Bacterial diversity along a 2600km river continuum

14 Pages·2015·1.38 MB·English

bs_bs_banner
EnvironmentalMicrobiology(2015)17(12),4994–5007 doi:10.1111/1462-2920.12886
Bacterial diversity along a 2600 km river continuum
DomenicoSavio,1,2LucasSinclair,3UmerZ.Ijaz,4 are commonly observed in freshwaters. Dominated
JurajParajka,1,5GeorgH.Reischer,2,6 by the acI lineage, the freshwater SAR11 (LD12)
PhilippStadler,1,7AlfredP.Blaschke,1,5 and the Polynucleobacter group, typical freshwater
GünterBlöschl,1,5RobertL.Mach,2 taxa increased in proportion downriver and were
AlexanderK.T.Kirschner,6,8 accompaniedbyadecreaseinsoilandgroundwater-
AndreasH.Farnleitner1,2,6andAlexanderEiler3* affiliated bacteria. Based on views of the meta-
1CentreforWaterResourceSystems(CWRS),Vienna community and River Continuum Concept, we
UniversityofTechnology,Vienna,Austria. interpret the observed taxonomic patterns and
2ResearchGroupEnvironmentalMicrobiologyand accompanying changes in alpha and beta diversity
MolecularEcology,InstituteofChemicalEngineering, with the intention of laying the foundation for a
ViennaUniversityofTechnology,Vienna,Austria. unifiedconceptforriverbacterioplanktondiversity.
5InstituteofHydraulicEngineeringandWaterResource
Management,ViennaUniversityofTechnology,Vienna,
Introduction
Austria.
7InstituteforWaterQuality,ResourceandWaste Streamsandriverslinkterrestrialandlenticsystemswith
Management,ViennaUniversityofTechnology,Vienna, theirmarinecounterpartsandprovidenumerousessential
Austria. ecosystemservices.Theysupplydrinkingwater,areused
3DepartmentofEcologyandGenetics,Limnology, forirrigation,industryandhydropowerandserveastrans-
ScienceforLifeLaboratory,UppsalaUniversity, portroutesorforrecreation.Ofgeneralimportanceisthe
Uppsala,Sweden. role of lotic systems in biogeochemical nutrient cycling.
4SchoolofEngineering,UniversityofGlasgow, Untilrecently,riversandstreamsweremainlyconsidered
Glasgow,UK. aspipesshuttlingorganicmaterialandnutrientsfromthe
6InteruniversityCooperationCentreWaterandHealth, landtotheocean(Coleetal.,2007).Thisviewhasbegun
www.waterandhealth.at. tochangeasloticandlenticsystemsarenowconsidered
8InstituteforHygieneandAppliedImmunology,Water more akin to ‘leaky funnels’ in regard to the cycling of
Hygiene,MedicalUniversityofVienna,Vienna,Austria. elements.Indeed,theyplayanimportantroleinthetem-
porary storage and transformation of terrestrial organic
matter (Ensign and Doyle, 2006; Cole etal., 2007;
Summary
Withers and Jarvie, 2008; Battin etal., 2009). As a
Thebacterioplanktondiversityinlargerivershasthus resultofrecognizingthespecificroleofstreamsandrivers
far been under-sampled despite the importance of in the carbon cycle (Richey etal., 2002; Battin etal.,
streams and rivers as components of continental 2009; Raymond etal., 2013), the study of the diverse
landscapes. Here, we present a comprehensive processesongoinginthewatercolumnandsedimentsof
datasetdetailingthebacterioplanktondiversityalong loticnetworkshasreceivedincreasinginterest(Kronvang
themidstreamoftheDanubeRiveranditstributaries. etal.,1999;Beaulieuetal.,2010;Seitzingeretal.,2010;
Using 16S rRNA-gene amplicon sequencing, our Aufdenkampe etal., 2011; Benstead and Leigh, 2012;
analysis revealed that bacterial richness and even- Raymondetal.,2013).
ness gradually declined downriver in both the free- Whenattemptingtomodelthemechanismsofnutrient
livingandparticle-associatedbacterialcommunities. processing in freshwater systems, bacteria are regarded
These shifts were also supported by beta diversity as the main transformers of elemental nutrients and
analysis, where the effects of tributaries were negli- viewed as substantial contributors to the energy flow
gible in regards to the overall variation. In addition, (Cotner and Biddanda, 2002; Battin etal., 2009; Findlay,
the river was largely dominated by bacteria that 2010;Madsen,2011).However,inthecaseofopenlotic
systems such as rivers, there remains a lack of knowl-
edge concerning the diversity of bacterial communities
Received20August,2014;accepted21April,2015.*Forcorrespond-
ence. E-mail [email protected]; Tel. [phone]; (Battinetal.,2009).Forexample,upuntilthepresentday
Fax[phone]. there is no agreement on the distinctness of river
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd.
ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttributionLicense,whichpermitsuse,distributionand
reproductioninanymedium,providedtheoriginalworkisproperlycited.
Riverbacterioplanktondiversity 4995
Fig.1. OverviewanddetailedmapoftheDanuberivercatchmentshowingallsamplingsitesduringtheJointDanubeSurvey2;reddots
indicatesamplingpointsinthemidstreamoftheDanubeRiver;bluedotsrepresentsamplingpointsintributariesbeforemerging
withtheDanubeRiver.Blue-shadedfontindicatesofficialnumberingofriverkilometres,startingwithrkm2600attheuppermostsitetorkm0
attherivermouth.Countryabbreviationsandlargecitiesarewritteninblack.ThemapwascreatedusingQuantumGIS(QuantumGIS
DevelopmentTeam,2011).
bacterioplanktoncommunitiesfromthatofotherfreshwa- to suggest a persistent and ubiquitous ‘core bacterial
ter systems or the variability of its diversity along entire community’ along a river stretch. Read and colleagues
rivernetworks. (2015) examined the longitudinal development of the
Untilrecently,itwasonlyknownthatthemostabundant bacterioplankton community at 23 sites along the river
taxacomprisingriverinebacterioplanktonseemtoresem- networkofthe9948km2Thamesbasin.Theyfoundashift
ble lake bacteria and can thus be regarded as ‘typical’ from a Bacteroidetes-dominated community in the head-
freshwater bacteria (Zwart etal., 2002; Lozupone and waters to an Actinobacteria-dominated community in the
Knight, 2007; Newton etal., 2011). In particular, bacteria lower reaches near the river mouth and location of the
affiliated with the phyla Proteobacteria (particularly samplingpointintherivernetworktobethemostpredic-
Betaproteobacteria), Actinobacteria, Bacteroidetes, tive parameter. These patterns they interpreted as evi-
CyanobacteriaandVerrucomicrobiawerefoundtodomi- denceforecologicalsuccessionalongtherivercontinuum.
nate the bacterial communities in rivers (Crump etal., However,theexistingstudiesfocusedonrelativelysmall
1999; Zwart etal., 2002; Cottrell etal., 2005; Winter river basins and/or a small number of sampling sites. In
etal., 2007; Lemke etal., 2008; Mueller-Spitz etal., large river basins, however, one would expect that the
2009; Newton etal., 2011; Liu etal., 2012). A recent spatialpatternsofbacterialcommunitycompositionsmani-
metagenome study corroborates a general dominance festthemselvesmoreclearlythaninsmallonesduetothe
of the phyla Proteobacteria and Actinobacteria, and largercontrastinenvironmentalconditions.Inthispaper,
morespecificallythecleardominanceofthecosmopolitan weanalysetheresultsfromasecond-generationsequenc-
freshwater lineage acI of the phylum Actinobacteria ing experiment by separately investigating the free-living
in the Amazon river (Ghai etal., 2011). The dominance and particle-associated bacterioplankton communities
of Actinobacteria and Proteobacteria in riverine along 96 sites in the network of the entire Danube basin
bacterioplanktonwasalsoconfirmedinthreerecenthigh- (Fig.1).TheDanubeRiveris2780kminlengthanddrains
throughput sequencing studies on the Upper Mississippi a catchment area of 801000km2 with 83 million inhabit-
River(USA;Staleyetal.,2013),theYeniseiRiver(RUS; ants (Schmedtje etal., 2004; Sommerwerk etal., 2010).
Kolmakovaetal.,2014)andtheRiverThames(UK;Read Basedonourresults,weproposethatthebacterioplankton
etal., 2015). Staley and colleagues (2013) were the first communitiesinthemidstreamofsuchalargeriverdevelop
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd,
EnvironmentalMicrobiology,17,4994–5007
4996 D.Savioetal.
illustrated in Fig.S1D–G. Total phytoplankton biomass
(Chla) showed a peak between river kilometre 1481 and
P
1107(sites38–55)withtotalbacterialproductionfollowing
asimilartrend,whereastotalsuspendedsolidconcentra-
tion increased considerably in the last 900 kilometres
beforereachingtheBlackSea(Fig.S1H–J).
Several geomorphological measures were calculated
suchas‘riverkilometre’,‘catchmentarea’,‘meandendritic
stream length’ and ‘cumulative dendritic distance
upstream’ and were compared with each other. For
example, ‘mean dendritic stream length’ correlated very
closelywith‘riverkilometre’(asdefinedbythedistanceto
P mouth; linear model R2=0.98; P<0.001; Fig.S1A). In
contrast, ‘cumulative dendritic distance upstream’ corre-
latedalmostperfectlywith‘catchmentarea’(linearmodel
R2>0.99;P<0.001;Fig.S1C).Fromahydrologicalpoint
of view, it can be argued that ‘mean dendritic stream
Fig.2. Gradualdevelopmentofthereadproportionsassignedto
length’ is a better proxy of stream residence time than
theoperationallydefined‘corecommunities’ofthefree-livingand
particle-associatedfraction(allOTUsoccurringinatleast90%of ‘cumulativedendriticdistanceupstream’,asitrepresents
allDanubeRiversamplesoftherespectivesizefraction)along the average travel time of a drop of water, assuming
‘meandendriticstreamlength’.Roundsymbolsrepresentsamples
fromthefree-livingsizefraction(0.2–3.0μm);squaresrepresent randomlydistributedspringdischargesandconstantflow
samplesfromtheparticle-associatedsizefraction(>3.0μm).Red velocities in the river system (Rodriguez-Iturbe etal.,
dotsindicatesamplesfromthefree-livingsizefraction(0.2–3.0μm) 2009).
oftheDanubeRiveronly;bluesquaresindicatesamplesfromthe
particle-associatedfraction(>3.0μm)oftheDanubeRiver;open
dotsandsquaresrepresenttributarysamplesfromthefree-living
Coremicrobialcommunity
andparticle-associatedsizefraction,respectively;darkbluelines
indicatefittedlinearmodelswithconfidenceintervalsof0.95inred
Intotal,sequencingresultedin1572361sequencereads
andbluefortherespectivesizefractionoftheDanubeRiver
samples.Detailedregressionstatisticsforthecorecommunity (further referred to as ‘reads’) after quality filtering and
developmentintheDanubeRiver(exclusivetributarysamples)are clustered into clustering into 8697 bacterial operational
showninthefigure.
taxonomic units (OTUs). The majority of bacteria-
assignedOTUs(4402outof8697)wereonlyrepresented
graduallyandincreasinglyindependentfromtributaryand
by less than 10 reads in the entire dataset.As a conse-
riparian influence as a result of the interplay between
quence,3243of8697OTUs(∼37%)werepresentinonly
dispersal-facilitated (‘mass effects’) and environmental
onetofoursamples,andanadditional2219OTUs(∼26%)
condition-based sorting (‘species sorting’; Leibold etal.,
were present in as few as five to nine samples. Besides
2004;Crumpetal.,2007;2012).Moreover,wearguethat
these rare OTUs, the core community of the Danube
these processes, represented in the meta-community
River,asoperationallydefinedbyallOTUsthatappeared
concept, can be linked to the River Continuum Concept
in at least 90% of all Danube River samples, comprised
(RCC;Vannoteetal.,1980),whichexplainstheroleofthe
89 OTUs in the free-living bacterioplankton (0.2–3.0μm)
hydrological flow conditions, the riparian zone, substrate
and 141 OTUs in the particle-associated fraction
and food as important factors in determining community
(>3.0μm).Onaverage,81%ofallreadsofthefree-living
structuresalongentireriversystems.
river community and 63% of all reads of the particle-
associated river community were part of their respective
Results corecommunity.Therelativeabundanceofthecorecom-
munities in both fractions increased significantly towards
Selectedenvironmentalandgeomorphological
therivermouthwithsimilarslopesrevealedbyregression
parameters
analysis(Fig.2).
In total, more than 280 individual parameters, including
chemical, microbiological, ecotoxicological, radiological
Variabilityinriverbacterioplanktonalphadiversity
and biological parameters were investigated within the
JointDanubeSurvey2(Liskaetal.,2008).Alkalinity,pH, To investigate alpha diversity, we calculated the Chao1
concentration of nitrate as well as dissolved silicates richness estimator and Pielou’s evenness index for both
exhibited a gradually decreasing trend along the river as size fractions after rarefying all samples down to 7000
previouslydescribedbyLiskaandcolleagues(2008)and reads and discarding 36 samples with fewer reads. We
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd,
EnvironmentalMicrobiology,17,4994–5007
Riverbacterioplanktondiversity 4997
P P
P
P
Fig.3. Thegradualdevelopmentof(A)thebacterialrichness(Chao1)and(B)Pielou’sevenness(J)alongthe‘meandendriticstreamlength’
ateachsamplingsite;reddotsindicatesamplesfromthefree-livingsizefraction(0.2–3.0μm)oftheDanubeRiveronly(n=27);bluesquares
indicatesamplesfromtheparticle-associatedfraction(>3.0μm)oftheDanubeRiver(n=40);opendotsandsquaresrepresenttributary
samplesfromthefree-livingandparticle-associatedsizefraction,respectively;darkbluelinesindicatefittedlinearmodelswithconfidence
intervalsof0.95inredandbluefortherespectivefractionofDanubeRiversamples.DetailedregressionstatisticsforDanubeRiversamples
(exclusivetributarysamples)areshowninthefigure.
observed the highest diversity of all samples in the described based on the observation of a higher richness
upstreampartoftheDanubeRiver,representingmedium- intheparticle-associatedfraction.
sized stream reaches according to the RCC definition.
Richnessandevennessthengraduallydecreaseddown-
Variabilityinriverbacterioplanktonbetadiversity
streaminbothsizefractions(Fig.3AandB)asconfirmed
by regression analysis using ‘mean dendritic stream While communities of both size fractions in the Danube
length’ as well as ‘river kilometre’, ‘cumulative dendritic Rivercovarysignificantlywithalkalinity,nitrateconcentra-
distance upstream’ and ‘catchment area’ (TableS1). For tion and dissolved silicates, the particle-associated com-
bacterial richness, the regressions revealed very similar munitiesadditionallycovariedhighlysignificantlywithtotal
slopesforthefree-livingandtheparticle-associatedfrac- bacterialproduction,nitriteconcentrations,phytoplankton
tions (Fig.3A). Besides the observed trends in the mid- biomassandtotalsuspendedsolids(Table1).Thesecor-
stream communities of the Danube River, the tributary relationsindicatethatchemicalpropertiesmaydetermine
communitiesfrequentlyformedoutlierswithconsiderably bacterioplankton community composition. Nevertheless,
lowerrichnessandevenness(Fig.3AandB). theobservedsignificantrelationshipbetweencommunity
Among the two size fractions, the estimated richness compositionand‘meandendriticstreamlength’(Table1)
wasconsistentlyhigherintheparticle-associatedcommu- emphasizes the underlying role of stream travel times.
nities than in the free-living fraction (Wilcoxon rank sum Still, as tributary samples were scattered in the multidi-
test;P-value<0.001)withmeansof2025OTUsand1248 mensional space (Fig.4), the midstream Danube River
OTUs, respectively (Fig.3A). Similar observations were communities are suggested to develop independently
reported in studies on (coastal) marine environments as from tributary communities, which is further supported
well as lentic freshwater environments (Bižic´-Ionescu when depicting tributaries according to their position of
etal.,2014;Mohitetal.,2014),ascribingthehigheralpha confluencewithintheDanubeRiver(seeFig.S2).
diversity in the particle-associated communities to the In addition to the observation that (i) tributary
high heterogeneity in the particle microenvironment. A bacterioplankton communities did not follow the general
large spectrum of niches can be provided by the high patterns of midstream Danube River communities and
heterogeneity among particles in rivers including mobi- oftenformedoutliersintheordinationspace,othervisual
lized sediments, living organisms such as planktonic impressionsfromthenon-metricmultidimensionalscaling
algaeorzooplanktonanddetritusderivedfromterrestrial (NMDS) are that (ii) there is a distinction in community
andaquaticsources.Bižic´-Ionescuandcolleagues(2014) compositionbetweenthetwosizefractionsasconfirmed
even suggested that the presence of diversely colonized by permutational multivariate analysis of variance
particles of different age, origin and composition can be (PERMANOVA) analysis (R2=0.156, P-value<0.01),
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd,
EnvironmentalMicrobiology,17,4994–5007
4998 D.Savioetal.
Table1. Summarystatisticsofcorrespondencebetweenenvironmentalvariablesandtheprojectionsofbacterioplanktoncommunitysamplesin
theNMDSordinationbasedoneitherfree-livingorparticle-associatedsizefractionsfortheDanubeRiversamplesonly(left)andtributarysamples
included(right).Theresultswereobtainedusingthefunction‘envfit’includedintheR-package‘vegan’(Oksanenetal.,2013).
DanubeRiveronly DanubeRiverandtributarysamples
Particle Particle
Freeliving associated Freeliving associated
R2 R2 R2 R2
OfficialDanubeRiverkilometrea(fortributariesrkmatconfluence) 0.840*** 0.832*** 0.241*** 0.236***
OfficialDanubeRiverkilometrea(calculatedfortributariesb) 0.840*** 0.832*** 0.628*** 0.574***
Meandendriticstreamlength(waterresidencetime) 0.799*** 0.808*** 0.620*** 0.579***
Mediandendriticlength 0.752*** 0.773*** 0.594*** 0.554***
Catchmentsize 0.766*** 0.803*** 0.559*** 0.524***
Cumulateddendriticdistance 0.772*** 0.807*** 0.568*** 0.527***
Nitrate 0.677*** 0.538*** 0.294** 0.099*
Alkalinity 0.605*** 0.430*** 0.329** 0.159*
Silicatesdissolved 0.500*** 0.589*** 0.200* 0.167*
Totalbacterialproduction 0.159* 0.440*** 0.399** 0.445***
Bacterialproductionfilteredfraction 0.059 0.468*** 0.379* 0.425***
Nitrite 0.220* 0.283*** 0.039 0.160*
Phytoplanktonbiomass(Chla) 0.016 0.396*** 0.039 0.222**
Totalsuspendedsolids 0.143* 0.347*** 0.117 0.166**
pH 0.140 0.175** 0.088 0.031
Watertemperature 0.100 0.029 0.103 0.002
Organicnitrogen 0.055 0.131* 0.032 0.091*
Conductivity 0.200* 0.107 0.131* 0.418***
Orthophosphatephosphorus 0.083 0.075 0.583*** 0.337***
Ammonium 0.073 0.026 0.445** 0.301**
Totalphosphorus 0.057 0.063 0.078 0.400***
Dissolvedoxygen 0.052 0.007 0.248** 0.069
Significantcodes:***≤0.001**≤0.01
*≤0.05.
a. rkm2600=upstreamregion[nearUlm(DE)];rkm0=rivermouth(BlackSea).
b. Toobtainthecorresponding’Danuberiverkilometres’accordingtotheofficialnumbering(seea)fortributaries,thelengthoftributarieswas
calculatedbysubtractingtheofficiallengthoftributaries(Schmedtjeetal.,2004)fromtheofficiallengthoftheDanubeRiver(2780km;Schmedtje
etal.,2004).
Fig.4. Non-metricmultidimensionalscaling
plotofthecompositionaldissimilarities
betweencommunities(Bray–Curtis
dissimilarities)ofallsamplesoftheDanube
Riveranditstributaries.Thestressvalueof
NMDSwas0.17.Dotsrepresentfree-living
bacterialcommunities(0.2–3.0μm);triangles
displayparticle-associatedbacterial
communities(>3.0μm).Opensymbols
representtributarysamples,whereasfull
Tributaries symbolsindicatecommunitiesintheDanube
River.Thegradientfromorangetobluevia
purpleindicatesthe‘meandendriticstream
length’attherespectivesamplingsiteinthe
DanubeRiver.
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd,
EnvironmentalMicrobiology,17,4994–5007
Riverbacterioplanktondiversity 4999
A
Danube River Tributaries
Free-living fraction Particle-associated fraction Free-l. P.-assoc.
16 %
LD12 14 %
acI-B1 12 %
acI-A7 10 %
acI-C2 8 %
Pnec
6 %
Luna1-A2
4 %
Algor
2 %
Iluma-A1
0 %
B 100%
80%
60%
40%
20%
0%
Flow direction Flow direction
Upstream River Upstream River Up RM Up RM
region mouth region mouth
(rkm 2600) (rkm 0) (rkm 2600) (rkm 0)
Freshwater taxa Freshwater 'clade' or 'lineage' non-typical freshwater taxa
Fig.5. Aheatmap(A)revealingthedynamicsoftheeightmostabundanttypicalfreshwatertribesalongtheDanubeRiverdefinedaccording
toNewtonandcolleagues(2011).Thegradientfromblackredviayellowtowhiteindicatestherelativequantitativecontributionofthe
respectivetribetoallsequencereadsinanyonesample,withamaximumof16%.
Panel(B)displaystheoverallcontributionoftypicalfreshwatertribes(darkblue)aswellascladesandlineages(turquoise)accordingtothe
definitionbyNewtonandcolleagues(2011)totheriverbacterioplanktonampliconsequencesalongtheriver;blackbarsrepresentreadsthat
couldnotbematchedtosequencesoftheusedfreshwaterdatabase(Newtonetal.,2011)neitherontribe,cladeorlineage-level(named
‘non-typicalfreshwatertaxa’).‘Freshwatertaxa’and‘freshwatercladeorlineage’representallreadsthatcouldbematchedtosequencesof
theusedfreshwaterdatabaseattherespectivesimilaritylevel.SamplesfromtheDanubeRiveraswellastheinvestigatedtributariesare
arrangedfromlefttotheright,withincreasingdistancefromthesourceandseparatedaccordingtotherespectivesizefractions.
and(iii)thereappearstobesynchronyinthecommunity thefree-livingfraction.Intheparticle-associatedfraction,
changesofthetwosizefractionsalongtheriver’scourse, these trends in phylum composition were less pro-
which we statistically verified using a Procrustes test nounced. In addition to assigning reads to the phylum
(r=0.96,P<0.001).Furthermore,(iv)apermutationtest level, we taxonomically annotated all 9322 OTUs using
on the beta dispersion values between the samples of similarity searches against the database of freshwater
eachsizefractionrevealedahighervariability(byafactor bacteria developed by Newton and colleagues (2011).
of0.06)inthe>3.0μmfractionwhencomparedwiththe Theanalysisrevealedthatupto80%ofthefree-livingand
0.2–3.0μm fraction (P-value=0.002) (see Fig.S3). This morethan65%oftheparticle-associatedbacterialpopu-
supportstheideaofalargerheterogeneityinnicheavail- lation inhabiting the midstream Danube River communi-
ability in the particle-associated communities based on tiescouldbeassignedtopreviouslydescribedfreshwater
higherdiversityinparticleageandorigin. taxa (Fig.5B). In particular, these included representa-
tives of the LD12-tribe belonging to the subphylum of
Alphaproteobacteria, as well as the acI-B1-, acI-A7- and
Typicalriverbacterioplankton
acI-C2-tribes belonging to the phylum Actinobacteria.
Along the river, the bacterioplankton community Such dominance of few freshwater taxa in riverine
was dominated by Actinobacteria, Proteobacteria, bacterioplankton communities (see also Zwart etal.,
Bacteroidetes, Verrucomicrobia and candidate division 2002; Lozupone and Knight, 2007; Newton etal., 2011;
OD1, with an increasing proportion of reads assigned to Read etal., 2015) corroborates the idea that river
the phylum Actinobacteria in the free-living size fraction bacterioplanktonresemblesthoseoflakes.
downriver (Fig.S4). In contrast, reads assigned to Interestingly, in the free-living size fraction, we
Bacteroidetes decreased significantly along the river in observedaclearincreaseintherelativeabundanceofthe
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd,
EnvironmentalMicrobiology,17,4994–5007
5000 D.Savioetal.
these OTUs may originate from non-aquatic sources. In
the particle-associated fraction, typical freshwater taxa
werelesscommon(Fig.5B).
To confirm the non-aquatic origin of certain OTUs, we
firstblastedarepresentativeforeachofthe8697bacterial
OTUs against the NCBI-Nucleotide database; next, any
environmental descriptive terms occurring in the search
results were retrieved and classified according to the
EnvironmentalOntology(EnvO;Buttigiegetal.,2013)ter-
minology.PermutationalanalysisofvarianceoftheEnvO-
classifieddatarevealedasignificantdifferenceinvariance
betweenthetwosizefractions(PERMANOVA;R2=0.42,
P<0.0001), supporting a high variability in origin and
particle age. Restricting the analysis to particular EnvO
terms such as ‘groundwater’ and ‘soil’ suggests that the
proportion of bacteria potentially originating from these
sourcesdecreasedtowardstheriver’smouth(Fig.6Aand
B), while ‘river’, ‘lake’ and ‘epilimnion’ terms exhibited
increasingtrendsdownriver(Fig.S5A–C).Inparticular,we
couldonlyidentifyfourOTUsreceivingaclassificationthat
wasdominatedbythe‘river’term.
As with every homology-based assignment, SEQENV
resultsareaffectedbydatabaseentries;forexamplethe
under-representationofentriesfromriverscomparedwith
lakes likely biases against the ‘river’term.Thus, a larger
number of typical river bacterioplankton may exist than
detectedbyouranalysis.
Discussion
The tremendous diversity within the microbial communi-
tiesinhabitingalltypesofenvironmentsisbeingrevealed
Fig.6. ResultsfromtheSEQENVanalysisscoringsequences by a rapidly increasing number of studies applying high-
accordingtotheirenvironmentalcontext(EnvO;environmental
ontology).TheY-axisrepresentstheproportionof(A)‘groundwater’ throughput sequencing technologies (e.g. Sogin etal.,
and(B)‘soil’termsassociatedwithsequencereadspersample 2006; Andersson etal., 2009; Galand etal., 2009; Eiler
alongthe‘meandendriticstreamlength’ateachsamplingsite
etal., 2012; Peura etal., 2012).At the same time, many
(X-axis).ReddotsindicateDanubeRiversamplesofthe
0.2–3.0μmsizefraction(n=42),andbluesquaresindicate mechanisms modulating this diversity have been sug-
samplesfromthe>3.0μmsizefraction(n=52).Opendotsand gested including ‘mass effects’ and ‘species sorting’,
squaresrepresenttributarysamplesofthefree-livingand
which vary widely in importance depending on the envi-
particle-associatedsizefractionsrespectively.Darkbluelines
representfittedlinearmodelsfortheDanubeRiversampleswith ronment (Leibold etal., 2004; Besemer etal., 2012;
confidenceintervalsof0.95inredandbluefortherespective Hanson etal., 2012; Lindström and Langenheder, 2012;
fractions.DetailedregressionstatisticsaregiveninTableS1.
Szekelyetal.,2013).
Regardingbacterioplanktoninlargerivernetworks,we
propose that the highest diversity exists in headwaters,
four above-mentioned tribes towards the river mouth and from thereon decreases towards river mouths. This,
(Fig.5A), contributing up to 35% of the community. The weargue,resultsfromtheinoculationofbacterioplankton
increasing relative abundance of these four tribes was byadvectionfromsurroundingenvironments(i.e.soiland
accompaniedbyageneralincreaseinrelativeabundance groundwater) in the headwaters as supported by our
of OTUs matching other freshwater tribes, lineages or SEQENVresultsandpreviousstudies(Crumpetal.,2007;
clades according to Newton and colleagues (2011) as 2012; Besemer etal., 2012; 2013). This initial pervasive
depicted in Fig.5B. In contrast, the number of OTUs not impact from the riparian zone on headwaters (‘mass
matchinganysequenceofthefreshwaterdatabaseeither effects’)cansimplybejustifiedasfollows:(i)inloticenvi-
at tribe, clade or at lineage-level decreased (Fig.5B; ronments, bacterioplankton is transported primarily pas-
labelled ‘non-typical freshwater taxa’), suggesting that sively, (ii) the large contact zone of small headwaters
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd,
EnvironmentalMicrobiology,17,4994–5007
Riverbacterioplanktondiversity 5001
(large surface-area-to-volume ratio) with the surrounding time for the Danube River of 32 days (Velimirov etal.,
environment (soil and groundwater) facilitates the contri- 2011),thisprovidessufficienttimefor‘successful’species
bution of allochthonous bacteria to the river community sorting.
(Crump etal., 2007; 2012; Besemer etal., 2012), (iii) IncontrasttoCrumpandcolleagues(2012),Readand
these source environments of inoculation (soils and colleagues(2015)putlittleemphasisonthe‘masseffects’
groundwater) harbour a much higher diversity than of the riparian zone and, instead, explained the down-
aquatic communities (e.g. Crump etal., 2012), and (iv) stream shift in the bacterial communities along the River
thesenewlyintroducedallochthonousbacteriashouldbe Thames as the result of ecological succession. They
at least temporarily capable of proliferating in their new arguedthatsamplingsitesnotonlyrepresentedaspatial
lotic environment, making them constitutive members of distributionbutalsoatimeserieswheredownstreamsites
the community. Overall, this process of allochthonous withlongerwaterresidencetimescontainolderriverwater
input can be described by the so-called ‘mass-effect’, and a planktonic community that is in the later stages of
where dispersal of organisms exceeds the rate of local ecologicalsuccession.However,aconceptlikeecological
extinction (Leibold etal., 2004; Crump etal., 2007; succession might not be the most suited to describe
2012). bacterioplankton diversity patterns in the Danube basin
Flowing downriver, with increasing river width and which is much larger, and the residence times are about
decreasing‘riparianinfluence’,weproposethat‘species- fivetimesaslongwhencomparedwiththeRiverThames.
sorting’ progressively prevails over ‘mass effects’ in First, we could not observe an initial active growth and
shaping the bacterioplankton composition. This is sup- propagation of new pioneer species (often r-strategists)
ported by the increase of the core communities’ relative as it would be expected during ecological succession,
abundance in both size fractions (Fig.2) as well as the but solely a decrease in proportions of Bacteroidetes-
rapidly decreasing number of first-time occurrences of affiliated cells. Second, the data clearly showed a
OTUsfromupstreamtodownstream(Fig.S6).Theasso- decreaseinalphadiversity,bothevennessandrichness,
ciated progressive rise of few and more competitive downriver. Assuming a dominance of fast-growing
taxa is further supported by the observed simultaneous r-strategists in the highly dynamic communities of the
decrease in evenness together with bacterial richness in upstream reaches, one would expect low evenness in
both size fractions downriver (Fig.3A and B). Finally, these reaches. Moreover, highest production rates, as
the decrease of cell volumes along the Danube River anticipated for a dominance of r-strategists, were not
(Velimirov etal., 2011) as well as a rise of typical fresh- observedintheupstreamreaches,butthemiddlesection
water bacteria (Fig.5A and B), representing small cells oftheDanubeRiver[betweenriverkilometre(rkm)1632-
with oligotrophic lifestyles (Salcher etal., 2011; Garcia 1071;Velimirovetal.,2011].Besides,elevatedproduction
etal.,2013),providesfurtherevidencefortheincreasing ratesintheuppermostsectionwerelargelyassignableto
importance of ‘species-sorting’. The increasing resem- particle-associated bacteria (Velimirov etal., 2011), indi-
blance with lake communities (Fig.S5B and C) further cating high-quality particulate substrate originating from
corroboratestheideathatlakeandriverbacterioplankton theriparianzone.
resembleeachother. Insteadofviewingthecommunityassemblyinheadwa-
The interplay between ‘species sorting’ and ‘mass tersasthecolonizationofalifelessarea(primarysucces-
effects’ is fundamental to the meta-community concept sion) or the development following a disturbance of an
(Leibold etal., 2004) and has previously been used to established community (secondary succession), we
explain the decreasing diversity along a path from simply regard streams as part of the hydrological cycle
upslope soils via headwater streams to a final lake suchassoilwatersandinterstitialgroundwaters,inwhich
(Crump etal., 2012). The latter study suggested advec- bacteria are transported passively. Thus, we conclude
tion of soil bacteria to strongly influence the receiving that dispersal-based concepts are more appropriate for
waterbodiesduetothe‘masseffect’ofdispersingorgan- describing the community assembly at the transition
ismsexceedingtherateof‘speciessorting’(Crumpetal., between the different compartments. The Danube data
2012). For large river networks, however, this raises the suggest that the ‘mass effects’ of soil and groundwater
question of when ‘species sorting’ will exceed ‘mass bacteriaacrossthestreambedcontactzoneareimportant
effects’. In contrast to ‘mass effects’, ‘species sorting’ processes, as reflected in the decreasing proportion of
presumes that bacterial growth rates have to be shorter soil and groundwater-associated bacteria.This decrease
than the residence time in the stream. Assuming a of the effect of the streambed contact zone can be illus-
maximum bacterial growth rate of about 1d−1 (as calcu- tratedbycomparingtheuppermostandlowestreachesof
lated for the bulk communities along the Danube River thesurvey,usingtheratioofwettedperimeterandcross-
withamaximumandmeanof0.92d−1and0.18d−1,respec- sectionalareaasameasureofriverbedcontactzone.At
tively)andconsideringanestimatedin-streamresidence the Upper Danube near Ulm, Germany, an average dis-
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd,
EnvironmentalMicrobiology,17,4994–5007
5002 D.Savioetal.
charge of 40m3s−1 translates into a ratio of about Geomorphological parameters including ‘catchment area’,
0.83m−1whileattheLowerDanubenearReni,Romania, ‘meandendriticstreamlength’and,forcomparison,‘cumula-
anaveragedischargeof6300m3s−1translatesintoaratio tive dendritic distance upstream’ were calculated based on
of 0.056m−1. This means that within the sampled reach, data from the Catchment Characterisation and Modelling
(CCM)RiverandCATCHMENTDATABASE,version2.1(DeJager
the effect of the contact zone decreases by more than
and Vogt, 2010). The ‘mean dendritic stream length’ was
10-folddownriver.
calculatedbyfirstidentifyingthestreampathsfromallsprings
Takentogether,thefindingsofthispaperandtheexisting inthecatchmentareaupstreamasamplingsitetothatsam-
literature(Besemeretal.,2012;2013;Crumpetal.,2012; plingsite.Thelengthsofthesepathswerethanaveraged.This
Staley etal., 2013) suggest that the diversity of impliesthatthesharedreacheswerecountedmultipletimes
bacterioplankton decreases from headwaters to the river consistentwiththemovementofwaterdropsintheriverbasin.
Thisparametergivestheaverageflowdistance(assumingthe
mouth due to the decreasing importance of the ‘riparian
springdischargesarerandomlydistributedinthecatchment)
influence’. This is consistent with the important role the
and therefore the average residence time in the stream
RCCassignstotheriparianzoneaswellastothephysical
(assumingconstantflowvelocities).Incontrast,the‘cumula-
driverssuchasriverflowandwettedperimeter.Moreover, tive dendritic distance upstream’ was calculated as
byreferringtodissolvedorganicmatter(DOM)intermsof the sum of the blue lines on the map (counting the lengths
quality, proposing highest DOM diversity in headwaters belowconfluencesonlyonce)whichgivesageometricparam-
andadownstreamexportofmorerefractorycompounds, eter indicative of the drainage density, but not of residence
times.
the RCC also allows us to incorporate a proposed DOM
Additionally,asameasureoftheriverbedcontactzone,the
(‘food’)-based‘speciessorting’,leadingtoanoligotrophic
ratioofwettedperimeterandcross-sectionalareawascalcu-
freshwaterbacteria-dominatedcommunitywithincreasing
lated for the long-term average discharge. Both the wetted
water residence time. Analogously, an increase in the perimeter and the areas are obtained from bathymetric
abundanceofafew,morecompetitivespeciesalongthe surveys.Theproductoftheareaandtheflowvelocitygivesthe
river is implied in the RCC for macroorganisms from discharge.Here,typicalvaluesoftheratiowerechosenfora
medium-sized reaches towards river mouths (Vannote shortreachupstreamthesamplingsites.
etal.,1980).However,tofurtherdevelopandincorporate
bacterioplankton into the RCC, future studies should
Studysitesandsamplecollection
addresstheroleofDOMqualitychangesasafunctionof
catchmentcharacteristicsaswellasseasonalchangesin Samples were collected within the frame of the JDS 2
project in 2007. The overall purpose of the Joint Danube
physicalandchemicalfeaturesofriverinesystems.
Surveys is to produce a comprehensive evaluation of the
In summary, the data from the Danube survey along a
chemical and ecological status of the entire Danube River
2600kmrivercontinuumindicatedthatbacterialrichness
onthebasisoftheEuropeanUnionWaterFrameworkDirec-
and evenness gradually declined downriver in both the tive (Liska etal., 2008). During sampling from 15August to
free-livingandparticle-associatedbacterialcommunities, 26September2007,75sitesweresampledalongthemain-
resulting in an increase in relative abundance of typical stream of the Danube River along its navigable way from
freshwatertaxadownstream.Thedecreasinginfluenceof riverkilometre(rkm)2600nearUlm(DE)totherivermouth
atrkm0(Kirschneretal.,2009)asshowninFig.1.Inaddi-
soil and groundwater bacteria downstream suggests the
tion, 21 samples from the Danube’s major tributaries and
RCCasavalidinterpretiveframeworkwhichstipulatesa
branches were included. At the most upstream sites, the
continuous gradient of physical conditions that elicit a
Danube River is representative of a typical stream of the
seriesofbiologicalresponses,resultinginconsistentpat- rhithron and characterized by its tributaries Iller, Lech and
ternsofcommunitystructureandfunctionalongtheriver Isar(KavkaandPoetsch,2002).Thetriptook43dayswhich
system. is a similar time period as the travel time of water in the
Danubeoverthisreach(seeVelimirovetal.,2011).Samples
were collected with sterile 1L glass flasks from a water
Experimentalprocedures
depth of approximately 30cm. Glass flasks were sterilized
Supportingdata by rinsing with 0.5% HNO3 and autoclaving them. For
deoxyribonucleic acid (DNA) extraction of the particle-
WithintheframeofthesecondJointDanubeSurvey(JDS2),a associated bacterioplankton depending on the biomass
wide range of chemical and biological parameters was col- concentration, 120–300ml river water was filtered through
lected(Liskaetal.,2008).Alldata,samplingmethodsaswell 3.0μm pore-sized polycarbonate filters (Cyclopore,
as analytical methods, are made publicly available via the Whatman,Germany)byvacuumfiltration.Thefiltrate,which
officialwebsiteoftheInternationalCommissionfortheProtec- represented the bacterioplankton fraction smaller than
tion of the Danube River (http://www.icpdr.org/wq-db/) and 3.0μm (later referred to as ‘free-living’ bacterioplankton),
datausedinthisstudyisprovidedinTableS3..Selecteddata was collected in a sterile glass bottle and subsequently fil-
from Joint Danube Survey (JDS) 1 and 2 were published tered through 0.2μm pore-sized polycarbonate filters
previouslyinseveralstudies(Kirschneretal.,2009;Janauer (Cyclopore, Whatman, Germany).The filters were stored at
etal.,2010;Velimirovetal.,2011;VonderOheetal.,2011). −80°C until DNAextraction.
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd,
EnvironmentalMicrobiology,17,4994–5007
Riverbacterioplanktondiversity 5003
DNAextractionandquantificationofbacterialDNA cation using the Agencourt AMPure XP purification system
usingquantitativepolymerasechainreaction (BeckmanCoulter,Danvers,MA,USA)andquantificationof
amplicon concentration using the Quant-iT PicoGreen
GenomicDNAwasextractedusingaslightlymodifiedproto- dsDNA Assay Kit (Life Technologies Corporation, USA).
col of a previously published phenol-chloroform, bead- Finally,atotalof137samplesincludingfivenegativecontrols
beating procedure (Griffiths etal., 2000) using isopropanol resultedinfourpoolsforsequencing.
instead of polyethylene glycol for DNA precipitation. Total
DNA concentration was assessed applying the Quant-iT
Illuminasequencing
PicoGreendsDNAAssayKit(LifeTechnologiesCorporation,
USA),and16SrRNAgeneswerequantifiedusingBacteria-
ThesequencingwasperformedonanIlluminaMiSeqatthe
specific quantitative polymerase chain reaction (qPCR).
SciLifeLabSNP/SEQsequencingfacilityhostedbyUppsala
QuantitativePCRreactionscontained2.5μlof1:4and1:16
University. For each pool, the library preparation was per-
diluted DNA extract as the template, 0.2μM of primers 8F
formedseparatelyfollowingtheTruSeqSamplePreparation
and338(Franketal.,2007;Fiereretal.,2008)targetingthe
Kit V2 protocol (EUC 15026489 Rev C, Illumina) with the
V1-V2 region of most bacterial 16S rRNA genes and iQ
exceptionoftheinitialfragmentationandsizeselectionpro-
SYBR Green Supermix (Bio-Rad Laboratories, Hercules,
cedures.Thisinvolvesthebindingofthestandardsequenc-
USA). All primer information are available in TableS2. The
ing adapters in combination with separate Illumina-specific
ratios of measured 16S rRNA gene copy numbers in the
multiplex identifier (MID) bar codes that enables the combi-
different sample dilutions that deviated markedly from one
nationofdifferentpoolsonthesamesequencingrun(Sinclair
after multiplication with the respective dilution factor were
etal., 2015). After pooling, random PhiX DNA was added
interpretedasanindicatorforPCRinhibition.
(5%)toprovidecalibrationandhelpwiththeclustergenera-
tionontheMiSeq’sflowcell.
Preparationof16SrRNAgeneampliconlibraries
For the preparation of amplicon libraries, 16S rRNA genes 16SrRNAgeneamplicondataanalysis
wereamplifiedandbarcodedinatwo-stepPCRprocedureto
reduce PCR bias that is introduced by long primers and The sequence data were processed as outlined by Sinclair
sequencing adaptor overhangs (Berry etal., 2011). We fol- andcolleagues(2015).Aftersequencingthelibrariesof16S
lowed the protocol as described by Sinclair and colleagues rRNA gene amplicons, the read pairs were de-multiplexed
(2015).Inshort,16SrRNAgenefragmentsofmostbacteria and joined using the PANDASEQ software v2.4 (Masella
were amplified by applying primers Bakt_341F and etal., 2012). Next, sequence reads (further referred to as
Bakt_805R (Herlemann etal., 2011; TableS2) targeting the ‘reads’)thatdidnotbearthecorrectprimersequencesatthe
V3-V4variableregions.In25μlreactionscontaining0.5μM startandendoftheirsequenceswerediscarded.Readswere
primer Bakt_341F and Bakt_805R, 0.2μM dNTPs thenfilteredbasedontheirPHREDscores.Chimeraremoval
(Invitrogen),0.5UQ5HFDNApolymeraseandtheprovided and OTU clustering at 3% sequence dissimilarity was per-
buffer (New England Biolabs, USA), genomic DNA was formed by pooling all reads from all samples together and
amplifiedinduplicatein20cycles.Touseequalamountsof applying the UPARSE algorithm v7.0.1001 (Edgar, 2013).
bacterial template DNA for increased comparability and Here, any OTU containing less than two reads was dis-
reduction of PCR bias, the final volume of environmental carded. Each OTU was subsequently taxonomically classi-
DNAextractusedforeachsamplewascalculatedbasedon fied by operating a similarity search against the SILVAMOD
16SrRNAgenecopyconcentrationintherespectivesample database and employing the CREST assignment algorithm
determined earlier by qPCR (see above). For 105 samples, (Lanzén etal., 2012). Plastid, mitochondrial and archaeal
the self-defined optimum volume of environmental DNA OTUswereremoved.Inaddition,OTUswerealsotaxonomi-
extract corresponding to 6.4×105 16S rRNA genes was cally annotated against the freshwater database (Newton
spiked into the first step PCR reactions; however, for 27 etal.,2011)usingthesamemethod.Ifnecessary,OTUrar-
samples, lower concentrations were used due to limited efying for the purpose of standardizing sequence numbers
amounts of bacterial genomic DNA or PCR inhibition wasperformedusingthe‘rrarefy’-functionimplementedinthe
detected by quantitative PCR (see above). These 132 Rpackagevegan(Oksanenetal.,2013).Foralphadiversity
samples included eight biological replicates. Prior to the analysis(Chao1richnessestimatorandPielou’sevenness),
analysis,weremovedfoursamplesduetotheirextremelylow werarefieddownto7000readspersample.Thiswasbased
genomic DNA concentrations and 16S rRNA gene copy ononestudyrevealingthatforwatersamples,asequencing
numbers.DuplicatesofPCRproductswerepooled,dilutedto depth of 5000 16S rRNA gene reads per sample captured
1:100 and used as templates in the subsequent barcoding morethan80%ofthetrendsinChao1richnessandPielou’s
PCR. In this PCR, diluted 16S rRNAgene amplicons were evenness(Lundinetal.,2012).Furthermore,thisstudycould
amplified using 50 primer pairs with unique barcode pairs show that for water samples, 1000 reads per sample
(Sinclair etal., 2015; TableS2). The barcoding PCRs for explained to 90% the trends in beta diversity (Bray–Curtis
mostsampleswereconductedintriplicatesanalogoustothe dissimilarityindex).Byrarefyingdownto2347,whichwasthe
firstPCR(n=100).Theremaining32samplesthathadweak readnumberofthesamplewiththelowestreads,allsamples
bands in first step PCR due to low genomic template DNA could be included in the beta-diversity analysis. Diversity
concentrationsorhighsampledilutionwereamplifiedin6–9 measures,statisticalanalysesandplotgenerationwerecon-
replicates to increase amplicon DNA yield. Barcoded PCR ducted in R (R Core Team, 2013) and using python scripts.
amplicons were pooled in an equimolar fashion after purifi- The habitat index for the top 5000 OTUs was determined
©2015TheAuthors.EnvironmentalMicrobiologypublishedbySocietyforAppliedMicrobiologyandJohnWiley&SonsLtd,
EnvironmentalMicrobiology,17,4994–5007