OPEN TheISMEJournal(2017)11,2691–2704 www.nature.com/ismej ORIGINAL ARTICLE Drought and host selection influence bacterial community dynamics in the grass root microbiome Dan Naylor1,2, Stephanie DeGraaf3, Elizabeth Purdom3 and Devin Coleman-Derr1,2 1DepartmentofPlantandMicrobialBiology,111KoshlandHall,UniversityofCalifornia,Berkeley,CA,USA; 2Plant Gene Expression Center, UC Berkeley, USDA-ARS, Albany, CA, USA and 3Department of Statistics, University of California, Berkeley, CA, USA Rootendophyteshavebeenshowntohaveimportantrolesindetermininghostfitnessunderperiods ofdroughtstress,andyettheeffectofdroughtonthebroaderrootendospherebacterialcommunity remains largely uncharacterized. In this study, we present phylogenetic profiles of bacterial communities associated with drought-treated root and rhizosphere tissues of 18 species of plants withvaryingdegreesofdroughttolerancebelongingtothePoaceaefamily,includingimportantcrop plants. Through 16S rRNA gene profiling across two distinct watering regimes and two developmental time points, we demonstrate that there is a strong correlation between host phylogenetic distance and the microbiome dissimilarity within root tissues, and that drought weakens this correlation by inducing conserved shifts in bacterial community composition. We identifyasignificantenrichmentinawidevarietyofActinobacteriaduringdroughtwithintherootsof allhosts,anddemonstratethatthisenrichmentishigherwithintherootthanitisinthesurrounding environments. Furthermore, we show that this observed enrichment is the result of an absolute increase in Actinobacterial abundance and that previously hypothesized mechanisms for observed enrichmentsinActinobacteriaindrought-treatedsoilsareunlikelytofullyaccountforthephenomena observedherewithin theplant root. TheISME Journal (2017) 11,2691–2704; doi:10.1038/ismej.2017.118; published online28July2017 Introduction the degree to which such changes are conserved across different plant hosts. Few environmental stresses are as omnipresent and Recent studies have noted enrichment of the devastating to agriculture as drought, which bacterial taxa Actinobacteria in drought-treated annually results in billions of dollars in losses soils across a range of environments (Bouskill worldwide (Lesk et al., 2016). While some crop et al., 2013, 2016) and in drought-treated rhizo- responses to drought are conserved across most spheres (soils adhering to the root surface) for species, including increased ABA production and several plant species (Nessner Kavamura et al., stomatal closure, other adaptations are host-speci- 2013; Taketani et al., 2016). It has been suggested fic, such as special carbon uptake pathways, that this observed relative enrichment is due to increased cuticle thickness and altered root mor- differing life strategies of soil microorganisms; phology (Fang and Xiong, 2015). As plants are specifically, the spore-forming ability of Actino- intimately intertwined with the communities of bacteria, which allows them to enter a stable and microbes living in and around them (Gaiero et al., quiescent state during periods of environmental 2013; Philippot et al., 2013; Berg et al., 2016), stress, would lead them to persist under drought perturbations in plant physiology and metabolism, conditions, while less fit bacterial lineages such as those resulting from drought response decrease in abundance (Nessner Kavamura et al., mechanisms, can be expected to alter the composi- 2013; Taketani et al., 2016). Whether Actinobac- tion of the plant microbiome with potential con- teria are also enriched within the plant root under sequences for host fitness (Berg et al., 2014). drought, and the degree to which such enrichment However, little information is available on how patterns are generalizable across a broad range of drought influences the plant root microbiome, and hosts and Actinobacterial taxa, remains unknown. We hypothesize that the plant root microbiome Correspondence:DColeman-Derr,PlantGeneExpressionCenter, will exhibit significant shifts during drought, UC Berkeley, USDA-ARS, 800 Buchanan Street, Albany, CA including an increase in Actinobacteria, and that 94710,USA. these changes will be, at least in part, host E-mail:[email protected] specific as a result of differing drought response Received3February2017;revised20April2017;accepted7June 2017;publishedonline28July2017 mechanisms. Droughtshapesgrassrootmicrobiome DNayloretal 2692 While the effect of drought on the root micro- Materials and methods biome remains largely unexplored, more effort has been made to understand the role of plant host Experimental design genotype in shaping root-associated bacterial com- The primary experimental field used in our study is munities (Kuske et al., 2002; Brusetti et al., 2005; an agricultural field site located in Albany, Califor- Aleklett et al., 2015; Bulgarelli et al., 2015). nia(37.8864°N,122.2982°W)characterizedbyasilty Different host species not only have unique loam soil with low pH (5.2) (Supplementary Table exudationpatterns (Cavaglierietal.,2009) but also S1). A smaller experiment at an additional field site root architecture, lifespan and rooting depth, all of was conducted to test the reproducibility of results which have been shown to influence microbiome observed at the first site (see Supplementary structure (Fierer et al., 2003; Shi et al., 2011; Information). The second site is located at the Berendsen et al., 2012; Badri et al., 2013; Chaparro Kearney Agricultural Research station (36.6008°N, et al., 2013). As these root phenotypes are under 119.5109°W), run by the University of California, genetic control of their hosts, genetically similar and has sandy loam soils with a silty substratum hosts might be expected to harbor comparable with neutral pH (7.37) (Supplementary Table S1). microbiome profiles (Parker and Spoerke, 1998; The nine C3 and nine C4 grasses and tomato Ushio et al., 2008). Indeed, host phylogenetic germplasm used in the primary study are described relationships were recently shown to be correlated in Supplementary Table S2, and include nine C3 with microbiome dissimilarity in the rhizosphere species(twovarietiesofTriticumaestivum(TaBand (Bouffaud et al., 2014), and among accessions of a TaG), Triticum monococcum (Tm), Triticum turgi- single species, similar hosts had more comparable dum(Tt),Secalecereale(Sc),Hordeumvulgare(Hv), microbiota in rhizosphere and roots (Peiffer et al., Avena sativa (As), Festuca arundinaceae (Fa), 2013; Bulgarelli et al., 2015). However, the relative Brachypodium distachyon (Bd)), nine C4 species strength of this correlation within roots and rhizo- (two varieties of Sorghum bicolor (SbA and SbH), spheres have yet to be compared within a phylo- Sorghum laxiflorum (Sl), Sorghastrum nutans (Sn), genetically diverse host framework. Additionally, Miscanthussinensis(Ms), Bothriochloa bladhii (Bb), it remains to be seen whether drought stress Zea mays (Zm), Pennisetum americanum (Pa), strengthens or weakens the correlation between Eragrostis tef (Et)) and an outgroup, Solanum host phylogeny and bacterial community composi- lycopersicum (Out). Seeds were surface sterilized, tion. We hypothesize that a positive correlation germinated and grown in growth chambers for between host phylogenetic and microbiome dis- 2 weeks as described in the Supplementary tance exists and will be strongest inside the root, Information,afterwhichallplantsweretransplanted where plants have greater influence over bacterial to the field in five full factorial replicate blocks. interactions and recruitment; furthermore, we Under control conditions, plants were watered for hypothesize that drought will strengthen this 1h three times per week using drip irrigation with correlation, because the metabolic and phenotypic 1.89Lh−1rateflowemittersundercontroltreatment, factorsdrivinghost-specificdifferencesinbacterial and no water was given during drought treatments. recruitment are, at least in part, adaptive traits to For the early (preflowering) time point, we imposed differing water availability in each host’s native drought 2 weeks post-transplantation and harvested environment. 5 weeks post-transplantation. For the late (postflow- To test these hypotheses, we analyzed bacterial ering)timepoint,weimposeddrought4weekspost- community composition in root endosphere and transplantation, and harvested at 12 weeks post- rhizosphere for 18 grass species in the Poaceae transplantation.Atcollectiontime,thepercentageof clade, grown in the field under drought conditions. plants that had transitioned from vegetative to We included nine species with C4 carbon metabo- reproductive growth was 2.1% at the early time lism and nine with C3 carbon metabolism, as plant point (almost entirely attributable to Solanum lyco- species using C4 metabolism are better adapted to persicum),incontrastto93.4%atthelatetimepoint arid environments (Fang and Xiong, 2015). The (Miscanthussinensiswastheonlyplantthathadnot grasses we chose included agronomically valuable flowered). For soil moisture content and plant C3 crops with moderate to poor drought tolerance, biomass, samples were collected at harvest time in such as wheat and barley (Nezhadahmadi et al., all five replicate blocks. For microbiome data, bulk 2013; Ghotbi-Ravandi et al., 2014), and also highly soil, rhizosphere and root samples were collected drought-tolerant C4 species, such as sorghum from three replicate blocks (due to limitations in (Sabadin et al., 2012). Our experimental design time and funding available, only three of the five allowed us to investigate the effect of drought on replicate blocks were used for microbiome analyses) the grass microbiome, to observe whether such andgenomicDNApreparedasinColeman-Derretal. effects are unique or conserved across a broad range (2016) with minor modifications, for 206, 207 and ofhostswithdifferingdroughttolerances,andassess 211 soil, root and rhizosphere samples. In brief, the degree of correlation between host phylogeny 15cm soil core samplers were used to collect bulk and the root microbiome under normal and drought soil up to ~20cm from the base of the plant. Root conditions. systems were excavated from the top 25cm of soil, TheISMEJournal Droughtshapesgrassrootmicrobiome DNayloretal 2693 shaken briefly to remove excess soil and root tissue of30sat95°C,1minat61.5°C,followedby5minat was collected and stored in epiphyte removal buffer 4°C, 5min at 90°C and holding at 4°C. All steps (0.75%KH PO ,0.95%K HPO ,1%TritonX-100in were performed with a slow ramp rate of 2°Cs−1. 2 4 2 4 ddH2O; filter sterilized at 0.2μM) on ice. Root Plates were promptly analyzed after conclusion of samples were then sonicated for 10min at ‘Low’ in PCR runs using the aforementioned droplet reader. aDiagenodeBioruptor(DiagenodeInc.,Denville,NJ, The Actinobacteria-specific primers used here have USA); residual soil remaining at the bottom of the reportedspecificitiesof91.8and99.4%,respectively tube after this step was retained as ‘rhizosphere’, (BacchettiDeGregorisetal.,2011;Yangetal.,2015), whereas the roots were transferred to a new tube, andinsilicoanalysesusingSILVATestPrime(Quast rinsed two times with sterile water and stored in et al., 2013) revealed similar specificity values of sterile epiphyte removal buffer. 87.8 and 99.3%. Plant phylogenetic distances Statistical analyses Phylogenetic distances between our 19 hosts were AllstatisticalanalyseswereperformedinRusingthe determined through the analysis of three genes rbcL normalized reduced data set with 2760 OTUs and (Rubisco large subunit), ndhF (NADH dehydrogen- 612samples,anddescriptionsofindividualanalyses ase subunit F) and matK (tRNA-Lys) intron; the anduseofspecificpackagesisdescribed indetailin primers used for the amplification of these regions the Supplementary Information. are described in Grass Phylogeny Working Group II (2012). Genomic DNA extractions, PCR and Sanger sequencing were carried out as described in the Results Supplementary Information. We planted 18 grass lineages (Supplementary Table S2) including varieties of wheat, rye, barley, oat, Amplicon data Brachypodium, tall fescue, sorghum, Indian grass, Samples from the primary data set were amplified Miscanthus, plume grass, maize, millet and tef, as using a dual-indexed 16S rRNA Illumina iTags well as one dicot species, tomato, as an outgroup. primer set specific to the V3–V4 region (341F, Plants were grown in five replicated randomized 5′-CCTACGGGNBGCASCAG-3′ and 785R, 5′-GAC blocks(SupplementaryFigureS1)withinourexperi- TACNVGGGTATCTAATCC-3′) as described in mental field. As plant developmental stage has been Takahashi et al. (2014) using reaction conditions implicated as a factor shaping the plant microbiome as described in Coleman-Derr et al. (2016). Raw (Edwards et al., 2015; Dombrowski et al., 2017), and reads were trimmed, quality-filtered, chimera- as drought has different effects on host physiology checked, clustered and assigned taxonomies as in depending on when it occurs in development (Fang Coleman-Derr et al. (2016), yielding 2760 high- and Xiong, 2015), we used two distinct treatments abundance operational taxonomic units (OTUs) for drought-treated replicates. In half of each of the after removing OTUs with o5 reads in at least five five full factorial replicate blocks, we imposed samples and then normalizing to 10000 reads per drought for non-control plants at week 4 and sample. Additional sequence data sets generated harvested 3 weeks later before most plants had for samples from our secondary field site, for a flowered; in the other half of the five blocks, we comparisonofDNAextractionmethodsandforthe imposed drought for non-control plants at week 6 fungal ITS2 sequencing, are described in the and harvested 8 weeks later after the majority of Supplementary Information. Raw sequencing data plants had flowered. This design allowed us to were deposited in the NCBI Short Read Archive, obtain samples from both preflowering (early) and BioProjectID PRJNA369551, accession number postflowering (late) time points. To test whether SRP098753. irrigation treatments had a significant effect on both plantgrowthandavailablesoilmoisture,allsamples inthefivereplicateblocksbelongingtothelatetime Actinobacterial-specific ddPCR point were used to measure above-ground shoot Digital droplet PCR using Bio-Rad QX200 Droplet biomass and soil water content. Both biomass and Reader(Bio-Rad,Hercules,CA,USA)wasperformed soilmoistureweresignificantlylowerunderdrought according to the manufacturer’s instructions. The (SupplementaryFigureS2),suggestingbothenviron- reaction volume was 20μl, including 900nM ment and plant were affected, even across the broad Actinobacteria-specific primers (Act920F3/ range of hosts. Act1200R (Bacchetti De Gregoris et al., 2011) and Toobtainmicrobiomedata,allsamplesfromthree Act664F/Act941R (Yang et al., 2015)), 1x EvaGreen replicate blocks, including both early and late time supermix and 1ng template DNA. Droplets were points,wereharvestedandbacterialcommunitiesin generated using 20μl sample and 70μl oil; 30μl root, rhizosphere and bulk soil samples were droplets were used for amplification, with the examined. We chose to focus on bacterial commu- following conditions: 5min at 95°C, then 40 cycles nities because of their demonstrated potential for TheISMEJournal Droughtshapesgrassrootmicrobiome DNayloretal 2694 Figure1 (a)DensityplotsofShannon’sDiversityindrought(yellow)andcontrol(samples)foreachcombinationofsampletype(soil, rhizosphere, root) and time point (early, late). Asterisks indicate drought treatment groups that were significantly different from the respective control group by anlaysis of variance (ANOVA) and Tukey's post hoc tests. (b) Principal coordinate plot for all samples generatedusingtheBray–Curtisdistance;samplesarecoloredforeachcombinationofsampletypeandtreatment.(c)Percentofvariance explained by each factor as determined by canonical analysis of principal coordinates (CAPS) using the Bray–Curtis distance for remainingthreeexperimentalfactors(species,timepointandreplicate),performedseparatelyforeachpairwisecombinationofsample typeandtreatment.They-axisindicatesthefractionofvarianceexplained,andtheshadeofthebarindicatesthesignificance.Errorbars indicatea95%bootstrapconfidenceintervalaroundeachfractionofvarianceexplained. plant growth promotion during periods of drought Bacterial diversity in the grass root microbiome is (Casanovas et al., 2002; Marasco et al., 2012), their influenced by drought and host species responsiveness to drought in soils (Bouskill et al., Toinvestigatehowbacterial α-diversitywasaffected 2013) and their relative abundance within the root by sample type, time point, species and watering microbiome (Shakya et al., 2013; de Souza et al., treatment, Shannon’s diversity index was calculated 2016). By comparison, fungal communities in soils for each sample. Analysis of variance revealed that and rhizosphere have been observed to remain all four factors had significant effects, and that relatively similar between drought and control sample type had the greatest effect (Supplementary conditions (Barnard et al., 2013; Blazewicz et al., Table S4). α-Diversity decreases significantly from 2014),andarchaeaarearelativelyminorcomponent soil to root, as found in similar previous studies of the total community in agricultural sites (Rascovan et al., 2016), although rhizosphere diver- (Taketani et al., 2015). Community analyses sity was not significantly lower than that found in were performed using paired-end sequencing of the the soil. Within the nine most abundant bacterial V3–V4 region of the ribosomal 16S rRNA on classes,α-diversitywasmoreaffectedbysampletype the Illumina MiSeq platform (Illumina Inc., for some than others. The class Actinobacteria San Diego, CA, USA), generating 57380917 reads displayedthemostsignificantdecreaseinα-diversity (Supplementary Table S3). These raw sequence data in roots (Supplementary Figure S3 and were demultiplexed, trimmed, quality-filtered, clus- Supplementary Table S5), and this was attributed tered and assigned taxonomies, resulting in a mainly to one genus, Streptomyces, which was reduction to 25694125 read counts, which mapped greatly enriched in roots relative to soils, represent- to 8434 distinct OTUs based on 97% similarity ing up to 50% of all Actinobacteria read counts threshold.AfterremovinglowabundanceOTUsand (Supplementary Figure S4). Of the remaining three normalizingtoaccountfordifferencesinreaddepth, factors,treatmenthadthegreatesteffect,followedby the data set included 2760 OTUs representing 32 speciesandtimepoint.Fortreatment,medianvalues distinct phyla of bacteria. of Shannon’s diversity were lower for drought than TheISMEJournal Droughtshapesgrassrootmicrobiome DNayloretal 2695 for control in all pairwise combinations of sample proportion of variance (37.6% and 17.4%, respec- type andtimepoint,althoughTukey’spost hoc tests tively). To investigate the amount of variance in the indicatedthatthesedifferenceswereonlysignificant data set attributable to each of our experimental variables, we used canonical analysis of principal at early time points (Figure 1a). coordinates,whichallowsconstrainedevaluationfor To determine which factors most influenced differences in community structure between sam- each experimental factor. After separating the data ples, we investigated β-diversity through principal set into groups based on sample type and watering coordinateanalysisusingBray–Curtisdissimilarities treatment,whicharethetwomostsignificantfactors generatedforallsamples(Figure1b).Takentogether, (Supplementary Table S6a), nearly all remaining factorsweresignificantinallsamplegroups,buthost the first two axes explained 35.4% of the variance, species explained a relatively large proportion of withtheprimaryaxis(27.6%)primarily distinguish- ing samples by sample type and the secondary axis variance in both rhizosphere and root under both watering treatments (Figure 1c and Supplementary (7.6%) by watering treatment (Figure 1b). Principal Table S6b). Interestingly, the level of variance coordinate analysis based on weighted UniFrac explained by host species was not significantly distances (Supplementary Figure S5a) supported similartrends,withthetwoaxesexplainingagreater greater in drought than control samples, Figure2 (a)Therelativeabundanceofeachofthetopninemostabundantbacterialclassesforeachpairwisecombinationofsample type,timepointandtreatment.(b)Thefoldenrichmentunderdroughtofeachofthetopninemostabundantbacterialclassesforeach pairwisecombinationofsampletypeandtimepoint.(c)Heatmapdisplayingthefoldenrichmentunderdrought(yellow)orcontrol(blue) inrootsforallbacterialgenera(x-axis)belongingtothetopninemostabundantclassesandidentifiedasstatisticallyenrichedineither drought or control treatments through indicator species analysis. Enrichment is shown independently for each of the 19 host species (y-axis). The class that each genus belongs to is indicated by the colored bar above its name (colors represent the same class as in (a) and (b)). Enrichment is shown independently for each of the 19 host species (y-axis). Abbreviations correspond to host species as follows (Out=Solanum lycopersicum, Et=Eragrostis tef, Pa=Pennisetum americanum, Zm=Zea mays, Bb=Bothriochloa bladhii, Ms=Miscanthus sinensis, Sn=Sorghastrum nutans, Sl=Sorghum laxiflorum, SbH=Sorghum bicolor H, SbA=Sorghum bicolor A, Bd=Brachypodium distachyon, Fa=Festuca arundinaceae, As=Avena sativa, Hv=Hordeum vulgare, Sc=Secale cereale, Tm=Triticummonococcum,Tt=Triticumturgidum,TaG=TriticumaestivumG,TaB=TriticumaestivumB). TheISMEJournal Droughtshapesgrassrootmicrobiome DNayloretal 2696 contradicting our hypothesis that drought might Relative abundance shifts indicate enrichment for increase the effect of host species on community Actinobacteria under drought composition. Indeed, when using weighted UniFrac We next investigated changes in relative abundance distances, the variance attributable to host species patterns associated with drought. As has been was lower under drought than control conditions reported for drought-treated soils and rhizospheres (SupplementaryFigureS5b),suggestingthatdrought (Blazewicz et al., 2014; Taketani et al., 2016), the may instead induce conserved shifts in the root and most striking change observed was enrichment for rhizosphere microbiome across our 18 grass hosts. Actinobacteria (Figure 2a). Notably, this enrichment Finally, a smaller experiment to test the effect of was most pronounced in the roots, where Actino- droughtonfungalcommunitieswasperformedusing bacteria exhibited a 3.1-fold increase, as compared fungal-specific ITS2 amplicon sequencing of sam- with 2.3- and 1.5-fold enrichment in rhizospheres ples belonging to only the early time point from one andsoils,respectively(Figure2b).Enrichmentunder replicate block. Through principal coordinate analy- drought was also observed for class Bacilli, whereas sis and canonical analysis of principal coordinate Acidobacteria and Deltaproteobacteria were rela- analysis,weobservedthatdroughthadnosignificant tivelydiminished.Todetermineifthesetrendswere effectoncommunitycompositioninanyofthethree driven by many or few members of each bacterial sample types (Supplementary Figure S6), as had class, we investigated enrichment patterns in roots been reported recently for comparative analyses of for all bacterial genera identified as indicators of dry and wet soils (Barnard et al., 2013). either drought or control treatments through Figure3 PhylogenetictreeofallActinobacteriageneraidentifiedasindicatorsofdroughttreatmentinanysampletype.Theinnerring indicatesthefiveordersthesegenerabelongto;thesecond,thirdandfourthinnermostringsindicatethedegreeoflog foldenrichmentin 10 thesoil,rhizosphereandroot,respectively(legendatbottom).Darkershadesofcolorindicategreaterenrichment,whereaswhiteindicates nosignificantenrichment.Theouterringindicatesthelog abundanceofeachgenusinallsamples. 10 TheISMEJournal Droughtshapesgrassrootmicrobiome DNayloretal 2697 indicator species analysis (Dufrene and Legendre, droughtinall sample types (32).A literature search 1997) (Figure 2c). All indicator genera belonging to indicated that only half of these lineages have been both Actinobacteria (52) and Bacilli (8) were observed to form spores (Supplementary enriched under drought relative to control—in fact, Figure S11). Similarly, a bioinformatic analysis of therewerenoActinobacteriaidentifiedasindicators representative sequenced genomes indicated that of the control treatment in any sample type only half of these genera contain copies of two (Supplementary Figure S7). By contrast, most indi- genes (ssgB and bldD) essential to Actinobacterial cator genera belonging to other classes exhibited sporulation (Bush et al., 2015). An alternative enrichment under control conditions (Figure 2c). explanation for the observed enrichment could be This enrichment pattern was found broadly across that Actinobacteria do not in fact increase in all host species, including the non-grass outgroup absolute abundance, but instead other bacterial tomato (Figure 2c and Supplementary Figure S8), lineages perish under these conditions, leading to suggesting that this phenomenon might be generally the observed relative increase. To test this hypoth- true for both monocots and dicots. While the largest esis, we performed quantification of absolute abun- effect was a conserved increase in Actinobacteria danceofActinobacteriainrootsunderbothdrought across all hosts, we did observe that Actinobacteria and control conditions using digital droplet PCR represented a larger fraction of the drought-treated (Supplementary Figure S12). The results demon- root microbiome in C4 than in C3 grasses strate that Actinobacteria exhibit a marked increase (Supplementary Figure S8), and relative enrichment in absolute abundance in drought-treated roots for between drought and control was on average more all C3 and C4 grass hosts tested (Supplementary pronouncedinC4grasses(3.4-foldenrichmentinC4 Figure S12). Taken together, the above analyses compared with 2.7 in C3). A phylogenetic recon- suggest that neither decreases in the abundances of struction of all indicator Actinobacterial lineages other taxa nor sporulation are likely to be the only (Figure 3) showed that while 33% of genera are factorsresponsibleforenrichmentofActinobacteria enriched under drought in all sample types, the under drought in roots observed here, although majority were specifically enriched in either roots furtherexperimentalevidenceisneededtotesthow and/orrhizosphere(Saccharopolyspora,Actinopoly- universal these patterns are. morpha, Glycomyces) or exclusively in soils (Krib- bella, Curtobacterium, Nonomurea), and that there was no overarching trend with respect to Actino- Phylogenetic distance between host species is bacterial taxonomy and sample type-specific enrich- correlated with root endosphere community distance ment patterns. Replication of this experiment with a Wenextusedthedatasettotestourhypothesesthat subset of host species at an additional field site host phylogenetic distance is positively correlated located at the Kearney Agricultural Research station with microbiome community distance inside the showedsimilartrendsinActinobacterialenrichment plant root, and that drought strengthens this correla- under drought (Supplementary Figures S9a and tion. To do so, we first determined plant host S10a). The effect was particularly pronounced in distances through sequencing three chloroplast the roots, and indeed most Actinobacterial genera genes, matK/trnK, rbcL and ndhF, for each host enrichedunderdroughtatourAlbanyfieldsitewere speciesandperformingphylogeneticanalysisonthe also enriched at the Kearney field site concatenated sequences (Figure 4a). We used Man- (Supplementary Figure S10b), with two exceptions tel’s tests to compare distances based on host (Catellatospora and Virgisporangium), which were phylogeny and Bray–Curtis dissimilarity enriched under control conditions. However, over- (Figure 4b) for each pairwise combination of sample lap in the enrichment patterns were not limited to type, time point, treatment and photosynthesis (C3 Actinobacteria; Although there were some discre- and C4); the latter segregation was introduced to aid pancies, most bacterial lineages identified as indetectionofthecorrelationbetweenspecieswitha enriched under drought or control in Albany had simple nested-monophyletic relationship, as is similar patterns when their respective abundance broadlytrueofthememberschosenheretorepresent trends at the Kearney site were examined the C3 and C4 clades. These analyses indicated that (Supplementary Figures S9b and S10b). root samples generally displayed a significant corre- AssomegeneraofActinobacteriasporulateunder lation between host phylogenetic distance and unfavorable environmental conditions (Bergey’s, microbiome distance, but that Mantel’s R-statistic 2010), the relative increase in abundance observed values for rhizosphere samples tended to be smaller hereandinrecentstudieshasbeenhypothesizedto andnotalwayssignificant.Soilmicrobiomesamples be due to DNA replication and cell division showed no significant correlation. Contrary to our inherent to sporulation (Taketani et al., 2016); if hypothesis that environmental stress might induce true, the resulting cells would be largely quiescent host-specific shifts in root-associated bacterial com- and perhaps unlikely to affect community function munities, at the early time point drought signifi- (McCormick andFlärdh, 2012). To test this hypoth- cantly reduced the size of the correlation in both esis, we documented all Actinobacterial lineages rhizosphere and roots. The early time point is also with genus level-classifications enriched under the time point in which we observe the greatest fold TheISMEJournal Droughtshapesgrassrootmicrobiome DNayloretal 2698 Figure4 (a)Phylogenetictreeforthe18grasslineagesusedinthisexperiment,constructedfromanalignmentofthreeconservedfull- lengthchloroplastgenesequences(rbcL,matK,trnK);nodesforC3grassesarecoloredinshadesofblue,andC4grassesarecoloredin shadesofred.(b)TheMantel’sRcoefficientisplottedforpairwisecomparisonsofhostphylogeneticdistanceandmicrobiomedistancefor each pairwisecombination of sample type, treatment and time point. To maintain sample evenness across the surveyed phylogenetic breadthofhostspecies,C3andC4cladeswereanalyzedseparatelywiththeinclusionofasinglememberfromtheothercladeandthe outgroupspecies.Theshadeofeachbarindicatesthesignificance.(c)TheMantel’sRcoefficientisplottedforpairwisecomparisonsof fielddistanceandmicrobiomedistanceforeachpairwisecombinationofsampletype,treatmentandtimepoint.Theshadeofeachbar indicatesthesignificance. enrichment in Actinobacteria between drought and for root samples for that host. To obtain the list of controlacrossbothrootsandrhizosphere(Figures2a taxafoundincommonacrossthebroadergrassroot and b). However, at the late time point, we observed microbiome, we identified those OTUs that were nosignificantreductioninthesizeofthecorrelation, sharedbetweenthe18listsoftaxacommontoeach which could be explained by the emergence of individual grass host (Supplementary Figure S13). additionaland/orstrongerhostselectionphenotypes Of these 224 OTUs, which represent 8.1% of all appearing later in plant development. Interestingly, OTUsand26.1%oftotalreadcountabundance,all when a Mantel’s test was used to compare pairwise but 6 were also identified as members of the taxa matrices for the microbiome and physical distances commonly found in tomato, suggesting that the between host plants, the opposite pattern was seen, microbes identified here as taxa common to the withsignificantandlargecorrelation insoils, butno grass root microbiome are not necessarily specific correlation observed in rhizosphere and roots to grasses but instead represent commonly (Figure 4c). Taken together, these results confirm enriched root endophytes of both monocots and our hypothesis that community composition is most dicots. Pairwise comparisons of individual lists of highly correlated with host phylogenetic distance common taxa for all 19 hosts (Figure 5) indicate a within the roots. However, in contrast to our varying degree of similarity (ranging from 42.2 to hypothesis that drought would strengthen this 73.5% overlap). Interestingly, we observed that the correlation, we observed evidence that drought three species of sorghum had larger sets of weakens it. commonly observed taxa than any other host. Notably, two of these hosts (Sorghum bicolor H and Sorghum laxiflorum) also had the lowest Common taxa in the grass microbiome median overlap with the lists of taxa commonly Our experimental design, incorporating 18 grass observed in the other hosts (55.8 and 58.3%), species grown in a single environment, allowed for including the outgroup tomato (59.4%). Taken exploration of the taxa found in common across together, these results suggest that while a collec- each of our grass root microbiomes. First, the taxa tivesetofcommonroottaxaexistsforgrassspecies, commonlyobservedineachindividualhostspecies these microbes represent asmall fractionofall root was determined by retaining only those OTUs that OTUs (6%) and are likely generalist colonizers of were present in at least 30% of samples and which the roots of most plants. However, as field site and represent at least 0.01% of read count abundance soil type can have a large effect on overall TheISMEJournal Droughtshapesgrassrootmicrobiome DNayloretal 2699 Figure5 (a)Heatmapdisplayingthepercentofoverlapbetweenthetaxacommontotherootmicrobiomeofeachhostwitheveryother host;lightershadesofbluerepresentlessoverlap,whiledarkershadesrepresentmore.Hostspeciesidentifiersarethefirstletterofthe genusandspecies(e.g.ZmforZeamays).(b)ThesizeofthelistofcommontaxaasthenumberofOTUsidentifiedasmemberofeachlist foreachofthe19hostspecies. microbiome composition, further work will be effect size observed in our study as compared with needed to determine the reproducibility of these the relatively low values observed in previous results in other systems. studies suggests that the influence of host species on the root microbiome should be considered in the context oftheworking phylogenetic framework, and a broader array of considered species will accord- Discussion inglyproduceagreaterhosteffect.Differencesinthe Cereal host root microbiome composition correlates developmental timelines between species may have with host phylogenetic distance introducedvariabilitywithrespecttodevelopmental Awiderangeofhostandenvironmentalfactorshave stage at the time of harvest between species; beenshowntoinfluenceplantmicrobiomecomposi- however, based on field observations for flowering tion (Berg and Smalla, 2009; Gaiero et al., 2013). time, the time points we selected broadly repre- Using canonical analysis of principal coordinates sented pre- and postflowering stages for nearly all analysis, we found that sample type, host species species. In contrast to our hypothesis that drought and watering treatment were the three most impor- might increase the differences in root community tant sources of compositional variance. Sample type composition between different hosts, we observed hasbeenimplicatedasthelargestsourceofvariation that the effect of host species was not substantially in multicompartment data sets (Lundberg et al., different between drought and control when using 2012; Marasco et al., 2012; Bulgarelli et al., 2015; Bray–Curtis distances, and was marginally reduced Cherif et al., 2015; Edwards et al., 2015; Coleman- when using weighted UniFrac distances. These Derr et al., 2016; Fonseca-García et al., 2016). results are consistent with the large shift in Actino- Previous studies investigating the effect of host bacteria we observe in roots, which is broadly geneticshavetendedtofocusonvarieties ofasingle conserved across all hosts we examined. Other given host species, such as barley (Bulgarelli et al., environmental stresses may provoke similar trends 2015), maize (Peiffer et al., 2013) or agave in conserved enrichment patterns across hosts, (Desgarennes et al., 2014; Coleman-Derr et al., whether in Actinobacteria or other bacterial taxa, 2016), or on a few distinct species (el Zahar and will need to be investigated. Haichar et al., 2008; Bouffaud et al., 2014; While previous studies have confirmed the role of Schlaeppi et al., 2014). In these studies, species genotype in determining microbiome composition, effects tended to be nonsignificant (Schlaeppi et al., very few have tested the correlation between host 2014) or small (Lundberg et al., 2012; Peiffer et al., phylogenetic and microbiome distances. In examin- 2013; Fonseca-García et al., 2016). By comparison, ing four Brassicaceae members, microbiota diversi- we found that host species was highly significant fication was not consistent with host distance and explained ~19.4 and 23.4% of variance within (Schlaeppi et al., 2014); similarly, two studies of the rhizosphere and root endosphere. The species maize inbred lines (Bouffaud et al., 2012; Peiffer TheISMEJournal Droughtshapesgrassrootmicrobiome DNayloretal 2700 et al., 2013) found that differences in rhizobacterial in determining community composition within root communities did not correlate with host genetic communities; drought was found to explain 8.8% distance. However, these studies may have been and 9.9% of variance within the root and rhizo- hampered by low sample size and close relatedness sphere, as compared to 5.6% within soils. While of host species. A study using five maize varietals drought induced many changes in microbiome and two additional Poaceae members (Bouffaud composition, the most significant at the class level et al., 2014) did detect a significant correlation wasanincreaseofActinobacteria.Whileenrichment between plant phylogenetic distance and micro- ofActinobacteriaunderdroughthasbeenreportedin biome distance in rhizospheres, but did not investi- soils(Haydenetal.,2012;Bouskilletal.,2013,2016) gate root endospheres. In this study, we expand on and rhizosphere (Nessner Kavamura et al., 2013; previous results by: (1) incorporating many distinct Taketani et al., 2016), enrichment has not been species from both C3 and C4 grasses; (2) examining reported in the root endosphere, nor have enrich- the relative effect size in the rhizosphere and root ment trends been compared between these distinct endosphere; and (3) investigating how environmen- microhabitats. Our observation that enrichment of talstressmightperturbtherelationshipbetweenhost Actinobacteria under drought is significantly higher phylogeny and microbiome composition. Our data in the root endosphere compared with either rhizo- confirm our hypothesis of a positive correlation sphere or bulk soil suggests that while the Actino- betweenhostphylogeneticandmicrobiomedistance bacterial enrichment under drought is not unique to that is strongest in roots, weaker in the rhizosphere roots, it is enhanced by them. Indeed, 24 genera, and nonexistent in the surrounding soil, indicating includingSaccharopolyspora,Glycomyces andActi- thathost-selective forces onthemicrobiomegrow in nopolymorpha, were exclusively enriched in roots magnitude with increased plant–microbe intimacy, and rhizosphere and not in surrounding soil. The whichhasbeenrecentlydemonstratedintwostudies observed enrichment in Actinobacteria under (Beckers et al., 2017; Samad et al., 2017). We also drought treatment within roots was found in two observed that drought generally decreased the effect field sites, suggesting that these results may be size and significance of the Mantel’s statistic, most generalizable not only across different plant hosts notably at the early time point, contradicting our but also across distinct environments as well. hypothesis that drought strengthens the correlation Additionally, while levels of enrichment for indivi- between host phylogeny and microbiome distance dualActinobacteriallineageswerelargelyconsistent due to host species-specific drought response across all hosts, there were notable exceptions. For mechanisms. Instead, these results are consistent example, an Actinobacteria genus with the most with the idea that conserved plant responses to significant enrichment under drought in grass roots drought, especially early in development (such as (20.3-fold), Glycomyces, had the opposite pattern in increased ABA production), have larger effects on tomato. Further comparative genomic analyses and microbiome composition than individual and functional studies using these lineages that exhibit species-specificadaptations.Additionally,theobser- sample type- and host-specific enrichment patterns vation that in thelate time point drought provokes a might help decipher underlying mechanisms less similar, and therefore a more host-specific, involved in this process. microbial community across hosts, as indicated by The general enrichment of nearly all Actinobac- the relatively larger size and significance of the teria under drought suggests that the force driving Mantel's correlation, is evidence that perhaps host thisenrichmentisrelatedtooneormoreconserved specificitydue to species-specific drought responses properties of the Actinobacterial lineage. For may take time to accumulate. Phylogenetic distance example, among bacteria, they are highly tolerant is not the only factor that differentiates our selected oflifeinaridenvironments(BullandAsenjo,2013; grasses: for example, root morphology can vary Pearce et al., 2013; Stevenson and Hallsworth, significantly between different grasses (Rich and 2014; Cherif et al., 2015) where other bacterial Watt, 2013) andcouldhave affected the microbiome lineages have difficulty surviving, owing in part to composition. To account for this factor, plant roots their ability to form spores. Our observation that were harvested to a consistent rooting depth (25cm) only about half of drought-enriched Actinobacteria for all species, athough it is possible that root in our study are likely capable of forming spores morphology or other species-specific root character- suggeststhatsporulationisnotfullyresponsiblefor istics that do not necessarily correlate with host their prevalence under drought observed here. phylogenetic relatedness may represent compound- Additionally, the two Actinobacteria genera ing factors in our analysis of the host effect. observed to be enriched under control conditions in the Kearney data set are both reported to be capable of forming spores, suggesting that the Actinobacteria enrichment under drought ability to form spores does not necessarily lead to To our knowledge this represents the first major enrichment under drought conditions (Bergey’s, investigation of the role of drought in determining 2010); however, the evidence presented in this root endosphere microbiome composition. Here we study is based on reports in the literature and identified watering treatment as a significant factor bioinformatic analyses, and further experimental TheISMEJournal
Description: