Rhizosphere and root endosphere present distinct microbial communities
In the root system (rhizosphere and root endosphere), 1450 bacterial OTUs were successfully affiliated to 16 phyla, 41 classes, 85 orders, 115 families, 167 genera and 70 species. The bacterial community associated to the globality of the root system was composed of Proteobacteria (43%), Actinobacteriota (32%), Acidobacteriota (7%), Chloroflexi (5%), Bacteroidota (3%), Verrucomicrobiota (3%), Firmicutes (2%), Myxococcota (2%), Planctomycetota (2%) and other phyla whose proportions were less than 1% (Dependentiae, Desulfobacterota, Gemmatimonadota, Latescibacterota, Methylomirabilota, Nitrospirota and RCP2-54) (Additional file 1: Fig. S1A). PCoA displayed a strong clustering of the bacterial communities according to the compartments and were primarily differentiated along the first axis (Fig. 1A). A higher dispersion of the individuals was observed in the root endophytes group along the second axis. Bacterial richness (Fig. 1B) and diversity (Fig. 1C), represented by Chao1 and Simpson indexes respectively, were significantly higher in the rhizosphere than in the root endosphere. Bacterial OTUs were primarily exclusives (66%) of the rhizosphere, while 31% were common between the root system compartments and less than 3% were root exclusives (Additional file 1: Fig. 2A). Root system compartments displayed bacterial dissimilarities at the phylum (Additional file 1: Fig. 1A) and the class (Additional file 1: Fig. 1B) levels. The LEfSe detected an enrichment of 7 phyla (Acidobacteriota, Chloroflexi, Verrucomicrobiota, Planctomycetota, Firmicutes, Actinobacteriota, Gemmatimonadota), 10 classes, 16 orders, 17 families, 14 genera and 13 species in the rhizosphere, compared to 3 phyla (Proteobacteria, Bacteroidota, Dependentiae), 5 classes, 8 orders, 14 families, 17 genera, 16 species in the root endosphere (Additional file 1: Fig. 3A).
Among the fungal community analyzed by ITS sequencing, 650 OTUs were affiliated to 12 phyla, 31 classes, 69 orders, 115 families, 174 genera and 302 species. The root system community was composed of Ascomycota (60%), Basidiomycota (22%), Rozellomycota (11%), Mortierellomycota (5%), Glomeromycota (2%) and other phyla whose proportions were less than 1% (Blastocladiomycota, Calcarisporiellomycota, Chytridiomycota, Kickxellomycota, Mucoromycota, Olpidiomycota, Zoopagomycota) (Additional file 1: Fig. S1C). PCoA enabled separation of the fungal community into two groups corresponding to the compartments of the root system, with the same profile as that observed for bacteria (Fig. 1D). Fungal richness (Fig. 1E) and diversity (Fig.1F) were significantly higher in the rhizosphere than in the root endosphere. Amongst the root system compartments, 62% of the fungal OTUs were shared, with 38% and less than 1% specific to the rhizosphere and the root endosphere, respectively (Additional file 1: Fig. S2A). Dissimilarities were observed at the phyla (Additional file 1: Fig. S1C) and the class (Additional file 1: Fig. S1D) levels between the root system compartments. LEfSe showed an enrichment of 2 phyla (Ascomycota, Mortierellomycota), 5 classes, 9 orders, 12 families, 12 genera and 10 species in the rhizosphere. In the root endosphere, 2 phyla (Glomeromycota, Basidiomycota), 5 classes, 7 orders, 9 families, 12 genera and 14 species were enriched (Additional file 1: Fig. S3B).
In the fungal sub-group of AMF, 230 OTUs were affiliated to 1 phylum, 2 classes, 3 orders, 5 families, 8 genera, and 40 species. The AMF genera were represented by Glomus (86%), Claroideoglomus (3%), Paraglomus (2%), Scutellospora (2%), multi-affiliated genera (6%), and other genera whose proportions were less than 1% (Acaulospora, Rhizophagus, and Septoglomus) (Additional file 1: Fig. S1E). No segregation of the root system compartments was observed with the PCoA (Fig. 1G). Although no significant difference was observed for the Chao1 index (Fig. 1H) between root system compartments, the Simpson index (Fig. 1I) was significatively higher in the rhizosphere than in the roots. Amongst both root and rhizosphere compartments, 88% of AMF OTUs were shared, with 11% and less than 1% specific to the rhizosphere and the root endosphere, respectively (Additional file 1: Fig. S2A). The abundance graph at the genus level showed different proportions of AMF between root endosphere and rhizosphere compartments (Additional file 1: Fig. S1E). LEfSe revealed an enrichment of 1 order (Paraglomerales), 2 families (Paraglomeraceae and Claroideoglomeraceae), 2 genera and 2 species in the rhizosphere, compared to 1 order (Glomerales) and 1 family (Glomeraceae) in the root endosphere (Additional file 1: Fig. S3C). Although the differences amongst the root system compartments were less evident for the AMF, we decided to separate the root endosphere and the rhizosphere in the following analyses for the three groups of microorganisms.
Rootstock and scion genotypes influence the bacterial community of the rhizosphere and the root endosphere
Table 1 Influence of the rootstock and the scion genotypes on the α-diversity (richness = Chao1, diversity= Simpson) and the β-diversity metrics (Bray-Curtis dissimilarity) amongst bacteria, fungi and AMF, in the rhizosphere (RH) and the root endosphere (RE)
|
|
|
Influence of the rootstock
|
Influence of the scion
|
|
|
|
Chao1
|
Simpson
|
Bray-Curtis
|
Chao1
|
Simpson
|
Bray-Curtis
|
|
|
|
p-value
|
PVE
|
p-value
|
PVE
|
p-value
|
PVE
|
p-value
|
PVE
|
p-value
|
PVE
|
p-value
|
PVE
|
Bacteria
|
RH
|
Genotype (G)
|
1.69e-08
|
***
|
73%
|
2.99e-4
|
***
|
53%
|
0.001
|
***
|
50%
|
0,574
|
ns
|
/
|
0.008
|
**
|
41%
|
0.002
|
**
|
25%
|
Block (B)
|
0.012
|
*
|
5%
|
0.485
|
ns
|
/
|
0.003
|
**
|
7%
|
0.001
|
***
|
36%
|
0.408
|
ns
|
/
|
0.019
|
*
|
6%
|
G x B
|
1.85e-05
|
***
|
15%
|
0.919
|
ns
|
/
|
0.001
|
***
|
17%
|
0.031
|
*
|
22%
|
0.749
|
ns
|
/
|
0.001
|
***
|
33%
|
RE
|
G
|
0.663
|
ns
|
/
|
0.018
|
*
|
29%
|
0.001
|
***
|
42%
|
0.469
|
ns
|
/
|
0.632
|
ns
|
/
|
0.004
|
**
|
24%
|
B
|
0.826
|
ns
|
/
|
0.002
|
**
|
19%
|
0.019
|
*
|
4%
|
0.861
|
ns
|
/
|
0.483
|
ns
|
/
|
0.044
|
*
|
6%
|
G x B
|
0.644
|
ns
|
/
|
0.512
|
ns
|
/
|
0.045
|
*
|
12%
|
0.476
|
ns
|
/
|
0.318
|
ns
|
/
|
0.001
|
***
|
25%
|
Fungi
|
RH
|
G
|
0.141
|
ns
|
/
|
0.007
|
**
|
40%
|
0.001
|
***
|
42%
|
0.990
|
ns
|
ns
|
0.232
|
ns
|
/
|
0.001
|
***
|
30%
|
B
|
0.008
|
**
|
18%
|
0.590
|
ns
|
/
|
0.002
|
**
|
6%
|
0.193
|
ns
|
ns
|
0.754
|
ns
|
/
|
0.001
|
***
|
7%
|
G x B
|
7.67e-06
|
***
|
45%
|
0.168
|
ns
|
/
|
0.001
|
***
|
19%
|
1.79e-4
|
***
|
60%
|
0.045
|
*
|
30%
|
0.001
|
***
|
29%
|
RE
|
G
|
0.001
|
**
|
49%
|
0.052
|
ns
|
/
|
0.001
|
***
|
33%
|
0.490
|
ns
|
/
|
0.727
|
ns
|
/
|
0.002
|
**
|
24%
|
B
|
0.897
|
ns
|
/
|
0.389
|
ns
|
/
|
0.008
|
**
|
5%
|
0.558
|
ns
|
/
|
0.318
|
ns
|
/
|
0.015
|
*
|
7%
|
G x B
|
0.729
|
ns
|
/
|
0.694
|
ns
|
/
|
0.001
|
***
|
19%
|
0.279
|
ns
|
/
|
0.136
|
ns
|
/
|
0.001
|
***
|
26%
|
AMF
|
RH
|
G
|
0.395
|
ns
|
/
|
0.239
|
ns
|
/
|
0.001
|
***
|
23%
|
0.409
|
ns
|
/
|
0.471
|
ns
|
/
|
0.001
|
***
|
22%
|
B
|
0.373
|
ns
|
/
|
0.033
|
*
|
12%
|
0.042
|
*
|
4%
|
0.284
|
ns
|
/
|
0.489
|
ns
|
/
|
0.438
|
ns
|
/
|
G x B
|
0.194
|
ns
|
/
|
0.056
|
ns
|
/
|
0.006
|
**
|
17%
|
0.374
|
ns
|
/
|
0.284
|
ns
|
/
|
0.031
|
*
|
16%
|
RE
|
G
|
0.007
|
**
|
41%
|
0.013
|
*
|
37%
|
0.031
|
*
|
19%
|
0.191
|
ns
|
/
|
0.516
|
ns
|
/
|
0.137
|
ns
|
/
|
B
|
0.385
|
ns
|
/
|
0.210
|
ns
|
/
|
0.523
|
ns
|
/
|
0.997
|
ns
|
/
|
0.174
|
ns
|
/
|
0.299
|
ns
|
/
|
G x B
|
0.131
|
ns
|
/
|
0.233
|
ns
|
/
|
0.278
|
ns
|
/
|
0.548
|
ns
|
/
|
0.098
|
ns
|
/
|
0.002
|
**
|
24%
|
When the bacterial communities of the six rootstocks grafted with CS were assessed, no effect of the rootstock genotype was observed on rhizosphere bacteria level with both cultivable and qPCR methodologies (Additional file 2: Table S4; Additional file 1: Fig. S4A). The rootstock genotype, the block, and the factor combinations significantly influenced the bacterial Bray-Curtis index in the rhizosphere compartment with a percentage of variance explained (PVE) of 50, 7, and 17%, respectively (Table 1). In the root endosphere, similar results were obtained with 42, 4, and 12% of PVE, respectively. The Chao1 and Simpson indexes were influenced by the rootstock genotype (73 and 53% of PVE respectively) in the rhizosphere. The Chao1 index was also influenced by the block (5% of PVE) and the factor interactions (15% of PVE). In the root endosphere, only the Simpson index was influenced by the genotype and the block (29 and 19 % of PVE respectively). As the effect of the genotype (expressed as the percentage of variance explained) was always stronger than that of the block or the factor interactions, we kept the genotype factor to compare rootstocks together with a suitable number of biological replicates (n = 6). PCoA showed a genotype-dependent clustering of individuals in both rhizosphere (Fig. 2A) and root endosphere (Fig. 2D) compartments. Interestingly, the Bray-Curtis index measured for 1103P rootstock was significantly different from those of all other genotypes in both root system compartments, except for SO4 rootstock in the root endosphere (Additional file 2: Table S5). Moreover, the Bray-Curtis index of 41B rootstock was significantly different from those of Nemadex and SO4 in the rhizosphere, as well as those of 3309C in both root system compartments. In the rhizosphere, 1103P rootstock had a significantly lower Chao1 index (Fig. 2B) than all the other genotypes, while RGM had the significantly lowest Simpson index (Fig. 2C). In the root endosphere, no significant differences were detected between bacterial richness amongst the rootstock (Fig. 2E), while the Simpson index of 3309C rootstock was significantly lower compared to Nem (Fig. 2F). Between the six different rootstock genotypes, 47% of the rhizospheric OTUs were common, while less than 1% were specific from one genotype (Additional file 1: Fig. S2B). Inversely in the root endosphere, only 6% of bacterial OTUs were common and 49% were genotype-exclusives. LEfSe detected an enrichment of 4 phyla (Acidobacteriota for SO4, Verrucomicrobiota for Nemadex, Planctomycetota for 3309C and Proteobacteria for 1103P- rootstocks), 7 classes, 7 orders, 6 families, 6 genus, 6 species in the rhizosphere amongst the six rootstocks (Fig. 3A). In the root endosphere, 3 classes, 10 orders, 11 families, 16 genus and 17 species were enriched (Fig. 3B).
Regarding the influence on microbial communities of five scion genotypes grafted onto RGM rootstock, an effect of the scion on the level of bacterial 16S rRNA gene from the rhizosphere was reported (Additional file 2: Table S4), but no significant difference was observed between groups with the adjusted P-values (Additional file 2: Fig. S4D). In the rhizosphere and the root endosphere, Bray-Curtis index was influenced by the scion genotype (25 and 24% of PVE), the block (6% of PVE) and their interaction (33 and 25% of PVE) (Table 1). Regarding the bacterial α-diversity, the effect of the scion genotype was only reported on the Simpson index in the rhizosphere, explaining 41% of PVE. Individuals were not well clustered by scion genotype according to PCoA plot for both rhizosphere (Additional file 1: Fig. S5A) and root endosphere (Additional file 1: Fig. S5D) compartments. Only the Bray-Curtis index of UB scion was significantly different from the one of Syrah grafted onto RGM in both root system compartments, as well as the Bray-Curtis indexes of UB and Grenache in the rhizosphere (Additional file 2: Table S5). Simpson index in the rhizosphere was significantly lower in CS than in Grenache and UB (Additional file 1: Fig. S5C). Bacterial OTUs were common between scion genotypes at 78% and less than 1% were genotype exclusive in the rhizosphere, compared to 18% and 40% in the root endosphere (Additional file 1: Fig. S2C). The LEfSe identified genotype-dependent enrichments of 2 phyla (Planctomycetota for UB, Proteobacteria for Grenache), 2 classes, 1 order, 1 family, 1 genus and 1 species in the rhizosphere (Additional file 1: Fig. S6A), compared to 2 phyla (Acidobacteriota for UB and Proteobacteria for Grenache) 2 orders, 2 families, 5 genera and 6 species in the root endosphere (Additional file 1: Fig. S6B).
Both genotypes of the grafted plant influence the root-associated fungal community
Quantitative PCR analysis revealed that the rootstock genotype affected the level of the fungal ITS gene from the rhizosphere (Additional file 2: Table S4) which was significantly higher for SO4 rootstock than for 1103P, 41B and RGM (Additional file 1: Fig. S4B). In the rhizosphere and the root endosphere, the Bray-Curtis index of fungal communities was influenced by the rootstock genotype (explaining 42% and 33% of PVE), the block (6% and 5% of PVE) and their interaction (19% of PVE) (Table 1). Regarding α-diversity metrics, the Simpson index in the rhizosphere and the Chao1 index in the root endosphere were influenced by the rootstock genotype only (40 and 49% of PVE, respectively). PCoA displayed strong genotype-dependent clustering in the rhizosphere (Fig. 4A). The fungal Bray-Curtis index of 1103P rootstock was significantly different from those of all other genotypes except for 41B (Additional file 2: Table S6). Other distinct fungal Bray-Curtis indexes were observed between Nemadex and 41B, as well as 3309C and RGM or 41B. The clustering of fungal communities according to rootstock genotype was less evident in the root endosphere (Fig. 4D). The fungal structure was significantly different between 1103P rootstock and Nemadex, 41B, and 3309C, as well as between 41B rootstock and Nemadex and 3309C (Additional file 2: Table S6). In the rhizosphere, the Simpson index was significantly higher in 1103P rootstock than in Nemadex and SO4 (Fig. 4C). In the root endosphere, the Chao1 index was significantly higher in Nemadex rootstock than in 1103P and 41B (Fig. 4E). In the rhizosphere, fungal OTUs were common and genotype-exclusive at 32% and 10%, respectively, while 7% were common and 34% were genotype-exclusive in the root endosphere (Additional file 1: Fig. S2B). Three phyla (Ascomycota for 1103P, Basidiomycota and Rozellomycota for SO4), 4 classes, 6 orders, 4 families, 2 genera, 2 species were enriched in the rhizosphere of rootstock genotypes (Fig. 3C). In the root endosphere, 2 classes, 7 orders, 14 families, 17 genera and 19 species were enriched (Fig. 3D).
Despite the reported effect of the scion on the level of fungal 18S rRNA gene from the rhizosphere (Additional file 2: Table 4), no significant differences were observed between scion genotypes using the adjusted P-values (Additional file 1: Fig. S4E). In the rhizosphere and the root endosphere, the fungal Bray-Curtis index was influenced by the scion genotype (30% and 24% of PVE), the block (7% of PVE) and their interaction (29% and 26% of PVE) (Table 1). None of the fungal α-diversity metrics were influenced by the scion genotype or the block. However, effects of their interaction were reported on the richness and the diversity of the rhizosphere (60% and 30% of PVE, respectively) (Table 1). According to PCoA, individuals were not clustered by genotypes in both rhizosphere (Additional file 1: Fig. S7A) and root endosphere (Additional file 1: Fig. S7D) compartments. The fungal Bray-Curtis index measured for UB was significantly different from those of PN, Syr and Gre in the rhizosphere, and from those of PN in the root endosphere (Additional file 2: Table S6). Syr and Gre were also significantly different in the rhizosphere. In this compartment, 40% of fungal OTUs were common and 11% were genotype-exclusive, while 17% were common and 33% were genotype-exclusive in the root endosphere. LEfSe analysis detected enrichments in 3 scion genotypes of 2 classes, 3 orders, 2 families, 2 genera and 2 species in the rhizosphere (Additional file 1: Fig. S6C), while enrichments of 3 classes, 2 orders, 2 families, 2 genera and 2 species were detected in the root endosphere (Additional file 1: Fig. S6D).
Rootstock genotypes drive the AMF community in the root endosphere
The Bray-Curtis index measured for AMF was influenced by the rootstock genotype, the block, and the interaction in the rhizosphere (23%, 4%, and 17% of PVE, respectively), as well as by the genotype in the root endosphere (19% of PVE) (Table 1). However, Chao1 and Simpson indexes were influenced by the genotype only in the root endosphere (41% and 37% of PVE, respectively). PCoA failed to cluster individuals according to rootstock genotypes in both rhizosphere (Fig. 5A) and root endosphere (Fig. 5D) compartments. Only the Bray-Curtis index of 3309C rootstock was significantly distinct from the indexes of 1103P and 41B in the rhizosphere (Additional file 2: Table S7). In the root endosphere, the Chao1 index was significantly higher in RGM than in 1103P and 41B (Fig. 5E). Likewise, the Simpson index was higher in RGM and Nemadex than in 41B (Fig. 5F). In the rhizosphere, 12% of the AMF OTUs were common between the rootstocks while 26% were genotype exclusive (Additional file 1: Fig. S2B). In the root endosphere, 20% were common and 30% were exclusive. LEfSe detected an enrichment of 2 orders, 1 family, 1 genus and 5 species in the rhizosphere (Fig. 3 F), compared to 1 phylum, 1 class and 1 species in the root endosphere (Fig. 3E).
In addition to rootstock, a scion genotype effect was reported on AMF communities, the Bray-Curtis index was influenced by both the genotype and its interaction with the block in the rhizosphere (22 and 16% of the PVE, respectively), as well as by the factor interactions in the root endosphere (24% of PVE) (Table 1). However, the α-diversity indices did not detect any effects of the scion genotype in either of the root system compartments. According to PCoA, individuals were not clustered by genotype in either the rhizosphere (Additional file 1: Fig. S8A) or the root endosphere (Additional file 1: Fig. S8D) compartments. No AMF communities were significantly distinct between the scion genotypes with the Bray-Curtis index in either of the root system compartments endosphere (Additional file 2: Table S7). In the rhizosphere, 14% and 27% of the AMF OTUs were common or exclusive between scion genotypes, respectively, while in the root endosphere, 20% were common and 28% were genotype exclusive (Additional file 1: Fig. S2C). LEfSe detected an enrichment of 1 phylum, 1 genus and 4 species in the rhizosphere (Additional file 1: Fig. S6E), compared to 1 phylum, 1 class and 2 species in the root endosphere (Additional file 1: Fig. S6F).
Predicted functions of the bacterial and fungal microbiomes and impact of the root system microbiomes on plant phenotypic traits
Bacterial potential metabolic pathways were predicted using PICRUSt2. All the abundances of pathways were significantly higher in the rhizosphere than in the root endosphere, except for the metabolism of lipids, the metabolism of other amino acids, and the xenobiotic biodegradation and metabolism (Additional file 2: Table S8). PCA performed with the 119 predicted functions showed high clustering dependent on the root system compartment (Fig. 6D), suggesting that the function of the bacterial microbiomes depends on their environment. An effect of the rootstock genotype was reported on the abundances of all the predicted bacterial metabolic pathways in the rhizosphere (Additional file 2: Table S9). Several biosynthetic pathways were very abundant in the rhizosphere of certain rootstock genotypes, such as the amino acid metabolism for 1103P, the carbohydrate metabolism for Nemadex, the signaling molecules interaction for 41B, and the xenobiotics biodegradation and metabolism for RGM (Fig. 6A). According to PCAs, the predicted functions of the rhizospheric bacterial community for 1103P rootstock were distinct from those of 3309C and Nemadex (Fig. 6E). In the root endosphere, the rootstock genotype influenced only the metabolisms of amino acids and lipids, the membrane transport, and the signal transduction of bacteria (Additional file 2: Table S9; Additional file 1: Fig. S6B). However, according to PCA, a slight segregation was detected between Nemadex and 41B (Fig. 6F). The abundances of six predicted bacterial metabolic pathways in the rhizosphere were influenced by the scion genotype (metabolism of carbohydrate, energy and lipid, membrane transport, signal transduction and xenobiotics biodegradation and metabolism), while the abundances of three and four pathways were influenced by the block and the factors combination, respectively (Additional file 2: Table S9; Fig. 6C). According to PCA, no scion genotype-dependent clustering of bacterial functions was observed in either the rhizosphere (Additional file 1: Fig. S9A) or root endosphere (Additional file 1: Fig. S9B) compartments. Here, the influence of the scion on the abundances of bacterial metabolic pathways is weaker than that of the rootstock.
Fungal trophic modes and guilds, predicted using the FUNGuild database, were successfully affiliated with 33% of fungal OTUs from the rhizosphere and 42% from the root endosphere. An effect of the root system compartments on the proportions of fungal trophic modes (Fig. 7A) and guilds (Fig. 7C) was reported (Chi Squared test: P-values < 2.2e-16). Fungi were mainly saprotroph in both root system compartments, even though a significant number of multi-trophic modes were observed in the rhizosphere. The highest proportions of pathotroph and plant pathogen were observed in the rhizosphere, whereas the highest proportions of symbiotroph or arbuscular mycorrhizal were observed in the roots. Rootstock genotype influenced the proportions of fungal trophic modes (Fig. 7B) and guilds (Fig. 7D) in both root system compartments (Chi Squared test: P-values < 2.2e-16). Interestingly, 1103P had over 3 times more pathogenic fungi than any other genotype in the root endosphere. Moreover, the highest proportion of AMF in roots was observed for Nemadex rootstock. An effect of the scion was also detected on the proportions of trophic modes and guilds in both root system compartments (Chi Squared test: P-values < 2.2e-16, Additional file 1: Fig. S10). The root system of RGM grafted with UB had over twice as much AMF in the roots than other scion genotypes.
To study the impacts of microbial communities on plant phenotypic traits, and mineral content in the roots and petioles, correlation matrices were established with the microbial variables (cultivable, qPCR and metabarcoding approaches). Between the root system microbiomes and the plant phenotypic traits, 22 significant correlations were established, with r values ranging from -0.49 to 0.45. The strongest correlations (P-values < 0.05) were obtained between the number of shoots and AMF richness in the rhizosphere (r = -0.45); the number of bunches and the cultivable fungi (r = -0.4) and the AMF richness in the rhizosphere (r = 0.41); the bunch pruning weight and the rhizosphere archaeal level measured by qPCR (r = 0.45); the δ13C and the rhizosphere archaeal level (r = -0.49) and the bacterial richness in the rhizosphere (r = 0.44) (Fig. 8A). In addition, 39 microbial variables were significantly correlated to mineral dosage in the petiole, with r values ranging from -0.44 to 0.45. The strongest correlations were obtained between the sulfur content and the AMF diversity in the roots (r = 0.45); the richness of rhizosphere bacteria and the calcium content (r = -0.46) and the manganese content (r = 0.43) (Fig. 8B). Finally, 29 microbial variables were significantly correlated to mineral content in the roots, ranging from -0.48 to 0.4. Interestingly, the calcium content was negatively correlated with the richness of bacteria (r = - 0.42), fungi (r = - 0.48), and AMF (r = - 0.42) in the rhizosphere. The manganese was also correlated with the AMF diversity (r = 0.4) and richness (r = 0.37) in the roots (Fig. 8C).