6. Y. Wu et al., Cell 126, 375 (2006).
7. A. Marson et al., Nature 445, 931 (2007).
8. Y. Zheng et al., Nature 445, 936 (2007).
9. M. Ono et al., Nature 446, 685 (2007).
10. P. Waterhouse et al., Science 270, 985 (1995).
11. E. A. Tivol et al., Immunity 3, 541 (1995).
12. The Wellcome Trust Case Control Consortium, Nature
447, 661 (2007).
13. D. R. Leach, M. F. Krummel, J. P. Allison, Science 271,
15. S. Read et al., J. Immunol. 177, 4376 (2006).
16. D. M. Sansom, L. S. Walker, Immunol. Rev. 212, 131
17. Materials and methods are available as supporting
material on Science Online.
18. S. Kawamoto et al., FEBS Lett. 470, 263 (2000).
19. M. Ono, J. Shimizu, Y. Miyachi, S. Sakaguchi, J. Immunol.
176, 4748 (2006).
20. Y. Y. Wan, R. A. Flavell, Nature 445, 766 (2007).
21. J. Shimizu, S. Yamazaki, S. Sakaguchi, J. Immunol. 163,
23. C. Oderup, L. Cederbom, A. Makowska, C. M. Cilio,
F. Ivars, Immunology 118, 240 (2006).
24. S. Yamazaki, K. Inaba, K. V. Tarbell, R. M. Steinman,
Immunol. Rev. 212, 314 (2006).
25. R. J. DiPaolo et al., J. Immunol. 179, 4685 (2007).
26. Y. Onishi et al., Proc. Natl. Acad. Sci. U.S.A. 105, 10113
27. We thank M. Ono for discussion and R. Ishii and
M. Matsushita for technical assistance. This work was
supported by Grants-in-Aid from the Ministry of
Education, Sports and Culture of Japan, Japan Science
and Technology Agency. Z.F. was a Japan Society for the
Promotion of Science fellow, and K.W. was granted a
fellowship by Astra-Zeneca, Loughborough, UK.
Supporting Online Material
Materials and Methods
Figs. S1 to S13
5 May 2008; accepted 15 August 2008
Reveals a Single-Species Ecosystem
Deep Within Earth
Dylan Chivian,1,2* Eoin L. Brodie,2,3Eric J. Alm,2,4David E. Culley,5
Paramvir S. Dehal,1,2Todd Z. DeSantis,2,3Thomas M. Gihring,6Alla Lapidus,7
Li-Hung Lin,8Stephen R. Lowry,7Duane P. Moser,9Paul M. Richardson,7
Gordon Southam,10Greg Wanger,10Lisa M. Pratt,11,12Gary L. Andersen,2,3
Terry C. Hazen,2,3,12Fred J. Brockman,13Adam P. Arkin,1,2,14Tullis C. Onstott12,15
DNA from low-biodiversity fracture water collected at 2.8-kilometer depth in a South African
gold mine was sequenced and assembled into a single, complete genome. This bacterium,
Candidatus Desulforudis audaxviator, composes >99.9% of the microorganisms inhabiting the
fluid phase of this particular fracture. Its genome indicates a motile, sporulating, sulfate-reducing,
chemoautotrophic thermophile that can fix its own nitrogen and carbon by using machinery
shared with archaea. Candidatus Desulforudis audaxviator is capable of an independent life-style
well suited to long-term isolation from the photosphere deep within Earth’s crust and offers an
example of a natural ecosystem that appears to have its biological component entirely encoded
within a single genome.
DNA from an environmental sample, a process
called environmental genomics or metagenomics
(1–8). This approach allows us to identify mem-
bers of microbial communities and to character-
ize the abilities of the dominant members even
when isolation of those organisms has proven
intractable. However, with a few exceptions (5, 7),
assembling complete or even near-complete ge-
nomes for a substantial portion of the member
speciesis usually hampered by the complexity of
natural microbial communities.
In addition to elevated temperatures and a
lack of O2, conditions within Earth’s crust at
depths >1 km are fundamentally different from
those of the surface and deep ocean environ-
ments. Severe nutrient limitation is believed to
result in cell doubling times ranging from 100s
to 1000s of years (9–11), and as a result sub-
surface microorganisms might be expected to
reduce their reproductive burden and exhibit the
streamlined genomes of specialists or spend
most of their time in a state of semi-senescence,
waiting for the return of favorable conditions.
more complete picture of life on, and
even in, Earth has recently become
possible by extracting and sequencing
Such microorganisms are of particular interest
because they permit insight into a mode of life
independent of the photosphere.
One bacterium belonging to the Firmicutes
phylum (Fig. 1A), which we herein name Can-
didatus Desulforudis audaxviator, is prominent
in small subunit (SSU or 16S) ribosomal RNA
(rRNA) gene clone libraries (11–14) from almost
all fracture fluids sampled to date from depths
greater than 1.5 km across the Witwatersrand basin
(covering 150 km by 300 km near Johannesburg,
South Africa). This bacterium was shown in a
previous geochemical and 16S rRNA gene study
(11) to dominate the indigenous microorga-
nisms found in a fracture zone at 2.8 km below
land surface at level 104 of the Mponeng mine
(MP104). Although Lin et al. (11) discovered
that this fracture zone contained the least-diverse
natural free-living microbial community reported
at that time, exceeding the ~80% dominance by
the methanogenic archaeon IUA5/6 of a com-
paratively shallow subsurface community in Idaho
(15), we were nonetheless surprised when the cur-
rent environmental genomics study revealed only
one species was actually present within the frac-
ture fluid. Furthermore, we found that the
genome of this organism appeared to possess
all of the metabolic capabilities necessary for
an independent life-style. This gene complement
was consistent with the previous geochemical
and thermodynamic analyses at the ambient
~60°C temperature and pH of 9.3, which indi-
cated radiolytically generated chemical species as
providing the energy and nutrients to the system
(11), with formate and H2 as possessing the
greatest potential among candidate electron
donors, and sulfate (SO42–) reduction as the
dominant electron-accepting process (11).
DNA was extracted from ~5600 liters of fil-
tered fracture water by using a protocol that has
been demonstrated to be effective on a broad
range of bacterial and archaeal species, includ-
ing recalcitrant organisms (16). A single, com-
plete, 2.35–megabase pair (Mbp) genome was
assembled with a combination of shotgun Sanger
sequencing and 454 pyrosequencing (16). Sim-
ilar to other studies that obtained near-complete
consensus genomes from environmental sam-
ples (5, 17), heterogeneity in the population of
the dominant species as measured with single-
nucleotide polymorphisms (SNP) was quite low,
showing only 32 positions with a SNP observed
1Physical Biosciences Division, Lawrence Berkeley National
Laboratory, Berkeley, CA 94720, USA.2Virtual Institute for
Microbial Stress and Survival, Berkeley, CA 94720, USA.
3Earth Sciences Division, Lawrence Berkeley National Lab-
oratory, Berkeley, CA 94720, USA.4Departments of Biological
and Civil and Environmental Engineering, Massachusetts
Institute of Technology, Cambridge, MA 02139, USA.5Energy
and Efficiency Technology Division, Pacific Northwest National
Laboratory, Richland, WA 99352, USA.
Oceanography, Florida State University, Tallahassee, FL
32306, USA.7Genomic Technology Program, U.S. Depart-
ment of Energy (DOE) Joint Genomics Institute, Berkeley, CA
94598, USA.8Department of Geosciences, National Taiwan
University, Taipei 106, Taiwan.
Ecosystem Sciences, Desert Research Institute, Las Vegas, NV
Western Ontario, London, ON N6A 5B7, Canada11Department
of Geological Sciences, Indiana University, Bloomington, IN
Initiative (IPTAI), NASA Astrobiology Institute, Bloomington, IN
47405, USA.13Biological Sciences Division, Pacific Northwest
National Laboratory, Richland, WA 99352, USA.14Department
of Bioengineering, University of California, Berkeley, CA 94720,
Princeton, NJ 08544, USA.
*To whom correspondence should be addressed. E-mail:
9Division of Earth and
10Department of Earth Sciences, University of
12Indiana Princeton Tennessee Astrobiology
15Department of Geosciences, Princeton University,
VOL 32210 OCTOBER 2008
on November 11, 2008
more than once (table S7), suggesting a recent
selective sweep or other population bottleneck.
The DNA recovered from the filter, assum-
ing the capture of cells and extraction of DNA
from those cells was indeed comprehensive, re-
vealed that this genome represented the only
species present in the fluid phase of the fracture.
Of the ~0.1% of microbial reads not belonging
to D. audaxviator (Fig. 1, C and D, and tables
S5 and S6), about one-half represented clear
laboratory contamination (table S6), the removal
of which resulted in only 22 of 29,179 Sanger
reads (0.075%) and 59 of 500,008 pyrosequenc-
ing reads (0.012%) that could be from other
microorganisms. Despite precautions taken in
collecting the sample, some of the trace reads
could come from microbial contaminants in the
mine. An upper-bound estimate of the con-
tribution of any microorganism other than D.
audaxviator to the community (table S6) offered
at most only five Sanger reads (0.017%) corre-
sponding to g-Proteobacteria and at most nine
pyrosequencing reads (0.0018%) corresponding
to a-Proteobacteria, both of which are common
in the mining water (11, 14). Even if these
Proteobacteria were not contaminants, it is
unlikely that D. audaxviator, and indeed the
functioning of the ecosystem, is metabolically
dependent on organisms that would be out-
numbered by about 5000 to 1 (or about 50,000
to 1 from the pyrosequencing data). However,
we could not rule out the presence of
organisms that might adhere to the surfaces
of the fracture or that were smaller than the
0.2-mm filter pore that might play a role in the
MP104 ecosystem, perhaps as reservoirs of
genetic variation (18).
We analyzed the genome of D. audaxviator
with use of MicrobesOnline (www.microbesonline.
org) (19). If D. audaxviator is indeed the solitary
resident of this habitat, then its genome should
contain the complete genetic complement for
maintaining the biological component of the
ecosystem, which would prohibit the extreme
reduction of its genome. The genome (Table 1)
at 2.35 Mbp was smaller than the 3 Mbp of its
nearest sequenced relative, Pelotomaculum ther-
mopropionicum. It contained 2157 predicted
protein coding genes, more than found in
streamlined free-living microorganisms, which
typically have fewer than 2000 genes (20). We
found all of the processes necessary for life en-
coded within the genome, including energy me-
tabolism, carbon fixation, and nitrogen fixation.
Consistent with the thermodynamic evalu-
ation (11) that SO42–offers the most energet-
ically favorable electron acceptor, the genome
possesses the capacity for dissimilatory sulfate
reduction (DSR) (Figs. 2 and 3 and table S13)
with a gene repertoire like that of other SO42–-
reducing microorganisms (21). These genes are
present in a set of operons (labeled SR1 to SR11
type sulfate adenylyltransferase (Sat) (fig. S5)
and a H+-translocating pyrophosphatase, both
of which appear to be a consequence of hori-
zontal gene transfer (HGT). High potential elec-
trons probably enter primarily via the activity
of a variety of hydrogenases acting upon H2
Carbon assimilation may be from a variety of
sources depending on local conditions. The ge-
nome contains sugar and amino acid transporters
(Fig. 3 and table S20), suggesting that, at loca-
tions where biodensity is high, heterotrophic
sources could be used, including recycling of
dead cells. At MP104, where biodensity is low,
carbon is fixed from inorganic sources. D.
audaxviator appeared not to use the reverse
TCA cycle (table S23) but did have all the ma-
chinery of the acetyl–coenzyme A (CoA) syn-
thesis (Wood-Ljungdahl) pathway (22, 23), which
uses carbon monoxide dehydrogenase (CODH)
for the assimilation of inorganic carbon (Figs. 2
and 3, fig. S7, and table S14). Entry of CO2sub-
strate into the cell may be accomplished by its
anionic species through a putative carbonate aden-
osine triphosphate (ATP)–binding cassette trans-
porter or a putative bicarbonate/Na+symporter
(Fig. 3 and table S20). Formate and CO may
serve as alternate, more direct, carbon sources
in other fractures when sufficiently abundant
The ambient concentration of ammonia in the
fracture water ([NH3] + [NH4+] = ~100 mM) (11)
appears sufficient for D. audaxviator (which has
an ammonium transporter as well as glutamine
Fig. 1. Phylogeny and population structure. (A) Phylogenetic placement of D.
audaxviator based on protein sequences of universal protein families (table S3).
High bootstrap value–supported nodes are indicated with circles. (B) Classifications
of SSU rRNA gene clones from polymerase chain reaction amplification of filter ex-
tract (fig. S3). (C) Proportions of Sanger sequencing reads from shotgun clone library
of filter extract. Reads classified as D. audaxviator by match to assembled genome or
by match to sequenced organisms (table S6). (D) Proportions of 454 pyrosequencing
reads directly from filter extract. Reads classified as D. audaxviator by match to
assembled genome or by match to sequenced organisms (table S6).
Table 1. General features of the D. audaxviator
Genome size (bp)
G+C content (%)
Predicted protein coding genes
Genes without homology to other
Pseudogenes derived from a protein
Average CDS/ORF length (bp)
Longest CDS/ORF length (bp)
Percent of genome protein
rRNA operons (16S-23s-5S)
Transfer RNAs (all amino acids
represented, including selenocysteine)
Other nonprotein coding RNAs
Other phage-associated genes
10 OCTOBER 2008 VOL 322
on November 11, 2008
synthetase) to obtain its nitrogen from ammonia
without resorting to an energetically costly ni-
trogenase conversion of N2to ammonia. None-
theless, a nitrogenase is present in the genome
(Fig. 2 and table S15) with a nifH subunit that is
more similar to archaeal types, including high-
temperature variants (24), than to the nitrogenase
of Desulfotomaculum reducens (figs. S4 and S8).
It may be that D. audaxviator is not always
presented with sufficient amounts of ammonia,
so the versatility provided by the horizontally
acquired nitrogenase may have contributed sub-
stantially to the success of D. audaxviator in
colonizing such habitats.
D. audaxviator shares other genes with archaea
that may confer benefits in extreme environments.
In addition to the unusual nitrogenase and sul-
fate adenylyltransferase, acquisitions by ancestors
of D. audaxviator (table S10) include a second
CODH system (CODH1 in Fig. 2 and fig. S7),
cobalamin biosynthesis protein CobN, and genes
for the formation of gas vesicles. It also has two
clustered regularly interspaced short palindromic
repeat (CRISPR) regions (table S12) that are
used for viral defense (25) and that occur in the
genome with adjacent CRISPR-associated genes
(CAS), some of which are horizontally shared
between D. audaxviator and archaea.
D. audaxviator’s ability to colonize indepen-
dently is also assisted by its possession of all of
the amino acid synthesis pathways (table S21).
Other factors that may confer fitness in this en-
vironment are the ability to form endospores (table
S16) and the potential for it to grow in deeper,
hotter conditions (table S9) than provided by
MP104. D. audaxviator appears capable of
sensing nutrients (table S19) in its environment
and possesses flagellar genes (table S18) to
permit motility along chemical gradients, such
as those that occur at the mineral surfaces of the
fracture (26). One ability that D. audaxviator is
lacking is a complete system for oxygen re-
sistance (table S25), suggesting the long-term
isolation from O2.
The MP104 fracture contains the simplest nat-
ural environmental microbial community yet de-
scribed and has yielded a single, complete genome
of an uncultured microorganism with the use of
environmental genomics. D. audaxviator’s ability
to reduce SO42–grants access to the most en-
ergetically favorable electron acceptor in the
fracture zones of the Witwatersrand basin (27).
Additionally, inherited characteristics of D. audax-
viator, such as motility, sporulation, and carbon
fixation, have been complemented by horizontally
acquired systems frequently found in archaea.
These abilities have enabled D. audaxviator to
colonize the deep subsurface, a process that, unlike
surface habitats which permit more immediate
access, has required fitness throughout the history
of the colonization. This bold traveler (audax
viator) has revealed a mode of life isolated from
for an independent life-style and showing that it is
possible to encode the entire biological component
of a simple ecosystem within a single genome.
Fig. 2. Genome of D. audaxviator, with key genes highlighted. (Innermost
ring) GC skew [average of (G-C)/(G+C) over 10,000 bases, plotted every
1000 bases]. Transition at the top (near dnaA) is origin of replication.
(Second ring) G+C content [average of (G+C) over 10,000 bases, plotted
every 1000 bases], with greater-than-average values (61%) in blue and
below average in red. Below-average G+C regions that result from CRISPR
sequences are indicated in gray. (Third and fourth rings) Predicted protein
coding genes on each strand. Genes with homologs only found within
closest clade species [including open reading frame (ORF)an genes] are in
cyan, genes that are found only within closest clade species and within
archaea (resulting from horizontal transfer) in magenta, and all other
genes in black. (Outer boxes) Genes of interest are shown around the ring
as operons for sulfate reduction (SR), carbon fixation via acetyl-CoA syn-
thesis pathway (CF), and nitrogen fixation (NF). Horizontally acquired genes
shared with archaea specific to D. audaxviator and its nearest relatives are
colored according to the key.
VOL 32210 OCTOBER 2008
on November 11, 2008
References and Notes
1. A. M. Deutschbauer, D. Chivian, A. P. Arkin, Curr. Opin.
Biotechnol. 17, 229 (2006).
2. O. Beja et al., Environ. Microbiol. 2, 516 (2000).
3. M. R. Rondon et al., Appl. Environ. Microbiol. 66, 2541
4. J. C. Venter, Science 304, 66 (2004); published online
4 March 2004 (10.1126/science.1093857).
5. G. W. Tyson et al., Nature 428, 37 (2004).
6. S. G. Tringe, Science 308, 554 (2005).
7. M. Strous et al., Nature 440, 790 (2006).
8. D. B. Rusch et al., PLoS Biol. 5, e77 (2007).
9. T. J. Phelps, E. M. Murphy, S. M. Pfiffer, D. C. White,
Microb. Ecol. 28, 335 (1994).
10. B. B. Jørgensen, S. D’Hondt, Science 314, 932 (2006).
11. L. H. Lin et al., Science 314, 479 (2006).
12. D. P. Moser et al., Appl. Environ. Microbiol. 71, 8773 (2005).
13. D. P. Moser et al., Geomicrobiol. J. 20, 517 (2003).
14. T. M. Gihring et al., Geomicrobiol. J. 23, 415 (2006).
15. F. H. Chapelle et al., Nature 415, 312 (2002).
16. Materials and methods are available as supporting
material on Science Online.
17. V. Zverlov et al., J. Bacteriol. 187, 2203 (2005).
18. M. L. Sogin et al., Proc. Natl. Acad. Sci. U.S.A. 103,
19. E. J. Alm et al., Genome Res. 15, 1015 (2005).
20. S. J. Giovannoni et al., Science 309, 1242 (2005).
21. M. Mussmann et al., J. Bacteriol. 187, 7126 (2005).
22. H. L. Drake, S. L. Daniel, Res. Microbiol. 155, 869 (2005).
23. M. Wu et al., PLoS Genet. 1, e65 (2005).
24. M. P. Mehta, J. A. Baross, Science 314, 1783 (2006).
25. R. Barrangou et al., Science 315, 1709 (2007).
26. G. Wanger, T. C. Onstott, G. Southam, Geomicrobiol. J.
23, 443 (2006).
27. T. C. Onstott et al., Geomicrobiol. J. 23, 369 (2006).
28. L. Lefticariu, L. M. Pratt, E. M. Ripley, Geochim.
Cosmochim. Acta 70, 4889 (2006).
29. We thank J. Banfield and G. Tyson for helpful discussion;
J. Bruckner and B. Baker for assistance with microscopy;
F. Warnecke for advice on 16S fluorescent in situ
hybridization; T. Kieft, G. Zane, and the MicrobesOnline
team (M. Price, K. Keller, and K. Huang) for advice; and
D. Kershaw and colleagues at the Mponeng mine and
AngloGold Ashanti Limited, RSA. This work was part of the
Virtual Institute for Microbial Stress and Survival (http://
vimss.lbl.gov), supported by DOE, Office of Science,
Office of Biological and Environmental Research,
Genomics Program:GTL through contract DE-AC02-
05CH11231 between Lawrence Berkeley National
Laboratory and DOE. This work was also supported by the
NASA Astrobiology Institute through award NNA04CC03A
to the IPTAI Team co-directed by L.M.P. and T.C.O. A.P.A.
received support from the Howard Hughes Medical
Institute. The genome sequence and 16S library
sequences reported in this study have been deposited in
GenBank under the accession numbers CP000860 and
EU730965 to EU731008, respectively.
Supporting Online Material
Materials and Methods
Figs. S1 to S8
Tables S1 to S26
22 January 2008; accepted 11 September 2008
is shown in a cartoon representation, including pathways for sulfate reduction,
nitrogen fixation, and carbon fixation. Signal transduction proteins are reported
includingthe number found in parentheses, with MCP indicating methyl-accepting
chemotaxis proteins; HPK, histidine protein kinases; and RR, response regulators.
Transporters include approximate substrates. Also shown are the radiolytically
generated sources of energy and nutrients for the ecosystem, as detailed in Lin
et al. (11), shown experimentally by Lefticariu et al. (28), and described in (16).
10 OCTOBER 2008VOL 322
on November 11, 2008