Biodiversidad Microbiana en Corrientes Que Alimntan Glaciares Wilhem Et Al 2013

7/27/2019 Biodiversidad Microbiana en Corrientes Que Alimntan Glaciares Wilhem Et Al 2013

1/10

OPEN

ORIGINAL ARTICLE

Microbial biodiversity in glacier-fed streams

Linda Wilhelm1, Gabriel A Singer1,2, Christina Fasching1, Tom J Battin1,2 and

Katharina Besemer1,21Department of Limnology, University of Vienna, Vienna, Austria and 2Wasser Cluster Lunz,Lunz am See, Austria

While glaciers become increasingly recognised as a habitat for diverse and active microbialcommunities, effects of their climate change-induced retreat on the microbial ecology of glacier-fedstreams remain elusive. Understanding the effect of climate change on microorganisms in theseecosystems is crucial given that microbial biofilms control numerous stream ecosystem processeswith potential implications for downstream biodiversity and biogeochemistry. Here, using a space-for-time substitution approach across 26 Alpine glaciers, we show how microbial communitycomposition and diversity, based on 454-pyrosequencing of the 16S rRNA gene, in biofilms ofglacier-fed streams may change as glaciers recede. Variations in streamwater geochemistrycorrelated with biofilm community composition, even at the phylum level. The most dominant phyladetected in glacial habitats were Proteobacteria, Bacteroidetes, Actinobacteriaand Cyanobacteria/chloroplasts. Microorganisms from ice had the lowest a diversity and contributed marginally tobiofilm and streamwater community composition. Rather, streamwater apparently collectedmicroorganisms from various glacial and non-glacial sources forming the upstream metacommu-nity, thereby achieving the highest a diversity. Biofilms in the glacier-fed streams had intermediate adiversity and species sorting by local environmental conditions likely shaped their communitycomposition. a diversity of streamwater and biofilm communities decreased with elevation, possiblyreflecting less diverse sources of microorganisms upstream in the catchment. In contrast, bdiversity of biofilms decreased with increasing streamwater temperature, suggesting that glacierretreat may contribute to the homogenisation of microbial communities among glacier-fed streams.The ISME Journal (2013) 7, 16511660; doi:10.1038/ismej.2013.44; published online 14 March 2013Subject Category: Microbial ecosystem impactsKeywords: Biofilm; glacier-fed streams; microbial diversity; climate change; metacommunity ecology

Introduction

Glacial ecosystems, at the interface betweenthe cryosphere, hydrosphere, pedosphere and thebiosphere, are particularly prone to impacts ofclimate change (Brown et al., 2007a; Milner et al.,2009; Finn et al., 2010; Woodward et al., 2010;Jacobsen et al., 2012). Glacier-fed streams constitutea prominent geomorphological and ecological com-ponent of the glacier foreland, and integrateupstream catchment processes. Changes in runofffollowing glacier retreat shift the relative contribu-tions of icemelt, snowmelt and groundwater to

stream discharge (Brown et al., 2007a; Milneret al., 2009), thereby influencing streamwater tem-perature, electrical conductivity and channel stabi-lity in glacier-fed streams (Hannah et al., 2007;Brown et al., 2007a; Milner et al. 2009). The closehydroecological coupling between glaciers andglacier-fed streams makes biodiversity in these

streams particularly susceptible to glacier retreat(Brown et al., 2007a; Finn et al., 2010; Jacobsenet al., 2012). For instance, reduced glacial runoffaccompanying glacier retreat facilitates the displa-cement of cold-adapted invertebrate species byrange expansion of less cryophilic species towardshigher elevations (Brown et al., 2007a; Milner et al.,2009; Finn et al., 2010). Ultimately, such shifts inspecies distribution may threaten the biodiversity inisolated high-elevation ecosystems with yetunknown implications for downstream biodiversity(Hughes et al., 2009; Finn et al., 2011).

Ice ecosystems harbour complex microbial com-

munities (Simon et al., 2009; Anesio and Laybourn-Parry, 2012), which have been proposed as sentinelsof climate change and the general attrition of thecryosphere (Vincent, 2010). The most abundantbacterial phyla detected in glacier ice are Proteo-bacteria, Actinobacteria, Bacteroidetes, Firmicutesas well as photosynthetic Cyanobacteria (Simonet al., 2009; Xiang et al., 2009; Anesio and Laybourn-Parry, 2012; Edwards et al., 2011). While themicrobial diversity and activity in glacier ice wasthe focus of several recent studies (Hodson et al.,2008; Simon et al., 2009; Xiang et al., 2009), theimplications of glacier retreat on the microbial

Correspondence: K Besemer, Wasser Cluster Lunz, Dr CarlKupelwieser Promenade 5, Lunz am See A-3293, Austria.E-mail: [email protected] 25 October 2012; revised 11 February 2013; accepted 12February 2013; published online 14 March 2013

The ISME Journal (2013) 7, 16511660

& 2013 International Society for Microbial Ecology All rights reserved 1751-7362/13

www.nature.com/ismej
http://dx.doi.org/10.1038/ismej.2013.44mailto:[email protected]://www.nature.com/ismejhttp://www.nature.com/ismejmailto:[email protected]://dx.doi.org/10.1038/ismej.2013.44


2/10

communities in glacier-fed streamsmost of themcontained in benthic biofilmsremain elusive.Filling this knowledge gap is important, given thatbiofilms, even in glacier-fed streams (Battin et al.,2004), orchestrate numerous ecosystem processesand contribute to large-scale carbon fluxes (Battinet al., 2008). Biofilms are remarkably diverse, partly

because they assemble from different sources ofbiodiversity (for example, soil and groundwater)within the catchment (Hullar et al., 2006; Besemeret al., 2012). Species sorting from these sources canoccur through local environmental conditionsand is increasingly understood as a key mechanismunderlying biofilm community assembly (Besemeret al., 2012).

As glaciers recede, they change the hierarchicalhabitat template of the riverine landscape whereenvironmental processes operating at local andregional scales differentially affect life in theglacier-fed streams (Ward et al., 2002; Brown et al.,2007a; Milner et al., 2009). Melting glaciers mobiliseice-locked organic matter with implications fordownstream carbon cycling (Singer et al., 2012)and heterotrophic activity (Milner et al., 2009). Toexplore how biofilm biodiversity and communitycomposition in glacier-fed streams might respond toenvironmental changes as induced by glacier retreat,we surveyed 26 glaciers and their streams in theAustrian Alps. Substituting space for time, weinvestigated the relationship between microbialcommunity composition and environmental vari-ables apparently linked to glacial retreat. Alludingto studies on invertebrate communities (Brownet al., 2007a; Milner et al., 2009; Finn et al., 2010),

we hypothesise an increase ofa diversitythe localrichness in a given glacier-fed streamwith increas-ing deviation of the stream environment from atypical glacier-fed stream. Along the same line, wealso postulate a decrease of b diversitythe varia-tion of communities among glacier-fed streams.

Material and methods

Study sites and sample collectionThe study sites covered 26 glaciers located along themain chain of the Austrian Alps (Silvretta, OtztalerAlps, Stubaier Alps, Zillertaler Alps, Venediger

Group, Granatspitz Group, Glockner Group,Hochalmspitze and Ankogel; geographically rangingfrom 10.16213.2781E, 46.77847.1351N and21002880m above sea level; SupplementaryFigure 1, Supplementary Table 1). Sampling wasperformed within 45 days in July and August 2010, aperiod with almost completely snow-free glaciers. Wesampled glacier-fed streams within 10 m downstreamof the glacier terminus to capture the physicochem-ical signatures from glacial and non-glacial sourcesupstream in the catchment (Ward et al., 2002; Brownet al., 2007a). Benthic biofilms were obtained byrigorously shaking B80 stones (diameter B2cm)

successively in sterile 50 ml tubes. This field proce-dure removed most of the biofilm from the stones.The supernatant was then filtered (0.2mm filters,GSWP Millipore, Billerica, MA, USA) using sterileequipment. Two to three litres of streamwater werefiltered through sterile 0.2-mm filters (GSWP) and allfilters were immediately frozen in liquid nitrogen in

the field. Streamwater pH, electrical conductivityand water temperature were measured in situ usingWTW probes (pH320, Cond340i).

As a putatively prominent source of microbialdiversity to the stream, we also sampled subsurfaceice (0.30.5 m depth) from 14 randomly chosen sitesacross the ablation zone of each of the 26 glaciers.We first removed the upper ice layer (B2030 cm)using sterile (ethanol-flamed) ice picks and thencollected ice from beneath into sterile Whirl-Paks(Nasco, Salida, CA, USA). These samples weretransported frozen to the next base camp. There,ice was melted within 1 hour and B10 l of icemeltwere filtered through sterile 0.2-mm filters (GSWP),which were immediately frozen in liquid nitrogen.

Streamwater dissolved organic carbon and nutrientsSreamwater was analysed for dissolved organiccarbon (DOC) concentration using a Sievers 900TOC Analyser (GE Analytical Instruments, Boulder,CO, USA) operated with an inorganic carbon removalunit. Prior to injection, DOC samples (GFF-filtered,pre-combusted) were automatically acidified in theanalyser as recommended by the manufacturer. Wedetermined a method detection limit of the Sievers900 according to US EPA guidelines and found it

typically o6mg C l1

. The detection limit was calcu-lated as three (approximate critical z-value for 99.9%confidence)-times the standard deviation of nine labreplicates of low DOC water (average DOC concentra-tion: 29.711.43mg C l1). Concentrations of N-NH4,N-NO2 and N-NO3 in the streamwater were deter-mined using Continuous Flow Analysis (Allianceinstruments, Salzburg, Austria). For further analyses,we used total nitrogen, which represents thesummation of N-NH4, N-NO2 and N-NO3.

DNA extraction and PCR amplificationGenomic DNA was extracted using the PowerSoil

DNA extraction kit (MoBio, Carlsbad, CA, USA)according to the manufacturers recommendations.The V4V5 regions of the 16S rRNA gene wasamplified in 25-ml PCR reactions containing0.5mmoll 1 of each primer (Thermo Scientific,Waltham, MA, USA), 0.2 mmol l1dNTPs (ThermoScientific), 40mg bovine serum albumin (ThermoScientific), 4 mmol l 1 MgCl2 (Thermo Scientific),1 U Taq-DNA Polymerase with the recommendedPCR buffer (Thermo Scientific) and 4ml DNA extract(24 ng DNA). Primers used for amplification of the16S rRNA gene were the universal 515F 50-GTGNCAGCMGCCGCGGTAA-30 and 926R 50-CCGY

Biodiversity in glacier-fed streamsL Wilhelm et al

1652

The ISME Journal


3/10

CAATTYMTTTRAGTTT-30 (Quince et al., 2011),containing the 454 Titanium A and B adaptors,respectively. For each sample, two different sample-specific barcodes contained in the forward primerwere employed to reduce barcode-specific bias(Berry et al., 2011).

Samples were amplified using an initial denatur-

ing step of 2min at 94 1C, followed by 30 cycles of30 s denaturation at 94 1C, 30 s annealing at 56 1C,1 min elongation at 72 1C and a final elongation for10 min at 72 1C. Each PCR reaction included anegative control. PCR products were run on a 1%agarose gel and purified using the Gel Extraction Kit(Qiagen, Hilden, Germany). The purified PCRproducts were quantified by gel using the Gel DocXR System (BioRad, Hercules, CA, USA) andpooled equimolar for pyrosequencing.

454-pyrosequencingAmplicons were sequenced on a GS FLX TitaniumSequencer in Liverpool (Centre for GenomicResearch, University of Liverpool, UK). Raw outputfiles were filtered and de-noised using the softwarepackage AmpliconNoiseV1.0 (Quince et al., 2011).After pre-clustering the sequences with PyroNoise(AmpliconNoiseV1.0), PCR single-base errors werecorrected via SeqNoise (AmpliconNoiseV1.0), asequence-based clustering method that performsalignment of the sequences. Chimeras were finallyidentified and removed with the Perseus algorithmat an intercept ofa 7.5 and a coefficient ofb 0.5(Quince et al., 2011). This filtering and de-noisingsteps reduced the number of 658592 flowgrams to

384 320 reads (read length 400 bp). A completelinkage algorithm on a 97% identity level clusteredthese reads to operational taxonomic units (OTUs)(AmpliconNoiseV1.0) and taxonomic assignmentswere determined by using a naive Bayesian rDNAclassifier (Ribosomal Database Project; Wang et al.,2007) at a 70% confidence threshold.

Data analysisDissimilarity matrices of the community compositionwere calculated by applying the relative abundance-based BrayCurtis dissimilarity index. Based onthese matrices, non-metric multidimensional scaling

analysis was used to visualise differences in com-munity composition, and a non-parametric permuta-tional analysis of variance Anderson (2001) to testdifferences in community composition among habi-tats (using the R-packages vegan and ellipse,Oksanen et al., 2011; Murdoch and Chow, 2007). Aconcomitant test for dispersion homogeneity (themultivariate analogue of variance homogeneity) isprovided by the analysis ofb diversity (see below).

A principal component analysis was performedto illustrate the relationship and importance ofz-transformed environmental variables. Further,after Hellinger-transformation of the microbial

community composition data, we performed aredundancy analysis to determine a potential effectof environmental variables on community composi-tion (Legendre et al., 2011; using the R-packagevegan: Oksanen et al., 2011). The explanatoryvalues of environmental variables were furtherexplored by running a forward selection test

included in the packfor package (Blanchet et al.,2008). Additionally, these variables were tested forcorrelations with the relative abundances of themost abundant bacterial families. These Pearsoncorrelation coefficients were visualised in aheat map.

We used Diversity Estimation software (Quinceet al., 2008), to estimate the true richness of thecommunities by fitting the 454-pyrosequencing-obtained OTU abundances to a Sichel distribution(Sichel, 1974). For comparison, all samples wererarefied to the lowest number of reads obtained froman individual sample (1007 reads; vegan: Oksanenet al., 2011). Using multiple linear regression, wetested for correlations between environmental vari-ables and microbial richness.

To estimate the probability that a biofilm commu-nity represented a random sample of the respectivestreamwater community, we performed a randomresampling procedure on the samples from eachglacier, using functions of the R-packages vegan,ecodist and gdata (Goslee and Urban, 2007; Oksanenet al., 2011; Warnes et al., 2011; Besemer et al.,2012). Each tested sample pair consisted of thebiofilm and the suspended community from oneglacier-fed stream. Individual reads were sampledfrom the suspended community (the probability of

each OTU to be sampled being its relative abun-dance) with replacement until the number of OTUsin this randomly assembled community equalledthe richness of the respective biofilm community.This procedure was repeated to yield 1000 randomassemblages. The probability of the biofilm commu-nity to fall within the distribution of these randomassemblages was calculated as the percentage of thedistances of the random assemblages to theircentroid, which were as high as or higher than thedistance of the biofilm community to the centroid.As a conservative approach, the biofilm communitydata set was reduced to those OTUs, which alsooccurred in the respective streamwater community.

Dispersion is an expression of average communitydissimilarity, equivalent to average BrayCurtisdissimilarity, and serves as a measure of spatialcommunity variation or b diversity (Anderson et al.,2006a). Dispersion was computed as the averagedistance of communities to the respective group(that is, the habitat) centroid in the ordination spaceresulting from a principal coordinate analysis (akametric scaling) of the BrayCurtis dissimilaritymatrix (Anderson, 2006b). Differences in dispersionamong habitats were tested using a permutationprocedure (Anderson, 2006b) with functions ofvegan in R (Oksanen et al., 2011).


1653

The ISME Journal


4/10

The relationship between selected environmentalvariables (for example, streamwater temperature)and b diversity of streamwater and biofilm commu-nities was tested by multiple linear regression ondistance matrices (Lichstein, 2006; Ptacnik et al.,2010). For this, the BrayCurtis dissimilarity matrixwas unfolded to a vector of pairwise dissimilarities

between any two communities and analysed asdependent on pairwise average temperature andpairwise temperature difference in a multiple linearregression model. In this analysis, the hypothesis ofinterest concerns the effect of average temperatureon dissimilarity, but temperature differences mustbe included as a covariate to account for environ-mental differentiation simultaneously affecting bdiversity; significances are computed by permuta-tion (see Supplementary Information for details).Additionally, as a graphical analysis, we adapted alocal polynomial regression fitting procedure(LOESS; Cleveland et al., 1992) to estimate a localb diversity at a given temperature of interest. Basedon a neighbourhood-approach, we identified subsetsof sites with similar temperatures, whose commu-nities were then included in the computation ofb diversity as an average dissimilarity in a weightedmanner (see Supplementary Information for details).In addition to temperature, these analyses wereperformed on elevation, electrical conductivity andpH, which were previously found to be significantlycorrelated to streamwater and biofilm communitycompositions.

Data analysis and visualisation was performed inSPSS (IBM, Armonk, NY, USA), SigmaPlot and R2.13.0 (R Development Core Team, 2011).

Results

Microbial community composition and taxonomyThe 454-pyrosequencing data set consisted of34891493 (averages.d.) reads per sample, whichclustered into 11 032 different OTUs. The originalpyrosequencing output files are available at the NCBISequence Read Archive under the accession numberSRX196420. Of all found OTUs, 2.9, 53.6 and 13.2%were unique to glacial ice, streamwater and biofilms,respectively (Supplementary Figure 2). In total, 7.4%of all OTUs were found in all three habitats, 19.8%were shared by biofilm and streamwater commu-

nities and 0.6% occurred exclusively in biofilms andglacial ice. Non-metric multidimensional scalingrevealed differences in community compositionamong all three habitats (Figure 1), which werefurther identified as significant by permutationalanalysis of variance (pseudo-F 11.326, d.f.1 2,d.f.2 75, Po0.001). However, these differences incommunity compositions were confounded withdifferences in dispersion among habitats (b diversity,see below). While glacial ice communities wererelatively constrained with little variation amongglaciers, communities in the streamwater and bio-films varied broadly and overlapped (Figure 1).

Taxonomic analyses of our sequence data revealedthat Proteobacteria, Bacteroidetes and Actinobacteriawere the dominant phyla in ice, streamwater andbiofilms, whereas Cyanobacteria/chloroplasts wereabundant only in glacial ice and biofilm communities(Figure 2a). Several less abundant phyla as Verruco-microbia or Nitrospira were primarily found in thestreamwater, while they were rarely or not detected inbiofilms and ice, respectively. Comamonadaceae,Chitinophagaceae and Oxalobacteraceae were themost abundant families detected in all three habitats,

while Sphingobacteriaceae and Acetobacteraceaewere primarily identified in the ice (Figure 2b).Streamwater communities were characterised bynumerous families, many of which occurred neitherin ice nor in biofilms (Figure 2b, SupplementaryFigure 2).

Relationship between environmental variables andmicrobial community compositionTo identify the potential effect of streamwatergeochemistry and glacial characteristics on micro-bial community composition and diversity in gla-cier-fed streams, we first tested the relationship of

the variables among study sites. Principal compo-nent analysis revealed streamwater electrical con-ductivity and pH to form a major environmentalgradient across the glacier-fed streams. This gradientwas also associated with differences in elevationand glacial coverage (% glaciation of the catchment)of the respective glaciers, and all four variablestogether accounted for most of the environmentalvariation among glacier-fed streams (SupplementaryFigure 3; Supplementary Table 2).

Using redundancy analysis, we found that thephysicochemical environment significantly explai-ned variation in community composition in

-2 -1 21

-2

-1

2

1

0

0

NMDS1

NM

DS2

Ice

Streamwater

Biofilm

Figure 1 Community composition in glacial ice, streamwaterand biofilms. Non-metric multidimensional scaling analysis ofmicrobial communities based on BrayCurtis dissimilarities in 26glacial ecosystems (stress 0.16). Ellipses represent the 95%confidence interval.


1654

The ISME Journal


5/10

streamwater (F 1.399, Po0.01, n 26) and inbiofilms (F 1.419, Po0.01, n 26). In the nextstep, forward variable selection found that biofilmcommunity composition was significantlyexplained by streamwater electrical conductivity(Po0.05), pH (Po0.05) and streamwater tempera-ture (Po0.05). Likewise, electrical conductivity(Po0.01) and pH (Po0.01) were related to stream-water community composition, whereas stream-water temperature was not. Elevation explained

variation in community composition of streamwater(Po0.01), but not of biofilms. Glacial coverage wasneither related to biofilm, nor to streamwatercommunities (data not shown).

Variation in electrical conductivity and pH corre-lated with the relative abundance of microbialfamilies and to some extent even of phyla in thebiofilms (Figure 3). For instance, Actinobacteria,Nitrospira and Verrucomicrobia varied with elec-trical conductivity, whereas Acidobacteria,

Figure 2 Distribution of taxonomic groups in glacial ice, streamwater and biofilm. The percentage of ( a) phyla and (b) the 50 mostabundant families associated with each habitat is visualised in ternary plots. The position in the triangle indicates the relative abundanceof each taxon among the three habitats; the size of the circle represents the relative abundance of taxa.


1655

The ISME Journal


6/10

Gemmatimonadetes and Proteobacteria were relatedto changes in pH (Supplementary Table 3).

To better understand these differing community

composition patterns, we looked for indications ofspecies sorting as a possible mechanism of commu-nity assembly. We tested whether biofilms were aproduct of purely stochastic immigration from thesource community suspended in the streamwater,by comparing biofilm communities with randomsamples of the streamwater communities. Resam-pling revealed that in all 26 glacial ecosystems, thebiofilm communities significantly differed from thecentroid of the random assemblages (probability ofthe biofilm community to fall within the distribu-tion of the random assemblages Po0.001;Supplementary Figure 4).

Relationship between environmental variables andmicrobial biodiversitya diversity was highest in streamwater, intermediatein biofilms and lowest in glacial ice, as revealed bytrue richness (Quince et al., 2008) and rarefiedrichness (Figure 4a). Based on our results fromcommunity composition, we investigated the relation-ship between elevation, electrical conductivity, pH,streamwater temperature and a diversity. Multiplelinear regressions revealed elevation and electricalconductivity as main variables explaining the truerichness in streamwater (adjusted r2 0.67,

F 10.508, Po0.001, n 26) and biofilms (adjustedr2 0.43, F 3.883, Po0.05, n 26). Using simplelinear regression, it was found that richness of both

streamwater and biofilm communities decreasedwith higher elevation, though for biofilms thisrelationship was weak (Figures 4b and c); electricalconductivity was not significantly related to therichness of streamwater communities, while biofilmrichness decreased with electrical conductivity(r2 0.18, Po0.05, n 26; data not shown). Thesimultaneous decrease of biofilm biodiversity withelevation and electrical conductivity seems counter-intuitive, as a higher contribution of glacial melt-water to streamflow results in lower electricalconductivity. This pattern is likely driven by animprint of geology, particularly the prevalence ofcarbonate minerals, among our study streams, as

indicated by a significant correlation betweenelectrical conductivity and pH (r 0.56, Po0.01,n 26).

In agreement with patterns of community compo-sition, b diversity among glacial ecosystems washighest in biofilms, intermediate in streamwater andlowest in ice (Figure 5a). We found a significantdecrease of biofilm b diversity with streamwatertemperature (Figure 5b); electrical conductivity, pHand elevation were not significantly correlated withbiofilm b diversity (Supplementary Figure 5). Micro-bial b diversity in streamwater did not correlate withany of the tested variables (data not shown).

Tempera

ture

pH

Conduct

ivity

Elatio

nev

Glacia

tion

TotalNitro

gen

D

OC

BacillariophytaStreptophytaChlorophytaFamilyI

AcetobacteraceaeRhodobacteraceaeSphingomonadaceaeCaulobacteraceaeComamonadaceaeOxalobacteraceaeBurkholder iales incertae sedisNitrosomonadaceaeRhodocyclaceaeNeisse riaceaeMethylophilaceaeHydrogenophilaceaePseudomonadaceaeMoraxellaceaeXanthomonadaceaeBdellovibr ionaceaeCystobacteraceaeDesulfobulbaceaeHelicobacteraceaeSphingobacteriaceaeChitinophagaceaeFlavobacteriaceaeCytophagaceaeIntrasporangiaceaeMicrobacte riacea eMicrococcaceaeNocardioidaceaeHolophagaceaeGp1Gp3Gp4Gp6OpitutaceaeSubdivision3 genera incertae sedisBacillaceaeClostridiacea e

AnaerolineaceaeNitrospiraceaeDeinococcaceaeGemmatimonadaceaePlanctom ycetaceae

Cyanobacteria/Chloroplasts

NitrospiraeChloroflexi

Firmicutes

Verrucomicrobia

Acidobacteria

Actinobacteria

Bacteroidetes

Proteobacteria

Planctomycetes

DeinococcusGemmatimonadetes

Pearson correlation

-0.75 0.75

Figure 3 Heat map of Pearson correlations between environmental variables and families detected in biofilms. Colours represent ther-values of Pearson correlations between relative abundances of the 45 most abundant bacterial families and environmental parameters.Families were ordered according to taxonomic affiliations. We found several families that were strongly correlated to pH and electricalconductivity. Positive correlations with DOC were only found with Proteobacteria. Streamwater temperature and total nitrogen were onlyweakly related to the abundance of taxonomic families.


1656

The ISME Journal


7/10

Discussion

Glaciers retreat worldwide, making it necessary tobetter understand the impacts of retreating glacierson glacial ecosystems. Our study streams

encompassed a gradient of elevation, glacial cover-age and streamwater electrical conductivity, suggest-

ing distinct hydrological settings of the glacier-fedstreams at the catchment scale as a result of variablecontributions of icemelt, snowmelt and groundwaterto streamflow (Brown et al., 2003; Brown et al.,2007a; Milner et al., 2009). Additionally, catchmentgeology and hydrochemical processes may imprintthe observed environmental gradient, as indicatedby the positive correlation of streamwater electricalconductivity and pH. At local scale, depending ongeomorphology (for example, slope and slopebreaks), the ablation zone of some of our studyglaciers was fragmented following thinning, andmeltwater therefore flowed over bare rock exposed

Figure 4 Microbial a diversity. (a) Boxplot representation ofrarefied richness (all samples rarefied to 1007 reads) and truerichness estimated by fitting Sichel distribution curves to theabundance distributions obtained from the 454-pyrosequencing

data. Median (line), 1st and 3rd quartile (box margins), 5 and 95%percentiles (whiskers), outliers (points). Different letters on eachbox represent significant differences between habitats as deter-mined by Wilcoxon tests followed by Bonferroni correction(Po0.01). (b,c) Microbial richness decreased with higher eleva-tion. Linear regression include 95% confidence intervals in the(b) streamwater and (c) biofilms.

OO

a

b

Figure 5 Spatial variation of microbial communities. (a) b

diversities of the three habitats as the average distance ofcommunities to the group (that is, habitat) centroid after principalcoordinate analysis of the BrayCurtis dissimilarity matrix.b diversity differed significantly among habitats (pseudo-F 17.082, d.f.1 2, d.f.2 75, Po0.001 for all pairwise compar-isons). (b) Graphical analysis of biofilm b diversity as a function ofstreamwater temperature by local polynomial regression fitting.Multiple linear regression on distance matrices revealed b diversityto decline significantly with increasing streamwater temperature(pseudo-t 4.77, Po0.05). The shaded area gives a bootstrapconfidence interval for the generated trendline. See Materials andmethods and Supplementary Information for computational details.


1657

The ISME Journal


8/10

to solar radiation (Supplementary Figure 6). Thiscaused elevated and outlying streamwater tempera-tures in these systems and weakened the relation-ship between streamwater temperature and glacialcoverage across all sites. The physicochemicalsignatures in glacier-fed streams thus reflectedcatchment scale and local processes operating along

the various flow paths (that is, supraglacial, engla-cial, subglacial and non-glacial).

The most abundant phyla identified in our studystreams are typical for freshwater ecosystems(Tamames et al., 2010). Their regular occurrence inother ice ecosystems (Simon et al., 2009; Xiang et al.,2009; Anesio and Laybourn-Parry, 2012) suggests thatthey comprise cryophilic taxa, as also supported bycultivation-dependent approaches (Cheng and Foght,2007; Loveland-Curtze et al., 2009). Streamwaterelectrical conductivity and pH were associated withcommunity composition in biofilms at the level ofOTUs, family and even of phyla (Figure 3,Supplementary Table 3). These results are in linewith soil studies (Fierer and Jackson, 2006) suggest-ing that pHa prime physiological control on single-celled organismsstructures microbial communitieseven at the phylum level (for example, Acidobacteria,Proteobacteria), and with studies on invertebratecommunities in glacier-fed streams (Brown et al.,2007b). Certain biofilm taxa, some of them (forexample, Nitrospira) having critical roles in biogeo-chemistry, seem particularly prone to physicochem-ical shifts. Streamwater temperature, known toinfluence benthic microbial community compositionin streams (Hullar et al., 2006), appeared to furtheraffect biofilm community composition, probably at a

more local scale given the pattern of streamwatertemperature among our study streams. Expected long-term changes of physiochemical conditions followingglacial retreat could potentially have widespreadimplications for glacial stream ecosystem functions.Owing to their sessile mode of life, biofilm micro-organisms are likely even more susceptible tosuch changes compared with transient cells in thestreamwater.

Despite the cold and mainly oligotrophic condi-tions in glacial ecosystems, our true richnessestimates for streamwater and biofilm communitieswere comparable to those from non-glacial streams(Besemer et al., 2012). In line with the present study,

Besemer et al. (2012) reported higher microbialrichness in streamwater than in biofilm commu-nities. We attribute the differences in a diversitybetween ice, streamwater and biofilms to variousdegrees of environmental harshness (sensu Jacobsenand Dangles, 2011) and metacommunity (that is, aset of local communities linked by dispersal) size(Leibold et al., 2004), which interact to determinebiodiversity. In fact, glacial ice is a harsh butcomparatively constant environment (Hodsonet al., 2008) harbouring a constrained microbialcommunity, which may be particularly well adaptedto its environment; cell immigration is assumedly

dominated by atmospheric deposition (Hervas andCasamayor, 2009). Glacier-fed streams, however, aredynamic with pronounced temporal fluctuations(Brown et al., 2007a; Milner et al., 2009), and collectmicroorganisms from various glacial (for example,subglacial, englacial and supraglacial runoff; Anesioand Laybourn-Parry, 2012) and non-glacial sources

(for example, groundwater; Brown et al., 2007a;Milner et al., 2009). In fact, microorganisms fromcryoconite holes (that is, holes with high microbialactivity at the glacier surface, which form aroundsolar-heated debris; Edwards et al., 2011), theenglacial environment (Anesio and Laybourn-Parry, 2012), and also from groundwater, atmo-spheric deposition (Hervas and Casamayor, 2009),residual snow (Hervas and Casamayor, 2009) andadjacent soils and rocks (Bardgett et al., 2007;Schutte et al., 2010) may form a metacommunitythat potentially contributes to the community in theglacier-fed streams. We suggest that decreasing adiversity in the streamwater with elevation may bedue, at least in part, to less diverse sources ofmicroorganisms upstream in the catchment. Appar-ently glacial ice contributed only marginally to themicroorganisms in the streamwater, whose a diver-sity may therefore increase as other compartments ofthe metacommunity (for example, groundwater,soils) gain importance. Similarly, the low common-ality of taxa both at the OTU (SupplementaryFigure 2) and family level (Figure 2b) between iceand biofilm communities suggests minor contribu-tions of the ice communities to biofilm assembly inthe glacier-fed streams.

Differing patterns of community composition and

diversity suggest deviating assembly mechanisms inbiofilm and streamwater communities. In the frame ofmetacommunity theory (Leibold et al., 2004), speciessorting, where the local environment and bioticinteractions select from the metacommunity,would be a candidate mechanism for biofilm comm-unity assembly (Besemer et al., 2012). In contrast,streamwater communities could likely be explainedby mass effects, determined by large cell influx ratesand short residence times limiting the influence ofenvironmental factors. The local environment in theglacier-fed streams would thus select microorganismsfrom the streamwater for biofilm formation, thediversity of which is influenced by the size of the

metacommunity upstream. The latter may be affected,for instance, by the diversity of hydrological flowpaths upstream and the various habitats they connect,and also by a possible catchment scale imprint ongeochemistry (for example, electrical conductivity,pH). We found that stochastic immigration of OTUsfrom the streamwater into the biofilms was unlikely toexplain the observed community composition ofbiofilms, thus supporting the assumption that speciessorting has a role in biofilm community assembly.Collectively, these findings suggest that the variouscomponents of the hierarchical habitat templatedifferentially influence the composition of microbial


1658

The ISME Journal


9/10

communities in glacier-fed streams, which complieswith general stream ecology (Ward et al., 2002).

Ice, streamwater and biofilm communities exhib-ited different levels of spatial variation among allstudy sites, likely reflecting community responses tovarious degrees of environmental variation in thesehabitats, and different assembly mechanisms. The

relatively small range of streamwater temperaturewas apparently sufficient to influence the spatialvariation of biofilm communities among glacier-fedstreams. We suggest therefore that sensitive warm-ing of glacier-fed streams, driven by streamflowshifts from glacial to non-glacial sources or localflow over exposed bedrock in a fragmentingglacial landscape, could have an impact on thedistribution of microbial diversity in these systems(Vincent, 2010).

Substituting space for time, our findings suggestbiofilms in glacier-fed streams as sentinels of glacierretreat and expand the current knowledge on thethreatened biodiversity in the European Alps (Brownet al., 2007a; Finn et al., 2010; Jacobsen et al., 2012).Invertebrate a diversity is thought to increase asglaciers recede, because of species migratingupstream (Brown et al., 2007a; Milner et al., 2009;Finn et al., 2010). Microorganisms, however, primar-ily disperse downstream with water flow. Therefore,we suggest that both shifts in the metacommunityupstream of the glacier-fed stream and changes of thelocal physicochemical environment, as induced byglacial retreat, alter biodiversity and composition ofthe streamwater communities and, via species sorting,also of biofilm communities. Glacier retreat mayincrease microbial a diversity in glacier-fed streams,

while concomitantly reducing b diversity, andthereby contributing to the homogenisation of biofilmcommunities among glacier-fed streamssimilar tothe patterns observed for invertebrates (Jacobsen andDangles, 2011; Jacobsen et al., 2012). Spatiallyisolated and sporadically occurring invertebratepopulations face an elevated risk of local extinctionas glaciers recede, which ultimately reduces theirspatial variation (Finn et al., 2012; Jacobsen andDangles, 2011; Jacobsen et al., 2012). The responsive-ness of biofilms to the local environment in glacier-fed streams invokes a similarly patchy distributionsusceptible to homogenisation following glacierretreat. These findings call for more research explor-

ing the implications of microbial biodiversity shifts inglacier-fed streams for ecosystem processes anddownstream biogeochemistry.

Conflict of Interest

The authors declare no conflict of interest.

Acknowledgements

We are grateful to K Wagner and T Urich for support in thelaboratory and to C Preiler, L Hartmann, B Eichelberger,

B Preiler, I Hodl, and L Nicklas for assistance in the field.C Kroisleitner provided glacier coverage data. C Quinceand M Bengtsson helped with data analysis. The manu-script benefited from comments by H Peter, R Sommarugaand three anonymous reviewers. Financial support camefrom the Austrian Science Fund (START Y420-B17) to TJB.

References

Anderson MJ, Ellingsen KE, Mcardle BH. (2006a). Multi-variate dispersion as a measure of beta diversity. EcolLett 9: 683693.

Anderson MJ. (2001). A new method for non-parametricmultivariate analysis of variance. Austral Ecol26: 3246.

Anderson MJ. (2006b). Distance-based tests for homogene-ity of multivariate dispersions. Biometrics 62: 245253.

Anesio AM, Laybourn-Parry J. (2012). Glaciers and icesheets as a biome. Trends Ecol Evol 27: 219225.

Bardgett RD, Richter A, Bol R, Garnett MH, Baumler R,Xu X et al. (2007). Heterotrophic microbial commu-nities use ancient carbon following glacial retreat.

Biol Lett 3: 487490.Battin JT, Wille A, Psenner R, Richter A. (2004). Large-scale environmental controls on microbial biofilms inhigh-alpine streams. Biogeoscience 1: 159171.

Battin TJ, Kaplan LA, Findlay S, Hopkinson CS, Marti E,Packman AI et al. (2008). Biophysical controls onorganic carbon fluxes in fluvial networks. Nat Geosci1: 95100.

Berry D, Mahfoudh KB, Wagner M, Loy A. (2011).Barcoded primers used in multiplex ampliconpyrosequencing bias amplification. Appl EnvironMicrobiol 77: 78467849.

Besemer K, Peter H, Logue JB, Langenheder S, LindstromES, Tranvik LJ et al. (2012). Unraveling assembly ofstream biofilm communities. ISME J 6: 14591468.

Blanchet FG, Legendre P, Borcard D. (2008). Forwardselection of explanatory variables. Ecology89: 26232632.Brown LE, Hannah DM, Milner AM. (2003). Alpine stream

habitat classification: An alternative approach incor-porating the role of dynamic water source contribu-tions. Arct Antarct Alp Res 35: 313322.

Brown LE, Hannah DM, Milner AM. (2007a). Vulnerabilityof alpine stream biodiversity to shrinking glaciers andsnowpacks. Glob Change Biol 13: 958966.

Brown LE, Milner AM, Hannah DM. (2007b). Groundwaterinfluence on alpine stream ecosystems. FreshwaterBiol 52: 878890.

Cheng SM, Foght JM. (2007). Cultivation-independentand dependent characterization of bacteria resident

beneath John Evans Glacier. FEMS Microbiol Ecol 59:

318330.Cleveland WS, Grosse E, Shyu WM. (1992). Localregression models. Chambers JM, Hastie TJ (eds)Statistical Models in S. Wadsworth & Brooks/Cole,Pacific Grove, CA, USA.

Edwards A, Anesio AM, Rassner SM, Sattler B, Hubbard B,Perkins WT et al. (2011). Possible interactions between

bacterial diversity, microbial activity and supraglacialhydrology of cryoconite holes in Svalbard. ISME J 5:150160.

Fierer N, Jackson RB. (2006). The diversity and biogeo-graphy of soil bacterial communities. Proc Natl AcadSci 103: 626631.

Finn DS, Bonada N, Murria C, Hughes JM. (2011). Smallbut mighty: headwaters are vital to stream network


1659

The ISME Journal


10/10

biodiversity at two levels of organization. J N A mBenthol Soc 30: 963980.

Finn DS, Khamis K, Milner AM. (2012). Loss of smallglaciers will diminish beta diversity in Pyreneanstreams at two levels of biological organization. GlobalEcol Biogeogr 22: 4051.

Finn DS, Rasanen K, Robinson CT. (2010). Physical andbiological changes to a lengthening stream gradient

following a decade of rapid glacial recession. GlobChange Biol 16: 33143326.

Goslee SC, Urban DL. (2007). The ecodist package fordissimilarity-based analysis of ecological data. J StatSoftw 22: 119.

Hannah DM, Brown LE, Milner AM, Gurnell AM,McGregor GR, Petts GE et al. (2007). Integratingclimatehydrologyecology for alpine river systems.Aquatic Conserv: Mar Freshw Ecosyst 17: 636656.

Hervas A, Casamayor EO. (2009). High similarity betweenbacterioneuston and airborne bacterial communitycompositions in a high mountain lake area. FEMSMicrobiol Ecol 67: 219228.

Hodson A, Anesio AM, Tranter M, Fountain A, Osborn M,Priscu J et al. (2008). Glacial ecosystems. Ecol Monogr78: 4167.

Hughes JM, Schmidt DJ, Finn DS. (2009). Genes instreams: using DNA to understand the movement offreshwater fauna and their riverine habitat. Bioscience59: 573583.

Hullar MAJ, Kaplan LA, Stahl DA. (2006). Recurringseasonal dynamics of microbial communities instream habitats. Appl Environ Microbiol 72: 713722.

Jacobsen D, Dangles O. (2011). Environmental harshnessand global richness patterns in glacier-fed streams.Global Ecol Biogeogr 21: 647656.

Jacobsen D, Milner AM, Brown LE, Dangles O. (2012).Biodiversity under threat in glacier-fed river systems.Nature Clim Change 2: 361364.

Legendre P, Oksanen J, Braak CFJ. (2011). Testing thesignificance of canonical axes in redundancy analysis.Methods Ecol Evol 2: 269277.

Leibold MA, Holyoak M, Mouquet N, Amarasekare P,Chase JM, Hoopes MF et al. (2004). The metacommu-nity concept: a framework for multi-scale communityecology. Ecol Lett 7: 601613.

Lichstein JW. (2006). Multiple regression on distancematrices: a multivariate spatial analysis tool. PlantEcol 188: 117131.

Loveland-Curtze J, Miteva VI, Brenchley JE. (2009).Herminiimonas glaciei sp. nov., a novel ultramicro-

bacterium from 3042 m deep Greenland glacial ice. IntJ Syst Evol Microbiol 59: 12721277.

Milner AM, Brown LE, Hannah DM. (2009). Hydroecolo-

gical response of river systems to shrinking glaciers.Hydrol Process 23: 6277.Murdoch D, Chow ED. (2007). Ellipse: functions for

drawing ellipses and ellipse-like confidence regions.R Package Version 0.3 5.

Oksanen J, Blanchet FG, Kindt R, Legendre P, OHara RB,Simpson GL et al. (2011). Vegan: community EcologyPackage. R Package Version 1.17-11.

Ptacnik R, Andersen T, Brettum P, Lepisto L, Willen E.(2010). Regional species pools control communitysaturation in lake phytoplankton. Proc R Soc B 277:37553764.

Quince C, Curtis TP, Sloan WT. (2008). The rational

exploration of microbial diversity. ISMEJ2: 9971006.Quince C, Lanzen A, Davenport RJ, Turnbaugh PJ. (2011).

Removing noise from pyrosequenced amplicons. BMCBioinformatics 12: 38.

R Development Core Team (2011). R: A Language andEnvironment for Statistical Computing. R Foundationfor Statistical Computing: Vienna, AT.

Schutte UME, Abdo Z, Foster J, Ravel J, Bunge J,Solheim B et al. (2010). Bacterial diversity in a glacierforeland of the high Arctic. Mol Ecol 19: 5466.

Sichel H. (1974). On a distribution representing sentence-length in written prose. J R Stat Soc A 137: 2534.

Simon C, Wiezer A, Strittmatter AW, Daniel R. (2009).Phylogenetic diversity and metabolic potentialrevealed in a glacier ice metagenome. Appl EnvironMicrobiol 75: 75197526.

Singer GA, Fasching C, Wilhelm L, Niggemann J, Steier P,Dittmar T et al. (2012). Biogeochemically diverseorganic matter in Alpine glaciers and its downstreamfate. Nature Geosci 5: 710714.

Tamames J, Abellan JJ, Pignatelli M, Camacho A, Moya A.(2010). Environmental distribution of prokaryotictaxa. BMC Microbiol 10: 85.

Vincent WF. (2010). Microbial ecosystem responses to rapidclimate change in the Arctic. ISME J 4: 10891091.

Wang Q, Garrity GM, Tiedje JM, Cole JR. (2007). NaveBayesian classifier for rapid assignment of rRNAsequences into the new bacterial taxonomy. ApplEnviron Microbiol 73: 52615267.

Ward JV, Tockner K, Arscott DB, Claret C. (2002). Riverinelandscape diversity. Freshwater Biol 47: 517539.

Warnes GR, Bolker B, Gorjanc G, Grothendieck G,Korosec A, Lumley T et al. (2011). gdata: Various Rprogramming tools for data manipulation. R PackageVersion 2.8.2..

Woodward G, Perkins DM, Brown LE. (2010). Climatechange and freshwater ecosystems: impacts acrossmultiple levels of organization. Phil Trans R Soc B365: 20932106.

Xiang S, Shang T, Chen Y, Jing Z, Yao T. (2009). Dominantbacteria and biomass in the Kuytun 51 Glacier. ApplEnviron Microbiol 75: 72877290.

This work is licensed under a Creative

Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License. To view a copyof this license, visit http://creativecommons.org/licenses/by-nc-nd/3.0/

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


1660

The ISME Journal
http://creativecommons.org/licenses/by-nc-nd/3.0/http://creativecommons.org/licenses/by-nc-nd/3.0/http://www.nature.com/ismejhttp://www.nature.com/ismejhttp://creativecommons.org/licenses/by-nc-nd/3.0/http://creativecommons.org/licenses/by-nc-nd/3.0/

Documents

Biodiversidad Microbiana en Corrientes Que Alimntan Glaciares Wilhem Et Al 2013