Coupling Spatiotemporal Community Assembly Processes to Changes in Microbial Metabolism

13 Pages·2016·3.46 MB·English

by  Emily B. Graham

fmicb-07-01949 December14,2016 Time:15:22 #1
ORIGINALRESEARCH
published:16December2016
doi:10.3389/fmicb.2016.01949
Coupling Spatiotemporal Community
Assembly Processes to Changes in
Microbial Metabolism
EmilyB.Graham*,AlexR.Crump,CharlesT.Resch,SarahFansler,EvanArntzen,
DavidW.Kennedy,JimK.FredricksonandJamesC.Stegen
BiologicalSciencesDivision,PacificNorthwestNationalLaboratory,Richland,WA,USA
Community assembly processes generate shifts in species abundances that influence
ecosystemcyclingofcarbonandnutrients,yetourunderstandingofassemblyremains
largely separate from ecosystem-level functioning. Here, we investigate relationships
between assembly and changes in microbial metabolism across space and time
in hyporheic microbial communities. We pair sampling of two habitat types (i.e.,
attached and planktonic) through seasonal and sub-hourly hydrologic fluctuation with
nullmodelingandtemporallyexplicitmultivariatestatistics.Wedemonstratethatmultiple
selectivepressures—imposedbysedimentandporewaterphysicochemistry—integrate
togeneratechangesinmicrobialcommunitycompositionatdistincttimescalesamong
Editedby:
AlisonBuchan, habitat types. These changes in composition are reflective of contrasting associations
UniversityofTennessee,USA ofBetaproteobacteriaandThaumarchaeotawithecologicalselectionandwithseasonal
Reviewedby: changesinmicrobialmetabolism.Wepresentaconceptualmodelbasedonourresults
SonjaKristineFagervold,
in which metabolism increases when oscillating selective pressures oppose temporally
UniversitéPierreetMarie
Curie–UPMC(Paris6),France stable selective pressures. Our conceptual model is pertinent to both macrobial and
AlexandreSoaresRosado,
microbial systems experiencing multiple selective pressures and presents an avenue
FederalUniversityofRiodeJaneiro,
Brazil for assimilating community assembly processes into predictions of ecosystem-level
*Correspondence: functioning.
EmilyB.Graham
[email protected] Keywords:niche,selection,dispersal,microbialcommunitystructure,aerobicrespiration,ammoniaoxidation,
hyporheic,Hanford
Specialtysection:
Thisarticlewassubmittedto
INTRODUCTION
AquaticMicrobiology,
asectionofthejournal
FrontiersinMicrobiology The collective effects of community assembly processes (e.g., dispersal, drift, and selection) on
Received:30August2016 microbial metabolism of carbon and nutrients in the environment are poorly understood, and
Accepted:21November2016 theyconstituteakeyknowledgegapinprocess-basedmodelingofbiogeochemicalcycles.Selection
Published:16December2016 and dispersal both have the potential to impact rates of microbial metabolism. For example,
Citation: selectioncanenhancemetabolismviaspeciessortingmechanismsthatoptimizethemicrobiome
Graham EB,Crump AR,Resch CT, for a given environment (Van der Gucht et al., 2007; Lindström and Langenheder, 2012), while
Fansler S,Arntzen E,Kennedy DW, dispersal limitation can inhibit immigration of metabolic diversity, and in some cases, lead to a
Fredrickson JKandStegen JC maladaptedandpoorlyfunctioningcommunity(Telfordetal.,2006;LindströmandÖstman,2011;
(2016)CouplingSpatiotemporal
Hansonetal.,2012;Peresetal.,2016).Theextenttowhichcommunityassemblyprocessesregulate
CommunityAssemblyProcesses
metabolismiscontingentonmyriadspatiotemporaldynamicsincludingthegeographicdistance
toChangesinMicrobialMetabolism.
separating communities, the rate of environmental change, and historical abiotic conditions
Front.Microbiol.7:1949.
doi:10.3389/fmicb.2016.01949 (Fukami et al., 2010; Graham et al., 2014; Nemergut et al., 2014; Hawkes and Keitt, 2015;
FrontiersinMicrobiology|www.frontiersin.org 1 December2016|Volume7|Article1949
fmicb-07-01949 December14,2016 Time:15:22 #2
Grahametal. CommunityAssemblyandMicrobialMetabolism
Graham et al., 2016). Yet, we lack a conceptual basis for groundwater. The groundwater-surface water mixing within
how multiple community assembly processes jointly influence the hyporheic zone can result in blending of complementary
microbialmetabolism(Prosseretal.,2007;Gonzalezetal.,2012; resources and, in turn, elevated rates of microbial metabolism
Shadeetal.,2013;Grahametal.,2016). relative to other systems (Hancock et al., 2005; Boulton et al.,
Community assembly processes act through space and 2010).
time to impact community membership, which then impacts We employ null modeling in conjunction with temporally
microbial metabolism (Vellend, 2010; Nemergut et al., 2013). explicitmultivariatestatisticstocharacterizeassemblyprocesses
For instance, communities experiencing a history of strong driving shifts in microbial communities and microbial
and consistent selection may contain taxa that are well- metabolism in the Columbia River hyporheic zone. We
adapted to their environment and exhibit high metabolic examine two co-occurring, yet ecologically distinct, habitat
rates. Alternatively, intense, unyielding selection may eliminate types—attachedandplanktoniccommunities.Further,weextend
microbialtaxathatmetabolizescarceresourcepoolsandimpair theanalysisofcommunityassemblyprocessestoindividualtaxa
community metabolic functioning (Grime, 1998; Loreau and thatlikelycontributetoobservedshiftsinmicrobialmetabolism.
Hector, 2001; Knelman and Nemergut, 2014). In this case, Ourresultsculminateinabroadlyapplicableconceptualmodel
more diverse communities (e.g., those experiencing higher couplingchangesinselectiveenvironments,traitabundance,and
rates of dispersal or counteracting selective pressures) would ecosystem-levelfunctioningthroughtime.
be expected to exhibit higher and more consistent rates of
metabolism than those structured by one dominant selective
MATERIALS AND METHODS
pressure.Dispersallimitationcaninhibittheabilityoforganisms
to reach their optimal environment, resulting in lower rates
Study Design
of community metabolism, while high rates of dispersal may
either reduce or enhance microbial metabolism, respectively, This study was conducted in Hanford Reach of the Columbia
by allowing for immigration of maladapted organisms or by River adjacent to the Hanford 300A (approximately 46◦22(cid:48)
increasing biodiversity (Hooper et al., 2012; Nemergut et al., 15.80(cid:48)(cid:48)N,119◦16(cid:48)31.52(cid:48)(cid:48)W)ineasternWashington,asdescribed
2014). elsewhere (Slater et al., 2010; Zachara et al., 2013; Stegen
Additionally, individual taxa are differentially impacted by et al., 2016), from March to November 2014. The hyporheic
community assembly processes. Only some taxa contain traits zone of the Columbia River experiences geographic variation
that are under selection in given environmental conditions, in groundwater-surface water mixing, porewater geochemistry,
and changes in the environment may affect some taxa andmicrobialcommunitycompositiononsub-hourlytoannual
to a greater extent than others (Poff, 1997; Lebrija-Trejos timescales (Peterson and Connelly, 2004; Arntzen et al.,
et al., 2010; Knelman and Nemergut, 2014; Krause et al., 2006; Slater et al., 2010; Lin et al., 2012; Stegen et al.,
2014). For example, a change in a certain nutrient should 2012, 2016; Zachara et al., 2013). Accordingly, the Hanford
have a larger influence on taxa that directly metabolize it Reach of the Columbia River embodies a model system to
than those that utilize it as a secondary resource. Similarly, facilitate the integration of community ecology and microbial
traits that facilitate dispersal (e.g., spore formation) are metabolism.
preferentially contained within certain taxa (Martiny et al., We monitored physicochemical conditions for three
2006; Tremlová and Münzbergová, 2007) such that changes in hydrologically-connected geographic zones (nearshore,
abiotic transport mechanisms should have contrasting effects inland, river) via aqueous sampling (Table S1). The inland
on taxa with or without traits that facilitate dispersal. Thus, environment is characterized by an unconfined aquifer within
taxon-specific relationships between assembly processes and the Hanford formation and more recent illuvial deposits,
microbial communities may inform our understanding of and it maintains a distinct hydrologic environment with
the ecological processes influencing changes in environmental stable temperatures (∼15◦C) and high concentrations of
microbiomes beyond trends we observe at the community- anions and inorganic carbon relative to the river. River water
level. contains high concentrations of organic material and low
Environmentaltransitionzonespresentauniqueopportunity concentrations of ions with seasonally variable temperatures.
for examining interactions between microbial metabolism The waters from these discrete hydrologic environments
and both long- and short-term assembly processes, as they experience dynamic mixing in a nearshore hyporheic zone
experienceextremespatiotemporalvariationinphysicochemical that is regulated by fluctuations in river stage; we focus on
characteristics and microbial community composition across ecologicaldynamicswithinthiszone.Tomonitorgroundwater-
tractable spatial and temporal scales. Here, we leverage surface water mixing across space and time, we utilize Cl−
inherent variation in hydrology, habitat heterogeneity, and as a conservative tracer for groundwater contributions to
aerobic respiration in a zone of subsurface groundwater- hyporheic porewater chemistry as employed by Stegen et al.
surface water mixing (hereafter termed “hyporheic zone”) (2016).
to examine the interplay of community assembly processes Detailed sampling and analytical methods are in the
and microbial metabolism through time. Hyporheic zones SupplementaryMaterial.Attachedandplanktoniccommunities
are subsurface regions below and adjacent to rivers and wereobtainedfromdeployedcolonizationsubstrateandaqueous
streams that experience mixing between surface water and samplesinthehyporheiczone.Samplestoconstructtheregional
FrontiersinMicrobiology|www.frontiersin.org 2 December2016|Volume7|Article1949
fmicb-07-01949 December14,2016 Time:15:22 #3
Grahametal. CommunityAssemblyandMicrobialMetabolism
species pool for null models were simultaneously obtained standarddeviationsfromthemeanofthenulldistribution—and
at three inland wells and at one location in the Columbia alpha values—|RC | = 0.95 reflects significance at the 0.05
bray
River (n = 0−4 at each sampling event, a full breakdown level. Inferences from both βNTI and RC have previously
bray
of sample size is listed in Table S2). These samples were beenshowntoberobust(Dini-Andreoteetal.,2015;Stegenetal.,
collectedatthree-weekintervalsfromMarchthroughNovember 2015).
2014, with the first planktonic samples collected in March and
Statistical Methods
the first attached samples collected after a 6-week incubation
period from piezometers installed to 1.2 m depth near the Regressions and one-sided Mann Whitney U tests were
riverbed. Samples collected at each time point were assumed conducted using the base statistics package in R. Variation in
to represent sub-hourly variation in hydrologic conditions, as community composition was assessed with PERMANOVA in
hydrologyfluctuatesatsub-hourlyrateswithinasingledayinour QIIME(Caporaso et al.,2010). We fitporewater characteristics
system (Arntzen et al., 2006), whereas samples collected across to NMDS plots of Bray-Curtis dissimilarities with and without
the full sampling period were assumed to represent seasonal stratifyingbytimewithinattachedandplanktoniccommunities
variation. using the ‘vegan’ package in R (999 permutations, Oksanen
Aqueous samples were obtained by pumping water from et al., 2013). Further details are available in the Supplementary
piezometers adjacent to colonization substrates and used Material.
to derive physicochemical conditions as well as to sample Because we observed large seasonal differences in species
planktoniccommunities.Attachedmicrobialcommunitieswere richness in both attached and planktonic communities, we
sampled by deploying mesh stainless steel incubators of performed similarity percentage analysis (SIMPER) in ‘vegan’
locally sourced colonization substrate in piezometers within to identify individual species driving community dissimilarity
one meter of piezometers from which aqueous samples (Clarke, 1993). SIMPER was conducted across time periods of
were obtained. All incubators were deployed 6 weeks prior high and low species richness within attached and planktonic
to removal. For each sample, we collected the following communitiesseparately(i.e.,communitiesfromsamplingpoints
data according to procedures outlined in the Supplementary with high vs. low richness in each habitat type). We grouped
Material: physicochemical characteristics (aqueous samples communities by sampling date (determined by the average
only), microbial metabolism (aerobic metabolism assayed with speciesrichnessincommunitiesateachtimepoint)tocontrolfor
Resazurin) per unit active biomass (ATP) defined by Raz:ATP seasonaleffects.Further,weusedSIMPERtoidentifyspeciesthat
(attachedsamplesonly),and16SrRNAampliconsequences(all differentiated attached and planktonic communities regardless
samples). Sequences are publically available at doi: 10.6084/m9. of differences in richness (i.e., all attached communities vs. all
figshare.4264148. planktoniccommunities).
Weextractedtaxonomicgroupsoforganismsattheclass-level
Null Modeling Approach containing at least one species identified as having a significant
We implemented null modeling methodology developed by impact on community composition by SIMPER (P < 0.05)
Stegen et al. (2013, 2015) using R software1 to disentangle for subsequent analyses. Organisms were grouped at the class-
community assembly processes (Supplementary Material). level to provide sufficient statistical power for analysis. Mantel
The approach uses pairwise phylogenetic distances between tests were used to compare the average relative abundance
communities,calculatedusingthemean-nearest-taxon-distance of taxonomic groups identified by SIMPER across samples to
(βMNTD) metric (Webb et al., 2008; Fine and Kembel, 2011), associated βNTI and RCbray values (‘vegan’, 999 permutations).
to infer the strength of selection. Communities were evaluated Finally, we compared dissimilarity in species richness between
for significantly less turnover than expected (βNTI < −2, sampleswithinandacrossattachedandplanktoniccommunities
homogeneous selection) or more turnover than expected toβNTIandfurther,if−2<βNTI<2,toRCbray usingMantel
(βNTI>2,variableselection)bycomparingobservedβMNTD tests to infer community assembly processes generating species
valuestothemeanofanulldistributionofβMNTDvalues—and turnoverbetweensamples.
normalizing by its standard deviation—to yield βNTI (Stegen
et al., 2012). Pairwise community comparisons that did not
RESULTS
deviate from the null βMNTD distribution were evaluated for
theinfluencesofdispersallimitationandhomogenizingdispersal
Hydrologic and Community Shifts
by calculating the Raup-Crick metric extended to account for
Through Time
species relative abundances (RC ), as per Stegen et al. (2013,
bray
2015).ObservedBray-Curtisdissimilaritieswerecomparedtothe Weobserveddistincttemporaltrendsinmicrobialcommunities
nulldistributiontoderiveRC .RC values>0.95,>−0.95 and in groundwater-surface water mixing, characterized by an
bray bray
and < 0.95, or < −0.95 were assumed to indicate dispersal abruptincreaseinCl− concentration(Figure1A)anddecrease
limitation, no dominant assembly process, or homogenizing in NPOC (Figure 1B) associated with a seasonal shift in water
dispersal, respectively. Significance levels for βNTI and RC stage (Figure S1 and Table S1). Temperature peaked during
bray
are based on standard deviations—|βNTI| = 2 denotes two August and followed a smooth temporal trend (Figure 1C).
The composition of both attached (PERMANOVA, R2 = 0.44,
1http://cran.r-project.org/ P=0.001)andplanktonic(PERMANOVA,R2=0.44,P=0.001)
FrontiersinMicrobiology|www.frontiersin.org 3 December2016|Volume7|Article1949
fmicb-07-01949 December14,2016 Time:15:22 #4
Grahametal. CommunityAssemblyandMicrobialMetabolism
FIGURE1|Changesin(A)chlorideconcentration,(B)NPOCconcentration,(C)temperature,and(D)speciesrichnessacrossoursamplingperiodaredepictedin
Figure1.ChlorideandNPOCconcentrationshowabruptshiftsbeginningatourJuly22samplingpoint(verticaldashedlines).P-valuesin(A)and(B)denote
one-sidedMann−WhitneyUtestresultsofsamplestakenbeforeversusonorafterJuly22,whiletrendsthroughtimein(A)and(B)aredisplayedusinglocally
weightedscatterplotsmoothing(LOWESS).Quadraticpolynomialswerefittotemperatureandspeciesrichnessdataandplottedin(C,D).Trianglesin(D)represent
planktoniccommunities;X’srepresentattachedcommunities.
communities also changed across our sampling period, but Spatiotemporal Assembly Processes
attached and planktonic communities remained taxonomically Dissimilarityinmicrobialcommunitiesindicatedapossiblerole
distinct through time (PERMANOVA, R2 = 0.19, P = 0.001). for different assembly processes governing species composition
MaintaxaineachenvironmentarepresentedinTableS3. among (i.e., within attached and planktonic separately) and
Speciesrichnessinbothattachedandplanktoniccommunities across (i.e., attached vs. planktonic) habitats. We investigated
mirroredtemperaturetrends,withthehighestnumberofspecies assembly processes across differences in richness to identify
observed during the warmest summer months (Figure 1D). processes that impacted species addition to each habitat, and
Richness was more tightly correlated with temperature (Figure wefoundthatassemblyvariedbetweenhabitats(Figures2B,C).
S2, regression, attached: R2 = 0.25, P = 0.001, planktonic: βNTI was positively correlated with differences in species
R2 =0.22,P=0.002)thanCl− (regression,attached:P=0.01, richness in planktonic samples (Mantel, P = 0.001, r = 0.41,
R2=0.14,planktonic:P>0.05)andNPOC(regression,attached: Figure 2C), with weaker correlations in attached communities
P=0.04,R2 =0.10,planktonic:P>0.05).Finally,withinboth (Mantel, P = 0.02, r = 0.24, Figure 2B) and between attached
environments, community dissimiliarity increased in concert and planktonic communities (Mantel, P = 0.006, r = −0.26,
withdifferencesinspeciesrichness(Figure2A). Figure2D).
FrontiersinMicrobiology|www.frontiersin.org 4 December2016|Volume7|Article1949
fmicb-07-01949 December14,2016 Time:15:22 #5
Grahametal. CommunityAssemblyandMicrobialMetabolism
FIGURE2|(A)Bray−Curtisdissimilaritywithineachsamplingtimepointincreasedasmeanspeciesrichnessincreasedinattached(X’s)andplanktonic(triangles)
communities.Inaddition,βNTIvaluesacrossdifferencesinspeciesrichnessareshownfor(B)attached,(C)planktonic,and(D)attachedvs.planktonic
communities.HorizontallinesatβNTI=−2andβNTI=2denotethresholdsforassemblyprocesses.βNTIvalueslessthan−2suggestassemblyisgovernedby
homogeneousselection,whilevaluesgreaterthan2suggestassemblyisgovernedbyvariableselection.Stochasticassemblyprocesses(dispersallimitation,
homogenizingdispersal)andundominatedassemblyprocessesliebetweenβNTI−2and2.TheproportionofβNTIvalueswithineachcategoryarelistedastextin
(B−D).Alinearregressiontrendlineisdepictedin(C)withsignificanceassessedviaManteltest.
Because selection was a substantial driver of all microbial time to represent dynamics occurring within each sampling
communities, we further examined the impact of aqueous date and those occurring in analyses without stratification
physicochemistry on communities at both sub-hourly (within to represent dynamics across sampling dates. Planktonic
sampling date) and seasonal (across all sampling dates) communities were correlated to more environmental variables
timescales using stratified and unstratified NMDS analysis. than attached communities at a sub-hourly timescale (NMDS
Permutations in stratified NMDS are constrained within stratifiedbytime,Figures3A,B;TableS4).Conversely,attached
the specified grouping (in our case, sampling date), thus communities correlated more tightly with physicochemical
controlling for the grouping factor (see Oksanen et al., 2013). attributes over a seasonal timescale (unstratified NMDS,
Here, we infer significant relationships when stratifying by Figures3A,B).
FrontiersinMicrobiology|www.frontiersin.org 5 December2016|Volume7|Article1949
fmicb-07-01949 December14,2016 Time:15:22 #6
Grahametal. CommunityAssemblyandMicrobialMetabolism
Phylogenetic Variability in Assembly and increasedseasonallywithinattachedcommunities,aneffectthat
Associated Shifts in Metabolism correlated with day of year (Figure 5C) but not temperature,
NPOC concentration, or hydrology (regression: temperature
Similaritypercentageanalysisrevealedspeciesdrivingdifferences P = 0.10, NPOC P = 0.21, log(Cl−) P = 0.15). The relative
among attached and planktonic communities during periods
abundance of Thaumarchaeota and Betaproteobacteria in
of high versus low species richness (Table S5). We extracted
attached communities also correlated positively and negatively,
phylogenetic classes of organisms containing species identified
respectively, with Raz:ATP (Figure 5D) and exhibited
by SIMPER and examined relationships between their relative
contrasting responses to porewater physicochemical properties
abundances and βNTI and RC . All significant correlations
bray (TableS7).
with absolute r values greater than 0.30 are listed in Table S5.
Taxainplanktoniccommunitiesexhibitednorelationshipswith
βNTI (Table S6), but the mean relative abundance of many DISCUSSION
taxa, including Thaumarchaeota (positive, Figure 4A), a class
ofAcidobacteria(positive,Figure4B),Actinobacteria(negative, Our results show pronounced seasonal changes in hydrologic
Figure 4C), and Alphaproteobacteria (negative, Figure 4D), and microbial characteristics within the Columbia River
displayedcorrelationswithRCbray.Themeanrelativeabundance hyporheic zone, as well as variation in the importance of
ofParvarcheotaandaclassofcandidatephylaOP3(koll11)were selection and dispersal in structuring attached vs. planktonic
positivelycorrelatedwithβNTIderivedfromcomparisonsacross communities.Thesecommunityassemblyprocessesoperatedat
attached and planktonic communities (Figures 4E,F). Finally, distincttimescalesineachhabitatandwerealsoassociatedwith
withinattachedcommunities,βNTIwascorrelatedwiththemean changes in the relative abundance of certain taxa. In particular,
relative abundance of Thaumarchaeota (positive, Figure 4G) changes in selection exhibited contrasting relationships with
and Betaproteobacteria (negative, Figure 4H). No correlations putative heterotrophic and autotrophic taxa in attached
were observed between RCbray and taxa within attached communities. Further, changes in community-level microbial
communities or across attached-vs.-planktonic communities metabolism correlated with the abundance of these same
(TableS6). taxa in attached communities. Based on our findings, we
We also observed a seasonal increase in the abundance of present a conceptual model applicable within both macrobial
Thaumarchaeota in attached communities and a concomitant and microbial systems that links trait selection, organismal
decrease in Betaproteobacteria (Figure 5A) that corresponded fitness, and ecosystem-level functioning in habitats that are
with shifts in hydrology (Figure 5B). Oxygenated conditions characterized by opposing selective pressures across multiple
persisted throughout our sampling period, and Raz:ATP timescales.
FIGURE3|Non-metricmultidimensionalscaling(NMDS)analysiswasconductedonBray-Curtisdistanceswithin(A)planktonicand(B)attached
communities.ColorsdenoteseasonalshiftsincommunitystructurealongagradientfromMarch(red)toNovember(blue).Physicochemicalcharacteristicswerefitto
eachplotwith(bluearrows)andwithout(redarrows)stratifyingpermutationsbysamplingtimetoassessshort-andlong-termcommunityresponses,respectively,to
theaqueousenvironment.
FrontiersinMicrobiology|www.frontiersin.org 6 December2016|Volume7|Article1949
fmicb-07-01949 December14,2016 Time:15:22 #7
Grahametal. CommunityAssemblyandMicrobialMetabolism
FIGURE4|RelationshipsbetweenβNTIorRCbrayandthemeanabundance(acrosssamples)ofselectedtaxaidentifiedbySIMPERanalysisare
depictedinFigure4.(A−D)DemonstraterelationshipsofThaumarchaeota,Acidobacteria-6,Actinobacteria,andAlphaproteobacteria,respectively,versusRCbray
inplanktoniccommunities.HorizontallinesatβNTI=−2(homogeneousselection)andβNTI=2(variableselection)andverticallinesatRCbray=−0.95
(homogenizingdispersal)and0.95(dispersallimitation)denotethresholdsforassemblyprocesses.Trendlinesinallpanelswerederivedfromlinearregressionsand
significancewasassessedviaManteltest.(E,F)ShowrelationshipsofβNTIwithParvarchaeotaandkoll11inattachedvs.planktoniccommunities;while(G,H)
denoterelationshipsofβNTIwithThaumarchaeotaandBetaproteobacteriawithinattachedcommunities,respectively.
Microbial Responses to Environmental dispersal processes may play a greater role in structuring
Change planktonic vs. attached communities. In contrast, attached
communities always yielded large negative βNTI values
Hyporheic microbial community composition shifted in
suggesting that features of the physical sediment environment
conjunction with seasonal changes in organic carbon
may impose stronger selective pressures than aqueous
concentration, temperature, and groundwater-surface water
physicochemistry.
mixing conditions. These variables each explained some
Because we observed pronounced seasonal trends in species
variation in temporal community dissimilarity within
richness associated with changes in the environment, we
attached and planktonic communities, indicating potential
used null modeling to assess the extent to which selection
influences of selection (e.g., mediated by environmental
versus dispersal influenced community composition across
change) and dispersal (e.g., mediated by hydrologic transport)
changesinspeciesrichness.Homogenousselectionprevailedin
over microbial community composition (Table S8). Both
attachedcommunitiesasdifferencesinspeciesrichnessincreased
selection by the geochemical environment and dispersal
(Figure2B),indicatingstrongandconsistentselectivepressures
from local sediment communities have been demonstrated
imposed by a relatively stable environment across temporal
within the groundwater aquifer in our system (Stegen
changes in microbial diversity. Strong consistent selection
et al., 2012); however, the balance of these two processes
throughtimeinattachedcommunitiessuggeststhatthephysical
in structuring hyporheic microbial communities remains
substratemayinherentlycontainalimitednumberofecological
unclear.
niches—potentiallyrelatedtomineralogyorphysicalstructure—
Our results indicate that homogeneous selection (i.e.,
withslowchangesinavailablenichespace.Inthiscase,temporal
consistently imposed selection for a given set of traits,
βNTI < −2) was the dominant assembly process in attached increasesinrichnesswerelikelyduetotheadditionoftaxathat
were ecologically similar to existing taxa and occupied similar
communities, while planktonic communities were influenced
nichespace.
by a combination of homogeneous selection, variable selection
(i.e., selective pressures that change through space or time, In contrast, differences in planktonic species richness were
βNTI>2),andspatialprocesses(i.e.,dispersal,−2<βNTI<2 positively correlated with βNTI, with βNTI values supporting
and |RC | > 0.95). Because planktonic communities a role for more variable selection as differences in richness
bray
consistently displayed higher βNTI values in comparison increased(Figure2C).Inthiscase,increasesinspeciesrichness
to attached communities, often with −2 < βNTI < 2, werelikelyduetotheadditionoftaxaoccupyingnewlyavailable
FrontiersinMicrobiology|www.frontiersin.org 7 December2016|Volume7|Article1949
fmicb-07-01949 December14,2016 Time:15:22 #8
Grahametal. CommunityAssemblyandMicrobialMetabolism
FIGURE5|(A,B)ChangesinThaumarchaeotaandBetaproteobacteriaacrosschangesintime(A)andchlorideconcentration(B).Trendlinesin(A)denotelinear
(Betaproteobacteria)andquadratic(Thaumarchaeota)regressions.Theverticallineandstatisticsin(B)denoteone-sidedMann-WhitneyUtestresultsof
BetaproteobacteriaandThaumarchaeotawhenchlorideconcentrationsareaboveorbelowthemaximumCl−concentrationintheColumbiaRiver(1.83mg/L).
(C)Showsincreasesinaerobicrespirationnormalizedtoactivebiomass(Raz:ATP)throughtime.Finally,(D)showsrelationshipsofBetaproteobacteriaand
ThaumarchaeotawithRaz:ATP.Trendlinesandassociatedstatisticsin(C,D)werederivedwithlinearregressions.ThaumarchaeotaandBetaproteobacteriaare
shownasclosedsquaresandopentriangles,respectively,in(A,B,D)withtrendsforeachgroupshownwithasolid(Thaumarchaeota)ordashed
(Betaproteobacteria)line.
niche space generated by changes in the selective environment, communities may also be reflective of differing rates
yielding new taxa that were ecologically dissimilar to existing of organismal response to fluctuations in the hyporheic
taxa. environment.
Variation in assembly processes between attached and
planktonic communities may be due to inherent differences Timescales of Selection
among these environments, such as influences of mineralogy The timescales at which selection imposes constraints on
(Carson et al., 2007; Jorgensen et al., 2012), physical matrix microbialcommunitycompositionarepoorlyunderstood(Shade
composition (Vos et al., 2013; Breulmann et al., 2014), etal.,2012;Nemergutetal.,2013).Here,weprovidenewinsights
and/or relative rates of change in environment characteristics into these timescales within the hyporheic zone, showing that
(discussed below). Further, differences in assembly processes selection on planktonic communities operates at the timescale
and niche dynamics between the attached and planktonic of shifting porewater conditions (sub-hourly to seasonal),
FrontiersinMicrobiology|www.frontiersin.org 8 December2016|Volume7|Article1949
fmicb-07-01949 December14,2016 Time:15:22 #9
Grahametal. CommunityAssemblyandMicrobialMetabolism
while selection on attached communities operates primarily at relationships with RC have physiologies that may diminish
bray
seasonaltimescales(Figure3).Bothcommunitiesexperienceda dispersalability.Alphaproteobacteriacanproducefilamentsthat
seasonalchangeincomposition;however,variationinplanktonic aidinattachment(Kragelundetal.,2006;Jonesetal.,2007),and
communitiescorrelatedwithporewaterphysicochemistryatboth dispersallimitationhasbeendemonstratedinsoilActinobacteria
the sub-hourly and seasonal timescales (Figure 3A, red and (Eisenlord et al., 2012). Thus, negative relationships between
blue arrows, respectively). In our system, planktonic organisms these taxa and RC may reflect an enhanced ability of
bray
may therefore be influenced by short-term fluctuations in these organisms to persist locally relative to other community
groundwater-surfacemixing,eitherthroughrapidchangesinthe members.
selectiveenvironmentordispersalviahydrologictransport. Additionally, when we examined assembly processes
Incontrast,selectiononattachedcommunitieswasdetectable governing differences between attached and planktonic
only at the seasonal timescale and exhibited resistance communities, we observed selection for microbial taxa in
to short-term hydrologic fluctuations (Figure 3B). This planktonic communities with unique ecological properties
short-term stability could be facilitated by a number of (Table S6). No correlations with RC were found in these
bray
potentially complementary mechanisms. For example, attached comparisons; however, we identified positive relationships
communities may reside within biofilms, whereby microbial betweenβNTIandtheaveragerelativeabundanceoftwoclasses
cells are imbedded in a matrix of extracellular polymeric of organisms—a candidate class of archaea (Parvarchaeota,
substances.Biofilmsareprevalentinaquaticsystemsandbuffer Figure 4D) and a class of the candidate phylum OP3 (koll11,
communitiesagainstfluctuationsinthehydrologicenvironment Figure 4E). Positive correlations between a particular taxon
(Battin et al., 2016). Attached microbial communities may andβNTIacrosshabitattypesimplythatthetaxonincreasesin
also have adhesion mechanisms (Hori and Matsumoto, 2010) abundanceasthehabitatsdivergeinselectiveenvironments(i.e.,
that confer stability. Community assembly processes, such as selectionbecomesmorevariable).
priority effects, may also contribute to relatively slow rates In our system, Parvarchaeota and koll11 were almost
of community turnover (Fukami, 2004; Fukami et al., 2010). exclusively found in planktonic communities suggesting
Attachedcommunitiescontainedmorespecies(onaverage)than that selection in the porewater environment favors these
planktonic communities (Figure 1D), further suggesting that organisms. Although the specific selective pressures regulating
temporal stability in attached communities may be enhanced the abundance of these organisms are unknown, archaea
by high species richness reducing susceptibility to invasion and members of the PVC superphyla to which OP3 belongs
(Stachowiczetal.,2002). have a cell membrane lacking peptidoglycan that conveys
resistance to common antibiotics and have the genetic
Taxon-Specific Assembly Processes
potential to metabolize C1 compounds such as methane
Taxon-specific assembly processes are masked when relating (Fuerst and Sagulenko, 2011). The distinctive features of these
βNTIandRC toenvironmentalvariables.Toelucidatetaxon- organisms and abundance within our system merits future
bray
specific selection and dispersal mechanisms, we compared the investigating into their role in carbon cycling in hyporheic
relative abundance of taxa identified by SIMPER to βNTI and environments.
RC values. Finally,weobservedchangesintheabundanceoftwomajor
bray
In light of rapid hydrologic fluctuations in the porewater taxa—BetaproteobacteriaandThaumarchaeota—withinattached
environment,relationshipsbetweenRC andtaxaabundances communities that correlated with changes in βNTI. Members
bray
inplanktoniccommunitiesprovideevidenceforaroleoftaxon- of Betaproteobacteria increased in relative abundance with
specific dispersal mechanisms (Table S6). In particular, positive increasesinthestrengthofhomogeneousselection(Figure4H),
relationships of Thaumarchaeota (Figure 4A) and a class of which occurred during times of low groundwater intrusion
Acidobacteria(Figure4B)withRC andnegativerelationships (Figures 5A,B). In contrast, members of Thaumarchaeota
bray
of Actinobacteria (Figure 4C) and Alphaproteobacteria (Figure4G) increased as homogeneous selection waned during
(Figure4D)withRC wereamongthestrongestcorrelations times of high groundwater intrusion (Figures 5A,B). In this
bray
(Table S6). These positive or negative relationships indicate scenariopositivecorrelationsbetweenataxonandβNTIindicate
higher or lower relative abundances, respectively, under higher thatthetaxonbecomesmoreabundantashomogeneousselection
levels of dispersal limitation. No relationships existed between wanes, and thus, that selection targets traits outside the taxon.
planktonictaxaabundancesandβNTI. In contrast, negative correlations between a taxon and βNTI
Althoughwecannotbecertainofthemechanismsresponsible should indicate that the primary selective pressure is for traits
forRC -abundancerelationshipsinplanktoniccommunities, contained within that taxon. Because organisms with wider
bray
the trends we observed help elucidate ecological dynamics niche breadths are favored by selection in a greater variety
impacting the abundance of microbial taxa within hyporheic of environmental conditions, positive relationships should be
zones. For example, taxa showing positive relationships with more probable for organisms occupying niches defined by a
RC —Acidobacteria and Thaumarchaeota—are widely narrow subset of environmental attributes, whereas negative
bray
distributed globally (Francis et al., 2005; Fierer and Jackson, relationshipsshouldbemoreprobablefororganismsoccupying
2006; Jones et al., 2009; Pester et al., 2011), suggesting these broaderniches.Indeed,BetaproteobacteriaandThaumarchaeota,
organisms may be able to disperse under community-level respectively, contain organisms with diverse and limited
dispersal limitation. Conversely, taxa exhibiting negative metaboliccapabilities(Amakataetal.,2005;Yangetal.,2005;Sato
FrontiersinMicrobiology|www.frontiersin.org 9 December2016|Volume7|Article1949
fmicb-07-01949 December14,2016 Time:15:22 #10
Grahametal. CommunityAssemblyandMicrobialMetabolism
FIGURE6|Figuredepictsaconceptualmodeldescribingrelationshipsbetweentraitselection,organismalfitness,andmicrobialmetabolismfor
communitiesexperiencingdualselectivepressures.(A)Selectionforatraitfollowsacontinuousgradientwithinastableenvironment(lightblue)andoscillating
environment(bluetoredgradient).Organismsthatcontainopposingtraits(dashedvs.solidlines)arefavoredateachendofthespectrum,delineatedhereastothe
left(selectionagainsttrait1andfortrait2)orright(selectionfortrait1andagainsttrait2)oftheverticalgrayline.Givenselectioninastableenvironmentdenotedby
theblackdotin(A),variationinhomogeneousselection(B)isdrivenbythemagnitudeanddirectionofselectionintheoscillatingenvironment.Whenselectioninthe
oscillatingenvironmentopposesselectioninthestableenvironment,homogeneousselectiondecreases(B)andmicrobialmetabolismincreases(C)duetoan
increaseinrealizednichespaceandbiodiversity.Blueandredtrianglesin(B,C)correspondtooscillatingselectionlocationsonthefitnesslandscapein(A).
et al., 2009; Pester et al., 2011; Beam et al., 2014; Weber et al., physicochemistry, favor diverse heterotrophs within
2015). Betaproteobacteria. However, when the porewater environment
changesduetoaseasonalchangeingroundwater-surfacewater
Functional Effects Through Time mixing,selectivepressuresshifttofacilitateammonia-oxidizing
Shifts in the relative abundances of Betaproteobacteria and Thaumarchaeota. More broadly, we propose that community
Thaumarchaeota also correlate with changes in groundwater- composition in dynamic environments is often the product
surface water mixing and in rates of microbial metabolism, of multiple selective pressures that operate across different
denoting a link between community assembly processes and timescales, resulting in an increase of specialist organisms
seasonal trends in autotrophic vs. heterotrophic metabolism. duringperiodsinwhichselectionbyanoscillatingenvironment
Betaproteobacteria is a metabolically diverse taxon, exhibiting (e.g., hydrologic change) opposes that of a temporally stable
a range of aerobic and facultative metabolisms including environment(e.g.,consistentsedimentchemistry).
methylotrophy (Kalyuzhnaya et al., 2006), ammonia-oxidation Importantly, changes in community composition in our
(Freitag et al., 2006), nitrogen fixation (Rees et al., 2009), system are associated with a seasonal increase in microbial
phototrophy(Giffordetal.,2013),andavarietyofheterotrophic metabolism, consistent with work in both micro- and
metabolisms(Amakataetal.,2005;Yangetal.,2005;Satoetal., macroecology demonstrating that productivity increases
2009).Althoughwecannotbecertainoftheprimarymetabolic with niche diversification (Hooper et al., 2005; Cardinale
role(s) of these organisms, a positive correlation between et al., 2007; Cardinale, 2011; Gravel et al., 2011; Hunting et al.,
their abundance and NPOC concentration supports their 2015). Specifically, we observed positive correlations between
important contribution to heterotrophy in this system (Table Raz:ATP and Thaumarchaeota abundance coincident with
S7). Betaproteobacteria remained in high abundance relative to negative correlations between Raz:ATP and Betaproteobacteria
otherorganismsthroughoutourstudyperiod,despiteaseasonal (Figure 5D). We therefore infer that seasonally fluctuating
decline, possibly indicating a broad metabolic role for these selective pressures from the porewater environment impact
organisms. In contrast, metabolic activity of Thaumarchaeota microbial metabolism via their influences on niche dynamics.
is primarily constrained to ammonia-oxidation (Pester et al., Weproposethatcommunity-levelnichediversificationgenerated
2011; Beam et al., 2014; Weber et al., 2015). Thaumarchaeota byaseasonalriseofspecializedautotrophsleadstoanincreasein
abundance was negatively correlated with NPOC, consistent communitymetabolicactivitydespiteselectionforheterotrophs
with their involvement in ammonia-oxidation (Table S7). imposedbytheconsistentsedimentenvironment.
The relative abundance of Thaumarchaeota also correlated
with physicochemical properties associated with increased
groundwater-surface water mixing (SO2, NO3, IC, Table Ecological Implications
4
S7), suggesting a heightened importance of Thaumarchaeota- Ourfindingsleadtoaconceptualmodeldescribingrelationships
mediated nitrification when enhanced groundwater discharge between trait selection, organismal fitness, and ecosystem
intothehyporheiczoneleadstoorganiccarbonlimitation(Taylor functioning for communities experiencing multiple selective
andTownsend,2010). pressures (Figure 6). The conceptual model focuses on the
Taken together, we hypothesize that selective pressures, combinedinfluencesofstableandoscillatingselectivepressures,
both from sediment composition and from porewater which should be prevalent across ecosystems. For example, in
FrontiersinMicrobiology|www.frontiersin.org 10 December2016|Volume7|Article1949