Plant Phylogeny and Life History Shape Rhizosphere Bacterial Microbiome of Summer Annuals in an Agricultural Field.

Rhizosphere microbial communities are critically important for soil nitrogen cycling and plant productivity. There is evidence that plant species and genotypes select distinct rhizosphere communities, however, knowledge of the drivers and extent of this variation remains limited. We grew 11 annual species and 11 maize (Zea mays subsp. mays) inbred lines in a common garden experiment to assess the influence of host phylogeny, growth, and nitrogen metabolism on rhizosphere communities. Growth characteristics, bacterial community composition and potential activity of extracellular enzymes were assayed at time of flowering, when plant nitrogen demand is maximal. Bacterial community composition varied significantly between different plant species and genotypes. Rhizosphere beta-diversity was positively correlated with phylogenetic distance between plant species, but not genetic distance within a plant species. In particular, life history traits associated with plant resource acquisition (e.g., longer lifespan, high nitrogen use efficiency, and larger seed size) were correlated with variation in bacterial community composition and enzyme activity. These results indicate that plant evolutionary history and life history strategy influence rhizosphere bacterial community composition and activity. Thus, incorporating phylogenetic or functional diversity into crop rotations may be a tool to manipulate plant-microbe interactions in agricultural systems.

Introduction

The rhizosphere is a hotspot of plant-microbe interactions with profound influence on plant productivity and ecosystem function (Philippot et al., 2013). Shaped by the release of labile class="Chemical">carbon (C) from placlass="Chemical">nt roots aclass="Chemical">nd root uptake of class="Chemical">nutrieclass="Chemical">nts aclass="Chemical">nd class="Chemical">n class="Chemical">water (Hinsinger et al., 2005), the physiochemical environment of the rhizosphere supports a microbial community compositionally and metabolically distinct from that found in bulk soil (Mendes et al., 2014). The resulting rhizosphere microbiome performs critical functions, modulating plant growth and development (Panke-Buisse et al., 2015), plant health (Mendes et al., 2011; Berendsen et al., 2012), and plant nutrient acquisition (Philippot et al., 2013; Pii et al., 2015).

class="Chemical">Nitrogen (class="Chemical">n class="Chemical">N) is a limiting nutrient in most terrestrial ecosystems and plant–microbe interactions in the rhizosphere govern many N transformations in soil. The C-rich and N-limited environment of the rhizosphere is a site of associative N-fixation (James, 2000), and also frequently a site of increased decomposition and subsequent N mineralization of soil organic N pools (Kuzyakov, 2002; Herman et al., 2006). While the details regulating this “rhizosphere priming effect” are still poorly understood, it is broadly thought improved C status of the rhizosphere relieves energetic constraints on microbial activity and production of extracellular enzymes that breakdown soil organic matter (Averill and Finzi, 2011; Dijkstra et al., 2013). The activity of these enzymes is a rate limiting step in decomposition and subsequent N mineralization (Schimel and Bennett, 2004) and increased rates of N cycling that follow can feed back and support plant N acquisition (Hamilton and Frank, 2001; Zhu et al., 2014), particularly when coupled with the turnover or predation of microbial populations (Clarholm, 1985).

The importance of plant-microbial collaborations in plant nutrient acquisition presents an opportunity to modify crop-breeding approaches to select genotypes that foster rhizosphere microbiomes that can decrease the need for surplus additions of class="Chemical">N fertilizer (Driclass="Chemical">nkclass="Chemical">n class="Chemical">water and Snapp, 2007; Wissuwa et al., 2009). Therefore there is considerable interest in understanding the factors governing the assembly and function of the rhizosphere microbiome. An emerging picture suggests soil background is a dominant force in shaping bacterial community composition (BCC) in the rhizosphere. Within a soil context, plant species and genotypes influence this community (Berg and Smalla, 2009; Peiffer et al., 2013; Bulgarelli et al., 2015; Edwards et al., 2015), which can in turn be modulated by plant developmental stage and plant health status (Zhang et al., 2011; Chaparro et al., 2014; Marques et al., 2014). However, while some studies observe strong plant identity effects, others report no or limited effects (Wagner et al., 2016; Leff et al., 2017), and therefore understanding the sources and extent of plant-driven variation in the composition and function of the rhizosphere bacterial community remains a critical research challenge.

Presumably, variation in rhizosphere community composition and function is most likely driven by the evolutionary and ecological differentiation of host plants. For instance, several host–microbe interactions display a phylogenetic signal, such that closely related species share more similar microbiomes than distantly related species (Ley et al., 2008; Brucker and Bordenstein, 2012b). Such a phylogenetic signal has been observed in the rhizosphere of the Poaceae (Bouffaud et al., 2014), and in the phyllosphere of a broad range of plants, where increasing beta-diversity can be observed at the species, order and division levels (Redford et al., 2010). These patterns can arise from either specific co-evolutionary processes (Brucker and Bordenstein, 2012a) or, as proposed by Bouffaud et al. (2014), from microbiome assembly driven by the ability of phylogenetically conserved plant traits to shape microbial niche space in the rhizosphere.

Conversely, ecological differentiation among closely related hosts may interact with such a phylogenetic signal. For instance, diet is a significant driver of the class="Species">mammalian class="Chemical">n class="Species">gut microbiome and only after controlling for diet is a phylogenetic relationship between mammalian hosts and microbiome composition evident (Ley et al., 2008). Plant uptake of N and release of C are among several factors that shape the rhizosphere physiochemical environment (Hinsinger et al., 2005; Bell et al., 2015), therefore plant traits governing N and C acquisition and use may be strongly linked with plant variation in rhizosphere composition (Zancarini et al., 2013).

Plants aclass="Chemical">dapt to varyiclass="Chemical">ng levels of class="Chemical">n class="Chemical">N availability through their competitive ability to acquire N from soil and their nitrogen use efficiency (NUE) defined broadly as the amount of C fixed per unit plant N (Vitousek, 1982). Both strategies, N-acquisition and NUE, may affect rhizosphere communities. The rate of plant N uptake likely shapes plant-microbe competition for N. Correspondingly, differences in rhizosphere BCC have been observed between genotypes or plant species that differ in rates of N uptake (Moreau et al., 2015; Pathan et al., 2015), and these differences extend to indicators of N-cycling and extracellular enzyme activity (Cantarel et al., 2015; Pathan et al., 2015). Conversely, NUE is often associated with improved N retention in plant tissues (Berendse and Aerts, 1987), and plant traits promoting tissue longevity and N retention (e.g., increased tissue thickness, lignin content and decreased N content) are associated with decreased rates of decomposition and nutrient-cycling in soils under high NUE plants (Diaz et al., 2004; Orwin et al., 2010). These nutrient-cycling effects may be an indirect consequence of variation in litter quality, but it is also possible that these effects are mediated by direct plant impacts on microbiome composition and function.

To investigate the sources and extent of plant variation in rhizosphere effects, we conducted a common garden experiment with a selection of n class="Species">maize iclass="Chemical">nbred liclass="Chemical">nes aclass="Chemical">nd summer aclass="Chemical">nclass="Chemical">nual species commoclass="Chemical">nly fouclass="Chemical">nd iclass="Chemical">n agricultural systems. We characterized BCC aclass="Chemical">nd eclass="Chemical">nzyme activity iclass="Chemical">n placlass="Chemical">nt rhizospheres to test hypotheses that (1) placlass="Chemical">nt rhizosphere effects vary accordiclass="Chemical">ng to the evolutioclass="Chemical">nary history of host species, aclass="Chemical">nd (2) that variatioclass="Chemical">n iclass="Chemical">n rhizosphere BCC aclass="Chemical">nd metabolism is associated with variatioclass="Chemical">n iclass="Chemical">n placlass="Chemical">nt growth characteristics aclass="Chemical">nd class="Chemical">n class="Chemical">nitrogen economy.

Materials and Methods

Experimental Design

A common garden experiment was conducted at the Musgrave Research Station in Aurora, class="Chemical">NY (42°44′11′′class="Chemical">n class="Chemical">N 76°39′05′′W). The soil at the site is classified as fine-loamy, mesic Oxyaquic Hapludalfs, with a circumneutral pH of 7.65 and consisted of 45.2% sand, 33.5% silt and 21.3% clay. The soil was 1.7 ± 0.14% carbon and 0.17 ± 0.14% N; inorganic N (NH4 + NO3) content at tillage was 7.1 ± 1.3 μg g-1. Mehlich extractable P and K concentrations were 19.5 ± 1.5 μg g-1 and 146 ± 16 μg g-1, respectively. The field was previously managed as a corn-soy rotation and had been planted to corn in the previous year. Prior to planting the field was moldboard plowed, disked, fit for planting and fertilized with 224 kg ha-1 of potassium phosphate (0-15-30).

Plants were selected to encompass a range of intra- and interspecific diversity found in agricultural fields. This included ten founding inbred lines of the class="Species">maize (class="Chemical">n class="Species">Zea mays subsp mays L.) Nested Association Mapping (NAM) population, which represents the genetic diversity of improved maize (Yu et al., 2008), as described by Peiffer et al. (2013). Lines were chosen to represent differences in growth, N uptake and yield under fertilized and unfertilized conditions (Meyer, 2006). Additionally, one inbred line (75-062) was included from a public organic breeding program. We broadened phylogenetic and functional variation by including eight C4 grasses [Echinochloa crus-galli (L.) P.Beauv, Setaria faberi R.A.W.Herrm., Eragrostis tef (Zucc.) Trotter, Sorghum bicolor (L.) Moench subsp. bicolor, Sorghum X drummondii (Nees ex Steud.) Millsp. & Chase, and Eleusine coracana (L.) Gaertn.], four dicots (Abutilon theophrasti Medik., Amaranthus powellii S.Watson, Helianthus annuus L., Fagopyrum esculentum Moench), and a legume [Glycine max (L.) Merr.] (Supplementary Table S1).

Replicated monocultures were planted on June 19th and 21st, 2013 in a split-plot randomized complete block design (n = 4). Plots consisted of eight 1.83 m rows spaced at 76 cm, with 23 cm between plants in a row, resulting in a final density of 57,500 plants ha-1. Each main plot was split such that half the rows received a class="Chemical">nitrogen applicatioclass="Chemical">n of 23.5 kg class="Chemical">n class="Chemical">N ha-1 at planting and two side-dress applications (July 11th and August 5th) totaling 95 kg N ha-1 as (NH4)2SO4, while the remaining rows received no N fertilizer. This fertilizer level was chosen to boost plant growth but not provide luxury N conditions. Granular side-dress N was hand applied throughout the plot and incorporated during cultivation. Plots were kept weed free through mechanical cultivation and hand weeding.

Plant and Rhizosphere Sampling

Plants were harvested when at least 50% of the flowers/tassels for that genotype were shedding pollen. Since the phenology of these species are not synchronized this resulted in eight harvests (Supplementary Table S1). By sampling at a common developmental stage we control for the effects of plant developmental stage on rhizosphere BCC (Chaparro et al., 2014; Marques et al., 2014) and by sampling at anthesis, when plant biomass accumulation and nutrient uptake are maximal, we are able to evaluate rhizosphere composition when it is most relevant for nutrient uptake of each species. Three to four adjacent and representative plants from an interior row of each plot were clipped at the first nodal roots and dried at 60°C for dry weight determination. Homogenized and ground tissue was analyzed for tissue C and class="Chemical">N coclass="Chemical">nteclass="Chemical">nt oclass="Chemical">n a PDZ Europa Aclass="Chemical">n class="Chemical">NCA-GSL elemental analyzer at the University of California Davis Stable Isotope Facility.

At sampling, root systems were loosened from the ground with a spade and soil loosely adhered to the root system was removed by massaging and gentle shaking and discarded. Soil that remained adhered to the roots was considered rhizosphere soil and gently removed with a gloved hand, passed through a 2 mm sieve, and bagged for downstream analysis of class="Chemical">inorganic N coclass="Chemical">nteclass="Chemical">nt aclass="Chemical">nd poteclass="Chemical">ntial extracellular eclass="Chemical">nzyme activity. Additioclass="Chemical">nally, iclass="Chemical">ntact roots with adhericlass="Chemical">ng rhizospheres were sampled by clippiclass="Chemical">ng raclass="Chemical">ndomly selected 4 cm segmeclass="Chemical">nts of root class="Chemical">n class="Disease">tips and parent 2nd order roots for nucleic acid analysis. On each sampling date, 2 cm diameter by 20 cm deep soil cores were collected from unplanted, weed-free plots to represent bulk/bare soil in downstream analyses. Multiple cores were combined, homogenized, subsampled and passed through a 2 mm sieve. All samples were immediately placed on ice and then stored at 4°C for downstream analysis of enzymes and inorganic N content and at -40°C for nucleic acid analysis.

Extracellular Enzyme Analysis

Potential activity of enzymes involved in degradation of class="Chemical">hemi-cellulose [β-xylosidase (BX)], class="Chemical">n class="Chemical">cellulose [cellobiohydrolase (CB)], protein [leucine aminopeptidase (LAP)] and chitin [β-N-acetyl-glucosaminidase (NAG)] were measured using standard fluorometric assays following German et al. (2011). Briefly, 2–3 g field moist soil was mixed with 150 ml of 50 mM sodium bicarbonate buffer adjusted to pH 8 for 60 s using an immersion blender. 200 μl of soil slurry was added to 8 replicate wells of a 96-well plate containing 50 μl of 200 μM substrate with attached fluorophore. LAP plates were incubated for 2 h and BX, CB, and NAG incubated for 4 h at 30°C. Fluorescence was measured on a BioTek Synergy HT microplate reader at 365 nm excitation and 450 nm emission. Enzyme activity for each soil was estimated using a standard curve (0–75 μM) prepared from the same homogenate to control for quenching and autofluorescence. Standard curves were made fresh daily. To fix a perceived degradation of our standard over the season, the curves from each date were scaled so the maximum florescence of the 50 μM standard was equal across dates. All enzyme analyses were completed within 48 hrs of sample collection. Subsamples of soil were dried at 60°C for 48 h for soil moisture determination. Enzyme activity is expressed on a soil dry weight basis (nmol g soil-1 hr-1).

Inorganic Nitrogen Determination

A subset of species and genotypes were chosen to collect sufficient rhizosphere soil for class="Disease">inorganic N determination (Supplemeclass="Chemical">ntary Table S2). class="Chemical">n class="Chemical">Inorganic nitrogen was extracted in duplicate from 8 to 10 g of rhizosphere or bulk soil in 40 ml of 2 M KCl, shaken for 1 h and filtered through pre-rinsed ashless Whatman filter paper. Extracts were analyzed colorimetrically for nitrate and ammonium concentration using a VCl3/Griess method (Miranda et al., 2001) and modified indophenol method (Kandeler and Gerber, 1988), respectively, in a 96-well microplate format following Doane and Horwáth (2003) and Hood-Nowotny et al. (2010). Plates were incubated at 37°C for 2 h for nitrate determination and 30 min at 21°C for ammonium determination. Absorbance of wells was analyzed on a BioTek Synergy HT microplate reader at 540 and 660 nm, respectively. Concentrations were calculated using a standard curve included on each plate and expressed on a soil dry weight basis (μg N g soil-1).

16S rRNA Gene Sequence Analysis

Root and rhizosphere and bare soil samples stored at -40°C were lyophilized for 24 h on a class="Chemical">LabConco FreeZone 2.5 freeze dry system. Roots were chopped to <1 cm leclass="Chemical">ngth segmeclass="Chemical">nts, mixed, aclass="Chemical">nd betweeclass="Chemical">n 0.01 aclass="Chemical">nd 0.05 g of freeze dried roots aclass="Chemical">nd adhericlass="Chemical">ng soil or 0.15 g of bare soil coclass="Chemical">ntrols were added directly to each well of a 96-well extractioclass="Chemical">n plate from the MoBio PowerSoil-htp Dclass="Chemical">n class="Chemical">NA kit (Carlsbad, CA, United States). Rhizosphere samples were added to duplicate wells to adequately capture heterogeneity of the root systems. Samples were homogenized on a BioSpec Mini-Beadbeater-96 (Bartlesville, OK, United States) for 2 min and extractions proceeded according to manufacturer’s instructions. The 0.7 mm bead size of the PowerSoil kit does not homogenize root tissue and roots remained largely intact following homogenization (Peiffer et al., 2013). However, there was likely disruption of epidermal and some cortical cells and it is likely that our extracts contained some DNA from root endophytes. Following bead beating, extraction proceeded according to the kit manufacturer’s instructions. DNA yields were quantified with the Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen). Extractions yielded a mean of 375 ± 224 ng DNA template for use in downstream applications.

Dual-barcoded MiSeq libraries of the SSU rRclass="Chemical">NA V4 regioclass="Chemical">n were prepared as iclass="Chemical">n Kozich et al. (2013) usiclass="Chemical">ng the forward (515F) (Whitmaclass="Chemical">n et al., 2016) aclass="Chemical">nd reverse primers (806R) aclass="Chemical">n class="Chemical">dapted from Caporaso et al. (2010). Amplicons were prepared in triplicate reactions. Each reaction included 5 ng of template DNA, 12.5 μl of 2x Q5 High Fidelity, Hot Start PCR Mastermix, 1 μM combined forward and reverse primer, 0.5 μg bovine serum albumin and 0.625 μl of 4x PicoGreen reagent to monitor DNA template production for a total volume of 25 μl. PCR conditions consisted of: 95°C for 2 min; 30 cycles of 95°C for 20 s, 55°C for 15 s and 72°C for 10 s; final extension 72°C for 5 min. Pooled triplicate reactions were standardized using the SequalPrep Normalization Plate Kit (Life Technologies). Standardized reactions were pooled then gel purified and extracted using the Wizard SV Gel and PCR Clean-Up System (Promega). The two resulting amplicon libraries were submitted for 2 × 250 bp paired-end sequencing on the Illumina MiSeq platform with the MiSeq Reagent v2 kit at the Cornell Biotechnology Resource Center Genomics Facility (Ithaca, NY, United States).

Resulting reads were processed in a custom bioinformatics pipeline as in Whitman et al. (2016). Overlapping paired-end reads were merged using class="Species">PEAR (v0.9.2) (Zhaclass="Chemical">ng et al., 2014). Merged reads were de-multiplexed with a custom pythoclass="Chemical">n script aclass="Chemical">nd those that did class="Chemical">not match a kclass="Chemical">nowclass="Chemical">n barcode were discarded. Remaiclass="Chemical">niclass="Chemical">ng reads were filtered to remove sequeclass="Chemical">nces with max expected error rates > 1 with USEARCH (Edgar, 2013), ambiguous base calls, ≥8 homopolymers aclass="Chemical">nd siclass="Chemical">ngletoclass="Chemical">ns (uclass="Chemical">nique). Sequeclass="Chemical">nces were clustered iclass="Chemical">nto operatioclass="Chemical">nal taxoclass="Chemical">nomic uclass="Chemical">nits (OTUs) at a 97% pairwise ideclass="Chemical">ntity cutoff with USEARCH (Edgar, 2013). Taxoclass="Chemical">nomic assigclass="Chemical">nmeclass="Chemical">nt of OTUs was performed with Qiime’s parallel taxoclass="Chemical">nomy assigclass="Chemical">nmeclass="Chemical">nt usiclass="Chemical">ng the uclust coclass="Chemical">nseclass="Chemical">nsus taxoclass="Chemical">nomy assigclass="Chemical">nmeclass="Chemical">nt fuclass="Chemical">nctioclass="Chemical">n (Caporaso et al., 2010) aclass="Chemical">nd the class="Chemical">n class="Disease">Silva reference database (v.111) (Quast et al., 2013). OTUs belonging to chloroplast, mitochondria, eukaryotes, archaea and unassigned sequences were removed. OTUs were aligned using SSU_align and poorly aligned positions masked based on posterior probabilities (Nawrocki, 2009). A phylogenetic tree was created and rooted to Sulfolobus (acc. X90478) using FastTree (Price et al., 2009) with default settings. The resulting OTU table contained 11,246 OTUs representing 7,517,735 mapped reads and was combined with the phylogenetic tree, taxonomic information and metadata for analysis using the phyloseq package in R (McMurdie and Holmes, 2013). This OTU table was further filtered using a sparsity threshold of greater than three reads in more than three samples in order to remove extremely rare taxa, but retain taxa that may be endemic in the rhizosphere of a particular genotype, which resulted in 4982 OTUs. Sequences and associated metadata were deposited in the NCBI sequence read archive under accession #SRP119673.

Phylogeny and Genetic Distance Matrices

Chloroplast class="Gene">rbcL aclass="Chemical">nd matK geclass="Chemical">nes were used to coclass="Chemical">nstruct a phylogeclass="Chemical">ny of the twelve placlass="Chemical">nt species. Represeclass="Chemical">ntative sequeclass="Chemical">nces were dowclass="Chemical">nloaded from the Geclass="Chemical">nBaclass="Chemical">nk class="Chemical">n class="Chemical">Nucleotide Database. Amaranthus powellii was not represented and sequences from congeneric A. viridis were used instead. Sequences were aligned and checked in Unipro UGENE (Okonechnikov et al., 2012) and a tree was constructed using phyloGenerator with Ginkgo biloba used as an outgroup (Pearse and Purvis, 2013). A distance matrix was derived using the cophenetic.phylo function in the R package “ape” (Paradis et al., 2004). We expected little intraspecific variation in chloroplast rbcL or matK genes and arbitrarily assigned a distance of 0.0002 to intraspecific comparisons. This approximates intraspecific distances among maize lines found in previous studies using chloroplast markers (Bouffaud et al., 2014). Genetic distance matrices for the ten NAM inbred lines were constructed using GBS markers build 2.7 available at panzea.org. Distance matrices were estimated using TASSEL version 5 (Bradbury et al., 2007).

Statistical Analysis

Statistical analyses were conducted in R (R Development Core Team, 2012). To compare growth of plants sampled on different days, log-transformed plant biomass, class="Chemical">N uptake aclass="Chemical">nd class="Chemical">n class="Chemical">NUE (g C g N-1) were modeled by days after planting (DAP) (Supplementary Figure S1) for all species with one representative of maize (cv. B73). Data from an early season biomass cut was included to improve the model of plant growth over the course of the season (Supplementary Figure S1). Inbred maize lines were modeled separately in order to avoid weighting the model of plant growth. Residuals from the best-fit line were used to estimate variation in plant growth characteristics independent of flowering time.

Univariate tests were conducted in the package “class="Chemical">lme4” (Bates et al., 2015) aclass="Chemical">nd p-values estimated with “lmerTest” (Kuzclass="Chemical">netsova et al., 2016). Sample type, placlass="Chemical">nt geclass="Chemical">notype, fertilizatioclass="Chemical">n aclass="Chemical">nd iclass="Chemical">nteractioclass="Chemical">ns were coclass="Chemical">nsidered fixed effects with the raclass="Chemical">ndom effects of replicate block aclass="Chemical">nd split-class="Chemical">n class="Chemical">N fertilization plots. A similar mixed model was used to test the influence of rhizosphere inorganic N concentration on potential extracellular enzyme activity. Here, plant genotype was included as a random effect to control for influence of plant genotype and date of sampling on enzyme activity. Post hoc tests were conducted using the glht function in the “multcomp” package (Hothorn et al., 2008).

Bacterial community beta-diversity was analyzed using weighted-UniFrac distance matclass="Species">rices (Lozupoclass="Chemical">ne et al., 2011) coclass="Chemical">nstructed usiclass="Chemical">ng aclass="Chemical">n OTU table rarified to 4989 reads per sample iclass="Chemical">n the class="Chemical">n class="Chemical">phyloseq package in R (McMurdie and Holmes, 2013). Treatment effects on beta-diversity were tested using permutational multiple analysis of variance (PERMANOVA) using the “adonis” function in the Vegan package (Oksanen et al., 2012). We tested the effect of phylogenetic distance, maize whole genome genetic distance, and variation in growth characteristics between plant hosts on rhizosphere BCC using a generalized least squares implementation of Clarke’s maximum likelihood population effects model (MLPE) (Clarke et al., 2002), using the R function corMLPE[1]. The MLPE allows correlation between distance matrices by using a random effect parameter to estimate residual covariance of observations sharing a common sample, which would otherwise violate the assumption of independent observations (Clarke et al., 2002). To avoid pseudo-replication in the analysis of phylogenetic and genetic distance, the weighted-UniFrac distance matrix was calculated on OTU tables averaged over each genotype. In the models of phylogenetic and genetic distance, the sampling dates of each pairwise plant comparison was included as a fixed effect to control for variation between sampling dates. These models were evaluated using a likelihood ratio test against the nested null model of sampling date. Analyses of interspecific variation were conducted using all species with one representative of maize (cv. B73), while intraspecific analyses were conducted using the maize inbred lines. When not explicitly stated, analyses were conducted using all samples. To further explore the role of plant growth characteristics in shaping interspecific variation in rhizosphere BCC we constrained the principle coordinate analysis of weighted-UniFrac distances to display only variation that could be explained by plant growth metrics, using the “CAP” method of the “ordinate” function in phyloseq.

The response of individual OTUs to treatments and correlation with covariates was calculated as log2-fold change using non-rarified OTU table in a negative binomial model within the DESeq2 package (Love et al., 2014) and Benjamini and Hochberg corrected p-values reported. Rhizosphere responders were identified as those OTUs with a significant positive log2-fold change greater than 0.5 between a genotype’s rhizosphere and the bare soil controls sampled on the same date. When testing the role of rhizosphere n class="Chemical">inorganic N coclass="Chemical">nceclass="Chemical">ntratioclass="Chemical">n or fertilizatioclass="Chemical">n oclass="Chemical">n OTU abuclass="Chemical">ndaclass="Chemical">nce placlass="Chemical">nt geclass="Chemical">notype was iclass="Chemical">ncluded iclass="Chemical">n the model to coclass="Chemical">ntrol for variatioclass="Chemical">n betweeclass="Chemical">n placlass="Chemical">nts aclass="Chemical">nd betweeclass="Chemical">n sampliclass="Chemical">ng dates. All figures were created iclass="Chemical">n the package “ggplot2” (Wickham, 2009) except the circular phylogeclass="Chemical">netic tree, which was created usiclass="Chemical">ng the iclass="Chemical">nteractive tree of life (iTOL) web server (Letuclass="Chemical">nic aclass="Chemical">nd Bork, 2016). Ficlass="Chemical">nal aclass="Chemical">nclass="Chemical">notatioclass="Chemical">n aclass="Chemical">nd formatticlass="Chemical">ng of figures was performed iclass="Chemical">n Iclass="Chemical">nkscape. Scripts for bioiclass="Chemical">nformatics pipeliclass="Chemical">ne, aclass="Chemical">nalysis aclass="Chemical">nd figure geclass="Chemical">neratioclass="Chemical">n are available at https://github.com/bdemmett/RhizCG.

Results

Variation in Plant Growth and Nitrogen Economy

Across plant species and genotypes we observed nearly 10-fold variation in biomass accumulation and class="Chemical">N uptake, aclass="Chemical">nd 4-fold variatioclass="Chemical">n iclass="Chemical">n class="Chemical">n class="Chemical">NUE at anthesis (Supplementary Figure S2). This variation derived primarily from differences in flowering time and sampling date. Flowering time ranged from 36 DAP for F. esculentum to 88 DAP for E. coracana, and from 72 to 88 DAP for short and long season maize lines (Supplementary Table S1). As a result, DAP captured 69 and 77% of the variation in log-transformed N uptake and NUE among the plant species sampled (91 and 86% among maize inbred lines, respectively) (Supplementary Figure S1). While this is expected, as longer-lived plants have more time to grow and acquire N from soil, it highlights the variation in resource demand among annual plants. The longer-lived plants had both greater N demand and greater NUE owing to increased effective retention time.

Residuals from the models above were used to evaluate differences in plant growth and class="Chemical">N ecoclass="Chemical">nomy, iclass="Chemical">ndepeclass="Chemical">ndeclass="Chemical">nt of pheclass="Chemical">nology aclass="Chemical">nd lifespaclass="Chemical">n. Iclass="Chemical">n iclass="Chemical">nterspecific comparisoclass="Chemical">ns, we observed sigclass="Chemical">nificaclass="Chemical">nt variatioclass="Chemical">n iclass="Chemical">n total class="Chemical">n class="Chemical">N uptake (p < 0.01), NUE (p < 0.01), and corresponding differences in biomass accumulation (p < 0.01; Supplementary Table S3). These results are well illustrated by contrasting E. crus-galli and A. powellii, which had considerable differences in N uptake despite being harvested on the same date; and are also illustrated by contrasting E. tef and S. × drummondii, which had a twofold difference in N uptake despite their similar phenology (Supplementary Figure S2). This variation may be partly attributed to seed size. For instance, the extremely small-seeded E. tef had relatively low biomass and N uptake residuals. Yet this was not a consistent trend as S. × drummondii and H. annuus had comparable biomass accumulation and N uptake despite large differences in seed size (Supplementary Figure S2). In contrast to interspecific comparisons, we did not observe significant variation in N uptake among maize inbred lines (p = 0.24; Supplementary Table S3). Rather, differences in biomass accumulation between maize genotypes (p < 0.01) were associated with differences in NUE (p < 0.01; Supplementary Table S2). As expected, nitrogen fertilizer significantly improved plant growth, N uptake and also lowered plant NUE (p < 0.05; Supplementary Figure S2 and Supplementary Table S2).

Extracellular Enzyme Activity in the Rhizosphere

We observed a significant stimulation of hydrolytic enzyme activity in the rhizosphere (Figure and Supplementary Table S4). This rhizosphere effect was modulated by class="Chemical">nitrogen fertilizer additioclass="Chemical">n, whereby CB, BX, aclass="Chemical">nd class="Chemical">n class="Chemical">NAG activity increased in the rhizosphere of plants receiving fertilizer, but not in fertilized bare soil plots. In contrast, there was a trend toward increased LAP activity in both bare soil and rhizosphere samples that received fertilizer (Figure and Supplementary Table S4). The fertilizer effect exhibited a positive correlation between inorganic N concentration and potential enzyme activity in the rhizosphere (p < 0.05; Supplementary Table S5). Enzyme activity in the rhizosphere also differed between plant genotypes (p < 0.05; Supplementary Table S6), however, this result was only observed when comparing plants with different sampling dates. These differences between dates were not associated with a trend toward increasing or decreasing activity over the growing season (data not shown).

Potential activity of extracellular beta-xylosidase (BX), cellobiohydrolase (CB), leucine aminopeptidase (LAP), and class="Chemical">N-acetyl-glucosamiclass="Chemical">nidase (class="Chemical">n class="Chemical">NAG) in bare soil and rhizosphere samples from plots receiving 0 kg N ha-1 (gray bars) and 95 kg N ha-1 (black bars). Letters indicate a significant difference between treatments (Tukey HSD p < 0.05). Note that scale of y-axis differs among plots.

Rhizosphere Effect on Bacterial Community Composition

In addition to shifts in enzyme activity, we observed a strong differentiation of BCC between bare soil and rhizosphere sample types (Figure ). In a PERMAn class="Chemical">NOVA of weighted-Uclass="Chemical">niFrac distaclass="Chemical">nce, sample type was the greatest source of variatioclass="Chemical">n (Table aclass="Chemical">nd Figures ). Of the 4982 OTUs, 1502 were sigclass="Chemical">nificaclass="Chemical">ntly eclass="Chemical">nriched iclass="Chemical">n the rhizosphere of at least oclass="Chemical">ne placlass="Chemical">nt geclass="Chemical">notype compared with bare soils (Figure ). Maclass="Chemical">ny of the rhizosphere respoclass="Chemical">nsive OTUs were at low abuclass="Chemical">ndaclass="Chemical">nce iclass="Chemical">n bulk soil, but obtaiclass="Chemical">ned high abuclass="Chemical">ndaclass="Chemical">nce iclass="Chemical">n the rhizosphere (Figure ) resulticlass="Chemical">ng iclass="Chemical">n a dramatic shift iclass="Chemical">n commuclass="Chemical">nity compositioclass="Chemical">n. These rhizosphere respoclass="Chemical">nsive taxa iclass="Chemical">ncluded 54 OTUs that were preseclass="Chemical">nt iclass="Chemical">n rhizosphere samples but class="Chemical">not detected iclass="Chemical">n bulk soil (Figure ), which could result from rhizosphere eclass="Chemical">nrichmeclass="Chemical">nt of extremely rare taxa or vertical traclass="Chemical">nsmissioclass="Chemical">n of root eclass="Chemical">ndophytes.

Bacterial community composition varies between rhizosphere and bulk soil and with respect to days after planting (A), plant genotype (B), and class="Chemical">nitrogen fertilizatioclass="Chemical">n (C). Chaclass="Chemical">nges iclass="Chemical">n bacterial commuclass="Chemical">nity compositioclass="Chemical">n are visualized as a priclass="Chemical">ncipal coordiclass="Chemical">nate aclass="Chemical">nalysis (PCoA) of weighted-Uclass="Chemical">niFrac distaclass="Chemical">nces betweeclass="Chemical">n samples. Geclass="Chemical">notype codes represeclass="Chemical">nt class="Chemical">n class="Species">maize inbred lines and species: E. crus-galli (ECHCG), S. faberi (SETFA), S. bicolor (SORBI), S. x drummondii (SORSUD), E. coracana (ELCOR), E. tef (ERATE), A. theophrasti (ABUTH), H. annuus (HELAN), F. esculentum (FAGES), A. powellii (AMAPO), and G. max (GLYMA).

Permutational multiple analysis of variance testing main effects of sample type (rhizosphere vs. bare soil), days after planting (class="Chemical">DAP), placlass="Chemical">nt geclass="Chemical">notype or species ideclass="Chemical">ntity (geclass="Chemical">notype), aclass="Chemical">nd class="Chemical">n class="Chemical">nitrogen fertilization treatment (0, 95 kg N ha-1) on bacterial community beta-diversity (weighted-UniFrac).

Phylogeny of 500 most abundant bacterial taxa in common garden experiment. From inner circle outward: red tiles indicate taxa significantly enriched in the rhizosphere of at least one genotype compared to bare soil controls collected on the same date (DESeq2: log2-fold change > 0.5; adjusted p < 0.05), black bars indicate mean relative abundance of OTUs in bare soil samples, green bars indicate mean relative abundance of OTUs in rhizosphere samples from n class="Species">Zea mays cv. B73, choseclass="Chemical">n to represeclass="Chemical">nt the rhizosphere effect iclass="Chemical">n geclass="Chemical">neral. Tree created usiclass="Chemical">ng the iclass="Chemical">nteractive tree of life (iTOL) web server.

(A) Variation in OTU mean relative abundance from bare soil to rhizosphere of three representative species sampled on a single date. Black line indicates rank abundance in bare soil and colored traces indicate shifts in OTU relative abundance in rhizosphere samples. Black ticks indicate positive rhizosphere responders identified on any sampling date (DESeq2: log2-fold change > 0.5; adjusted p < 0.05), highlighting the enrichment in the rhizosphere of taxa at low abundance in bulk soil. (B) OTUs detected in bare soil and rhizosphere samples in full dataset.

Taxa enriched in the rhizosphere belonged to diverse phyla (Figure and Supplementary Table S7), but clustered within several groups. The Proteobacteria and Actinobacteria accounted for 23.6% and 16% of median relative abundance in rhizosphere samples and taxa from the Bacteroidetes, Chloroflexi, Verrucomicrobia, and Firmicutes also comprised a substantial fraction of the rhizosphere community (Figure and Supplementary Table S7). Within these phyla, some families showed a strong rhizosphere preference. The branching depth of clades sharing a phenotype can indicate the degree of phylogenetic conservation of that trait (Martiny et al., 2013). In our dataset, clades with more than 90% of OTUs displaying a rhizosphere response had a deeper average branching depth than expected under a permuted null model (Tau D = 0.02, p = 0.04; consenTRAIT), indicating phylogenetic conservation of the rhizosphere response within these families. These included families within the Proteobacteria (Comamonadaceae, Oxalobacteraceae, Caulobacteraceae, and Sphingomonadaceae), Actinobacteria (Streptomycetaceae), Firmicutes (Bacillacae and Paenibacillaceae), Bacteroidetes (Flavobacteriaceae and Chitinophagaceae), Verrucomicrobia (Opitutaceae), and Chloroflexi (Chloroflexaceae).

class="Chemical">Nitrogen fertilizatioclass="Chemical">n had a statistically sigclass="Chemical">nificaclass="Chemical">nt, but very small, iclass="Chemical">nflueclass="Chemical">nce oclass="Chemical">n BCC, accouclass="Chemical">nticlass="Chemical">ng for <1% of the variatioclass="Chemical">n iclass="Chemical">n the PERMAclass="Chemical">n class="Chemical">NOVA of rhizosphere samples, which is not easily perceptible in the ordination (Table and Figure ). Nitrogen fertilization led to the enrichment of 118 OTUs and decline of 45 OTUs in relative abundance within plant rhizospheres (Supplementary Figure S3; log2-fold change ≠ 0, p < 0.05). Many of the OTUs responding positively to N fertilization were from the Proteobacteria, Actinobacteria, and Bacteroidetes, though these phyla also had representatives that decreased in abundance in response to N fertilization (Supplementary Figure S3). Several Nitrospiraceae and a single Nitrosomonadaceae increased in abundance, while several other Nitrosomonadaceae decreased in abundance in rhizosphere samples from fertilized plots (Supplementary Figure S3). In comparison, relatively few OTUs were correlated with inorganic N concentration in the rhizosphere, which suggests that the effects of fertilization on BCC in the rhizosphere are indirect, being driven less by the availability of mineral N and more by changes in plant growth and physiology, which occur in response to fertilizer. This result is not unexpected since plants were sampled at anthesis and not immediately after fertilization. In addition, fertilization did not have a significant impact on BCC in bare soil (Table and Figure ), further emphasizing the role of plants in mediating the observed response of BCC to fertilization.

Plant Genotype Shapes Rhizosphere Bacterial Community Composition

Plant genotype and flowering time strongly shaped rhizosphere BCC as indicated by the PERMAn class="Chemical">NOVA (Table ). Variatioclass="Chemical">n betweeclass="Chemical">n geclass="Chemical">notypes sampled oclass="Chemical">n differeclass="Chemical">nt dates accouclass="Chemical">nted for 33% of the variatioclass="Chemical">n iclass="Chemical">n rhizosphere BCC (Table aclass="Chemical">nd Figure ). Sampliclass="Chemical">ng date also explaiclass="Chemical">ned a sigclass="Chemical">nificaclass="Chemical">nt portioclass="Chemical">n of variaclass="Chemical">nce iclass="Chemical">n bare soil samples (Table aclass="Chemical">nd Figure ), however, both the total variaclass="Chemical">nce aclass="Chemical">nd the proportioclass="Chemical">n of variaclass="Chemical">nce explaiclass="Chemical">ned by sampliclass="Chemical">ng date were greater iclass="Chemical">n rhizosphere samples thaclass="Chemical">n iclass="Chemical">n bulk soils (Figure ) aclass="Chemical">nd rhizosphere BCC varied dramatically from bulk soils (Figure ). Thus, the variatioclass="Chemical">n betweeclass="Chemical">n geclass="Chemical">notypes sampled oclass="Chemical">n differeclass="Chemical">nt dates caclass="Chemical">nclass="Chemical">not be explaiclass="Chemical">ned by temporal variatioclass="Chemical">n iclass="Chemical">n bulk soils aclass="Chemical">nd must be due to either placlass="Chemical">nt geclass="Chemical">notype effects or uclass="Chemical">nmeasured iclass="Chemical">nteractioclass="Chemical">ns betweeclass="Chemical">n placlass="Chemical">nt rhizosphere effects aclass="Chemical">nd time.

When added sequentially to the PERMAclass="Chemical">NOVA to coclass="Chemical">ntrol for sampliclass="Chemical">ng date, placlass="Chemical">nt geclass="Chemical">notype accouclass="Chemical">nted for 19% of the variatioclass="Chemical">n iclass="Chemical">n rhizosphere samples (Table ). Iclass="Chemical">n this model, some placlass="Chemical">nt geclass="Chemical">notype effects are attributed to sampliclass="Chemical">ng date, which provides a coclass="Chemical">nservative estimate of placlass="Chemical">nt geclass="Chemical">notype iclass="Chemical">nflueclass="Chemical">nce oclass="Chemical">n the rhizosphere commuclass="Chemical">nity. It is possible to elimiclass="Chemical">nate sampliclass="Chemical">ng date effects oclass="Chemical">n those dates wheclass="Chemical">n multiple geclass="Chemical">notypes were sampled. Wheclass="Chemical">n assessed withiclass="Chemical">n a sampliclass="Chemical">ng date, placlass="Chemical">nt geclass="Chemical">notype explaiclass="Chemical">ned betweeclass="Chemical">n 13% (p = 0.09) aclass="Chemical">nd 43% (p < 0.01) of variaclass="Chemical">nce iclass="Chemical">n rhizosphere BCC (Supplemeclass="Chemical">ntary Table S8). Geclass="Chemical">notype was also aclass="Chemical">n importaclass="Chemical">nt predictor of iclass="Chemical">ntraspecific variatioclass="Chemical">n iclass="Chemical">n rhizosphere BCC. Wheclass="Chemical">n evaluaticlass="Chemical">ng just the class="Chemical">n class="Species">maize inbred lines, sampling date accounted for 10% of the variance [F(2,76) = 6.44; p < 0.01] and plant genotype explained an additional 26% [F(8,76) = 3.95; p < 0.01] of the variance in rhizosphere BCC.

Plant Phylogeny and Growth Characteristics Explain Differences in Rhizosphere Bacterial Community Composition

To evaluate the influence of plant phylogeny on rhizosphere BCC we constructed an MLPE model using a weighted-UniFrac distance matrix of average OTU relative abundance for each plant genotype. Beta-diversity increased in relation to plant phylogenetic distance through the taxonomic rank of family (Figures ), explaining 8% of the variation between plant rhizosphere communities ( = 19.17, p < 0.01). This effect was robust when controlling for variation in BCC attributed to pairwise sampling date comparisons ( = 13.50, p < 0.01) and when intraspecific comparisons were removed from the dataset ( = 7.93, p < 0.01). Beta-diversity apclass="Species">peared to plateau beyoclass="Chemical">nd the raclass="Chemical">nk of family, such as wheclass="Chemical">n compariclass="Chemical">ng betweeclass="Chemical">n grasses aclass="Chemical">nd class="Chemical">n class="Species">dicots, suggesting that plants with similar characteristics had similar effects on rhizosphere BCC. The phylogenetic signal was evident in overall beta-diversity and at the level of individual OTUs (Figure ). Non-maize species had more differentially abundant OTUs than maize genotypes when compared to maize reference line B73 (DESeq2: log2-fold change ≠ 0; BH adjusted p < 0.05). In addition, changes in relative abundance for the differentially abundant OTUs were greater for non-maize species than for other maize inbred lines when compared to maize reference line B73 (Figure and Supplementary Figure S4).

Plant phylogeny and flowering time influence rhizosphere bacterial community composition. Rhizosphere bacterial community beta-diversity (weighted-UniFrac distance) is positively correlated with plant host phylogenetic distance (A), days between sampling (B), and increases with plant host taxonomic rank (C). Colored bars indicate correspondence of plant phylogenetic distances and taxonomic rank between (A,C). Influence of time and plant phylogeny on bacterial taxa abundance is illustrated with two Streptomyces OTUs (D). OTUs increase in abundance over time as a result of selective enrichment in n class="Species">maize (top paclass="Chemical">nel) or eclass="Chemical">nrichmeclass="Chemical">nt over time iclass="Chemical">ndepeclass="Chemical">ndeclass="Chemical">nt of placlass="Chemical">nt phylogeclass="Chemical">ny (bottom paclass="Chemical">nel) (DESeq2: log2-fold chaclass="Chemical">nge per day = 0.08 ± 0.01 aclass="Chemical">nd 0.07 ± 0.01 iclass="Chemical">n top aclass="Chemical">nd bottom paclass="Chemical">nels, respectively; p < 0.05). Weighted-Uclass="Chemical">niFrac distaclass="Chemical">nces calculated oclass="Chemical">n meaclass="Chemical">n geclass="Chemical">notype OTU abuclass="Chemical">ndaclass="Chemical">nces (A) aclass="Chemical">nd plot level OTU abuclass="Chemical">ndaclass="Chemical">nces (B,C).

Differential abundance between rhizosphere of class="Species">Zea mays cv. B73 aclass="Chemical">nd other class="Chemical">n class="Species">maize genotypes and annual species. Tiles arranged by plant phylogeny (top tree) and bacterial phylogeny (left tree). Colored tiles indicate significant difference between listed genotype and reference B73 (DESeq2: log2-fold change ≠ 0, adjusted p < 0.05), color and intensity indicate direction and magnitude of log2-fold change. Green bars represent mean relative abundance in rhizosphere samples of B73. Genotype codes represent maize inbred lines and species: E. crus-galli (ECHCG), S. faberi (SETFA), S. bicolor (SORBI), S. x drummondii (SORSUD), E. coracana (ELCOR), E. tef (ERATE), A. theophrasti (ABUTH), H. annuus (HELAN), F. esculentum (FAGES), A. powellii (AMAPO), and G. max (GLYMA).

Flowering time, which explained most variation in plant growth and class="Chemical">N ecoclass="Chemical">nomy (Supplemeclass="Chemical">ntary Figure S1), was also the best coclass="Chemical">nticlass="Chemical">nuous predictor of beta-diversity iclass="Chemical">n plot level data. Rhizosphere beta-diversity iclass="Chemical">ncreased with time betweeclass="Chemical">n sampliclass="Chemical">ng dates iclass="Chemical">n a polyclass="Chemical">nomial fashioclass="Chemical">n (Figure ), accouclass="Chemical">nticlass="Chemical">ng for approximately 28% of the variatioclass="Chemical">n iclass="Chemical">n rhizosphere BCC ( = 77.63, p < 0.01). Beta-diversity also iclass="Chemical">ncreased with differeclass="Chemical">nces iclass="Chemical">n placlass="Chemical">nt class="Chemical">n class="Chemical">NUE and seed size ( = 37.08, p < 0.01), but these factors only explained an additional 2% of the variation in rhizosphere BCC. The effect of sampling date could have multiple drivers including differences in physiology of plants with different lifespan, seasonal variation in soil characteristics, and temporal autocorrelation between sampling dates. There was also a relationship between flowering time and plant phylogenetic distance (r = 0.56, p < 0.01). Maize and many of the C4 grasses flowered later in the season while four of five dicots flowered early in the season. As a result, temporal variation in the relative abundance of OTUs will be driven both by plant species specific rhizosphere effects and by temporal variation in background soils. We highlight two Streptomyces OTUs to illustrate these patterns (Figure ). One is responsive to maize and related crop plants in the subfamily Andropogoneae (S. bicolor and Sorghum × drummondii). This maize responsive OTU is found in highest relative abundance during anthesis for maize and Sorghum, but it remains in low abundance in the rhizospheres of other plants sampled on these same dates (Figure , top panel). In contrast, a second OTU from Streptomyces increases in abundance less specifically, responding to a range of plant genotypes including both C4 grasses and H. annuus (Figure , bottom panel).

Intraspecific variation in rhizosphere BCC was not correlated with genetic distance ( = 1.51, p = 0.22). class="Chemical">Nor was beta-diversity correlated with variatioclass="Chemical">n iclass="Chemical">n fuclass="Chemical">nctioclass="Chemical">nal measures iclass="Chemical">ncludiclass="Chemical">ng flowericlass="Chemical">ng time, placlass="Chemical">nt class="Chemical">n class="Chemical">NUE or N uptake (p > 0.05).

Changes in Bacterial Community Composition and Activity Associated with Plant Resource Acquisition and Use Strategies

To further explore the role of plant growth characteristics in shaping variation in rhizosphere BCC between species, a principal coordinate ordination of weighted-UniFrac distances was constrained by explanatory growth characteristics including: days to flowering, seed size, class="Chemical">N uptake aclass="Chemical">nd class="Chemical">n class="Chemical">NUE. The constrained ordination explained roughly 22% of the variation in BCC (p < 0.01) (Figure and Supplementary Table S9). The primary axis was negatively correlated with traits defining a resource intensive life history: longer lifespan, larger seed size and higher NUE. Notably, grasses and dicots were intermixed along this axis, with H. annuus, maize, and Sorghum occupying one end of the spectrum while E. crus-galli and A. theophrasti occupied the opposite end. Additionally, flowering time was not the sole driver of differentiation (Supplementary Table S9). Instead, long-lived but low NUE plants such as E. coracana and E. tef grouped with early season E. crus-galli to the exclusion of late season but high NUE plants such as maize and Sorghum. This axis captured marked compositional changes in rhizosphere BCC, characterized by the enrichment of many Actinobacteria OTUs in association with longer season and higher NUE plants (Figure ). The second CAP axis represented variation in total plant N uptake independent of flowering time and explained a small portion (3.7%) of variance in rhizosphere BCC. This axis was correlated with several Bacillus OTUs as well as a few Acidobacteria and Cyanobacteria. Neither axis separated the legume, G. max, from the other species, possibly indicating the plant’s life history traits were more important to its placement on this axis than its ability to fix nitrogen.

Relationship between plant life history strategy and bacterial community composition (BCC) in the rhizosphere. (A) Constrained analysis of principle coordinates (CAP) displaying variation in BCC explained by plant growth characteristics: days to flowering, seed size, plant class="Chemical">nitrogen use efficieclass="Chemical">ncy (class="Chemical">n class="Chemical">NUE) (g C g N-1) and nitrogen uptake (g N). Numbers in parentheses alongside species names in figure legend refer to sampling date as days after planting. Nitrogen use efficiency and N uptake are corrected for differences in sampling date by using residuals of model relating growth characteristics to date of harvest. (B) Estimates of log2-fold change in bacterial OTU abundance per unit shift in sample score on CAP1. Points colored in red are OTUs significantly correlated with CAP1 (DESeq2: log2-fold change ≠ 0; adjusted p < 0.05), gray points indicate OTUs not significantly correlated with CAP1 (p > 0.05). Point size proportional to relative abundance.

The correspondence of rhizosphere BCC with plant growth characteristics coincided with shifts in enzyme activity in the rhizosphere. The potential activity of BX, LAP, and n class="Chemical">NAG were class="Chemical">negatively correlated with sample scores oclass="Chemical">n the primary CAP axis, while the secoclass="Chemical">ndary axis was positively correlated with the poteclass="Chemical">ntial activity of BX aclass="Chemical">nd CB (Table ). This ficlass="Chemical">ndiclass="Chemical">ng liclass="Chemical">nks placlass="Chemical">nt growth characteristics to variatioclass="Chemical">n iclass="Chemical">n BCC aclass="Chemical">nd eclass="Chemical">nzyme activity such that placlass="Chemical">nts with resource iclass="Chemical">nteclass="Chemical">nsive life history traits had higher eclass="Chemical">nzyme activity aclass="Chemical">nd differeclass="Chemical">nces iclass="Chemical">n BCC compared to placlass="Chemical">nts with less resource iclass="Chemical">nteclass="Chemical">nsive life history traits.

Correlations between principle coordinates that explain plant life history and bacterial community composition relationships and potential activity of extracellular beta-xylosidase (BX), cellobiohydrolase (CB), leucine aminopeptidase (LAP) and class="Chemical">N-acetyl-glucosamiclass="Chemical">nidase (class="Chemical">n class="Chemical">NAG) and inorganic N concentration in the rhizosphere.

Discussion

In a common garden experiment, we investigated the sources and extent of plant variation in rhizosphere community composition and activity. We observed distinct changes in BCC and enzyme activity, reflecting the different C and n class="Chemical">N status betweeclass="Chemical">n rhizosphere aclass="Chemical">nd bulk soil. Withiclass="Chemical">n this coclass="Chemical">ntext, we show that rhizosphere BCC aclass="Chemical">nd eclass="Chemical">nzyme activity is modulated by placlass="Chemical">nt species aclass="Chemical">nd geclass="Chemical">notype aclass="Chemical">nd that this effect is related to placlass="Chemical">nt phylogeclass="Chemical">ny aclass="Chemical">nd life history strategy.

Rhizosphere Effect on Bacterial Community Composition and Metabolism

Shifts in BCC and enzyme activity from bare to rhizosphere soils reflect the altered energy status of the rhizosphere environment. Consistent with other studies, rhizosphere samples were dominated by Proteobacteria and Actinobacteria (Bulgarelli et al., 2012, 2015; Peiffer et al., 2013), which include many bacterial species that grow rapidly in response to the availability of labile n class="Chemical">carbon substrates (Goldfarb et al., 2011). Additioclass="Chemical">nally, maclass="Chemical">ny OTUs eclass="Chemical">nriched iclass="Chemical">n the rhizosphere were phylogeclass="Chemical">netically clustered aclass="Chemical">nd fouclass="Chemical">nd at low abuclass="Chemical">ndaclass="Chemical">nce iclass="Chemical">n bare soil, which suggests that rhizosphere competeclass="Chemical">nce requires traits that are evolutioclass="Chemical">narily coclass="Chemical">nserved aclass="Chemical">nd which may class="Chemical">not be aclass="Chemical">n class="Chemical">daptive in bulk soil (Barret et al., 2011; Ofek-Lalzar et al., 2014; Shi et al., 2015).

The potential activity of class="Chemical">cellulose, class="Chemical">n class="Chemical">hemi-cellulose, protein and chitin degrading enzymes was consistently greater in the rhizosphere compared to bare soil, which is consistent with studies showing a positive rhizosphere effect on enzyme activity, SOM decomposition, and N mineralization in the rhizosphere (Herman et al., 2006; Zhu et al., 2014). Controls on enzyme production in soil can include nutrient demand, target substrate availability, energetic constraints, and nutrient constraints (Sinsabaugh and Moorhead, 1994; Allison and Vitousek, 2005; Geisseler and Horwath, 2008). Accordingly, increased enzyme activity in the rhizosphere could reflect substrate flow from plant roots and release of C limitation. Nitrogen fertilizer further increased activity in the rhizosphere for all enzymes assayed, which was not observed in bare soil. These results could indicate that microbes experience greater N limitation in the rhizosphere, or that labile C from the plant roots is necessary to take advantage of the increased nutrient availability (Averill and Finzi, 2011).

A surprising result is that, while fertilizer addition increased plant growth and rhizosphere enzyme activity, its addition explained little variation in BCC relative to the effects of plant species and genotype. There are well documented effects of class="Chemical">N fertilizatioclass="Chemical">n oclass="Chemical">n soil BCC (Ramirez et al., 2012; Leff et al., 2015) aclass="Chemical">nd evideclass="Chemical">nce that class="Chemical">n class="Chemical">N fertilization can shift rhizosphere BCC (Zancarini et al., 2012). Inorganic N fertilizer can influence BCC through a variety of mechanisms including: immediate direct responses to inorganic N availability (Verhamme et al., 2011), short term indirect responses caused by the effect of fertilizer on plant growth (Paterson et al., 2006), and long term indirect effects of fertilizer on soil properties such as pH (Hallin et al., 2009). Furthermore, these mechanisms may interact such that short-term effects vary depending on the fertilization history of the site. We propose two explanations for the minimal effect of N fertilization on BCC that we observed. First, temporal decoupling between fertilizer application and sampling may minimize detection of direct fertilization effects on BCC. Second, we propose that long-term use of mineral fertilizer at this site has minimized the responsiveness of BCC to short term fertilization effects. We note that those OTUs that increased in abundance in response to fertilizer were not tightly coupled with inorganic N concentration in soil. This result suggests that the fertilizer effects we did observe on BCC were mediated indirectly by plant response to fertilizer, such as increased root growth and exudation. This would explain why fertilization enhanced enzyme activity without causing substantial changes in rhizosphere BCC and why fertilizer had little effect on BCC and enzyme activity in bare soils.

Plant Identity Shapes Rhizosphere Bacterial Community

Ourclass="Chemical">n clclass="Chemical">ass="Chemicclass="Chemical">al"> class="Chemical">an>ficlass="Chemical">nding class="Chemical">thaclass="Chemical">n class="Chemical">t 13–43% of the variation in beta-diversity on a single sampling date could be attributed to plant species or genotype is consistent with previous reports of variation within a single field (Peiffer et al., 2013; Edwards et al., 2015), yet some studies have reported little or no plant identity effect on rhizosphere BCC (Bulgarelli et al., 2012; Wagner et al., 2016; Leff et al., 2017). To some degree these conflicting reports are expected. Genotype influences are less apparent in analyses where multiple fields or sample types increase total variance of the BCC (Peiffer et al., 2013; Bulgarelli et al., 2015; Edwards et al., 2015). Heritable plant phenotypes that influence rhizosphere communities may also be most influential during specific growth stages (İnceoğlu et al., 2010) or within a particular soil context. For example, Bell et al. (2014) observed that willow cultivars grown in contaminated soils selected distinct rhizosphere fungal communities, while those grown in non-contaminated soils did not. By situating our study in a single field it is implicit that the genotype effects we observed may not always emerge. However, reduced environmental variation allows a deeper look at the factors driving plant genotype and species variation in rhizosphere BCC. Here we investigated the strength of two plant factors—plant evolutionary history and variation in growth and N economy—in predicting variation in rhizosphere BCC.

Plant Phylogeny Shapes Rhizosphere Bacterial Community

Plant evolutionary history explained a significant portion of variation in rhizosphere BCC. This adds to a growing body of studies detailing a link between host phylogeny and microbiome composition (Ley et al., 2008; Redford et al., 2010; Brucker and Bordenstein, 2012b). Similar to Bouffaud et al. (2014) our experimental design centered around class="Species">maize aclass="Chemical">nd the Poaceae. Here we democlass="Chemical">nstrate that a phylogeclass="Chemical">netic sigclass="Chemical">nal is evideclass="Chemical">nt iclass="Chemical">n the rhizosphere of field growclass="Chemical">n placlass="Chemical">nts, whereby iclass="Chemical">ncreasiclass="Chemical">ng phylogeclass="Chemical">netic distaclass="Chemical">nce leads to a more dissimilar bacterial commuclass="Chemical">nity. This relatioclass="Chemical">nship has importaclass="Chemical">nt implicatioclass="Chemical">ns. First, it suggests the phylogeclass="Chemical">netic coclass="Chemical">nservatioclass="Chemical">n of placlass="Chemical">nt traits that iclass="Chemical">nflueclass="Chemical">nce BCC. As discussed by Bouffaud et al. (2014), several traits that exhibit phylogeclass="Chemical">netic coclass="Chemical">nservatism are likely to iclass="Chemical">nflueclass="Chemical">nce rhizosphere commuclass="Chemical">nities. For iclass="Chemical">nstaclass="Chemical">nce, root morphology displays a phylogeclass="Chemical">netic sigclass="Chemical">nal coiclass="Chemical">ncideclass="Chemical">nt with mycorrhizal associatioclass="Chemical">n (Bruclass="Chemical">ndrett, 2002; Comas et al., 2014). Iclass="Chemical">n additioclass="Chemical">n, secoclass="Chemical">ndary metabolite pathways, which may serve as sigclass="Chemical">nals iclass="Chemical">n host–microbe commuclass="Chemical">nicatioclass="Chemical">n, are ofteclass="Chemical">n coclass="Chemical">nserved at the family level (Wiclass="Chemical">nk, 2003). Furthermore, host immuclass="Chemical">ne respoclass="Chemical">nses directly iclass="Chemical">nflueclass="Chemical">nce the compositioclass="Chemical">n of the root microbiome aclass="Chemical">nd caclass="Chemical">n be coclass="Chemical">nserved phylogeclass="Chemical">netically (Brucker aclass="Chemical">nd Bordeclass="Chemical">nsteiclass="Chemical">n, 2012b; Lebeis et al., 2015). A secoclass="Chemical">nd implicatioclass="Chemical">n is that the phylogeclass="Chemical">netic structure of placlass="Chemical">nt commuclass="Chemical">nities would be expected to cause loclass="Chemical">ng-term chaclass="Chemical">nges iclass="Chemical">n soil BCC (Barberáclass="Chemical">n et al., 2015). This may serve as a mechaclass="Chemical">nism uclass="Chemical">nderlyiclass="Chemical">ng the positive relatioclass="Chemical">nship betweeclass="Chemical">n placlass="Chemical">nt phylogeclass="Chemical">netic diversity aclass="Chemical">nd ecosystem fuclass="Chemical">nctioclass="Chemical">ns (Flyclass="Chemical">nclass="Chemical">n et al., 2011). Iclass="Chemical">ntroduciclass="Chemical">ng phylogeclass="Chemical">netic diversity to agricultural systems, either through rotatioclass="Chemical">ns or iclass="Chemical">ntercroppiclass="Chemical">ng, could therefore represeclass="Chemical">nt a maclass="Chemical">n class="Chemical">nagement tool to influence rhizosphere and soil BCC and ultimately influence nutrient cycling in these soils (Berthrong et al., 2013).

Plant Life History Strategy Shapes Rhizosphere Bacterial Community

Beta-diversity in rhizosphere BCC varied in response to plant flowering time, seed size, and life-span independent variation in class="Chemical">NUE. These observatioclass="Chemical">ns are coclass="Chemical">nsisteclass="Chemical">nt with the hypothesis that variatioclass="Chemical">n iclass="Chemical">n placlass="Chemical">nt growth aclass="Chemical">nd class="Chemical">n class="Chemical">N economy influence rhizosphere microbiome composition. Within annual agricultural fields, where fitness is limited to ruderal plants with high growth rates, life history strategies are primarily differentiated by lifespan, which leads to differences in plant biomass accumulation, N demand, and NUE. These terms, along with seed size, reflect key dimensions of plant form and function (Moles and Westoby, 2006; Díaz et al., 2016) and could be linked with variation in rhizosphere BCC through multiple mechanisms.

Variation in class="Chemical">N-uptake betweeclass="Chemical">n placlass="Chemical">nt species has stroclass="Chemical">ng impacts oclass="Chemical">n class="Chemical">n class="Chemical">nitrogen cycling dynamics in soil (Tilman and Wedin, 1991), and may contribute to variation in rhizosphere BCC between plant genotypes (Bell et al., 2015; Moreau et al., 2015; Pathan et al., 2015). In our study, extended N uptake of longer-lived plants may exacerbate N limitation within rhizosphere bacterial communities. It is possible that actinobacterial OTUs, enriched in long-lived and high NUE plants, such as H. annuus, maize and Sorghum, have adaptations to withstand N limitation in the rhizosphere. Actinomycetes produce a range of extracellular enzymes to degrade organic matter in soil (McCarthy and Williams, 1992). This could provide access to soil N pools in an otherwise N limited environment and underlie the increase in putative N-accessing enzymes observed in the rhizosphere of longer-lived, high NUE plants.

Alternately, plant traits correlated with lifespan and class="Chemical">NUE may alter BCC. For example, placlass="Chemical">nts classified as class="Chemical">n class="Chemical">nitrogen competitive (high uptake) or conservative (high NUE) have been found to vary in the quantity and composition of their root exudates (Kaštovská et al., 2015; Guyonnet et al., 2017). In turn, species with higher rates of exudation supported increased microbial growth, turnover and high rates of N transformations in the rhizosphere (Blagodatskaya et al., 2014; Kaštovská et al., 2015). Thus, it is possible that shifts in rhizosphere C flows in long-lived high NUE plants alter the rhizosphere bacterial community. There is a rich literature connecting plant growth strategies to litter quality and subsequent impacts on nitrogen cycling in soil (Diaz et al., 2004; Hawkes et al., 2005; Orwin et al., 2010). Our findings suggest that plant life history strategy can also have direct impacts on rhizosphere BCC and activity.

By sampling at the onset of flowering, we captured a primary dimension of plant variation while limiting the effects of plant development on BCC. Since the phenology of these species was not synchronized, we cannot rule out that temporal shifts in eclass="Chemical">daphic factors coclass="Chemical">ntributed to our results. class="Chemical">n class="Chemical">Nevertheless, if seasonal effects, rather than endogenous plant effects, are the source of changes in rhizosphere BCC observed here, they remain directly related to realized rhizosphere communities as they impact and interact with plants in the field. Furthermore, the grouping of long-lived, low NUE species with short-lived species in the constrained ordination supports the interpretation of a strong plant life history mediated effect on rhizosphere bacterial communities. Sequential sampling or staggered plantings to synchronize developmental stage (e.g., Wagner et al., 2016) in similar field experiments will be necessary to disentangle the interrelated effects of plant variation in growth, life history, and temporal variation.

class="Chemical">Neither geclass="Chemical">netic relatedclass="Chemical">ness class="Chemical">nor growth aclass="Chemical">nd class="Chemical">n class="Chemical">N economy successfully described intraspecific variation among maize lines, despite differences in rhizosphere BCC between genotypes. In this regard our results are consistent with previous work where genetic distance, plant height and plant size have not predicted intraspecific variation in rhizosphere BCC (Peiffer et al., 2013; Leff et al., 2017). In contrast, ecophysiological measures related to carbon and nitrogen acquisition did parse variation in rhizosphere BCC between Medicago genotypes (Zancarini et al., 2013). While it seems clear from our data that different plant species have different impacts on rhizosphere BCC, which are associated with differences in life history traits and rhizosphere function, it is less clear how intraspecific variation in plant traits influences microbiome composition and function.

Conclusion

We demonstrate that both plant phylogeny and life history traits, including variation in lifespan, growth and class="Chemical">N ecoclass="Chemical">nomy, explaiclass="Chemical">n sigclass="Chemical">nificaclass="Chemical">nt variatioclass="Chemical">n iclass="Chemical">n rhizosphere BCC aclass="Chemical">nd eclass="Chemical">nzyme activity. These results suggest that differeclass="Chemical">nces iclass="Chemical">n placlass="Chemical">nt fuclass="Chemical">nctioclass="Chemical">nal traits drive variatioclass="Chemical">n iclass="Chemical">n BCC aclass="Chemical">nd impact resource acquisitioclass="Chemical">n from soil, which likely has both short aclass="Chemical">nd loclass="Chemical">ng-term coclass="Chemical">nsequeclass="Chemical">nces for soil BCC aclass="Chemical">nd class="Chemical">n class="Chemical">N-cycling dynamics. Crop selection, cover cropping and crop rotation are key management interventions in below ground processes in agricultural systems. If the rhizosphere phenotypes observed in this study are repeatable in other fields, then incorporating phylogenetic and functional diversity into crop rotations may provide a mechanism to manipulate plant-microbe interactions over time. Fully understanding the implications of plant-induced shifts in the rhizosphere and soil microbiome will be critical in selecting plants and beneficial rotations for maximal agronomic benefit.

Author Contributions

BE, LD, and DB designed the research. BE performed the research. BE and n class="Chemical">NY performed the sample aclass="Chemical">nd data aclass="Chemical">nalysis. BE, LD, aclass="Chemical">nd DB wrote the maclass="Chemical">nuscript with assistaclass="Chemical">nce from class="Chemical">n class="Chemical">NY.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Table 1

Permutational multiple analysis of variance testing main effects of sample type (rhizosphere vs. bare soil), days after planting (DAP), plant genotype or species identity (genotype), and nitrogen fertilization treatment (0, 95 kg N ha-1) on bacterial community beta-diversity (weighted-UniFrac).

Factor	SS	DF	F	R2	p∗
Full dataset
Sample type	2.24	1	141.49	0.25	<0.01
DAP	1.47	7	13.20	0.17	<0.01
Genotype	1.80	21	5.39	0.20	<0.01
N treatment	0.04	1	2.65	0.005	0.03
Residuals	3.28	207		0.40
Rhizosphere
DAP	1.97	7	14.87	0.33	<0.01
Genotype	1.16	14	4.37	0.19	<0.01
N treatment	0.06	1	2.92	0.01	<0.01
Residuals	2.85	151		0.47
Bare soil
DAP	0.13	1	2.61	0.25	<0.01
N treatment	0.009	1	1.19	0.02	0.19
Residuals	0.41	61		0.73

Table 2

Correlations between principle coordinates that explain plant life history and bacterial community composition relationships and potential activity of extracellular beta-xylosidase (BX), cellobiohydrolase (CB), leucine aminopeptidase (LAP) and N-acetyl-glucosaminidase (NAG) and inorganic N concentration in the rhizosphere.

	Pearson correlation coefficients
	BX	CB	LAP	NAG	Inorganic N
CAP1∗	-0.38	-0.26	-0.55	-0.29	0.35
CAP2	0.51	0.53	0.10	-0.14	-0.11