Novel Microbial Diversity Retrieved by AutonomousRobotic Exploration of the World’s Deepest

Vertical Phreatic Sinkhole

Jason W. Sahl,1 Nathaniel Fairfield,2 J. Kirk Harris,3 David Wettergreen,2

William C. Stone,4 and John R. Spear1

Abstract

The deep phreatic thermal explorer (DEPTHX) is an autonomous underwater vehicle designed to navigate anunexplored environment, generate high-resolution three-dimensional (3-D)maps, collect biological samples basedon an autonomous sampling decision, and return to its origin. In the spring of 2007, DEPTHX was deployed inZacaton, a deep (*318m), limestone, phreatic sinkhole (cenote) in northeastern Mexico. As DEPTHX descended,it generated a 3-D map based on the processing of range data from 54 onboard sonars. The vehicle collected watercolumn samples and wall biomat samples throughout the depth profile of the cenote. Post-expedition sampleanalysis via comparative analysis of 16S rRNA gene sequences revealed a wealth of microbial diversity. Tradi-tional Sanger gene sequencing combined with a barcoded-amplicon pyrosequencing approach revealed novel,phylum-level lineages from the domains Bacteria and Archaea; in addition, several novel subphylum lineageswere also identified. Overall, DEPTHX successfully navigated and mapped Zacaton, and collected biologicalsamples based on an autonomous decision, which revealed novel microbial diversity in a previously unexploredenvironment. Key Words: Autonomous—Robotics—Microbial diversity—16S rRNA. Astrobiology 10, 201–213.

1. Introduction

In the search for extraterrestrial microbial life, there isemergent interest in the field of exploratory autonomous

robotics (Diaz-Calderon et al., 2007). To search for microbiallife on planetary bodies that contain liquid water, such asthe jovian moon Europa (Squyres et al., 1983), a robotic ve-hicle would be required to navigate through space, land onthe moon’s surface, penetrate the ice sheet, explore the alienocean for microscopic life, and transmit its findings back toEarth. In an effort to develop tools for autonomous naviga-tion and exploration of unexplored water bodies, a vehicleknown as the deep phreatic thermal explorer (DEPTHX) wasdesigned (Krajick, 2007; Stone, 2007). The deployment goals ofDEPTHX were to map and navigate autonomously throughan unexplored environment and collect biological samplesthrough autonomous decisions based on a set of programmedcriteria.

Located in northeastern Mexico (Fig. 1A), Zacaton, thedeepest water-filled vertical sinkhole (cenote) in the world(Gary and Sharp, 2006), provided the site for deployment of the

DEPTHX vehicle (Fig. 1B). Zacaton (Fig. 1C) is fed by subsur-face hydrothermal input from nearby volcanic activity, whichprovides a spatially and temporally stable temperature profileof*308C throughout the year, and a sulfur input that suppliesthe surrounding area its name: Rancho laAzufrosa (azufrosa isSpanish for sulfurous). The cenote is thought to have beencreated as hydrothermal water supplied from nearby volcanicactivitywas acidifiedwith carbon dioxide or hydrogen sulfide,or both, to dissolve the host limestone as it moved upwardthrough the subsurface (Gary et al., 2008). In a preliminaryequipment test for DEPTHX, a sonar sonde dropped intoZacaton in May 2005 revealed that the cenote was at least290m deep and potentially much deeper. Zacaton was chosenas the test site for the deployment of DEPTHX because thecenote was deep, likely mappable, and unexplored.

2. Methods and Materials

2.1. Site description

Sistema Zacaton is located in a localized karst regionof northeastern Mexico (22859035.10@N, 98809055.96@W). The

1Colorado School of Mines, Environmental Science and Engineering, Golden, Colorado.2Carnegie Mellon Robotics Institute, Pittsburgh, Pennsylvania.3Pediatrics, University of Colorado, Denver, Colorado.4Stone Aerospace, Del Valle, Texas.

ASTROBIOLOGYVolume 10, Number 2, 2010ª Mary Ann Liebert, Inc.DOI: 10.1089=ast.2009.0378

201

geological formations of this system are part of a larger sys-tem of travertine-capped sinkholes, dry karst cave passages,and dolines in the vicinity (Gary et al., 2008). Zacaton is a105m diameter, *318m deep sinkhole with a 17m deeplimestone rim=cliff band above the water. Two meter-thick,round, floating reed islands, known as zacates, float and areblown around on the surface of the cenote. In direct sunlight,the clear blue water can turn milky-white as sulfur cloudsdevelop in the top 5m of the surface water and disappearwhen the Sun sets.

2.2. Vehicle engineering

A detailed description of the DEPTHX vehicle has beenreported previously (Stone, 2007). In short, DEPTHX is ap-proximately 1.5m tall and 1.9m in length and width and ispowered by lithium-ion batteries. The vehicle has a full suiteof underwater navigation sensors, including a HoneywellHG2001 Inertial Measurement Unit (IMU), two ParoscientificDigiquartz depth sensors, and an RDI Navigator 600 DopplerVelocity Log (DVL). There is also a conductivity, temperature,and depth sensor for measuring the speed of sound for cor-rections to sonar-based measurements. The specifications forthe inertial navigation system are roll=pitch 0.28 2s, yaw 0.482s, for the DVL velocities 0.3 cm=s 1s, and for the depthsensors 0.01% of full range (10 cm for our 1000m rated sen-

sor). Fifty-four narrow-beam long-range sonar sensors allowthe vehicle to detect subsurface features and generate maps in3-D. The vehicle is equipped with a science package for mi-crobiological sampling (see below).

The System Executive (i.e., the multiprocessor autonomous‘‘brain’’ of the robot) allocates out steps in a dive specificationbased on the state of the system and its location, which isestimated using a combination of dead-reckoning and sonar-based localization. In the DEPTHX system, there is no on-board planning; rather, the Executive chooses actions from aprimary or cascading set of contingency plans. Dive specifi-cations are composed intomission plan files that encode, in aninterpreted command language, the sequence of behaviors toexecute. These behaviors may initiate sensing, navigate tolocations, trigger on conditions, perform maneuvers, or ac-quire samples. The primary plan file is backed up by a numberof contingency plans that are triggered on conditions encodedin the currently executing plan description.

2.3. Vehicle navigation and SLAM

The high-grade IMU had negligible yaw drift over time,and under nominal conditions the onboard sensors provideddead-reckoned navigation on the order of 0.5% of distancetraveled. Under certain conditions [e.g., when the vehiclegets too close to a wall (<*2m)], the DVL cannot providevelocity measurements, at which point the quality of thedead-reckoned solution degrades significantly (*10% of dis-tance traveled). To estimate velocity, the DVL requires thateach of its four sonars receive a reflected signal from eitherthe water column or a surface. At times of transition, how-ever, as when moving away from a surface into open wateror when the orientation of the surface deflects the returnaway, the unit can ‘‘lose lock’’ and become unable to estimatevelocity. As a result, dead-reckoning accuracy is reduceduntil velocity lock is re-established. Underwater, there isno absolute position information (like GPS), so this dead-reckoning position error is cumulative.

The vehicle can bound this position error by performing itsown localization estimate (using the onboard sonars) rela-tive to a map. However, exploration implies that there is nomap available, thus the vehicle has to build a map with newdata and simultaneously localize itself using that map; this isknown as the simultaneous localization andmapping (SLAM)problem. The approach we describe below builds highly ac-curate 3-Dmaps; but, due to the coupled nature of localizationand mapping, it is susceptible to unavoidable error (or drift)over long distances, except when the vehicle can close loopsand return to previously mapped areas. We developed a real-time probabilistic SLAM algorithm that provides an estimateof the vehicle trajectory as well as a 3-D metric map of theenvironment [see Fairfield et al. (2006, 2007) for a more com-plete description of our SLAM approach].

For SLAM, the DEPTHX vehicle has an array of 54 narrow-beam (24 at 300 kHz, 1.758 beam width; and 32 at 600 kHz, 28beam width with ranges of 200 and 100m, respectively) so-nars that provide a constellation of range measurementsaround the vehicle at about 1Hz. This array is in the shape ofthree great circles, a configuration that was arrived at afterstudying the suitability of various sonar geometries for 3-Dmapping (Fairfield et al., 2006). The sonars have long ranges(100 or 200m), and the accuracy of the range measurements

FIG. 1. A map showing the location of Zacaton (A); thedeep phreatic thermal explorer (DEPTHX) (B); an aerial viewof Zacaton (C); biomats in Zacaton from depths of 11m (D),273m (E), 10m (F).

202 SAHL ET AL.

is approximately 10 cm; however, the low resolution, updaterate, and point density make the mapping problem under-water significantly more difficult than it is with rangingsensors, like a laser scanner, that provide fast, accurate, high-resolution ranges in typical air-filled environments.

2.4. Sample collection and preparation

Wall biomat sampleswere collectedwith an onboard spring-loaded sampling arm and stored in a stainless-steel borerwith a sealable door; once a single wall sample was taken, thedoor closed on the collection borer. Immediately after thevehicle surfaced, wall biomat samples were ejected into 2mlcryovials and frozen in liquid nitrogen. Between samplingtrips, the borer was cleaned with a dilute bleach solution (3%)and thoroughly rinsed with sterile, double-distilled water.

Prior to water column sample collection, DEPTHX ran itspump for 2min, which allowed for the sample collectiontubing to be completely flushed with in situ water. Sampleswere then collected into sterile urinary collection bags (2 Lcapacity). Upon surfacing, the collection bags were connectedto a peristaltic pump, and water was filtered directly onto0.2 mm polyethersulfone filters (Pall Corporation, East Hills,NY) housed in a stainless steel high-pressure filter manifold(Millipore, Billerica, MA). One quarter of each filter was cutwith flame-sterilized scissors and frozen in liquid nitrogen.New collection bags were used for each sampling trip.

2.5. Water and wall biomat chemistry

Values for pH, temperature, dissolved oxygen, and specificconductivity were recorded with a modified Hydrotech HT6meter (Hydrotech, Hutto, TX) mounted on the DEPTHX ve-hicle. Sulfide concentrations were calculated with a modifiedlaboratory sensor mounted on the vehicle. For shallow watercolumn samples, concentrations of ammonia and hydrogensulfide were manually recorded with CHEMetrics kits (CHE-Metrics, Calverton, VA) and a field photometer. Prior to fil-tering for microbiological analysis, water for the DEPTHXsampling missions was collected for laboratory water analyteanalysis. Fifteen milliliters of water was syringe filtered(0.45mm) and acidified (pH< 2) for analysis of metals andmajor cations with inductively coupled plasma atomic emis-sion spectroscopy. An additional 15ml ofwaterwas filtered foranalysis of major anions with ion chromatography, and ap-proximately 200ml of water was collected in glass amber vialsfor analysis of total organic carbon and alkalinity. All sampleswere stored at 48C until analyzed. A Spearman correlation test(Spearman, 1904) of specific water chemistry species withdepth was calculated with XLSTAT (www.XLSTAT.com).DEPTHX had the ability to collect five individual 2L watersamples per dive, and site-dependant replicate sampling waspossible based on the generated map by SLAM.

For wall biomat compositional analysis, between 17 and125mg of mat material was dried for 24 h at 558C. The driedmat material was digested in 100% nitric acid on a hot platefor 24 hours. The digest was centrifuged, the supernatant wasdiluted 1:10 in Milli-Q water (Millipore, Billerica, MA), andsyringe filtered (0.45 mm). Samples were then run with in-ductively coupled plasma atomic emission spectroscopy forcompositional analysis. Aqueous concentrations were con-verted to a percentage of each element, considering the massof the entire extraction.

2.6. DNA extraction, PCR, cloning, sequencing

DNA was extracted from wall samples and filter sectionswith the Mobio Powersoil Kit (Mobio, Carlsbad, CA) andeluted in a volume of 30 ml TE. Polymerase chain reaction(PCR) of the 16S rRNA gene was performed on all samplesby using universal forward primer 515F (50-GTG CCA GCMGCC GCG GTA A-30) (Lane, 1991), bacterial-specific forwardprimer 8F (50-AGA GTT TGA TCC TGG CTC AG-30) (Lane,1991), and archaeal-specific forward primers 4Fa (50-TCCGGT TGA TCC TGC CRG-30) (Hershberger et al., 1996) and21Fa (50-TTC CGG TTG ATC CYG CCG GA-30) (Reysenbachand Pace, 1995). Reverse primers included the nominallyuniversal 1391R (50-GAC GGG CGG TGW GTR CA-30) and1492R (50-GGT TAC CTT GTT ACG ACT T-30) (Lane, 1991).PCRs were performed with Promega PCR Master Mix (Pro-mega, Madison, WI) and a primer concentration of 1 mM. PCRthermal cycling conditions consisted of an initial denaturationat 948C for 2min, followed by 30 cycles of denaturation (948Cfor 1min), annealing (50–558C for 1min) and elongation (728Cfor 1min). PCR amplicons were gel purified (Millipore, Bill-erica, MA), TOPO TA (Invitrogen, Carlsbad, CA) cloned andsequenced on a MegaBACE 1000 dye-terminating sequencer.

2.7. Sanger sequence analysis

Sequences were processed by PHRED and PHRAP withuse of the XplorSeq interface (Frank, 2008). Multi-directionreads were performed on each sample, and contiguous readswere compiled. Sequences were screened for chimeras withMallard (Ashelford et al., 2006), and non-chimeric sequenceswere aligned with the NAST aligner (DeSantis et al., 2006a).Sequences were inserted into the ARB dendogram by parsi-mony insertionwith use of the Lanemask filter (Lane, 1991). Aglobal sequence comparison was made by building a dis-tance matrix between the query and target sequence in ARB;the sequence distance was then converted to a percent relat-edness.

2.8. Phylogenetic analyses

Sequences that did not group with known phyla in ARBwith use of the current Greengenes (DeSantis et al., 2006b)database release were analyzed in a phylogenetic tree thatcontained representative sequences from currently recog-nized bacterial phyla and candidate phyla. Sequences wereexported unaligned and subsequently aligned with the In-fernal aligner (Nawrocki et al., 2009). Covariancemodel (Eddyand Durbin, 1994) files were created with a seed alignmentdownloaded from the comparative RNA web site (Cannoneand Subramanian, 2002), and a consensus secondary structurewas calculated with RNAalifold (Hofacker, 2003). Alignmentconfidence estimates were calculated with Infernal, with useof the–p flag, for each residue of each sequence. Columnswithhigh alignment confidence values (*100%) were consideredto be conserved, and a filter mask, which masks out variablecolumns, was created with custom Perl script. The conservedcolumn alignment mask, along with all aligned sequences,was imported back into ARB.Manual adjustments weremadeto the alignment, and sequences were exported with use ofthe custom conserved filter mask. For the Archaea, a phylo-genetic tree was inferred via the RAxML web server (Stama-takis et al., 2008). For the Bacteria, the number of taxa

MICROBIAL DIVERSITY IN ZACATON 203

analyzed (*36,000) exceeded the capabilities of themaximum-likelihood algorithm, and a neighbor-joiningmethod, which uses a minimum evolution principle (Priceet al., 2009), was used to infer the tree topology.

2.9. Pyrosequencing and analysis

Polymerase chain reactions were conducted with PromegaPCR Master Mix and the bacterial-specific primers 8F and338R (50-CAT GCT GCC TCC CGT AGG AGT-30) (Amannet al., 1990); these priming sites span the hypervariable V1 andV2 regions of the 16S rRNA gene (Van de Peer et al., 1996).The reverse primer contained a Hamming barcode, whichallowed for the pooling of multiple samples in a single high-throughput sequencing run (Hamady et al., 2008). PCR am-plicons were normalized, quantified on aNanoDropND-1000(NanoDrop, Wilmington, DE), diluted in molecular-gradewater to a concentration of *0.01 ng=ml, and sequenced ona 454 GS FLX machine (454 Life Sciences, Branford, CT).Sequences were sorted by barcode and quality filtered to re-move sequences with ambiguous characters (Ns) and se-quences that were of an insufficient length (<2 standarddeviations of the mean) with custom Java script. The barcodeand primer sequence were trimmed prior to analysis.

Sequences were aligned with use of the ribosomal databaseproject (RDP) pyrosequencing pipeline (Cole et al., 2009),which uses the Infernal aligner in conjunction with a con-sensus secondary structure alignment. The pyrosequencingpipelinewas also used to cluster sequences based on a furthestneighbor algorithm; operational taxonomic units (OTUs)were then formed based on a 97% sequence identity thresh-old. A representative sequence from each OTU was alignedwith the NAST aligner and classified with the Greengenesclassifier (DeSantis et al., 2006b). Taxonomic classificationswere made according to the Hugenholtz taxonomic system(DeSantis et al., 2006b).

2.10. qPCR

The total number of 16S rRNA gene copies in extractedsamples was estimated with quantitative PCR (qPCR) on aLightcycler 480 (Roche). Degenerate primers for conservedregions flanking the hypervariable V4 region of the 16S rRNAgene (Raue et al., 1988) were forward primer (50-AYT GGGYDT AAA GNG-30) and reverse primer (50-TAC NVG GGTATC TAA TCC-30) (Cole et al., 2009); these primers cover>94.5% of long (>1200 nucleotides), high-quality sequencesin the RDP database, using the probe match tool (Cole et al.,2009). Primers were also designed specifically for the C2Crenarchaeota; these include forward primer 303F (50-GGGAGC CCG GAG ATG-30) and reverse primer 1069R (50-ACCTCA CGG CAC GAA-30) (Dr. Bradley Stevenson, unpub-lished). A calibration curve was calculated by using sevenstandards, in triplicate, and purified and serial diluted am-plicons. Gradient PCR revealed an optimum annealing tem-perature for each primer of 488C.

The reverse primer 494R (50-CTG AGG GTA CCG TCAATA-30) was also designed for the putative candidate phylumAzufrosa group 1 (AG1). The primer was compared againstthe RDP database with the probe match tool, which only re-turned six positive matches; all these sequences grouped withAG1 and only two were >1200 nucleotides in length. Thisreverse primer was paired with the bacterial primer 8F to

determine the distribution of this clade throughout the depthprofile of Zacaton.

2.11. Species richness estimators

Parametric species richness estimates were calculated withmethods reported previously ( Jeon et al., 2006). Six differentparametric models [single-point Poisson, negative binomial,inverse Gaussian, lognormal, Pareto, mixture of two expo-nentials (M.E.)] were used to fit the observed data based onthe frequency of sequences at the designated percent identitylevel. The best performing model estimate was chosen basedon a biologically relevant standard error [<(estimate=2)], thehighest goodness-of-fit value, and a high right truncationpoint. For comparison, the non-parametric species richnessestimates Chao1 (Chao, 1987) and ACE (Chao and Lee, 1992)were calculated with EstimateS (Colwell, 2005). Coverageestimates were calculated by dividing the number of observedOTUs by the best parametric estimate.

2.12. Secondary structure analysis

A comparative secondary structure analysis was per-formed by aligning a consensus sequence from AG1 against a16S rRNA sequence from E. coli (accession #J01695). The se-quence from AG1 was then manually mapped onto the well-established E. coli 16S rRNA secondary structure (Woese et al.,1980) to identify regions of conservation and divergence.

2.13. Nucleotide accession numbers

The 1588 Sanger sequences generated in this study weresubmitted to Genbank under accession numbers FJ484010–FJ485598. Pyrosequences were submitted to the Genbankshort read archive under accession number SRA008128.3.

3. Results

3.1. Vehicle navigation and SLAM

With SLAM, we built a three-dimensional 0.5m resolutionmetric map of Zacaton. Without ground truth, it is difficult tomake strong statements about localization and map accu-racy, but the maps showed consistent findings independentlyproduced across many dives. Our method was able to usesparse, noisy, low-resolution, and low-rate sonar data to build3-D maps, localize, run SLAM on board the DEPTHX vehicle,and generate the first ever maps of this large-scale, complex,natural formation.

3.2. Water chemistry

As DEPTHX maneuvered through the cenote, the vehiclerecorded thousands of water chemistry values with the Hy-drotech sensor. The data reveal an incredibly homogeneouswater body, with nearly identical values for pH and temper-ature throughout the depth profile of the cenote (Table 1).Major ion analyses confirmed this trend, with uniform con-centrations of sulfate, nitrate, chloride, and sodium. However,statistically significant (P< 0.03) negative Spearman correla-tions (r<!0.7) of zinc, iron, and manganese with depth wereobserved. No clear chemistry boundaries were detectable; forexample, even the surface water sample showed reducingconditions, with no clear oxic=anoxic boundary or chemoclineobserved. Sulfate was the most abundant common terminal

204 SAHL ET AL.

Tabl

e1.

Chem

icalConstituen

tsin

Zacato

n

Water

Wallbiom

at

Param

eter

Surface

22m

30m

60m

90m

150m

198m

273m

285m

4m

32m

114m

273m

Tem

p(8C)

30.10

30.03

30.02

30.03

30.00

30.04

30.10

30.10

30.06

Ph

6.67

6.58

6.73

6.72

6.72

6.52

6.54

6.53

6.50

Eh(M

v)

!17

5!16

8!17

7!18

5!20

0!27

4!28

2!28

2!26

3DO

(mM)

00

00

00

00

0TOC

(mM)

0.16

0.15

0.18

0.16

0.18

0.24

0.14

0.14

0.39

Spec

Cond(mS=

cm)

759

767

767

767

767

760

761

761

761

Fe-T(mM)

0.01

1.63

0.52

1.04

0.38

10.20

10.20

0.20

0.08

NH

" 4(m

M)

0.05

0.07

NM

NM

NM

NM

NM

NM

NM

Major

anions(m

M)

PO

3!

4BDL

BDL

00

0BDL

BDL

0BDL

CI!

1.01

0.97

0.98

0.97

0.95

0.97

0.96

0.98

0.95

NO

! 30

00

00

00

00

Br

00

00

00

00

0HCO

! 36.24

6.32

6.28

6.20

6.17

6.17

6.30

6.29

6.32

SO2 4

0.15

0.15

0.15

0.15

0.15

0.15

0.16

0.16

0.14

Major

cations(m

M)

Ca

3.11

3.37

2.79

3.06

2.80

2.94

2.98

2.96

2.71

6.57

%6.05

%0.14

%6.45

%K

0.08

0.07

0.08

0.08

0.08

0.08

0.08

0.08

0.09

0.07

%0.10

%0.03

%0.05

%Mg

0.49

0.44

0.44

0.46

0.46

0.49

0.47

0.49

0.47

0.08

%0.10

%0.40

%0.07

%Na

1.38

1.20

1.27

1.30

1.31

1.72

1.21

1.32

2.30

0.04

%0.11

%0.02

%0.05

%H

2S

0.01

0.01

NM

NM

NM

NM

NM

NM

NM

NM

NM

NM

NM

Trace

elem

ents

(mM)

Zn

13.31

37.17

7.80

12.39

14.84

4.13

7.04

1.84

1.07

0.00

19%

0.01

18%

0.01

20%

0.01

80%

Ni

00

00

00

00

00.00

01%

0.00

09%

0.00

11%

0.00

26%

Mn

0.73

0.73

0.73

1.09

1.46

0.55

0.55

0.55

0.36

0.00

27%

0.00

52%

0.01

10%

0.00

49%

Scations(m

eq)

8.27

8.20

0.23

0.25

0.24

0.26

0.24

0.25

0.28

San

ions(m

eq)

7.44

7.81

0.13

0.13

0.13

0.13

0.13

0.13

0.13

Chargeim

balan

ce(%

)5.28

2.44

1.70

5.96

3.23

7.65

3.94

4.58

7.43

BDL,below

detectionlimit;NM,notmeasu

red.

DO,dissolved

oxy

gen

;TOC,totalorgan

iccarbon.

electron acceptor, which, combined with lineages associatedwith sulfate reduction (see below), suggests that this may bean important mode of metabolism throughout the depthprofile of the cenote.

Inductively coupled plasma atomic emission spectroscopydata were only obtained for four wall samples in Zacaton(Table 1). Only zinc and nickel showed a correlation (r# 1.0)of concentration with depth (Table 1). Approximately 8% ofthe biomat mass consisted of extractable metals; the re-maining mass of the mat is likely non-digestible minerals.Calcium comprised the highest mass percent of the mats.

3.3. 16S rRNA gene sequence analysis

To sample the microbiology of Zacaton, the DEPTHXvehicle was equipped with five sterile collection bags foraqueous samples and a spring-loaded extendable arm de-signed to sample the extensive biomats that coat the walls ofthe cenote (Fig. 1D, 1E, 1F).

Culture-independent gene sequence analysis (Pace, 1997;Spear et al., 2005) allowed for the analysis of microbial com-munity structure in 3-D space, all in the context of the geo-chemistry of the cenote. Visualization of the mats by anonboard camera on DEPTHX revealed thick, complex pho-tosynthetic mats in the shallow subsurface (<30m) (Fig. 1D,1F) and delicate, striated mats 273m deep in the cenote(Fig. 1E). The initial survey of 16S rRNA gene sequences re-sulted in 1588 Sanger sequence reads that guided a morecomprehensive sampling of the microbial community with abarcoded-amplicon pyrosequencing approach (Hamady et al.,2008).

Results of the 16S rRNA gene sequence analysis of the fil-terable fraction of the water column suggest that the com-munity is dominated by a few phylogenetic types (phylotypes)such as Arcobacter spp. and Sulfuromales spp., which groupwith the Epsilonproteobacteria. Sequences related to theseArcobacter and Sulfuromales phylotypes have been isolatedfrom other sulfidic caves (Macalady et al., 2008; Porter andEngel, 2008) and are thought to be important microbes in the

cycling of carbon, nitrogen, and sulfurous compounds(Campbell et al., 2006). Even within the Epsilonproteobacteria,clear differences were observed with depth; for example, deepin the water column (273m), the epsilon sequences werephylogenetically related (98% global identity) to Sulfuricurvumkujiense, a chemoautotrophic microbe capable of anaerobicsulfide oxidation (Kodama and Watanabe, 2004).

Other sequences from the water column grouped with theGammaproteobacteria and were related (96% global identity)to sequences amplified from deep subsurface mine fluids inSouth Africa (Gihring et al., 2006). In general, the water col-umn microbial diversity was low (Table 2); in fact, out of*7500 full- and partial-length gene sequences screened froma depth profile of the water column, only 850 OTUs werecompiled with a 97% sequence identity threshold. Even withlow observed diversity and dominance of a few phylotypes,clear trends in phylotype distribution and diversity wereobserved in the depth profile of the water column (Table 2).

In contrast, sequences obtained from the walls of the cenoteshowed high microbial diversity within the bacterial and ar-chaeal domains (Figures 2 and 3); for example, out of 105bacterial phyla currently recognized by the Hugenholtz tax-onomic system (DeSantis et al., 2006b), 74% were found inthe mats of Zacaton (more than 10 representative sequencesfrom each clade) based on *39,000 pyrosequences, as deter-mined by the Greengenes classifier. Furthermore, a cleardepth profile of bacterial phylotypes was observed, with no-vel cyanobacterial sequences (95% global identity with near-est neighbor) dominant in shallow regions of the cenote andsequences associated with sulfate reduction (Deltaproteo-bacteria, Thermodesulfovibrio) dominant below the photiczone (>40m) (Fig. 4). While some Deltaproteobacterial se-quences showed relatedness to cultured taxa, such as Syn-trophobacter pfennigii (93%) (Wallrabenstein et al., 1995) andDesulfomonile tiedjei (95%) (DeWeerd et al., 1990), manyDeltaproteobacterial sequences (*45%) in deeper samples(>100m) belonged to 15 OTUs that were not closely related(<95%) to Deltaproteobacterial sequences in public databasesfrom environmental samples. Sequences associated with

Table 2. Microbial Diversity and Coverage Estimates for Samples in Zacaton

MatrixSample

depth (m) SequencesOTUs(97%)

Best parametricestimate

coverage(%) ACE Chao 1

Shannonindex

Simpsonindex (1=D)

mat1 2490 584 1552 (se 128) 38 1187 1187 5.1 41.884 3680 1209 4062 (se 309) 30 2659 2659 6.23 147.947 2697 718 2656 (se 267) 27 1689 1700 4.32 6.8310 3306 1444 4670 (se 259) 31 3228 3237 6.51 129.9520 3948 1753 5389 (se 270) 33 4149 4003 6.94 578.932 2705 901 4011 (se 507) 22 2280 2293 5.64 55.958 1459 923 5482 (se 754) 17 2991 3011 6.54 623.09

100 1761 655 3022 (se 430) 22 1642 1654 5.45 55.55114 1898 1129 4984 (se 475) 23 3025 3038 6.72 712.97195 2119 1056 4305 (se 460) 25 2295 2303 6.52 348.34273 2338 1186 4485 (se 395) 26 2686 2696 6.69 607.78

water0 2662 140 563 (se 134) 25 297 305 2.29 5.5117 1810 204 674 (se 94) 30 513 527 3.24 10.5

273 2396 392 1697 (se 248) 23 951 962 3.69 12.66

se, standard error.

206 SAHL ET AL.

photosynthetic anoxygenic sulfide oxidizers (Chlorobi) werefound throughout the depth profile of the cenote, even inzoneswell below the photic zone (Fig. 4). At depths of>100m,the community was dominated by sequences that groupedwith Deltaproteobacteria and Chloroflexi (Fig. 4). Epsilon-proteobacterial sequences, which are ubiquitous in the watercolumn, were almost entirely absent from all wall biomatsamples, regardless of depth.

Archaeal diversity was dominated (58% of archaeal se-quences) by the C2 Crenarchaeota (Fig. 4), which has beenrecognized as a monophyletic group by some taxonomic

curators (DeLong and Pace, 2001; Robertson et al., 2005;DeSantis et al., 2006b); this clade is also associated with the so-called miscellaneous Crenarchaeota group (Nunoura et al.,2005). Sequences that group with this clade have been am-plified from deep subsurface gold mines (Nunoura et al.,2005), geothermal wells (Rogers and Amend, 2005), deepmarine anoxic basins ( Jeon et al., 2008), and a hydrothermalvent in the Mariana Trough (unpublished data).

Clone libraries constructed with universal primers suggestthat C2 sequences were prevalent at depth in the cenote and inlower abundance, or absent altogether, in shallow regions.

FIG. 2. Neighbor-joining phylogenetic tree of the bacterial domain. Numbers in parentheses indicate accession numbers.Dark wedges indicate phyla that are represented in either biomat or water column samples.

MICROBIAL DIVERSITY IN ZACATON 207

qPCR confirmed this trend, with an increased number of C216S rRNA gene copy numbers, compared to the total 16SrRNA gene copy numbers, with sample depth (Fig. 5). Twonucleotide mismatches were found between the V4 forwardprimer and sequences of the C2 Crenarchaeota on the 30 end.To allow for the potential of the primer not annealing to C2gene templates, the percentage of C2 gene copy numberswas calculated by comparison of the C2 gene copy num-bers to the V4 gene copy numbers plus the C2 gene copynumbers. This calculation is considered even more conser-vative, given the fact that bacteria generally contain more rrnoperon copies than archaea (Klappenbach et al., 2001; Acinaset al., 2004).

Additional archaeal sequences grouped with known me-thanogens (Methanomicrobia) as well as anaerobic methaneoxidizers (ANME-1). Singleton, non-chimeric, sequenceswereobserved on the archaeal tree (Fig. 3) that showed low globalidentity to nearest neighbors (*85% relatedness). Additional

sequencing with archaeal-specific primers may help to furtherdefine these novel lineages.

3.4. Azufrosa candidate phyla

Initial phylogenetic analyses using ARB parsimony inser-tion suggested that some phylotypes did not group withknown bacterial phyla at a phylogenetic distance (*0.20)consistent with phylum-level delineations (Schloss and Han-delsman, 2005) (Fig. 2). In total, three novel candidate phylawere identified, although two of these phyla were only re-presented by two sequences from Zacaton mats greater than1250 nucleotides in length. Additional phylogenetic analysesby maximum likelihood (Stamatakis, 2006) and neighbor-joining algorithms (Price et al., 2009) support the presence ofthese novel lineages.

Azufrosa group 1 (AG1) was the most represented novelcandidate phylum identified. Sequences from deep aquifers

FIG. 3. Maximum-likelihood phylogenetic tree of the archaeal domain, including sequences from both wall biomat andwater column samples. Numbers in parentheses following accession numbers indicate the number of sequences associatedwith that phylotype in this environment.

208 SAHL ET AL.

in Japan (Shimizu et al., 2006) and New York (Bakermansand Madsen, 2002) are also affiliated with AG1 (Fig. 3). Asecondary structure analysis of AG1 sequenceswas comparedto the established secondary structure of E. coli, and a trun-cation of Helix 17 was identified (Fig. 6), which has also beenseen in archaea and other bacteria (Raue et al., 1988). The PCR

primer 494R spans the unique truncation in the helix, whichallows for high specificity to this group.

All sequences that group with AG1 were amplified fromdeep samples in Zacaton (>82m). A qPCR reaction was runto determine the relative abundance and distribution of thisclade throughout the cenote water column; however, no cleartrend in abundance was observed with depth (Fig. 5). Thedistribution of these sequences, based on these results, doesnot appear to be confined to deep environments.

3.5. Diversity and species richness

A positive and significant (P# 0.001) Spearman correlation(r# 0.83) of biomat microbial diversity (1=D) with depth wasidentified (Table 2). Parametric species richness estimatorswere higher than non-parametric estimators for all samples,which is consistent with other studies (Hong et al., 2006; Jeonet al., 2006).

Species richness coverage estimates ranged from 17–38%.However, richness estimates have been shown to scale withsequencing effort (Morales et al., 2009), which suggests thatadditional sequencing may reveal these estimates to be anoverestimation of the calculated community coverage.

3.6. Autonomous sampling

One of the two primary missions of DEPTHXwas to collecta biological sample based on an autonomous decision.

FIG. 4. Vertical profile of Zacaton and a heat map showing the relative distribution of selected phyla at various depths.Abbreviations include Proteob., Proteobacteria (a# alpha, d#delta, g# gamma, e# epsilon); Bactero., Bacteroidetes; Cya-noba., Cyanobacteria; Nitrosp., Nitrospirae; Firmicu., Firmicutes; Spiroch., Spirochaetes; Acidoba., Acidobacteria; Chlorob.,Chlorobi; Chlorof., Chloroflexi. ‘‘Others’’ represents the summed remainder of bacterial phyla.

FIG. 5. Relative abundance of C2 Crenarchaeota andAzufrosa group 1 (AG1) gene copies in the depth profile ofZacaton compared to the total gene copy number.

MICROBIAL DIVERSITY IN ZACATON 209

DEPTHX was programmed to detect chemical interfaces; wehypothesized that these interfaces provide sources of chemicaldisequilibrium and were the most likely location of variedmicrobial diversity and metabolism. However, due to thehomogeneous nature of detectable chemical parameters inZacaton, DEPTHX was unable to locate a clear chemical in-terface in the water column for biological sampling. In fact,only the sulfide concentration appeared to vary significantlywith depth. Based on this observation, a voltage thresholdthat corresponded to an in situ sulfide concentration wasprogrammed into DEPTHX. In one sampling mission,DEPTHX detected the programmed threshold, approachedthe wall of the cenote, collected a wall biomat sample, andreturned to its origin based on the SLAMmethod. Although asulfide calibration was never successfully calculated, the au-tonomous decision to sample was successfully made byDEPTHX.

4. Discussion

4.1. Water chemistry

The DEPTHX vehicle was programmed to detect chemicalinterfaces. Currently within the field of astrobiology, thepresence of known electron donors and acceptors may indi-cate that the energetics are amenable to microbial metabolism(Hand et al., 2007; Hoehler et al., 2007).

However, an analysis of water chemistry parameters inZacaton revealed an extremely homogeneous water body.This homogeneity is likely caused by convective mixing ashydrothermal groundwater from the deep subsurface movesupward through the water column. The hydrothermal sourceis likely deep, as no discernable differences in temperaturewere observed between the surface and at 300m depth.Three-dimensional maps of the cenote also failed to revealany physical inflow to the water body, which suggests thatthe opening is too small (<1m) to be discerned by the lowresolution of the narrow-beam sonars on DEPTHX.

Dissolved oxygenwas not detected at any depth in Zacaton(Table 1). Even just below the surface, oxygen appeared tobe depleted, perhaps due to biotic or abiotic (Luthy et al.,2000) sulfide oxidation. Elemental sulfur clouds, whichemerge in full sunlight in the top 5m of the water column, arelikely formed from incomplete phototrophic sulfide oxidation(Amend et al., 2004), due to oxygen-limited conditions (vanden Ende and vanGemerden, 2003). In the depth profile of thecenote, no significant trends in concentration were observed

from major cations, anions, or total organic carbon; these re-sults correspond well with the homogeneous behavior ofother abiotic parameters.

4.2. Microbial diversity

The geochemistry and morphology of Zacaton select for aunique and diverse microbial community. A comparison ofmicrobial communities between Zacaton and other hydro-thermal travertine systems (Breitbart et al., 2008), as well asfreshwater sulfurous lakes (Casamayor et al., 2000), showedthat the microbial communities in Zacaton are significantlydifferent than these comparable systems.

Initial Sanger sequencing of wall biomat samples revealeda diverse microbial community. Subsequent directed pyro-sequencing provided a higher-resolution analysis of the mi-crobial community structure in Zacaton. This method hasbeen used previously to identify the low-abundance micro-bial phylotypes in an environment, which constitute the‘‘rare biosphere’’ (Sogin et al., 2006).

Overall, pyrosequencing of wall biomat DNA revealedhigher microbial diversity than Sanger sequencing. Althoughthis is likely a result of the greater sequencing effort, results ofa recent study suggest that shorter sequences reveal highermicrobial diversity than longer sequences due to more effi-cient amplification (Huber et al., 2009). In this study, longerSanger sequences were used for accurate phylogenetic anal-ysis and to anchor shorter pyrosequences on phylogenetictrees, while shorter sequences were used for assessments ofmicrobial diversity and distribution. Such an approach maytake full advantage of the benefits and limitations of short andlong gene sequences.

A positive correlation of biomat microbial diversity withdepth was observed in the depth profile of Zacaton. Highermicrobial diversitywith depth has also been observed in otheranoxic systems (Madrid et al., 2001) and may be attributed tomore varied metabolisms below the photic zone.

4.3. 16S rRNA gene sequence analysis

Water column samples showed low microbial diversity,with Epsilonproteobacteria and Chlorobi phylotypes domi-nating the clone libraries (Fig. 4). Both these groups have beenassociated with sulfide oxidation (Bergstein and Cavari, 1983;Wagner et al., 1998); their presence, combined with elemen-tal sulfur clouds that develop in sunlight, suggests that thesemicrobes may oxidize sulfide in Zacaton as well. Microbestypically associated with oxygenic photosynthesis, such ascyanobacteria and green algae, were absent from water col-umn clone libraries. Epsilonproteobacteria sequences domi-nated water column samples and were absent or in lowabundance in wall biomat samples; cultivated members ofrelated Epsilonproteobacteria are motile (Miller et al., 2007;Ho et al., 2008), which suggests that the Epsilonproteobacteriain Zacaton prefer an unattached lifestyle.

Shallow biomat samples in the cenote contained sequencesfrom known phototrophic organisms (e.g., Cyanobacteria)and photosynthetic sulfide oxidizers (e.g., Chlorobi). How-ever, even deeper samples (>100m) contained sequences thatgroup with Chlorobi. Microorganisms that group withChlorobi have been shown to oxidize sulfide in the presenceof geothermal radiation (Beatty et al., 2005), which may helpexplain the presence of sequences associated with obligate

FIG. 6. Helix 17 of the bacterial secondary structure model.The structure from E. coli is in gray, and the truncated helixfrom AG1 is in black. The bracketed section indicates theposition of primer 494R.

210 SAHL ET AL.

photosynthetic microbes 273m below the surface in this hy-drothermally fed system (Fig. 4).

Archaeal sequences in both biomat and water columnsamples were dominated by the C2 Crenarchaeota clade.In general, the majority of Crenarchaeota genomes containgenetic machinery needed for sulfur metabolism (Amendet al., 2004), although no closely related genome to the C2Crenarchaeota has currently been sequenced. 16S rRNA genesequencing and qPCR revealed the composition and distri-bution of this clade throughout the depth profile of Zacaton.The increase in C2 sequence abundance with depth (Fig. 5)suggests that in the photic zone these archaea are outcom-peted, perhaps by organisms that metabolize photosynthateexuded by anoxygenic phototrophs (Medina-Sanchez andVillar-Argaiz, 2006). The ecological role of these Archaeacannot be determined through phylogenetic analyses, al-though the types of environments where C2 have previouslybeen observed may help to identify samples of significant C2abundance, which can provide targets for exploratory me-tagenomics.

Comparative 16S rRNA gene sequence analysis revealedthe presence of three well-supported, novel candidate phyla.In cenote Zacaton, two of these phyla are only representedby two sequences, although a comparative molecular analysisof four adjacent cenotes has revealed additional sequencesthat groupwith these phyla (unpublished data). Although thepresence of 16S rRNA gene sequences from novel lineagesdoes not reveal their ecological function, their identificationmay represent a fertile ground for bioprospecting for novelproteins, hydrogenases, and lipids, for example. Metage-nomic analyses of samples containing significant percentagesof novel microbes may allow for the assembly and interpre-tation of genomes (Kunin et al., 2008) from microbes thatrepresent phylum-level novelty (Hugenholtz and Kyrpides,2009).

Deeper samples in Zacaton well below the photic zonemaybe more representative of an astrobiological community thanshallow samples that receive full or partial sunlight. For ex-ample, sequences amplified from deep in the water column(273m) that show high relatedness to chemoautotrophic, an-aerobic sulfide oxidizers suggest that these areas may be in-dependent of photosynthetically derived carbon. This type ofmetabolism may be applicable to deep, dark, extraterrestrialaqueous systems with a hydrothermal feed source.

4.4. Autonomous robotic exploration

During the course of the field campaign, DEPTHXwas ableto implement SLAM, detect a programmed chemical inter-face, sample both the water column and wall biomat, andreturn to its origin, all based on decisions executed on boardwithout any human guidance. The sampling and subsequentmolecular analysis of diver-inaccessible samples, with corre-lated spatial and geochemical information, have allowed fora more thorough understanding of microbial diversity anddistribution in hydrothermal, phreatic sinkholes.

As robotic technology advances, the integration of theautonomous sampling and maneuvering methods discussedin this study with advanced onboard chemistry and molec-ular analyses tools will enable the development of a vehiclecapable of searching for microbial life on planetary bodiesthat contain liquid water, such as Europa.

Acknowledgments

The authors would like to graciously thank Alejandro Da-vila for access to Rancho la Azufrosa and the entire DEPTHXteam including George Kantor, Antonio and Nora Fregoso,John Kerr, Marcus and Robin Gary, Dom Jonak, and theSouthwest Research Institute. Additional thanks to the PaceLab at the University of Colorado for their assistance oper-ating the MegaBACE 1000 and to the Consortium for Com-parative Genomics located at the University of ColoradoDenver for operating the 454 GS FLX machine. Custom Javascript to decode barcodes was written by Erik Sahl and thecustom Perl infernal script was written by Chuck Pepe-Ranney. The authors would also like to thank Dr. BradleyStevenson for providing the C2 primer set and Dr. Laura Beerfor her valuable comments. Funding for this project wasprovided by a NASA ASTEP grant NNG04GC09G.

Abbreviations

AG1, Azufrosa Group 1; DEPTHX, deep phreatic thermalexplorer; DVL, Doppler Velocity Log; IMU, Inertial Mea-surement Unit; OTUs, operational taxonomic units; PCR,polymerase chain reaction; qPCR, quantitative PCR; RDP,ribosomal database project; SLAM, simultaneous localizationand mapping.

References

Acinas, S.G., Marcelino, L.A., Klepac-Ceraj, V., and Polz, M.F.(2004) Divergence and redundancy of 16S rRNA sequences ingenomes with multiple rrn operons. J. Bacteriol. 186:2629–2635.

Amann, R., Binder, B.J., Olson, R.J., Chisholm, S.W., Devereux, R.,and Stahl, D.A. (1990) Combination of 16S rRNA-targeted oli-gonucleotide probes with flow cytometry for analyzing mixedmicrobial populations. Appl. Environ. Microbiol. 56:1919–1925.

Amend, J.P., Rogers, K.L., and Meyer-Dombard. D.R. (2004)Microbially mediated sulfur-redox: energetics in marine hy-drothermal vent systems. In Sulfur Biogeochemistry Past andFuture, GSA Special Paper 379, edited by J.P. Amend, K.J. Ed-wards, and T.W. Lyons, Geological Society of America,Boulder, CO, pp 17–34.

Ashelford, K.E., Chuzhanova, N.A., Fry, J.C., Jones, A.J., andWeightman, A.J. (2006) New screening software shows thatmost recent large 16S rRNA gene clone libraries contain chi-meras. Appl. Environ. Microbiol. 72:5734–5741.

Bakermans, C. and Madsen, E.L. (2002) Diversity of 16S rDNAand naphthalene dioxygenase genes from coal-tar-waste-contaminated. Microb. Ecol. 44:95–106.

Beatty, J.T., Overmann, J., Lince, M.T., Manske, A.K., Lang, A.S.,Blankenship, R.E., Van Dover, C.L., Martinson, T.A., andPlumley, F.G. (2005) An obligately photosynthetic bacterialanaerobe from a deep-sea hydrothermal vent. Proc. Natl. Acad.Sci. U.S.A. 102:9306–9310.

Bergstein, T. and Cavari, B.Z. (1983) Sulfide utilization by thephotosynthetic bacterium Chlorobium phaeobacteroides. Hydro-biologia 106:241–246.

Breitbart, M., Hoare, A., Nitti, A., Siefert, J., Haynes, M., Dins-dale, E.A., Edwards, R.A., Souza, V., Rohwer, F., and Hol-lander, D. (2008) Metagenomic and stable isotopic analyses ofmodern freshwater microbialites in Cuatro Cienegas, Mexico.Environ. Microbiol. 11:16–34.

Campbell, B.J., Engel, A.S., Porter, M.L., and Takai, K. (2006) Theversatile e-proteobacteria: key players in sulphidic habitats.Nat. Rev. Microbiol. 4:458–468.

MICROBIAL DIVERSITY IN ZACATON 211

Cannone, J.J. and Subramanian, S. (2002) The comparative RNAweb (CRW) site: an online database of comparative sequenceand structure information for ribosomal, intron, and otherRNAs. BMC Bioinformatics 3:2.

Casamayor, E.O., Schafer, H., Baneras, L., Pedros-Alio, C., andMuyzer, G. (2000) Identification of and spatio-temporal dif-ferences between microbial assemblages from two neighboringsulfurous lakes: comparison by microscopy and denaturinggradient gel electrophoresis. Appl. Environ. Microbiol. 66:499–508.

Chao, A. (1987) Estimating the population size for capture-recapture data with unequal catchability. Biometrics 43:783–791.

Chao, A. and Lee, S.M. (1992) Estimating the number of classesvia sample coverage. J. Am. Stat. Assoc. 87:210–217.

Cole, J.R., Wang, Q., Cardenas, E., Fish, J., Chai, B., Farris, R.J.,Kulam-Syed-Mohideen, A.S., McGarrell, D.M., Marsh, T.,Garrity, G.M., and Tiedje, J.M. (2009) The Ribosomal DatabaseProject: improved alignments and new tools for rRNA anal-ysis. Nucleic Acids Res. 37:D141–D145.

Colwell, R.K. (2005) EstimateS: Statistical estimation of speciesrichness and shared species from samples, Version 8. User’sguide and application available online at http:==purl.oclc.org=estimates.

DeLong, E.F. and Pace, N.R. (2001) Environmental diversity ofBacteria and Archaea. Syst. Biol. 50:470–478.

DeSantis, T.Z., Hugenholtz, P., Keller, K., Brodie, E.L., Larsen,N., Piceno, Y.M., Phan, R., and Andersen, G.L. (2006a) NAST:a multiple sequence alignment server for comparative analysisof 16S rRNA genes. Nucleic Acids Res. 34, doi:10.1093=nar=gkl244.

DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie,E.L., Keller, K., Huber, T., Dalevi, D., Hu, P., and Andersen,G.L. (2006b) Greengenes, a chimera-checked 16S rRNA genedatabase and workbench compatible with ARB. Appl. Environ.Microbiol. 72:5069–5072.

DeWeerd, K.A., Mandelco, L., Tanner, R.S., Woese, C., andSuflita, J.M. (1990) Desulfomonile tiedjei gen. nov. and sp. nov.,a novel anaerobic, dehalogenating, sulfate-reducing bacte-rium. Arch. Microbiol. 154:23–30.

Diaz-Calderon, A., Backes, P.G., and Bonitz, R.G. (2007) Anautonomous robotic scooping approach for planetary sampleacquisition. In IEEE=RSJ International Conference on IntelligentRobots and Systems (IROS), Vol. 10, doi:10.1109=IROS.2007.4398966.

Eddy, S.R. and Durbin, R. (1994) RNA sequence analysis usingcovariance models. Nucleic Acids Res. 22:2079–2088.

Fairfield, N., Kantor, G.A., and Wettergreen, D. (2006) Towardsparticle filter SLAM with three dimensional evidence grids ina flooded subterranean environment. In IEEE Internal Con-ference on Robotics and Automation (ICRA).

Fairfield, N., Kantor, G.A., and Wettergreen, D. (2007) Real-timeSLAM with octree evidence grids for exploration in under-water tunnels. Journal of Field Robotics 24:3–21.

Frank, D.N. (2008) XplorSeq: a software environment for inte-gratedmanagement and phylogenetic analysis ofmetagenomicsequence data. BMC Bioinformatics 9, doi:10.1186=1471-2105-9-420.

Gary, M.O. and Sharp, J.M., Jr. (2006) Volcanogenic karstificationof Sistema Zacaton, Mexico. In Perspectives on Karst Geomor-phology, Hydrology, and Geochemistry—A Tribute Volume to DerekC. Ford and William B. White, edited by R.S. Harmon and C.M.Wicks, GSA Special Paper 404, Geological Society of America,Boulder, CO, pp 79–89.

Gary, M.O., Fairfield, N., Stone, W.C., Wettergreen, D., Kantor,G.A., and Sharp, J.M., Jr. (2008) 3D mapping and character-ization of Sistema Zacaton from DEPTHX (DEep PhreaticTHermal eXplorer). In Proceedings of KARST08: 11th Sinkholeconference ASCE, pp 1–11.

Gihring, T.M., Moser, D.P., Lin, L.-H., Davidson, M., Onstott,T.C., Morgan, L., Milleson, M., Kieft, T.L., Trimarco, E.,Balkwill, D.L., and Dollhopf, M.E. (2006) The distribution ofmicrobial taxa in the subsurface water of the Kalahari Shield,South Africa. Geomicrobiol. J. 23:415–430.

Hamady, M., Walker, J.J., Harris, J.K., Gold, N.J., and Knight, R.(2008) Error-correcting barcoded primers for pyrosequencinghundreds of samples in multiplex. Nat. Methods 5:235–237.

Hand, K.P., Carlson, R.W., and Chyba, C.F. (2007) Energy, che-mical disequilibrium, and geological constraints on Europa.Astrobiology 7:1006–1022.

Hershberger, K.L., Barns, S.M., Reysenbach, A.-L., Dawson, S.C.,and Pace, N.R. (1996) Wide diversity of Crenarchaeota. Nature384:420.

Ho, H.T.K., Lipman, L.J.A., Wosten, M.S.M., and van Asten,A.J.A.M. (2008) Arcobacter spp. possess two very short flagel-lins of which FlaA is essential for motility. FEMS Immunol.Med. Microbiol. 53:85–95.

Hoehler, T.M., Amend, J.P., and Shock, E.L. (2007) A ‘‘follow theenergy’’ approach for astrobiology. Astrobiology 7:819–823.

Hofacker, I.L. (2003) Vienna RNA secondary structure server.Nucleic Acids Res. 31:3429–3431.

Hong, S.-H., Bunge, J., Jeon, S.-O., and Epstein, S.S. (2006) Pre-dicting microbial species richness. Proc. Natl. Acad. Sci. U.S.A.103:117–122.

Huber, J.A., Morrison, H.G., Huse, S.M., Neal, P.R., Sogin, M.L.,and Welch, D.B.M. (2009) Effect of PCR amplicon size on as-sessments of clone library microbial diversity and communitystructure. Environ. Microbiol. 11:1292–1302.

Hugenholtz, P. and Kyrpides, N.C. (2009) A changing of theguard. Environ. Microbiol. 11:551–553.

Jeon, S.-O., Bunge, J., Stoeck, T., Barger, K.J.-A., Hong, S.-H., andEpstein, S.S. (2006) Synthetic statistical approach reveals ahigh degree of richness of microbial eukaryotes in an anoxicwater column. Appl. Environ. Microbiol. 72:6578–6583.

Jeon, S.-O., Ahn, T.-S., and Hong, S.-H. (2008) A novel archaealgroup in the phylum Crenarchaeota found unexpectedly in aneukaryotic survey in the Cariaco basin. J. Microbiol. 46:34–39.

Klappenbach, J.A., Saxman, P.R., Cole, J.R., and Schmidt, T.M.(2001) rrndb: the ribosomal RNA operon copy number data-base. Nucleic Acids Res. 29:181–184.

Kodama, Y. and Watanabe, K. (2004) Sulfuricurvum kujiensegen. nov., sp. nov., a facultatively anaerobic, chemolithoau-totrophic, sulfur-oxidizing bacterium isolated from an un-derground crude-oil storage cavity. Int. J. Syst. Evol. Microbiol.54:2297–2300.

Krajick, K. (2007) Robot seeks new life—and new funding—inthe abyss of Zacaton. Science 315:322–324.

Kunin, V., Copeland, A.C., Lapidus, A., Mavromatis, K., andHugenholtz, P. (2008) A bioinformatics guide to metage-nomics. Microbiol. Mol. Biol. Rev. 72:557–578.

Lane, D.J. (1991) 16S=23S rRNA sequencing. In Nucleic AcidTechniques in Bacterial Systematics, edited by E. Stackebrandtand M. Goodfellow, John Wiley & Sons, Chichester, UK.

Luthy, L., Fritz, M., and Bachofen, R. (2000) In situ determinationof sulfide turnover rates in a meromictic alpine lake. Appl.Environ. Microbiol. 66:712–717.

Macalady, J.L., Dattagupta, S., Schaperdoth, I., Jones, D.S.,Druschel, G.K., and Eastman, D. (2008) Niche differentiation

212 SAHL ET AL.

among sulfur-oxidizing bacterial populations in cave waters.ISME J. 2:590–601.

Madrid, V.M., Taylor, G.T., Scranton, M.I., and Chistoserdov,A.Y. (2001) Phylogenetic diversity of bacterial and archaealcommunities in the anoxic zone of the Cariaco Basin. Appl.Environ. Microbiol. 67:1663–1674.

Medina-Sanchez, J.M. and Villar-Argaiz, M. (2006) Solarradiation–nutrient interaction enhances the resource and pre-dation algal control on bacterioplankton: a short-term exper-imental study. Limnol. Oceanogr. 51:913–924.

Miller, W.G., Parker, C.T., Rubenfield, M., Mendz, G.L., Wosten,M.M., Ussery, D.W., Stolz, J.F., Binnewies, T.T., Hallin, P.F.,Wang, G., Malek, J.A., Rogosin, A., Stanker, L.H., and Man-drell, R.E. (2007) The complete genome sequence and analysisof the epsilonproteobacterium Arcobacter butzleri. PLoS ONE26:e1358.

Morales, S.E., Cosart, T.F., Johnson, J.V., and Holben, W.E. (2009)Extensive phylogenetic analysis of a soil bacterial communityillustrates extreme taxon evenness and the effects of ampliconlength, degree of coverage, and DNA fractionation on classi-fication and ecological parameters. Appl. Environ. Microbiol.75:668–675.

Nawrocki, E.P., Kolbe, D.L., and Eddy, S.R. (2009) Infernal 1.0:inference of RNA alignments. Bioinformatics 25:1335–1337.

Nunoura, T., Hirayama, H., Takami, H., Oida, H., Nishi, S., Shi-mamura, S., Suzuki, Y. Inagaki, F., Takai, K., Nealson, K.H.,and Horikoshi, K. (2005) Genetic and functional properties ofuncultivated thermophilic crenarchaeotes from a subsurfacegold mine as revealed by analysis of genome fragments. En-viron. Microbiol. 7:1967–1984.

Pace, N.R. (1997) A molecular view of microbial diversity andthe biosphere. Science 276:734–740.

Porter, M.L. and Engel, A.S. (2008) Diversity of uncultured ep-silonproteobacteria from terrestrial sulfidic caves and springs.Appl. Environ. Microbiol. 74:4973–4977.

Price, M.N., Dehal, P.S., and Arkin, A.P. (2009) FastTree: com-puting large minimum evolution trees with profiles instead ofa distance matrix. Mol. Biol. Evol. 26:1641–1650.

Raue, H.A., Klootwijk, J., and Musters, W. (1988) Evolutionaryconservation of structure and function of high molecularweight ribosomal RNA. Prog. Biophys. Mol. Biol. 51:77–129.

Reysenbach, A.-L. and Pace, N.R. (1995) Protocol 16. Reliableamplification of hyperthermophilic archaeal 16S rRNA genesby the polymerase chain reaction. In Thermophiles. Archaea: ALaboratory Manual 3, Cold Spring Harbor Laboratory Press,New York, 101–107.

Robertson, C.E., Harris, J., Spear, J.R., and Pace, N.R. (2005)Phylogenetic diversity and ecology of environmental archaea.Curr. Opin. Microbiol. 8:638–642.

Rogers, K.L. and Amend, J.P. (2005) Archaeal diversity andgeochemical energy yields in a geothermal well on VulcanoIsland, Italy. Geobiology 3:319–332.

Schloss, P.D. and Handelsman, J. (2005) Introducing DOTUR, acomputer program for defining operational taxonomic unitsand estimating species richness. Appl. Environ. Microbiol. 71:1501–1506.

Shimizu, S., Akiyama, M., Ishijima, Y., Hama, K., Kunimaru, T.,and Naganuma, T. (2006) Molecular characterization of

microbial communities in fault-bordered aquifers in theMiocene formation of northernmost Japan. Geobiology 4:203–213.

Sogin, M.L., Morrison, H.G., Huber, J.A., Welch, D.M.,Huse, S.M., Neal, P.R., Arrieta, J.M., and Herndl, G.J.(2006) Microbial diversity in the deep sea and the under-explored ‘‘rare biosphere.’’ Proc. Natl. Acad. Sci. U.S.A.103:12115–12120.

Spear, J.R., Walker, J.J., McCollom, T.M., and Pace, N.R. (2005)Hydrogen and bioenergetics in the Yellowstone geothermalecosystem. Proc. Natl. Acad. Sci. U.S.A. 102:2555–2560.

Spearman, C.E. (1904) The proof and measurement of associa-tion between two things. Am. J. Psychol. 15:72–101.

Squyres, S.W., Reynolds, R.T., Cassen, P.M., and Peale, S.J.(1983) Liquid water and active resurfacing on Europa. Nature301:225–226.

Stamatakis, A. (2006) RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa andmixedmodels. Bioinformatics 22:2688–2690.

Stamatakis, A., Hoover, P., and Rougemont, J. (2008) A rapidbootstrap algorithm for the RAxML web servers. Syst. Biol.57:758–771.

Stone, W.C. (2007) Design and deployment of a 3D autonomoussubterranean submarine exploration vehicle [Proceedings#UUST07]. In Conference on Unmanned, Untethered SubmersableTechnology, Durham, NH, August 20–22.

Van de Peer, Y., Chappelle, S., and de Wachter, R. (1996) Aquantitative map of nucleotide substitution rates in bacterialrRNA. Nucleic Acids Res. 24:3381–3391.

van den Ende, F.P. and van Gemerden, H. (2003) Sulfide oxi-dation under oxygen limitation by Thiobacillus thioparus iso-lated from a marine microbial mat. FEMS Microbiol. Ecol.13:69–77.

Wagner, M., Roger, A.J., Flax, J.L., Brusseau, G.A., and Stahl,D.A. (1998) Phylogeny of dissimilatory sulfite reductasessupports an early origin of sulfate respiration. J. Bacteriol.180:2975–2982.

Wallrabenstein, C., Hauschild, E., and Schink, B. (1995) Syn-trophobacter pfennigii sp. nov., new syntrophically propionate-oxidizing anaerobe growing in pure culture with propionateand sulfate. Arch. Microbiol. 5:346–352.

Woese, C., Magrum, L.J., Gupta, R., Siegel, R.B., Stahl, D.A.,Kop, J., Crawford, N., Brosius, J., Gutell, R., Hogan, J.J., andNoller H.F. (1980) Secondary structure model for bacterial 16Sribosomal RNA: phylogenetic, enzymatic and chemical evi-dence. Nucleic Acids Res. 8:2275–2293.

Address correspondence to:John R. Spear

Colorado School of MinesEnvironmental Science and Engineering

Golden, CO 80401

E-mail: jspear@mines.edu

Submitted 22 April 2009Accepted 28 December 2009

MICROBIAL DIVERSITY IN ZACATON 213