from AY191466 to AY191497.

RESULTS

Mat distribution and structure.

Water coming into the cave had no detectable dissolved oxygen. Dissolved sulfide averaged 32 μmol liter⁻¹ in the three orifice pools and decreased to nondetectable levels beyond the terminus of the mats. Along the Upper Spring outflow channel (Fig. 2a), dissolved oxygen concentrations progressively increased to 45 μmol liter⁻¹, corresponding to a decrease in dissolved sulfide. Cave water had a temperature of ≈22°C and pH 7.2. All spring orifices had long white filament bundles suspended in discharging water. Some filamentous structures stretched up to 1 m in length. An associated microbial mat formed in spring outflow channels, having an average thickness of 5 cm in water 8 to 10 cm deep (Fig. 2b). During all the sampling periods, the mats in the Fissure Spring and Upper Spring streams were nearly 20 m in total length, while the Lower Spring extended for only a meter (Fig. 1).

FIG. 2.

(a) Photograph of cave passage showing microbial mats growing in sulfidic stream channel formed downstream of the Upper Spring orifice in Lower Kane Cave (light area at lower right). The stake in the center of the view is approximately 25 cm high. (b) Filamentous microbial mats in cave stream. Scale bar, 10 cm.

Phylogenetic analysis of clone sequences.

From six samples, 44 clones were randomly selected for sequencing in order to conduct a broad survey of the microorganisms present in mat communities. Nearly full-length 16S rRNA gene fragments from the clones were amplified and sequenced. None of the sequences were identified as chimeras from RDP analysis. All clone sequences belonged to the Proteobacteria phylum, with 85% of the clones affiliated with the “Epsilonproteobacteria,” 11% of the clones belonged to the Betaproteobacteria, and 4% of the clones clustered with the Gammaproteobacteria.

Within the “Epsilonproteobacteria,” the Lower Kane Cave clones clustered into two different clades (Fig. 3), referred to as LKC group I and LKC group II, with high bootstrap values supporting their phylogenetic position. The closest relatives to both LKC group I and group II were two environmental clones, sipK119 and sipK94, obtained from microbial aggregates with a string-of-pearls-like morphology in sulfidic springs at the Sippenauer Moor, Regensburg, Germany (35, 47). Group I clones, identified from Fissure and Upper Spring orifice samples, clustered closely (98 to 99% similar in nucleotide sequence) with clone sipK119 and Cesspool Cave clone CC-4 (14). The closest cultured representative of LKC group I sequences is Sulfuricurvum kujiense (22). LKC group II clones, obtained from Upper and Lower Spring orifice and white mat samples, were closely related to the sipK94 clone (99% similar in nucleotide sequence), and more distantly to various marine and hydrothermal vent clones (91 to 94% similarity), as well as to epibionts of the polychaetous annelid Alvinella pompejana (89% similarity) (18), ectosymbionts of the shrimp Rimicaris exoculata (90% similarity) (43), and Parker Cave clones (92 to 94% identical; SRang names on Fig. 3) (5).

FIG. 3.

16S rRNA gene-based phylogenetic tree showing the phylogenetic position of 32 clones from Lower Kane Cave (designated LKC1 and having sequences >1,120 nucleotides long) within the “Epsilonproteobacteria.” The topology of the tree was inferred from the results of the maximum-likelihood analysis, and the phylogenetic affiliations of the LKC clones were confirmed by comparison with different reconstruction methods (data not shown). Clones are labeled in bold with corresponding sample and clone numbers. Reference sequences (with GenBank accession numbers) were chosen from the RDP to represent the diversity of “Epsilonproteobacteria” members and specifically the Thiovulum phylogenetic group. The specificity of the “Epsilonproteobacteria” FISH probes applied in this study is shown. The tree was rooted with the sequences of Desulfocapsa thiozymogenes and Hydrogenophaga pseudoflava, shown as an arrow labeled “to outgroups.” Numbers at branch intersections refer to bootstrap values for each node from maximum likelihood, maximum parsimony, minimum evolution, and Bayesian inference posterior probabilities (values below 50% or where only one method supports a node are not shown).

Fluorescence in situ hybridization.

FISH probes were used to identify and to quantify specific microorganisms in natural white filamentous microbial mat samples from Lower Kane Cave. For each of the two LKC clone groups clustering with the “Epsilonproteobacteria,” a specific probe was designed. Probes LKC59 and LKC1006 targeted clone sequences from group I and group II, respectively (Table 2). Although LKC group I is closely related to other environmental clones from groundwater and caves (Fig. 3), probe LKC59 has at least one mismatch with these and all other rRNA gene sequences in the database. Probe LKC1006 also did not target any other sequences in the database, including clone sipK94 which has 99% 16S rRNA gene sequence similarity with LKC clone group II. However, it should be mentioned that three of the LKC group II clones possess a single mismatch within the target site of probe LKC1006 and might not be detectable by this probe (Table 2). These mismatches either indicate actual genetic microheterogeneity or originate from PCR and/or sequencing artifacts.

TABLE 2.

Difference alignment of the target region of the 16S rRNA for LKC-specific probes

Probe and target	Target sequencea
LKC59 probe sequence (5′-3′)	TCCTCTCATCGTTCGACT
Target sequence	AGUCGAACGAUGAGAGGA
LKC group I clonesb	------------------
Uncultured Cesspool Cave clone CC-4 (AF207530)	-------------U----
Uncultured groundwater clone 1023 (AB030610)	-------------U----
LKC1006 probe sequence (5′-3′)	CTCCAATGTTTCCATCGG
Target sequence	CCGAUGGAAACAUUGGAG
Most LKC group II clonesc	------------------
LKC1.114_5 (AY191480)	---G--------------
LKC1.199_5 (AY191494)	-U----------------
LKC1.199_6 (AY191495)	---------U--------

^a—, identical to the probe sequence.

^bBased on nine LKC group I clones with 16S rRNA gene sequences at this position.

^cBased on 15 LKC group II clones.

Optimal hybridization stringency was determined for the newly designed probes LKC59 and LKC1006, as well as for the previously published probe EPS710 (64) that targets environmental clones within the “Epsilonproteobacteria” and strain YK-1 (22). Since no cultured strains possessing the target sites for probes LKC59 and LKC1006 are available, optimal hybridization stringency was inferred by visual comparison of filament fluorescence from cave samples. For all three probes, mat filaments showed bright fluorescence if hybridization buffers with up to 30% formamide were used. At more stringent conditions, signal intensity decreased sharply (results not shown).

In the six samples analyzed, between 68% and 88% of the cells stainable with a general nucleic acid dye could be detected with the bacterial probe set (Table 3). Of the nine different group- and genus-specific probes applied (Table 1), positive hybridization signals were only observed with the probes BET42a, GAM42a, and G123T. Probe BET42a labeled small rods specifically but weakly in all the mat samples, while long filamentous cells hybridized strongly with probes GAM42a and G123T, indicating the presence of Thiothrix spp. in some samples.

TABLE 3.

Quantification of epsilonproteobacterial filaments of LKC groups I and II with specific FISH probes^a

Group	Relative biovolume, % (SE)
	Fissure Spring orifice	Upper Springb			Lower Spring
		White mat 1	White mat 2	White mat 3	Orifice	White mat
EUB338I-III mix/SYBR Green I ratio	88 (2)	87 (3)	77 (3)	80 (3)	74 (1)	68 (4)
LKC group I/EUB338I-III mix	<8c (2)	<10c (1)	<1	<1	2 (1)	<1
LKC group II/EUB338I-III mix	64 (5)	67 (4)	70 (2)	57 (2)	50 (1)	4 (1)

^aRelative biovolumes are given as percentages, and the number in parentheses is the standard error.

^bWhite mat 1 and white mat 2 were separated by 4 m, and white mat 2 and white mat 3 were separated by 5 m.

^cThis value is most likely overestimated due to relatively weak filament signal intensity and high background.

The three probes targeting subgroups within the “Epsilonproteobacteria” were used in different combinations. All probes exclusively hybridized to filamentous microbes and conferred very bright signals to their target cells, indicating high rRNA contents (4). Filaments detected by probe LKC59 had an average diameter of 1 μm and appeared kinked or twisted, while straighter, longer, and slightly thicker filaments were detected by LKC1006 (Fig. 4). Simultaneous hybridization with the epsilonproteobacterial probes LKC59 and EPS710 showed that they stained the same filamentous cell morphotypes (Fig. 4, row I), and as expected, neither probe EPS710 nor probe LKC59 overlapped with probe LKC1006 (Fig. 4, rows II and III).

FIG. 4.

Fluorescence in situ hybridization of Lower Kane Cave microbial mat samples with probes EUB338I-III mix, newly designed probes LKC59 and LKC1006, and probe EPS710. Rows: I, Fissure Spring orifice filaments; II, Upper Spring white mat 1; III, Upper Spring white mat 2. Columns: A, EUB338I-III mix (labeled with FluosPrime, colored in green); B and C, epsilonproteobacterial LKC group probe or EPS710 probe (labeled with Cy3 and colored in red, or labeled in Cy5 and colored in light blue); D, overlap of columns B and C. If Cy3- and Cy5-labeled probes overlap, the filaments appear pink. Scale bar, 20 μm.

Epsilonproteobacterial filaments belonging to LKC group II dominated five of the six microbial mats examined and made up to 70% of the biovolume of those cells detectable by FISH with the bacterial probe set (Table 3). Only in the Lower Spring white mats, which were dominated by Thiothrix spp. (data not shown), epsilonproteobacterial filaments occurred at relatively low numbers (4% of the bacterial biovolume). In contrast, epsilonproteobacterial filaments of LKC group I were below the detection limit in three of the six samples investigated, and made up less than 10% of the bacterial biovolume in the other three samples.

DISCUSSION

Defining the composition of microbial communities can aid in our understanding of biogeochemical cycling that occurs in remote and difficult-to-characterize habitats. Particularly for the terrestrial subsurface, however, microbial community structures are poorly understood due to a limited number of investigations done on such systems. In this study we applied the full-cycle rRNA approach to characterize microbial mats from sulfidic cave springs and to assess dominant microbial populations integral to biogeochemical cycling in this system.

The majority of 16S rRNA gene clones were assigned to two evolutionary lineages within the “Epsilonproteobacteria,” designated LKC groups I and II, neither of which possesses closely related cultured representatives. However, each of the clone groups also contains a molecular clone, either sipK119 or sipK94 (Fig. 3), from microbial communities growing in a string-of-pearls-like morphology (35, 47). Several 16S rRNA sequences retrieved from groundwater at an underground petroleum storage cavity (64) and S. kujiense, an anaerobic sulfur oxidizer recently isolated from this habitat (22), cluster with LKC group I. Two clone groups (CC-4 and CC-9) from the small sulfidic cave, Cesspool Cave (14), are also closely related to group I (Fig. 3). In contrast, Lower Kane Cave group II forms a monophyletic grouping with the 16S rRNA gene sequence of an uncultured filamentous epibiont associated with the vent annelid Alvinella pompejana (7), while clone groups from phylogenetic studies of Parker Cave, another sulfidic cave system, are found in sister clades (5) (Fig. 3).

Quantitative FISH analysis with two newly developed 16S rRNA-targeted oligonucleotide probes specific for LKC group I and II revealed that both groups are filamentous bacteria. Filaments for LKC group I were detected in rather low numbers in three of the six mats analyzed and were below detection in other mat samples. In contrast, filaments of LKC group II dominated five of the six mats and represented 50 to 70% of the bacterial biovolume in these communities. Bright FISH signals observed for both filament groups suggest that these microorganisms were physiologically active in the mats.

While high in situ abundance of free-living and eukaryote-associated “Epsilonproteobacteria” have been described from many marine environments, including hydrothermal vent communities (1, 7-9, 18, 25-27, 29, 36, 37, 43, 45, 58), there is generally no evidence available that members of the “Epsilonproteobacteria” are also numerically important in terrestrial systems. The only exceptions currently known are from two engineered systems and sulfidic cave microbial mats in which “Epsilonproteobacteria” exist in significant numbers. Watanabe et al. (64) report between 12 and 24% of the total prokaryotic cells in petroleum-contaminated groundwater could be identified via FISH as epsilonproteobacterial curved rods, and relatively low abundances of Arcobacter (4%) were identified from activated sludge in municipal wastewater treatment plant with FISH (53).

Epsilonproteobacterial abundance estimates from 16S rRNA clone libraries in Parker and Cesspool Caves, at 73% and 47% “Epsilonprotoebacteria,” respectively, are similar to the FISH biovolume values for LKC group I and group II (5, 14). Although the microbial aggregates in a string-of-pearls-like morphology harbor “Epsilonproteobacteria” closely related to those in Lower Kane Cave, FISH analysis revealed that the string-of-pearls are dominated not by “Epsilonproteobacteria,” but by Thiothrix spp. filaments and novel Archaea (35, 47). Consequently, the mats from the sulfidic springs in Lower Kane Cave represent the first nonmarine natural system that is demonstrably driven by the activity of “Epsilonproteobacteria.”

Since the filamentous “Epsilonproteobacteria” dominating the mats from Lower Kane Cave have not yet been cultured, we can only speculate about the ecophysiology of these microorganisms. However, because most microbial mats in sulfide-containing marine and freshwater environments are dominated by sulfur bacteria that form the bulk of the mats as filaments or webs and veils, including Beggiatoa (39), Thiothrix (35), Thiomargarita (50), Thioploca (41), and Thiovulum (15, 36, 52), it is tempting to suggest that the two cave filamentous groups are associated with sulfur cycling. Consistent with this hypothesis, all “Epsilonproteobacteria” presently isolated are associated with sulfur metabolism, specifically the oxidation of reduced sulfur compounds with oxygen (27, 29, 43), nitrate (17, 22, 57), various sulfur species (16), or metals (51, 54) as electron acceptors. Some epsilonproteobacterial strains also grow anaerobically by reducing sulfur to sulfide (2, 7, 34).

In Lower Kane Cave the two groups of “Epsilonproteobacteria” may be associated with anaerobic sulfur oxidation, as incoming sulfidic groundwater is anoxic and dissolved oxygen concentrations were generally low along the stream channel where the mats occurred. LKC group I had highest abundances in orifice filament bundles and proximal mats, and the phylogenetic affiliation of LKC group I to anaerobic groundwater clones (63, 64) and S. kujiense (22) is consistent with anaerobic sulfur oxidation. However, while the availability of alternative electron acceptors, particularly sulfate, in the anaerobic to disaerobic water in Lower Kane Cave is high, the concentration of nitrate (the electron acceptor used by S. kujiense) is generally low or below detection (P. C. Bennett, unpublished results). LKC group II filaments dominated all but one mat sample and may also oxidize sulfur compounds anaerobically. Interestingly, the abundance of LKC group II filaments in the mats decreased with increasing distance from the orifice (Table 3), which might suggest that these microorganisms prefer more anaerobic and microaerophilic conditions. At elevated dissolved oxygen concentrations, LKC group II may be outcompeted by Thiothrix (Table 3; white mats from Lower Spring), a typical aerobic, sulfur-oxidizing bacterium frequently found in freshwater habitats and activated sludge (21, 52, 61).

In addition to sulfur cycling, carbon fixation by chemolithoautotrophic microorganisms is of major importance for the microbial mat communities in Lower Kane Cave because chemolithoautotrophs would provide an energy source for their respective ecosystem. The phylogenetic relatives to the LKC clones, such as isolates of A. pompejana epibionts (7), T. denitrificans (57), and S. kujiense (22), grow as autotrophs. Because the concentration of available dissolved organic carbon is low (<80 μmol liter⁻¹, including dissolved methane gas) in the incoming spring water, the filamentous “Epsilonproteobacteria” in Lower Kane Cave are presumably also chemolithoautotrophs. Future characterization of the filamentous populations with stable and radiolabeled isotopes and culturing work will address this issue, for instance by combining fluorescence in situ hybridization with microautoradiography to elucidate both sulfur and carbon metabolism (24).

The ecological significance of “Epsilonproteobacteria” is becoming more apparent, as these microorganisms are found associated with cycling carbon and sulfur compounds in extreme environments, including the sulfidic and oligotrophic groundwater in Lower Kane Cave. In conclusion, this study expands the known diversity of “Epsilonproteobacteria” in the terrestrial subsurface and provides information about the distribution of these microbes relative to habitat geochemistry, which can be used to aid future cultivation attempts of these environmentally important bacteria.