Population of EB
The population range of aerobic EB (log cfu/g) in dry seeds (DS) across the seven media was different depending upon the genotypes of rice. EB population in different media from Maguri bao seeds ranged from 5.59-8.98, followed by Kekua bao (6.95-7.43) in different media except in R2A-PEC. The population of EB from hill landraces Idaw, Fazai, and Taiklwangh ranged from 3.00-7.16 cfu/g DS in different media, except on PKA no colony was detected from Idaw. Range of log cfu/g DS of EB from Kalajoha and Ranjit was lowest (3.00-6.11) among the genotypes (Fig. 1a).
Identification, phenotype and metabolic function of EB
The identity of randomly selected 40 isolates based on their 16S rRNA gene sequences is provided in Additional file 1: Table S1. The 16S rRNA gene sequences of isolates from six genotypes (sequences from Maguri bao were poor quality) were submitted to GenBank with accession numbers KY486204-KY486232, KY019244-KY019246, KY013009 - KY013011 and KY003114. Based on alignment against NCBI database, the 40 sequences were found to be representative of 3 phyla, 11 genera, and 16 different species. Distribution of the 16 species within genotypes and agroecosystems is shown in Fig. 1b,c. Kalajoha seeds contained highest (9) and Kekua bao had lowest (2) number of species (Fig. 1c). Among agroecosystems, midland contained highest (12) and deep-water had lowest (2) number of species (Fig. 1b). Proteobacteria was the most dominant phylum comprising five genera, eight species and 72% sequences, followed by Actinobacteria comprising three genera, four species and 18% sequences and Firmicutes comprising three genera and four species and only 10% sequences. 19 out of 40 isolates were 98-99% similar to Pantoea agglomerans and were detected in all the seven growth media from surface sterilized seeds of the six genotypes. Phylogenetic relationship among the 40 isolates is shown in (Fig. 2). We found that the culturable EB population of midland agroecosystem is more phylogenetically diverse than upland and deep-water agroecosystem. Same is also true for genotype Kalajoha than the remaining six genotypes. Brevibacillus borstelensis 37, Knoellia flava 06 and Curtobacterium luteum 41 of families Paenibacillaceae, Intrasporangiaceae and Microbacteriaceae, isolated from Kalajoha are clumped together in a clade. Similarly, Xanthomonas sacchari 05, Microbacterium proteolyticum 34 and Bacillus marisflavi 36, also isolated from Kalajoha and belonging to distinct families cluster in a clade (Fig. 2).
A total of 436 metabolic pathways were assigned based on database search for the 40 EB of the seeds. There were a large number of assigned central pathways of which several pathways are for growth and development of host plant. These pathways were found to vary in their abundance (number of times a pathway was present in an agroecosystem) depending upon agroecosystem. For example, (R)-acetoin biosynthesis I& II, responsible for the synthesis of plant beneficial volatile organic compound (VOC), acetoin [16] was found to be present in most of the EB isolated from landraces of upland agroecosystem. Similarly, pathways for siderophore production such as 2,3-dihydroxybenzoate biosynthesis and siroheme biosynthesis were also found more prominent in isolates of upland agroecosystem followed by midland agroecosystem. These two pathways had very low confidence score (0.714) in the deep-water agroecosystem. In contrast, three pathways for 4-aminobutanoate (GABA) degradation were found to be present in all the isolates of the deep-water agroecosystem. Interestingly, adenosylcobalamin (vitamin B12) biosynthesis pathways were assigned to isolates of the midland genotype Kalajoha and upland genotypes Idaw and Taiklwangh. Selenate reduction pathway was assigned with a higher confidence score (4.22) to the EB of midland agroecosystem followed by those in upland and deep-water (Fig. 3). A full list of the pathways with their confidence score values is given in Additional file 2: Table S2.
General features of sequences and distribution of taxa
522,134 combined reads were obtained after quality filtering and chimera removal with an average of 9323 reads per sample (min=366, max=45574, SD=11608). 4061 OTUs were obtained across the 56 samples (7 genotypes x 8 hills) which were assigned to 29 phyla and 291 genera. Proteobacteria, Actinobacteria, Bacteroides and Firmicutes were the most dominant phyla (relative abundance >3%) covering 98.8% of total sequences. Achromobacter, Agrobacterium, Bifidobacterium, Erwinia, Microbacterium, Ochrobactrum, Pseudomonas, Sphingomonas and Xanthomonas were the nine most prevailing genera (relative abundance >2%) accounting for 74.4% of total reads.
Assembly pattern of taxa at individual and population level
The most dominant among the 29 detected phyla with their abundance values and genera within each phyla with relative abundance >0.7% are shown in Additional file 3: Table S3. Alpha diversity of EB was measured by observed OTU richness, Simpson’s reciprocal index, and Chao1 index (Fig. 4). Diversity indices range of the EB in the seven genotypes is shown in Additional file 4: Table S4. The observed OTU richness was significantly higher in Idaw (1985) than in the other six genotypes as tested by student’s t-test (P≤0.05). It was followed by Taiklwangh (1701), Kekua bao (1597), Kalajoha (1509), Maguri bao (1427), Ranjit (949) and Fazai (896). Chao1 index was also significantly higher in Idaw than in the remaining genotypes (P≤0.05) and was followed by Kekua bao, Taiklwangh, Kalajoha, Maguri bao, Ranjit and Fazai. On the contrary, Simpson’s reciprocal index was significantly higher (P < 0.05) in Fazai, followed by Ranjit, Taiklwangh, Maguri bao, Kalajoha, Kekua bao and Idaw. Number of EB phyla inside seeds of different genotypes varied. Kalajoha had the highest (22), followed by Idaw (21), Taiklwangh (20), Kekua bao (20), Maguri bao (19), Fazai (16) and Ranjit (14). Similarly, total number of genera inside seed of the genotypes also varied. Of the 291 identified genera, Kekua bao had the highest (178), followed by Kalajoha (174), Idaw (165), Maguri bao (158), Taiklwangh (153), Fazai (130) and Ranjit (127). Irrespective of genotype, phyla Proteobacteria, Actinobacteria, Bacteroides and Firmicutes were profuse throughout the seed EB population. The most abundant Proteobacteria included Pseudomonas and Agrobacterium as the prevailing genera (relative abundance >0.7%) found in all the seven genotypes. The number of identified species within each of the four phyla ranged from 45-65. Unassigned OTUs were also present in a substantial number in the landraces (19.79-43.44%) (Additional file 3: Table S3). There was significant difference (ANOVA, F=8.95, Fcrit=2.18, p=2.32E-07) in bacterial community at OTU level between seeds of the seven genotypes of the three agroecosystems.
Apart from variations between the seven genotypes, we found differences in EB taxa in rice seeds of eight replicate hills sampled within each of the genotypes. The eight replicate plants of each rice genotype had dissimilar EB community composition and abundance. Only a few members of total genera were shared by the seeds of sampled eight hills of each genotype. For example, Kekua bao, Idaw, Taiklwangh, Maguri bao, Kalajoha, Fazai, and Ranjit has 18, 18, 18, 15, 14, 7 and 4 common genera respectively in each of their hills and thus forming their core seed EB flora. Among the core OTUs of each genotype, there were members unique to a genotype and members present also in the remaining genotypes. We found Taiklwangh with the most number of unique core OTUs (17), followed by Idaw, Kekua bao, Maguri bao, Kalajoha with only one unique core OTU in each of them. Genotype Fazai and Ranjit, on the contrary had no unique core OTUs (Fig. 5b). List of OTUs detected in the eight hills of each genotype is shown in Additional file 5: Table S5. A heat map analysis of top OTUs across the individual hills of a genotype with hierarchical clustering shows relatedness in seed EB diversity among the eight hills of the seven rice genotypes. Interestingly we found, genotype Fazai and Ranjit formed a single cluster and there were similarities among their hills (Fig. 5a).
There were additional rare genera (relative abundance <0.7%) ranging from 105 (Fazai) to 162 (Kekua bao) in number which formed a distinctive mix in each of the 8 hill seeds of the seven genotypes. We found significant differences in OTU richness, chao1 index, and Simpson’s reciprocal index among the eight hills of each genotype (Fig. 4, Additional file 4: Table S4). Analysis of variance at OTU level of the 8 sampled hills showed that there was no significant difference (ANOVA, F=2.38, Fcrit=4.44, p=0.001) in EB community within the genotypes.
Pearson’s correlation analysis of the most abundant EB at genus level (relative abundance >1%) among the eight hills of each genotype was performed. In Idaw, while Achromobacter was the most dominant in six hills, Erwinia was dominant in the remaining two hills (IDO3 and IDO7) and they were negatively correlated (r = -0.4; Fig. 6a). In TaiklwanghErwinia was negatively correlated with most genera except Xanthomonas (r = 0.96) (Fig. 6b). In Fazai, the dominant Prevotella showed a strong positive correlation with Succinivibrio and Streptococcus (r=0.99) (Fig. 6c). In Ranjit there was a positive correlation between all genera except Brevibacillus (Fig. 6d). In Kalajoha, the most abundant Erwinia displayed positive correlation with all except Achromobacter and Ochrobactrum, with which it showed no correlation (Fig. 6e). In Maguri bao, Achromobacter was positively correlated to most genera except Acinetobacter (r=-0.6) (Fig. 6f). In Kekua bao, Pseudomonas, the most dominant organism has strong positive relation with Agrobacterium (r=0.96) and Erwinia (r=0.99) only (Fig. 6g).
Differential abundance of EB
A few members of the rice seed microbiota showed an interesting pattern of their distribution among the seven genotypes. Significant differences in their population were determined by t-test/ANOVA at FDR corrected P-value cutoff of 0.05. Planctomycetes was the only phylum present at notably distinct numbers (ANOVA, F=4.18, PFDR=0.043) across the seeds of seven genotypes. Its population was lowest in Fazai and Ranjit as compared to the remaining genotypes. At the class level Betaproteobacteria showed significant variation in abundance (ANOVA, F=6.95, PFDR=0.001) (not shown in figure). Orders Actinomycetales, Burkholderiales, Caulobacterales, Rhizobiales, families Alcaligenaceae, Brucellaceae, Propionibacteriaceae, Xanthomonadaceae and genera Achromobacter, Ochrobactrum and Propionibacterium are among the other taxa differentially abundant across the genotypes (Fig. 7). We could not detect any significant OTUs differentially abundant across the agroecosystems.
Agroecosystem effects on rice seed EB diversity
Among the three agroecosystems, highest number of unique OTUs was found in upland (2979), followed by deep-water (2366) and midland (1912). In terms of OTU richness and diversity, upland agroecosystem was richest and most diverse (D=0.96) followed by midland (D=0.95) and deep-water (D=0.93). The upland landrace Fazai and midland variety Ranjit formed a single cluster in the principal coordinate analysis (PCoA) plot obtained using weighted unifrac distance and binary sorensen’s dice index (Fig. 8). Beta diversity analysis clearly showed overlap in seed EB diversity across agroecosystems. This was further confirmed by ANOSIM (R=0.37, p<0.001) and adonis (R2=0.28, p<0.001). Phylogenetic distance of the seed EB diversity showed similarity in EB community of genotypes Kekua bao, Maguri bao, Kalajoha and Idaw (Fig. 8a).
There were 27 unique OTUs common in the two landraces i.e., Kekua bao and Maguri bao forming the core community of deep-water agroecosystem. This comprised of Pseudomonas, Agrobacterium, Achromobacter, Bifidobacterium, Erwinia and three unidentified genera of family Xanthomonadaceae, AlcaligenaceaeEnterobacteriaceae. Inside the upland agroecosystem landrace seeds, 6 OTUs belonging to phyla Actinobacteria and Proteobacteria formed the core microbiota. Inside seeds of all representative genotypes of midland agroecosystem (100%),only 3 OTUs belonging to genera Pseudomonas and Delftia were found to form the core EB (Fig. 5c). However when 90% genotype replicate samples were considered, 8 OTUs belonging to genera Pseudomonas, Delftia, Stenotrophomonas, Propionibacterium and Enhydrobacter were found to form the core microflora.