De novo metagenomic assembly reveals abundant novel major lineage of Archaea in hypersaline ...

13 Pages·2011·1.28 MB·English

by  Priya Narasingarao

TheISMEJournal(2012)6,81–93
&2012InternationalSocietyforMicrobialEcology Allrightsreserved1751-7362/12
www.nature.com/ismej
ORIGINAL ARTICLE
De novo metagenomic assembly reveals abundant
novel major lineage of Archaea in hypersaline
microbial communities
Priya Narasingarao1,8, Sheila Podell1,8, Juan A Ugalde1, Ce´line Brochier-Armanet2,
Joanne B Emerson3, Jochen J Brocks4, Karla B Heidelberg5, Jillian F Banfield3,6
and Eric E Allen1,7
1MarineBiologyResearchDivision,ScrippsInstitutionofOceanography,UniversityofCalifornia,SanDiego,
La Jolla, CA, USA; 2Universite´ de Provence, Aix-Marseille Universite´, CNRS, UPR 9043, Laboratoire de
Chimie Bacte´rienne, Institut de Microbiologie de la Me´diterrane´e (IFR88), Marseille, France; 3Department
of Earth and Planetary Sciences, University of California, Berkeley, Berkeley, CA, USA; 4Research School
of Earth Sciences, The Australian National University, Canberra, ACT, Australia; 5Department of Biological
Sciences, University of Southern California, Los Angeles, CA, USA; 6Department of Environmental Science,
Policy, and Management, University of California, Berkeley, Berkeley, CA, USA and 7Division of Biological
Sciences, University of California, San Diego, La Jolla, CA, USA
This study describes reconstruction of two highly unusual archaeal genomes by de novo
metagenomicassemblyofmultiple,deeplysequencedlibrariesfromsurfacewatersofLakeTyrrell(LT),
a hypersaline lake in NW Victoria, Australia. Lineage-specific probes were designed using the
assembledgenomestovisualizethesenovelarchaea,whichwerehighlyabundantinthe0.1–0.8lm
size fraction of lake water samples. Gene content and inferred metabolic capabilities were highly
dissimilar to all previously identified hypersaline microbial species. Distinctive characteristics
included unique amino acid composition, absence of Gvp gas vesicle proteins, atypical archaeal
metabolic pathways and unusually small cell size (approximately 0.6lm diameter). Multi-locus
phylogenetic analyses demonstrated that these organisms belong to a new major euryarchaeal
lineage,distantlyrelatedtohalophilicarchaeaofclassHalobacteria.Consistentwiththesefindings,
weproposecreationofanewarchaealclass,provisionallynamed‘Nanohaloarchaea’.Inadditionto
their high abundance in LT surface waters, we report the prevalence of Nanohaloarchaea in other
hypersaline environments worldwide. The simultaneous discovery and genome sequencing of a
novel yet ubiquitous lineage of uncultivated microorganisms demonstrates that even historically
well-characterizedenvironmentscanrevealunexpecteddiversitywhenanalyzedbymetagenomics,
andadvancesourunderstandingoftheecologyofhypersalineenvironmentsandtheevolutionary
historyof the archaea.
TheISME Journal(2012)6, 81–93; doi:10.1038/ismej.2011.78; published online30June 2011
SubjectCategory: integrated genomics and post-genomics approaches in microbial ecology
Keywords: assembly; halophile; hypersaline; metagenome; Nanohaloarchaea
Introduction cingisanappealingrouteforinvestigatingmicrobial
community composition because it provides simul-
Cultivation-independent molecular ecology techni-
taneous insight into phylogenetic composition and
ques currently used tosurvey environmental micro-
metabolic capabilities of uncultivated populations
biotaincludeanalysisofphylogeneticmarkergenes,
(Allen and Banfield, 2005; Wilmes et al., 2009).
targeted functional gene inventories and direct
Gene fragments from individual sequencing reads
sequencing of DNA recovered from environmental
and small assembled contigs can be annotated and
samples (reviewed in Hugenholtz and Tyson, 2008;
assignedtoapproximatephylogeneticbinsbasedon
Wooley et al., 2010). Direct metagenomic sequen-
comparison with databases of known reference
genomes (Mavromatis et al., 2007). However, culti-
Correspondence: EE Allen, Marine Biology Research Division, vation biases limit the phylogenetic and physio-
Scripps Institution of Oceanography, University of California, logical breadth of available reference genomes (Wu
SanDiego,LaJolla,CA92093-0202,USA. et al., 2009). Single cell genomics can potentially
E-mail:[email protected]
broaden genomic databases, but often provides
8Theseauthorscontributedequallytothiswork.
highly fragmented data because of amplification
Received13January2011;revised10May2011;accepted14May
2011;publishedonline30June2011 biases(Lasken,2007;Woykeetal.,2009).Asaresult
DenovometagenomicassemblyofNanohaloarchaea
PNarasingaraoetal
82
ofskewedgenomicrepresentationsinreferencedata extreme hypersaline habitats are dominated by
sets, metagenome analysis methods that rely on halophilic archaea belonging to the monophyletic
previouslydescribedsequenceexamples(forexample, class Halobacteria (phylum Euryarchaeota), includ-
fragmentrecruitment approaches) share an inherent ing members of the genera Haloquadratum,
potential bias against novel findings. This anti- Halobacterium, Halorubrum and Haloarcula (Oren,
novelty bias can be overcome by de novo sequence 2008). Pure isolates of halophilic archaea currently
assembly, which does not rely on external reference include 496 species distributed among 27 genera,
sequences, and can facilitate resolution of phylo- with genome sequence information available for
geny-to-function linkagesforindividualcommunity more than a dozen species (Oren et al., 2009).
members. Yet de novo sequence assembly techni- Numerous cultivation-independent biodiversity
ques are rarely applied to metagenomic sequences surveys have been performed in hypersaline envir-
because of sampling deficiencies and/or computa- onments using PCR amplification of archaeal and
tional challenges (Allen and Banfield, 2005; Baker bacterial16SribosomalRNA(rRNA)genes,aswellas
et al., 2010). direct metagenomic sequencing of community DNA
Habitats characterized by low diversity microbial (Grantetal.,1999;Benllochetal.,2001;Ochsenreiter
communities have proven useful for validating et al., 2002; Burns et al., 2004; Demergasso et al.,
molecular (eco-)systems biology approaches to ex- 2004;Jiangetal.,2006;Maturranoetal.,2006;Mutlu
amine the genetic and functional organization of et al., 2008; Pagaling et al., 2009; Sabet et al., 2009;
native microbial consortia (Tyson et al., 2004; Allen Oh et al., 2010; Rodriguez-Brito et al., 2010). These
andBanfield,2005;Rametal.,2005;Loetal.,2007; studies confirm high abundance of a few dominant
Raes and Bork, 2008; Wilmes et al., 2009). High species with widespread geographical distribution,
salt-impacted habitats are distributed globally in the but the intermittent recovery of atypical, uncon-
form of hypersaline lakes, salt ponds and solar firmed sequence fragments hints at additional,
(marine)salterns,whereevaporativeprocessesresult unrecognized diversity among halophilic archaea
in salt concentrations close to and exceeding satura- (Grantetal.,1999;Lopez-Garciaetal.,2001;Pagaling
tion. These environments contain microbial commu- etal.,2009;Ohetal.,2010;Sime-Ngandoetal.,2010).
nities of intermediate complexity (Oren, 2008), The lure of uncovering biological novelty is a
providing excellent model systems for developing major incentive driving metagenomic investigations
scalable analytical techniques applicable to environ- in many habitats worldwide. This study demon-
ments with greater species richness and evenness. strates that even historically well-characterized
The biochemical and physiological challenges habitats like extreme hypersaline lakes and solar
faced by extremely halophilic organisms have salterns can reveal unexpected genes, metabolic
resulted in unique adaptations to maintain osmotic features and entire lineages overlooked previously.
balance,overcomereducedwateractivitybecauseof The ‘assembly-driven’ community metagenomic
the hygroscopic effects of saturating salt concentra- approach applied in the current study has led to
tions, and deter DNA damage induced by intense the discovery and reconstruction of near-complete
solar irradiation (Bolhuis et al., 2006; Hallsworth genomes for two new archaeal genera representing
et al., 2007). The most extreme halophiles maintain the first members of a previously undescribed
osmoticbalanceusinga‘highsalt-in’strategy,which taxonomic class of halophilic archaea. We demon-
allows intracellular salt concentrations to reach strate that members of this new archaeal class are
levels approximately isosmotic with the external present in high abundance and broadly distributed
environment (Oren, 2008). Microorganisms using in other hypersaline habitats worldwide.
the salt-in strategy not only endure extreme ionic
strength,theyrequireitforgrowth.Althoughsalt-in
adaptation can be energetically more favorable than Materials and methods
transporting salt out and the accumulation of
compatible solutes (Oren, 1999), it requires signifi- Sample collection
cant modifications to the intracellular machinery, Surface water samples (0.3m depth) were collected
including specialized protein amino acid composi- fromLakeTyrrell(LT),Victoria,Australiaandahigh
tions to maintain solubility, structural flexibility, salinity crystallizer pond at South Bay Salt Works,
and water availability necessary for enzyme func- Chula Vista (CV) California. Detailed locations,
tion (Fukuchi et al., 2003; Bolhuis et al., 2008; sampling dates, and physical characteristics of the
Paul et al., 2008; Rhodes et al., 2010). collection sites are provided in Supplementary
The study of microbial populations in extreme Figure S1.
hypersaline environments is well established; the Water samples of 20l each were passed through a
first cultivated halophilic microorganism appeared 20mm Nytex prefilter, followed by sequential filtra-
inBergey’smanualoveracenturyago(Oren,2002a). tion through a series of polyethersulfone, 142mm
Despite the extreme conditions in salt-saturated diameter membrane filters (Pall Corporation, Port
habitats, microbial cell densities often exceed Washington, NY, USA) of decreasing porosities
107cellsml–1 (Oren, 2002b). Although salt-adapted (3mm40.8mm40.1mm) using a peristaltic pump.
organismsderivefromallthreedomainsoflife,most After each stage of filtration, filters were frozen for
TheISMEJournal
DenovometagenomicassemblyofNanohaloarchaea
PNarasingaraoetal
83
future DNA extraction, 16S rRNA gene analysis and Genome annotation
metagenomic sequencing. Aliquots of filtered water J07AB43 and J07AB56 draft genomes were anno-
werefixedwithformaldehyde(7%finalconcentration) tated using the Integrated Microbial Genome Expert
overnight at 41C. Fixed water samples were collected Review service of the Joint Genome Institute
on 0.2mm polycarbonate GTTP filters (Millipore, (Markowitz et al., 2009b). Genome completeness
Billerica,MA,USA)forfluorescenceinsituhybridiza- wasestimatedfortheJ07AB56andJ07AB43scaffold
tion (FISH) and direct count microscopy. groups by comparing genes involved in transcrip-
tion, translation and replication to those identified
Library construction and assembly as highly conserved in previously sequenced
Genomic DNA was extracted from individual, archaeal genomes (Ciccarelli et al., 2006; Wu and
bar-coded 0.8 and 0.1mm filters. Filter-specific Eisen, 2008; Puigbo et al., 2009). Orthologs shared
DNA libraries were constructed with insert sizes of between the J07AB43 and J07AB56 proteomes were
8–10kbpand/or40kb(fosmids)attheJCraigVenter detected using the reciprocal smallest distance
Institute, as described previously (Goldberg et al., algorithm (threshold e-value¼1e-05; sequence
2006). Details of genomic DNA sequence libraries divergence¼0.4) (Wall and Deluca, 2007).
are provided in Supplementary Table S1.
16S rRNA gene clone libraries were constructed
by amplification of LT metagenomic DNA using Amino acid composition analysis
universal archaeal primer sequences Arc21F and Amino acid frequencies in predicted proteins from
Arc529R (Table 1), as previously described (Bik J07AB56, J07AB43 and 1455 archaeal and bacterial
et al., 2010). A group-specific primer for Nanoha- genomes were compared using the Primer 6 software
loarchaea (LT_1215R) was designed using the NCBI program (Clarke and Gorley, 2006) to perform
primerdesigntool,andusedtogetherwithuniversal Non-Metric Multidimensional Scaling (NM-MDS)
archaeal primer Arc21F to amplify both LTand CV analysis (Ramette, 2007). For each genome, the
community DNA. Amplification products were frequencyofeachaminoacidforallpredictedproteins
cloned using the TOPO TA cloning kit (Invitrogen, was calculated using a custom perl script. These
Carlsbad, CA, USA) and sequenced bi-directionally values were standardized by Z-score, then used to
with M13F and M13R primers. calculate a Euclidean distance similarity matrix.
Sanger and pyrosequencing read libraries were NM-MDS analysis was performed using default
assembled both individually and in various combina- program parameters (25 random starts, Krustal fit
tions, using Celera Assembler software version 5.4 scheme of 1 and a minimum stress value of 0.01).
(Myers et al., 2000), in a series of iterative assemblies In addition to NM-MDS analysis, a cluster analysis
guided by phylogenetic binning. Detailed genome was performed to define groups within the NM-MDS
assembly procedures are provided in Supplementary plot using a multidimensional distance parameter
Information. of 4%.
Table1 Primersandprobesfordetecting16SrRNAsequences
Use Target Name Sequence(50 to30) Reference
PCR NHA LT_1215R ggccgcgtgtatcccagagc Thisstudy
A Arc21F ttcCggttgatccygccCga DeLong(1992)
A Arc529R accgcggckgctggc DasSarmaandFleischman(1995)
PCRa A ArcF1 attcCggttgatcctgc Iharaetal.(1997)
A Arc27Fa tcyggttgatcctgscGg RaesandBork(2008)
U Univ515F tgccagcAgccgcggtaa Lane(1991)
A Arc751F CcGAcggtgAgRgrygaa Bakeretal.(2003)
A Arc958R yCcGgcgttGAmtcCaatt DeLong(1992)
U Univ1390R acGggcGgtgtgtrcaa BrunkandEis(1998)
A UA1406R acGggcGgtgwgtrcaa Bakeretal.(2003)
A Arc1492R ACGGhTACCttgTtaCgactt Grantetal.(1999)
U Univ1492R GGTTACCttgTtaCgactt Lane(1991)
FISH A Arc915 gtgctcccccgccaattcct Amannetal.(1995)
NHA Narc_1214 ccgcgtgtatcccagagc Thisstudy
NHA LT_1198h1 attcgggccatactgacct Thisstudy
NHA LT_976-h2 ggctctggtagrgtrc Thisstudy
NHA LT_1237h3 tytstttgthccggccattg Thisstudy
B Eub338 gctgcctcccgtaggagt Amannetal.(1990)
B Eub338plus gcwgccacccgtaggtgt Daimsetal.(1999)
Target specificity abbreviations: A, archaea; B, bacteria; NHA, nanohaloarchaea; U, universal. aPCR primer mismatches are capitalized. Bold
indicatesprimermismatchestoJ07AB43only,underlinetoJ07AB56only,andboxedtobothJ07AB43andJ07AB56.
Lowercaseitaliclettersindicateexactmatchestotheorganismsdescribedinthetext.Uppercasenon-italiclettersindicatemismatchesofthree
differenttypes(bold,underline,orboxed).
aTheseprimerswerenotusedinthisstudy;sequencesareshownforcomparisononly.
TheISMEJournal
DenovometagenomicassemblyofNanohaloarchaea
PNarasingaraoetal
84
Phylogenetic analysis Genome Shotgun projects under accession numbers
16S rRNA sequences and ribosomal proteins from AEIY01000000 (J07AB43) and AEIX01000000
euryarchaeal genomes in the JGI-IMG database (J07AB56).
(Markowitz et al., 2009a) and GenBank were
compared with metagenomic gene sequences
Results
obtained by (i) extraction from assembled scaffolds
and (ii) amplification and sequencing of 16S rRNA
Metagenomic assembly
genes from LT and CV clone libraries. Maximum Seven independent DNA sequencing libraries were
likelihood trees were constructed using TreeFinder constructed from size-fractionated surface water
v.10.08(Jobbetal.,2004)andPhyMLv.3.0(Guindon samples collected at LT, Australia (Supplementary
and Gascuel, 2003). The robustness of each Figure S1 and Supplementary Table S1). Initial
maximum likelihood tree was estimated using assembly of the combined 632903 Sanger sequen-
non-parametric bootstrap analysis. Details of align- cing reads produced 15008 scaffolds (maximum
mentcurationandtree constructionareprovided in length¼2764168bp; scaffold N50¼29346bp).
Supplementary Information. These scaffolds included at least six different
Predicted proteins in assembled genomes were relatively abundant microbial populations, each
evaluated for phylogenetic relatedness to known with a distinct nucleotide percent GþC composi-
sequencesinNCBIGenBanknrusingtheDarkHorse tion. A length-weighted histogram of percent GþC
program, version 1.3, with a threshold filter setting versus total assembled scaffold nucleotides showed
of 0.05 (Podell and Gaasterland, 2007; Podell et al., peakscorrespondingtothesepopulations(Figure1).
2008). Minimum quality criteria for match inclusion The largest peak in this histogram, at 48% GþC,
in the DarkHorse analysis were that BLASTP align- included scaffolds containing 16S rRNA sequences
mentstoGenBanknrsequencescoveratleast70%of from multiple strains of Haloquadratum walsbyi,
total query length and have e-value scores of 1e-5 or consistent with previous observations noting the
better. dominance of this species in similar hypersaline
environments (Cuadros-Orellana et al., 2007; Oh
Fluorescence in situ hybridization etal.,2010).Threeadditionalpeaksat60%GþCor
Fluorophore-conjugated custom 16S rRNA probes higherincludedscaffoldscontaining16SrRNAgenes
(Table 1) were designed using ARB (Ludwig et al., with 89–99% identity to clone sequences annotated
2004), screened for specificity in silico using as uncharacterized halophilic archaea (class Halo-
ProbeCheck (Loy et al., 2008) and synthesized by bacteria).Microbialpopulationsassociatedwiththese
Integrated DNA Technologies Inc. (San Diego, CA, peaks are currently under investigation, but fall
USA). FISH was performed on CV and LT water outside the scope of the present report.
samples collected on 0.2mm polycarbonate GTTP Two groups of scaffolds, with peaks at 43% and
filters (Millipore) at every stage of filtration (post 56% GþC, shared an intriguing pattern of unusual
20mm, post 3mm and post 0.8mm). The Nanoha- characteristics. In addition to distinctive GþC
loarchaea-specificprobeNarc_1214conjugatedwith content, 490% of the reads that co-assembled in
Cy3alongwithunlabeledhelperprobesLT_1198h1,
LT_976h2 and LT_127h3 (Fuchs et al., 2000) were
7,000,000 Assembled scaffolds
used for FISH analysis. Universal probes Arc915 HQ
HA
(archaeal) and EubMix (a bacterial probe consisting n 6,000,000 SR
ofanequimolarmixtureofEub338andEub338plus) es/bi 5,000,000 HHRS
were also used for the purpose of cell counts. otid
Hybridization conditions were optimized at 461C cle 4,000,000
u
for 2h, as previously described (Pernthaler et al., n
m 3,000,000
2001). Filters were mounted with Vectashield u
n
medium (Vector Laboratories, Burlingame, CA, otal 2,000,000 43% 56%
USA), and imaged at 1000(cid:2) with a Nikon Eclipse T
1,000,000
TE-2000U inverted microscope (Nikon Instruments
Inc., Irvine, CA, USA). Cell counts were performed 0
40 45 50 55 60 65 70
onmultiplefieldsperslide,normalizing16SrRNA-
Percent GC (bin width = 1%)
specificprobecountstototalnumberofcellsstained
with the DNA-binding dye 40,6-diamidino-2-pheny- Figure 1 Length-weighted histogram of percent GþC for all
scaffolds assembled from the LTcommunity, binned in 1% GC
lindole.
increments.Symbolsrepresentreferencecontrolpoints,indicat-
ing where five previously sequenced halophile genomes would
havefallen,iftheyhadbeenpresentinthisdataset.Datapoints
Accession numbers areplottedbasedontotalnumberofnucleotidesineachscaffold
16S rRNA gene sequences have been deposited to (yaxis)versusaveragepercentGCfortheentirescaffold(xaxis).
HA, Haloarcula marismortui; HQ, Haloquadratum walsbyi; HR,
DDBJ/EMBL/GenBank under accession numbers
Halorubrumlacusprofundi;HS,HalobacteriumsalinarumR1;SR,
HQ197754 to HQ197794. Assembled genomes
Salinibacter ruber. Peaks labeled at 43% and 56% GC are the
with annotations have been deposited as Whole focusofthisstudy.
TheISMEJournal
DenovometagenomicassemblyofNanohaloarchaea
PNarasingaraoetal
85
Table 2 General features of the J07AB43 and J07AB56 draft proteins, amino acid tRNA synthetases, translation
genomes initiation and elongation factors, molecular chaper-
onesandproteinsessentialforDNAreplicationand
J07AB43 J07AB56
repair. All 53 of the universal archaeal housekeeping
proteins were identified in J07AB56 while 44/53
Genomesize,bp 1227157 1215802
(83%) were found in the J07AB43 draft genome
G+Cpercentage 44% 56%
Scaffoldnumber 7 3 (SupplementaryTableS3).Thepresenceofthesecore
rRNAoperons 1 1 proteins, a single rRNA operon and transfer RNAs
tRNAs 59 38 enabling translation of all 20 amino acids, suggests
PredictedCDSs 1678 1411
that both draft genomes are nearly complete.
CDSsw/func.Pred. 773 719
Abbreviations:CDS,codingsequence;rRNA,ribosomalRNA;tRNA,
transferRNA. Community abundance
Community abundance of J07AB43 and J07AB56
these scaffolds were obtained from microorganisms wasinitiallyassessedbysequencing16SrRNAgene
that had passed through a 0.8mm filter, but were clone libraries, constructed by amplifying LT com-
retainedona0.1mmfilter,suggestingsmallcellsize. munity DNA with universal archaeal primers
The 16S rRNA gene sequences contained in these Arc21FandArc529R(Table1).Amplifiedsequences
scaffolds were o78% identical to any previously with 491% identity to the J07AB43 and J07AB56
known cultured isolate, although they did resemble draft genomes were found in 124/315 (39%) of
16S rRNA gene fragments periodically recovered in archaealclonesobtainedfromorganismsretainedon
culture-independent surveys of microbial diversity 0.1mm pore filters, but only 24/254 (9%) of clones
in hypersaline waters (Grant et al., 1999; Oh et al., retained on 0.8mm pore filters. These results are
2010; Sime-Ngando et al., 2010). consistentwiththeobservedenrichmentofJ07AB43
Tooptimizeassemblyefficiencyfortheseunusual and J07AB56 reads specifically derived from 0.1mm
populations, the full set of metagenomic reads were filter fractions in the assembled data set.
subjected to a series of iterative assemblies guided As a second, independent test of community
by phylogenetic binning. The 43% GþC peak was abundance, new lineage-specific 16S rRNA probes
thereby consolidated into seven major scaffolds were designed to visualize J07AB43 and J07AB56
(J07AB43) and the 56% GþC peak into three major cells in environmental samples by FISH (Table 1).
scaffolds (J07AB56) (Supplementary Table S2). The These probes were used in combination with the
J07AB43 and J07AB56 scaffold groups were subse- DNA-binding dye 40,6-diamidino-2-phenylindole
quently analyzed as draft genomes, each represent- and universal bacterial and archaeal probes to
ing the consensus sequence of an individual obtaindirectcellcountsinLTandCVwatersamples
microbial population. Overall properties of these (Figure 2). Cells approximately 0.6mm in diameter
draft genomes are summarized in Table 2. These werelabeledwithlineage-specificprobeNArc_1214
properties differ substantially from previously se- in samples from both locations. These results are
quenced extreme halophiles in both nucleotide consistent with size estimates of o0.8mm but
composition, expressed as percent GþC, and total 40.1mm based on filter-specific composition for
genome size (Markowitz et al., 2009a). With the both amplified 16S rRNA clones and metagenomic
exception of H. walsbyi, at 48% GþC, all other reads. Direct counts of fluorescently labeled cells
previously described halophilic archaea, as well as indicated that the combined abundance of strains
the halophilic bacterium Salinibacter ruber, have matching the new, lineage-specific probes was
nucleotide compositions of 60% or greater GþC, approximately 14% of all 40,6-diamidino-2-pheny-
compared with 43% and 56% for these new lindole-labeled cells in water samples from LT,
organisms. Estimated total genome size and pre- and 8–11% in samples from CV (Supplementary
dicted number of coding sequences for J07AB43 Table S4).
and J07AB56 (Table 2) were also considerably Community abundance of the organisms respon-
smaller than other known extreme halophiles, sible for the J07AB43 and J07AB56 draft genomes
which currently range from 2.7 to 5.4Mbp. was further examined using statistical properties of
the assembled metagenomic sequence data. The
number of reads that co-assembled to create each
Genome completeness compositepopulationscaffoldgroupwasdividedby
To estimate the extent of genome completeness of the total number of reads available and normalized
J07AB43andJ07AB56,functionalannotationsforall for estimated genome size. Assuming the two new
predicted proteins were searched against a set of 53 genomesareapproximately1.2Mbpeach,andother
housekeeping genes, previously identified as uni- microbial species sampled from LT have an average
versally present in all archaeal genomes sequenced genome size of 3Mbp, J07AB43 was estimated to
as of 2009 (Puigbo et al., 2009). These highly represent at least 6.7% of the LT sampled commu-
conserved genes are physically dispersed throughout nity (17066 reads) and J07AB56 at least 3.4% (8652
the genome (non-clustered) and include ribosomal reads), totaling approximately 10% for the two
TheISMEJournal
DenovometagenomicassemblyofNanohaloarchaea
PNarasingaraoetal
86
Figure3 Unrootedmaximum-likelihood16SrRNAgenephylo-
genetictreeoftheEuryarchaeota.Treeisbasedon48sequences,
1275positions.Numbersofsequencesineachcollapsednodeare
indicated in parentheses. Numbers at nodes represent bootstrap
values inferred by TreeFinder/PhyML. Bootstrap values o50%
areindicatedbya‘–’sign.Scalebarrepresents0.1substitutions
per site. A full, uncollapsed version of this tree is presented in
SupplementaryFigureS2a.
Concatenated ribosomal protein data sets have
been shown to be particularly useful for resolving
deep evolutionary relationships (Brochier et al.,
2002; Matte-Tailliez et al., 2002; Rokas et al., 2003;
Figure 2 FISH micrographs. (a) LT (0.1 to 3mm filter fraction), Rannala and Yang, 2008). Phylogenetic analysis of
(b)CVSouthBaySaltWorks(0.1to0.8mmfilterfraction).Allcells 57 ribosomal proteins from the J07AB43 and
are stained with 40,6-diamidino-2-phenylindole (blue). Nanoha- J07AB56 draft genomes showed, like the 16S rRNA
loarchaeacellsshownarestainedwithlineage-specificCy3probe
tree, robust placement of these genomes as a deeply
Narc_1214(red).Scalebar¼2mm.
branching sister group of class Halobacteria, with
populations combined (3.0/1.2*25718/632903). bootstrap values of 98% (PhyML) and 74% (Tree-
Calculations based on metagenomic assembly most Finder). This relationship was corroborated using
likely underestimate true population abundance, Dayhoff04recodingofribosomalproteinalignments
because they may exclude closely related poly- (Hrdy et al., 2004; Susko and Roger, 2007), to rule
morphic strains containing DNA sequence varia- out possible artifacts of biased amino acid composi-
tions that were not incorporated into the consensus tionorfast-evolvinglineages(SupplementaryFigure
population assembly. S2b). The long branch lengths separating J07AB43
and J07AB56 from members of class Halobacteria
indicate that these two sister-lineages are only
distantly related, consistent with the average diver-
Taxonomic position of J07AB43 and J07AB56 gence of 35% observed between Halobacteria and
J07AB43 and J07AB56 16S rRNA shared sequence J07AB43 and J07AB56 16S rRNA gene sequences
identities of 68% to 75% with previously sequenced, (Supplementary Table S5). By contrast, 16S rRNA
cultured representatives of class Halobacteria (Sup- variability within the Halobacteria is o16%.
plementary Table S5). An unrooted maximum like- Nearly60% ofpredictedproteinsinJ07AB43 and
lihood phylogenetic tree of euryarchaeotal 16S rRNA J07AB56 had no GenBank database matches close
gene sequences placed J07AB43 and J07AB56 as a enough to enable confident phylogenetic assign-
deepsistergroupofclassHalobacteria(Figure3),with ment. Of those that could be assigned, fewer than
significant bootstrap support. 20% matched proteins from members of class
TheISMEJournal
DenovometagenomicassemblyofNanohaloarchaea
PNarasingaraoetal
87
Figure4 Phylogeneticdistributionofnon-self-proteinBLASTmatchesforeuryarchaeotalgenomes.SearchesagainsttheGenBanknr
databasewereclassifiedbyeuryarchaeotalclass,archaealphylum,domainornomatchusingtheDarkHorsealgorithm,asdescribedin
Materialsandmethodssection.
Halobacteria (Figure 4). In contrast, 480% of strategy of maintaining osmotic balance, as evi-
predicted proteins in the genomes of previously denced by the over-representation of amino acids
sequenced Halobacteria had closest non-self with negatively charged side chains (aspartic and
matches to other members of their own class, glutamic acid) and the under-representation of
leaving fewer than 20% unmatched. Similar pat- residueswithbulkyhydrophobicsidechains(trypto-
terns of protein sequence conservation were ob- phan, phenylalanine and isoleucine), to enhance
served in organisms with many sequenced database protein structural flexibility and solubility under
relatives, including Methanocaldococcus janaschii, intracellular conditions of high ionic strength and
Methanospirillum hungatei and Salinibacter ruber, low water availability. Although a similar salt-in
but not in sparsely sampled species that are only strategy is employed by other extreme halophiles,
distantly related to other known lineages, such as J07AB43 and J07AB56 use their own, distinct
NanoarchaeumequitansandMethanopyruskandleri combination of amino acids to achieve this end,
(Branciamore et al., 2008). preferring glutamic to aspartic acid, serine to
threonine, and reduced frequencies of alanine,
proline and histidine (Supplementary Table S6).
Genome characteristics of J07AB43 and J07AB56 The large number of proteins annotated with
Although the J07AB43 and J07AB56 genomes are ‘hypothetical’ functions in the J07AB43 and
more closely related to each other than to any J07AB56 genomes may be at least partially because
previously sequenced organisms, gene content ana- of their unusual amino acid compositions, which
lysis identified only 480 (30%) shared protein can hinder recognition of database homologs in
orthologpairsbetweenthem.Ofthese,143(approxi- sequence-based similarity searches.
mately 10% of each genome) were not found in ThepeculiaraminoacidcompositionsofJ07AB43
other halophilic archaea. The majority of these and J07AB56 compared with other halophilic
shared lineage-specific sequences were too dissim- archaea are highlighted in a NM-MDS plot of
ilar to previously characterized proteins to assign a intergenomic distances based on frequencies for all
functional annotation. The remainder was domi- 20standardaminoacids(Figure5).Thedatausedto
natedbyhousekeepingproteinsinvolvedintransla- constructthismatrixincludedallproteinsequences
tionandribosomalstructure.Eachgenomeincluded from euryarchaeal genomes used to build the
onlyonerhodopsin-likegene,comparedwithmulti- phylogenetic tree in Figure 3, supplemented
pleparalogspresentinthegenomesofotherextreme with four bacterial species found in high salt
halophilic archaea (Ihara et al., 1999), and the environments: Salinibacter ruber (Bacteroidetes),
extremely halophilic bacterium Salinibacter ruber Halorhodospira halophila (Gammaproteobacteria),
(Mongodin et al., 2005). Notably absent from both Chromohalobacter salexigens (Gammaproteobacter-
genomes were homologs to the highly conserved ia) and Halothermothrix orenii (Firmicutes).
Gvp family of gas vesicle proteins found in most Although genome percent GþC compositions
halophilic archaea, Cyanobacteria and purple were not explicitly included as one of the factors
photosynthetic bacteria (Walsby, 1994). in this analysis, there is a trend for microorganisms
Both J07AB43 and J07AB56 have highly unusual withlowerGþC(denotedwithlowerlabelnumbers
aminoacidcompositionscomparedwithpreviously in Figure 5) to be located further to the right along
sequenced archaeal and bacterial genomes. These thehorizontalaxis.Thistrendisconsistentwiththe
unusual compositions appear to support a ‘salt-in’ known influence of GþC composition on usage
TheISMEJournal
DenovometagenomicassemblyofNanohaloarchaea
PNarasingaraoetal
88
Figure5 NM-MDScomparisonofaminoacidcompositions.Euryarchaealgenomesweresupplementedwithfourhalophilicbacteria
genomes.Symbolsdenotetaxonomicclassifications.NumbersrankgenomesinincreasingorderofGþCcontent(1–10:29–38%,11–20:
38–43%,21–30:43–50%,31–40:50–60%,41–53:60–67%).Greycirclesindicatehierarchicalclustering,basedona4%distancesetting
todefinegroups.AcompletelistofthesegenomesandtheiraminoacidcompositionsispresentedinSupplementaryTableS6.
frequency for some amino acids because of codon pathway were observed in both genomes, including
bias(Liuetal.,2010).Incontrast,positionalongthe both oxidative and non-oxidative branches. The
vertical axis of Figure 5 was unrelated to percent presence of a complete pentose phosphate pathway
GþC. Instead, amino acid composition differences has not been demonstrated previously in any other
captured along this axis appear to correlate more archaea, by either biochemical or bioinformatic
closely with common ancestry and/or shared envir- methods(Verheesetal.,2003).Thekey,rate-limiting
onmental adaptations. The outlier positions of enzyme for this pathway is glucose-6-phosphate
J07AB43 (#19) and J07AB56 (#39) along the vertical dehydrogenase, which converts glucose-6-phos-
axis of Figure 5 clearly demonstrate their unusual phate into 6-phosphoglucono-d-lactone. Although
amino acid compositions relative to other archaea. bothJ07AB43andJ07AB56appeartohavecomplete
Similar outlierpositions were observed for these two genomic copies of this gene, the closest database
genomes when analyzed in the context of a much relativestotheirsequencesareallbacterial,suggest-
larger microbial genomic data set, including 1382 ing this functionality may have been acquired by
bacterial and 73 archaeal species (data not shown). ancient horizontal gene transfer. The nearest homo-
InferredmetaboliccapabilitiesoftheJ07AB43and log of the glucose-6-phosphate dehydrogenases in
J07AB56 genomes are consistent with a predomi- J07AB43 and J07AB56 is from the genome of
nately aerobic, heterotrophic lifestyle. The absence Salinibacter ruber, a common bacterial inhabitant
of identifiable anaerobic terminal reductases of hypersaline environments believed to have
suggests they are incapable of anaerobic respiration experiencedfrequenthorizontalgeneexchangewith
although the presence of lactate dehydrogenases archaea (Mongodin et al., 2005).
suggests possible fermentative metabolism under
microaerophilic conditions. Both genomes contain
enzymes necessary to support glycolysis, as well as Geographical distribution and diversity
operons encoding key enzymes for glycogen synth- Lineage-specific PCR primer, LT_1215R (Table 1)
esis and catabolism. Several of these enzymes, and general archaeal primer Arc21F were used to
including a glycogen debranching enzyme and construct clone libraries from environmental DNA
predicted alpha-1,6-glucosidase activity, are not samples collected from both LTand CV, yielding 43
present in any other known members of class new 16S rRNA gene sequences. Additional 16S
Halobacteria. However, these enzyme activities are rRNA gene sequences, with 485% identity to
frequently found in archaea from classes Methano- J07AB43 and J07AB56, were identified in public
cocci and Thermoplasmata that utilize starch as an databases. These published sequences originated in
internal storage molecule (Ko¨nig et al., 1985, 1982). environmentalsamplesfromAfrica,AsiaandSouth
This suggests a possible common ancestral origin, America, as well as Australia and North America
withsubsequentgenelossintheHalobacterialineage. (Supplementary Table S7). The phylogenetic analysis
In addition to the Embden–Meyerhoff pathway, of these 16S rRNA gene sequences reveals at least
genes supporting the entire pentose phosphate eight distinct clades with strong bootstrap support
TheISMEJournal
DenovometagenomicassemblyofNanohaloarchaea
PNarasingaraoetal
89
Figure6 16SrRNAgenemaximumlikelihoodtreeofNanohaloarchaeasequencesrecoveredfromworldwidehypersalinehabitats.Tree
isbasedon709nucleicacidpositionsin77sequences.Numbersatnodesrepresentbootstrapvalues(valueso50%notshown).Scalebar
showsaveragenumberofsubstitutionspersite.
(bootstrapvalues487%,Figure6).Basedondegree using a combination of strategic environmental
of sequence divergence, each clade most likely sampling,deepsequencing,anddenovometagenomic
represents a new genus or higher taxonomic level. assembly can reveal significant new information.
ClassificationofJ07AB43andJ07AB56intoseparate We have discovered and characterized nearly
genera is strongly supported by tree topology, complete genomes representing a novel archaeal
16% sequence divergence in the 16S rRNA gene lineage prevalent in hypersaline systems world-
(Supplementary Table S5) and a 13% difference in wide, yet very different from all previously
genomic GþC content. described members of class Halobacteria.
We propose the creation of a new class ‘Nanoha-
loarchaea’ within phylum Euryarchaeota to
Discussion accommodate this new lineage. We further propose
partitioning class Nanohaloarchaea to place
This study has demonstrated that re-examination of J07AB43 and J07AB56 into distinct genera, Candi-
a fairly simple, well studied environmental habitat datus ‘Nanosalina sp. J07AB43’ and ‘Candidatus
TheISMEJournal
DenovometagenomicassemblyofNanohaloarchaea
PNarasingaraoetal
90
Nanosalinarum sp. J07AB56’. Evidence supporting et al., 2006). If extremely high environmental
these proposals includes: (i) comprehensive eur- magnesium cannot be adequately excluded from
yarchaeotal phylogenetic analyses based on 16S the cell, lower genomic GþC helps maintain DNA
rRNA genes and ribosomal proteins; (ii) lineage- structuralflexibilityandavoidsdifficultiesinstrand
specificfeatures,includingnumerousgeneswithout separationcausedbyelevatedmeltingtemperatures.
previously described close relatives; and (iii) These same principles could apply to J07AB43,
significant intra-lineage diversity and abundance providingapossible selectiveadvantage underhigh
within geographically distinct hypersaline habitats magnesiumconditionsexpectedinevaporativehigh
worldwide. Evolutionary distinctness of J07AB43 salt environments.
and J07AB56 from other halophilic archaea is Nanohaloarchaea are estimated to represent at
reinforced by taxonomic patterns of BLASTP least 10–25% of the total archaeal community in
matches for their predicted proteomes against surface water samples from LT, Australia and CV,
GenBank nr, as well as amino acid composition- California,USA.Webelievethese valuesarerobust,
based clustering. The sister-grouping of class based on agreement of three independent analysis
Halobacteria and class Nanohaloarchaea reflects techniques: amplification of environmental 16S
probable derivation from an ancient common rRNA gene sequences; statistical analysis of meta-
halophilic ancestor with a ‘high salt-in’ osmotic genomic sequencing reads assembled into near-
regulation strategy, followed by subsequent complete draft genomes; and quantitative FISH of
divergence along separate evolutionary paths. cells from natural water samples labeled with
Lineage-specific characteristics that distinguish lineage-specific probes. Microscopic counts reveal
‘CandidatusNanosalinasp.’and‘CandidatusNano- that Nanohaloarchaea are present at cell concentra-
salinarum sp.’ from most other extreme halophiles tions exceeding 106cellsml–1 in hypersaline habitats
includetheirsmallphysicalsize,compactgenomes, of Australia and North America. The sporadic
single-copy rRNA operon, low GþC composition, identification of Nanohaloarchaea in other surveys
unique proteome amino acid composition, absence of hypersaline communities worldwide suggests that
of conserved gas vesicle genes and atypical Nanohaloarchaearepresentasignificantyetneglected
predicted pathways associated with carbohydrate fractionofthebiomassanddiversityinthesehabitats.
metabolism. Small compact genomes, as well as The inability of earlier studies to recognize the
single-copy rRNA operons, have been proposed to significant contribution of Nanohaloarchaea to
minimize metabolic costs in habitats where neither hypersaline community composition is likely due
broad metabolic repertoire nor high numbers of to limitations of the tools routinely used to assess
paralogous proteins are needed to accommodate environmentalmicrobialdiversity,includinglabora-
rapid growth under fluctuating environmental toryculture,microscopy,amplificationof16SrRNA
conditions (Klappenbach et al., 2000). Small cell gene fragments, and sequence database similarity
size, which increases surface to volume ratio, could searches for unassembled metagenomic reads. The
be an adaptation for optimizing nutrient uptake isolation of cultured strains from environmental
capacity. Alternatively it is possible that small habitats is known to exclude many organisms that
physical size allows Nanohaloarchaea to remain are highly successful in their native habitats. It is
suspended in oxygenated surface waters to support therefore not surprising the 96 hypersaline archaeal
aerobic metabolism, thus eliminating the need for isolates described to date do not include any
gas vesicles to provide buoyancy. Nanohaloarchaea. Repeated efforts to culture these
The low GþC compositions of the two Nanoha- microorganisms in our own laboratory have also
loarchaea genomes, especially J07AB43 (43%), are been unsuccessful. Furthermore, cultivation-inde-
surprisingconsideringtheirprevalenceinhighlight pendent microbial diversity studies based on 16S
habitats. In the absence of compensatory mechan- rRNA gene amplification are known to suffer from
isms, lower GþC would be expected to increase primerbias(Siposetal.,2007).Mismatchesbetween
susceptibility to ultraviolet-induced DNA damage. Nanohaloarchaea and many commonly used
One possible explanation is that the low GþC universal primers may have impeded detection in
composition of J07AB43 is related to ecological earlier studies. Primers likely to have been particu-
lifestyle. Low GþC composition and genomic larlyproblematicarehighlightedinTable1(Amann
streamlining have been associated with decreased et al., 1990, 1995; Lane, 1991; DeLong, 1992;
nitrogen requirements and a slow-growing, energy- DasSarma and Fleischman, 1995; Ihara et al., 1997;
conservativelifestyleinmarinebacteria(Giovannoni BrunkandEis,1998;Daimsetal.,1999;Grantetal.,
etal.,2005).However,thehabitatsfromwhichthese 1999; Baker et al., 2003; Raes and Bork, 2008).
Nanohaloarchaea were isolated are not generally The exceptionally small size of Nanohaloarchaea
considered to be nutrient-limited (Oren, 2002b). compared with other halophilic microorganisms
Alternatively, it has been proposed that the low makes them difficult to visualize by microscopy in
GþC composition of H. walsbyi (48%) compared the absence of selective enrichment techniques or
with other halophiles is a specific adaptation to group-specific probes, and can prevent recovery
counteract the over-stabilizing effect of high magne- during sample concentration procedures targeting
sium concentrations on DNA structure (Bolhuis largermicroorganismsorsmaller viruses(Rodriguez-
TheISMEJournal