lable at ScienceDirect

Soil Biology & Biochemistry 42 (2010) 739e747

Contents lists avai

Soil Biology & Biochemistry

journal homepage: www.elsevier .com/locate/soi lb io

A phylogenetic microarray targeting 16S rRNA genes from the bacterial divisionAcidobacteria reveals a lineage-specific distribution in a soil clay fraction

Mark R. Liles a, Ozgur Turkmen b, Brian F. Manske c, Mingzi Zhang c, Jean-Marie Rouillard d,Isabelle George e, Teri Balser f, Nedret Billor b, Robert M. Goodman g,*

aDepartment of Biological Sciences, Auburn University, Auburn, AL, USAbDepartment of Mathematics and Statistics, Auburn University, Auburn, AL, USAcDepartment of Plant Pathology, University of Wisconsin, Madison, WI, USAdDepartment of Chemical Engineering, University of Michigan, Ann Arbor, MI, USAeUnité de Génie Biologique, Université Catholique de Louvain, Louvain-la-Neuve, BelgiumfDepartment of Soil Sciences, University of Wisconsin, Madison, WI, USAg School of Environmental and Biological Sciences, Rutgers, The State University of New Jersey, 88 Lipman Drive, Suite 104, New Brunswick, NJ 08901, USA

a r t i c l e i n f o

Article history:Received 30 October 2009Received in revised form8 January 2010Accepted 14 January 2010Available online 4 February 2010

Keywords:Acidobacteria16S rRNAMicroarrayPhylogenyClayPhylochip

* Corresponding author. þ1 732 932 9000x500.E-mail address: execdean@aesop.rutgers.edu (R.M

0038-0717/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.soilbio.2010.01.007

a b s t r a c t

We designed an oligonucleotide microarray using probe sequences based upon a phylogenetic analysis of16S rRNA genes recovered from members of the bacterial division Acidobacteria. A total of 42,194oligonucleotide probes targeting members of the Acidobacteria division at multiple phylogenetic levelswere included on a high-density microarray. Positive control hybridizations revealed a linear relationshipbetween hybridization signal and template concentration, and a substantial decrease in non-specifichybridization was achieved through the addition of 2.5 M betaine to the hybridization buffer. A meanhybridization signal value was calculated for each Acidobacteria lineage, with the resultant lineage-specific hybridization data revealing strong predictive value for the positive control hybridizations. TheAcidobacteria phylochip was then used to evaluate Acidobacteria rRNA genes from a Wisconsin soil andwithin a soil clay fraction. The Acidobacteria hybridization profile revealed the predominance of Acid-obacteria subdivisions four and six, and also suggested a decrease in the abundance of subdivision sixrelative to subdivision four in the soil clay fraction. The change in relative abundance of these subdivi-sions in a soil clay fraction was supported by data from quantitative PCR. These results support the utilityof a phylogenetic microarray in revealing changes in microbial population-level distributions ina complex soil microbial assemblage.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The vast majority of prokaryotes live within oligotrophic naturalenvironments and are poorly represented in culture collections(Amann et al., 1995; Torsvik et al., 1990, 1994, 1996). This is espe-cially true of soil microorganisms, which have contributed greatlyto our arsenal of antimicrobial agents, yet every census of soilmicroorganisms to date has revealed only fragmentary evidence ofthe extant phylogenetic and metabolic diversity present in any soil.Efforts to understand the distribution and ecological roles ofenvironmental microorganisms have been aided by the molecularsuite of tools now available, particularly those based on analysis ofthe small subunit ribosomal RNA gene (16S rRNA). Polymerase

. Goodman).

All rights reserved.

chain reaction (PCR) amplification of 16S rRNA genes from naturalenvironments reveals many 16S rRNA gene sequences that arehighly divergent from known cultured phyla (Woese, 1987;Hugenholtz et al., 1998a,b; Meier et al., 1999; Dojka et al., 2000;Wilson et al., 2002). Many of the highly divergent 16S rRNA genesequences together comprise entirely new monophyletic prokary-otic lineages, forming newly recognized divisions (Barns et al.,1994; Head et al., 1998; Hugenholtz et al., 1998a,b; Pace et al., 1986;Ward et al., 1990). Of the bacterial divisions revealed primarily byrRNA gene sequence data, the division Acidobacteria is ubiquitousin soils and sediments (Liesack et al., 1994; Kuske et al., 1997; Barnset al., 1999). All of the cultured Acidobacteria isolates to date fall intofour of the proposed subdivisions (Kishimoto et al., 1991; Ludwiget al., 1997; Hugenholtz et al., 1998a,b; Barns et al., 1999; Janssenet al., 2002), the number of which has recently been expanded toinclude up to 26 subdivisions (Zimmermann et al., 2005; Barnset al., 2007). A phylogenetic analysis of 16S rRNA gene sequences

M.R. Liles et al. / Soil Biology & Biochemistry 42 (2010) 739e747740

recovered from a soil at the West Madison Agricultural ResearchStation (WMARS) revealed that 25% of the rRNA genes were affili-ated with the Acidobacteria division (Liles et al., 2003).

Phylogenetic microarrays, or “phylochips”, have been used todiscriminate rapidly between diverse 16S rRNA genes presentwithin cultured or environmental microorganisms (Wilson et al.,2002; Brodie et al., 2006; Palmer et al., 2006; Huyghe et al., 2008).Compared to the labor- and resource-intensive efforts to clone andsequence a representative number of clones from a 16S rRNA geneclone library, phylogenetic microarrays can provide an efficientreadout of the phylogenetic diversity present in an environmentalsample. Furthermore, a hierarchical design allows probing formicrobial taxa at different phylogenetic levels (Huyghe et al., 2008),giving information on the presence or absence of the branches andthe twigs on the tree of life. In the present study, the divisionAcidobacteriawas targeted by a phylogenetic microarray approach,with oligonucleotide probes targeting multiple phylogenetic levels.A series of positive control hybridizations were used to validate thephylogenetic microarray design and hybridization conditions,followed by experiments to determine the distribution of Acid-obacteria taxa in a soil sample and a clay fraction from this same soilsample. Results from the Acidobacteria phylogenetic microarraywere then tested by an independent molecular analysis to validatemicroarray hybridization results.

2. Materials and methods

2.1. Soil collection

Soil cores were collected from an undisturbed site at the WestMadison Agricultural Research Station (WMARS) (Bintrim et al.,1997; Rondon et al., 2000). The ecosystem is a turfgrass understorywithout organic amendments. The soil type is a Plano silt loamcontaining 61% sand, 23% silt, and 16% clay, with 1.7% organicmatter (pH 7.0). The top 10 cm of soil were sampled and sievingwasused to remove roots and debris. Genomic DNA was isolatedimmediately after soil collection. The remainder of the soil samplewas frozen at �80 �C.

2.2. DNA isolation from soil or Escherichia coli cultures

A bead-beating method (Bio101, Inc., La Jolla, CA) was used toisolate genomic DNA from soil microorganisms. The genomic DNAisolated via a Bio101 kit is generally less than 20 kb in size, yet theharsh lysis conditions ensure that the genomic DNA is broadlyrepresentative of the soil microbial community (Burgmann et al.,2001). Samples were stored at �20 �C until further analysis.

A set of positive control Acidobacteria rRNA genes was availablefrom a previous study, with recombinant clones containing Acid-obacteria rRNA operons and associated genes cloned withina bacterial artificial chromosome (BAC) vector (Liles et al., 2003).BAC DNAwas isolated from E. coli cultures grown in Luria Broth (BDDiagnostic Systems, Sparks, MD) overnight while shaking at 37 �C,using a Large-Construct DNA Isolation kit (Qiagen, Inc., Valencia,CA). The use of an Acidobacteria-specific primer set prevented E. coli16S rRNA gene contamination of the amplicons.

2.3. PCR amplification of rRNA genes

AnAcidobacteria16S rRNAgene-specificprimer (31F)wasused toamplify 16S rRNA genes from soil genomic DNA or positive control16S rRNA gene clones by PCR (Barns et al., 1999). It should be notedthat although this division-level primer is not predicted to PCRamplify 16S rRNA genes from all of the extant Acidobacteria subdi-visions (Barns et al., 2007), the Acidobacteria subdivisions present

within WMARS soil were identified using a universal bacterialprimer set (Liles et al., 2003) and each of the Acidobacteria subdivi-sions identified in WMARS may be PCR amplified using the Acid-obacteriadivision-level primer 31F. Ampliconswere produced usingapproximately 100 ng DNA template, 1 unit Taq polymerase(Promega,Madison,WI),1� Taqpolymerase reaction buffer, 200 mMdNTPs, and 200 nM of each of the primers 31F (50-GATTCTGAGC-CAAGGATC, Acidobacteria-division specific)(Barns et al., 1999) and1492R (50-ACGGYTACCTTGTTACGACTT, universal Bacteria domain)(Medlin et al., 1988) in 50-ml. Reactions to be used for microarrayhybridization directly incorporated the Cy3-dCTP dye (GE Health-care, Piscataway,NJ) into the PCRproduct using afinal concentrationof 40 mM Cy3-dCTP (20% of total dCTP) according to manufacturer'sinstructions. The reactionwasperformedwith3mindenaturationat95 �C, 30 cycles of 95 �C for 1 min, 55 �C annealing for 90 s, 72 �Cextension for 150 s, followed by 7 min extension at 72 �C. All reac-tions were carried out in a Robocycler 96 (Stratagene, La Jolla, CA)with 50 ml mineral oil added to each tube. Reactions were analyzedby agarose gel electrophoresis to confirm production of a single(heterogeneous in the case of soil genomic DNA template) amplicon.The resulting PCR products were fragmented to an approximateaverage size of 200 bp using diluted DNAse I (final concentration0.004 U/ml) for 30 min at 37 �C. Consistent DNA fragmentation wasmonitored by agarose gel electrophoresis. To remove unincorpo-rated Cy3-dCTP, the fragmented PCR products were purified andconcentrated over Centricon 3000 NMWL centrifuge concentrators(Millipore, Inc., Billerica, MA). The concentration of fragmented,labeled PCR product was determined using a Nanodrop ND-1000spectrophotometer (Thermo Fisher Scientific, Wilmington, DE).

2.4. Phylogenetic analysis

Acidobacteria 16S rRNA gene sequences were aligned to a data-set of 6883 bacterial sequences (courtesy of Dr. Phillip Hugenholtz,http://rdp.cme.msu.edu/html/alignments.html) using the ARBsoftware package (http://www.arb-home.de/) and refined manu-ally to remove regions of ambiguous homology. Alignments usedfor phylogenetic analysis were minimized by the Lane mask (Laneet al., 1985) for bacterial data or an Acidobacteria filter (phylumspecific 50% filter by base frequency) prepared in ARB. Phylogenetictrees for near full-length sequences (>1400 nt) were inferredwithin the ARB package using evolutionary distance (neighbor-joining algorithms with Felsenstein correction) and the PHYLIPprogram for maximum parsimony (Felsenstein, 1993). Partialsequences (<1400 nt) were inserted into trees without brancharrangement of full-length sequences using the parsimony inser-tion tool of ARB. The robustness of the tree topology was tested bybootstrap resampling with multiple outgroups.

2.5. Probe design

Acidobacteria-specific oligonucleotide probes were designedusing the ARB software package and the ssujun02 dataset ofsequences (http://www.arb-home.de/). Additional Acidobacteriasequences deposited in GenBank were also included. The datasetwas limited to only Acidobacteria sequences larger than 500 bp andall potential chimeras were removed by partial treeing of the finalalignment. Bulk probe design was accomplished through the ARBprobe design function on two separate trees, one consisting of onlyfull-length sequences (>1300 bp) and the other of full and partialsequences (>500 bp). Probes were designed for all clades sup-ported by bootstrap values >¼ 85% (parsimony), supporting thepresence of 10 monophyletic Acidobacteria subdivisions. Valuesused in the ARB probe design function were as follows: Length ofOutput ¼ 100; Max. non-group hits ¼ 0; Max. hairpin bonds ¼ 4;

M.R. Liles et al. / Soil Biology & Biochemistry 42 (2010) 739e747 741

Min. group hits (%) ¼ 75; Length of probe ¼ 18, 19, and 20;Temperature ¼ 30e80; G þ C content (%) ¼ 50e80. Reversecomplements were generated and added to the probe list, sinceeach strand of a rRNA gene amplicon is capable of hybridization tothe microarray. All probes were assembled into a list with associ-ated target clade, sequence, probe length, predicted Tm, G þ Ccontent and corresponding E. coli alignment position.

A set of 70 additional oligonucleotide probes targeting variousdivisions of Bacteria were obtained from probeBase (http://www.microbial-ecology.net/probebase/) and were included in theoligonucleotide probes on the microarray. All oligonucleotideprobes used in this study, as well as their predicted phylogeneticspecificity, E. coli relative position, length in base pairs, % G þ Ccontent, predicted Tm, and all hybridization signals for eachphylogenetic group are deposited within the Gene ExpressionOmnibus (GEO) as accession GSE18711 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token¼plcldieqayqccju&acc¼GSE18711).

2.6. Spiked-in oligonucleotides

In initial hybridization experiments, the 70 probe sequencesspecific to non-Acidobacteria 16S rRNA genes were tested forhybridization to an Acidobacteria rRNA gene PCR product. Threeprobes (DHP1006, 50-CCTCTCGATCTCTCTCAAGT (McSweeney et al.,1993); GSB532, 50-TGCCACCCCTGTATC (Tuschak et al., 1999),LGC354A, 50-TGGAAGATTCCCTACTGC (Meier et al., 1999)) thatshowed the least background hybridization were chosen for use asoligonucleotide probes to include in each hybridization to helpcontrol for microarray-to-microarray variability. For each experi-ment, Cy3-labeled oligonucleotides complementary to these threeprobes were added to the hybridization reaction, and hybridizationsignals for each control probe were averaged among replicates.

2.7. Microarray design and construction

A custom microarray containing a total of 182,318 total featureswas synthesized onto a glass slide using maskless array synthesis(MAS) (Roche Nimblegen, Inc., Madison, WI), with three replicatesof each oligonucleotide probe. Probe lengths ranged from 15 to20 bp, with all Acidobacteria-specific oligonucleotides in the18e20 bp size range. Slides were kept desiccated at roomtemperature prior to hybridization.

2.8. Microarray hybridization

Each hybridization reaction had a total volume of 20 ml whichcontained bovine serum albumin (0.2 mg/ml, Sigma Chemical Co.),sheared salmon sperm DNA (0.14 mg/ml, Invitrogen, Inc.), 1� Nim-blegen hybridization buffer (Roche Nimblegen, Inc.), spiked-in oli-gos (10 nM each, see sequences above), and 100 ng fragmented anddenatured PCR product. Hybridizations were conducted at 45 �C for16h in aNimblegen12BayHybridization System(RocheNimblegen,Inc.). Most reactions (except for first control hybridizations) con-tained 2.5 M betaine to enhance hybridization stringency. Micro-arrays were washed 3 times with wash solutions I, II, and III (RocheNimblegen, Inc.) before being dried using N2 gas.

2.9. Microarray scanning

Each hybridized microarray was scanned using an Axon4000Bscanner (Molecular Devices, Sunnyvale, CA) at 10 mm resolution.GenePix� Pro 6.0 software was used for the analysis and processingof signal strength from each scanned microarray. Overall signalstrength for each microarray was adjusted to prevent inclusion of

saturated signals. Signal intensity for each feature was determinedand exported into a tab-delimited file for subsequent analysis.

2.10. Microarray data analysis

Data were imported into an Excel spreadsheet and sorted byphylogenetic group. Data were then normalized in relation toa mean hybridization signal observed from the set of Acidobacteriadivision-level probes. Since this set of Acidobacteria division-levelprobes was universal for every hybridized PCR amplicon, and thesignal strength was consistently strong for each of these division-level probes, microarray normalization relative to the Acidobacteriadivision-level probes provided the best array-to-array consistency,compared to normalization using spiked-in oligonucleotide probes.A mean hybridization signal for each phylogenetic group relative tothe Acidobacteria division-level probes was determined for everyexperiment, and data were analyzed for statistical significance asdescribed below.

2.11. Statistical analysis of oligonucleotide probethermodynamic characteristics

Several outlier detection methods were used to identify probeswith extreme values of %G þ C content, DG62, Tm, and/or DG62-foldto eliminate probes that would not be predicted to perform well inthe hybridization reaction. DG62, Tm and DG62-fold were predictedusing the hybrid-min and hybid-ss-min software from the UNAfoldpackage (Markham and Zuker, 2005, 2008) using a temperature of62 �C (hybridization temperature) and sodium and DNA concen-trations of 1 M and 1 mM respectively. Since there was a sufficientnumber of features on the microarray to include all oligonucleotideprobes regardless of predicted thermodynamic characteristics, allprobe sequences identified by ARB were included on the micro-array and outlier sequences were analyzed and removed post-hybridization. This permitted a comparison between the dataset ofprobe hybridizations with and without removal of outlier probes.The Mahalanobis distance based outlier detection methods such asthe minimum covariance determinant (MCD) (Rousseeuw and vanDriessen,1999) and BACON (Billor et al., 2000) were used for 42,142observations to identify “bad” probes. Any observation thatexceeded c2 based cutoff value was identified as a “bad” probe. Therest of the statistical analyses were carried out on the clean dataafter eliminating w7% “bad” probes with in silico predicted poorhybridization characteristics.

2.12. Statistical analysis of the effect of betaineon hybridization stringency

A series of positive control hybridizations was conducted first toidentify hybridization conditions for optimal resolution betweenphylogenetic groups. An experiment was conducted with positivecontrol Acidobacteria subdivision 4 rRNA clone 220H (Liles et al.,2003) with and without the addition of 2.5 M betaine in thehybridization buffer, using the exact same PCR amplicon forhybridization to the microarrays with and without betaine addi-tion. The mean hybridization signal for all true-positive phyloge-netic groups (i.e., corresponding to the predicted phylogeny fromAcidobacteria subdivision 4 containing sequence 220H) werecompared to all other phylogenetic groups based on a Tukey'sStudentized Range (HSD) test.

M.R. Liles et al. / Soil Biology & Biochemistry 42 (2010) 739e747742

2.13. Statistical analysis of phylogenetic microarrayhybridization experiments

Within the Acidobacteria division-level phylogeny, there were76 non-overlapping lineages (i.e., terminal branches) that had over75% bootstrap support. For every microarray hybridization, pair-wise tests between each Acidobacteria phylogenetic lineage wereperformed to compare each lineage to the others. The TukeyeK-ramer method was used to compare all pairwise differences inphylogenetic group mean hybridization signals. A false discoveryrate was also utilized to control the expected proportion of incor-rectly rejected null hypotheses (type I errors) in a list of rejectedhypotheses.

2.14. Soil fractionation

Soil from WMARS was wet sieved through a fine mesh, andfractionated into sand, silt, and clay fractions by differentialcentrifugation (Poll et al., 2003). The sand fraction was washedrepeatedly to remove any other soil particles.

2.15. Quantitative PCR

Acidobacteria subdivision 4 (50-GGCGAAAGCCTGACCCAGCA, atequivalentE. coli16S rRNApositionbp376) and subdivision6-specific(50-ACGCGGCGTTGCGTCGTC, at equivalent E. coli 16S rRNA positionbp 387) oligonucleotide primers were designed to PCR amplify thesesubdivisions from soil community genomic DNA, in conjunctionwithan Acidobacteria division-level primer (e.g., 31F) to provide sufficientphylogenetic specificity. In the case of the subdivision 4-specificprimer, this is also predicted in silico (Ribosomal Database Project,http://rdp.cme.msu.edu/) to amplify rRNA genes from Acidobacteriasubdivision 21, but using the forward primer 31F should restrict thisprimer pair to amplify Acidobacteria subdivision 4 rRNA templates.These subdivision-specific primer pairs produced amplicons ofequivalent size (345 bp or 356 bp, respectively) and with equivalentthermodynamics based on amplification reactions using positivecontrol templates (PCR amplified rRNA gene clones) at knownconcentrations. In each set of reactions a serial dilution of positivecontrol templates was included to provide quantitative comparisonsbetween samples (limit of detection 0.01 ng rRNAgene templateDNAfor both subdivision 4 and 6 primer sets) and establish amplificationefficiency. Metagenomic DNA isolated from soil samples or soil frac-tions was used as a template in a quantitative PCRwith 100 ng of soilgenomicDNA in a 50ml reaction including 1� iQ SYBRGren Supermix(BioRad Laboratories, Hercules, CA) containing 100 nM of eachsubdivision-specific primer. The quantitative comparison betweenAcidobacteria subdivisions4 and6 indifferent soil fractionswasbasedupon the calibrated curve of DNA (cloned 16S rRNA genes) from therespective Acidobacteria subdivisions used as template in positivecontrol reactions (n ¼ 4 for all reactions).

3. Results

3.1. Phylogenetic analysis of Acidobacteria rRNA genes

Every Acidobacteria 16S rRNA gene sequence available in theGenBank nr/nt database was imported into the ARB phylogeneticanalysis software package. Of the 73 Acidobacteria sequencesobtained from a phylogenetic analysis of WMARS soil, 58 affiliatedwith Acidobacteria subdivision 6 (Hugenholtz et al., 1998a,b), whichwas the dominant Acidobacteria subdivision in this soil in multiplephylogenetic surveys (Liles et al., 2003). No chimeric rRNA genesequences were identified from this study. A total of 10 Acid-obacteria subdivisions were supported by bootstrap analysis, with

subdivisions 9 and 10 with limited numbers of affiliated sequences(note that the uranium-contaminated soil sequences which indi-cated additional Acidobacteria subdivisions (Barns et al., 2007) hadnot yet been included within this database and were also not foundwithin WMARS soil). Using the probe-finding functions of ARB, theAcidobacteria clades with at least 75% bootstrap support werechosen to identify 18-, 19-, and 20-bp oligonucleotide probes thatwould be predicted to hybridize only with the specific phylogeneticlineage. In total, 76 Acidobacteria lineages were supported bybootstrap support, and 42,194 oligonucleotide probes were chosenon the basis of phylogenetic specificity.

3.2. Effect of betaine on non-specific hybridization signal

The effect of adding 2.5 M betaine to the microarray hybrid-ization buffer in increasing hybridization stringency was assessedusing a positive control Acidobacteria 16S rRNA gene template (i.e.,clone P220H). First, a range of concentrations of the positive controltemplate was hybridized to the Acidobacteria phylochip, using10 ng, 50 ng, 100 ng, or 250 ng of PCR amplicon, respectively,without including betaine in the hybridization buffer. Theseexperiments indicated that 100 ng of rRNA gene amplicon providedthe best signal to noise from the microarray hybridization (data notshown).

To test the effect of adding betaine to the hybridization buffer onthe degree of background hybridization, 100 ng of a single, labeledrRNA gene amplicon (i.e., clone 220H) was hybridized under twoconditions, with and without 2.5 M betaine, with all otherhybridization conditions remaining identical. The subset of probespredicted to be positive had a mean normalized (relative to theuniversal Acidobacteria probe set) hybridization signal of 80% withbetaine (Fig. 1A), and 78% without added betaine (Fig. 1B), a differ-ence that was not statistically significant. However, the subset ofprobes predicted to be negative had a mean normalized hybrid-ization signal of 13% in the absence of betaine (Fig. 1B), whichdropped to only 5% when 2.5 M betaine was added to the hybrid-ization buffer, a reduction of 62% of the non-specific backgroundhybridization (Fig. 1A). Using nonparametric hypothesis testingprocedures we tested if expected-negative probes with andwithout the addition of betaine would cause the same backgroundhybridization, and concluded that there is strong statisticalevidence that the expected-negative probes with and without theaddition of betaine are different (P < 0.0001) at a ¼ 0.05 signifi-cance level. As a result, the inclusion of betaine in the hybridizationbuffer results in much lower background hybridization for allexpected-negative probes.

3.3. Positive control phylochip hybridizations

A set of positive control hybridizations was conducted usingsingle, cloned Acidobacteria 16S rRNA genes under the higherstringency hybridization conditions achieved by including 2.5 Mbetaine within the hybridization buffer. Each of the “expectedpositive” phylogenetic groups produced a hybridization signal(mean for all probe hybridization signals per group) that wassignificantly greater than the mean value for all “expected-nega-tive” with each of the positive control rRNAs, indicating the pres-ence of an Acidobacteria taxa in subdivision 5 (clone P17F) (Fig. 2A)or an Acidobacteria taxa in subdivision 6 (clone P147G) (Fig. 2B).Other positive control 16S rRNA gene amplicons from clones 16H1(subdivision 4) and clone 85E8 (subdivision 6) also revealed levelsof hybridization signal for all expected-positive lineages that weregreater than the mean hybridization signal for the microarray (datanot shown); specifically, for positive control amplicons from clones16H1 and 85E8 the mean normalized hybridization signal for all

Fig. 1. Betaine improves hybridization stringency. Normalized hybridization signal froman Acidobacteria phylogenetic microarray using as template a single cloned 16S rRNAgene (clone P220H, subdivision 4) in the presence (A) or the absence (B) of 2.5M betainein the hybridization buffer. Themean hybridization signal from each phylogenetic probe(in triplicate)was groupedwith everyother probe from the samephylogenetic level, andthe mean hybridization signal for each phylogenetic group was then normalized bycomparison to the set of universal Acidobacteria probes. Each phylogenetic group isrepresented by a separate bar on the X-axis, and ten of the Acidobacteria subdivisions areindicated below their respective clades. The bar in blue indicates the universal probe setthat were used to normalize the microarray data, and the bars in red represent theexpected-positive phylogenetic groups for this hybridization.

Fig. 2. Control 16S rRNA gene microarray hybridizations yield expected phylogeny.Normalized hybridization signal from a phylogenetic microarray using as templatea single cloned 16S rRNA gene from (A) Acidobacteria subdivision 5 (clone P17F) orfrom (B) Acidobacteria subdivision 6 (clone P147G). The hybridization included 2.5 Mbetaine to achieve higher stringency hybridization, and all hybridization signals werenormalized by comparison to universal probes. The bar in blue indicates the universalprobe set that were used to normalize the microarray data, and the bars in redrepresent the expected-positive phylogenetic groups for each hybridization.

M.R. Liles et al. / Soil Biology & Biochemistry 42 (2010) 739e747 743

expected-negative features was 0.098 and 0.093, and the meannormalized hybridization signal for all expected-positive featureswas 1.14 and 0.54, respectively. As the template concentration wasidentical in all hybridizations, the differences in the normalizedhybridization signal for expected-positive features betweendifferent rRNA gene sequences (e.g., clone 16H1 versus clone 85E8)is a consequence of different probe hybridization kinetics. As ex-pected with large numbers of distinct oligonucleotide probes, theabsolute signal intensity is not a quantitative indicator of rRNAtemplate concentration, rather it is the change in the normalizedsignal strength of specific phylogenetic groups relative to otherphylogenetic groups from array-to-array that is predicted to revealchanges in the relative abundance of bacterial taxa. In some cases,an expected-negative phylogenetic group also produced a signifi-cant hybridization signal; however, upon checking the individualprobes that produced higher signals, thesewere always observed tocorrespond to probe sequences with only a single base pairmismatch with the positive control rRNA. In other words, the vastmajority of probes that were expected to be negative in the controlhybridization reactions did not give a significant hybridizationsignal, and only the probes that had a sequence with a single base

pair mismatch from the positive control rRNA sequence werecapable of a significant degree of mismatch hybridization.

3.4. WMARS soil phylochip hybridizations

The soil sample from which the positive control rRNAs werederived was used as a template to produce a soil Acidobacteria PCRamplicon. Since a phylogenetic analysis had been previously per-formed on this soil sample (Liles et al., 2003) to some extent thediversity of Acidobacteria subdivisions present in this soil samplecould be predicted and compared to the resultant phylogeneticresults from microarray hybridization. Furthermore, the Acid-obacteriadivision-level probe thatwasused, althoughnot capable ofencompassing all describedAcidobacteria subdivisions, did howevercompletely cover the Acidobacteria taxa known to predominatewithin this specific soil sample. To encompass an exhaustivesampling ofAcidobacteria taxawithin a rRNA amplicon derived froma previously uncharacterized environmental sample, it wouldsimply require using multiple Acidobacteria division-level primers(or universal Bacteria primers, in the case of a domain Bacteriaphylochip). Two replicate soil metagenomic DNA extractions hadbeen performed on this soil sample, and were used independentlyfor two separate PCR amplifications of Acidobacteria rRNA genes andsubsequent microarray hybridizations. The hybridization signals

Fig. 3. Two distinct Acidobacteria lineages observed in WMARS soil and a soil clayfraction. Normalized hybridization signal from a phylogenetic microarray using astemplate metagenomic DNA extracted from (A) a WMARS soil sample, or (B) a clayfraction of this same WMARS soil sample. Each phylogenetic group was normalized bycomparison with the set of universal probes, and the mean value of all (non-universal)probes is indicated by a dotted line. The phylogenetic lineage from subdivision 4(4.1.2.2.2.2.2.2.1) that exceeded the mean probe value is indicated in red, whereas thesubdivision 6 lineage (6.2.3.1.1.2.1) is indicated in yellow. Other phylogenetic groupswith hybridization signals exceeding the mean (non-universal) probe signal are indi-cated in white. Note that of the phylogenetic groups indicated in white, other phylo-genetic clades within the same subdivision do not show significant strength of signal,resulting in less support for their presence within the soil sample.

M.R. Liles et al. / Soil Biology & Biochemistry 42 (2010) 739e747744

resulting from the soil Acidobacteria amplicons were highly similarto eachother (data not shown), had significantlygreater backgroundhybridizations signals compared to control hybridizations, and bothindicated the high relative abundance of Acidobacteria taxa insubdivisions 4 and 6, respectively (Fig. 3A). An Acidobacteria lineagewas determined to be present within the soil sample if there wasa strong hybridization signal from a particular phylogenetic lineagebeginning with the subdivision level. A strong hybridization signalwas defined as a normalized,meanhybridization signal for a specificphylogenetic group that is greater than the mean hybridizationsignal for all probes (indicated on Fig. 3 as a dashed blue line). TheWMARS soil microarray results indicated that two distinct Acid-obacteria lineages, in subdivision 4 (lineage 4.1.2.2.2.2.2.2.1) andsubdivision 6 (lineage 6.2.3.1.1.2.1), were present in WMARS soil,with strong hybridization signals observed atmultiple phylogeneticlevels for each lineage (Fig. 3A). Other phylogenetic groups (i.e.,lineages 1.1, 1.3, 1.25, 3.1, 3.12, and 5.3 indicated in white on Fig. 3A)did exceed the minimum cutoff value for a strong hybridizationsignal, but were rejected since the subdivision-level probes did notsupport their presence (i.e., subdivisions 1, 3, and 5 signals wereinsufficiently strong, respectively). Subdivision 2 probes did givea mean value greater than the cutoff level for significance, yet no

other subdivision 2-specific clade gave a strong hybridization signal.The presence of subdivisions 4 and 6 inWMARS soil was supportedby the hierarchical lineages indicated by probe hybridizations, andby the strength of the hybridization signals for some lineages insubdivisions4 and6,whichexceeded themeanvalueof all probes bymore than 3-fold in some cases. Some of the positive control rRNAsused to validate the microarray hybridization conditions (i.e., cloneP16H in subdivision 4, and clone P147G in subdivision 6)were also inthese same respective Acidobacteria lineages, again reflecting thehigh relative abundance of these bacterial taxa in this WMARS soil.No significant difference was observed between the replicate soilsample hybridizations, as the signal strength for both subdivision 4and 6 lineages were strongly supported.

3.5. WMARS soil clay fraction phylochip hybridizations

The microarray hybridization signals observed when meta-genomic DNA isolated from the WMARS soil clay fraction was usedas a template closely resembled the hybridization pattern observedfrom the whole soil sample. Both of the lineages that were stronglysupported to be present in theWMARS soil (i.e., 4.1.2.2.2.2.2.2.1 and6.2.3.1.1.2.1) were also observed with the two replicate clay frac-tions (Fig. 3B). The other similarity to the entire soil hybridizationincluded a greater degree of non-specific hybridization ascompared to positive control hybridizations. However, the magni-tude of the signals from the lineages in subdivisions 4 and 6,although both strongly supported, were observed to vary relative toone another by comparisonwith thewhole soil sample. Specifically,the normalized signal strength of lineage 4.1.2.2.2.2.2.2.1 wasconsistently stronger than lineage 6.2.3.1.1.2.1 in the whole soilsample relative to the clay fraction in both of the replicate samples.The punctured normal distribution (Lai et al., 2004) which is thespecial case of a truncated normal distributionwas used to test if ananalysis of the Acidobacteria community within the clay fraction ofthis soil suggested a decrease in the abundance of subdivision 6relative to subdivision 4. We concluded that the mean ratio of thesubdivision 4 lineage relative to the subdivision 6 lineage for clay isgreater than this same ratio in bulk soil, supporting the indicationof a decrease in the abundance of the subdivision 6 lineage relativeto the subdivision 4 lineage. There was a consistent difference inthe ratio of subdivision 4 and 6 in the replicate microarrayhybridizations, with the ratio of subdivision 4 to subdivision 6lineages (using a mean value for all respective sub-lineages) being86.3% and 87.9% for the two replicate soil samples, and 108.3% and106.6% for the two replicate clay fraction samples.

3.6. Quantitative PCR of Acidobacteria subdivisions in soil fractions

Subdivision 4- and subdivision 6-specific primer sets were cali-brated using known concentrations of cloned and PCR amplified 16SrRNA genes, from the respective subdivisions, for quantitativeassessmentof these two subdivisions in soil samples and soil fractions.Both of the subdivision-specific primer sets yielded PCRproductswithat least 0.01ngof added16S rRNAgene template andgave comparableyield of amplicons at equal template concentrations.

The subdivision-specific quantitative PCR data from whole soiland soil fractions revealed a significantly higher amount of subdi-vision 6 template in all samples, relative to subdivision 4, and thatin the clay fraction ofWMARS soil the subdivision 6 templatewas atits lowest concentration relative to subdivision 4 taxa compared toall other soil samples and fractions (P < 0.01 by Student's t-test).Therefore, the ratio of subdivision 4 template levels relative tosubdivision 6 template levels were at their very highest within theclay fraction compared to whole soil or any other soil fraction(Fig. 4).

Fig. 4. Quantitative PCR analysis indicates an increased ratio of Acidobacteria subdi-vision 4 relative to subdivision 6 within the clay fraction of WMARS soil. Subdivision-specific primer sets were used to quantitatively determine the amount of subdivision 4and subdivision 6-specific templates, respectively, relative to total metagenomic DNAextracted from WMARS soil and different soil fractions. The ratio of subdivision 4template DNA compared to subdivision 6 template DNA was determined for each soilsample or soil fraction, with a statistically significant increase in this ratio observed inthe soil clay fraction relative to all other samples (P < 0.01 by Student's t-test). Lettersindicate statistical groupings and error bars indicate standard deviations.

M.R. Liles et al. / Soil Biology & Biochemistry 42 (2010) 739e747 745

4. Discussion

The massively parallel nature of microarray hybridizations,combined with a large database of 16S rRNA gene sequences,enables rapid identification of bacterial phylotypes present withinpure cultures or environmental samples. This ability is especiallyuseful when the bacterial taxa may be identified by a phylogeneticsignature, yet be recalcitrant to cultivation. In our study thebacterial division Acidobacteria was chosen as the focus ofa phylogenetic microarray to target its diverse taxa that had beenshown to be present at a high relative abundance in 16S rRNA geneclone libraries in many soils, and had been extensively studied froma soil in Madison, WI (WMARS soil). Due to the difficulties inherentin conducting a molecular phylogenetic census from a soil sample,and the potential advantages that would accrue from being able toapply a phylogenetic microarray approach to previously unchar-acterized soil microbial assemblages, this study was initiated to testspecific hypotheses regarding the predictive power of a phyloge-netic microarray in reflecting the bacterial taxa present withina soil microbial assemblage, using a well defined (as best as wasavailable) soil sample for this analysis.

The first hypothesis tested in this study was that the use ofa hierarchical design of oligonucleotide probes, targeting bacterialtaxa at different phylogenetic levels, could be used to identifyspecific lineages of the Acidobacteria division. In many studies usingphylogenetic microarrays, only previously characterized oligonu-cleotide probes were included that had been validated againsta cultured bacterial target. In our study, due to the lack of culturedisolates for many clades of the division Acidobacteria, this was notfeasible. Furthermore, the very large number of unique oligonu-cleotide probes identified from an ARB database phylogeneticanalysis of the Acidobacteria division precluded individual assess-ment of each probe prior to its inclusion on the microarray. Rather,three different approaches were taken to validate the oligonucle-otide probes on the phylogenetic microarray: 1) A set of controlhybridizations with 16S rRNA genes of known sequence were usedto evaluate the degree of expected-positive probe hybridizationsignal strength, relative to expected-negative signals from knownnegative probes (for their respective control rRNA gene sequences).From evaluating the positive control hybridizations using

increasing concentrations of betaine, a higher degree of hybrid-ization stringency was achieved which was then used in allsubsequent experiments; 2) Every oligonucleotide probe includedon the microarray was evaluated using in silico predictions ofthermodynamic properties (i.e., % G þ C content, DG, Tm, and DG-fold), and probe outliers were identified. A comparison of resultsfrom each of the positive control and soil sample experiments, withand without exclusion of the outlier probes, did not yield anystatistically different results. Likely, this is a function of the largenumber of separate probes included on the microarray that targeteach phylogenetic clade, so that any small number of outlier probes(e.g., resulting in self-complementarity and/or decreased binding torRNA gene targets) did not significantly affect the overall meanhybridization results; 3) The presence of a particular phylogeneticlineage was strongly supported by the presence of a stronghybridization signal at multiple phylogenetic levels. For example,from the soil hybridization results the lineage 6.2.3.1.1.2.1 wasevident from both of the replicatemicroarray hybridizations. In thiscase, the probes at phylogenetic levels 6, 6.2, 6.2.3, 6.2.3.1, 6.2.3.1.1,6.2.3.1.1.2, and 6.2.3.1.1.2.1 each gave a mean hybridization signalsignificantly greater than was observed for the mean value for allprobes included on the microarray; thus, the cumulative data forthe presence of lineage 6.2.3.1.1.2.1 within the WMARS soil wasderived from multiple phylogenetic levels, as well as the previousphylogenetic analysis of WMARS soil that had demonstrated theabundance of this subdivision 6 clade. In conclusion, the datasupport the utility of a hierarchical design for a phylogeneticmicroarray in enabling the inclusion of previously uncharacterizedoligonucleotide probes with phylogenetic predictive power.

In interpreting data from a phylogenetic microarray hybridiza-tion, it is important to consider the potential biases inherent in a PCRand hybridization-based approach. To begin with, PCR of 16S rRNAgenes extracted from a microbial assemblage will preferentiallyamplify the most abundant bacterial taxa. This biased representa-tionwithin a heterogeneous rRNA gene ampliconwill be reflected inthe downstream analysis, whether that is rRNA gene clone librarysequencing, denaturing gradient gel electrophoresis, or phylochiphybridization (Holben et al., 2004). The selection of primers withwhich to amplify the targeted bacterial taxa is also a significantvariable. In this study, the combined use of Acidobacteria-specificprimer 31F with universal Bacteria primer 1492R was appropriategiven that the WMARS soil sample had been previously character-ized. In futurework using previously uncharacterized soils samples,it is advisable to use newly developed Acidobacteria-specific primersets that will be inclusive of more Acidobacteria subdivisions (Barnset al., 2007). The potential for chimeric rRNA gene amplicons is alsoa concern, given that use of a Chimera Check program is not possibleusing phylochip hybridization data. As in this study, validation ofhybridization results using quantitative PCR or other independentmethod is advisable to confirm and extend the hybridizationobservations. Lastly, using a massively parallel hybridization designrequires use of thousands of oligonucleotide probes each withdistinct hybridization kinetics. Given that it is not feasible to inde-pendently validate each probe using a positive control target rRNAgene, this study adopted an in silico analysis of probe predictedthermodynamic characteristics, the results ofwhich did not supportthe rejection of probes based solely on in silico analyses. This latterobservation is likely the beneficial result of including many probestargeting each phylogenetic clade, which reduces the respectivecontribution of probes that may have some degree of self-comple-mentarity or inefficient hybridization kinetics under the selectedhybridization conditions.

The second hypothesis tested in this study was that a phyloge-netic microarray could be used to identify changes in the relativeabundance of bacterial taxa within a soil fraction. Since the relative

M.R. Liles et al. / Soil Biology & Biochemistry 42 (2010) 739e747746

abundance of Acidobacteria taxa may be altered in different soilmicrohabitats, it was of interest to compare the results of anAcidobacteria microarray hybridization from a whole soil sampleversus the clay fraction from that same soil sample. The microarrayhybridizations from both whole soil and a clay fraction of this soilrevealed many similarities, including a higher non-specifichybridization and two distinct Acidobacteria lineages (i.e.,4.1.2.2.2.2.2.2.1 and 6.2.3.1.1.2.1) whose presence was stronglysupported both by themagnitude of themean hybridization signalsrelative to the mean value of all probes, and by the number ofseparate sub-lineages that were positive in either case. Also inter-esting was the apparent decrease in the hybridization signal foreach of the phylogenetic groups in lineage 6.2.3.1.1.2.1 from the clayfraction hybridization relative to the whole soil results. It should benoted that the mean probe hybridization signals cannot be taken tobe indicative of the quantitative abundance of any phylogeneticgroup present in an environmental sample, due to potential kineticdifferences between individual probes; however, if a comparison isconducted of the same set of probes with different environmentalsamples, a shift in probe hybridization strength from one sample toanother could indicate a real shift in bacterial relative abundance.These data therefore predict that, within the clay fraction ofWMARS soil, a lower relative abundance of subdivision 6 bacterialtaxa would be observed.

To test the hypothesis that there is a difference in the relativeabundance of Acidobacteria subdivisions within a clay fraction of soilversusbulk soil, quantitativePCRwasemployed to compare the levelsof subdivision 4 and subdivision 6 rRNA gene templates within thewhole WMARS soil and in the different soil fractions. In the bulk soiland in all soil fractions the quantitative PCR results indicated thatsubdivision 6 taxa were present at higher levels than subdivision 4taxa. These results are supported by previous results from a 16S rRNAgene survey from this same soil sample, which recorded a higherrelative abundance of subdivision 6 rRNA gene clones than subdivi-sion 4 rRNAgene clones (Liles et al., 2003). Even though subdivision 6taxa were present at high levels in every soil fraction, the soil clayfraction revealed a significant decrease in the quantitative PCR signalfor subdivision 6 taxa, so that the ratio of subdivision 4 taxa tosubdivision 6 taxa increased substantially in the clay fraction.Therefore, the predictive power of the phylogenetic microarray inrevealing a previously unpredicted distribution of Acidobacteria taxawithin a clay fractionwas supported by the quantitative PCR analysis.

Our collective understanding of soil Acidobacteria taxa will beenhanced both through recent culture-based and genomic studies(Ward et al., 2009), and through a better understanding of theirdistribution in soils and soil microhabitats. Severe reductions in theabundance and/or diversity of soil Acidobacteria have beenobserved in response to soil pollutants such as 2,4,6-trinitrotoluene(George et al., 2009), phenylurea herbicides (El Fantroussi et al.,1999), or hydrocarbons (2005). In contrast, soil Acidobacteria taxahave been observed to be abundant in PCB- and parathion-pollutedsoils (Nogales et al., 1999; Debarati et al., 2006), suggesting thecapacity for some Acidobacteria taxa to degrade specific soilpollutants. Furthermore, it would be of interest to use a phylochipapproach to investigate the prevalence of Acidobacteria and otherbacterial taxa in the inner- and outer-microaggregate soil fractions(Kim et al., 2008), considering that Acidobacteria were found tohave a reduced abundance within the inner-aggregate fraction inmultiple soils (Mummey et al., 2006). Both human activities andnatural ecosystem processes may have substantial effects on theabundance, diversity, and/or metabolic activities of soil Acid-obacteria, and the availability of such a rapid means to detectdiverse Acidobacteria (and other bacterial) taxa present in soilmicrobial assemblages can be a powerful tool to investigatechanges in soil bacterial population structure. This study

demonstrates the feasibility of this phylogenetic microarrayapproach using positive control amplicons and a soil sample thathad been previously studied. Future work will expand beyond thedivision Acidobacteria to encompass a broader phylogenetic diver-sity of soil bacteria, and may be applied to many different studies ofsoil microbial ecology.

Acknowledgements

We thank the other members of the Goodman and Liles labo-ratories for their support and critical analysis of this work, and Dr.Ena Urbach for useful scientific discussions. The gene expressioncenter at the University of Wisconsin is thanked for all of theirtechnical help in this work. This research was funded by grantsfrom the David and Lucille Packard Foundation, the McKnightFoundation, NSF DEB-0213048, and Auburn University's Depart-ment of Biological Sciences, College of Sciences and Mathematics,and Vice President for Research.

References

Amann, R.I., Ludwig, W., Schleifer, K.H., 1995. Phylogenetic identification and in situdetection of individual microbial cells without cultivation. MicrobiologicalReviews 59, 143e169.

Barns, S.M., Fundyga, R.E., Jeffries, M.W., Pace, N.R., 1994. Remarkable Archaealdiversity detected in a Yellowstone national park hot spring environment.Proceedings of the National Academy of Sciences of the United States ofAmerica 91, 1609e1613.

Barns, S.M., Takala, S.L., Kuske, C.R., 1999. Wide distribution and diversity ofmembers of the bacterial kingdom Acidobacterium in the environment. Appliedand Environmental Microbiology 65, 1731e1737.

Barns, S.M., Cain, E.C., Sommerville, L., Kuske, C.R., 2007. Acidobacteria phylumsequences in uranium-contaminated subsurface sediments greatly expand theknown diversity within the phylum. Applied and Environmental Microbiology73, 3113e3116.

Billor, N., Hadi, A., Velleman, P., 2000. BACON: blocked adaptive computationally-efficient outlier nominators. Computational Statistics and Data Analysis 34,279e298.

Bintrim, S.B., Donohue, T.J., Handelsman, J., Roberts, G.P., Goodman, R.M., 1997.Molecular phylogeny of Archaea from soil. Proceedings of the NationalAcademy of Sciences of the United States of America 94, 277e282.

Brodie, E.L., DeSantis, T.Z., Joyner, D.C., Baek, S.M., Larsen, J.T., Andersen, G.L.,Hazen, T.C., Richardson, P.M., Herman, D.J., Tokunaga, T.K., Wan, J.M.,Firestone, M.K., 2006. Application of a high-density oligonucleotide microarrayapproach to study bacterial population dynamics during uranium reductionand reoxidation. Applied and Environmental Microbiology 72, 6288e6298.

Burgmann, H., Pesaro, M., Widmer, F., Zeyer, J., 2001. A strategy for optimizingquality and quantity of DNA extracted from soil. Journal of MicrobiologyMethods 45, 7e20.

Debarati, P., Pandey, G., Meier, C., van der Meer, J.R., Jain, R.K., 2006. Bacterialcommunity structure of a pesticide-contaminated site and assessment ofchanges induced in community structure during bioremediation. FEMSMicrobiology Ecology 57, 116e127.

Dojka, M.A., Harris, J.K., Pace, N.R., 2000. Expanding the known diversity andenvironmental distribution of an uncultured phylogenetic division of bacteria.Applied and Environmental Microbiology 66, 1617e1621.

El Fantroussi, S., Verschuere, L., Verstraete, W., Top, E.M., 1999. Effect of phenylureaherbicides on soil microbial communities estimated by analysis of 16S rRNAgene fingerprints and community-level physiological profiles. Applied andEnvironmental Microbiology 65, 982e988.

Felsenstein, J., 1993. PHYLIP (Phylogeny Inference Package), Version 3.5c. Distrib-uted by the Author. Department of Genetics, University of Washington, Seattle.

George, I.F., Liles, M.R., Hartmann, M., Ludwig, W., Goodman, R.M., Agathos, S.N.,2009. Changes in soil Acidobacteria communities after 2,4,6-trinitrotoluenecontamination. FEMS Microbiology Letters 296, 159e166.

Head, I.M., Saunders, J.R., Pickup, R.W., 1998. Microbial evolution, diversity, andecology: a decade of ribosomal RNA analysis of uncultivated microorganisms.Microbial Ecology 35, 1e21.

Holben, W.E., Feris, K.P., Kettunen, A., Apajalahti, J.H.A., 2004. GC fractionationenhances microbial community diversity assessment and detection of minoritypopulations of bacteria by denaturing gradient gel electrophoresis. Applied andEnvironmental Microbiology 70, 2263e2270.

Hugenholtz, P., Goebel, B.M., Pace, N.R., 1998a. Impact of culture-independentstudies on the emerging phylogenetic view of bacterial diversity. Journal ofBacteriology 180, 4765e4774.

Hugenholtz, P., Pitulle, C., Hershberger, K.L., Pace, N.R., 1998b. Novel division levelbacterial diversity in a Yellowstone hot spring. Journal of Bacteriology 180,366e376.

M.R. Liles et al. / Soil Biology & Biochemistry 42 (2010) 739e747 747

Huyghe, A., Francois, P., Charbonnier, Y., Tangomo-Bento, M., Bonetti, E., Paster, B.J.,Geneva Study Group on Noma (GESNOMA), et al., 2008. Novel microarraydesign strategy to study complex bacterial communities. Applied and Envi-ronmental Microbiology 74, 1876e1885.

Janssen, P.H., Yates, P.S., Grinton, B.E., Taylor, P.M., Sait, M., 2002. Improved cultur-ability of soil bacteria and isolation in pure culture of novel members of thedivisions Acidobacteria, Actinobacteria, Proteobacteria, and Verrucomicrbia.Applied and Environmental Microbiology 68, 2391e2396.

Kim, J.S., Dungan, R.S., Crowley, D., 2008. Microarray analysis of bacterial diversityand distribution in aggregates from a desert agricultural soil. Biology andFertility of Soils 44, 1003e1011.

Kishimoto, N., Kosako, Y., Tano, T., 1991. Acidobacterium capsulatum gen. nov., sp.Nov.: an acidophilic chemoorganotrophic bacterium containing menaquinonefrom acidic mineral environment. Current Microbiology 22, 1e7.

Kuske, C.R., Barns, S.M., Busch, J.D., 1997. Diverse uncultivated bacterial groupsfrom soils of the arid southwestern United States that are present in manygeographic regions. Applied and Environmental Microbiology 63,3614e3621.

Lai, C.D., Wood, G.R., Qiao, C.G., 2004. The mean of the inverse of a puncturednormal distribution and its application. Biometrical Journal 46 (4), 420e429.

Lane, D.J., Pace, B., Olsen, G.J., Stahl, D.A., Sogin, M.L., Pace, N.R., 1985. Rapiddetermination of 16S ribosomal RNA sequences for phylogenetic analysis.Proceedings of the National Academy of Sciences of the United States ofAmerica 82, 6955e6959.

Liesack, W., Bak, F., Kreft, J.U., Stackebrandt, E., 1994. Holophaga foetida gen. Nov., sp.Nov., a new, homoacetogenic bacterium degrading methoxylated aromaticcompounds. Archives of Microbiology 162, 85e90.

Liles, M.R., Manske, B.F., Bintrim, S.B., Handelsman, J., Goodman, R.M., 2003. Acensus of rRNA genes and linked genomic sequences within a soil metagenomiclibrary. Applied and Environmental Microbiology 69 (5), 2684e2691.

Ludwig, W., Bauer, S.H., Bauer, M., Held, I., Kirchhof, G., Schulze, R., Huber, I.,Spring, S., Hartmann, A., Schleifer, K.H., 1997. Detection and in situ identificationof representatives of a widely distributed new bacterial phylum. FEMS Micro-biology Letters 153, 181e190.

Markham, N.R., Zuker, M., 2005. DNAMelt web server for nucleic acid meltingprediction. Nucleic Acids Research 33, W577eW581.

Markham, N.R., Zuker, M., 2008. DNAFold: software for nucleic acid folding andhybriziation. In: Keith, J.M. (Ed.), Bioinformatics. Structure, Functions andApplications, Number 453 in Methods in Molecular Biology, vol. II. HumanaPress, Totowa, NJ, ISBN 978-1-60327-428-9, pp. 3e31 (Chapter 1).

McSweeney, C.S., Mackie, R.I., Odenyo, A.A., Stahl, D.A., 1993. Development of anoligonucleotide probe targeting 16S rRNA and its application for detection andquantitation of the ruminal bacterium Synergistes jonesii in a mixed-populationchemostat. Applied and Environmental Microbiology 59, 1607e1612.

Medlin, L., Elwood, H.J., Stickel, S., Sogin, M.L., 1988. The characterization of enzy-matically amplified eukaryotic 16S-like ribosomal RNA-coding regions. Gene 71,491e500.

Meier, H., Amann, R., Ludwig, W., Schleifer K.-, H., 1999. Specific oligonucleotideprobes for in situ detection of a major group of gram-positive bacteria with lowDNA G þ C content. Systematic and Applied Microbiology 22, 186e196.

Mummey, D., Holben,W., Six, J., Stahl, P.D., 2006. Spatial stratification of soil bacterialpopulations in aggregates of diverse soils. Microbial Ecology 51, 404e411.

Nogales, B., Moore, E.R.B., Abraham, W.R., Timmis, K.N., 1999. Identification of themetabolically active members of a bacterial community in a polychlorinatedbiphenyl polluted moorland soil. Environmental Microbiology 1, 199e212.

Pace, N.R., Stahl, D.A., Lane, D.J., Olsen, G.J., 1986. The analysis of natural microbialpopulations by ribosomal RNA sequences. Advances inMicrobial Ecology 9,1e55.

Palmer, C., Bik, E.M., Eisen, M.B., Eckburg, P.B., Sana, T.R., Wolber, P.K., Relman, D.A.,Brown, P.O., 2006. Rapid quantitative profiling of complex microbial pop-ulations. Nucleic Acids Research 34, e5.

Poll, C., Thiede, A., Wermbter, N., Sessitch, A., Kandeler, E., 2003. Micro-scaledistribution of micro-organisms and microbial enzyme activities in a soil withlong term organic amendment. European Journal of Soil Science 54, 715e724.

Rondon, M.R., August, P.R., Bettermann, A.D., Brady, S.F., Grossman, T.H., Liles, M.R.,et al., 2000. Cloning the soil metagenome: a strategy for accessing the geneticand functional diversity of uncultured microorganisms. Applied and Environ-mental Microbiology 66, 2541e2547.

Rousseeuw, P., van Driessen, K., 1999. A fast algorithm for the minimum covariancedeterminant estimator. Technometrics 41, 212e223.

Torsvik, V., Goksoyr, J., Daae, F.L., 1990. High diversity in DNA of soil bacteria.Applied and Environmental Microbiology 56, 782e787.

Torsvik, V., Goksøyr, J., Daae, F.L., Sørheim, R., Michalsen, J., Salte, K., 1994. Use ofDNA analysis to determine the diversity of microbial communities. In: Ritz, K.,Dighton, J., Giller, K.E. (Eds.), Beyond the Biomass. John Wiley and Sons, Chi-chester, pp. 39e48.

Torsvik, V., Sorheim, R., Goksoyr, J., 1996. Total bacterial diversity in soil and sedi-ment communities e a review. Journal of Industrial Microbiology 17, 170e178.

Tuschak, C., Glaeser, J., Overmann, J., 1999. Specific detection of green sulfur bacteriaby in situ hybridization with a fluorescently labeled oligonucleotide probe.Archives of Microbiology 171, 265e272.

Ward, D.M., Weller, R., Bateson, M.M., 1990. 16S rRNA sequences reveal numerousuncultured microorganisms in a natural community. Nature 345, 63e65.

Ward, N.L., Challacombe, J.F., Janssen, P.H., et al., 2009. Three genomes from thephylum Acidobacteria provide insight into the lifestyles of these microorgan-isms in soils. Applied and Environmental Microbiology 75 (7), 2046e2056.

Wilson, K.H., Wilson, W.J., Radosevich, J.L., DeSantis, T.Z., Viswanathan, V.S.,Kuczmarski, T.A., Andersen, G.L., 2002. High-density microarray of small-subunit ribosomal DNA probes. Applied and Environmental Microbiology 68,2535e2541.

Woese, C.R., 1987. Bacterial evolution. Microbiological Reviews 51, 221e271.Zimmermann, J., Gonzalez, J.M., Saiz-Jimenez, C., Ludwig, W., 2005. Detection and

phylogenetic relationships of highly diverse uncultured acidobacterialcommunities in Altimira Cave using 23S rRNA sequence analysis. Geo-microbiology Journal 22, 379e388.