SoilBiology&Biochemistry112(2017)140e152 ContentslistsavailableatScienceDirect Soil Biology & Biochemistry journal homepage: www.elsevier.com/locate/soilbio Soil depth and crop determinants of bacterial communities under ten biofuel cropping systems Bangzhou Zhang a,b, C. Ryan Penton b,c, Chao Xue b,d, John F. Quensen b, Sarah S. Roleye, Jiarong Guo b, Aaron Garoutte b, Tianling Zheng a,**, James M. Tiedje b,* aStateKeyLabofMarineEnvironmentalScienceandKeyLaboftheMOEforCoastalandWetlandEcosystems,SchoolofLifeSciences,XiamenUniversity, Xiamen,China bCenterforMicrobialEcology,MichiganStateUniversity,EastLansing,MI,USA cCollegeofLettersandSciencesandJulieAnnWrigleyGlobalInstituteofSustainability,ArizonaStateUniversity,Mesa,AZ,USA dJiangsuCollaborativeInnovationCenterforSolidOrganicWasteUtilizationandNationalEngineeringResearchCenterforOrganic-BasedFertilizers, DepartmentofPlantNutrition,NanjingAgriculturalUniversity,Nanjing,China eGreatLakesBioenergyResearchCenterandW.K.KelloggBiologicalStation,MichiganStateUniversity,HickoryCorners,MI,USA a r t i c l e i n f o a b s t r a c t Articlehistory: Biofuel-croppingsystems,projectedforlargelandareas,canpotentiallychangetheirsoilmicrobiome Received23October2016 and the ecosystem services they catalyze. We determined the bacterial community composition and Receivedinrevisedform relevantsoilpropertiesforsamplescollectedafter6cropyearsat0e10cm,10e25cm,25e50cm,and50 30March2017 e100 cm under corn, switchgrass, Miscanthus, and restored prairie, as well as 0e10 cm under six Accepted17April2017 additionalcandidatebiofuelcropsinreplicateside-bysideplots.Deepsequencingofthe16SrRNA-V4 Availableonline16May2017 regionestablishedthatsoilbacterialcommunitiesweresignificantlydifferentiatedbydepthasdeter- minedbyproportionalOTUabundanceandcomposition,UniFracdistance,andtaxonomicandindicator Keywords: analyses.Thecroppingsystemsignificantlyimpactedbacterialcommunitycompositionwithinthetop Bacterialcommunity Soildepth three layers, with corn and switchgrass communities the most different within the 0e25 cm and 25 Biofuelcrops e50cmdepths,respectively.Theeffectsofcroptypeanddepthco-mingled,likelyattributedtodiffer- Soilproperties encesinrootingdepthandbiomassamongcrops.Individualphylademonstratedvaryingpatternswith depth, with significant proportional decreases of Proteobacteria, Actinobacteria, Planctomycetes, and Bacteroidetes but proportional increases of Firmicutes from shallow to deep soils. The Acidobacteria, Verrucomicrobia, and Chloroflexi peaked in abundance in the middle layers, whereas Thaumarchaeota decreasedinabundance.Importantly,someclasseswithin theAcidobacteria,Verrucomicrobia,andFir- micutes followed contrasting patterns with depth suggesting that they have different ecological spe- cializations.Poplar,followedbysoilswithperennialcropscontainedthemostCinthesurfacesoils,with dataindicatingthatthesedifferenceswillbecomemorepronouncedwithtime. ©2017ElsevierLtd.Allrightsreserved. 1. Introduction abundant biological component (Daniel, 2005; Gans et al., 2005; Rodriguez and Konstantinidis, 2014). Soil bacteria play a critical Due to spatially and temporally heterogeneous chemical and role in biogeochemical cycling and are keystones to overall structuralproperties,soilisthemostcomplexanddiversemicro- ecosystemfunction.Itiswelldocumentedthatsoilbacterialcom- bial environment on earth, with microorganisms as its most munity composition is influenced by soil properties, geographic distance, plant species, land use/management, and environment type(Daniel,2005;Acosta-Martinezetal.,2008;Ushioetal.,2010; Caporaso et al., 2011; Rodrigues et al., 2013). However, most soil * Corresponding author. 540 Plant and Soil Sciences Building, Michigan State microbialecologystudieshavefocusedsolelyonthesurfacehori- University,EastLansing,MI48824,USA. zons and were previously limited by low-resolution genetic ** Correspondingauthor.C-318,SchoolofLifeSciences,XiamenUniversity,Xia- profiling methodologies (Muyzer et al.,1993; Fierer and Jackson, men361102,China. E-mail addresses: [email protected] (T. Zheng), [email protected] 2006; Hartmann and Widmer, 2006; Wakelin et al., 2008). As (J.M.Tiedje). http://dx.doi.org/10.1016/j.soilbio.2017.04.019 0038-0717/©2017ElsevierLtd.Allrightsreserved. B.Zhangetal./SoilBiology&Biochemistry112(2017)140e152 141 such, there is currentlylittle knowledge concerning soil bacterial and species are given in Table S1. These cropping systems were communitycompositionanddiversityindeepersoillayers,wherea selectedtohavearangeofplantdiversitiesandofexternal/man- significant proportion of microbial biomass resides (Fierer et al., agementinputs. 2003). Three 2.5-cm diameter intact soil cores weretaken fromeach Biofuelsareasustainablealternativetofossilfuelsandhavethe plotwithaGeoprobe540MT(GeoprobeSystems,Salinas,KS)and potential to address emerging energy demands while providing transportedtoa4(cid:3)Ccoolerwithin3hofcollectionandstoredthere otherecosystemservices,suchascarbonsequestration,enhanced until processing, which usually occurred within one week of biodiversity, and reduced greenhouse gas emissions (Lemus and collection. Upon processing, soil cores were sectioned into Lal, 2005; Hill et al., 2006; Robertson et al., 2008; Gelfand et al., 0e10cm,10e25cm,25e50cm,and50e100cmslicesfromG1,G5, 2013; Werling et al., 2014; Oates et al., 2015). Biofuel crops are G6 and G10. Soils of the other systems were collected only at projectedtocoverlargeareasoflesser-studiedlandscapesunsuited 0e10cm.Eachcoresectionwassieved(4mm)toremoverootsand for food crop production (Gelfand et al., 2013), with ensuing stonesandthenthethreesamplesfromthesameplotanddepth changes in the soil microbiome. Meanwhile, soil microbes can were combined into a single composite sample. Each composite promote biofuel feedstock yields directly or indirectly by fixing samplewasdivided,withportionsfrozenforlaterDNAextraction, atmosphericnitrogen,increasingphosphorusacquisition,recycling air-dried for nutrient analysis, and oven-dried at 60 (cid:3)C for total nutrients, improving soil aggregation, and suppressing plant dis- carbon(C)andnitrogen(N)analysis.Intotal,110compositesam- ease(Swiftetal., 2004;TiedjeandDonohue,2008). Near-surface ples((4treatments(cid:1)4depthsþ6treatments(cid:1)1depth)(cid:1)5plots) and root zone soil microbial community composition is influ- werecollected. enced by plant type due to the annual/perennial nature of the plants, root depth, differing root exudates, plant residue decom- 2.2. Soilproperties posability, and active recruitment of specific microbial taxa. In contrast, sub-surface soil microbial communities are likely more TodeterminethetotalCandNcontent,sub-samplesfromeach relatedtolong-termCsequestrationbecausesubsoilorganicmat- depthofeachcropthathadbeenoven-driedwerepulverizedand terischaracterizedbyhighmeanresidencetimesandenrichedin combusted in a Costech Elemental Combustion System 4010 microbially-derived carbon compounds (Rumpel and Kogel- (Costech Analytical Technologies, Valencia CA). To determine soil Knabner, 2011; Liang et al., 2012a). Thus characterizing the im- pH,potassium (K),phosphorus(P),calcium(Ca),andmagnesium pactsofbiofuelcroppingsystemsonbothsurfaceandsub-surface (Mg) concentrations, sub-samples that had been air-dried were microbial communities is important in understanding long-term analyzedatMichiganStateUniversitySoilandPlantNutrientLab- systemsustainabilityandenvironmentalimpacts. oratory using standard methods (http://extension.missouri.edu/ Our objectives were to: i) determine shifts in the microbial explorepdf/specialb/sb1001.pdf). community structure and diversity due to crop and depth, ii) determine taxonomic patterns, especially with depth since that 2.3. DNAextraction,PCRamplification,andMiSeqsequencing presumably reflects the physiologies of the favored taxa, iii) determineindicatorspeciesforcropanddepthandthedriversthat Soil samples were thawed and total DNA was extracted from correlatewiththem,andiv)wherepossible,inferpotentialfunc- 0.3 gportionsofeachsoil sampleusing theMoBioPowerSoil Kit tionoftheselectedtaxa.Todothis,wedeterminedsoilmicrobial (MoBio, Carlsbad, CA), according to the manufacturer's protocol. community composition after 6 crop years by Illumina MiSeq TheresultingDNAyieldandqualitywerecheckedwithanND1000 sequencingof16S-V4rRNAgenesatfoursoildepthsthatcovered device(NanoDrop,Wilmington,DE),followedbyre-extractionand the rooting zone and beyond of four primary candidate biofuel pooling of some samples with lower yields, which occurred for crops:corn,switchgrass,Miscanthus,andrestoredprairieaswellas some50e100cmsamples. thesurfacesoilsfromthreeothercorn-basedcroppingsystemsand Bacterial 16S rRNA PCR amplification using a dual-index threeotherperennialbiofuelcrops. sequencing strategy (Kozich et al., 2013) was conducted to generate the amplicon library for MiSeq sequencing. Briefly, 2. Methods ampliconsweregeneratedinthereactionsystemconsistingof17ml of AccuPrime Pfx SuperMix (Invitrogen, CA, USA), 1 ml of DNA 2.1. Samplecollection (around20ng/ml),and1mlofeachbarcodedprimer(10mM).The fusionprimersetwithIlluminaadapterandbarcodestargetedthe Samples were collected post-harvest in November 2013 from hypervariable V4 region with specific forward primer 50- ten biofuel cropping systems (G1 - G10) of the Great Lakes Bio- GTGCCAGCMGCCGCGGTAA and reverse primer 50-GGAC- energyResearchCenter(GLBRC)BiofuelsCroppingSystemExper- TACHVGGGTWTCTAAT (Caporaso et al., 2011). The cycling condi- iment (BCSE) at Kellogg Biological Station in southwest MI, USA. tionswere95(cid:3)Cfor2min,30cyclesof95(cid:3)Cfor20s,55(cid:3)Cfor15s, Firstestablishedin2008,theBCSEexperimentaldesignutilizesa 72(cid:3)Cfor1min,and72(cid:3)Cfor10min.PCRproductswerepurified randomized complete block design with 5 replicate blocks and normalized using the SequalPrep Normalization Plate Kit (30(cid:1)40m).Thesitehadthesameorverysimilarcroppinghistory (Invitrogen,CA,USA),followedbypoolingof5mlofeachsampleto prior to establishing the current plot design in 2008. The soil is produce a sample library. Sequencing was performed by the predominantly Kalamazoo loam (Fine-Loamy, Mixed, Semiactive, ResearchTechnologySupportFacility,MSU,forMiSeqpaired-ends Mesic Type Hapludalf), a sandy loam with 47e56% sand. These sequencing with the 250 bp kit (Standard v2 flow cell with 500 cropping systems are continuous corn (G1), continuous cycles).RawsequencesweredepositedintheNCBISequenceRead cornþcovercrop(G2),soybeaninasoybean-cornrotationþcover ArchiveunderaccessionnumberPRJNA296796. crop (G3), corn in a corn-soybean rotation þ cover crop (G4), switchgrass (Panicum virgatum, G5), Miscanthus (Miscanthus x 2.4. Bioinformaticanalyses giganteus,G6),nativegrassmix(G7),hybridpoplar(Populusnigrax Populus maximowiczii, G8), early successional (G9), and restored Rawpaired-endreadswereassembledusingtheRDPAssembler prairie(G10).Previously,G2,G3,andG4wereinacorn-soybean- (Cole et al., 2014) with a read Q score cutoff of 28. Primers were canola rotation from 2008 to 2011. Details on cropping histories trimmedandtheresultingreads(averaging253bp)werechimera 142 B.Zhangetal./SoilBiology&Biochemistry112(2017)140e152 checkedusingUCHIME(Edgaretal.,2011)indenovomode.OTU (ANOVA,p<0.001),whileCaandMgfirstincreasedwithdepthand clustering was performed by UPARSE Following the pipeline, all thendecreased(highestat25e50cm,ANOVA,p <0.001).More- chimera-freereads weredemultiplexed intoone file, clusteredat over,continuouscorn(G1)generallyhadsignificantlylowerC,N, 97% similarity, representative sequences generated, singletons pH,Ca,andMgthantheotherbiofuelcropsystems(especiallythe removed, and a final OTU table was created. Representative se- poplar, G8) at 0e10 cm depth (Table 1), while switchgrass (G5), quences of OTUs were submitted for taxonomic classification by Miscanthus(G6),nativegrassmix(G7),andrestoredprairie(G10) RDP Classifier at a confidence of 50% (Wang et al., 2007). OTUs containedsignificantlylowerKthancorn-relatedsoils(especially classified as chloroplasts or unclassified at the domain level as thecontinuouscorn,G1).Therewerenosignificantdifferencesin Bacteria or Archaea were removed (For simplicity, “bacterial” is soil characteristics among the continuous corn, switchgrass, Mis- used throughout since Archaea accounted for less than 3% of the canthusandtherestoredprairieinthedeeperlayers. sequences). This did leave OTUs unclassified to a bacterial or archaealphylum.Allsampleswerere-sampledto25,000sequences 3.2. Diversitymeasures per sample for downstream analyses. All bioinformatic analyses wereperformedusingtheHighPerformanceComputationalCenter Atotalof5,006,088highqualityreadswereobtainedafteras- facilitiesatMichiganStateUniversity.Alldatafileswereorganized sembly which produced 20,179 unique OTUs at a 97% sequence usingRpackagephyloseq(McMurdieandHolmes,2013). similaritycutoff,18,354ofwhichoccurredinsamplesfromG1,G5, G6,andG10(Fig.S1).Shared“universal”OTUs(foundinsamples 2.5. Statisticalanalyses fromallfourof thesecropsat all fourdepths)accountedfor 31% (5655) of the total (Fig. S1). Surface soils contained the highest Differences in communitystructure werevisualized using db- proportion(12%)ofexclusiveOTUs,followedbythebottomlayer RDA based on generalized UniFrac distances (a ¼ 0.5), which (8%),whiletheintermediatelayersaccountedfortheleast(5and weregeneratedfromrepresentativesequencesofOTUsbyFastTree 6%each). andtheRpackageGUniFrac(Chenetal.,2012a).Thesignificance Forthefourcropssampledoverdepth,bacterialdiversity(H0), was tested by PERMANOVA with 999 permutations (using the observednumberofspecies(Sobs),andevenness(J’)werehighestin function adonis in R package vegan), and the correlations of soil the top 10 cm of soil and decreased significantly in each deeper propertiestothedb-RDApatternswerecalculatedusingtheenvfit layer(Fig.2,leftpanels,ANOVA,a<0.05).Comparingcropswithin functionintheRpackage vegan(Oksanenetal.,2015).Alphadi- eachsoillayer,bacterialdiversityindexesweresignificantlylower versity indexes, including observed richness (Sobs), Shannon di- incornsoilsthanswitchgrassinthe10e25cmlayer(Fig.2,right versity(H0),andPielou'sevenness(J0),werecomputedbasedonthe panels), butotherwise no marked differences were found among OTU table using the vegan package as well. Shared OTUs were crops(Fig.2andFig.S2).Foralldatacombined,bacterialdiversity calculated and plotted using the R package VennDiagram (Chen, indexes were strongly associated with soil total C and N content 2014). The indicator analysis based on genera (singleton and (Fig.S3). doubleton genera removed first) was conducted using R package indicspecies(DeCaceresandLegendre,2009)with999permuta- 3.3. Communitystructures tionsandallowingforcombinationsofdepthsandcrops,andtheP- values were corrected for multiple comparison using R package There were significant differences in microbial community qvalue(DabneyandStorey,2015)withafalsediscoveryrateof10% structure among the ten crops in the 0e10 cm layer (p < 0.001, (q (cid:4) 0.1). Other measurements, including soil properties, were Table 2). On ordination, annual corn treatments were separated analyzed and plotted using the R package ggplot2 (Wickham, from the perennial crop treatments by the first db-RDA axis 2009), while NMDS of soil properties was performed using the (Fig.3A).Continuouscorn(G1)wasseparatedfromcorntreatments veganpackage.Allone-wayANOVAtestswereconductedbyone- withcovercropbyaxis2(Fig.3A),andcontinuouscornwithcover way.test and pairwise.t.test with P-value adjustment method of crop(G2)wasseparatedfromcornwithcovercropinrotation(G3 “BH” for multiple comparisons. Data were square-root- or log- andG4)byaxis3(Fig.3B).Similarly,twoclusterswereobserved transformed to meet normality and homogeneity of variance. within the six perennial systems: monocultures G5, G6, and G8 Otherwise,kruskalinRpackageargicolae(Mendiburu,2015)was (switchgrass,Miscanthus,poplar)andpolyculturesG7,G9,andG10 used. Pearson's correlations and their significances of a-diversity (nativegrassmix,earlysuccessional,andrestoredprairie)(Fig.3A). indexes, abundances of individual phyla and classes to soil prop- Furthermore, G8 was separated from the other monocultures by ertieswerecalculatedbyfunctionsofcorandcor.testinRpackage axis3(Fig.3B).Thus,sixgroups intotalare distinguishable. Cor- stats(RCoreTeam,2014).TheseanalyseswereperformedusingR relations between site scores and environmental variables were 3.1.2(RCoreTeam,2014). relativelyweakbutstatisticallysignificantexceptforK;valuesfor allbutPwerelowerforthecornsites(Fig.3andTableS2). 3. Results Ordinationofallsamplesforthefourcropssampledwithdepth (G1,G5,G6,andG10)indicatedthattheirmicrobialcommunities 3.1. Soilproperties wereshapedprimarilybyfactorsassociatedwithdepth(Fig.4A). Site scores were significantly correlated with all measured envi- Soilchemicalpropertiesdifferedwithdepthandamongcrop- ronmental variables, especially C (r2 ¼ 0.91) and N (r2 ¼ 0.88) pingsystems(Fig.1andTable1).NMDSordination(Fig.1)showed (TableS2).Whilesuchdepthrelatedfactorsplayedadominantrole that samples separated primarily by depth (PERMANOVA, in shaping communities, PERMANOVA revealed a significant P<0.001).Cropwasaminorfactor(PERMANOVA,P¼0.037),and interactionbetweendepthandcrops(Table2),leadingustomake there was no significant interaction between crop and depth furthertestsforeachseparatesoillayer.Theseresultsindicatedthat (PERMANOVA,P¼0.750).Fromthesurfacesoil(0e10cm)tothe cropsaffectedsoilcommunitiesforallbutthedeepest(50e100cm) bottom (50e100 cm), soil total C and total N contents decreased layer(Table2). (ANOVA,p<0.001)by12and7times,respectively.ThepHwasalso Forthe0e10cmlayer,ordinationseparatedsamplesforthefour significantlylowerinthebottomlayer(ANOVA,p<0.01).SoilKand cropsinthesamewayaswhenall10cropswereincluded.G5and P concentrations were significantly higher in the surface soil G6clusteredtogether,G1andG10separately(Fig.4B).Onlyfourof B.Zhangetal./SoilBiology&Biochemistry112(2017)140e152 143 Fig.1. Comparisonsofsoilchemicalpropertiesamongdepthsunderfourcrops(G1,G5,G6,andG10),andNMDSordinationbasedontheseproperties.Thebottomandtopofabox arethe25thand75thquartiles,thehorizontallinewithinaboxisthemedian,andtheendsofthewhiskersarethelimitsofthedistributionasinferredfromtheupperandlower quartiles.Dotsareoutliers.Blackstarsaremeans.LettersindicatetheANOVAgroupingamongdepths.BothsoilCandNareinunitsofg/100g(weight%),whileP,K,Ca,andMgare allinunitsofmg/g(ppm). Table1 Comparisonsofsoilchemicalpropertiesamongcropsat0e10cm.LettersindicatetheANOVAgroupingamongcrops. Crops# C* N** pH** P K** Ca Mg*** G1 1.148±0.118b 0.111±0.009c 5.9±0.2c 57±43 173±56a 748±148 85±11c G2 1.265±0.172ab 0.120±0.013bc 6.4±0.7ab 48±24 148±35ab 992±248 131±40abc G3 1.208±0.115b 0.117±0.013bc 6.2±0.3abc 45±16 140±42abc 914±159 126±24ab G4 1.195±0.172b 0.116±0.015bc 6.2±0.2abc 50±15 149±26ab 938±163 125±12b G5 1.141±0.170b 0.110±0.013c 6.3±0.1abc 33±13 97±17c 863±167 124±12b G6 1.257±0.152ab 0.114±0.013bc 6.1±0.1bc 40±9 111±62c 935±83 126±15b G7 1.292±0.126ab 0.124±0.011bc 6.4±0.5abc 33±7 111±22bc 1056±57 156±16a G8 1.558±0.295a 0.148±0.024a 6.4±0.1ab 54±10 182±45a 1000±107 145±3ab G9 1.365±0.135ab 0.136±0.010ab 6.5±0.2ab 47±30 169±51a 996±225 172±17a G10 1.208±0.109b 0.116±0.012bc 6.5±0.1a 35±12 110±29bc 915±138 154±37ab #G1andG2areannuallycontinuouscornandcontinuouscornþcovercrop,respectively;whileG3andG4aresoybeanþcovercropandcornþcovercrop,respectively,ina soybean-cornrotation.G5-G10areperenniallyswitchgrass,Miscanthus,nativegrassmix,poplar,earlysuccessional,andrestoredprairie,respectively.Significance:*P<0.05, **P<0.01,***P<0.001. the environmental variables were significantly (a (cid:4) 0.05) associ- distinctinthetopthreelayers(0e50cm), andbecamemoreho- atedwiththesitescores.Kwaspositivelyassociatedwiththecorn mogeneous(withlessvariation)atthe50e100cmdepth(Fig.S4B). sampleswhilepH,Mg,andCawerepositivelyassociatedwiththe othercrops.Clusteringofcropswasdifferentforthedeeperdepths. 3.4. Taxonabundances Inthe10e25cmlayer,G5andG10clusteredtogetherwhileG1and G6 formed separate clusters (Fig. 4C). In the 25e50 cm layer, G5 OTUswereclassifiedto10majorphyla(eachmorethan1%of formed a separate cluster, and while the other crops showsome thetotalsequences,Table3)thattogetheraccountedfor81.0%of separation in Fig. 4D, no differences among themwere found by the total sequences, and 22 minor phyla accounting for an addi- PERMANOVA(datanotshown).Therewerenosignificantcorrela- tional3.7%.Atotalof14.2%ofthesequenceswereunclassifiedat tionsbetweenenvironmentalvariablesandsitescoresfordepths the phylum level (unclassified_Bacteria, Table 3). Relative abun- otherthan0e10cm,andnosignificantdifferencesamongcropsfor dances of individual phyla were highly variable with depth and the50e100cmlayer(Table2). among cropping systems, and were significantly correlated with Generalized UniFrac distances between pairs of soil layers soil properties (Table 3, Fig. 5, Tables S3 and S4). Proteobacteria, increased with the distance between the layers being compared Actinobacteria,Planctomycetes,Bacteroidetes,andcandidateWPS-2 (Fig. S4A). Pairwise comparisons of distances between cropping decreased from the surface to deeper soils, while Firmicutes and systems also showed that bacterial communities were more unclassified Bacteria increased. The most abundant phylum, 144 B.Zhangetal./SoilBiology&Biochemistry112(2017)140e152 andTableS5).Thesecontrastingpatternsofclasseswithinthesame phylawith soil depth were also observed in Verrumicrobia (Spar- tobacteriavsSubdivision3)andChoroflexi(KtedonobacteriavsCal- dilineae)andevenintheconsistently-increasingphylumFirmicutes (ClostridiavsBacilli)(TableS5),butwerenotdetectedinclassesof consistently-decreasing phyla Proteobacteria and Bacteroidetes. Similarly,corn-relatedsoilshadlowerabundancesofGp6andGp17 butgreaterabundancesofGp1andGp3at0e10cm.Switchgrass soilsaccumulatedmoreDelta-andGamma-butlessAlpha-proteo- bacteriaat25e50cm,aswellashigherpercentageofsubdivision3 and Verrucomicrobiae but lower Spartobacteria of Verrucomicrobia (Fig.5andTableS6). 3.5. Indicators Indicator analysis identified 271 genera significantly (q < 0.1) associated with a specific depth or a combination of depths (Table S7), of which 91 genera were clearly classified and had relativeabundancesover0.005%(Fig.7).Usingthesameanalysis, 20, 5, and 7 indicator generawere characteristic (q < 0.1) of one crop or a combination of crops at 0e10 cm, 10e25 cm, and 25e50cm,respectively(TableS8).Ofthese,27generawereunique and clearly classified (Fig. 8). Indicator genera characteristic of 0e10cmand10e25cmwererelativelysimilar(moreshared),but differentfromthoseatthetwodeeperlayersthatweresimilarto each other as well. Corn-related soils were characterized by in- dicators different from those associated with perennial crops at 0e10 cm and 10e25 cm, while switchgrass soils had more distinctive indicators at 25e50 cm. Overall, most of indicator Fig.2. Comparisonsofbacterialdiversityindexesamongsoildepthsandamongcrops generabelongedtoProteobacteriaandActinobacteria,followedby atthedepthof10e25cm.ThefourcropabbreviationsaredefinedinTable1.Letters BacteroidetesandAcidobacteria. indicatetheANOVAgroupingamongdepthsandcrops(a¼0.05).Explanationofboxes andstarsasforFig.1.Sobs,observednumberofOTUs;H0,Shannondiversityindex;J0 Pielou'sevennessindex. 4. Discussion In this study, we observed significant differences in soil pro- Table2 karyotic communities and soil properties among four soil depths Effectofsoildepth,crop,andtheirinteractionsonbacterialcommunitystructures (down to 100 cm) under four different biofuel crop systems and basedonPERMANOVAandgeneralizedUniFracdistances(a¼0.5). among10cropsinnearsurfacesoilsafter6cropgrowingseasons. Depth Crop Depth(cid:1)Crop Asaproxyforchangingsoiledaphicfactors,especiallytotalCandN, F R2 F R2 F R2 soil depth was a major driver of bacterial community structure 0-10cm,10crops e e 1.64*** 0.28 e e among thefourcropping systemsexamined overdepth, whereas 4depths(cid:1)4crops 19.51*** 0.42 2.43** 0.05 1.47* 0.10 crop effects were secondary and significantly correlated with 0-10cm,4crops 2.14*** 0.31 severalsoilvariablesat0e10cm(Fig.3).Moreover,ourindicator 10-25cm,4crops e e 1.39** 0.24 e e analyses identified exclusive and shared genera characteristic of 25-50cm,4crops e e 2.20*** 0.29 e e soilsindifferentlayersandsoilsunderdifferentcropsatthesame 50-100cm,4crops e e 1.24ns 0.19 e e depth,whichsuggestspatternstobetestedmorebroadlyaswellas Significance:*P<0.05,**P<0.01,***P<0.001. forconsistencyoftheirdrivingecologicalfeatures. The significant decrease in alpha diversity with depth is Acidobacteria, as well as Verrucomicrobia and Chloroflexi were mirrored in previous studies using pyrosequencing (Will et al., 2010; Eilers et al., 2012; Tsitko et al., 2014), clone libraries generallymoreabundantinthemiddlelayers.Therelativeabun- (Hansel et al., 2008), and fingerprinting approaches (Fiereret al., danceofThaumarchaeota,themostabundantArchaea,showedabi- 2003; Agnelli et al., 2004). Eilers et al. (2012) observed that mi- modal distribution with depth. Among cropping systems, corn- crobialcommunitiesbasedonweightedUniFracdistancesamong relatedsoilsharboredmoreProteobacteria(especiallyAlpha-)and Chloroflexi, but less Acidobacteria than perennial crop soils at sites were most variable in the near-surface horizons, relatively 0e10 cm. Less abundant Actinobacteria and Planctomycetes but uniformatintermediatedepths,andagainhighlyvariableatdeep moreunclassifiedBacteriawerepresentincornsoilsat10e25cm. layers.Indeed,inourstudysamplevariabilitywithinthesamelayer basedongeneralizedUniFracdistanceswashighestinthesurface SwitchgrasssoilscontainedmoreBacteroidetesinlayersfrom25to 100cm,andmoreunclassifiedBacteriabutlessThaumarchaeotaat anddeepestsoilsformostcases(Fig.S4B).Thesoildepthassociated 25e50cmaswellaslessSpartobacteriabutmoresubdivision3of withcommunitychangeappearstoextendtoapproximately50cm, with communities becoming less similar (greater phylogenetic theVerrucomicrobiathantheotherthreecrops. distance) with increasing physical distance. Indicator analyses Although the predominant classes of Acidobacteria (Gp6, Gp4, revealedmanyspecificgeneraforsoildepthswithdiverseputative Gp16,Gp17)displayedsimilartrendsasthephylumwithsoildepth, functions: for example, biodegradation/decomposition (Actino- someclasses(Gp7,Gp1,Gp2)wereproportionallymoreabundant planes(GoodfellowandWilliams,1983),Cellvibrio(Mergaertetal., withsoildepth,whileGp3remainedquitestablewithdepth(Fig.6 2003)) and nitrogen-fixation (Microvirga (Ardley et al., 2012), B.Zhangetal./SoilBiology&Biochemistry112(2017)140e152 145 Fig.3. Differencesinsoilbacterialcommunitystructuresamongcropsforthe0e10cmlayer.db-RDAconstrainedoncropswasfittedwithsignificantlycorrelatedsoilproperties. Fig.4. Differencesinsoilbacterialcommunitystructuresamongdepths(A)andundercropsatthesamedepth(BeD).Thevariationsexplainedbydepthsorcropsandstatistics frompermutationtestsaregiveninTable2db-RDAsconstrainedbydepthsorcropswerethenfittedwithsignificantlycorrelatedsoilproperties.Thereisnoplotforthe50e100cm layerbecausetherewerenosignificantdifferencesinsoilbacterialcommunitiesamongcrops(Table2).Datawerebasedonthe4cropswithdepth. Azoarcus(Reinhold-Hureketal.,1993))inshallowsoils,aswellas waterholdingcapacity,andredoxstatuscanvaryamongsoilho- denitrification(Microvirgula(Patureauetal.,1998))andutilization rizons and also impact which species thrive or senesce at each of recalcitrant compounds (Rubrobacter (Carreto et al., 1996)) in depth.Paststudiesingrasslands(Willetal.,2010)andpeat(Tsitko deeperlayers.WhileCandNwerehighlycorrelatedtomicrobial etal.,2014)havealsodocumentedastrongcorrelationtosoiltotal communitychanges, other soil properties such as aggregate size, C and N. Microbial biomass is well known to decline with depth 146 B.Zhangetal./SoilBiology&Biochemistry112(2017)140e152 Table3 Relativeabundancesofmajorphyla(>1%)andtheirchangeswithsoildepthandcorrelationstosoilproperties.MoredetailsinTableS3. Phylum Abundance F-value Changes Correlationto soil variables with depth C N pH P K Ca Mg Acidobacteria 28.68±4.57 24.21*** 0.47*** 0.43*** 0.59*** -0.09 0.09 0.26* 0.15 Proteobacteria 14.28±3.82 118.24*** 0.89*** 0.88*** 0.38*** 0.44*** 056*** -0.01 -0.161 unclassified_Bacteria 14.24±4.34 106.22*** -0.89*** -0.88*** -0.30** -0.32*** -0.51*** -0.13 -0.04 Verrucomicrobia 9.49±2.01 17.08*** -0.46*** -0.49*** -0.37*** -0.27* -0.09 0.22 0.39*** Actinobacteria 6.86±1.85 9.53*** 0.45*** 0.47*** 0.31** 0.35** 0.27* 0.03 -0.04 Firmicutes 5.42±4.57 75.10*** -0.73*** -0.70*** -0.53*** -0.17 -0.35** -0.21 -0.02 Chloroflexi 4.91±2.39 27.32*** -0.60*** -0.65*** -0.42*** -0.24* -0.14 0.31** 0.36** Planctomycetes 4.58±1.18 46.49*** 0.74*** 0.77*** 0.33** 0.33** 0.38*** 0.09 -0.15 Bacteroidetes 2.90±1.84 60.72*** 0.81*** 0.81*** 0.38*** 0.26* 0.48*** 0.01 -0.07 Thaumarchaeota 2.26±0.92 25.41*** 0.50*** 0.55*** -0.02 0.32** 0.15 -0.36** -0.40*** candidateWPS-2 1.64±0.93 248.30*** 0.90*** 0.91*** 0.40*** 0.25* 0.46*** 0.12 -0.08 Significance: *P<0.05,**P<0.01,***P<0.001. (Agnellietal.,2004;Eilersetal.,2012),consistentwithourlowDNA 2010;Shietal.,2011). extraction yield, especially at 50e100 cm. Overall, the effects of Differencesinmicrobialcommunitystructureamongcropshave depth(0e10cmvs.10e25cm)hadagreaterinfluenceonbacterial alsobeenattributedtotheperennialnatureandrootingsystemsof communities than selective pressures from different cropping the crops. Corn has a shorter photosynthetic life cycle, shallower systemsat0e10cm.Similarly,itwasreportedthatthedeptheffects rooting depth, lower root densities, and less root biomass than onmicrobialcommunities,evenas littleas10e20cmwithin the perennial grasslands, including switchgrass (Jackson et al.,1996; samecore,canreachasimilarextentasthatofsurfacesoilssepa- Tufekcioglu et al.,1999; Culman et al., 2010). Indeed, we identi- rated by thousands of kilometers (Eilers et al., 2012). Together, fiedabundantrootsinswitchgrasssoildownto100cm,butalmost these results suggest that soil depth integrates environmental norootsincornsoildeeperthan25cm.Furthermore,arecentstudy gradientsalongsoilprofilesandformsastrongecologicalfilterfor demonstratedthatswitchgrassrootlengthdensitiesandrootdry selecting/shapingmicrobialcommunitieswithin different depths, weight (both peaking at 30e45 cm) are much higher than Mis- asproposedbyEilersetal.(2012). canthus(peakingat0e10cm,90%concentratinginthetop35cm) In addition to the overwhelming depth influences, we also inrelativelydeepersoil(MontiandZatta,2009),althoughwehave identifiedsignificanteffectsofcropsonsoilmicrobialcommunities. notobservedsuchdifferenceshere(Sprungeretal.,unpublished). These crop effects were constrained within the top 50 cm, pre- Together,theseobservationssuggestthatswitchgrassmaypossess sumablyduetorhizosphericinfluencessuchasplantrootexudates, themostuniquerootstructureand/orqualityandquantityofroot root cap sloughing, and root turnover (Acosta-Martinez et al., exudatesandhencesupportthehighestuniquenessoftheassoci- 2008).Perennialplantsareknowntoproducemoreabundantmi- atedsoilmicrobialcommunitiesatthe25e50cmdepthasseenin crobial resources through rhizodeposition and root turnover and thisstudy.Switchgrassaccumulatesgreatermicrobialbiomassthan can thus accommodate a larger population of more diverse mi- corn and this influence of resource availability can move deeper croorganisms(Liangetal.,2012b).Inaddition,croprotation,tillage within the soil over time (to 30 cm after 9 years) (Stewart et al., (including annual planting), and other management influences 2015). We detected significantly higher diversity in switchgrass (suchasfertilizationandherbicides)servetodifferentiateperen- thancontinuouscornat10e25cmafter6croppingyears.There- nial from annual crops. The distant grouping of continuous corn fore,giventheinherentlydifferentrootingsystems,wepredictthat awayfromtheotherthreecorn-basedsystemsandtheseparation switchgrasswilleventuallysignificantlyshapemicrobialcommu- of perennial monocultures from multispecies grasses can infer nities as deep as 100 cm over time through root-mediated C these impacts (Fig. 4). Specifically, corn soils harbored the most allocation. distinct communities,compared totheperennial biofuel cropsin ThedominancesofAcidobacteria,Proteobacteria,Verrucomicro- thetop25cm,similartoresultscomparingcorn,switchgrass,and bia,Actinobacteria,Firmicutes,andChloroflexiobservedinthisstudy Miscanthussurfacesoils(downto24cmdepth)(Liangetal.,2012b; havebeendetectedinvarioussoiltypes,biomes,anddepths(Jesus Mao et al., 2013; Jesus et al., 2016). This microbial pattern under etal.,2010;Willetal.,2010;Eilersetal.,2012;Liangetal.,2012b; differentcropswasalsoreflectedinindicatorgenera.Forexample, Tsitko et al., 2014), which indicates the ubiquity and important continuous corn soils at 0e10 cm, which were lower in pH than rolesofthesephyla.However,therelativeabundancesofthesetaxa othercropsoils(Table1),werecharacterizedbyRhodanobacterthat changed distinctly with soil depth and among cropping systems. iscapableofdenitrificationanddominateatlowpH(Greenetal., The most striking of these changes was associated with Proteo- 2012; Kostka et al., 2012). In contrast, perennial switchgrass and bacteria,whoserelativeabundancedecreasedwithdepth,atrend Miscanthus soils were strongly associated with lignin-degrading observedinotherstudiesaswell(Eilersetal.,2012;Tsitkoetal., Sphingobacterium (Wang et al., 2013), while switchgrass soils 2014). Given the strong correlation to soil C (Table 3), the were also characterized by lignin-degrading Novosphingobium decreaseofproteobacterialabundanceswithdepthislikelydueto (Chen et al., 2012b) at 10e50 cm. The plant growth-promoting their lifestyle, as they are more abundant in soil with higher C Acinetobacter, an indicator at lower depth layers, especially in availability (Fierer et al., 2007; Eilers et al., 2010; Goldfarb et al., switchgrass soils might provide benefits by fixing nitrogen, solu- 2011). This decreasing trend was consistent for all four proteo- bilizing P, producing siderophores and indole-3-acetic acid, and bacterialclasses.Croptype,possiblyinconjunctionwithsampling promoting root-growth (Huddedar et al., 2002; Sachdev et al., time,alsoinfluencedProteobacteria,withAlphaproteobacteriaand B.Zhangetal./SoilBiology&Biochemistry112(2017)140e152 147 Fig.5. Differencesinrelativeabundancesofmajorphyla(>1%)undercropsatthesamedepths.Starsanderrorbarsarethemeansandstandarddeviations,respectively.Firstletters ofthephylumnameswereusedtosavespace.LettersindicatetheANOVAgroupingamongdepthsandcrops.Moredetailsandmorephylaaswellastheircorrelationstosoil propertiesareinTableS4. Gammaproteobacteria more abundant in corn-based soil at Granulicellaat0e10cm,whichhydrolyzesawiderangeofsugars 0e10cm.ThismayreflectmorelabileorganicCinthesurfacesoil, andcomplexpolysaccharides(Rawatetal.,2013).Additionally,this perhapsduetothedegradationofdeadcornroots,asthesampling ecologicallifestyle-basedhypothesisexplainsthedeclinesofActi- timewasapproximatelyonemonthafterharvest.Moredeadroots nobacteria and Bacteroidetes (Fierer et al., 2007) with depth. in corn soils were also supported by the observed indicator Planctomycetes(IvanovaandDedysh,2012)andcandidatedivision 148 B.Zhangetal./SoilBiology&Biochemistry112(2017)140e152 Fig.6. Relativeabundancesofbacterialclasseswithinthesamephylumwithsoildepth.Starsanderrorbarsarethemeansandstandarddeviations,respectively.Lettersindicate theANOVAgroupingamongdepthsandcrops.MoredetailsandmoreclassesaswellastheircorrelationstosoilpropertiesareinTableS5. WPS-2(Tsitkoetal.,2014)havealsobeenreportedtodecreasewith abundant in the four corn systems than in the perennial ones, depth, consistent with our observation. Recently, Planctomycetes suggestingthatpossiblyCquality,soilproperties(e.g.significantly wereshowntobecapableofaerobicdegradationofplantsaccha- positive correlation to pH, Table S6), or other management prac- rides (Erbilgin et al., 2014), which agrees with their decreasing ticesmayimpacttheirstatus.Moreover,acidobacterialabundances abundanceswithdepth. have been found to be negatively correlated with pH (Hartman Otherbacteriallineagesincreasedwithdepthorpeakedinthe etal.,2008;Jonesetal.,2009).Thiscorrelationwasnotobserved middlelayers.Acidobacteria(themostabundant),Verrucomicrobia, at the phylum level in this study. Rather we found some classes andChloroflexipeakedatmid-layers,whileFirmicutesandunclas- (especiallyGp1 and Gp2) werenegativelycorrelatedtopH while sifiedBacteriaconsistentlyincreasedwithdepth.Acidobacteriaare some(e.g.Gp6,Gp4)werepositivelycorrelatedwithpH,similarto wellknowntobewidelydistributedalongthesoilprofiles(Hansel earlier findings (Jones et al., 2009). Distinct genera within Acid- etal.,2008;Eilersetal.,2012),andareconsideredtobeoligotrophs obacteriaidentifiedas indicatorsassociated with different depths withhigherabundancesassociatedwithlesserCresources(Fierer canalsobeexplainedbytheirvariousphysiologies. et al., 2007). This may explain increases in relative abundances The peak in Verrucomicrobia abundance in the middle depths fromthesurfacetothemid-layer(higherCtolowerC).However,it wasalsoshownpreviously,andwassuggestedtobepartiallylinked is not clear what drives a decrease in relative abundance in the totheiroligotrophicclassification(Eilersetal.,2012).Similartothe deeper layers. Although this is an overall trend at the coarse Acidobacteria, we also observed variations in trends between the phylumlevel,subgrouppopulationsmaypossessvaryingphysiol- twomajorclasses,SpartobacteriaandSubdivison3,withthelatter ogies(Fiereretal.,2007)thatdrivevariationswithsoildepthand more abundant in switchgrass soil at 25e50 cm. The Firmicutes edaphic properties. For example, most subgroups (e.g. Gp7, Gp1, werepredominatelyrepresentedbytheobligatelyanaerobicClos- and Gp2) increased with depth, reflecting oligotrophic status. In tridiainthisstudy.Theirincreasingproportionswithdepthwere contrast,themostabundantGp6decreasedwithdepth,appearing notexpectedsincethesearewell-drained,sandysoils,andlimited toexhibitamorecopiotrophic-likephysiology.However,asimple oxygenwouldnotbeexpected.However,theabilityofClostridiato C-based impact on Gp6 abundance is unlikely as they were less sporulate as well as restricted metabolism due to substrate B.Zhangetal./SoilBiology&Biochemistry112(2017)140e152 149 Fig.7. Indicatorgenerasignificantly(q<0.1)associatedwithonesoildepthoracombinationofsoildepths.Thebarindicatestherelativeabundanceofeachindicatorgenus,while thesizeofeachcirclerepresentstheindicatorvalue(associationstrength)ofaspecificgenuswiththedifferentsoildepths:0e0.25,notcharacteristic;0.25e0.5,weaklychar- acteristic;0.5e0.75,characteristic;and0.75e1.0,stronglycharacteristic.Taxonomicinformation,indicatorvalues,P-values,andq-valuesofallindicatorgeneraaregiveninTableS7. Datawerebasedon4cropswithdepth.
Description: