Influences of plant organ, genotype, and cultivation site on the endophytic bacteriome of maize (Zea mays L.) in the semi-arid region of Pernambuco, Brazil

Abstract

Endophytic bacteria play a crucial role in plant development and adaptation, and the knowledge of how endophytic bacteria assemblage is influenced by cultivation site and plant genotype is an important step to achieve microbiome manipulation. This work aimed to study the roots and stems of endophytic bacteriome of four maize genotypes cultivated in two regions of the semi-arid region of Pernambuco - Brazil. Our hypothesis is that the endophytic community assemblage will be influenced by plant genotypes and cultivation region. Metabarcoding sequencing data revealed significant differences in alfa diversity in function of both factors, genotypes, and maize organs. Beta diversity analysis showed that the bacterial communities differ mainly in function of the plant organ. The most abundant genera found in the samples were Leifsonia, Bacillus, Klebsiella, Streptomyces, and Bradyrhizobium. To understand ecological interactions within each compartment, we constructed co-occurrence network for each organ. This analysis revealed important differences in network structure and complexity and suggested that Leifsonia (the main genera found) had distinct ecological roles depending on the plant organ. Our data showed that root endophytic maize bacteria would be influenced by cultivation site, but not by genotype. We believe that, collectively, our data not only characterize the bacteriome associated with this plant and how different factors shape it, but also increase the knowledge to select potential bacteria for bioinoculant production.

Supplementary Information

The online version contains supplementary material available at 10.1007/s42770-023-01221-w.

Introduction

Maize (Zea mays L.) is a widely cultivated crop that is used for human and animal nutrition and as a raw material for biofuels. As a tropical grass with C4 photosynthetic metabolism, it has a high efficiency and growth rate. However, maize production is limited by factors such as nitrogen (N) and phosphorus (P) availability [1–3], as well as drought and salinity, especially in arid and semi-arid regions [4, 5].

Beneficial bacteria have been shown to promote productivity in maize [6–8], as they can provide significant benefits to plant growth by mechanisms such as biological nitrogen fixation, inorganic phosphate solubilization, regulation of plant hormones, and protection against phytopathogens [9–11]. To be successful in promoting plant growth, microbial inoculants must have beneficial characteristics and be able to colonize the target environment (plant interior or rhizosphere) and interact with the plant and other microorganisms [12].

Next-generation sequencing technologies have enabled better descriptions of the plant microbiome, which is the collection of microorganisms and their microbial molecules, associated with plants [6–8, 13]. This has helped to identify the most important factors shaping the expected community within a plant, including variations among cultivation sites, plant organs, genotypes, and growth stages [6].

In this study, we aimed to characterize the endophytic maize bacteriome cultivated in different sites in the Brazilian semi-arid region, understanding how different genotypes and regions influence this set of bacteria. To achieve this, we sampled roots and culms of four genotypes cultivated in the areas of Serra Talhada and Ararapina, and submitted these to 16S rRNA gene sequencing, after superficial disinfestation, to access endophytic communities. Our hypotheses were (i) the sampled plant organ would be the main factor determining the bacterial composition, followed by cultivation site and genotype, with minor effects; (ii) different indicators or ecologically important bacteria would be associated with each factor, helping understand selection and community assemble in this plant species.

Materials and methods

Site description and experimental design

The experiment was carried out in two experimental stations of the Pernambuco Agronomic Institute (IPA), located in the cities of Serra Talhada and Ararapina, in the Brazilian semi-arid region. Soil physical-chemical analysis of the sites was performed according to Embrapa [14] and Unkovich et al. [15], and the data is compiled in Table 1. The experiment adopted identical designs in each site, with three replications in randomized blocks, and was carried out during the rainy season (November of 2018 to March of 2019). The treatments consisted of four maize genotypes (BRS4107 - G1, BRS4105 - G3, BRS5036 - G6, and BRS5037 - G7), selected from the maize breeding program of the IPA and Brazilian Agricultural Research Corporation (EMBRAPA). These four genotypes were chosen because they have proven to be the most promising in terms of productivity, resistance to pests and diseases, and adaptability to the soil and climate conditions of the Brazilian semi-arid region. The experimental plot consisted of 4 rows of 4 m in length, spaced 0.80 m apart, in which the maize was sown in holes spaced 0.20 m apart. The two central rows were considered the useful area of the plot, disregarding 1 m at each end, to avoid the border effect.

Table 1.

Characteristics of the two experimental sites, fertilization procedure and soil’s physical-chemical characteristics used in the experiments in two municipalities in Pernambuco state, Brazilian semi-arid

Site characteristics	Serra Talhada			Araripina
Coordinates	07°59′00″S, 38°19′16″W			7°27′50″S; 40°24′38″W
Altitude (m)	500			828
Annual rainfall (mm)	709			713
Average temperature (°C)	24.5			22.9
Soil type	Ultisols			Oxisols
Soil fertilization	Foundation: 70 kg ha−1 of urea, 10 kg ha−1 of Ca(H2PO4)2 + CaSO4. 2H2O) and 20 kg ha−1 of KCl. 40 days after sowing: more 40 kg ha−1 of urea			Foundation: 70 kg ha−1 of urea, 30 kg ha−1 of Ca(H2PO4)2 + CaSO4. 2H2O) and 30 kg ha−1 of KCl. 40 days after sowing: more 40 kg ha−1 of urea
Soil characteristics	0–20 cm	20–40 cm	40–60 cm	0–20 cm	20–40 cm	40–60 cm
pH (H2O, 1:2.5 v/v)	6.5	6.4	6.9	4.4	4.4	4.6
N (%)	0.06	0.04	0.02	0.12	0.09	0.06
C (%)	0.71	0.50	0.27	1.67	1.40	0.99
C:N	11.8	12.5	13.5	13.9	15.5	16.5
δ 13C (‰)	−17.78	−18.1	−17.2	−25.01	−24.96	−24.42
δ 15N (‰)	10.57	9.25	8.9	16.73	17.16	16.93
P (mg/dm3)	98.7	90.0	86.0	2.0	1.0	1.0
Ca2+ (cmolc/dm3)	2.3	2.9	2.3	0.4	0.3	0.2
Mg2+ (cmolc/dm3)	1.7	1.9	1.2	0.5	0.6	0.5
Na+ (cmolc/dm3)	0.1	0.1	0.3	0.0	0.0	0.0
K+ (cmolc/dm3)	0.9	0.5	0.4	0.0	0.0	0.0
Al3+ (cmolc/dm3)	0.0	0.0	0.0	0.0	0.0	0.0
H+ (cmolc/dm3)	1.2	0.5	0.5	2.8	2.2	1.5
S (cmolc/dm3)	5	5.4	4.1	1	0.9	0.7
CEC (cmolc/dm3)	6.1	5.9	4.6	4.3	3.8	2.7
Coarse sand (%)	31.7	38.3	38.7	53.3	53.3	48.7
Fine Sand (%)	43.0	41.0	43.0	26.0	23.7	26.3
Silt (%)	15.3	10.7	8.3	4.0	2.3	3.0
Clay (%)	10.0	10.0	10.0	16.7	20.7	22.0

CEC cation exchange capacity

Plant sampling and surface disinfestation

Plant samples were collected in the intermediate phase of the vegetative stage (4 months), by selecting plants of homogeneous sizes and at least ten expanded leaves, within each experimental plot. At this stage, named V10, maize plants begin rapid and continuous growth, with accumulation of nutrients and dry weight, which will continue until the reproductive stages. From each plot, five plants were collected to form a composite sample. To do this, the plants were carefully removed from the soil, using a hoe, to preserve as much of the roots as possible. The plants were washed with running water and then with sterilized distilled water, removing all dirt and soil adhering to the roots, and cut in the root insertion region using sterilized knives. Roots and stems were processed separately. From each plant, non-lignified roots were collected, from which three sample fragments of 2 cm in length were separated. A composite sample of root fragments was formed from three fragments of each maize plant. The stems were sectioned at the insertion of the fourth leaf, fragmented into small pieces of approximately 0.125 cm³. To form the composite stem samples, two fragments were collected from each sampled plant. The samples were stored in sterile flasks containing silica gel, transported to the laboratory, and kept at room temperature until processing began, 2 weeks later.

In the laboratory, the plant material was rehydrated for 1 h in sterile mineral water. Subsequently, surface sterilization was performed by immersing the material in 70% ethanol for 1 min, followed by immersion in a 2.5% sodium hypochlorite solution for 1 min. After disinfestation, the samples were washed with sterile distilled water five times [16].

DNA extraction and 16S rRNA gene sequencing:

Samples of plant organs were frozen in liquid nitrogen and macerated using autoclaved pestles and mortars. DNA was extracted from each sample using the Qiagen DNeasy PowerSoil kit (ID: 47017; QIAGEN, Germany) following the manufacturer’s instructions. The resulting DNA was checked for integrity by electrophoresis on a 0.8% agarose gel. Sequencing library was prepared according to the manufacturer’s instructions, and sequenced with an Illumina MiSeq (2x300 pb; Illumina, USA), at the Laboratory of Bioinformatics and Evolutionary Biology (LABBE), Department of Genetics, Health Sciences Center, Federal University of Pernambuco, using the Nextera XT Index kit (Illumina, USA; FC-131-1001), Miseq Reagent Kit V3 (Illumina, USA; MS-102-3003) and Phix Control v3 (Illumina, USA; FC-110-3001). Briefly, the 341F-805R primer pair (CCTACGGGNGGCWGCAG; GACTACHVGGGTATCTAATCC) [17], which targeted the V3 and V4 hypervariable regions of the 16S rRNA gene was used, along with standard appended adapters. PCR reactions were cleaned by utilizing AMPure XP beads (Illumina, USA), and after they were quantified, normalized, and pooled, before loading in the MiSeq platform.

Bioinformatic analysis

Raw DNA sequences were processed using Mothur v.1.44.3 software [18] to evaluate and analyze the total bacterial diversity of the samples. The forward and reverse sequences were combined using the make.contigs command. Sequences outside the range of 440–465 nucleotides, with any ambiguity and/or with a number greater than 8 homopolymers, were removed using the screen.seqs command. Identical sequences were grouped using the unique.seqs command. The sequences were aligned with the SILVA 138.1 [19] database after performing a virtual PCR with the 341F-805R primers. The resulting alignment was further processed using the screen.seqs and filter.seqs commands to remove misaligned sequences and uninformative columns. The sequences were then subjected to the pre-cluster command, with the parameter “diffs=2.” Chimeric sequences were identified and removed using the chimera.vsearch method, with the parameter “dereplicate=T.” Classification was performed using the RDP database v.18 (July 2020) as a reference, using an 80% confidence limit. Sequences classified as chloroplasts, mitochondria, Archaea, Eukarya, or unknown were removed. The remaining high-quality sequences were grouped into operational taxonomic units (OTUs) using dist.seqs, followed by the cluster command with a sequence dissimilarity cutoff of 3%. Singletons were removed using the split.abund command.

Sequence normalization was performed using the sub.sample method, and all samples were rarified to 4126 sequences. Each dataset was organized into OTU abundance matrices using the make.shared function. From this point, alpha diversity, rarefaction curves, and relative abundance indices of taxonomic classifications were generated.

Despite the experiment was designed to evaluate 48 samples (24 from each experimental site), only 36 could be evaluated due to high co-amplification of chloroplasts DNA in some samples. Therefore, we tested two models. To evaluate the influence of genotypes and organs, we used a subset having samples from roots and culm cultivated in Serra Talhada only. To evaluate the influence of cultivation site, we used root samples of two different genotypes (G6 and G7) from Araripina and Serra Talhada (Table S1).

Statistical procedures

Statistical analyses regarding the relative abundance of taxa and alpha diversity indices were conducted using the type II two-way ANOVA in the R software (Team, 2021) with the CAR (Companion to Applied Regression) [20] package. For model 1, the effects of genotype and organ were tested, and for model 2, the effects of genotype and cultivation area were assessed. Normality was evaluated using the Shapiro-Wilk test, and homoscedasticity was assessed using the Levene test. Tukey’s post hoc test was employed for data with homoscedastic results, while the Games-Howell post hoc test was used for heteroscedastic data. To analyze the beta diversity of the datasets, a non-metric multidimensional scaling (nMDS) was performed using the Bray-Curtis dissimilarity index and the metadata recorded during sample collection. Multidimensional differences under different factors were tested using a two-way PERMANOVA. Beta diversity analyses were conducted using the Past v4.06 software [21].

A linear discriminant effect size analysis (LEfse) [22] was conducted to identify community indicator organisms and explain differences in taxa identified for each pair of factors. The analysis utilized the OTU matrix exported from Mothur as input to the LEfSE software, which was hosted on the Huttenhower lab Galaxy server (available at: http://huttenhower.sph.harvard.edu/galaxy/).

Furthermore, co-occurrence networks were generated using the Pearson and Spearman correlation methods, along with the Bray-Curtis dissimilarity, applied to the OTU matrix. The final networks were assembled using the Cytoscape software [23] along with the CoNet plugin [24]. The ReBoot correction method for compositional data was applied. Raw sequences were deposited in the NCBI Sequence Read Archive (SRA) and are available under Bioproject PRJNA846726.

Results

We obtained 11,957,828 sequences from the Miseq’s sequencing, with a mean of 441 bp. After quality filtering and subsampling all samples to the size of the smallest sample (4126 sequences), we obtained a total of 132,032 high-quality sequences and 5011 OTUs based on 97% similarity index, with an average length of 416 bp. Rarefaction curves achieved a plateau, suggesting good coverage of the diversity in all samples (Figure S1).

Alpha diversity analysis revealed a significant effect of both factors, genotypes, and maize organs, on OTU richness in model 1 (samples from Serra Talhada only). However, differences in the Shannon index were observed only between organs (Fig. 1A and B). An interaction effect was observed in the richness analysis, which could be attributed to a higher value in the G7 genotype. Overall, roots exhibited higher alpha diversity compared to culms. In model 2 (root samples only), the genotype factor also showed statistically significant differences in OTU richness, with the G7 genotype displaying higher values.

Fig. 1.

Bacterial alfa diversity (richness of OTUs and Shannon Index) in different maize genotypes (BRS4107 - G1, BRS4105 - G3, BRS5036 - G6 and BRS5037 - G7), organs (root and culm) and cultivation sites (Serra Talhada and Araripina). A - OTU richness in model 1, B - Shannon Index in model 1, C - OTU richness in model 2, and D - Shannon Index in model 2

Beta diversity analysis based on model 1 (Fig. 2A) demonstrated that the bacterial communities primarily differed based on the plant organ (p<0.01, two-way PERMANOVA), with no significant differences among genotypes. Culm samples exhibited a narrow range of dispersion, while root samples displayed greater dispersion along the ordination (higher beta diversity).

Fig. 2.

Non-metric multidimensional scaling (nMDS) with the Bray-Curtis dissimilarity index of the OTU distribution from different maize genotypes (BRS4107 - G1, BRS4105 - G3, BRS5036 - G6 and BRS5037 - G7), organs (root - . R, and culm - .C) and cultivation sites (Serra Talhada - . St and Araripina - . A). A - model 1 (genotype × organ, in Serra Talhada only). B - model 2 (genotype × cultivation site, only roots). The angles and the length of radiating lines indicate the direction and strength of the relationship between the chemical and biological variables with the ordination scores. %Ndda: percentage of nitrogen derived from the air

Analysis of the second model (Fig. 2B) revealed a significant influence of the plantation site on the endophytic root bacteriome. Araripina samples clustered closely together, while Serra Talhada exhibited higher dispersion. The differences among the bacteriomes correlated with %C, %N, 13C, 15N, C/N, and %Ndda (percentage of nitrogen derived from the air).

In our samples, the most abundant phyla were Actinobacteria, Proteobacteria, Firmicutes, and Bacteroidetes, respectively. Only Firmicutes was statistically different. In model 1, this phylum was different among organs, with higher abundance in culm samples. In model 2, it was different between collection sites, with higher relative abundance in Araripina (Table S2 and S3).

The main genera identified in the samples were Leifsonia, Bacillus, Klebsiella, Streptomyces, Bradyrhizobium, and others, as summarized in Fig. 3. Among the 15 most abundant genera, 10 differed between culms and roots (model 1). Only the Ralstonia genus displayed a difference based on genotype, particularly due to its depletion in G7 samples. In the second model, comparing the same genera in relation to cultivation site and genotype, none of them was affected by the cultivation site. However, the genera Bacillus, Streptomyces, Geobacillus, and Enterobacter showed statistically significant differences between genotypes (root samples only).

Fig. 3.

Heatmap of z-scores based on the relative abundance of main bacterial genera in different maize genotypes ((BRS4107 - G1, BRS4105 - G3, BRS5036 - G6, and BRS5037 - G7), organs (root and culm), and cultivation sites (Serra Talhada and Araripina). Model 1: A (genotype × organ). Model 2: B (genotype × cultivation site). Data submitted to a two-way ANOVA. * means significant differences for organs comparison, + means significant differences for site comparison, and i means interaction between factors

Out of the 4126 OTUs generated from all samples, only 65 OTUs were highly prevalent (>80% of the samples) in the culm samples, and 54 in the root samples (model 1). There was a strong agreement between the most abundant genera depicted in Fig. 3 and the genera represented in the core (those present in >80% of the samples), indicating that they are not only abundant but also colonize the majority of the samples [25].

To gain insights into ecological interactions within each compartment, co-occurrence networks were constructed for each organ using model 1 (Fig. 4). The bacteriome of the culm samples formed a simpler network compared to the root samples. Additionally, only positive relationships were found in the root network, while several negative relationships were observed in the culm.

Fig. 4.

Co-occurrence networks with OTUs from culm and root samples (model 1) from different maize genotypes. Each node size represents the relative abundance of the OTU, while its color the community cluster it belongs to inside the network. Border width details the degree of connections of a node, and green lines represent positive associations and red negative ones. The thickness of each line connecting two nodes depict the intensity of this interaction (d – domain, p – phylum, c – class, o – order, f – family, g – genus)

Discussion

Given the economic and social significance of maize, extensive research has been conducted on its microbiome. Many studies have delved into understanding the patterns of microbial communities in the maize rhizosphere [7, 26–28], maize endophytes [8, 28, 29], and the response of the maize microbiome to inoculation [29–31].

Our study contributes interesting and novel data to this existing body of literature. While previous studies have demonstrated that different maize organs harbor distinct bacteriomes [8], and that cultivation site influences these bacteriomes [8], to the best of our knowledge, this is the first report showing no influence of genotype on the maize endophytic bacteriome, as shown by beta diversity. We believe that, even though there were limitations because of chloroplast amplification, resulting in the exclusion of some samples, important data derived from different factors in this semi-arid environment contribute to a more comprehensive understanding of endophytic microbial selection in crops. Robust studies have been conducted on the influence of maize genotypes on rhizosphere microbial assembly or response to inoculation [7, 26]. Favela et al. [7] demonstrated that microbial communities undergo changes with genotype selection of maize, with different profiles observed when comparing genotypes from different decades of breeding and selection. The study revealed that genotype selection promoted an increase in microbial beta-diversity and alterations in the relative abundance of genes related to the nitrogen cycle. Walters et al. [28] conducted a large-scale longitudinal field study, comparing 27 modern maize genotypes, and found that plant genotype can explain a small but significant proportion of bacterial community structure. Different genotypes may exhibit differences in root architecture, development time, and exudation patterns, among various complex quantitative traits. However, these attributes may have a greater impact on the surrounding soil of the roots (rhizosphere) rather than the plant interior.

Both alpha and beta diversity analyses demonstrated that the plant organ plays a major role in selecting the plant bacteriome. Roots exhibited higher alpha diversity indices and higher beta diversity compared to culm samples. The complexity of the root network was also more pronounced. These findings are consistent with those reported by Zhang et al. [8], who studied different maize plantations in China. The authors compared bacterial communities in bulk soil, rhizosphere soil, root endosphere, stem endosphere, and other compartments using 16S rRNA amplicon sequencing. They concluded that plant compartments were the primary drivers of the maize microbiome, compared to environmental and edaphic factors. They also reported a decreasing diversity gradient among bulk soil, rhizosphere soil, root endosphere, and stem endosphere. According to these authors, 60.61% of bacteria in the root endosphere were derived from rhizosphere soils, while 24.21% of bacteria in the stem xylem were derived from the root endosphere. Roots have the potential to recruit bacteria from the entire soil, while the culm can only recruit those adapted to the interior of the roots, resulting in a reduced set of bacteria that can thrive in this compartment.

Our study revealed a predominance of the Leifsonia genus (Figs. 3 and 4) in the microbial community of both roots and culm. Species of this genus are usually present in the xylem vessels of plants and can even be found in the mesophyll and sheath cells surrounding the vascular system [30]. There are no reports in the literature of harmful effects of the Leifsonia genus on maize plants. However, the fastidious bacterium Leifsonia xyli subsp. xyli is the pathogen responsible for ratoon stunting disease in sugarcane, a significant stem disease that causes a significant reduction in plant height, stem diameter, and stem weight, directly impacting the productivity of sugarcane plantations worldwide [30–34]. Interestingly, in our network analysis, Leifsonia showed positive relationships in roots but mainly negative relationships in the culm, where it causes diseases in other plants. Therefore, it is plausible to believe that the differences in the micro-environment of plant compartments lead to different interactions involving this taxon. The same process was observed for Bacillus, Streptomyces, and Bradyrhizobium.

The genera Bacillus, Streptomyces, and Bradyrhizobium also played significant roles in the evaluated samples. These genera are commonly described as plants, including maize. Several studies have demonstrated the beneficial effects of species within these genera in promoting the growth of maize plants. These effects include contributions to biological nitrogen fixation [35–37], production of indoleacetic acid and phosphate solubilization [38–41], suppression of phytopathogens [42–46], and tolerance to drought and saline stress [5, 47, 48]. In culm, all of them present only positive interactions in the network analysis; however, in root they showed both positive and negative interactions, including between each other. While it is hard to stablish a casual effect of negative interactions from network analysis, especially with compositional data [49], it suggests that main members of community can interact differently depending on the environment. One main factor contributing to this difference in interaction type is the difference in competition in each habitat. Root microbiome is richer than culm, which will result in higher functional redundancy and therefore increased competition (negative interaction).

Our data demonstrated that the composition of root endophytic bacteria in maize is primarily influenced by the plant organ, which may have an impact on the interactions among the major taxa. Additionally, to a lesser extent, the cultivation site was found to influence the assemblage of endophytic bacteria. Interestingly, genotype differences among the tested plants did not have a significant effect on the microbiome. These findings collectively contribute to our understanding of the bacteriome associated with maize and shed light on how various factors shape its composition. Moreover, this knowledge can be valuable in the selection of potential bacteria for bioinoculant production.

Author contribution

Study conception and design: A.D.S.F, J.N.T, L.R.C.S, and C.T.R.C.C. Field experimentation: L.R.C.S, A.F.M, J.N.T, M.C.C.P.L, and A.D.S.F. Laboratory procedures: L.R.C.S, A.F.M, and M.C.C.P.L. Bioinformatic and statistical analysis: P.S.R.B, D.A.M, and C.T.R.C.C. Drafting of the manuscript: L.R.C.S, P.S.R.B, D.A.M, M.C.C.P.L, and C.T.R.C.C. Critical revision of the manuscript: all authors.

Funding

The authors thank the Brazilian Council for Scientific and Technological Development (CNPq; Projetos Universal 2018, Grant Numbers 426655/2018-4 and 409519/2018-9). This work is part of the National Observatory of Water and Carbon Dynamics in the Caatinga Biome - NOWCDCB, supported by Fundação de Amparo à Ciência e Tecnologia de Pernambuco (FACEPE; APQ-0498-3.07/17 ONDACBC), CNPq (grants: 441305/2017-2; 465764/2014-2), and CAPES (grants: 88887.136369/2017-00). Part of this study was financed by FACEPE (APQ-0420-5.01/18) and Fundação Carlos Chagas Filho de Apoio à Pesquisa do Estado do Rio de Janeiro (FAPERJ). Freitas ADS and Costa CTRC are CNPq Research Productivity Scholars of CNPq (classes 1D and 2, respectively)

Data availability

Raw sequences were deposited in the NCBI Sequence Read Archive (SRA) and are available under Bioproject PRJNA846726.

Declarations

Competing interests

The authors declare no competing interests.