Volcanic calderas delineate biogeographic provinces among ...

Volcanic calderas delineate biogeographic provincesamong Yellowstone thermophiles

Cristina Takacs-Vesbach,1* Kendra Mitchell,1

Olan Jackson-Weaver1 andAnna-Louise Reysenbach2

1Department of Biology, University of New Mexico,MSC03 2020, 1 UNM, Albuquerque, NM 87131, USA.2Department of Biology, Portland State University,PO Box 751 Portland, OR 97207, USA.

Summary

It has been suggested that the distribution of microor-ganisms should be cosmopolitan because of theirenormous capacity for dispersal. However, recentstudies have revealed that geographically isolatedmicrobial populations do exist. Geographic distanceas a barrier to dispersal is most often invoked toexplain these distributions. Here we show that uni-que and diverse sequences of the bacterial genusSulfurihydrogenibium exist in Yellowstone thermalsprings, indicating that these sites are geographicallyisolated. Although there was no correlation with geo-graphic distance or the associated geochemistry ofthe springs, there was a strong historical signal.We found that the Yellowstone calderas, remnantsof prehistoric volcanic eruptions, delineate biogeo-graphical provinces for the Sulfurihydrogenibiumwithin Yellowstone (c2: 9.7, P = 0.002). The pattern ofdistribution that we have detected suggests that majorgeological events in the past 2 million years explainmore of the variation in sequence diversity in thissystem than do contemporary factors such as habitator geographic distance. These findings highlight theimportance of historical legacies in determining con-temporary microbial distributions and suggest that thesame factors that determine the biogeography of mac-roorganisms are also evident among bacteria.

Introduction

Many studies have examined the extent to which theidentity, abundances and distributions of microorganismsvary with contemporary environmental conditions, but fewhave explored how their distribution may be influenced by

historical factors (Martiny et al., 2006). Baas-Becking’sstatement ‘everything is everywhere, but the environ-ment selects’ held that the distributions of microorganismsare cosmopolitan because of their enormous capacityfor long-distance dispersal, whereas influences of habitatand environmental variables determine where local popu-lations occur (de Wit and Bouvier, 2006). However, recentstudies have revealed the existence of geographicallyisolated microbial populations (Papke et al., 2003;Whitaker et al., 2003). Geographic distance as a barrier todispersal is most often invoked to explain these distribu-tions (Martiny et al., 2006).

The geographic distributions of organisms reflect boththe effects of contemporary environmental conditions andthe legacies of historical geological and climatic condi-tions on the origins and spread of lineages (for a reviewsee Martiny et al., 2006). A large body of literature showshow the distributions of macroorganisms can be classifiedinto biogeographic provinces, which reflect the historiceffects of plate tectonics, sea-level changes and barriersto dispersal such as mountain ranges (Hedges et al.,1996; Wares and Cunningham, 2001; de Bruyn et al.,2005). Methodological difficulties have hindered a com-parable treatment of microbial biogeography. Advances inmolecular techniques have found significant correlationsof microbial distributions with environmental factors suchas salinity or oxygen (Franklin et al., 1999; Casamayoret al., 2002), but there is little evidence of the latitudinalpatterns of diversity or geographic provinces seen inmacroorganisms (Morris et al., 2002; Knittel et al., 2005).If dispersal of microorganisms is limited by geographicalbarriers, biogeographical provinces should be apparent,such as those reported for terrestrial thermophiles fromseparate continents (Papke et al., 2003; Whitaker et al.,2003).

Recently, a framework for microbial biogeography wasproposed to consider the relative influence of habitat (con-temporary environmental factors) and province (historicallegacies) by evaluating four alternatives hypotheses: (i)the null hypothesis that microorganisms are randomlydistributed, (ii) that contemporary environmental variationis necessary and sufficient to account for current distri-butions, (iii) that contemporary distributions show clearlegacies of historical events, such as tectonic eventsand geographic barriers, and (iv) that the distribution ofmicroorganisms is due to both environmental effects and

Received 13 September, 2007; accepted 25 January, 2008. *Forcorrespondence. E-mail [email protected]; Tel. (+1) 505 2773418; Fax (+1) 505 277 0304.

Environmental Microbiology (2008) doi:10.1111/j.1462-2920.2008.01584.x

© 2008 The AuthorsJournal compilation © 2008 Society for Applied Microbiology and Blackwell Publishing Ltd

historical events (Martiny et al., 2006). The apparent influ-ence of habitat and province is affected by the scale ofsampling (Martiny et al., 2006). On smaller scales, envi-ronmental factors are believed to be of primary impor-tance in determining presence or absence of patterns ofgenetic variation, whereas on an intercontinental scale,isolation distance is believed to override environmentaleffects (Papke et al., 2003). It has been suggested thatthe joint influence of both environmental factors and his-torical contingencies should be most apparent at interme-diate scales (10–3000 km) (Martiny et al., 2006).

Here we apply this framework to determine the relativeinfluence of habitat and province on the distribution ofmembers of the bacterial genus Sulfurihydrogenibium inthermal springs of Yellowstone National Park. Usingmolecular phylogenetic approaches, combined withenvironmental data and dispersal-vicariance analyses,we found geographically isolated microbial populationswithin the Yellowstone thermal ecosystem. The distributionpattern we detected was correlated with the boundary ofYellowstone’s calderas (or volcanic craters) and suggeststhat volcanic eruptions in the past 2 million years explainmore of the DNA sequence divergence than do contempo-rary factors such as habitat or geographic distance. Thesefindings highlight the importance of historical legacies indetermining contemporary microbial distributions andsuggest that the same factors that determine the biogeog-raphy of macroorganisms are also evident among bacteria.

Results and discussion

We surveyed 18 thermal springs (Table 1) from essentiallyevery major thermal area in the Yellowstone. Sites sepa-rated by 5–79 km were screened for bacterial 16S rRNAgene sequences belonging to the Sulfurihydrogenibium,which are members of the order Aquificales. Sulfurihydro-genibium spp. often dominate near-neutral Yellowstonesprings (Barns et al., 1994; Hugenholtz et al., 1998;Reysenbach et al., 2000; Spear et al., 2005) and havebeen found in many terrestrial thermal systems world-wide, often contributing up to 95% of the biomass(Reysenbach et al., 2000; Takacs et al., 2001). Thirty-three unique 16S rRNA gene sequences (of 1728screened clones) were retrieved from eight springs. Simi-larity of the Sulfurihydrogenibium sequences ranged from97.1% to > 99% over 1446 nucleotides (based on uncor-rected distance). One explanation for the observedsequence divergence is that the unique sequences reflecttemperature ecotypes. However, there was no relationbetween the GC content of the sequences and the tem-perature of the spring from which they were collected(r2 = 0.062, n = 33, P = 0.14).

Sulfurihydrogenibium sequences from the eight siteswere distinct, suggesting that geographic isolation exists Ta

ble

1.Lo

catio

nan

dse

lect

edge

oche

mic

alpa

ram

eter

sof

ther

mal

sprin

gssu

rvey

edin

this

stud

y.

Sam

ple

IDA

rea

Nor

thin

gbE

astin

gbTe

mpe

ratu

re(°

C)

pHC

ondu

ctiv

ity(m

S)

Alk

alin

ityc

SO

4=A

s(t

otal

)F

e(t

otal

)S

iO2

H2S

03Y

MA

M00

1aM

amm

oth

4979

165.

252

2886

.673

.56.

4322

3078

254

60.

186

0.00

851

0.00

003

YG

IB02

3G

ibbo

nH

ill49

4881

1.6

5207

89.8

73.8

8.51

2230

115

107

2.45

60.

0222

20.

023

03Y

LOW

027

Low

erG

eyse

rB

asin

4934

077.

751

3245

.886

.28.

4314

8721

029

1.12

0.00

226

50.

079

03Y

CO

F03

5C

offe

eP

ots

4955

983.

655

4771

.265

.25.

7555

58

195

0.00

25.

4291

0.09

203

YJO

S03

8aJo

seph

’sC

oat

4954

349.

255

3827

.667

.36.

0416

8014

424

21.

240.

023

134

0.00

003

YH

SB

042a

Hot

Spr

ing

Bas

in49

5565

3.2

5588

89.2

64.5

5.90

1678

4070

10

0.00

723

44.

650

04Y

LST

010a

Lone

star

4918

086.

051

5156

.473

.67.

9012

5355

170.

874

0.01

518

40.

021

04Y

UP

P01

2U

pper

Gey

ser

Bas

in49

2339

8.0

5130

06.0

78.4

8.80

2260

407

191.

860

416

0.25

504

YM

ID01

4M

idw

ayG

eyse

rB

asin

4928

874.

851

4753

.574

.68.

4714

7538

014

1.5

0.00

423

90.

034

04Y

SM

H02

4aS

even

Mile

Hol

e49

5577

4.6

5471

65.5

69.7

7.21

2050

172

161

3.7

0.01

718

71.

167

04Y

SM

J030

aS

mok

eJu

mpe

r49

1809

8.3

5035

49.8

70.5

7.34

2130

072

00.

042

223

2.41

704

YH

RT

036

Hea

rtLa

ke49

0582

6.8

5380

70.8

69.3

8.93

1500

251

146

1.06

0.00

830

80.

654

04Y

HR

T03

9H

eart

Lake

4904

299.

053

9046

.284

.89.

0616

6025

410

40.

446

0.00

917

70.

035

04Y

SH

O04

1aS

hosh

one

4910

920.

551

6275

.777

.06.

4911

2225

150

0.4

0.01

825

64.

583

04Y

SH

O04

4S

hosh

one

4910

969.

251

5601

.580

.98.

3213

6546

036

0.6

0.00

527

50.

072

04Y

SH

O04

6S

hosh

one

4910

961.

651

5877

.072

.58.

4914

0238

235

0.82

40.

0129

20.

243

04Y

BE

C05

0B

echl

er49

0364

1.8

5096

96.1

80.6

7.81

1003

454

210.

178

0.03

717

10.

065

04Y

CA

S05

5aC

asca

deC

orne

r49

0556

6.2

4967

45.8

52.6

5.63

160

6611

00.

086

94.8

0.00

0

a.S

ites

whe

reS

ulfu

rihyd

roge

nibi

umse

quen

ces

wer

ede

tect

ed.

b.

Nor

thin

gan

dE

astin

ggi

ven

inU

TM

s,N

AD

1983

,G

rid12

.c.

As

HC

O3- .

Con

cent

ratio

ngi

ven

asm

gl-1

unle

ssin

dica

ted

othe

rwis

e.

2 C. Takacs-Vesbach, K. Mitchell, O. Jackson-Weaver and A.-L. Reysenbach

© 2008 The AuthorsJournal compilation © 2008 Society for Applied Microbiology and Blackwell Publishing Ltd, Environmental Microbiology

among the sites we assayed. The level of genetic varia-tion among the eight sites was computed by Weir andCockram’s FST parameter (Wier and Cockerham, 1984;Ronquist, 1997), which is a measure of differentiationamong samples. All FST values were large and signifi-cantly different, indicating that each of the sites wasgenetically distinct (P < 0.05; FST values ranged from0.420 to 0.938, Table 2). The amount of genetic variationamong sites is consistent with a hypothesis of isolationand diversification. Our FST values were in the same rangeas values computed between regions separated by 254–6305 km in a biogeographical study of the archaeonSulfolobus (Whitaker et al., 2003), and higher than valuesreported for bacteria associated with recent volcanicdeposits 42–300 years old (Dunfield and King, 2004;2005).

Spatial differentiation among the surveyed sites is sup-ported by phylogenetic analysis. Sequences from six geo-graphically distinct thermal areas of the park (Mammoth,Hot Spring Basin, Seven Mile Hole, Smoke Jumper, Shos-hone and Cascade Corner) formed clades well supportedby bootstrap analysis regardless of whether parsimony,maximum likelihood or neighbor-joining algorithms wereused to construct the tree. The remainder of the treetopology was composed of sequences that did not form

geographically distinct clades, which suggests that thesesequences are cosmopolitan or that dispersal may occurbetween these sites (Fig. 1).

Unlike previous biogeographical studies of micro-organisms, we did not find that geographical distance orgeochemistry correlated significantly with the distributionof sequence types (Mantel test of FST values or geneticdistance, compared with geographical distance and majorion chemistry, P = 0.174, pH, temperature, alkalinity, con-ductivity and sulfate combined, P = 0.691 or temperaturealone, P = 0.231). From these results it would seem thatthere was no geographical or ecological pattern in thedistribution of Sulfurihydrogenibium. However, a plot ofgenetic divergence as a function of geographic distanceindicated a strong linear relationship (r2 = 0.55, P <0.0001) among sites up to 49 km apart, compared with amuch weaker pattern across all sites (r2 = 0.13 P <0.0001, Fig. 2). Recognizing that this pattern may resultbecause environmental parameters may be autocorre-lated to distance at sites < 49 km, we repeated the Manteltests described above for these sites, but did not finda significant relationship (P = 0.128–0.882). Therefore,these genetic–geographic patterns suggest that (i) dis-persal is largely responsible for the genetic divergence weobserved at distances < 49 km, and (ii) there is a thresh-

Table 2. FST values estimated from 16S rRNA gene sequences and tested for significance against 1000 randomized bootstrap replicates withARLEQUIN 3.1.

FST

Between areas within regionsInside the calderas Lonestar Shoshone 0.420

Smoke Jumper 0.642Cascade Corner 0.651

Shoshone Smoke Jumer 0.682Cascade Corner 0.691

Outside the calderas Seven Mile Hole Hot Spring Basin 0.468Joseph’s Coat 0.666Mammoth 0.674

Hot Spring Basin Joseph’s Coat 0.672Mammoth Hot Spring Basin 0.680

Joseph’s Coat 0.914Between regionsInside:outside Lonestar Seven Mile Hole 0.422

Hot Spring Basin 0.430Joseph’s Coat 0.620Mammoth 0.632

Shoshone Seven Mile Hole 0.459Hot Spring Basin 0.466Joseph’s Coat 0.663Mammoth 0.672

Smoke Jumper Seven Mile Hole 0.684Hot Spring Basin 0.690Mammoth 0.894Joseph’s Coat 0.926

Cascade Corner Seven Mile Hole 0.693Hot Spring Basin 0.670Mammoth 0.903Joseph’s Coat 0.938

Significant (P < 0.001) differentiation was detected between all sites.

Calderas delineate biogeographic provinces 3


old beyond which dispersal no longer dominates com-munity assembly (although some dispersal is evidentamong sites separated by > 49 km, e.g. the ‘cosmopoli-tan’ sequences in Fig. 1 and the less divergent sequencesin Fig. 2). This led us to explore alternate explanations for

the distribution of sequence types, especially among sitesgreater than 49 km apart.

Geographic barriers and geological events offer twopotential explanations for the abrupt breakdown of thegenetic divergence–geographic distance relationship for

Fig. 1. Phylogenetic tree of sequencesdetected in this study. Trees were constructedby maximum parsimony, neighbor-joining andmaximum likelihood methods, which resultedin the same overall topology. Tree given wasgenerated by maximum parsimony, numbersat nodes represent bootstrap proportions forclades with significant support based on 1000or 100 resamplings for maximum likelihood(parsimony/neighbor-joining/maximumlikelihood, the asterisks represent bootstrapproportions below 50%). Tree was rooted withAquifex aeolicus (AJ309733) andThermocrinis ruber (AJ005640). Accessionnumbers for the remainder of sequencesanalysed are DQ906005 to DQ906039 andAJ113542 (pBB Calcite Springs clone). Scalebar represents 0.5 fixed mutations per 100nucleotide positions. Clades that werecomprised solely of sequences from a singlegeographic region are indicated by the barswith region names on the side of the figure.Sequences that did not cluster geographicallywere interpreted to be cosmopolitan.

Fig. 2. Pairwise comparisons of uncorrectedgenetic divergence plotted against geographicdistance (km). Regression line was fitted todata points < 49 km apart (r2 = 0.55,P < 0.001). Dashed portion of line is given toillustrate the lack of correlation for distances> 49 km apart. Note that sequencecomparisons with low divergence from sites> 49 km apart result from the cosmopolitansequence types found in Fig. 1.



the more distantly separated sites. However, there wereno apparent geographic barriers or corridors, such asmountain ranges or rivers, that explained our data, nordid the prevailing south-west wind across Yellowstone(Houston, 1973) or rhyolite lava flows and faults (Chris-tiansen, 2001) correlate with the distributions. Subterra-nean dispersal is one factor that is not possible to assess,but is not believed to be important across the Yellowstoneecosystem because of the ring fracture system formedby the calderas, the high temperatures at shallow depth(~350°C) and lack of hydraulic connectivity betweenrecharge zones (Fournier, 1989; Bonheyo et al., 2005).The eight sites where Sulfurihydrogenibium sequenceswere detected occur in either the northern or southernregion of Yellowstone, suggesting a latitudinal trend.However, analysis of molecular variance (AMOVA) indi-cated no genetic structure among sites when placed intonorth or south groups (Fixation index = -0.004, P =0.565). More of the molecular variation among sites couldbe attributed to between and within site variation, ratherthan variation between north and south groups of sites.

Inspection of Yellowstone geological maps revealedthat the distance between sample sites that are inside andoutside of the historic caldera boundaries was greaterthan 49 km, which suggested a relationship betweengenetic divergence among sequences and the calderas.Most geothermal activity in Yellowstone occurs withincalderas that formed from volcanic eruptions during thepast 2 million years. The most recent was 640 000 yearsago, following earlier eruptions 1.3 and 2 million years ago(the 1.3 million year caldera lies to the west of the Parkboundaries, Fournier, 1989). We found a significantrelationship (c2: 9.7, P = 0.002, Pearson’s phi = 0.113,n = 528) between the level of genetic divergence amongthe sequences (i.e. pairwise comparison of sequences

that had a low < 1% and high > 1% difference) and theorigin of the sequence with respect to the calderas (i.e.inside or outside of the calderas). Four sites were insidethe caldera (Cascade Corner, Smoke Jumper, Lone Starand Shoshone) and four sites were outside the calderas(Mammoth, Seven Mile Hole, Joseph’s Coat and HotSpring Basin). The c2 result was confirmed by performinga Monte Carlo simulation upon the contingency table(5000 permutations, P = 0.001). Randomization proce-dures such as Monte Carlo provide a means for verifyinga statistical pattern that is not dependent on the number ofsamples or any of the underlying assumptions of a para-metric test such as the c2 (Eddington, 1995). Examinationof the pairwise comparison of the sequences showed thatthe more divergent sequences most often resulted whenone sequence was from outside the calderas (Fig. 3) andgenetic divergence was the greatest when sites on oppo-site sides of the caldera boundaries were compared.Thus, it appears that historic volcanic eruptions haveleft their imprint on the genetic variation of thermophilicSulfurihydrogenibium within Yellowstone. Calderas, sur-face expressions of past volcanic activity delineate bio-geographical provinces among these microorganisms,although dispersal at sites < 49 km apart also seems toplay a role in the distribution of Yellowstone Sulfurihydro-genibium sequences.

The phylogeographic history of a taxon may be inferredfrom the distribution of the most basal members. Twocontrasting conclusions have been drawn: either the mostbasal members should occupy the region of its origin orthey survive only in the peripheral areas (reviewed inBriggs, 2004). We reconstructed the history of Yellow-stone Sulfurihydrogenibium sequence distribution by adispersal vicariance approach based on the distributionand phylogeny of the 16S rRNA gene sequences using

Fig. 3. Plot of observed frequencies used inc2 test. Number of pairwise comparisons of16S rRNA gene sequences with low and highgenetic divergence from sites found within thecalderas (Same Region) or when at least oneof the sites was located outside the calderas(Different Region). Four sites were inside thecalderas (Cascade Corner, Smoke Jumper,Lone Star and Shoshone) and four sites wereoutside the calderas (Mammoth, Seven MileHole, Joseph’s Coat and Hot Spring Basin). Asignificant relationship was found betweenlevel of genetic divergence and geographicposition with respect to calderas (c2: 9.7,n = 528, P = 0.002) and was supported by aMonte Carlo simulation based on 5000permutations (P = 0.001).



the DIVA software package (Ronquist, 1997). DIVA, aparsimony method, was used because it does not makeassumptions about the existence or shape of a bio-geographic pattern and is appropriate even when arearelationships may be reticulate, rather than hierarchic.Furthermore, DIVA searches for optimal reconstruction ofancestral distributions by assuming vicariant speciation,but incorporating the potential for dispersal and extinctionevents in shaping contemporary distributions. The DIVAanalysis suggested two possible ancestral distributionsof the Sulfurihydrogenibium sequences: (i) within theyoungest caldera, specifically the Shoshone spring, or (ii)within both the youngest caldera and totally outside thecalderas, specifically the Shoshone and Mirror Plateausprings. Shoshone appears to be a special case within thecalderas; all the other springs within the calderas werefounded by dispersal events from outside the calderas.Additionally, the DIVA analysis indicated that three dis-persal events were responsible for the distribution(Fig. 4). One dispersal event occurred from out of thecalderas to the ancestor of the communities at CascadeCorner and Smoke Jumper, followed by a second dis-persal back out of the caldera to Mammoth. The thirddispersal was from out of the caldera to Lone Star. Eitherthe ancestor of the Shoshone sequences is now extincteverywhere except Shoshone or it now exists in someother spring that has not been surveyed yet.

One scenario that could explain the DIVA results andthat is consistent with our hypothesis that the calderasdelineate biogeographical provinces for Sulfurihydro-genibium sequences in Yellowstone is that with each ofthe volcanic eruptions in Yellowstone, ancestral thermo-philes went extinct within the calderas and as new springsformed, they were subsequently colonized from periph-eral sites that survived outside the calderas. While werecognize that infinite dispersal and extinction events maynot be accounted for in the static approach of using phy-logeographic analysis tools such as DIVA, this approachprovides a means for developing and testing phylogeo-graphic hypotheses of microbial distributions. An alternateexplanation for the relationship between Sulfurihydro-genibium sequence distribution and the calderas is thatthis observation reflects some as yet unrecognized under-lying factor that is autocorrelated with the outline of thecalderas.

Environmental factors appear to have little influence onthe distribution of Sulfurihydrogenibium in the thermalsprings that were the focus of this study. This finding maybe an artefact of our site selection (near-neutral thermalsprings only), or the limited number of sites assayed inthis study. We recognize that the patterns we havedetected are based on the current data set only, but thedata and analyses presented here provide strong evi-dence that Yellowstone calderas delineate Sulfurihydro-

Fig. 4. Map of sampling sites whereSulfurihydrogenibium 16S rRNA genesequences were detected. Exact geographiclocation of the sites is given in Table 1. Greenand yellow lines represent the approximatetopographic rim of the 640 ka and 2 mycalderas today respectively. Ring fracture andrim of the actual calderas lie within theseboundaries. Therefore, Seven Mile Hole andJoseph’s Coat are classified as outside of thecalderas. Four sites were inside the caldera(Cascade Corner, Smoke Jumper, Lone Star,and Shoshone) and four sites were outsidethe calderas (Mammoth, Seven Mile Hole,Joseph’s Coat and Hot Spring Basin). Thesites circled on the map indicate locations thatare phylogenetically similar and collectivelyare areas of genetic sources and sinks(i.e. the Mirror plateau sequences clusteredphylogenetically and were the ancestralsource of the population now found in theLone Star site). Backbone of the overallphylogeny among the sequences is used toillustrate the biogeographic history of the sitessampled, including dispersal events (redlines), inferred by DIVA analysis.



genibium biogeographical provinces. Although futureanalysis of additional sites will be necessary to furtherexplore our findings, this report provides a framework forhypothesis development and future study. The precisemechanism by which Yellowstone calderas form barriersbetween the thermal springs is not clear, but Sulfurihydro-genibium sequences from these regions appear to reflectthe historical events that formed them. If similar rela-tionships are found for other groups of thermophiles inYellowstone, then vicariance rather than a barrier to dis-persal would be indicated for thermal communities.Finally, our study provides evidence that a legacy of geo-logical history, which has been instrumental in describingthe biogeography of macroorganisms, is also evidentamong present-day distributions of microorganisms.

Experimental procedures

Sampling sites

The microbial communities of 18 thermal springs withinYellowstone National Park were sampled during 2003 and2004 (Table 1). Prior to collecting biomass, spring water wascollected for geochemical analysis by sampling the waterdirectly overlying the biomass. All water was filtered (0.2 mm)and preserved as appropriate for the analysis to beperformed. Geochemical analysis, including anions, cations,trace metals, sulfide and various gases, was conducted usingstandard USGS methods (McCleskey et al., 2005). Environ-mental DNA was extracted from samples as described pre-viously (Auchtung et al., 2006).

Phylogenetic analysis of 16S rRNA gene sequences

Amplification of the 16S rRNA gene from environmentalDNAs was performed with universal primers 8F (5′-AGAGTTTGAT-CCTGGCTCAG-3′) and 1492R (5′-GGTTAC CTT GTT ACG ACT-3′) and cloned using the TOPO-TAcloning kit (Invitrogen, Carlsbad, CA). Libraries werescreened (96 clones each) by restriction fragment lengthpolymorphism analysis and clones with unique patternswere fully sequenced (both strands, 2X coverage).Assembled sequences were submitted to Check Chimera(http://rdp8.cme.msu.edu/cgis/chimera.cgi) to test for theexistence of PCR artefacts. GenBank accession numbers ofthe sequences determined in this study are DQ906005 toDQ906039. The 16S rRNA gene sequences were alignedwith published sequences from Thermocrinis ruber(AJ005640), Aquifex pyrophilus (M83548), and the uncul-tured Yellowstone relative, pBB (AJ113542). The alignmentwas manually checked using established secondary struc-ture and conserved sequence regions of the 16S rRNA,thus ensuring that only homologous regions were com-pared. Of the approximately 1600 nucleotides sequencedfor each sequence type, 1446 bases were used in the dis-tance matrices, phylogenetic analysis and ARLEQUIN analy-sis; sites that were not conserved in more than 50% of thealigned sequences or that were ambiguously aligned were

not considered. Phylogenetic tree topologies were exploredby parsimony, neighbor-joining and maximum likelihoodanalyses using heuristic search in PAUP (Swofford, 2003). Ageneral time reversible model, plus gamma = 0.5 was usedfor the neighbor-joining and maximum likelihood analyses.Bootstrap proportions were determined on the final treefrom 1000 resamplings of the data (100 resamplings for themaximum likelihood tree). We were concerned that the highlevel of within-site sequence variation was contributing toour results and repeated all analyses possible using onerepresentative sequence from each site, which did notchange or dampen the conclusions made here. Additionally,we conducted the complete analysis presented here usingsequences downloaded from public sequence databasesfrom other Yellowstone sites such as Joseph’s Coat (W.Inskeep, unpublished) and Washburn and Obsidian springs(Spear et al., 2005). Again, the overall results of our analy-ses were the same, but these data were not included herebecause they tended to heavily skew the data set towardssites outside the calderas. Furthermore, the methods foridentifying unique sequences among our data sets were notidentical and a portion of the database sequences were notcompletely sequenced. Although our results are based onone sample per spring, Yellowstone Sulfurihydrogenibium(Reysenbach et al., 2005) communities are known to betemporally stable and the sequence types we detectedwere consistent with previous studies (Barns et al., 1994;Hugenholtz et al., 1998; Reysenbach et al., 2000; Bonheyoet al., 2005; Spear et al., 2005).

Geographic positioning and visualization

Geographic location of sample sites was determined inthe field using a Trimble GeoXM (< 3 m accuracy) and dif-ferentially corrected to the GPS base station at Old Faithfulusing Pathfinder Office. Caldera layers were obtainedfrom the Yellowstone Volcano Observatory (http://volcanoes.usgs.gov/yvo/) and represent the approximate topographicrim of the 640 ka and 2 my calderas today. The ring fractureand actual rim of the calderas lie within these boundaries(J.B. Lowenstern, USGS, pers. comm.). Therefore, SevenMile Hole and Joseph’s Coat are classified as outside thecalderas.

Statistical analysis

Chi-squared test was performed on uncorrected pairwisesequence differences (1446 bp) grouped into categories(columns) that were distinguished by whether the twosequences came from sites within the same region (inside –Cascade Corner, Smoke Jumper, Lone Star and Shoshone,or outside of the calderas – Mammoth, Seven Mile Hole,Joseph’s Coat and Hot Spring Basin) or different regionswhere at least one sequence was from outside the calderasand into groups of low (< 1%) and high (> 1%) levels ofgenetic divergence (rows). Monte Carlo simulations wereperformed to estimate a significance level of the chi-squared statistic. Contingency tables with the same mar-ginal sums were generated 5000 times and the computedchi-squared statistic was compared with the observedchi-squared statistic. Site pairwise differences (FST



values, shown in Table 2) and AMOVA were performed inARLEQUIN version 3.1 (Schneider et al., 2005), a softwareprogram for population genetics analysis that is appropriatefor FST estimation regardless of the ploidy of an organism(L. Excoffier, pers. comm.). FST values were estimated foreach site and were tested for significance against 1000randomized bootstrap resamplings. The Mantel test wasperformed on matrices of (i) FST values or a matrix of uncor-rected distance of one sequence representative from eachsite, (ii) geographical distance between sites and (iii)geochemistry. Geochemistry included major ions, pH, tem-perature, conductivity, alkalinity, and sulfate or temperaturealone. These parameters were selected using non-metricmultidimensional scaling analysis and principal componentsanalysis of over 40 geochemical parameters measured forthe sites, and by examining a piper diagram of major iongeochemistry.

History of the observed distribution was reconstructedusing a dispersal vicariance approach with the DIVA soft-ware package (Ronquist, 1997) based on the distributionand phylogeny of the 16S rRNA gene sequences (Fig. 1).Biogeographic methods most often search for a singlebranching relationship among areas of endemism althoughthe sites may have different histories and reticulate relation-ships. Additionally, dispersal and extinction are rarely incor-porated into these models. DIVA is a unique tool in that itdoes not rely on area cladograms. DIVA searches foroptimal reconstruction of ancestral distributions by assum-ing vicariant speciation, but incorporates the potential fordispersal and extinction events in shaping contemporarydistributions. Ancestral states of a character at each nodeare inferred based on parsimony, but unlike other methods,it does not restrict widespread distributions to terminals orforce ancestral distributions to one area. As with all biogeo-graphic inference methods, it is a static approach based onthe data at hand. The addition of data from other sites mayradically change the outcome, but nevertheless, DIVA is arobust hypothesis testing tool. In this case the ancestralstate of interest was the geographic location of the ances-tor. The three possible states were (i) within the 2 mycaldera, (ii) within the 600 k caldera or (iii) outside of thecalderas. The phylogenetic tree in Fig. 1 was used for theanalysis, each terminal node was assigned one of the threepossible geographic states. This kind of analysis may, andin this case did, result in different but equally parsimoniousstates at a given node.

Acknowledgements

We thank J.H. Brown for comments and discussions regard-ing the manuscript. The Yellowstone Center for Resources(NPS) provided logistical and GIS support. This work wasfunded by a grant from the National Science Foundation’s(NSF) Biotic Surveys and Inventories programme (DEB02030677) to C.T.-V. and A.-L.R., a UNM RAC to C.T.-V., andNSF PEET (0328326) to A.-L.R. An American Society forMicrobiology undergraduate fellowship provided additionalfunding to O.J.-W., and the UNM sequencing facility is oper-ated in part by a grant from the National Institutes of Health(Grant 1P20RR18754).

References

Auchtung, T.A., Takacs-Vesbach, C.D., and Cavanaugh,C.M. (2006) 16S rRNA phylogenetic investigation of thecandidate division ‘Korarchaeota’. Appl Environ Microbiol72: 5077–5082.

Barns, S.M., Fundyga, R.E., Jeffries, M.W., and Pace, N.R.(1994) Remarkable archaeal diversity detected in a Yellow-stone National Park hot spring environment. Proc NatlAcad Sci USA 91: 1609–1613.

Bonheyo, G.T., Frias-Lopez, J., and Fouke, B.W. (2005) Atest for airborne dispersal of thermophilic bacteria fromhot springs. In Geothermal Biology and Geochemistry inYellowstone National Park. Inskeep, W.P., and McDermott,T.R. (eds). Bozeman, MT, USA: Thermal Biology Institute,pp. 327–340.

Briggs, J.C. (2004) Older species: a rejuvenation on coralreefs? J Biogeogr 31: 525–530.

de Bruyn, M., Nugroho, E., Hossain, M.M., Wilson, J.C., andMather, P.B. (2005) Phylogeographic evidence for theexistence of an ancient biogeographic barrier: the Isthmusof Kra Seaway. Heredity 94: 370–378.

Casamayor, E.O., Massana, R., Benlloch, S., Ovreas, L.,Diez, B., Goddard, V.J., et al. (2002) Changes inarchaeal, bacterial and eukaryal assemblages along asalinity gradient by comparison of genetic fingerprintingmethods in a multipond solar saltern. Environ Microbiol 4:338–348.

Christiansen, R.L. (2001) The Quaternary and Pliocene Yel-lowstone Plateau Volcanic Field of Wyoming, Idaho, andMontana. Geology of Yellowstone National Park, Profes-sional Paper 729-G. Reston, VA, USA: Department of theInterior, U.S. Geological Survey.

Dunfield, K.E., and King, G.M. (2004) Molecular analysis ofcarbon monoxide-oxidizing bacteria associated with recentHawaiian volcanic deposits. Appl Environ Microbiol 70:4242–4248.

Dunfield, K.E., and King, G.M. (2005) Analysis of the distri-bution and diversity in recent Hawaiian volcanic deposits ofa putative carbon monoxide dehydrogenase large subunitgene. Environ Microbiol 7: 1405–1412.

Eddington, E.S. (1995) Randomization Tests. New York, NY,USA: Marcel Dekker.

Fournier, R.O. (1989) Geochemistry and dynamics of theYellowstone National Park hydrothermal system. Annu RevEarth Planet Sci 17: 13–53.

Franklin, R.B., Taylor, D.R., and Mills, A.L. (1999) The distri-bution of microbial communities in anaerobic and aerobiczones of a shallow coastal plain aquifer. Microbial Ecol 38:377–386.

Hedges, S.B., Parker, P.H., Sibley, C.G., and Kumar, S.(1996) Continental breakup and the ordinal diversificationof birds and mammals. Nature 381: 226–229.

Houston, D.B. (1973) Wildfires in northern YellowstoneNational Park. Ecology 54: 1111–1117.

Hugenholtz, P., Pitulle, C., Hershberger, K.L., and Pace, N.R.(1998) Novel division level bacterial diversity in a Yellow-stone hot spring. J Bacteriol 180: 366–376.

Knittel, K., Losekann, T., Boetius, A., Kort, R., and Amann, R.(2005) Diversity and distribution of methanotrophicarchaea at cold seeps. Appl Environ Microbiol 71: 467–479.



McCleskey, R.B., Ball, J.W., Nordstrom, D.K., Holloway,J.M., and Taylor, H.E. (2005) Water Chemistry Data forSelected Hot Springs, Geysers, and Streams in Yellow-stone National Park, Wyoming, 2001–2002: U.S. Geologi-cal Survey Open-File Report 2004–1316. Boulder, CO,USA: U.S. Department of the Interior, U.S. GeologicalSociety.

Martiny, J.B., Bohannan, B.J., Brown, J.H., Colwell, R.K.,Fuhrman, J.A., Green, J.L., et al. (2006) Microbial bioge-ography: putting microorganisms on the map. Nat RevMicrobiol 4: 102–112.

Morris, R.M., Rappe, M.S., Connon, S.A., Vergin, K.L.,Siebold, W.A., Carlson, C.A., and Giovannoni, S.J. (2002)SAR11 clade dominates ocean surface bacterioplanktoncommunities. Nature 420: 806–810.

Papke, R.T., Ramsing, N.B., Bateson, M.M., and Ward, D.M.(2003) Geographical isolation in hot spring cyanobacteria.Environ Microbiol 5: 650–659.

Reysenbach, A.-L., Banta, A., Civello, S., Daly, J., Mitchell,K., Lalonde, S., et al. (2005) The Aquificales in YellowstoneNational Park. In Geothermal Biology and Geochemistry inYellowstone National Park. Inskeep, W.P., and McDermott,T.R. (eds). Bozeman, MT, USA: Thermal Biology Institute,pp. 129–142.

Reysenbach, A.L., Ehringer, M., and Hershberger, K. (2000)Microbial diversity at 83 degrees C in Calcite Springs,Yellowstone National Park: another environment where theAquificales and ‘Korarchaeota’ coexist. Extremophiles 4:61–67.

Ronquist, F. (1997) Dispersal-vicariance analysis: a new

approach to the quantification of historical biogeography.Syst Biol 46: 195–203.

Schneider, S., Kueffer, J.-M., Roessli, D., and Excoffier, L.(2005) ARLEQUIN Ver. 3.0: an integrated software packagefor population genetics data analysis. Evol BioinformOnline 1: 47–50.

Spear, J.R., Walker, J.J., McCollom, T.M., and Pace, N.R.(2005) Hydrogen and bioenergetics in the Yellowstonegeothermal ecosystem. Proc Natl Acad Sci USA 102:2555–2560.

Swofford, D.L. (2003) PAUP*. Phylogenetic Analysis UsingParsimony (*and Other Methods). Version 4. Sunderland,MA, USA: Sinauer Associates.

Takacs, C.D., Ehringer, M., Favre, R., Cermola, M., Eggerts-son, G., Palsdottir, A., and Reysenbach, A.-L. (2001) Phy-logenetic characterization of the blue filamentous bacterialcommunity from an Icelandic geothermal spring. FEMSMicrobiol Ecol 35: 123–128.

Wares, J.P., and Cunningham, C.W. (2001) Phylogeographyand historical ecology of the North Atlantic intertidal.Evolution 55: 2455–2469.

Whitaker, R.J., Grogan, D.W., and Taylor, J.W. (2003)Geographic barriers isolate endemic populations ofhyperthermophilic archaea. Science 301: 976–978.

Wier, B.S., and Cockerham, C.C. (1984) EstimatingF-statistics for the analysis of population structure. Evolu-tion 38: 1358–1370.

de Wit, R., and Bouvier, T. (2006) ‘Everything is everywhere,but, the environment selects’; what did Baas Becking andBeijerinck really say? Environ Microbiol 8: 755–758.



Volcanic calderas delineate biogeographic provinces among ...

Documents

Transcript of Volcanic calderas delineate biogeographic provinces among ...