General traits of the rhizosphere bacterial and archaeal community of field-grown rice
We used metabarcoding targeting the V4 hypervariable region of the 16S rRNA gene to assess the effects of long-term (> 25 years) inorganic fertilization on bacterial and archaeal community inhabiting the rice rhizosphere in paddy fields at three growth stages (tillering, panicle initiation and booting). A total of 31 bacteria phyla including 66 classes were found across all soil samples. The most dominant bacteria phyla (classes for Proteobacteria) in terms of relative abundance across fertilizations and growth stages were Chloroflexi (13.34–25.62%), Deltaproteobacteria (18.21–19.72%) and Firmicutes (8.67–12.01%), which together accounted for 51.23% of the bacteria sequences (Fig. 1a). They were followed by Betaproteobacteria (6.28–14.00%), Actinobacteria (5.63–8.78%), Alphaproteobacteria (5.35–11.40%), and Acidobacteria (4.66–6.35%), which contributed to 27.98% of the bacteria sequences. For archaea, we recorded 5 phyla and 13 classes across all rhizosphere soil samples. Methanomicrobia (39.54–58.19%) was the most dominant class, followed by Crenarchaeota (12.89–21.85%), Methanobacteria (12.05–15.69%) and Thaumarchaeota (6.10–20.70%); together they accounted for 92.64% of the archaea sequences (Fig. 1b).
Alpha diversity (Shannon index) of the bacterial community was greater than those of the archaeal community across all samples (Fig. 1c-d). For bacteria, we detected a significant interaction effect between fertilization and growth stage on Shannon diversity index (Chi2 = 12.71, df = 4, P = 0.013, Table S1). However, compared to unfertilized treatment, the long-term application of N and NPK-fertilizer did not significantly affect the bacterial alpha diversity at the three growth stages (Fig. 1c). For archaea, we detected significant effects of fertilization (Chi2 = 14.77, df = 2, P = 0.001) and growth stage (Chi2 = 20.28, df = 2, P < 0.001) on Shannon diversity index, while there were no significant interaction effects (Table S1), indicating that inorganic fertilization and growth stage independently affected alpha diversity. In general, archaeal alpha diversity increased in response to N and NPK-fertilization and from tillering to booting stage (Fig. 1d).
We next used unconstrained principal coordinate analysis (PCoA) and PERMANOVA based on Bray-Curtis dissimilarities to visualize and quantify the differences between microbial communities (beta-diversity). For both bacteria and archaea, the NPK-fertilized soil was separated from the N-fertilized and unfertilized soil along the axis 1 (Fig S1a-b). Interestingly, this observation was confirmed by PERMANOVA (F = 3.760, R2 = 0.133, P < 0.001 for bacteria and F = 7.039, R2 = 0.239, P < 0.001 for archaea; Table S2). Moreover, Mantel test showed significant moderate to strong correlations between soil properties especially soil pH, NH4+-N, total C, total N and total P and both bacterial and archaeal community (Table S3). We further observed a growth stage effect on microbial communities, especially bacteria across the axis 3 of the PCoA plot (Fig. S1c-d). PERMANOVA again revealed that the effect of growth stage was significant for both bacteria (F = 2.293, R2 = 0.081, P < 0.001; Table S2) and archaea (F = 1.371, R2 = 0.047, P = 0.023; Table S2).
Long-term Fertilization Effects On Microbial Communities Inhabiting The Rice Rhizosphere At Different Growth Stages
We used constrained analysis of principal coordinates (CAP) to further track the effects of long-term fertilization and sampling time-point on microbial communities at each growth stage of field-grown rice. The partial CAP, constrained by both long-term fertilization and sampling time-point, showed long-term fertilization effects on both bacterial and archaeal community at the tillering, panicle initiation and booting stage (Fig. 2).
No sampling time effects were observed on both bacterial and archaeal community at the different growth stages. This observation was confirmed by PERMANOVA (Table S4). The pairwise comparisons revealed that, at the tillering stage, the bacterial communities in unfertilized and N-fertilized soil were not significantly different, while they differed significantly from those in NPK-fertilized soil. Meanwhile, the three soil types (unfertilized, N-fertilized and NPK-fertilized) harbored dissimilar bacterial communities at the panicle initiation as well as at the booting stage (Table S4). For archaeal community, significant differences were observed between unfertilized, N-fertilized and NPK-fertilized soil at the tillering and booting stage and between NPK-fertilized soil and both unfertilized and N-fertilized soil at the panicle initiation (Table S4). Furthermore, the Betadisper analysis revealed significant differences in group dispersion for the archaeal community at the tillering stage (F = 4.024, df = 2, P = 0.028) and panicle initiation (F = 3.237, df = 2, P = 0.009). Conversely, no significant differences in group dispersion were observed for bacterial community at the different growth stages, suggesting that the differences between fertilization regimes were mainly driven by true biological differences (Table S4). Overall, our results indicate that the effect of long-term fertilization on rhizosphere microbial communities varied depending on the growth stage, and that the bacterial and archaeal community differed in their response to long-term N and NPK-fertilization.
Distribution Of Long-term Inorganic Fertilization Sensitive Taxa Across Growth Stages Of Field-grown Rice
We next tracked the bacteria and archaea taxa sensitive to long-term inorganic fertilization at each growth stage. For this, indicator species analysis was performed to identify bacteria and archaea OTUs whose abundances varied between unfertilized, N-fertilized and NPK-fertilized soil at each growth stage and then validated those OTUs by using likelihood ratio tests (LRT). At each growth stage, the validated OTUs (identified by both methods) were defined as long-term inorganic fertilization sensitive (lifs) OTUs (Fig. S2). 3849 and 986 lifs OTUs of bacteria and archaea, respectively, were identified across the different growth stages (Fig. 3a-b). The bacterial community included 946, 2387 and 1415 lifs OTUs contributing to 7.81, 16.67 and 8.22% of the total abundance at the tillering, panicle initiation and booting stage, respectively (Fig. 3a-c). For archaea, we found 407, 466 and 415 lifs OTUs accounting for 22, 23.40 and 17.30% of the total abundance at the tillering, panicle initiation and booting stage, respectively (Fig. 3b-d).
A statistically significant effect of growth stage on the relative abundance of lifs OTUs was detected for both kingdoms, with a larger magnitude for bacteria (Chi2 = 62.73, df = 2, P < 0.001) compared to archaea (Chi2 = 13.42, df = 2, P = 0.001). In addition, dynamic enrichment/depletion patterns of bacteria and archaea lifs OTUs in unfertilized, N-fertilized and NPK-fertilized soil were observed across growth stages (Fig. 4). For instance, the bacteria lifs OTUs were preferentially enriched in NPK-fertilized soil, followed by N-fertilized soil at the panicle initiation, whereas they were preferentially depleted in NPK-fertilized soil at the booting stage (Fig. 4a-c). For archaea, the lifs OTUs were also preferentially depleted in NPK-fertilized soil as compared to unfertilized and N-fertilized soil at the tillering and booting stage (Fig. 4d-f).
On the other hand, 7 bacteria phyla and 6 archaea classes were identified as sensitive to long-term inorganic fertilization across the different growth stages (Fig. S3). Among them, Epsilonproteobacteria was significantly depleted, whereas Parvarchaeota was enriched in NPK-fertilized soil at the tillering stage. At the panicle initiation, Alphaproteobacteria and Betaproteobacteria were significantly enriched, whereas Chlorobi, Chloroflexi, Ignavibacteriae, Verrucomicrobia, Methanomicrobia and Thermoplasmata were depleted in NPK-fertilized soil. At the booting stage, Thaumarchaeota were enriched, whereas Halobacteria and Methanococci were depleted in N-fertilized soil (Fig. S3). Furthermore, 57 and 3 bacteria and archaea families, respectively, were identified as sensitive to long-term inorganic fertilization across the different growth stages (Fig. S4). Although some of those families exhibited their sensitivity to long-term inorganic fertilization either at one, two or three growth stages, most of them were preferentially sensitive at the panicle initiation (Fig. S4).
Taken together, our results reveal differential microbial sensitivity to long-term inorganic fertilization across the different growth stages, with a higher relative abundance of lifs taxa at the panicle initiation compared to at the tillering and booting stage. However, the effect of growth stage on microbial sensitivity to long-term inorganic fertilization was more strongly pronounced for bacterial than archaeal community.
Microbial Inter-kingdom Co-occurrence Patterns In The Rhizosphere Core Microbiome Across Growth Stages Of Field-grown Rice
We first determined the core bacterial and archaeal microbiomes inhabiting the rice rhizosphere at each growth stage by using a 75% prevalence threshold. We next analyzed the differences in the core microbiomes at the different growth stages by investigating the taxa co-occurrence patterns using microbial inter-kingdom network analysis. Hence, the microbial inter-kingdom network obtained at the tillering stage contained 814 nodes, those at the panicle initiation contained 1040 nodes and those at the booting stage contained 910 nodes (Fig. 5, Table S5). Furthermore, the microbial inter-kingdom network obtained at the panicle initiation displayed a higher proportion (28.56%) of lifs OTUs than those obtained at the booting (19.67%) and the tillering stage (18.80%). The proportion of bacteria nodes increased from tillering to panicle initiation and decreased to booting stage, while those of archaea nodes followed opposite trends. Similarly, the proportion of bacteria – bacteria edges, bacteria – archaea edges and negative correlations increased from tillering to panicle initiation and decreased to booting stage, whereas those of archaea – archaea edges and positive correlations followed opposite trends (Fig. 5, Table S5).
In addition, 17 potential hub OTUs were identified across the three growth stages (Table S6). Of those potential hub OTUs, 2 belonging to archaea (Methanobacterium and Methanosaeta) were identified at the tillering stage, 9 including 2 bacteria (Gemmatimonas and Pirellula) and 7 archaea (four Methanobacterium, two Methanosaeta and one Methanosphaerula) at the panicle initiation, and 6 belonging to archaea (five Methanosaeta and one Methanobacterium) at the booting stage (Table S6). Together these results reveal dynamics of bacteria and archaea co-occurrence patterns in the rice rhizosphere, with differentiated bacterial and archaeal pivotal roles in the microbial inter-kingdom networks across growth stages.