Caves are sometimes associated with geothermally warmed subsurface waters, which can accelerate cave formation by increased calcium carbonate dissolution (26). Hot springs within cave passages fill the cave with geothermally warmed gases and steam to produce “vapor caves.” The vapor/aerosol can be rich not only in water but also in gases such as H₂ and H₂S, which in combination with mineral surfaces are potential energy sources to promote lithotrophic microbial life (9, 19, 20). Such microbial life can be evidenced by “slime” on cave walls, discoloration of cave limestone matrix, powderization of the cave matrix, and fabric-like speleothem formations (6). Microbial growth on cave walls is evident in both sulfidic and nonsulfidic caves (6, 15, 36, 39-42).

We report here an analysis of novel biofabrics, complex associations of microbes in a weave of mineralization, associated with two kinds of speleothems from a geothermally heated 50°C mine adit near Glenwood Springs, on Colorado's western slope. The walls and ceilings of the adit have patches of a slimy microbial mat, from which descend short (2- to 3-cm) leather-like stalactites. Floating in pools on the floor of the adit are dumpling-like “cave rafts” of various sizes and shapes, which can be as thick as 10 cm. Both of these features presumably formed from mineralogical precipitation of calcite and/or gypsum, but a biological role in the process seemed possible, and preliminary microscopy indicated the occurrence of conspicuous microbiological activity in both structures. Little is known about the kinds of organisms that are associated with the mineralization processes within caves. Consequently, we sought to explore the biological component of these novel speleothems.

Sample collection. The mine adit lies approximately 3 m below the surface; it is 50 m in length and is filled with 49.6°C water in spring-fed pools. The source of this hot water is the Leadville aquifer, which is responsible for the water chemistry observed: pH 6.4, 1.65 mg/liter dissolved hydrogen sulfide, and 2.5 mg/liter dissolved oxygen (44% saturation at this temperature and atmospheric pressure) (8). Due to reduced oxygen, at 7.5% atmospheric oxygen, limited visibility from the high vapor/fog content, an air temperature of 49.6°C, and difficulties associated with walking/standing in hot water, sampling trips were short in duration. In August 2004, sampling trips were made to retrieve two “stalactites,” a sample of the ceiling microbial mat, and a sample of a floating “dumpling” for phylogenetic and microscopic analysis. Samples for phylogenetic analysis were transported back to the lab at 4°C (<4 h) and stored at −80°C. Samples for microscopic analysis were placed in 4% paraformaldehyde-phosphate-buffered saline at room temperature for 2 h, washed in 1× phosphate-buffered saline, and stored in 50% ethanol at 4°C.

Molecular phylogenetic analyses. Total community DNA was obtained and prepared from ~1 g of the frozen samples as previously described (18, 45), with an average yield of 600 ng/μl in 30 μl of extract. Initial PCR amplifications were performed with the universal small-subunit rRNA degenerate primers 515F (5′GTGCCAGCMGCCGCGGTAA3′) and 1391R (5′GACGGGCGGTGWGTRCA3′) (29). PCRs were carried out in an Eppendorf gradient thermocycler at 94°C for 2 min followed by 29 cycles at 94°C for 30 s, 55.5°C for 1 min, and 72°C for 1.5 min followed by a 72°C step for 12 min. Each 25-μl reaction mixture contained 30 mM Tris (pH 8.4), 50 mM KCl, 1.5 mM MgCl₂, 0.05% bovine serum albumin (Sigma), 1 M betaine, 0.2 mM of each deoxynucleotide triphosphate, 50 ng of each oligonucleotide primer, 0.5 units of Taq polymerase, and approximately 600 ng of environmental DNA template. PCR products were gel purified using a QIAGEN gel extraction kit (QIAGEN, Inc., Valencia, CA) as per the kit instructions. Before cloning, 30 μl of gel-purified PCR product was incubated with 2 μl of deoxynucleotide triphosphates for 20 min at 72°C in 1× buffer, 1 U of Taq polymerase, and 1.5 mM MgCl₂. Reactions were cloned with the TOPO TA cloning kit for sequencing (Invitrogen, Carlsbad, CA), and unique restriction fragment length polymorphisms were sequenced with the DYEnamic ET Terminator cycle sequencing kit (Amersham Biosciences, Piscataway, NJ). Sequences were assembled with XplorSeq 1.0 (in-house software by Daniel Frank), and an initial identification was made via a BLAST search (www.ncbi.nlm.nih.gov/BLAST) (1). Clone libraries were compared in a phylogenetic context via the UniFrac method of Lozupone and Knight (31).

The software programs Mallard (2) and Bellerophon (25) were used to screen 16S rRNA gene sequence data for potential chimeras. Sequences flagged by either program as potentially chimeric were manually evaluated and rejected if suspect. Sequences that passed the screening process were aligned with the NAST aligner (17), and the aligned sequences were inserted, by parsimony insertion, into trees in the 16S rRNA database from greengenes.lbl.gov (16) that is compatible with the ARB (32) software package; ARB was used to construct all phylogenetic trees in this work. Sequence alignments used in phylogenetic analysis were minimized/masked using the Lane mask, with ~876-bp homologous nucleotide positions for phylogenetic analyses (29).

Functional gene analysis. For analysis of putative ammonia-oxidizing archaea (AOA), amoA gene fragments of ~635 bp were amplified using the PCR primers Arch-amoAF (5′-STAATGGTCTGGCTTAGACG-3′) and Arch-amoAR (5′-GCGGCCATCCATCTGTATGT-3′) (21) by use of the following protocol: 95°C for 5 min and 30 cycles consisting of 94°C for 45 s, 53°C for 60 s, and 72°C for 1 min, followed by 72°C for 15 min. Triplicate PCRs were pooled, gel purified, cloned, and sequenced as described above. Archaeal amoA nucleotide sequences were assembled and edited using Sequencherv.4.2 (GeneCodes, Ann Arbor, MI). Amino acid-based alignment of amoA DNA sequences was performed with the software transAlign (10). Maximum likelihood phylogenetic trees of the aligned amoA amino acid data set were constructed with RAxML (46). Bootstrap analysis was used to estimate the reliability of phylogenetic reconstructions (100 replicates). Branches with less than 70% support were deleted. To evaluate the richness within the archaeal amoA clone library, rarefaction analysis and Chao1 nonparametric richness estimations were performed using DOTUR (43). Operational taxonomic units (OTUs) were defined as sequence groups in which sequences differed by ≤5%.

SEM and energy-dispersive spectroscopy (EDS). Samples were washed in 70% ethanol and dehydrated in an ethanol series to 100%. Samples were dried in a critical point dryer with liquid CO₂. Half of the samples were examined under environmental scanning electron microscopy (SEM) conditions by use of an FEI Quanta 200 environmental SEM (ESEM) with a Princeton Gamma Tech Avalon microanalysis using secondary and backscatter electron detectors. The remaining samples were sputter coated with Au/Pd.

Nucleotide sequence accession numbers. 16S rRNA gene sequences have been deposited in GenBank under the accession numbers EF378701 to EF378943, while the amoA sequences have been deposited under accession numbers DQ672634 to DQ672715 and EF032877. The clone names for the 16S rRNA sequences reflect location as follows: GSST indicates Glenwood Springs, speleothem tip; GSSB indicates Glenwood Springs, speleothem base; GSCM indicates Glenwood Springs, ceiling mat; and GSCR indicates Glenwood Springs, cave raft.

Microscopy of stalactite biofabric. An ~2-cm leather-like soda straw stalactite, shown in Fig. 1A, was cut from the ceiling biofilm, and three segments were taken for analysis: the ceiling mat/base; the base to middle ~1 cm, presumably the more mature component of the stalactite; and the ~1-cm “growing” tip. Phase-contrast microscopy and DAPI (4′,6′-diamidino-2-phenylindole) fluorescence microscopy suggested that the wall of the stalactite was significantly biological in nature and complex in structure. Obvious variation included an ~10-fold-higher concentration of microbial cells along the outside of the stalactite than on the inside wall, adjacent to the lumen of the straw. To gain a more detailed view, we analyzed samples with an ESEM fitted with EDS (ESEM-EDS) for elemental analyses.

FIG. 1. (A) Photograph of a sectioned stalactite (tip removed) on a razor blade for scale (base, near blade). (B) ESEM image of the outer, leather-like wall of the “stalactite” with an outer “sheath” of structured material (center) (more ...)

The microbial assemblage presumably responsible for the rubber-like structure of the speleothem is evident in the ESEM image mosaic shown in Fig. 1B to I. The central region of diffuse microbial cells (Fig. 1B to E) is surrounded by a sheath-like structural material. This sheath is comprised of what appear to be laminated, densely packed microbial filaments that are interwoven to form the biofabric (Fig. 1B to E). Embedded in the surfaces of these laminated layers are numerous crystals (Fig. 1D and E), indicated by ESEM-EDS to be comprised primarily of sulfur granules, consistent with the presence of these speleothems in a hydrogen sulfide-rich atmosphere. This biofabric-mineral matrix is consistent down the length of the soda straw and is similar in appearance to fabrics formed by some endolithic communities (50).

The structure of the inner wall of the stalactite is distinct from that of the outer surface, as shown in Fig. 1F to I. The inside is less densely packed but still forms a fabric comprised of two distinct morphotypes as follows: small, irregular spheres attached to filaments. Most of the fabric mass is associated with the wall, but some structures bridge hollows in the wall (Fig. 1G) along the entire length of the speleothem, possibly contributing to the structural stability of the formation.

Microscopy of adit wall. These stalactites descend from a biofilm on the limestone ceiling. Imaging of this biofilm by ESEM revealed finely structured biofabric-like layers (Fig. 1J to L) and, on the wall-proximal side, a red/brown corrosion residue possibly due to microbial activities. To gain insight into the mineralogy of the corrosion process, we examined the residue by ESEM-EDS, which detected iron, silica, and oxygen, suggesting the presence of iron oxides and silica-iron clays (see Fig. S1A to S1D in the supplemental material). The Leadville limestone in which the mine adit is found is nearly pure calcium carbonate (23), so we surmised that the iron oxides and clays represented insoluble material in the rock matrix that accumulates due to rock dissolution. To test this theory, we dissolved ~5 g of Leadville limestone (from the same location as the mine adit) in hydrochloric acid and determined that all nonsoluble particles had the same EDS profile as the red/brown corrosion residue on the wall adjacent to the biofilm. The results collectively suggest that biofilm actively dissolves the host rock.

We also used ESEM-EDS to inspect the host rock of the biofilm, a thin veneer of host limestone peeling away from the ceiling in locations where biofilm occurs. The results (see Fig. S1E to S1J in the supplemental material) suggested a significant structural change in the rock in direct contact with the biofilm, with a secondary mineralization on the surface of the host rock and the deposition of a sulfur-rich mineral. The high-sulfur material suggests the activity of sulfuric acid dissolution. The sulfur signature in the ESEM-EDS could be traced down into the rock, following miniscule cracks in the limestone host. Beyond the zones of apparent sulfur dissolution, the mineral was highly enriched in phosphate, which is not normally found in high concentrations in the Leadville limestone (23) and so could potentially serve as a biomarker.

Microscopy of “dumplings.” Floating on top of the 0.3-m-deep, 50°C, spring-fed pools along the length of the passage were 3- to 10-cm-thick mineral deposits we referred to as “dumplings” due to their unusual appearance. Unlike what is the case for typical mm-thick cave “rafts,” to our knowledge such thick dumplings have not been described previously. ESEM-EDS analysis of the dumpling speleothem (Fig. 2) showed that these structures are accumulations of amorphous gypsum surrounded by a morphologically complex community. EDS indicated a high carbon content in the outer fabric that surrounds the dumpling with a morphological appearance similar to that of extracellular polymeric substance (EPS) and a gypsum-like elemental signature from the inside. Different crystal shapes are evident under the biofabric, with various cellular morphotypes evident (Fig. 2).

FIG. 2. ESEM images of the mineral consortia associated with the surface and inner portions of the floating “dumpling” structures. (A to C) Images demonstrate the presence of microbial consortia within the “dumpling” structures (more ...)

Phylogenetic composition of biofabrics. To identify the organisms associated with these biofabrics, we conducted phylogenetic analyses using rRNA gene sequences. Total DNA was extracted from three sections of the stalactite along its length (mat at base, base [~ 1 cm], and tip [~1 cm]) and from one bulk dumpling sample. PCR was carried out with the DNAs by use of the “universal” rRNA primer pair 515F/1391R, PCR products were cloned, and 96 randomly selected clones were sequenced per location. Although this number of clones likely would not sample all of the diversity of the communities, it provides an estimate of the phylogenetic composition of the main biomass.

Figure 3 shows the microbial community composition at the phylum (phylogenetic division) level for all four samples. The stalactite and dumpling communities differ considerably, with the stalactite community being less diverse. Generally, the same kinds of bacterial and archaeal phyla are located in all three regions of the stalactite, although the proportions differ markedly. No eucaryal sequences were detected among clones analyzed. In contrast to the stalactite community, the dumpling community is highly diverse (Fig. 3). The dumpling community contains many close relatives to known thermophilic heterotrophs, some related to organisms isolated from caves from many global locations, and no archaeal or eucaryal sequences were found.

FIG. 3. Cumulative rRNA gene analyses with universal PCR primers (515F and 1391R) reflect the distribution of sequences by phylogenetic group, as identified by RAxML analysis and replacement in ARB.

The stalactite community exhibits a marked shift in phylogenetic makeup along its length from the cave ceiling mat to the tip of the stalactite (Fig. 3). Since our analysis probably undersamples the clone libraries, we tested if the observed results are statistically significant using the computer program UniFrac (31). This analysis revealed marginally significant differences between the tip and ceiling mat (UniFrac P value, 0.018), while differences between base and ceiling mat (P = 1.0) and between base and tip (P = 0.884) were not significant (31).

Representatives of the Thermus-Deinococcus phylum are abundant in clone libraries from throughout the length of the stalactite. Phylogenetic analysis (Fig. 4A) shows that these sequences are of limited diversity, that they are the same group at a species-level proxy (≥97% rRNA sequence identity) throughout the stalactite, and that members correspond to the Meiothermus genus. Cultured examples of such organisms are moderately thermophilic heterotrophs. The other dominant (42%) group in the ceiling mat library, representatives of Chloroflexi, are almost absent from the tip community (2%). Phylogenetic analysis (Fig. 4B) shows that the stalactite sequences are mostly not specifically related to known sequences and are only distantly related to the few cultured representatives of the Chloroflexi phylum. Gammaproteobacteria rRNA genes are enriched severalfold from the mat (4%) to the tip (18%) and have closest relatives in sequences of organisms from soils and sediments. Chloroflexi sequences also are abundant in the dumpling library but were phylogenetically distinct from those in the stalactite libraries. As with soils and sediments, what metabolic capabilities these organisms are capable of, or engaged in, within these speleothem environments are unknown.

FIG. 4. Diagrammatic phylogeny constructed to show organisms from this study compared to other key sequences. (A) Phylogenetic tree of the Thermus-Deinococcus group with sequences from this study represented by Glenwood Springs speleothem tip (GSST), speleothem (more ...)

Another compositional difference between the stalactite libraries is the proportion of mesophilic Crenarchaeota 16S rRNA sequences from the ceiling mat (4%), the base (17%), and the tip (23%) of the stalactite (Fig. 3). Phylogenetic analysis (Fig. 4C) shows that these sequences are most closely related to clones previously recovered from service water obtained from subsurface depths (>3 km) within deep South African gold mines (47). The next closest relatives to the cave crenarchaeal sequences are various marine “group 1a” Crenarchaeota, some of which are capable of ammonia oxidation (27).

The energy sources for organisms in this environment are obscure. To evaluate the possibility that these speleothem-associated crenarchaeotes might have the genetic capacity for ammonia oxidation (AOA), we screened the community DNA from all three samples for putative crenarchaeal ammonia monooxygenase subunit A (amoA) genes. These amoA genes have recently been identified in metagenomic libraries (24, 48, 49), in the cultured ammonia-oxidizing archaeon Nitrosopumilus maritimus (27), in the sponge symbiont Cenarchaeum symbiosum (24), and in a variety of soil (30), water column (14, 21, 28, 35, 52), and sedimentary (21) environments, including an Austrian subsurface radioactive thermal spring with water temperatures of 42°C (51).

Although archaeal 16S rRNA sequences were found in all three stalactite samples, archaeal amoA genes were successfully amplified only from the stalactite tip. Phylogenetic analysis of 82 amoA sequences from the stalactite tip revealed one distinct and dominant sequence type (65 of 82 clones with >98% nucleotide identity; clone group A, a geothermal AOA cluster, in Fig. 5), which is closely related to 3 of 4 sequences reported from the Austrian subsurface thermal spring (51) as well as database sequences from Nevada hot springs (Q. Ye, H. Jiang, H. Dong, and C. Zhang, unpublished data). Two other closely related groups of adit amoA sequences, of 5 (group B) and 11 (group C) sequences, respectively, fall into a cluster comprised primarily of sequences from soils and estuarine sediments and of a single Austrian thermal spring sequence. Archaeal amoA richness, using 5% and 15% OTU cutoffs at the DNA level, showed only seven and three OTUs, respectively, associated with the stalactite (Chao1-estimated richness analysis in DOTUR [43]). The AOA richness in this unusual environment is fairly low relative to that seen in most other environments analyzed to date (21). However, the temperature at which these organisms are found (50°C), coupled with the similarity of the cave amoA gene sequences to those from a subsurface thermal spring (51) and Nevada hot springs, show that crenarchaeota putatively capable of ammonia oxidation are not limited to low-temperature environments. The similarity of gene sequences from these geographically disparate geothermal environments (Colorado, Austria, and Nevada) is also consistent with the notion that geochemically and physically similar environments select for functionally, and in some cases phylogenetically, similar kinds of organisms (45).

FIG. 5. Diagrammatic phylogenetic tree of the crenarchaeal amoA gene diversity associated with the tip of the geothermal speleothem, based on RAxML analysis of nucleotide sequences. Representative database sequences from all known environments where amoA has (more ...)

Microscopic analyses. The ~2-cm soda straw stalactite (Fig. 1A) descends from a dark-colored, microbial biofilm that covers the cave ceiling. The stalactites are hollow (Fig. 1A) and appear to grow downward from the ceiling by a combination of microbial growth and mineral precipitation that forms a fine biofabric around the edge of a water droplet at the tip. We surmise that the soda straw-like speleothems originate as an extension of the microbial community that makes up the biofilm on the adit walls. We postulate that water drops on the ceiling-associated microbial mat initially propagate the mat. It is possible that cellular metabolites and gas exchange alter chemical solubilities in the water and cause mineral precipitation at the air-water interface. With time, microbial cells would become intertwined with these growing crystals, and they indeed may be responsible for their propagation; thus, a cell-mineral-based biofabric thickens into visible layers (Fig. 1B and C). With this process, the community composition shifts in terms of ecological evenness, likely due to changing environmental conditions along the length of the stalactite. Throughout this development, water continues to flow through the center of the cellular and mineral layer outside and the process continues. Eventually, structures of multiple centimeters in length are formed.

Biofabrics and biomineralization. One phenomenon in the stalactite is the intertwining of the mineral matrix and layers of cells (Fig. 1). This might occur as a result of chemical precipitation of dissolved compounds in waters or of active biomineralization processes that result from microbial activities. ESEM-EDS shows a disproportionate amount of gypsum and/or sulfate present on and around the cellular forms within the stalactite compared to what is seen for the ceiling biofilm. A second phenomenon is that separate processes seem to occur between the stalactite (with sulfur granules) and the ceiling biofilm (with iron, silica, and oxygen dominant [see Fig. S1 in the supplemental material]). ESEM-EDS shows the presence of both iron oxides and clays between cells, and these minerals accumulate between the biofilm and the hard limestone wall in particular (see Fig. S1 in the supplemental material). The chemical composition of the clays indicates they likely result from acidic dissolution of the limestone matrix due to microbial activities; organic acids of metabolic origin chemically weather the limestone basement rock. Surprisingly, altered rock underneath the biofilm is enriched in both calcium sulfate (gypsum) and calcium phosphate (see Fig. S1 in the supplemental material). Secondary gypsum mineralization processes in karst zones are well documented; however, in the Leadville limestone of central Colorado, calcium phosphates are uncommon (23).

Phylogenetic analyses. The phylogenetic results provide some insight into the processes that likely underlie this ecosystem. If the relative abundances of particular physiological types are taken to reflect the relative amounts of different energy sources that are drawn on for primary productivity, then a census of the microbial community would correspond to the energy demands of the particular environmental niche. The census conducted here yields information about what kinds of cells are present, and from relationships to cultivated relatives, potential metabolic themes for this ecosystem can be proposed (45). For instance, the finding of rRNA genes representative of both Thermus spp. and Hydrogenophilus spp., whose cultivated relatives oxidize molecular hydrogen (H₂), indicates that H₂ may be a significant electron donor in this subaerial ecosystem. However, most of the organisms detected by sequence are not sufficiently closely related to cultivars with known properties to merit speculation on metabolic lifestyle.

Potential AOA. The identification of putative crenarchaeal amoA gene sequences from the speleothem tip indicates that ammonia may serve as an electron donor in this microenvironment. We acknowledge, however, that the detection of these putative amoA genes (by PCR) does not confirm activity. The pronounced increase in the abundance of crenarchaeal 16S rRNA gene sequences along the axis of the stalactite suggests that their distribution is controlled by geochemical gradients (Fig. 3). We speculate that ammonia from surface organic matter above could reach the tip of the speleothem as gaseous aerosols in this closed, humid environment. The detection of putative AOA in this unusual low-oxygen environment is consistent with their previous detection in suboxic water columns (e.g., the Black Sea) (14, 21, 28), activated sludge (37), and soil and sediments (21). This study continues to expand not only the range of habitats where AOA occur but also their temperature range up to 50°C.

Although it is not possible to link definitively the crenarchaeal 16S rRNA gene and amoA sequences obtained without cultivation or metagenomic approaches, the dominant group of amoA sequences from the stalactite tip (geothermal AOA cluster, group A; Fig. 5) possibly corresponds to the sole South African gold mine Archaea-8/group 1.1a-like crenarchaeal 16S rRNA phylotype recovered in this study (Fig. 4C). Support for this stems from the fact the dominant amoA sequence type (group A) is more closely related to various marine “water column” sequences (marine group 1.1a Crenarchaeota) than to soil-derived sequences likely to be affiliated with the group 1.1b or group 1.1c Crenarchaeota (not found in our 16S rRNA clone libraries). In contrast, the other 17 amoA sequences (groups B and C; Fig. 5) fall into a cluster comprised primarily of closely related soil and a few sediment sequences, as well as a lone Austrian thermal spring clone sequence (51), consistent with the recovery of group 1.1b-like 16S rRNA sequences from that system. In addition, 16S rRNA sequences closely related to various South African gold mine Archaea-like or group 1.1b Crenarchaeota were recovered in a previous study of Lechuguilla Cave, New Mexico (36), which suggests that crenarchaeal lineages may be common inhabitants of caves. Of note is the fact that the archaeal sequences recovered from both Lechuguilla Cave and the Austrian thermal spring were obtained with archaea-specific primers (21F and 958R). Thus, although only one dominant crenarchaeal 16S rRNA phylotype was detected within our “universal” clone libraries, it is possible that additional sequence types (e.g., group 1.1b) that correspond to these other AOA would be detected using archaea-specific primers or with more clones examined. Nevertheless, the role of the Crenarchaeota and other cave microorganisms in subsurface nitrogen cycling, even at high temperatures, will be the subject of future investigations.

Cave dumplings. Phylogenetically, the most complex community in this vapor cave surrounds the dumplings that float on the surface of cave floor pools (Fig. 3). The complexity confounds efforts to interpret a nutritional basis for the dumpling community. Of note, however, is the fact that 14% of the library is composed of relatives (99% rRNA sequence identity) of heterotrophic organisms isolated from other caves. This subspecies relationship suggests that much of the community thrives on organics provided in the hot water. ESEM-EDS indicates that the dumplings have a gypsum core wrapped within a complex biofabric of microbiota (Fig. 2) embedded in what appears to be EPS wrapped around gypsum crystals and between cells. In addition to the thickness of the biofabric around the mineral matrix, this biofabric surface enclosure likely prevents the highly soluble gypsum from dissolving in the 50°C water. Such a hypothesis is supported by the observation that when a dumpling is physically damaged, disrupting the microbial fabric, dissolution of the precipitated gypsum is immediate.

Conclusions. We have found that microbial mats line nearly every surface of this geothermal mine adit and form distinct communities comprised of novel organisms, many only distantly related to known organisms. Thus, the study expands known phylogenetic diversity. We focused on communities that occur as biofabrics, organized in arrays of cellular structures that lend structural strength to the mineralized component. Such fabric-like structures probably are common in mineral-based environments such as soils and sediments and are little studied. Fabric-like structure could play a significant role in the maintenance and propagation of the communities that make up the natural microbial world.

Our results also confirm the known temperature range where putative AOA are likely to be found to 50°C and expand the broad range of habitats where these organisms may persist if not thrive. Finally, we suggest that given the amount and extent of microbial diversity present in the examined speleothems, microbiota may play a more significant role in many speleothem formations than is currently acknowledged.

Support for this work to J.R.S. came from a geobiology postdoctoral fellowship from the Agouron Institute. Support for H.A.B. comes from the Center for Integrated Natural Science and Mathematics at NKU with infrastructure support from the NIH KY-INBRE program. N.R.P. is supported by funds from the NASA Astrobiology Institute.

We thank members of the Colorado caving community who assisted with this project, including Donald Davis, Douglas Medville, Richard Rhinehart, and Fred Luiszer. At NKU, we thank Nicholas Taylor for assistance with rock extraction techniques and Karl Hagglund for assistance with ESEM-EDS imaging. Thanks to Katie Roberts for her early contribution to this project and to members of the Pace and Spear labs and Jason Sahl for assistance and thoughtful reviews of this work.