Author response: Association analyses of host genetics, root-colonizing microbes, and plant phenotypes under different nitrogen conditions in maize
Year: 2022
Type: peer-review
Abstract: Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Appendix 1 Data availability References Decision letter Author response Article and author information Metrics Abstract The root-associated microbiome (rhizobiome) affects plant health, stress tolerance, and nutrient use efficiency. However, it remains unclear to what extent the composition of the rhizobiome is governed by intraspecific variation in host plant genetics in the field and the degree to which host plant selection can reshape the composition of the rhizobiome. Here, we quantify the rhizosphere microbial communities associated with a replicated diversity panel of 230 maize (Zea mays L.) genotypes grown in agronomically relevant conditions under high N (+N) and low N (-N) treatments. We analyze the maize rhizobiome in terms of 150 abundant and consistently reproducible microbial groups and we show that the abundance of many root-associated microbes is explainable by natural genetic variation in the host plant, with a greater proportion of microbial variance attributable to plant genetic variation in -N conditions. Population genetic approaches identify signatures of purifying selection in the maize genome associated with the abundance of several groups of microbes in the maize rhizobiome. Genome-wide association study was conducted using the abundance of microbial groups as rhizobiome traits, and n=622 plant loci were identified that are linked to the abundance of n=104 microbial groups in the maize rhizosphere. In 62/104 cases, which is more than expected by chance, the abundance of these same microbial groups was correlated with variation in plant vigor indicators derived from high throughput phenotyping of the same field experiment. We provide comprehensive datasets about the three-way interaction of host genetics, microbe abundance, and plant performance under two N treatments to facilitate targeted experiments toward harnessing the full potential of root-associated microbial symbionts in maize production. Editor's evaluation It is widely assumed that plants actively manipulate their root associated microbiomes although the genetic factors that contribute to this process have yet to be discovered. The authors catalog the root-associated bacteria that associate with diverse maize lines, grown under low and high nitrogen treatments and through a genome-wide association study, identify candidate genetic loci that influence the plant associated microbiome. In addition to the interesting insights reported in the present manuscript, these data are likely to be used for and/or compared to, in many future studies. https://doi.org/10.7554/eLife.75790.sa0 Decision letter Reviews on Sciety eLife's review process Introduction Symbiotic relationships between plant hosts and root-associated microbes have been shaped through natural selection over millions of years of coevolution (Limborg and Heeb, 2018), and have been a driver of crop performance and yield in agricultural production since the beginning of plant domestication (Yadav et al., 2018). Microbial actors in the rhizosphere have been shown to promote plant growth (Saleem et al., 2019), improve nutrient use efficiency (Gomes et al., 2018; Zhu et al., 2016), and reduce abiotic stress response (Hussain et al., 2018). The promise of high throughput screens for plant growth promoting activity in isolated microbial strains or synthetic communities (Singer et al., 2021; Yee et al., 2021) is the potential discovery of microbial agents that can be used as seed or soil additives to improve crop performance under field conditions. Promising results have been observed in controlled environments (Van Gerrewey et al., 2020; Xi et al., 2020; Yu et al., 2021), but it remains a challenge to achieve similar outcomes in crops under agriculturally relevant field conditions (Eida et al., 2017; Kaur et al., 2020; Sessitsch et al., 2019), as many microbial inoculants struggle to compete with naturally occurring microbes in the rhizosphere and rarely survive for extended periods of time in the field (Piromyou et al., 2011). An improved understanding of how plants shape the composition of their rhizobiomes under diverse field conditions would make it more feasible to identify beneficial plant-microbe interactions that will be persistent and replicable in field environments. Moreover, studying the effects of plant genetics on microbial communities may identify opportunities to breed crop plants that recruit and maintain superior growth-conducive microbial communities from the natural environment. Few studies to date have addressed the relationship between plant genetics and rhizobiomes in field settings, mainly because large-scale rhizosphere sampling (as opposed to leaf microbiome sampling) and DNA sequence analysis of microbial communities in diverse plant hosts is time-consuming, expensive, and poses significant logistical and technical challenges. It has been shown that plant genetics can explain variation in both root architecture (Bray and Topp, 2018) and exudation (Mönchgesang et al., 2016). If these factors in turn shape microbial communities (Sasse et al., 2018), variation in the root-associated microbial groups (hereafter referred to as rhizobiome traits) may also result from genetic factors. Recent studies suggested that the variation in the composition of rhizobiomes is likely controlled by plant genetic factors (i.e., heritable) in maize (Peiffer et al., 2013), sorghum (Deng et al., 2021), and switchgrass (Sutherland et al., 2021). However, it remains unclear to what extent these heritable microbes are affected by the plant host and contribute to variation in the crop phenotype. Like any other trait under heritable genetic control, rhizobiome traits can be targeted in selective breeding experiments. To explore this idea, previous efforts have been directed towards identifying plant genetic loci that are associated with rhizobiome traits. Initial studies have shown that rhizosphere microbial communities differ between distinct genotypes of the same host species, which has been shown in a study on 27 maize genotypes (Peiffer et al., 2013; Walters et al., 2018) and more recently, in a diversity panel of 200 sorghum lines (Deng et al., 2021). Genome-wide association study (GWAS) has successfully revealed associations between plant genes and rhizobiome traits at high-level measures of rhizosphere community dissimilarity (i.e. using principal components) in an Arabidopsis diversity panel (Bergelson et al., 2019) or at order level (derived from operational taxonomic units [OTUs]) in a sorghum diversity panel (Deng et al., 2021). However, previous attempts at linking plant genes to the abundance of specific groups of microbes have had limited success due to small population size, limited host genetic diversity, or due to insufficient taxonomic resolution (Beilsmith et al., 2019; Liu et al., 2021). It was observed previously (Zhu et al., 2016) that soil microbial communities drastically change in response to N fertilization. In bulk soil, this is likely due to a direct effect of N application or lack thereof. In rhizospheres, however, only a subset of the observed changes can be attributed to direct effects of nitrogen (N) fertilization, while particular microbial groups may be subject to indirect effects induced by the plant host in response to N availability or deficiency (Meier et al., 2021). A possible explanation for this could be that during most of the interval between maize domestication and the present, beneficial plant-microbe interactions have evolved in low-input agricultural systems characterized by relative scarcity of nutrients, predominantly nitrogen (Brisson et al., 2019). This is in stark contrast to the modern agricultural environment that has been the norm since the 1960s, in which plants are supplied with large quantities of inorganic N fertilizer (Cao et al., 2018). As a consequence, previous selection pressure to retain traits favorable under low N conditions, including plant growth-promoting microbes, has been largely reduced in modern maize breeding programs (Haegele et al., 2013; Zhu et al., 2016). Thus, if a microbial group is indeed under host genetic control and has an effect on plant fitness (i.e. promotes plant development or increases crop yield) under either N condition, we would expect the rhizobiome trait to be under host selection. In the present study, we evaluate the role that selection on plant genetic factors has played in shaping the maize rhizobiome under different N conditions. We employ the Buckler-Goodman maize diversity panel, a set of maize lines selected for maximum representation of genetic diversity and growth in temperate latitudes (Flint-Garcia et al., 2005). This population has previously been used to determine the heritability of leaf microbiome traits and to perform genome-wide association studies (GWAS) on a number of different phenotypes (Wallace et al., 2018). We collected replicated data on the rhizobiome of 230 lines drawn from this panel when grown under either high nitrogen (+N) and low nitrogen (-N) conditions in the field. For 150 microbial groups present in the rhizosphere (referred to as ‘rhizobiome traits’), which were abundant and consistently reproducible, we quantify the degree to which variation is subject to plant genetic control, and test for evidence of selection under either or both N conditions. Using a set of 20 million high-density single-nucleotide polymorphisms (SNPs), we perform GWAS for each rhizobiome trait identifying genomic loci that are associated with one or more rhizobiome traits. Through comparison with gene expression data generated for the same population, we determine whether genes near microbe-associated plant loci are preferentially expressed in root tissue. Lastly, we evaluate whether the abundance of each microbial group in the rhizosphere is correlated with plant performance traits measured in the field, and whether microbe abundance and plant performance depend on the allele variant at selected microbe-associated plant loci. The results presented in this study lay the groundwork for future endeavors to investigate the molecular mechanisms of specific plant-microbe interactions under agronomically relevant conditions. Results Characterization of the rhizobiome for diverse maize genotypes under two different N conditions 3,313 rhizosphere samples from 230 replicated genotypes of the maize diversity panel (Flint-Garcia et al., 2005) were collected from field experiments conducted under both +N and -N conditions (Materials and methods). At the time of sampling, visible phenotypic differences were observable between +N and -N plots as measured through aerial imaging (details are reported in Rodene et al., 2022 using the same experimental field). Paired-end 16 S sequencing produced 216,681,749 raw sequence reads representing 496,738 unique amplicon sequence variants (ASVs) (Materials and methods). Raw reads were subjected to a series of quality checks and abundance filters following a workflow for 16 S sequencing data analysis by Callahan et al., 2016a, which resulted in a curated dataset of 3626 ASVs for 3306 samples, and 105,745,986 total ASV counts (Supplementary file 1). This dataset includes ASVs that are highly abundant across the maize diversity panel and reproducible in both years of sampling. Constrained Principal Coordinates analysis calculated from the abundances of 3626 ASVs shows divergence of samples collected under either -N or +N treatment (Figure 1A), which indicates that the microbiomes differ between these two experimental conditions (PERMANOVA p-value for N treatment <0.001). Figure 1 with 3 supplements see all Download asset Open asset Diversity, phylogenetics, and heritability of rhizobiome traits. (A) Constrained ordination analysis showing the largest two principal coordinates calculated from the abundances of 3626 ASVs. Each diamond represents one sample collected from plants under +N (blue) and -N (red) treatment, respectively. Note the separation of N treatments along PCo1. (B) Phylogenetic tree of 150 taxonomic groups of rhizosphere microbiota (‘rhizobiome traits’) generated by clustering 3626 ASVs. Families are prefixed with ‘f_’, genus and species names are given where known. Numbers at tree tips indicate distinct ASVs in each group. Label colors indicate heritability of each rhizobiome trait as in panel C. (C) Heritability (h2) calculated for all 150 rhizobiome traits under +N and -N treatments. Green line indicates linear regression with 95% confidence interval, r2=0.104. Diagonal dashed line denotes identity. Gray lines mark density of data points. Colors indicate whether traits are significantly heritable under either or both N treatments, as determined through a permutation analysis using 1000 permutations. An initial analysis looking at high-level rhizobiome traits (Principal Components and alpha diversity metrics derived from the ASV table) shows the same pattern of divergent microbial communities between N treatments, and in particular under the -N treatment there is evidence for the association of plant genomic loci and microbiome composition (Figure 1—figure supplement 1). To study changes in rhizobiome composition more accurately, the final 3626 ASVs were clustered into n=150 distinct microbial groups (‘rhizobiome traits’), spanning 19 major classes of rhizosphere microbiota (Figure 1B, Supplementary files 2 and 3) using a method previously described (Meier et al., 2021, Supplementary Methods). Of these rhizobiome traits, 79/150 (52.7%) groups were significantly more abundant in samples collected from the +N condition (t-test, p<0.05), 53/150 (35.3%) significantly more abundant in samples collected from the -N condition, and 18/150 (12.0%) showed no significant difference in abundance between the two treatments. In several cases, more closely related microbial groups exhibit shared patterns of differential abundance between N treatments (Figure 1—figure supplement 2A). Rhizobiome traits are more heritable under -N conditions The abundance of each of the 150 rhizobiome traits was assessed separately for +N and -N conditions, and the heritability (proportion of total variance explicable by genetic factors) was estimated using an approach following a previous study (Deng et al., 2021) (Materials and methods). Rhizobiome traits were comparatively more heritable under -N than +N conditions (paired Student’s t-test, p=0.021, Figure 1C). We found 34/150 (22.7%) microbial groups to be significantly heritable (permutation test, p<0.05, Materials and methods) under both N conditions, 18/150 (12%) only under +N conditions, and 27/150 (18%) only under -N conditions. Twelve rhizobiome traits exhibited estimated h2 >0.6 in both +N and -N conditions (Figure 1—figure supplement 3). These include four groups of ASVs assigned to the order Burkholderiales (the genus Pseudoduganella, an unknown genus in the Comamonadaceae family, the family A21b, and Burkholderia oklahomensis) and two groups in the Sphingomonadales order (Sphingobium herbicidovorans 1 and an unknown genus in the Sphingomonadaceae family). Notably, closely related microbial groups did not exhibit obvious patterns of shared high or low estimated heritabilities (Figure 1B). As heritabilities and responses to treatments can vary considerably within families, genera, and lower taxonomic ranks, this underscores the importance of adequate taxonomic resolution when analyzing rhizosphere microbial communities. We further observed that more abundant microbes in the rhizosphere also tend to be more heritable. The correlation of relative abundance vs. heritability was r=0.29 (Pearson’s correlation test, p=3.4 × 10–4) for +N and r=0.39 (Pearson’s correlation test, p=1.1 × 10–6) for -N (Figure 1—figure supplement 2B). Rhizobiome traits are related with plant fitness and predominantly under purifying selection Under the hypothesis that the rhizobiome traits have effects on plant fitness, we sought to estimate the selection differentials under different N treatments (Rausher, 1992). To reduce biases due to environmental covariances (Lande and Arnold, 1983), the standardized BLUP values of the microbial traits were fitted into the fitness function (See Materials and methods). For the selection differential estimation, the canopy coverage (CC) obtained from UAV imaging was used as a proxy for plant fitness. As a result, we identified 58 unique rhizobiome traits exhibiting significant linear selection differentials (bootstrapping p-value <0.05) under +N (28 traits) and -N (46 traits) treatments (Figure 2—figure supplement 1). Additionally, four rhizobiome traits showed significant quadratic selection differentials (+N: Luteolibacter pohnpeiensis [–2.627913e-05, p-value = 0.044], -N: Blastococcus [8.516159e-06, p-value = 0.03], Pseusomonas umsongensis [–2.003792e-05, p-value = 0.04], Chthoniobacter flavus [–5.807404e-05, p-value = 0.028]). Selection acting on rhizobiome traits can happen either by purging deleterious alleles (purifying selection) or by elevating the frequencies of advantageous alleles (positive selection). To evaluate the mode of selection at the genomic level, a Bayesian-based method (Genome-wide Complex Trait Bayesian analysis, or GCTB) was used to test for each rhizobiome trait (Materials and methods). A set of n=834,975 independent SNPs was used to estimate their effects on 150 rhizobiome traits as well as 17 conventional plant traits collected from the same population in the same field experiments (Materials and methods, Supplementary file 4). Using the relationship between effects of non-zero SNPs and their minor allele frequencies (MAFs) as a proxy for the signature of selection (Zeng et al., 2018), the S parameter was jointly estimated from the GCTB analysis for rhizobiome traits and plant traits. According to Zeng (Zeng et al., 2018), if S=0 (i.e. the posterior distribution of S is insignificantly different from zero), the SNP effect is independent of MAF, suggesting neutral selection. If there is selection acting on the trait, the SNP effect can be positively (S>0) or negatively (S<0) related to MAF, indicating positive and purifying selection, respectively. We report 10 rhizobiome traits that showed both significant linear selection differentials and significant S parameters (Figure 2A). Under these stringent criteria, nine rhizobiome traits show evidence of purifying selection under +N or under -N. One microbial group (Bacillus fumarioli) showed positive S values indicating that this trait might have been a target of positive selection. Relative to rhizobiome traits, plant leaf traits and nutrient traits were both more likely to exhibit evidence of selection within this maize population. Three out of 15 plant leaf traits, that is leaf area (LA), leaf fresh weight (FW), and leaf dry weight (DW) (Materials and methods), exhibited S>0 values under the +N condition, consistent with positive selection, while only one of the three exhibited a slightly negative S value in the -N condition and in that case exhibited a pattern consistent with weak purifying selection (Figure 2B). Note that the three leaf-related traits are not independent. The pairwise correlation coefficients are 0.92, 0.91, and 0.94, for LA and FW, LA and DW, FW and DW, respectively. Of the 11 micronutrient traits evaluated, 9/11 and 4/11 showed significantly negative S values in trait data collected under +N and -N conditions, respectively. From the same GCTB analysis, estimates of the number of SNPs with non-zero effects were substantially lower for rhizobiome traits than for conventional plant traits, whereas the differences were insignificant between the two N treatments for both rhizobiome and plant traits (Figure 2C). Using these non-zero effect SNPs, we plotted their minor allele frequency vs. the minor allele effect. As expected, in the case of positive selection (Bacillus fumarioli), we observed a skew towards higher MAF and in the case of purifying selection (f_Comamonadaceae Unknown Genus), a skew towards lower MAF (Figure 2D). Figure 2 with 1 supplement see all Download asset Open asset Population parameters estimated from genome-wide SNPs for plant and rhizobiome traits. Selection coefficients (S value) of rhizobiome (A) and plant (B) traits calculated for both N treatments using genome-wide independent SNPs. A negative S value indicates negative (purifying) selection, and a positive S value indicates positive (directional) selection. Traits are shown that show significant selection under one or both N treatments. (C) Number of SNPs showing non-zero effects for both plant and rhizobiome traits. (D) Examples of positive (Bacillus fumarioli) and purifying selection (f_Comamonadaceae Unknown Genus) showing minor allele effect vs. allele 1 frequency with data skew to the right and to the respectively. microbe-associated plant loci are preferentially expressed in root The that many rhizobiome traits are both under significant host genetic control and of selection it may be possible to large effect loci rhizobiome traits. To investigate we GWAS using each of the 150 rhizobiome traits. This analysis was separately for the -N and +N conditions, as N deficiency changes in plant including changes in root gene expression et al., and root exudation (Zhu et al., 2016), and because N to the field soil and rhizosphere microbiomes (Meier et al., 2021). We on along the genome where we the density of significant associations between SNPs and any rhizobiome traits under either N treatment, because (i.e. root or changes may several rhizobiome traits. For this we the maize genome into 10 genomic and the number of significant GWAS in each This analysis revealed genomic at one significant GWAS and we to these as microbe-associated plant loci (Materials and methods). We report these genes in Supplementary file of 150 microbial were associated with at one of the loci. To reduce we to a subset of that had at two significant GWAS these were observed under +N treatment and under -N treatment (Figure Supplementary file Of the 150 rhizobiome traits were associated with at one of the with rhizobiome traits associated with under the +N treatment and 17 rhizobiome traits with under the -N Three rhizobiome traits Unknown and showed significant associations under both N treatments, with different plant loci. loci were found that had associations with rhizobiome traits under both N treatments, which is expected as GWAS were separately for different N treatments and the microbial groups were on differential abundance in response to N treatments. Figure 3 Download asset Open asset associated plant loci genes expressed in (A) GWAS of 150 rhizobiome traits microbe-associated plant loci across the maize line indicates the level for GWAS on of at each indicate the number of rhizobiome traits associated with each Each is with the associated rhizobiome that showed significant GWAS (B) gene expression of genes and other genes in measured in genotypes of the maize diversity panel et al., 2018). (C) gene expression of genes and other genes in two measured in the present study in four maize genotypes under +N and -N treatments. We that many plant genes may control over the rhizosphere microbiome changes in root and composition et al., and may be preferentially expressed in root tissue. of gene were within of the identified where between 1 and We the expression of these genes relative to the patterns exhibited by all genes the in using expression data from the same maize population et al., 2018). data was in this dataset for out of genes across maize genotypes from samples collected at and during These when compared to other genes in the show on significantly higher expression in the the and the leaf (Figure To the gene expression data by we selected four diverse and well characterized maize genotypes and were grown in a controlled environment under N and N conditions and gene expression was in and of to et al., In with the dataset by et significantly higher expression of genes was observed in root but not leaf compared to other genes in this dataset (Figure response to N deficiency was expected for and no significant differences were observed in the pattern of gene expression between the two N treatments. these data are consistent with the hypothesis that rhizobiomes are at in controlled by the host plant in a process by plant gene and selected are associated with plant phenotypes We the correlation of microbe abundance with 17 plant traits, including leaf leaf micronutrient traits, and traits from aerial (Materials and methods) to identify potential plant phenotypic of variation in the abundance of specific rhizosphere rhizobiome traits were significantly correlated with measures of plant as leaf leaf dry weight and fresh and with several measures of leaf as and (Figure supplement 1). The trait that was most linked to microbe abundance was leaf canopy coverage In microbial groups more than expected by (permutation test, were significantly correlation test, associated with in Figure in for positive correlation and in for negative microbial groups under +N and under -N were positively correlated with groups under +N and under -N were negatively correlated with 15 microbial groups were associated with under +N treatment, under -N treatment, and showed a significant association under both N treatments (Figure Under both N treatments, we an association between heritability and the correlation with which was significant correlation × 10–6) for +N and more significant × under the -N condition (Figure Figure with 2 supplements see all Download asset Open asset microbial groups tend to be correlated with plant canopy (A) of and correlation values for the relationship between canopy coverage (CC) and each of 150 microbial groups under either +N or -N conditions. line indicates level (B) between the estimated heritability of rhizobiome traits and correlation of the same rhizobiome traits with variation in line indicates level We the relationship of the conducted in this study under either N treatment for the microbial groups that are correlated with are heritable and are under selection under either or both N treatments (Figure supplement 2 show GWAS in of the identified which of microbe-associated microbial Unknown and Burkholderia are of particular as in all showing evidence of heritability and selection, a GWAS in associated plant genomic and positive correlation with canopy The data for all 150 microbial groups are in Supplementary file data show a that the microbial groups associated with plant performance also tend to be associated with host genetics, and the datasets can be used to more targeted experiments to associations of rhizosphere microbial groups with plant genetics and
Source: eLife
Institution: University of Nebraska–Lincoln
Language: English
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: gold
Estimated APC paid: $2,500
OpenAlex ID: W4297794490