Determinants of the distribution of nitrogen-cycling microbial communities at the landscape scale

ORIGINAL ARTICLE

Determinants of the distribution of nitrogen-cyclingmicrobial communities at the landscape scale

D Bru1,2, A Ramette3, NPA Saby4, S Dequiedt1,2,5, L Ranjard1,2,5, C Jolivet4, D Arrouays4

and L Philippot1,2

1INRA, Department of Soil and Environmental Microbiology, UMR 1229, Dijon, France; 2University ofBurgundy, Department of Soil and Environmental Microbiology, Dijon, France; 3Microbial Habitat Group,Max Planck Institute for Marine Microbiology, Bremen, Germany; 4INRA, US 1106 InfoSol, CS 40001, Ardon,Orleans , France and 5Platform GenoSol, Dijon, France

Little information is available regarding the landscape-scale distribution of microbial communitiesand its environmental determinants. However, a landscape perspective is needed to understand therelative importance of local and regional factors and land management for the microbialcommunities and the ecosystem services they provide. In the most comprehensive analysis ofspatial patterns of microbial communities to date, we investigated the distribution of functionalmicrobial communities involved in N-cycling and of the total bacterial and crenarchaealcommunities over 107 sites in Burgundy, a 31 500 km2 region of France, using a 16� 16 km2

sampling grid. At each sampling site, the abundance of total bacteria, crenarchaea, nitrate reducers,denitrifiers- and ammonia oxidizers were estimated by quantitative PCR and 42 soil physico-chemical properties were measured. The relative contributions of land use, spatial distance, climaticconditions, time, and soil physico-chemical properties to the spatial distribution of the differentcommunities were analyzed by canonical variation partitioning. Our results indicate that 43–85% ofthe spatial variation in community abundances could be explained by the measured environmentalparameters, with soil chemical properties (mostly pH) being the main driver. We found spatialautocorrelation up to 739 km and used geostatistical modelling to generate predictive maps of thedistribution of microbial communities at the landscape scale. The present study highlights thepotential of a spatially explicit approach for microbial ecology to identify the overarching factorsdriving the spatial heterogeneity of microbial communities even at the landscape scale.The ISME Journal (2011) 5, 532–542; doi:10.1038/ismej.2010.130; published online 12 August 2010Subject Category: microbial ecology and functional diversity of natural habitatsKeywords: nitrogen cycle; landscape; denitrifiers; ammonia oxidizers; nitrate reducers; biogeography

Introduction

Spatial patterns have long been of concern inecology and have changed the manner in whichstudies of plant and animal ecology are designedand analyzed. Characterization of the patterns ofspecies diversity is central for understandingthe underlying evolutionary and ecological pro-cesses that shape biodiversity across spatialand temporal scales (Levin, 1992). Patterns alsohave implications for applied ecology, as under-standing and predicting spatial patterns are the keys

for developing ecosystem management strategies(Levin, 1992).

In contrast to plants and animals, studying spatialpatterns is recent for microorganisms (Hughes-Mar-tiny et al., 2006; Ramette and Tiedje, 2007a) and anincreasing body of literature supports the idea thatmicrobial communities exhibit spatial pattern atdifferent scales. Thus, in terrestrial ecosystems,several studies reported spatial patterns from thecentimetre to the meter scale (Nunan et al., 2002;Franklin and Mills, 2003; Ritz et al., 2004; Philippotet al., 2009a). In contrast, only a few studies haveinvestigated spatial patterns of microbial commu-nities over broad spatial scales even though spatialdependence was also observed at the kilometre scale(Cho and Tiedje, 2000; Dequiedt et al., 2009; Yergeauet al., 2009). However, such investigations at broaderspatial scales are of importance as it is well knownthat patterns can change with the scale of description(Hutchinson, 1953). Indeed a landscape perspective is

Received 13 April 2010; revised 14 June 2010; accepted 6 July2010; published online 12 August 2010

Correspondence: L Philippot, Department of Soil and Environ-mental Microbiology, INRA, UMR 1229, 17 rue Sully, Dijon Cedex21065, France.E-mail: [email protected]

The ISME Journal (2011) 5, 532–542& 2011 International Society for Microbial Ecology All rights reserved 1751-7362/11

www.nature.com/ismej

needed to understand the impact of human activities,geomorphology or climate on microbial communitydistribution. Thus, how microorganisms are spatiallydistributed at the landscape scale and which thefactors, among land management, soil physico-chemical properties and local climate, governing theirdistribution are therefore central, yet unanswered,questions despite the fact that microbial communitiesare essential for biogeochemical cycling and ecosys-tem functioning.

In this study, we investigated microbial distribu-tion at the landscape scale by focusing on thefunctional communities involved in nitrogencycling because traits rather than taxa were sug-gested to be the fundamental units of biodiversityand biogeography (Weiher and Keddy, 1995). Thepotential of such a functional trait-based approachto microbial biogeography has recently been furtheremphasized by Green et al. (2008). Nitrogen-cyclingmicrobial communities such as the ammonia oxidi-zers, nitrate reducers and denitrifiers have beendescribed as excellent models of functional com-munities (Kowalchuk and Stephen, 2001; Philippotand Hallin, 2005), of both agronomic and environ-mental importance. Thus, microbial transformationswithin the nitrogen cycle affect the bioavailability ofnitrogen, which is one of the nutrients that limitplant growth most often limiting for plant growth.Denitrification and ammonia oxidation are alsomajor contributors to the emission of N2O, a green-house gas with ca 300 times the global warmingpotential of CO2 (Forster et al., 2007) and thedominant ozone-depleting substance (Ravishankaraet al., 2009).

Here, we characterize and explain the spatialvariability in the distribution of microbial commu-nities that are involved in nitrogen cycling at thelandscape scale. The abundance of the nitrate-reducing, denitrifying and ammonia-oxidizing com-munities in soil samples, collected using a gridcovering 31 500 km2, was quantified by real-timePCR. Canonical variation partitioning was used toexamine the relative contributions of land manage-ment, spatial distance, climatic conditions, time andmore than 40 soil physico-chemical properties tothe distribution of each microbial community. Wealso used geostatistical modelling to investigatethe spatial correlations of the microbial commu-nities and produce maps of their distribution at thelandscape scale.

Materials and methods

Experimental site and samplingSoil sampling was performed using a systematic gridapproach. For this purpose, the Burgundy regionwas divided into 118 cells of about 16� 16 km2 andthe soil was collected at the center of 107 out of the118 cells (Supplementary Figure S1). This scale wasselected according to the minimum sampling density

recommended to monitor soils across Europe(Morvan et al., 2008) and is fully compatible withthe unique existing pan-European soil-monitoringnetwork (Lacarce et al., 2009). At each sampling sitelocated in the center of the cell, 25 individual soilcores were collected in the topsoil (0–30 cm), usingan unaligned sampling design within a 20� 20 m2

area. The 25 core samples were then composited foreach site. Samples of known volume were taken forbulk density determination. Soil samples wereair-dried and sieved to 2 mm before analysis. Soilsampling was achieved thanks to the French SoilQuality Monitoring Network, which collectedsoil throughout France over a 10-year period usingthe same 16� 16 km2 sampling grid. In Burgundythe 107 soil samples were collected from October2002 to October 2008 at all seasons (37 in winter, 39in spring, 13 in summer and 19 in fall).

Soil, climate and land use dataThe following soil characteristics were measured:(i) total organic carbon content and nitrogen mea-sured by dry combustion, (ii) particle-size distribu-tion using five classes (clay (0–2 mm), fine silt(2–20 mm), coarse silt (20–50mm), fine sand(50–200 mm) and coarse sand (200–2000 mm) usingwet sieving and the pipette method (NF X 31–107),(iii) cation exchange capacity and Ca, Mg, K, Na, Al,Mn exchangeable cations (cobaltihexamin method),(iv) total K, Ca, Mg, Na, Fe, Al, Cd, Co, Cr, Cu, Mn,Ni, Pb, Tl, Zn, (v) pH in water (1:5 soil:water ratio),(vi) extractable boron (boiling water method)and (vii) EDTA-extractable Cd, Cr, Cu, Ni, Pb andZn. Analyses were performed by the Soil AnalysisLaboratory of INRA in Arras, France, which isaccredited for soil and sludge analysis. Climate datacame from the SAFRAN database (Quintana-Seguiet al., 2008) and included 1992–2004 averagesof monthly and yearly evapotranspiration (ETP),temperature (1C) and rainfall (mm), interpolated onthe basis of a 8� 8 km2 grid. Land use was classifiedaccording to the Corine Land Cover databaseClassification (Heymann et al., 1994) and groupedin the following broad classes: grasslands, forest,agricultural soil, vineyard and orchards.

DNA extractionFor each of the 107 samples, DNA was extractedfrom 250 mg to 1 g of soil based on the methoddeveloped by Martin-Laurent et al. (2001), which iscurrently under final evaluation by national bodymembers of the ISO before being published asthe ISO standard 11063 ‘Soil quality—Method todirectly extract DNA from soil samples’. Eventhough the comparison for the ISO standardizationof DNA extraction from air dried, fresh, and frozensoils from different soils did not show any signifi-cant effect on 16S rRNA gene copy number per ng ofDNA (unpublished results), we cannot exclude the

Distribution of N-cycling microbial communitiesD Bru et al

533

The ISME Journal

possibility that our results might have been differentfor some soils with a different procedure. Briefly,samples were homogenized in 1 ml of extractionbuffer for 30 s at 1600 r.p.m. in a minibead beater celldisrupter (Mikro-DismembratorS; B. Braun BiotechInternational, Melsungen, Germany). Soil and celldebris were eliminated by centrifugation (14 000 gfor 5 min at 4 1C). After precipitation with ice-coldisopropanol, nucleic acids were purified using bothpolyvinylpyrrolidone and Sepharose 4B spin col-umns. Quality and size of soil DNA were checked byelectrophoresis on 1% agarose. DNA was quantifiedusing the Quant-iT dsDNA Assay Kit (Invitrogen,Paisley, UK) and a plate reader (Berthold MithrasLB940, Thoiry, France).

Real-time PCR quantification (qPCR)The total bacterial and crenarcheal communitieswere quantified using 16S rRNA primer-based qPCRassays described previously (Ochsenrelter et al.,2003). Quantification of the bacterial and crenarch-aeal ammonia oxidizers was performed according toLeininger et al. (2006) and Tourna et al. (2008)whereas quantification of nitrate reducers anddenitrifiers was performed according to Bru et al.(2007) and Henry et al. (2004, 2006), respectively.For this purpose, the genes encoding catalyticenzymes of ammonia oxidation (bacterial andcrenarchaeal amoA), nitrate reduction (narG andnapA) and denitrification (nirK, nirS and nosZ) wereused as molecular markers. Reactions were carriedout in an ABI prism 7900 Sequence DetectionSystem (Applied Biosystems, Carlsbad, CA, USA).Quantification was based on the increasing fluores-cence intensity of the SYBR Green dye duringamplification. The real-time PCR assay was carriedout in a 20 ml reaction volume containing the SYBRgreen PCR Master Mix (Absolute QPCR SYBR GreenRox, ABgene, Courtaboeuf, France), 1mM of eachprimer, 100 ng of T4 gene 32 (QBiogene, Illkrich,France) and 0.5 � 2 ng of DNA. Two independentquantitative PCR assays were performed for eachgene. Standard curves were obtained using serialdilutions of linearized plasmids containing thestudied genes. PCR efficiency for the different assaysranged between 86 and 99%. Two to three no-template controls were run for each quantitativePCR assay and no template controls gave null ornegligible values. The presence of PCR inhibitorsin DNA extracted from soil was estimated by(i) diluting the soil DNA extract and (ii) mixing aknown amount of standard DNA with soil DNAextract prior to qPCR. No inhibition was detected ineither case.

As the number of 16S rRNA operons per cell isvariable (Klappenbach et al., 2001) the 16S rRNAgene copy data were not converted into cell numbersand the results were expressed as 16S rRNA genecopy numbers per ng of extracted DNA. Calculationof the gene copy number per ng of DNA rather than

per gram of soil minimizes the bias related topossible differences in the DNA extraction yieldbetween samples. To obtain an estimate of therelative abundance of the different functional com-munities within the total bacterial or crenarchaealcommunities in the samples, we calculated ratiosbetween the ammonia-oxidation, nitrate reductionand denitrification gene copy numbers and the totalbacterial 16S rRNA gene copy number ratios, andbetween the crenarchaeal ammonia oxidation genecopy numbers and the crenarchaeal 16S rRNA genecopy number ratio.

Statistical analysesAll quantitative (response and explanatory) datawere transformed using Box�Cox transformationprior to analyses (the corresponding lamba para-meters were estimated by maximum likelihood(Cook and Weisberg, 1999)). Qualitative explanatoryvariables were transformed into dummy variables.Spatial vectors were constructed by using thePrincipal Coordinate of a Neighbor Matrix (PCNM)approach (Borcard and Legendre, 2002). Thisspatial decomposition method was applied to thegeographic coordinates of the samples (data werespatially detrended if necessary), which yielded 62spatial variables that represented all spatial scalespresent in the sampling scheme. The order of thePCNM variables corresponds to a progression fromlarger to smaller spatial scales (Borcard et al., 2004).For each response data model, the most significantPCNM variables were chosen by permutationalforward model selection and by ensuring that theadjusted R2 of the reduced models did not exceedthe adjusted R2 of the global models. Explanatoryvariables were then selected by multiple regressionanalysis using stepwise selection and by minimizingthe Akaike Information Criterion. Statistical signi-ficances were assessed by 1000 permutation ofthe reduced models. The respective effects of eachexplanatory variable, or combinations thereof, weredetermined by canonical variation partitioning(Borcard et al., 1992; Ramette and Tiedje, 2007b).P values were Bonferroni-corrected to maintain thefamily-wise error level in multiple testing. Allstatistical calculations were performed with the Rstatistical platform using the vegan, PCNM andMASS packages.

Geostatistical interpolationKriging or geostatistical interpolation aims topredict the unknown value of a variable Z at anon-observed location xi using the value zi atsurrounding locations. For this purpose, a stochasticfunction was used as a model of spatial variationso that the actual but unknown value z(xi) and thevalue at the surrounding location were spatiallydependent random variables. A Box–Cox trans-formation was applied to our data (Box and Cox,

Distribution of N-cycling microbial communitiesD Bru et al

534

The ISME Journal

1964) so that z was a realization of a Gaussianrandom function with a covariance matrix V.

Z� ¼ logðzÞ if t ¼ 0zt�1

t otherwise;

�ð1Þ

where t is the parameter of the transformation.The elements of V are expressed as a function

of the distance separating two observations (h).The elements of V are obtained from a parametricfunction C(h), where h is the lag vector separatingtwo observations. In general this parametric func-tion may vary with both the length and directionof h, but here we assumed that the function isisotropic and varies only according to the lengthof h, which we denote by h. It is common in thegeostatistical literature for the spatialcovariance ofa random variable to be expressed in terms of thevariogram

gðhÞ ¼ 0:5EðfðxÞ � Zðx þ hÞg2Þ ð2Þ

The full details for the calculation of V are given inWebster and Oliver (2007). To model the spatialcovariance, we used the Matern function, which hasa smoothness parameter n. When n is small thespatial process is rough, whereas for large n it issmooth. We calculated an effective range, whichdepends both on a and n, by using the distance atwhich the Matern semi-variance equalled 95% ofthe partial sill variance. The parameters ofthe Matern function were obtained by maximumlikelihood estimation (Lark, 2000). The validityof the fitted geostatistical models was confirmedby leave-one-out cross-validation. For each sam-pling site location i¼ 1,y , n, the value of theproperty at site xi is predicted by simple krigingupon z*(�i), the vector of observations excludingz*(xi). The statistic

yi ¼fz�ðxiÞ � ~Z�ð�iÞg

2

s2ð�iÞ

; ð3Þ

where ~Z�ð�iÞ and s2(i) denote the kriging prediction

and kriging variance at xi when z*(xi) is omittedfrom the transformed observation vector. If the fittedmodel is a valid representation of the spatialvariation of the soil property, then y¼ (y1,y ,yn)has a w2 distribution with mean y¼ 1.0 and mediany¼ 0.455 (Lark, 2002). The mean and median valuesof y were also calculated for 1000 simulatedrealizations of the fitted model to determine the90% confidence limits. Moreover, the evaluationof the model was also verified by performing alikelihood ratio test. This test was used to comparethe fits of the spatial and non-spatial models.The spatial analysis GeoR package was used forthe spatial analyses (Ribiero and Diggle, 2001).

Results and Discussion

The largest variations in gene copy numbersacross the Burgundy region were observed for theammonia-oxidizing crenarchaeota (AOA) and totalcrenarchaeota, with densities ranging between lessthan 102 (detection limit) to 9.8� 104 and from4.7� 101 to 5.9� 104 gene copies per ng of DNA,respectively (Figure 1, Supplementary Figure S2). Incomparison, the abundance of ammonia-oxidizingbacteria (AOB) varied over two orders of magnitude,whereas the abundances of nitrate reducers anddenitrifiers mostly varied within one order ofmagnitude. In accordance with the study of Leiningeret al. (2006), which showed a good correlationbetween a membrane lipid biomarker of archaeaand the amoA gene copy numbers, we found that theabundance of the AOA and the total crenarchaeotawere highly correlated (R2¼ 0.72, Po0.001; Supple-mentary Figure S3). In most soils of the Burgundyregion, the AOA were largely predominant overthe AOB with a ratio of archaeal to bacterial amoAcopy numbers ranging from 10 to 400 in 77 out of 107sites, as observed in other studies (Leininger et al.,2006; Nicol et al., 2008; Jia and Conrad, 2009).However, 9 sites without common characteristics hada AOA:AOB ratio ranging between 0.2 and 1.

The relative abundance of coexisting communitiesis of fundamental interest in ecology (Weiher andKeddy, 1999). Therefore, we also calculated theratios of the different bacterial N-cycling genes tobacterial 16S rRNA copy numbers and that of theAOA to total crenarchaeal 16S rRNA copy numbersto examine how the proportions of the differentN-cycling communities within the prokaryoticcommunity vary across landscapes (Supplementary

4

5

6

1

2

3

4

Lo

g g

ene

cop

y n

um

ber

per

ng

DN

A

Bac

teri

a

narG

nap

A

nirK

nirS

nosZ

Cre

na.

AOA

AOB

Figure 1 Variation in the abundance of different microbialcommunities across the Burgundy region. The upper and lowerboundaries of each box indicate the 75th and 25th percentile,respectively, and the mid-line marks the median of the distribu-tion of the corresponding qPCR values. Whiskers above and belowthe box indicate the 90th and 10th percentiles, respectively, whileblack dots indicate outliers.

Distribution of N-cycling microbial communitiesD Bru et al

535

The ISME Journal

Figure S4). Interestingly, the AOA to crenarchaeotaratio varied from 0.08 to 2.7, which suggests that(i) not all crenarchaea have the amoA gene and aretherefore capable of ammonia oxidation and (ii) theproportion of ammonia oxidizers within the crenar-chaea is not constant in terrestrial environments andis influenced by environmental changes. However,this might also be partly explained by a variation inthe number of amoA and 16S rRNA gene copies percell and/or by the specificity of the primers used.Thus, ratios higher than 1 are likely due to thefact that the crenarchaea primers are not trulyuniversal and are underestimating the total numberof crenarchaea. We found that the nitrate reducers,denitrifiers and AOB represented around 5–20%,1–5% and 0.05–1% of the total bacterial community,respectively, as previously reported (Okano et al.,2004; Henry et al., 2006; Philippot et al., 2009b).It is noteworthy that the percentage of bacteriapossessing the nosZ gene, which encodes the N2Oreductase, within the denitrifying community (thatis, those possessing the nirK or nirS genes encodinga nitrite reductase) varied within one order ofmagnitude and was never higher than 56%. Thisis consistent with the work of Jones et al. (2008),which showed that out of approximately 68 com-plete prokaryotic genomes in the database witheither nirS or nirK, only 43 had the nosZ gene. Ourfindings support the mounting evidence that asignificant fraction of the denitrifying communitymight lack the genetic ability to perform the last stepof the reduction pathway, that is, reduction of the

potent greenhouse gas N2O into harmless N2 (Henryet al., 2006; Richardson et al., 2009; Philippot et al.,2009b).

Using a dataset describing 49 different soil andenvironmental variables at each sampling site(Supplementary Table S1), we found that between43 and 85% of the biological variance in thedistribution of the studied communities could beexplained (Table 1 and Supplementary Table S2).The amounts of explained variation belonged tothe upper range of what has been evidenced in otherstudies in microbial ecology (Yergeau et al., onlinefirst; Ramette and Tiedje, 2007b), or in classicalcommunity ecology (Cottenie, 2005) using compa-rable statistical approaches. To better understandthe mechanisms driving the spatial distributionsobserved in this study, all variables were groupedinto five categories (spatial effects, land use, climate,soil physics and soil chemistry), and partial regres-sion models were calculated for each dataset(Table 1 and Supplementary Table S2). In almostall cases, soil chemistry was the strongest predictorand explained between 20 and 68% of the totalvariance (Table 1). When separating the effect ofeach variable, pH emerged as either an important orthe strongest single soil chemistry predictor for mostcommunities (Table 2). Thus, differences in soil pHalone could explain up to 17.8% of the variabilityin abundance of the total bacterial community,between 15.6 and 21.4% for the denitrifier commu-nity, 8.5% for the AOA and 2.9–7.1% for the nitrate-reducing community. The importance of soil pH has

Table 1 Partitioning of the biological variation of different microbial communities as a function of contextual parameters

Overall model a Respective contribution of contextual variables (% explained variance) b

N F-ratio Total explainedvariance (%)

Space Landuse

Climate Time Soilphysics

Soilchemistry

Total bacteria 16 14.81*** 73.1 7.70*** 19.7*** 6.5*** 20.8***Total crenarchaea 16 27.01*** 85.1 1.3** 1.6* 0.3 NS 1.4* 25.2***

Nitrate reducersnarG 25 17.33*** 55.0 14.6*** 1.4* 0.6 NS 2.3** 39.1***napA 16 10.89*** 66.7 6.0** 6.6** 3.9** 49.5***

DenitrifiersnirK 12 18.64*** 71.1 2.8** 59.3***nirS 21 16.99*** 83.0 2.2** 0.5 NS 1.6** 2.1** 1.2* 35.5***nosZ 10 17.08*** 64.7 3.8** 5.2** 2.3* 41.3***

NitrifiersAOB 8 22.59*** 70.8 18.6*** 1.3* 16.9***AOA 15 25.65*** 83.5 1.1 NS 0.9* 26.9***

aFor each quantitative response variable, redundancy analysis models were assessed for significance after selecting the most parsimoniousexplanatory variables by minimizing the Akaike Information Criterion. N, number of retained explanatory variables (details about the individualvariables can be found in Table 2). Total explained variance consists of the sum of the respective contributions of each contextual category and oftheir overall covariation.bThe biological variation in each model was partitioned into the respective effects of different categories of explanatory variables by using partialredundancy analyses and by adjusting R2 values (that is, the percentage of explained variation) to obtain unbiased estimates (Peres-Neto et al.,2006).Significance was determined by 1000 permutations with ***Po0.001; **Po0.01; *Po0.05; and NS (not significant), PX0.05.

Distribution of N-cycling microbial communitiesD Bru et al

536

The ISME Journal

been widely documented for both the total bacterialcommunity and the microbial communitiesinvolved in N-cycling (Fierer and Jackson, 2006;Philippot et al., 2007; Hartman et al., 2008; Nicolet al., 2008; Hallin et al., 2009). Despite the fact thatsoil pH is now recognized as a driver of changes inAOA and AOB communities (Prosser and Nicol,2008; Erguder et al., 2009), the way in which pHaffects AOA communities is still debated andcontroversial. Thus, decreasing of AOA abundancehas been reported both with decreasing soil pH(He et al., 2007; Hallin et al., 2009; Jia and Conrad,2009) and with increasing soil pH (Nicol et al.,2008). In our study, which included 107 soils withpH ranging from 4.2 to 8.3, we found that the AOAwere below the detection limit only in acidic soilsand that soil pH was positively correlated with AOAabundance (R2¼ 0.424, Po0.001). Soil pH was alsothe best predictor of the AOB/AOA ratio with 12%of the variance explained (Po0.001). Although soilpH has been shown to influence the abundance ofAOB (He et al., 2007; Hallin et al., 2009), it was nota significant factor across the large range of soilsexamined here. This suggests that pH may beimportant only in regulating AOB on smaller scalesor across specific fertilization regimes (Fierer et al.,2009). Altogether, our results indicate that nichepartitioning between AOB and AOA is largelyattributable to soil pH, with AOA being moreaffected by acidic pH than AOB. Interestingly, we

also found that the 24% spatial variability of theAOA/crenarchaea ratio could also be explained bychanges in pH, which suggests a stronger effect ofsoil pH on the crenarchaeal ammonia oxidizers thanon the rest of the crenarchaeal community. Evidenceof AOA and AOB specific niches in terrestrialenvironments is strengthened by the findings thatnone of the 42 measured soil properties at thesampling sites could explain the variation in boththe abundances of AOA and AOB (Table 2). Amongthe soil chemical properties other than pH thatexplained the variance in distribution of the differ-ent N-cycling communities, exchangeable manga-nese availability was a significant predictor of theabundance of the nitrate reducers (4.2–7.4%) and ofthe denitrifiers (5.0–8.5%) (Table 2). In contrast tosoil chemistry, the relative contribution of soilphysics was never higher than 6.5% of the varia-bility, although significant for several functionalcommunities (Table 1). Altogether, these findingsdemonstrate that a very significant proportion of thevariation in the distribution of microbial guilds canbe predicted across terrestrial ecosystems at thelandscape scale.

Three dominant types of ecosystems were distin-guished across the 107 Burgundy sites, with forests(with 21 out of 26 being deciduous forests), grass-lands and agricultural soils dominating. We foundthat changes in land use did not strongly influencethe abundance of any of the studied communities

Table 2 Contribution of the first five most important explanatory variables to the variation in microbial community abundances

% Variance explained by:

Total bacteria pH (17.8) TpM (13.1) Sp. Dist.y (7.1) Rain (6.0) Crext (6.0)Total crenarchaea pH (9.6) Ktot (5.6) Mgtot (4.5) Res. water (3.8) Cdext (3.3)

Nitrate reducersNarG Carbon (7.6) pH (7.1) Sp. Dist.V11 (5.6) Sp. Dist.V7 (4.5) Mnexch (4.2)napA Mnexch (7.4) Time (6.6) Pb (6.2) Sp. Dist.x (6.0) Cuext (5.2)

DenitrifiersnirK pH (21.4) Cutot (7.3) Cr (6.0) Fetot (5.4) B (5.1)nirS pH (15.6) Mnexch (6.4) Mntot (4.8) Catot (4.5) Natot (2.8)nosZ pH (15.9) Mnexch (8.5) Naexch (5.7) Ktot (5.5) TpM (5.2)

NitrifiersAOB Land use (18.6) Carbon (8.4) Ni (6.6) Naexch (5.3) Kexch (3.3)AOA pH (8.5) Ktot (5.5) Mgtot (3.6) Pbtot (3.2) Pass (2.7)

Relative abundancesnarG/16S Carbon (23.0) Sp. Dist.y (15.1) Natot (8.9) Mntot (6.1) Rainyr (5.5)napA/16S Caexch (32.0) Res. water (5.4) ETPM (4.3) — —nirK/16S Caexch (16.8) Res. water (10.2) TpM (9.8) — —nirS/16S Natot (9.8) Pass (8.0) Rainyr (6.8) Mntot (5.4) Ktot (5.3)nosZ/16S Sp. Dist.y (9.6) — — — —AOB/16S Land (18.0) Niext (9.3) Catot (6.3) Nitrogen (5.7) Naexch (5.7)AOA/16S pH (23.9) Res. water (12.6) Sand (8.3) Fetot (7.2) Mnexch (5.3)AOA/AOB pH (12.0) Ktot (9.9) Nitrogen (7.8) Kexch (6.6) Mntot (4.1)

The respective contributions were calculated by taking into account all other significant variables in the model using partial redundancy analysesand adjusting the R2 values (see Table 1). Only significant fractions based on 1000 data permutations and subsequent multiple-comparisonBonferroni corrections are reported. Subscripts are: M (monthly), Sp. Dist. (spatial distance with the corresponding spatial variables indicated assubscripts), Res. water (residual water), ass (assimilable), tot (total), ext (extractable), exch (exchangeable) and yr (year).

Distribution of N-cycling microbial communitiesD Bru et al

537

The ISME Journal

other than the AOB, for which 18.6% of the vari-ation could be explained by the pure effects of thatfactor alone (Table 1). Changes in land use alsoaffected the proportion of AOB within the totalbacterial community, further suggesting an inherentsensitivity of this community to land management(Table 2). This coupling between land use manage-ments and abundance of AOB indicates that AOBabundance could be used as a pertinent biologicalindicator for monitoring soils. Likewise, AOBdiversity is affected by land use (Carney et al.,2004) and was recently selected as a top candidatebiological indicator of soil quality for national-scalesoil monitoring (Ritz et al., 2004).

To examine the relative contribution of climaticvariables to the landscape distributions of thefunctional microbial communities, we used regionalpatterns of precipitation, net evapotranspiration andtemperature data (Supplementary Table S1). Despiteimportant variations, local climate mostly had asignificant influence on the distribution pattern ofthe total bacterial community, with temperature,precipitation and evapotranspiration significantlyexplaining 13.1, 6.0 and 5.6% of the total variance,respectively (Table 2). We also found that localclimate explained around 10–13% of the variabilityin the relative abundance of nirS and nirK deni-trifiers within the total bacterial community (Sup-plementary Table S2). When considered separatelyfrom the other factors, the sampling period (time)had a weak influence and affected significantlyonly the napA and nirS communities (% varianceexplained of 6.6 and 2.1, respectively), thus indicat-ing that the large time period needed to sampleall 107 sites did not strongly affect our resultsby masking the effects of other environmentalvariables. Likewise, a recent study reported thattemporal variation also had little impact on thedistribution of the microbial community composi-tion, despite being sampled in different seasons anddifferent years (Drenovsky et al., 2010). Altogether,the results show that neither local climate nor thesampling time was a major factor influencing thedistribution patterns of the studied N-cycling com-munities over the 31 500 km2 Burgundy region.

Geographical distance, when separated fromthe other environmental variables, was a weakbut significant predictor of the total bacterialand crenarchaeal communities and of the nitrate-reducing and -denitrifying communities (Table 1).However, the explanatory power of the spatialdistance dramatically increased when spatial auto-correlation was explicitly modelled without dissect-ing the environmental variables and incorporatingcovariation. Thus, by investigating the spatialcorrelation of microbial abundance using a geosta-tistical approach, we found strong spatial patternsin the distribution of some communities, withautocorrelation ranging between 22 and 739 km(practical ranges in supplementary Table S3). Threemajor types of spatial distributions were found.

The predicted map of the distribution of totalbacteria was quite smooth with a high density inthe north and a low density in the south, while thecrenarchaea exhibited a more patchy distribution(Figure 2). Finally, the maps of nirK and nirS are

a

1.8 105

0 20 40

6.7 104 Kilometers

b

0 20 405.3 104

3.0 102 Kilometers

Figure 2 Maps of the abundances of total bacteria andcrenarchaea in Burgundy. (a) Bacterial 16S rRNA, (b) crenarchaeal16S rRNA. The color scale to the left of each map indicates theextrapolated abundance values (gene copy number per ng ofDNA).

Distribution of N-cycling microbial communitiesD Bru et al

538

The ISME Journal

somewhat ‘spotty’. These three types of maps aredirectly related to the roughness of the spatialdistributions, which were modelled thanks to theflexibility of Matern variogram. Interestingly, thesignificant latitude effect observed both for thedistribution of the total bacterial community andalso for the relative abundance of the narG and nosZgenes (Table 2, Supplementary Figure S5) wasrelated to the distribution of soil parental materialwith limestone plateau in northern Burgundy andcrystalline rocks in southern Burgundy. Few differ-ences were observed between the distributions ofthe nirS and nosZ denitrifiers, while the distributionof the nirK denitrifiers was more related to thatof the total bacteria (Figure 3). In contrast, the

predicted map of AOB distribution differed con-siderably from that of all the other studied commu-nities (Figure 3). Those maps supported the resultsof the canonical variation partitioning analyses,indicating that the AOB was the only communityfor which soil chemistry was not the main determi-nant of the spatial distribution (Table 1). Althoughwe know of no other directly comparable studies, afew articles have reported spatial dependence of thedistribution of microbial abundance at much lowerscales ranging from centimetres to tens of meters(Franklin et al., 2002; Ritz et al., 2004; Philippotet al., 2009b; Enwall et al., 2010). At larger scales,studies using a spatially explicit approach havefocused on microbial diversity rather than on

0 20 40 0 20 402.3 104

6.1 103 Kilometers

0 20 403.8 103

1.3 103 Kilometers

a b c

d e

1.1 104

7.5 102 Kilometers

0 20 407.1 103

1.1 102 Kilometers

0 20 406.5 102

94 Kilometers

Figure 3 Maps of the abundances of N-cycling genes in Burgundy. (a) narG, (b) nirK, (c) nirS, (d) nosZ, (e) AOB. The color scale to theleft of each map indicates the extrapolated abundance values (gene copy number per ng of DNA).

Distribution of N-cycling microbial communitiesD Bru et al

539

The ISME Journal

microbial abundance. Thus, spatial dependence ofmicrobial diversity at a kilometer scale was observedby Dequiedt et al. (2009) and Cho and Tiedje (2000),while Fierer and Jackson (2006) found that microbialdiversity was not related to geographic distanceacross North and South America. Although spatialvariability in the distributions of soil microorgan-isms is generally regarded as random noise, ourresults revealed that this variability can be ex-plained even at the landscape scale.

In conclusion, the present study provides anoverview of the factors driving the spatial distribu-tion of microbial communities involved in N-cyclingand of the total bacterial and crenarchaeal commu-nities across a 31 500 km2 terrestrial landscape. Ourspatially explicit approach showed that no singlebiogeographical distribution was shared by all thestudied microbial communities. However, somecommon features emerged and soil chemistry—withpH as an overarching controlling factor—was themost important predictor of the distribution of themicrobial communities in many cases. Thus,although many environmental variables weresignificant predictors, only a few accounted for alarge amount of the total variance in the distributionof the studied microbial guilds and we couldexplain between 43 and 85% of this spatial variationin community abundances. Furthermore, our find-ings illustrate the potential of geostatistic methods,which were successfully used to produce the firstmaps of the distribution of microbial guilds at ascale of relevance to policy makers and stakeholdersfor ecosystem management.

Acknowledgements

Sampling and soil analyses were supported by a FrenchScientific Group of Interest on Soils: the ‘GIS Sol’,involving the French Ministry for Ecology and SustainableDevelopment, the French Ministry of Agriculture, theFrench Agency for Energy and Environment (ADEME),the National Institute for Agronomic Research (INRA), theInstitute for Research and Development (IRD) andthe National Forest Inventory (IFN). We are grateful toall the soil surveyors and technical assistants involved insampling the sites: Chambers of Agriculture of Cote-d’Or,Nievre, Saone-et-Loire, Yonne and National Institute forAgronomy, Food and Environment (AgroSup Dijon). Thisproject was funded by the Burgundy region and theFrench National Research Agency (ANR).

References

Borcard D, Legendre P. (2002). All-scale spatial analysis ofecological data by means of principal coordinates ofneighbour matrices. Ecol Model 153: 51–68.

Borcard D, Legendre P, Avois-Jacquet C. (2004). Dissectingthe spatial structure of ecological data at multiplescales. Ecology 85: 1826–1832.

Borcard D, Legendre P, Drapeau P. (1992). Partialling outthe spatial component of ecological variation. Ecology73: 1045–1055.

Box GEP, Cox DR. (1964). An analysis of transformation.J Royal Stat Soc 26: 211–243.

Bru D, Sarr A, Philippot L. (2007). Relative abundancesof proteobacterial membrane-bound and periplasmicnitrate reductases in selected environments. ApplEnviron Microbiol 73: 5971–5974.

Carney KC, Matson PA, Bohannan BJM. (2004). Diversityand composition of tropical soil nitrifiers acrossa plant diversity gradient and among land-use types.Ecol Lett 7: 684–694.

Cho J-C, Tiedje JM. (2000). Biogeography and degree ofendicity of fluorescent Pseudomonas strains in soil.Appl Environ Microbiol 66: 5448–5456.

Cook RD, Weisberg S. (1999). Wiley series in probability andstatistics. In: Statistics, W.S.i.P.a. (ed) Applied Regression,Including Computing and Graphics. Wiley-Interscience.

Cottenie K. (2005). Integrating environmental and spatialprocesses in ecology community dynamics. Ecol Lett8: 1175–1182.

Dequiedt S, Thioulouse J, Jolivet C, Saby N, Lelievre M,Maron P-A. et al. (2009). Biogeographical patterns ofsoil bacterial communities. Environ Microbiol Rep 1:251–255.

Drenovsky RE, Steenwerth KL, Jackson LE, Scow KM.(2010). Land use and climatic factors structureregional patterns in soil microbial communities.Global Ecol Biogeogr 19: 27–39.

Enwall K, Throback N, Stenberg M, Soderstrom M, Hallin S.(2010). Soil resources influence spatial patterns ofdenitrifying communities at scale compatible with landuse management. Appl Environ Microbiol 76: 2243–2250.

Erguder TH, Boon N, Wittebolle L, Marzorati M,Verstraete W. (2009). Environmental factors shapingthe ecological niches of ammonia-oxidizing archaea.FEMS Microbiol Rev 33: 855–869.

Fierer N, Carney KM, Horner-Devine MC, Megonigal JP.(2009). The biogeography of ammonia-oxidizingbacterial communities in soil. Ecol 58: 435–445.

Fierer N, Jackson R. (2006). The diversity and biogeogra-phy of soil bacterial communities. Proc Natl Acad SciUSA 10: 626–631.

Forster P, Ramaswamy V, Artaxo P, Berntsen T,Betts R, Fahey DW et al. (2007) In: Solomon S, et al.(eds). Climate Change 2007: Changes in atmosphericconstituents and in radiative forcing. The PhysicalScience Basis. Contribution of Working Group Ito the Fourth Assessment Report of the Inter-governmental Panel on Climate Change. CambridgeUniversity Press: Cambridge, UK,and New York, NY,USA.

Franklin R, Mills A. (2003). Multi-scale variation in spatialheterogeneity for microbial community structure in aneastern Virginia agricultural field. FEMS MicrobiolEcol 44: 335–346.

Franklin RM, Blum LK, McComb AC, Mills AL. (2002).A geostatistical analysis of small-scale spatial varia-bility in bacterial abundance and community structurein salt marsh creek bank sediments. FEMS MicrobiolEcol 42: 71–80.

Green JL, Bohannan BJM, Whitaker RJ. (2008). Microbialbiogeography: from taxonomy to traits. Science 320:1039–1043.

Hallin S, Jones CM, Schloter M, Philippot L. (2009).Relationship between N-cycling communities and

Distribution of N-cycling microbial communitiesD Bru et al

540

The ISME Journal

ecosystem functioning in a 50-year-old fertilizationexperiment. ISME J 3: 597–605.

Hartman WH, Richardson CJ, Vilgalys R, Bruland GL.(2008). Environmental and anthropogenic controlsover bacterial communities in wetland soils. Proc NatlAcad Sci USA 105: 17842–17847.

He JZ, Shen JP, Zhang LM, Zhu YG, Zheng YM, Xu MGet al. (2007). Quantitative analyses of the abundanceand composition of ammonia-oxidizing bacteria andammonia-oxidizing archaea of a Chinese upland redsoil under long-term fertilization practices. EnvironMicrobiol 9: 2364–2374.

Henry S, Baudoin E, Lopez-Gutierrez JC, Martin-Laurent F,Brauman A, Philippot L. (2004). Quantification ofdenitrifying bacteria in soils by nirK gene targetedreal-time PCR. J Microbiol Methods 59: 327–335.Corrigendum (2005) 61:289–290.

Henry S, Bru D, Stress B, Hallet S, Philippot L. (2006).Quantitative detection of the nosZ gene, encodingnitrous oxide reductase, and comparison of the abun-dances of 16S rRNA, narG, nirK, and nosZ genes insoils. Appl Environ Microbiol 72: 5181–5189.

Heymann Y, Steenmas C, Croissille G, Bossard M.(1994). Corine Land Cover-Technical Guide. Officefor Official Publications of the European Commu-nities: Luxemberg, p 144, ISBN 92-826-2578-8.

Hughes-Martiny JB, Bohannan BJM, Brown JH, Colwell RK,Fuhrman JA, Green JL et al. (2006). Microbial bio-geography: putting microorganisms on the map. RevMicrobiol 4: 102–112.

Hutchinson GE. (1953). The concept of pattern in ecology.Proc Acad Nat Sci Phila 105: 1–12.

Jia Z, Conrad R. (2009). Bacteria rather than Archaeadominate microbial ammonia oxidation in an agricul-tural soil. Environ Microbiol 11: 1658–1671.

Jones CM, Stres B, Rosenquist M, Hallin S. (2008).Phylogenetic analysis of nitrite, nitric oxide, andnitrous oxide respiratory enzymes reveal a complexevolutionary history for denitrification. Mol Biol Evol5: 1955–1966.

Klappenbach JA, Saxman PR, Cole JR, Scmidt TM. (2001).rrndb: the ribosomal RNA operon copy numberdatabase. Acid Res 29: 181–184.

Kowalchuk GA, Stephen JR. (2001). Ammonia-oxidizingbacteria: a model for molecular microbial ecology.Annu Rev Microbiol 55: 485–529.

Lark RM. (2000). A comparison of some robust estimatorsof the variogram for use in soil survey. Eur J Soil Sci51: 137–157.

Lacarce E, Le Bas C, Cousin JL, Pesty B, Toutain B,Houston Durrant T et al. (2009). Data management formonitoring forest soils in Europe for the Biosoilproject. Soil Use Manag 25: 57–65.

Lark RM. (2002). Modelling complex soil properties ascontaminated regionalized variables. Geoderma 106:173–190.

Leininger S, Urich T, Schloter M, Schwark L, Qi J,Nicol GW et al. (2006). Archaea predominate amongammonia-oxidizing prokaryotes in soils. Nature 442:806–809.

Levin SA. (1992). The problem of pattern and scale inecology. Ecology 73: 1943–1967.

Martin-Laurent F, Philippot L, Hallet S, Chaussod R,Germon JC, Soulas G et al. (2001). DNA extrac-tion from soils: old bias for new microbial diversityanalysis methods. Appl Environ Microbiol 67:2354–2359.

Morvan X, Saby NPA, Arrouays D, Le Bas C, Jones RJA,Verheijen FGA et al. (2008). Soil monitoring inEurope: a review of existing systems and requirementsfor harmonisation. Sci Tot Environ 391: 1–12.

Nicol G, Leininger S, Schleper C, Prosser JI. (2008). Theinfluence of soil pH on the diversity, abundance andtranscriptional activity of ammonia oxidizing archaeaand bacteria. Environ Microbiol 10: 2966–2978.

Nunan N, Wu K, Young IM, Crawford JW, Ritz K. (2002).In situ spatial patterns of soil bacterial populations,mapped at multiple scales, in an arable soil. MicrobiolEcol 44: 296–305.

Ochsenrelter T, Selezi D, Qualser A, Bonch-OsmolovskayaL, Schleper C. (2003). Diversity and abundance ofCrenarchaeota in terrestrial habitats studied by 16SRNA surveys and real time PCR. Environ Microbiol 5:787–797.

Okano Y, Hristova KR, Leutenegger CM, Jackson LE,Denison RF, Gebreyeys B et al. (2004). Applicationof real-time PCR to study effects of ammonium onpopulation size of ammonia-oxidizing bacteria in soil.Appl Environ Microbiol 70: 1008–1016.

Peres-Neto P, Legendre P, Dray S, Borcard D. (2006).Variation partitioning of species data matrices: estima-tion and comparison of fractions. Ecol 87: 2614–2625

Philippot L, Bru D, Saby NPA, Cuhel J, Arrouays D,Simek M et al. (2009a). Spatial patterns of bacterialtaxa in nature reflect ecological traits of deep brancesof the 16S rRNA bacterial tree. Environ Microbiol 11:1518–1526.

Philippot L, Cuhel J, Saby NPA, Cheneby D, ChronakovaA, Bru D et al. (2009b). Mapping field-scale spatialdistribution patterns of size and activity of thedenitrifier community. Environ Microbiol 11: 1518–1526.

Philippot L, Hallin S. (2005). Finding the missing linkbetween diversity and activity using denitrifyingbacteria as a model functional community. Curr OpinMicrobiol 8: 234–239.

Philippot L, Hallin S, Schloter M. (2007). Ecology ofdenitrifying prokaryotes in agricultural soil. AdvAgronomy 96: 135–190.

Prosser JI, Nicol GW. (2008). Relative contributions ofarchaea and bacteria to aerobic ammonia oxidation inthe environment. Environ Microbiol 10: 2931–2941.

Quintana-Segui P, Moigne PL, Durand Y, Martin E, HabetsF, Baillon M et al. (2008). Analysis of near-surfaceatmospheric variables: validation of the SAFRANanalysis over France. J Appl Meteo Climat 47: 92–107.

Ramette A, Tiedje JM. (2007a). Biogeography: an emergingcornerstone for understanding prokaryotic diversity,ecology and evolution. Microbiol Ecol 53: 197–207.

Ramette A, Tiedje JM. (2007b). Multiscale responses ofmicrobial life to spatial distance and environmentalheterogeneity in a patchy ecosystem. Proc Natl AcadSci USA 104: 2761–2766.

Ravishankara AR, Daniel JS, Portmann RW. (2009).Nitrous oxide (N2O): the dominant ozone-depletingsubstance emitted in the 21st Century. Science 326:123–125.

Ribiero PJ, Diggle PJ. (2001). geoR: a package forgeostatistical analysis. R news 1: 8–15.

Richardson D, Felgate H, Watmough N, Thomson A,Baggs E. (2009). Mitigating release of the potentgreenhouse gas N2O from the nitrogen cycle—couldenzymic regulation hold the key? Trends Biotech 27:388–397.

Distribution of N-cycling microbial communitiesD Bru et al

541

The ISME Journal

Ritz K, McNicol JW, Nunan N, Grayston S,Millard P, Atkinson D et al. (2004). Spatialstructure in soil chemical and microbiologicalproperties in an upland grassland. FEMS MicrobiolEcol 49: 191–205.

Tourna M, Freitag TE, Nicol GW, Prosser JI. (2008).Growth, activity and temperature responses of ammo-nia-oxidizing archaea and bacteria in soil microcosms.Environ Microbiol 10: 1357–1364.

Webster R, Oliver M (eds) Geostatistic for EnvironmentalScientists. John Wiley & Sons Ltd.: Chichester, UK,2007.

Weiher E, Keddy PA. (1995). Assembly rules, null models,and traits dispersion: new questions from old patterns.Oikos 74: 159–164.

Weiher E, Keddy PA. (1999). Relative abundance andevenness patterns along diversity and biomass gradi-ents. Oikos 87: 355–361.

Yergeau E, Bezemer TM, Hedlund K, Mortimer SR,Kowalchuk GA, van der Putten WH. (2009). Influenceof space, soil, nematode and plants on microbialcommunity composition of chalk grassland soils.Environ Microbiol (e-pub ahead of print 16 September2009) doi: 10.1111/j.1462-2920.2009.02053.x.

Supplementary Information accompanies the paper on The ISME Journal website (http://www.nature.com/ismej)

Distribution of N-cycling microbial communitiesD Bru et al

542

The ISME Journal